17.11.2012 Views

Raw Reading Times (ms) - Generative Linguistics in the Old World

Raw Reading Times (ms) - Generative Linguistics in the Old World

Raw Reading Times (ms) - Generative Linguistics in the Old World

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Marta Abrusan (Gött<strong>in</strong>gen) & Kriszta Szendrői (UCL)<br />

Experiment<strong>in</strong>g with <strong>the</strong> K<strong>in</strong>g of France: topics, verifiability and def<strong>in</strong>ite descriptions<br />

1. Theoretical situation The Fregean position is widely accepted <strong>in</strong> <strong>the</strong> semantics literature:<br />

def<strong>in</strong>ite descriptions trigger an existential presupposition (Heim & Kratzer 1998). However, this<br />

existential presupposition see<strong>ms</strong> not to appear equally strongly <strong>in</strong> every sentence (Strawson 1969,<br />

Re<strong>in</strong>hart 1981, von F<strong>in</strong>tel 2004, etc.). E.g. accord<strong>in</strong>g to Strawson, <strong>in</strong> a context where it is known<br />

that France has no k<strong>in</strong>g, (1a) is felt as “nei<strong>the</strong>r false nor true”, but (1b) is judged as simply false.<br />

(1) a. The k<strong>in</strong>g of France is bald. b. The exhibition was visited yesterday by <strong>the</strong> k<strong>in</strong>g of France.<br />

The solution of Strawson (1969) (also Re<strong>in</strong>hart 1981) was that def<strong>in</strong>ite descriptions trigger an<br />

existential presupposition only <strong>in</strong> topic position. The NP ‘<strong>the</strong> k<strong>in</strong>g of France’ is <strong>in</strong> topic position<br />

<strong>in</strong> (1a) but not <strong>in</strong> (1b). Thus (1a) but not (1b) leads to a presupposition failure. Lasersohn (1993)<br />

and von F<strong>in</strong>tel (2004) dismissed <strong>the</strong> importance of topichood and argued that ra<strong>the</strong>r than<br />

topicality, verifiability is important: If a sentence conta<strong>in</strong>s an <strong>in</strong>dependent NP such that <strong>the</strong><br />

sentence could be verified based on <strong>the</strong> properties of this NP, speakers might have enough<br />

grounds to accept or reject <strong>the</strong> sentence whe<strong>the</strong>r or not <strong>the</strong> presupposition of <strong>the</strong> def<strong>in</strong>ite<br />

description is satisfied. This happens <strong>in</strong> The k<strong>in</strong>g of France is sitt<strong>in</strong>g <strong>in</strong> this chair, where <strong>the</strong> sentence<br />

could be verified based on <strong>the</strong> properties of <strong>the</strong> chair. Fur<strong>the</strong>r, as von F<strong>in</strong>tel (2004) has shown, <strong>in</strong><br />

some cases even def<strong>in</strong>ite descriptions that are uncontroversially topics do not seem to trigger <strong>the</strong><br />

“squeamishness” associated with presupposition failure: Let me tell you about my friend, <strong>the</strong> k<strong>in</strong>g of<br />

France. I had breakfast with him this morn<strong>in</strong>g. Thus accord<strong>in</strong>g to both Lasersohn and von F<strong>in</strong>tel,<br />

def<strong>in</strong>ite descriptions always trigger an existential presupposition, hence all <strong>the</strong> above examples<br />

<strong>in</strong>volve a presupposition failure. But if <strong>the</strong> speaker has <strong>in</strong>dependent footholds for verification, he<br />

might judge <strong>the</strong> sentence as false.<br />

2. Summary of <strong>the</strong> experiment and results All <strong>the</strong> data <strong>in</strong> <strong>the</strong> above papers are based on <strong>the</strong><br />

<strong>in</strong>tuitions of <strong>the</strong> authors. We designed an experiment to verify <strong>the</strong> alleged differences between <strong>the</strong><br />

various (local) l<strong>in</strong>guistic contexts and to test <strong>the</strong> predictions of <strong>the</strong> <strong>the</strong>ories. We found that <strong>the</strong><br />

situation is more complex than appears from any of <strong>the</strong> above papers. Both topicality and<br />

verifiability is a factor <strong>in</strong> its own right. We also discuss how to reconcile <strong>the</strong>se seem<strong>in</strong>gly divergent<br />

factors. We dist<strong>in</strong>guish two concepts: topic, what <strong>the</strong> sentence is pragmatically about, and what we<br />

call pivot, which is <strong>the</strong> constituent <strong>in</strong> <strong>the</strong> sentence based on which <strong>the</strong> sentence is verified. We<br />

propose, follow<strong>in</strong>g Lasersohn (1993) and von F<strong>in</strong>tel (2004), that verifiability is <strong>the</strong> primary factor<br />

that <strong>in</strong>for<strong>ms</strong> pragmatic truth value <strong>in</strong>tuitions: <strong>in</strong> particular we suggest that sentences are verified<br />

based on <strong>the</strong>ir pivots. The topic-effect noticed by Strawson and Re<strong>in</strong>hart comes about because<br />

topics are default pivots. Overall, our data suggests <strong>the</strong> follow<strong>in</strong>g picture: We should comb<strong>in</strong>e <strong>the</strong><br />

pragmatic <strong>in</strong>sights of <strong>the</strong> various Strawsonian <strong>the</strong>ories with a bivalent, but presuppositional<br />

approach to def<strong>in</strong>ite descriptions, such as that of Stalnaker (1974, 1978).<br />

3. The experimental setup We tested 33 native speakers of English (mostly British English),<br />

aged 22-55. The participants first read <strong>in</strong>structions on <strong>the</strong> computer screen, reproduced <strong>in</strong> (2).<br />

(2) In this experiment, statements will appear on your screen. If you th<strong>in</strong>k a statement is true, you should click on<br />

<strong>the</strong> 'TRUE' button. If you th<strong>in</strong>k a statement is false, you should click on <strong>the</strong> 'FALSE' button. Sometimes,<br />

it may happen that you cannot decide. In those cases, you should click on <strong>the</strong> 'CAN'T SAY' button. Please<br />

do not dwell on your decision for too long. There is no right or wrong answer!<br />

After a short practice session, <strong>the</strong>y were left alone with a program which presented <strong>the</strong> ite<strong>ms</strong><br />

described below one by one <strong>in</strong> pseudo-random order, 253 <strong>in</strong> total. An example of an item:<br />

The k<strong>in</strong>g of France is bald.<br />

FALSE CAN’T SAY TRUE<br />

1


The test ite<strong>ms</strong> were created by plac<strong>in</strong>g 8 def<strong>in</strong>ite descriptions that lack referents (listed <strong>in</strong> (3)) <strong>in</strong> 11<br />

sentential contexts, called test conditions, illustrated with ite<strong>ms</strong> <strong>in</strong>volv<strong>in</strong>g <strong>the</strong> k<strong>in</strong>g of France <strong>in</strong> (4).<br />

(3) <strong>the</strong> k<strong>in</strong>g of France; <strong>the</strong> emperor of Canada; <strong>the</strong> Pope’s wife; Pr<strong>in</strong>cess Diana’s daughter; <strong>the</strong> beaches of<br />

Birm<strong>in</strong>gham; <strong>the</strong> Belgian ra<strong>in</strong>forest; <strong>the</strong> coral reefs of Brighton; <strong>the</strong> volcanoes of Kent.<br />

(4) C(ondition) 0 [ref. failure NP, no extra NP] The k<strong>in</strong>g of France is bald.<br />

C1 [no presupposition] France has a k<strong>in</strong>g and he is bald.<br />

C2 [von F<strong>in</strong>tel, <strong>in</strong>dep. unknown NP] The k<strong>in</strong>g of France is on a state visit to Australia this week.<br />

C3 [Lasersohn, <strong>in</strong>dep. known NP] The k<strong>in</strong>g of France is married to Carla Bruni.<br />

C4 [ref. failure NP <strong>in</strong> topic] The k<strong>in</strong>g of France, he was <strong>in</strong>vited to have d<strong>in</strong>ner with Sarkozy.<br />

C5 [ref. failure NP not <strong>in</strong> topic] Sarkozy, he was <strong>in</strong>vited to have d<strong>in</strong>ner with <strong>the</strong> k<strong>in</strong>g of France.<br />

C6 [negation of 0] The k<strong>in</strong>g of France isn't bald.<br />

C7 [negation of 2] The k<strong>in</strong>g of France is not on a state visit to Australia this week.<br />

C8 [negation of 3] The k<strong>in</strong>g of France is not married to Carla Bruni.<br />

C9 [negation of 4] The k<strong>in</strong>g of France, he wasn’t <strong>in</strong>vited to have d<strong>in</strong>ner with Sarkozy.<br />

C10 [negation of 5] Sarkozy, he wasn’t <strong>in</strong>vited to have d<strong>in</strong>ner with <strong>the</strong> k<strong>in</strong>g of France.<br />

C(ondition) 1 was <strong>in</strong>cluded to test that <strong>the</strong> participants did know that our tested def<strong>in</strong>ite<br />

descriptions were referentially challenged. The difference between C2 and C3 (and similarly<br />

between C7 and C8) was that <strong>in</strong> <strong>the</strong> latter <strong>the</strong> participants were expected to have knowledge about<br />

<strong>the</strong> relevant properties of <strong>the</strong> <strong>in</strong>dependent NP <strong>in</strong> <strong>the</strong> sentence (Carla Bruni, <strong>in</strong> C3 and C8 <strong>in</strong> (7)—<br />

this we call Lasersohn-effect), but <strong>in</strong> <strong>the</strong> former <strong>the</strong>y were expected not to have such knowledge<br />

(Australia, <strong>in</strong> C2 and C7 <strong>in</strong> (7)—this we call <strong>the</strong> von F<strong>in</strong>tel-effect). The item list also conta<strong>in</strong>ed 165<br />

filler sentences, 8 of which were controls for <strong>the</strong> ite<strong>ms</strong> <strong>in</strong> C3 and C8, i.e. <strong>the</strong> conditions that tested<br />

for <strong>the</strong> Lasersohn-effect. E.g., for C8 <strong>in</strong> (7), <strong>the</strong> control item was Carla Bruni is married to Sarkozy.<br />

4. Results We compared <strong>the</strong> proportion (%) at which subjects replied ‘False’ to <strong>the</strong> test ite<strong>ms</strong> <strong>in</strong><br />

Conditions 0-5 with ANOVA, and also compared <strong>the</strong> proportion (%) at which subjects replied<br />

‘True’ to <strong>the</strong> test ite<strong>ms</strong> <strong>in</strong> Conditions 6-10. We found only weak or nearly significant differences<br />

between any of <strong>the</strong> conditions 0-5, i.e. <strong>the</strong> positive conditions; our subjects said ‘False’ to most of<br />

<strong>the</strong>se most of <strong>the</strong> time. However, we found clear significant differences among <strong>the</strong> negative<br />

conditions: As Figure 1 shows, we found that speakers responded ‘True’ at a significantly higher<br />

proportion to <strong>the</strong> ite<strong>ms</strong> <strong>in</strong> C7, C8, C9 and C10 than to <strong>the</strong> ite<strong>ms</strong> <strong>in</strong> C6 (p


Stavroula Alexandropoulou & Bert le Bruyn (Utrecht)<br />

Inalienable possession: a semantic take<br />

In most analyses of <strong>in</strong>alienable possession <strong>in</strong> French (see 1) it is standard to assume that <strong>the</strong><br />

def<strong>in</strong>ite article is simply not <strong>the</strong>re (e.g. Vergnaud & Zubizarreta 1992) or that it is some k<strong>in</strong>d of<br />

pronoun (e.g. Guéron 1985). We ma<strong>in</strong>ta<strong>in</strong> that expletive analyses should be a last resort option<br />

and that pronom<strong>in</strong>al analyses only make sense if we can show that <strong>the</strong>re is some k<strong>in</strong>d of<br />

agreement between <strong>the</strong> anaphor and its antecedent. Given that this last condition is not met (see<br />

1) we propose an alternative semantic analysis under <strong>the</strong> assumption that <strong>the</strong> def<strong>in</strong>ite article is<br />

what it looks like: a def<strong>in</strong>ite article.<br />

(1) Je 1st person, masc/fem lève la fem ma<strong>in</strong>. (2) ??Jean lève sa ma<strong>in</strong>.<br />

I raise <strong>the</strong> hand Jean raises his hand<br />

‘I raise my hand’<br />

From a semantic viewpo<strong>in</strong>t, we face two challenges: (i) How come we can use a def<strong>in</strong>ite article <strong>in</strong><br />

this construction (despite <strong>the</strong> non-uniqueness of hands)?; (ii) Why don’t we use a possessive<br />

pronoun ra<strong>the</strong>r than a def<strong>in</strong>ite article (see 2) ? Given that <strong>the</strong> first challenge has been treated for<br />

o<strong>the</strong>r relational uses of <strong>the</strong> def<strong>in</strong>ite article (see e.g. Barker <strong>in</strong> press) we focus on (ii). We propose an<br />

analysis with a m<strong>in</strong>imum of stipulations that derives <strong>the</strong> facts for French as well as for English.<br />

The key to understand<strong>in</strong>g <strong>in</strong>alienable possession is that <strong>the</strong> possessum <strong>in</strong> this construction<br />

presupposes possession: a hand e.g. is always someone’s hand. This has two important<br />

consequences that lead to partially conflict<strong>in</strong>g requirements on <strong>the</strong> <strong>in</strong>alienable possession<br />

construction. We will exploit <strong>the</strong>se as constra<strong>in</strong>ts <strong>in</strong> an OT account. The first consequence is that<br />

an <strong>in</strong>alienable possessum is def<strong>in</strong>ed only if we know who <strong>the</strong> possessor is. We <strong>the</strong>n expect<br />

speakers to make sure that <strong>the</strong>re is no doubt about who <strong>the</strong> possessor is. The most<br />

straightforward way to do so is to use a possessive pronoun. This leads a speaker to prefer (2)<br />

over (1). The second consequence is that <strong>the</strong> overt mark<strong>in</strong>g of possession is superfluous: given<br />

that possession is presupposed, it is uneconomic to mark it explicitly with a possessive pronoun<br />

that – due to <strong>the</strong> fact that it imposes a b<strong>in</strong>d<strong>in</strong>g relation – is syntactically more complex than a<br />

def<strong>in</strong>ite article. We consequently expect speakers to avoid us<strong>in</strong>g overt possessive pronouns and<br />

to prefer (1) over (2).<br />

In (3) we re<strong>in</strong>terpret <strong>the</strong> two consequences we have <strong>in</strong>troduced as an OT faithfulness and an OT<br />

markedness constra<strong>in</strong>t.<br />

(3) FaithPoss(essor):Make sure an <strong>in</strong>alienable possessum has a determ<strong>in</strong>ed possessor.<br />

D(on’t)U(se)P(ossessive)P(ronouns): Don’t use possessive pronouns with <strong>in</strong>alienable<br />

possessum nouns<br />

The <strong>in</strong>terpretation of DUPP is unproblematic but FaithPoss does need fur<strong>the</strong>r explanation. We<br />

def<strong>in</strong>e <strong>the</strong> notion of determ<strong>in</strong>ed possessor <strong>in</strong> DRT:<br />

(4) A possession relation have(x,y) conta<strong>in</strong>s a determ<strong>in</strong>ed possessor y iff for every f that<br />

embeds <strong>the</strong> <strong>in</strong>put DRS K, it holds that for every f’, f’’ that extend f, f’(y)=f’’(y) and<br />

have(f’(x),f’(y))=have(f’’(x),f’’(y)) (cf. Farkas 2002 on determ<strong>in</strong>ed reference).<br />

Above, we <strong>in</strong>dicated one way of satisfy<strong>in</strong>g FaithPoss, viz. to use possessive pronouns. Here, we<br />

discuss <strong>the</strong> two o<strong>the</strong>r ways we know of. Note that nei<strong>the</strong>r <strong>in</strong>volves explicit mark<strong>in</strong>g of<br />

possession. We <strong>in</strong>troduce <strong>the</strong> first on <strong>the</strong> basis of (5) and make <strong>the</strong> assumptions <strong>in</strong> (6):<br />

(5) Marie a frappé Jean sur la ma<strong>in</strong>.<br />

Marie has hit Jean on <strong>the</strong> hand<br />

(6) a. sur <strong>in</strong> comb<strong>in</strong>ation with frapper refers ei<strong>the</strong>r to <strong>the</strong> location or specifies <strong>the</strong> goal of <strong>the</strong><br />

action of hitt<strong>in</strong>g (semantics of <strong>the</strong> preposition)<br />

b. if hitt<strong>in</strong>g an object A and hitt<strong>in</strong>g an object B form one and <strong>the</strong> same event, <strong>the</strong>n it has<br />

to be <strong>the</strong> case that ei<strong>the</strong>r A is part of/equal to B or B is part of/equal to A (semantics of<br />

<strong>the</strong> verb)<br />

c. hands are parts of humans (semantics of <strong>in</strong>alienable possession)<br />

From <strong>the</strong> assumptions <strong>in</strong> (6) it follows that <strong>the</strong> hand <strong>in</strong> (5) has to be John’s. More precisely, on<br />

<strong>the</strong> goal read<strong>in</strong>g of sur <strong>the</strong> hitt<strong>in</strong>g of John is associated with <strong>the</strong> hitt<strong>in</strong>g of <strong>the</strong> hand, while, by b.


and c., it is ensured that <strong>the</strong>re is a part-of relation between <strong>the</strong> hand and John. Similar analyses<br />

can be presented for several preposition/verb/<strong>in</strong>alienable possessum noun triplets (see Koenig<br />

1999 for a similar yet less flexible analysis).<br />

The second way FaithPoss can be satisfied without an overt possessive pronoun is through <strong>the</strong><br />

use of datives that can and sometimes must undergo so-called possessor rais<strong>in</strong>g (see Szabolsci<br />

1983, Nicol 1997 and Landau 1999). This is illustrated <strong>in</strong> (7):<br />

(7) Le frère k de Jean l lui l a pris la ma<strong>in</strong> t l.<br />

The bro<strong>the</strong>r of John to_him has taken <strong>the</strong> hand<br />

‘John’s bro<strong>the</strong>r took his hand’<br />

The reason<strong>in</strong>g here is similar to <strong>the</strong> one we made on <strong>the</strong> basis of (6): tak<strong>in</strong>g <strong>the</strong> hand to someone<br />

entails that <strong>the</strong> hand <strong>in</strong> question has to belong to this someone.<br />

We are now ready to put our constra<strong>in</strong>ts to work <strong>in</strong> an OT analysis and to discuss its predictions.<br />

Note that <strong>the</strong> facts we discuss are <strong>the</strong> same as <strong>the</strong> ones that have been discussed <strong>in</strong> <strong>the</strong> syntactic<br />

literature as basic evidence <strong>in</strong> favour of an analysis <strong>in</strong>volv<strong>in</strong>g b<strong>in</strong>d<strong>in</strong>g.<br />

Given that we have two constra<strong>in</strong>ts, we expect to have two types of languages: one <strong>in</strong> which<br />

FaithPoss is ranked below DUPP and one <strong>in</strong> which <strong>the</strong> reverse is <strong>the</strong> case. In<br />

DUPP>>FaithPoss languages we expect <strong>in</strong>alienable possessum nouns – wherever possible – to<br />

occur <strong>in</strong> prepositional phrases or with a dative, satisfy<strong>in</strong>g FaithPoss without violat<strong>in</strong>g DUPP.<br />

Wherever this is impossible, we expect <strong>the</strong>m to occur with <strong>the</strong> def<strong>in</strong>ite article, violat<strong>in</strong>g FaithPoss<br />

but satisfy<strong>in</strong>g DUPP. This derives <strong>the</strong> facts for French <strong>in</strong> (1), (7), (8), (9) and (10).<br />

(8) *Il se lève la ma<strong>in</strong>. (9) *Le frère k de Jean l a pris la ma<strong>in</strong> t l.<br />

He to_hi<strong>ms</strong>elf raises <strong>the</strong> hand The bro<strong>the</strong>r of John to_him has taken <strong>the</strong> hand<br />

(10) Le frère k de Jean l l l’a pris par la ma<strong>in</strong> l.<br />

The bro<strong>the</strong>r of John him has taken by <strong>the</strong> hand<br />

The acceptability of (1) follows from <strong>the</strong> unacceptability of (8) and <strong>the</strong> unavailability of a<br />

prepositional variant, <strong>the</strong> unacceptability of (9) follows from <strong>the</strong> availability of (7) and (10). The<br />

crucial <strong>in</strong>sight we propose here is that <strong>the</strong> acceptability of <strong>the</strong> def<strong>in</strong>ite article <strong>in</strong> <strong>the</strong> <strong>in</strong>alienable<br />

possession construction depends on <strong>the</strong> availability of o<strong>the</strong>r constructions. We don’t deny that<br />

<strong>the</strong>re are <strong>in</strong>terest<strong>in</strong>g generalizations that govern <strong>the</strong> use of <strong>the</strong>se o<strong>the</strong>r constructions but we do<br />

claim that <strong>the</strong> <strong>in</strong>tricacies that come with <strong>the</strong>se should not be <strong>in</strong>cluded <strong>in</strong> <strong>the</strong> analysis of<br />

<strong>in</strong>alienable possession as <strong>the</strong>y might ultimately reside <strong>in</strong> <strong>the</strong> lexicon. A clear argument <strong>in</strong> favour<br />

of this strategy comes from <strong>the</strong> fact that similar verbs <strong>in</strong> languages that generally allow for<br />

<strong>in</strong>alienable possession with <strong>the</strong> def<strong>in</strong>ite article need not allow for similar constructions. Compare<br />

<strong>in</strong> this respect <strong>the</strong> marg<strong>in</strong>ality of French la tête me fait mal (‘<strong>the</strong> head to me makes ache’) to <strong>the</strong> full<br />

productivity of Spanish me duele la cabeza (‘to me makes_ache <strong>the</strong> head’).<br />

In FaithPoss>>DUPP languages we expect <strong>in</strong>alienable possessum nouns – wherever possible –<br />

to occur <strong>in</strong> prepositional phrases or with a dative, satisfy<strong>in</strong>g FaithPoss without violat<strong>in</strong>g DUPP.<br />

Wherever this is impossible, we expect <strong>the</strong>m to take a possessive pronoun, violat<strong>in</strong>g DUPP but<br />

satisfy<strong>in</strong>g FaithPoss. This straightforwardly derives <strong>the</strong> facts for English:<br />

(11) Mary k hit John l on <strong>the</strong> hand l. (12) *John k raised <strong>the</strong> hand k.<br />

The acceptability of (11) follows from <strong>the</strong> fact that both FaithPoss and DUPP are satisfied, <strong>the</strong><br />

unacceptability of (12) follows from a violation of FaithPoss.<br />

In this paper, we have proposed an OT analysis that comb<strong>in</strong>es <strong>in</strong>sights about <strong>the</strong> semantics of<br />

verbs, prepositions, datives and <strong>in</strong>alienable possessum nouns with a straigthforward markedness<br />

(DUPP) and faithfulness (FaithPoss) constra<strong>in</strong>t. The result is an elegant analysis with a m<strong>in</strong>imum<br />

of stipulations and a maximum of data coverage. We <strong>in</strong> particular showed how we can avoid<br />

stipulat<strong>in</strong>g that <strong>the</strong> def<strong>in</strong>ite article functions as an expletive or a pronoun <strong>in</strong> disguise.<br />

References | Barker (<strong>in</strong> press) ‘Possessive weak def<strong>in</strong>ites’ | Farkas (2002), Journal of Semantics 19:<br />

213-243. | Guéron (1985), <strong>in</strong> Guéron et al. (eds.), Grammatical Representation, Dordrecht: Foris. |<br />

Koenig (1999), NLLT 17: 219-265. | Landau (1999), L<strong>in</strong>gua 107: 1-37. | Nicol (1997), PhD<br />

dissertation, Paris X. | Szabolcsi (1983), The L<strong>in</strong>guistic Review 3: 89-102. |Vergnaud & Zubizarreta<br />

(1992), LI 213: 595-652.


Artemis Alexiadou (Stuttgart) & Elena Anagnostopoulou (Crete)<br />

Verb mean<strong>in</strong>g, local context, and <strong>the</strong> syntax of roots <strong>in</strong> alternations<br />

1. While it is uncontroversial that extra-grammatical context (l<strong>in</strong>guistic and non-l<strong>in</strong>guistic)<br />

can <strong>in</strong>fluence <strong>the</strong> form and mean<strong>in</strong>g of sentences, research <strong>in</strong> <strong>the</strong> area of verbal alternations<br />

has revealed that <strong>the</strong>re is ano<strong>the</strong>r notion of context important for mean<strong>in</strong>g, which we call here<br />

local context. Mean<strong>in</strong>g of verbs is computed on <strong>the</strong> basis of a ‘core element’ (<strong>the</strong> ‘root’)<br />

which is <strong>in</strong>tegrated <strong>in</strong> a structural template (local context). On this view, verb classes are sets<br />

of verbs shar<strong>in</strong>g a local context while <strong>the</strong> idiosyncratic properties of a given root determ<strong>in</strong>e<br />

whe<strong>the</strong>r and how this root can be <strong>in</strong>tegrated <strong>in</strong> a structural template. This basic approach has<br />

been advocated both <strong>in</strong> models of lexical semantic decomposition (e.g. Rappaport Hovav &<br />

Lev<strong>in</strong> RH&L 1998) and <strong>in</strong> models of morphological-syntactic decomposition (e.g. work <strong>in</strong><br />

Distributed Morphology (DM) follow<strong>in</strong>g Marantz 1997, 2001). From <strong>the</strong> DM perspective, a<br />

syntactic representation is <strong>the</strong>n <strong>in</strong>terpreted <strong>in</strong> consultation with <strong>the</strong> encyclopedia. Crucially,<br />

encyclopedia entries <strong>in</strong>terpret structures <strong>in</strong> very specific contexts and <strong>in</strong> comb<strong>in</strong>ations. A<br />

question aris<strong>in</strong>g <strong>in</strong> such models is to what extent <strong>the</strong> determ<strong>in</strong>ants of argument realization are<br />

part of <strong>the</strong> root <strong>in</strong>formation and to what extent <strong>the</strong>y are part of <strong>the</strong> event/syntactic template<br />

(see Borer 2005 for a radically exo-skeletal view). In this paper, we explore how <strong>the</strong>se two<br />

mean<strong>in</strong>g components <strong>in</strong>teract on <strong>the</strong> basis of a case study, namely <strong>the</strong> clear-alternation. We<br />

argue that object alternations relate to whe<strong>the</strong>r verbs are classified as manner or result, which<br />

depends on <strong>the</strong> ontological types of <strong>the</strong> morphological roots <strong>the</strong>y are based on and <strong>the</strong> local<br />

contexts permissible for different ontological types of roots.<br />

2. In English, locative verbs of plac<strong>in</strong>g (spray, load) and detach<strong>in</strong>g (clear) alternate between<br />

two frames (1), illustrated <strong>in</strong> (2), see Lev<strong>in</strong> (1993), L& RH (1991):<br />

(1) a. Frame A: DPAgent V DPStuff PPLoc b. Frame B: DPAgent V DPLoc PPStuff<br />

(2) a. Lucy sprayed <strong>the</strong> pa<strong>in</strong>t on <strong>the</strong> wall b. Lucy sprayed <strong>the</strong> wall with pa<strong>in</strong>t<br />

c. Henry cleared dishes from <strong>the</strong> table d. Henry cleared <strong>the</strong> table of dishes<br />

Many locative verbs do not allow <strong>the</strong> alternation, and choose ei<strong>the</strong>r only Frame A (e.g. steal<br />

Change of Location COL) or Frame B (e.g. cover, Change of State COS). The <strong>in</strong>terpretation<br />

of <strong>the</strong> two frames is given <strong>in</strong> (3):<br />

(3) a. COL <strong>in</strong>terpretation (Frame A): [X CAUSE [Y BECOME [AT Z/AWAY FROM Z]]]<br />

b. COS <strong>in</strong>terpretation (Frame B): [X CAUSE [Z BECOME [WITH Y/WITHOUT Y]]]<br />

The question thus is what underlies <strong>the</strong> dist<strong>in</strong>ction between alternat<strong>in</strong>g and non-alternat<strong>in</strong>g<br />

verbs and why <strong>the</strong> latter choose <strong>the</strong> COL or <strong>the</strong> COS frame. In <strong>the</strong> recent literature, it has been<br />

proposed that <strong>the</strong> (un-)availability of object alternations can be understood <strong>in</strong> ter<strong>ms</strong> of <strong>the</strong><br />

Manner vs. Result hypo<strong>the</strong>sis (RH&L 1998, 2008) accord<strong>in</strong>g to which: i) Verbs lexicalize<br />

ei<strong>the</strong>r manner or result (manner or result, respectively, are entailed by all uses of verbs). ii)<br />

Verbs alternate when <strong>the</strong>y lexicalize manner and <strong>the</strong>y don’t when <strong>the</strong>y lexicalize result.<br />

3. Mavropoulou, Moschou, Tsikala & Anagnostopoulou (MMTA) (2011) show that Greek<br />

has both <strong>the</strong> spray- and <strong>the</strong> clear- alternation:<br />

(4) a. psekasa boja ston tiho b. psekasa ton tiho me boja<br />

sprayed-1sg pa<strong>in</strong>t to <strong>the</strong> wall sprayed-1sg <strong>the</strong> wall with pa<strong>in</strong>t<br />

c. katharisa ta psihula apo to trapezi d. katharisa to trapezi apo psihula<br />

cleaned-1sg <strong>the</strong> crumbles from <strong>the</strong> table cleaned-1sg <strong>the</strong> table from crumbles<br />

These authors provide lists of locative verbs that alternate <strong>in</strong> Greek, and establish <strong>the</strong><br />

follow<strong>in</strong>g: i) <strong>in</strong> Greek <strong>the</strong> alternation is more systematic with verbs of detach<strong>in</strong>g. This<br />

property is partially attributed to <strong>the</strong> fact that <strong>the</strong> preposition used with this class is apo<br />

‘from’, which is clearly directional (unlike se ‘to’) and can easily comb<strong>in</strong>e with manner roots<br />

express<strong>in</strong>g COL (Frame A). ii) More verbs of detach<strong>in</strong>g alternate <strong>in</strong> Greek than <strong>in</strong> English,<br />

where accord<strong>in</strong>g to L&RH (1991) only four verbs alternate (but o<strong>the</strong>rs claim that several<br />

wipe- verbs may alternate as well; see Lev<strong>in</strong> 1993). In <strong>the</strong>se two respects, Greek is similar to<br />

Hebrew, as described by Segal & Landau (S&L 2009). S&L (2009) argue that alternat<strong>in</strong>g


clear verbs are not specified for <strong>the</strong> type of change <strong>in</strong>volved (<strong>the</strong>y entail nei<strong>the</strong>r COL nor<br />

COS), while those that do not alternate are specified ei<strong>the</strong>r entail<strong>in</strong>g COL and appear <strong>in</strong><br />

Frame A or COS (and appear <strong>in</strong> Frame B). MMTA (2011) show that this holds <strong>in</strong> Greek too.<br />

4. The morphological make-up of Greek verbs of detachment strongly suggests that a large<br />

class of alternat<strong>in</strong>g verbs (i.e. verbs not specified for COL/COS) are manner verbs and nonalternat<strong>in</strong>g<br />

ones (ei<strong>the</strong>r only COL or only COS) are result verbs, as expected by <strong>the</strong> Manner<br />

vs. Result Hypo<strong>the</strong>sis. Alternat<strong>in</strong>g verbs typically <strong>in</strong>volve manner/<strong>in</strong>strument roots and a<br />

verbalizer e.g. -on- <strong>in</strong> (5a) (<strong>the</strong>se correspond to <strong>the</strong> English wipe-class). Verbs only allow<strong>in</strong>g<br />

COL conta<strong>in</strong> prepositions/prefixes/particles that overtly encode location (5b); verbs only<br />

allow<strong>in</strong>g COS are based on an adjective, plus a verbalizer, thus overtly encod<strong>in</strong>g state (5c).<br />

(5) a. sider-on-o lit. iron-v-1sg ‘iron’ b. af-er-o lit. from-lift-1sg ‘remove’<br />

c. atho-on-o lit. <strong>in</strong>nocent-v-1sg ‘acquit’<br />

Apply<strong>in</strong>g a series of manner vs. result diagnostics (see Koontz Garboden & Beavers 2011)<br />

confir<strong>ms</strong> that (5a)-type verbs are manner verbs and (5b,c)-type verbs are result verbs. These<br />

can be morphologically decomposed by us<strong>in</strong>g <strong>the</strong> tools of DM: verbs consist of categoryneutral,<br />

idiosyncratic roots which are merged with categoriz<strong>in</strong>g heads (Marantz 2001) ei<strong>the</strong>r<br />

as <strong>the</strong>ir modifiers, direct Merge, or as <strong>the</strong>ir complements (Embick 2004; Harley 2005).<br />

(6) a. modifiers of v, direct Merge: [v v √ ] b. complements of v [v √ v]<br />

For Embick, direct merge specifies <strong>the</strong> means component of <strong>the</strong> complex predicate, and (6a)<br />

can feed secondary resultative predication. Thus <strong>the</strong> type of merge is sensitive to <strong>the</strong> manner<br />

vs. result/state classification of roots: manner roots merge as modifiers of v, state roots merge<br />

as complements of v. Comb<strong>in</strong><strong>in</strong>g RH&L’s <strong>in</strong>sights with Embick’s structures, we propose that<br />

direct merge applies to manner/<strong>in</strong>strument roots with alternat<strong>in</strong>g verbs of <strong>the</strong> (5a)-type. S<strong>in</strong>ce<br />

<strong>the</strong>se roots merge as modifiers of v, <strong>the</strong>y can be <strong>in</strong>volved <strong>in</strong> resultative secondary predication,<br />

yield<strong>in</strong>g <strong>the</strong> COL and <strong>the</strong> COS frames. Follow<strong>in</strong>g Hale & Keyser (2002), S&L (2009) and<br />

o<strong>the</strong>rs, we assume that <strong>in</strong> COL, a locative (source) PP is <strong>the</strong> sister of v and <strong>the</strong> stuff DP as <strong>the</strong><br />

specifier of P (7). In COS, a ResultP is <strong>the</strong> sister of v and <strong>the</strong> DPlocation is <strong>the</strong> specifier of v (8):<br />

(7) [v v � [PP DPstuff [P DPlocation ]]] (8) [vP DPlocation [v [ResultP]]]<br />

Turn<strong>in</strong>g to non-alternat<strong>in</strong>g verbs, Greek verbs show<strong>in</strong>g only frame A conta<strong>in</strong> prefixes, drawn<br />

from <strong>the</strong> prepositional <strong>in</strong>ventory. Most of <strong>the</strong>m belong to <strong>the</strong> <strong>in</strong>ventory of Classical Greek and<br />

are no longer productive. They cannot be separated from <strong>the</strong> verbal stem, predom<strong>in</strong>antly have<br />

spatial mean<strong>in</strong>gs, and <strong>in</strong> comb<strong>in</strong>ation with <strong>the</strong> morphological root give <strong>the</strong> mean<strong>in</strong>g of what<br />

L&RH label “root” <strong>in</strong> <strong>the</strong> lexical semantic sense, e.g. ek- top- iz-o ‘from -place-verbalizer-<br />

1sg.’. We propose that <strong>the</strong> prefixes ‘lexicalize’ path and select a PlaceP complement<br />

(Svenonius 2008), lexicalized by �top <strong>in</strong> (9). The apo phrase is <strong>in</strong> Spec,Path, and <strong>the</strong>re is a<br />

Spec-head Agreement relationship, <strong>in</strong> this case Source. Support for this analysis comes from<br />

<strong>the</strong> fact that when Path = Goal, <strong>the</strong> verbal prefix is ‘eis’ (to) and <strong>the</strong> Goal-PP is a se-‘to’ PP.<br />

(9) [vP iz [PathP ek [PlaceP �top ]]]<br />

Greek non-alternat<strong>in</strong>g verbs show<strong>in</strong>g only COS are built on <strong>the</strong> basis of an adjectival base.<br />

The set of adjectives related to <strong>the</strong>se verbs can each take a complement expressed by means<br />

of an apo phrase <strong>in</strong>troduc<strong>in</strong>g <strong>the</strong> stuff argument (adio apo nero ‘empty from water’), which is<br />

<strong>the</strong>n <strong>in</strong>herited by <strong>the</strong> verb derived from <strong>the</strong>se adjectives. Thus <strong>the</strong> verbs appear only <strong>in</strong> COS.<br />

5. We provided evidence that verbal alternations depend on both <strong>the</strong> ontological types of <strong>the</strong><br />

verbal roots <strong>in</strong>volved and on <strong>the</strong> local contexts permissible for <strong>the</strong>se different types. This <strong>in</strong><br />

turn suggests that a) roots do have some mean<strong>in</strong>g <strong>in</strong> isolation (contra e.g. Acquaviva 2009,<br />

Borer 2008, de Belder & van Craenenbroeck 2011) which consists of a limited number of<br />

contrastive properties (manner, state and place <strong>in</strong> our case), b) <strong>the</strong>se dist<strong>in</strong>ctions <strong>in</strong>fluence <strong>the</strong><br />

syntactic merger possibilities of roots, c) <strong>the</strong> syntactic template roots are <strong>in</strong>serted <strong>in</strong> fully<br />

specifies <strong>the</strong> realization of verbal arguments, and d) mean<strong>in</strong>g is assigned by <strong>the</strong> encyclopedia<br />

that calculates both <strong>the</strong> <strong>in</strong>put of <strong>the</strong> root and that of <strong>the</strong> syntactic template.


Artemis Alexiadou & Florian Schäfer (Stuttgart)<br />

Naturally reflexive verbs revisited: relations between syntax, semantics, and <strong>the</strong> lexicon<br />

1. In <strong>the</strong> literature, naturally reflexive verbs ((NRVs) e.g. wash, Kemmer 1993) have been <strong>the</strong><br />

topic of much controversial discussion. Here we focus on one particular aspect of this debate,<br />

namely <strong>the</strong> question whe<strong>the</strong>r or not NRVs <strong>in</strong>clude an empty category at some level of<br />

representation, ei<strong>the</strong>r a null reflexive or a trace/copy, <strong>in</strong> languages like English and Greek. In<br />

<strong>the</strong>se languages, NRVs are, at <strong>the</strong> surface, <strong>in</strong>transitive ei<strong>the</strong>r because no overt reflexive form<br />

is present (English: John washed) or because, <strong>in</strong> addition, de-transitiviz<strong>in</strong>g morphology is<br />

present (Greek: O Janis plithike 'John washed-non-active-3sg'). As <strong>the</strong> class of verbs that are<br />

treated as naturally reflexive see<strong>ms</strong> quite stable across languages, a uniform analysis has been<br />

argued to be desirable (see e.g. Alexiadou & Doron 2011). In this paper, we argue for a nonuniform<br />

analysis of NRVs across languages (cf. Re<strong>in</strong>hart & Siloni 's 2004 lexicon vs. syntax<br />

parameter; see Decha<strong>in</strong>e & Wiltschko 2011 for discussion of heterogeneous approaches to<br />

reflexives). We first show that under closer <strong>in</strong>spection, English NRVs are syntactically<br />

<strong>in</strong>transitive, provid<strong>in</strong>g thus fur<strong>the</strong>r support for <strong>the</strong> view that <strong>the</strong>se predicates lack a<br />

syntactically projected object (Re<strong>in</strong>hart & Siloni 2004). While an <strong>in</strong>transitive analysis is<br />

available for English, it cannot extend to German, Dutch and Greek. In <strong>the</strong> former case<br />

(German, Dutch), this is so because a syntactically projected object is overtly present, <strong>in</strong> <strong>the</strong><br />

latter case (Greek) because NRVs, unlike ord<strong>in</strong>ary unergatives, show de-transitivization<br />

morphology. We thus propose a tri-partition of <strong>the</strong> way languages resolve <strong>the</strong> requirement<br />

that <strong>the</strong> two <strong>the</strong>matic roles associated with NRVs make reference to <strong>the</strong> same entity: a) <strong>in</strong><br />

English, this is done at <strong>the</strong> conceptual <strong>in</strong>terface, b) <strong>in</strong> German and Dutch, this is done via<br />

b<strong>in</strong>d<strong>in</strong>g of an object anaphor, and c) <strong>in</strong> Greek, this is done via movement <strong>in</strong>to a <strong>the</strong>ta-position<br />

(syntactic bundl<strong>in</strong>g). This reflects <strong>the</strong> type of morphology we f<strong>in</strong>d <strong>in</strong> <strong>the</strong> three language<br />

groups, zero/none, reflexive pronoun, and an affix, respectively. Our analysis correctly<br />

predicts <strong>the</strong> productivity of reflexivization noted for Greek, but not for e.g. English, which is<br />

not captured under <strong>the</strong> homogeneous approach.<br />

2. Some authors argued that NRVs are actually transitive, i.e. <strong>the</strong>y select a specific reflexive<br />

null morpheme, which is overt <strong>in</strong> languages like German (Hans wäscht sich ‘Hans washes<br />

hi<strong>ms</strong>elf’), see Safir (2004) and references <strong>the</strong>re<strong>in</strong>, Marantz (2009). Support for this comes<br />

from <strong>the</strong> observation that <strong>the</strong>se predicates do have transitive construals, e.g. John washed<br />

Mary/John didn't wash Mary but hi<strong>ms</strong>elf. O<strong>the</strong>rs proposed that NRVs are actually<br />

unaccusatives, <strong>the</strong> <strong>the</strong>me argument is a derived subject and leaves a trace/copy <strong>in</strong> its <strong>the</strong>matic<br />

position (Marantz 1984, and subsequent work, McG<strong>in</strong>nis 1998, Embick 2004). Support for<br />

this analysis comes from languages such as Greek that mark <strong>the</strong>ir NRVs with <strong>the</strong> same nonactive<br />

morphology as <strong>the</strong>ir passives or uncontroversial unaccusatives. Both <strong>the</strong>se analyses are<br />

problematic for English. Sells, Zaenen & Zec (1987) showed that NRVs behave unlike<br />

predicates that take an overt reflexive with respect to ‘strict’ and ‘sloppy’ read<strong>in</strong>gs. (1), with<br />

an overt reflexive, is ambiguous between a strict and sloppy read<strong>in</strong>g. (2), an NRV, has only<br />

<strong>the</strong> sloppy read<strong>in</strong>g, suggest<strong>in</strong>g that <strong>the</strong> predicate is semantically <strong>in</strong>transitive:<br />

(1) John defends hi<strong>ms</strong>elf better than Peter<br />

a. John defends hi<strong>ms</strong>elf better than Peter defends hi<strong>ms</strong>elf sloppy<br />

b. Johni defends hi<strong>ms</strong>elf better that Peter defends himi strict<br />

(2) John washes more than George.<br />

a. John washes hi<strong>ms</strong>elf more than George washes hi<strong>ms</strong>elf. sloppy<br />

b. John i washes hi<strong>ms</strong>elf more <strong>the</strong>n George washes him i *strict<br />

(1) has a fur<strong>the</strong>r read<strong>in</strong>g, <strong>the</strong> so-called object comparison read<strong>in</strong>g (Johni defends hi<strong>ms</strong>elf<br />

better that he i defends Peter), which (2) lacks, as expected, see Dimitriadis & Que (2009).<br />

Re<strong>in</strong>hart & Siloni (2004) po<strong>in</strong>ted out that NRVs <strong>in</strong> English behave unlike one-place<br />

unaccusatives. As is well-known, agentive -er nom<strong>in</strong>als can only be derived from predicates<br />

with an external argument (Lev<strong>in</strong> & Rappaport Hovav 1992). Re<strong>in</strong>hart & Siloni show that


NRVs form agentive nom<strong>in</strong>als (3c), similar to unergatives (3a):<br />

(3) a. She runs so fast because she is an experienced runner.<br />

b. * She moves so gracefully because she is an experienced mover.<br />

c. She dresses slowly because she is an elegant dresser.<br />

Fur<strong>the</strong>r support for an unergative analysis comes from <strong>the</strong> application of o<strong>the</strong>r transitivity<br />

tests (Bresnan 1982, Kratzer 2004, Lev<strong>in</strong> 1999, Rappaport Hovav and Lev<strong>in</strong> 1998, 2008).<br />

Reduplication to get an iterative read<strong>in</strong>g, out-prefixation and ‘fake’ reflexives are possible<br />

only with non-core transitive (and unergative) verbs (4), but not with core-transitives (5). (6),<br />

an NRV, behaves like (4), <strong>in</strong> both its read<strong>in</strong>gs, <strong>the</strong> reflexive and <strong>the</strong> disjo<strong>in</strong>t one:<br />

(4) a. John run and run and run/John ate and ate. (5) a.*John broke and broke and broke<br />

b. John out-run/out-ate Mary. b. *John out-broke Mary.<br />

c. John run/ate hi<strong>ms</strong>elf dead. c. *John broke-hi<strong>ms</strong>elf dead.<br />

(6) a. John washed and washed and washed.<br />

b. John out-washed his sister.<br />

c. John washed hi<strong>ms</strong>elf raw.<br />

3. To expla<strong>in</strong> <strong>the</strong> behavior of English NRVs, it has been proposed that <strong>the</strong>se have two lexical<br />

entries, a reflexive and a disjo<strong>in</strong>t one (see e.g. Moulton 2005). Assum<strong>in</strong>g that <strong>the</strong> external<br />

argument is severed from <strong>the</strong> VP (Kratzer 1996), and is <strong>in</strong>troduced <strong>in</strong> Voice, it is not clear<br />

how to formally encode that such a lexical entry br<strong>in</strong>gs about a reflexive <strong>in</strong>terpretation.<br />

Fur<strong>the</strong>rmore, such a move see<strong>ms</strong> undesirable, as it would lead to an explosion of <strong>the</strong> lexicon.<br />

We showed that NRVs are similar to o<strong>the</strong>r non-core transitive predicates like eat. An option<br />

that is available for <strong>the</strong>se predicates, under <strong>the</strong> assumption that <strong>the</strong>matic roles belong to <strong>the</strong><br />

conceptual level (Cho<strong>ms</strong>ky 1995), is to <strong>in</strong>terpret <strong>the</strong>m as two-place predicates at <strong>the</strong> syntaxconceptual<br />

level <strong>in</strong>terface. This means that <strong>the</strong> <strong>in</strong>ternal argument is added post-syntactically.<br />

With verbs of consumption, <strong>the</strong> added argument is <strong>the</strong> most prototypical object for this<br />

predicate-class, an amount of food/fluid that one can consume. NRVs are def<strong>in</strong>ed as carry<strong>in</strong>g<br />

“<strong>in</strong>herent <strong>in</strong> <strong>the</strong>ir mean<strong>in</strong>g [...] <strong>the</strong> lack of expectation that <strong>the</strong> two semantic roles <strong>the</strong>y make<br />

reference to will refer to dist<strong>in</strong>ct entities” (Kemmer 1993:58). The most prototypical object is<br />

thus identical to <strong>the</strong> subject, and this is computed at <strong>the</strong> conceptual <strong>in</strong>terface.<br />

4. Dutch and German have an overt object for <strong>the</strong> reflexive <strong>in</strong>terpretation, and allow only <strong>the</strong><br />

disjo<strong>in</strong>t <strong>in</strong>terpretation to be resolved at <strong>the</strong> conceptual <strong>in</strong>terface. In <strong>the</strong>se languages, NRVs are<br />

syntactically transitive (e.g. Ste<strong>in</strong>bach 2002). As an <strong>in</strong>ternal argument is projected <strong>in</strong> <strong>the</strong><br />

syntax to be bound by <strong>the</strong> external one, <strong>the</strong>se predicates lack disjo<strong>in</strong>t read<strong>in</strong>gs. We relate this<br />

to <strong>the</strong> special (light) reflexives for<strong>ms</strong> used, an option unavailable <strong>in</strong> English.<br />

5. Nei<strong>the</strong>r <strong>the</strong> unergative nor <strong>the</strong> transitive analysis is applicable to Greek due to <strong>the</strong> presence<br />

of non-active morphology with NRVs. Embick (2004) proposed that NRVs comb<strong>in</strong>e with a<br />

Voice that is agentive, but lacks a specifier, and this is realized via non-active morphology at<br />

morphological structure. This cannot expla<strong>in</strong> <strong>the</strong> absence of disjo<strong>in</strong>t reference with reflexives,<br />

but not with passives. Doron (2003) argued that NRVs comb<strong>in</strong>e with a special k<strong>in</strong>d of Voice<br />

(<strong>the</strong> host of non-active morphology), which does not trigger <strong>the</strong> disjo<strong>in</strong>t reference effect. But<br />

this raises <strong>the</strong> question what regulates <strong>the</strong> cross-l<strong>in</strong>guistic distribution of this special Voice.<br />

We propose <strong>in</strong>stead that <strong>in</strong> Greek, <strong>the</strong> <strong>in</strong>ternal argument moves to Spec,Voice, comb<strong>in</strong><strong>in</strong>g<br />

two <strong>the</strong>matic roles (<strong>the</strong>me & agent). This expla<strong>in</strong>s Papangeli’s (2004) observation that <strong>the</strong>re<br />

are many newly formed reflexives <strong>in</strong> Greek (e.g. hapakonete = to give pills to hi<strong>ms</strong>elf,<br />

derived from <strong>the</strong> transitive verb via <strong>the</strong> addition of non-active morphology) that would be<br />

hard to expla<strong>in</strong>, if we assumed that only a closed set of verbs has a reflexive variant. We view<br />

non-active morphology as similar to be-selection <strong>in</strong> Romance (cf. Haider & R<strong>in</strong>dler-Schjerve<br />

1987): it is <strong>the</strong> morphological reflex of an agreement cha<strong>in</strong> between T and an element with<strong>in</strong><br />

VP, as <strong>in</strong> <strong>the</strong> case of passives and unaccusatives. With reflexives, T agrees with <strong>the</strong> whole<br />

cha<strong>in</strong> between <strong>the</strong> object and <strong>the</strong> subject position, lead<strong>in</strong>g aga<strong>in</strong> to non-active morphology.


Anke Assmann, Doreen Georgi, Fabian Heck, Gereon Müller & Philipp Weisser (Leipzig)<br />

Ergatives move too early<br />

Claim: We claim that <strong>the</strong> <strong>in</strong>ability of ergative arguments (DPerg) to undergo Ā-movement<br />

(wh-movement, focuss<strong>in</strong>g, relativization) <strong>in</strong> many morphologically ergative languages should<br />

not be brought about by restrictions on <strong>the</strong> movement of ergative DPs. Ra<strong>the</strong>r, we argue that<br />

movement of <strong>the</strong> ergative argument is not prohibited per se but if it applies, it applies too early,<br />

<strong>the</strong>reby underm<strong>in</strong><strong>in</strong>g case assignment to <strong>the</strong> absolutive argument (DPabs). As a consequence,<br />

<strong>the</strong> derivation crashes.<br />

Background & Data: It has been noted <strong>in</strong> <strong>the</strong> literature that <strong>in</strong> many ergative languages DPerg<br />

cannot be Ā-moved, <strong>in</strong> contrast to DPabs, cf. (1).<br />

(1) Katuk<strong>in</strong>a-Kanamari—Wh-Movement (Queixalos 2010)<br />

a. hanian tu Nodia nah=hoho-n<strong>in</strong>?<br />

b. *hanian tan dyuman tahi yu?<br />

whom Q Nodia ERG=call-DURATIVE who here spread water Q<br />

‘Whom is Nodia call<strong>in</strong>g?’<br />

‘Who spread water here?’<br />

There are at least two possible analyses: (i) The trace of DPerg is not licensed (e.g. <strong>in</strong> GB<br />

ter<strong>ms</strong>, it is not strictly governed); (ii) (covert) movement of DPabs blocks movement of DPerg,<br />

ei<strong>the</strong>r due to m<strong>in</strong>imality (Campana 1992), or DPabs blocks <strong>the</strong> only escape hatch with<strong>in</strong> vP<br />

(Aldridge 2004, Coon 2010). Both approaches suffer from technical and empirical proble<strong>ms</strong>.<br />

As for (i), constra<strong>in</strong>ts on traces cannot be formulated <strong>in</strong> <strong>the</strong> M<strong>in</strong>imalist framework. As for (ii),<br />

Campana’s analysis is based on a non-standard concept of <strong>in</strong>tervention while Aldridge (2004)<br />

and Coon (2010) must stipulate a ban on multiple specifiers. All three accounts must resort to<br />

covert movement of DPabs, which is hardly motivated on <strong>in</strong>dependent grounds. Empirically, <strong>the</strong><br />

analyses <strong>in</strong> (ii) make <strong>the</strong> wrong prediction that DPabs not only blocks movement of DPerg but<br />

also movement of o<strong>the</strong>r vP-<strong>in</strong>ternal elements like adjuncts or obliques (2). They also predict that<br />

similar movement asymmetries between coarguments should be found <strong>in</strong> nom<strong>in</strong>ative-accusative<br />

languages, contrary to fact.<br />

(2) Mam —Wh-Movement of Passive Agent (England 1983)<br />

al uPn xhi kub’ tzy-eet qa-cheej?<br />

Q RN DEP-3PL.ABS DIR GRAB-pass PL-horse<br />

‘By whom were <strong>the</strong> horses grabbed?’<br />

Assumptions: [A1] Follow<strong>in</strong>g Müller (2009), we assume that (i) case is assigned under Agree,<br />

(ii) <strong>in</strong> all languages, T assigns unmarked case (NOM/ABS) and v marked case (ERG/ACC),<br />

(iii) <strong>the</strong> only difference between accusative and ergative case syste<strong>ms</strong> is <strong>the</strong> order <strong>in</strong> which<br />

<strong>the</strong> basic operations Merge and Agree apply with<strong>in</strong> vP. S<strong>in</strong>ce v triggers two operations (it assigns<br />

case and selects an external argument), languages must give preference ei<strong>the</strong>r to Agree or<br />

Merge, given Earl<strong>in</strong>ess (Pesetsky 1989) and <strong>the</strong> tenet that operations cannot apply simultaneously.<br />

[A2] Instead of m<strong>in</strong>imality, we assume a Specifier-Head-Bias (SHB) which states that<br />

agreement between a head and its specifier is preferred to agreement between a head and an element<br />

its c-command doma<strong>in</strong> (Cho<strong>ms</strong>ky 1986, 1995, Koopman 2006; see Béjar & ˇRezáč 2009<br />

for a similar idea with a bias <strong>in</strong>versed). The SHB is compatible with equi-distance effects, which<br />

pose a problem for path-based def<strong>in</strong>itions of m<strong>in</strong>imality. [A3] Movement to SpecCP must make<br />

an <strong>in</strong>termediate stop <strong>in</strong> SpecTP. This is ensured by assum<strong>in</strong>g that ei<strong>the</strong>r TP is a phase (Richards<br />

2011), any XP is a phase (Müller 2010) or strict locality (Cho<strong>ms</strong>ky 2005). [A4] Edge features<br />

on phase heads are freely available (Cho<strong>ms</strong>ky 2000). [A5] Arguments may receive more than<br />

one case (cf. approaches to case stack<strong>in</strong>g: Andrews 1996, Nordl<strong>in</strong>ger 1998, Richards 2007).<br />

[A6] Agree is not sensitive to <strong>the</strong> PIC (see e.g. Boˇskovič 2007).<br />

Analysis: In ergative languages, preference is given to Merge over Agree with <strong>the</strong> result that v<br />

first selects <strong>the</strong> external argument (DPext) before it assigns case. Due to <strong>the</strong> SHB, <strong>the</strong> marked<br />

case of v ends up on its specifier while <strong>the</strong> <strong>in</strong>ternal argument (DP<strong>in</strong>t) receives case from T,


given [A6]. In accusative languages, preference is given to Agree over Merge, such that v has<br />

to assign case to DP<strong>in</strong>t before Merge of DPext, which <strong>in</strong> turn receives case from T (cf. Müller<br />

2009). Now, on <strong>the</strong> T head, <strong>the</strong> same <strong>in</strong>determ<strong>in</strong>acy as on v arises: it assigns case and must<br />

have an edge feature if one argument is to undergo Ā-movement. Assum<strong>in</strong>g that <strong>the</strong> languagespecific<br />

order of operations rema<strong>in</strong>s constant throughout <strong>the</strong> derivation, we can derive <strong>the</strong> ban on<br />

ergative movement and <strong>the</strong> asymmetry between ergative and accusative languages, consider<strong>in</strong>g<br />

two different derivations: [D1] DPext has an additional [wh]-feature and moves to an operator<br />

position <strong>in</strong> <strong>the</strong> left periphery via SpecTP ([A3]; [A4]) to check this feature; [D2] DP<strong>in</strong>t has a<br />

[wh]-feature and moves to an operator position. As for [D1], <strong>in</strong> accusative languages DP<strong>in</strong>t has<br />

received accusative with<strong>in</strong> vP. DPext receives case via Agree as soon as T is merged because<br />

Agree is preferred over Merge. Afterwards, DPext is free to move and check <strong>the</strong> edge feature on<br />

T. In ergative languages, DPext receives case with<strong>in</strong> <strong>the</strong> vP; DP<strong>in</strong>t rema<strong>in</strong>s unvalued. As soon as<br />

T is merged, it first attracts DPext to satisfy its edge feature (preference for Merge over Agree).<br />

Due to <strong>the</strong> SHB ([A2]), DPext checks case on T ([A5]) with <strong>the</strong> result that DP<strong>in</strong>t does not<br />

receive any case. The derivation crashes because DPext has absorbed <strong>the</strong> case features provided<br />

for DP<strong>in</strong>t. As for [D2], <strong>in</strong> accusative languages, DP<strong>in</strong>t receives its case from v while DPext is<br />

still unvalued. When T enters <strong>the</strong> structure, it first assigns case to DPext before it attracts DP<strong>in</strong>t<br />

to its specifier (preference for Agree). Thus, <strong>the</strong> context for multiple case-check<strong>in</strong>g does not<br />

arise <strong>in</strong> <strong>the</strong> first place. In ergative languages, DP<strong>in</strong>t is attracted to SpecTP where it receives its<br />

case due to <strong>the</strong> SHB while DPext receives case from v. The derivation converges. In summary,<br />

out of <strong>the</strong> four derivations <strong>in</strong> which one argument is extracted, three derivations converge. Only<br />

<strong>the</strong> derivation <strong>in</strong> which DPext <strong>in</strong> an ergative language is extracted is ungrammatical because<br />

DPerg moves before T has assigns case to DPabs. Hence, ergatives move too early.<br />

Conclusion: The present analysis has several advantages: (i) It captures <strong>the</strong> <strong>in</strong>tuition by Pol<strong>in</strong>sky<br />

et al (2011) that movement of DPerg is not problematic per se but leads to a problem for <strong>the</strong><br />

identification of DPabs. (ii) It is not forced to assume unmotivated covert movement steps of <strong>the</strong><br />

absolutive. (iii) It avoids stipulations about <strong>the</strong> different nature of ergative and accusative syntax.<br />

One parameter, namely <strong>the</strong> order<strong>in</strong>g of Merge and Agree, derives <strong>the</strong> difference <strong>in</strong> argument<br />

encod<strong>in</strong>g as well as <strong>the</strong> difference that <strong>the</strong>re is a movement asymmetry between coarguments <strong>in</strong><br />

ergative but not <strong>in</strong> accusative languages. (iv) It is empirically superior to o<strong>the</strong>r exist<strong>in</strong>g <strong>the</strong>ories<br />

because it makes <strong>the</strong> correct prediction that <strong>in</strong> languages which exhibit a ban on movement of<br />

<strong>the</strong> ergative o<strong>the</strong>r elements like adjuncts, passive agents and obliques are still free to move (cf.<br />

(2)). It should also be clear that our analysis implies a strictly derivational syntax <strong>in</strong> which <strong>the</strong><br />

order of operations plays an important role <strong>in</strong> deriv<strong>in</strong>g properties of <strong>the</strong> grammar.<br />

Outlook: F<strong>in</strong>ally, we address <strong>the</strong> related question of why not all ergative languages <strong>in</strong>stantiate<br />

a ban on ergative movement and discuss several possible answers: (i) The order of operations<br />

on T differs from <strong>the</strong> order on v. (ii) T is not a phase head. (iii) DPs cannot check multiple<br />

case features. Fur<strong>the</strong>rmore, we exam<strong>in</strong>e repair strategies for <strong>the</strong> ban on ergative movement<br />

such as antipassive or agent focus (Stiebels 2006, Aissen 1999). We argue that <strong>the</strong> agent focus<br />

morpheme is <strong>the</strong> morphological realization of an added probe which assigns case to DP<strong>in</strong>t,<br />

<strong>the</strong>reby prevent<strong>in</strong>g a crash of <strong>the</strong> derivation (cf. Béjar & ˇRezáč 2009, Coon 2010).<br />

Selected References: Aldrige, E. (2008): <strong>Generative</strong> Approaches to Ergativity, Language and <strong>L<strong>in</strong>guistics</strong> Compass<br />

2(5), 966-995. • Béjar, S. & M. ˇRezáč (2009): Cyclic Agree, L<strong>in</strong>guistic Inquiry 40(1), 35-73. • Boˇskovič,<br />

ˇZ. (2007): Agree, Phases, and Intervention Effects. In L<strong>in</strong>guistic Analysis 33: 54-96. • Campana, M. (1992):<br />

A Movement Theory of Ergativity, PhD <strong>the</strong>sis, McGill University, Montréal. • Coon, J. (2010): A-Bar Extraction<br />

Asymmetries. Manuscript. McGill University. • Müller. G. (2009): Ergativity, Accusativity, and <strong>the</strong> Order of<br />

Merge and Agree. In K. Grohmann (ed.), Explorations of Phase Theory. Features and Arguments, 269-308. Berl<strong>in</strong>:<br />

Mouton de Gruyter. • Pesetsky, D. (1989): The Earl<strong>in</strong>ess Pr<strong>in</strong>ciple. Ms. • Pol<strong>in</strong>sky, M., C. Gómez Gallo, P. Graff<br />

& E. Kravtchenko (to appear): ‘Subject Preference and Ergativity’, L<strong>in</strong>gua. • Richards, M. (2011): Deriv<strong>in</strong>g <strong>the</strong><br />

edge: What’s <strong>in</strong> a phase? Syntax. 14.1, 74-95.


Anke Assmann & Fabian Heck (Leipzig)<br />

Opaque Intervention <strong>in</strong> German Scrambl<strong>in</strong>g<br />

Claim: Based on hi<strong>the</strong>rto unnoticed facts from German, we claim that an argument’s capacity to b<strong>in</strong>d a<br />

parasitic gap (PG) or to associate with a float<strong>in</strong>g quantifier (FQ) is not dependent on <strong>the</strong> argument’s surface<br />

position. Ra<strong>the</strong>r, <strong>the</strong> distribution of PGs and FQs is often opaque: (i) There are cases where b<strong>in</strong>d<strong>in</strong>g of<br />

a PG by an argument is impossible although no co-argument <strong>in</strong>tervenes on <strong>the</strong> surface (counter-feed<strong>in</strong>g);<br />

(ii) and <strong>the</strong>re are cases where a wh-phrase can associate with a FQ although a surface <strong>in</strong>tervener is present<br />

(counter-bleed<strong>in</strong>g). We suggest that opacity arises because <strong>the</strong>re is an earlier derivational stage where (i ′ )<br />

a co-argument b<strong>in</strong>ds <strong>the</strong> PG and <strong>the</strong>n scrambles to a position where it ceases to <strong>in</strong>tervene on <strong>the</strong> surface;<br />

and where (ii ′ ) <strong>the</strong> wh-phrase occupies a position such that it is <strong>the</strong> closest element to associate with <strong>the</strong><br />

FQ, and later it moves on to a higher position, giv<strong>in</strong>g <strong>the</strong> surface impression to <strong>in</strong>volve <strong>in</strong>tervention.<br />

Observations: (i) An <strong>in</strong>direct object cannot b<strong>in</strong>d a PG by scrambl<strong>in</strong>g if a direct object co-argument has<br />

scrambled, too, not even if <strong>the</strong> co-argument does not <strong>in</strong>tervene on <strong>the</strong> surface between <strong>the</strong> <strong>in</strong>direct object<br />

and <strong>the</strong> PG (1-a). A direct object can b<strong>in</strong>d a PG even if <strong>the</strong> <strong>in</strong>direct object co-argument <strong>in</strong>tervenes (1-b).<br />

(1) a. *wenn jemand das Buch der Anette2 [ ohne PG2 zu vertrauen ] ausleiht<br />

if someone <strong>the</strong>acc book <strong>the</strong>dat Anette without to trust<br />

“if someone lends Anette <strong>the</strong> book without trust<strong>in</strong>g her”<br />

lends<br />

b. wenn jemand das Buch2 der Anette [ ohne PG2 durchzulesen ] zurückgibt<br />

if someone <strong>the</strong>acc book <strong>the</strong>dat Anette without through-to-read<br />

“if some returns <strong>the</strong> book to Anette without hav<strong>in</strong>g read it”<br />

returns<br />

(ii) A wh-subject cannot associate with <strong>the</strong> FQ alles if an <strong>in</strong>def<strong>in</strong>ite object <strong>in</strong>tervenes on <strong>the</strong> surface (2-a).<br />

The same holds for an <strong>in</strong>direct wh-object and an <strong>in</strong>def<strong>in</strong>ite direct object (2-b). However, if <strong>the</strong> grammatical<br />

functions of wh-phrase and <strong>in</strong>def<strong>in</strong>ite are reversed, surface <strong>in</strong>tervention is unproblematic (3-a,b).<br />

(2) a. *Wer1 hat e<strong>in</strong>en Koch alles1 bezahlt?<br />

whonom has aacc cook all<br />

“Who all paid a cook?”<br />

paid<br />

b. *Wem1 hat er e<strong>in</strong>en Koch alles1 gezeigt?<br />

whodat has er aacc cook all shown<br />

“To whom all did he show a cook?”<br />

(3) a. Wen1 hat e<strong>in</strong> Koch alles1 beleidigt?<br />

whoacc has a cooknom all<br />

“Who all did a cook <strong>in</strong>sult?”<br />

<strong>in</strong>sulted<br />

b. Was1 hat er e<strong>in</strong>em Koch alles1 geborgt?<br />

whatacc has he adat cook all lent<br />

“What all did he lend to a cook?”<br />

Assumptions: Our analysis is embedded <strong>in</strong> <strong>the</strong> <strong>the</strong>ory of Cho<strong>ms</strong>ky (2001) (thus presuppos<strong>in</strong>g phases,<br />

<strong>the</strong>ir impenetrability, and Agree). We follow Richards (2004) <strong>in</strong> assum<strong>in</strong>g that scrambl<strong>in</strong>g is triggered by<br />

edge features (EFs, Cho<strong>ms</strong>ky 2007). In a scrambl<strong>in</strong>g language like German, such EFs can be <strong>in</strong>serted (at<br />

least) on v and T. Follow<strong>in</strong>g Müller (2010), we assume that EF-<strong>in</strong>sertion on a head H is only possible as<br />

long as H is active, i.e., H still bears some o<strong>the</strong>r feature that is to trigger some syntactic operation. As<br />

Müller (2011) shows, this leads to <strong>the</strong> Intermediate Step Corollary (ISC), which states that <strong>in</strong>ternal Merge<br />

to SpecX triggered by an EF precedes external Merge (triggered by a selectional feature). Toge<strong>the</strong>r with<br />

<strong>the</strong> Extension Condition (Cho<strong>ms</strong>ky 1995), this implies that <strong>in</strong>ternal Merge triggered by an EF always<br />

targets an <strong>in</strong>ner SpecX while <strong>the</strong> outermost SpecX is created by external Merge (provided external Merge<br />

applies at all). If EFs are undeletable (see Cho<strong>ms</strong>ky 2007), it is plausible to assume that <strong>the</strong>y are <strong>in</strong>serted<br />

on a head H only once per derivation. Once an EF has been <strong>in</strong>serted on H, it attracts all <strong>the</strong> categories<br />

C1, C2, . . ., Cn that are supposed to reach H’s specifier doma<strong>in</strong>. Suppose now that <strong>the</strong> EF first attracts<br />

C1, C2, . . ., Cn <strong>in</strong> a top down fashion, start<strong>in</strong>g with <strong>the</strong> closest Ci (closeness be<strong>in</strong>g def<strong>in</strong>ed <strong>in</strong> ter<strong>ms</strong> of<br />

asymmetric c-command). Every time a Ci is attracted, it is placed on a stack. Only after all of C1, C2, . . .,<br />

Cn have been placed on <strong>the</strong> stack, <strong>the</strong>y are subsequently merged <strong>in</strong> H’s specifier doma<strong>in</strong>. Note that Merge<br />

exclusively has access to <strong>the</strong> Ci on top of <strong>the</strong> stack. Once Ci is removed, Ci−1 occupies <strong>the</strong> top and thus<br />

becomes accessible. As a consequence, C1, C2, . . ., Cn are moved <strong>in</strong> parallel: due to <strong>the</strong> first <strong>in</strong> – last out<br />

procedure, <strong>the</strong>y are merged <strong>in</strong> <strong>the</strong> <strong>in</strong>verse order <strong>in</strong> which <strong>the</strong>y have been attracted, see also Stroik (2009).<br />

Our assumptions about PGs and FQs are as follows. Adjunct clauses conta<strong>in</strong><strong>in</strong>g PGs are merged <strong>in</strong> Specv<br />

(above or below <strong>the</strong> subject). PGs must establish Agree with <strong>the</strong>ir b<strong>in</strong>der (see Assmann 2010). Once<br />

an argument has established Agree with a PG, it does not <strong>in</strong>duce any M<strong>in</strong>imal L<strong>in</strong>k Condition effects


(MLC, Fanselow 1991, Cho<strong>ms</strong>k 1995) with respect to o<strong>the</strong>r Agree-relations <strong>in</strong>volv<strong>in</strong>g PGs. FQ alles<br />

must associate with an <strong>in</strong>def<strong>in</strong>ite (see Reis 1992). Association <strong>in</strong>volves Agree. Alles is merged <strong>in</strong> <strong>the</strong> TP.<br />

Analysis: In (1-a), <strong>the</strong> direct object is merged as sister of V, <strong>the</strong> <strong>in</strong>direct argument as its specifier. The<br />

adjunct clause (conta<strong>in</strong><strong>in</strong>g <strong>the</strong> PG) is merged <strong>in</strong> Specv. As long as <strong>the</strong> subject is not merged, v rema<strong>in</strong>s<br />

active and an EF can be <strong>in</strong>serted on it. Accord<strong>in</strong>g to <strong>the</strong> assumptions <strong>in</strong> Müller (2010), an EF <strong>in</strong>serted<br />

on H must serve its purpose before o<strong>the</strong>r features of H can be accessed. Therefore both objects are first<br />

attracted by <strong>the</strong> EF to Specv. Movement proceeds <strong>in</strong> parallel (via <strong>the</strong> stack), result<strong>in</strong>g <strong>in</strong> <strong>the</strong> objects’<br />

relative order der Anette ≻ das Buch. Both c-command <strong>the</strong> PG. But s<strong>in</strong>ce das Buch is closer to <strong>the</strong> PG,<br />

der Anette cannot b<strong>in</strong>d it without violat<strong>in</strong>g <strong>the</strong> MLC (see also Fanselow 1993). Next, <strong>the</strong> subject is merged<br />

<strong>in</strong> Specv. When T is merged, it is active (due to its Φ-probes) and <strong>the</strong>refore accessible to EF-<strong>in</strong>sertion.<br />

EF on T attracts (aga<strong>in</strong> <strong>in</strong> parallel) subject and direct object to SpecT, result<strong>in</strong>g <strong>in</strong> (1-a). Although, on <strong>the</strong><br />

surface, der Anette is closest to <strong>the</strong> PG, b<strong>in</strong>d<strong>in</strong>g is impossible as <strong>the</strong> PG has already established Agree<br />

with das Buch. By <strong>the</strong> same reason<strong>in</strong>g (1-b) is grammatical. In (2-a), <strong>the</strong> direct object is merged as sister<br />

of V. In order to rema<strong>in</strong> PIC-accessible (to ultimately reach SpecT) it is attracted by an EF to Specv. Due<br />

to <strong>the</strong> ISC, this happens before <strong>the</strong> subject is merged, yield<strong>in</strong>g <strong>the</strong> order wer ≻ e<strong>in</strong>en Koch with<strong>in</strong> vP.<br />

Next, alles enters <strong>the</strong> TP. S<strong>in</strong>ce T is active (Φ-probes), an EF can be <strong>in</strong>serted. Subject and object are<br />

attracted <strong>in</strong> parallel, conserv<strong>in</strong>g <strong>the</strong>ir relative order. As a result, e<strong>in</strong>en Koch ends up closer to alles than<br />

wer. The MLC prevents wer from establish<strong>in</strong>g Agree with alles. Instead, alles enters Agree with <strong>the</strong><br />

<strong>in</strong>def<strong>in</strong>ite, which does not qualify for an associate, presumably due to semantic reasons. This accounts for<br />

<strong>the</strong> ungrammaticality of (2-a). Similar considerations apply to (2-b). In (3-a) <strong>the</strong> direct object is attracted<br />

to Specv (before <strong>the</strong> subject is merged), yield<strong>in</strong>g <strong>the</strong> order e<strong>in</strong> Koch ≻ wen with<strong>in</strong> vP. After alles is merged<br />

<strong>in</strong> TP and an EF is <strong>in</strong>serted on T, <strong>the</strong> arguments are moved (<strong>in</strong> parallel) to SpecT. As <strong>the</strong>ir relative order<br />

is not altered, wen is closer to alles and establishes Agree with it, <strong>the</strong>reby creat<strong>in</strong>g <strong>the</strong> association. In<br />

a later step, <strong>the</strong> wh-phrase moves away, giv<strong>in</strong>g <strong>the</strong> superficial impression of be<strong>in</strong>g separated from alles<br />

by a potential <strong>in</strong>tervener. Aga<strong>in</strong>, <strong>the</strong> derivation of (3-b) is analog. The analysis also accounts for facts<br />

about German PGs that have been noted elsewhere. Thus, it has been observed that no <strong>in</strong>tervention <strong>in</strong><br />

PG-b<strong>in</strong>d<strong>in</strong>g arises (a) if <strong>the</strong> potential <strong>in</strong>tervener is <strong>the</strong> subject (Fanselow 1993), (b) if it is a pronoun (den<br />

Dikken & Mulder 1991, Fanselow 1993), or (c) if it b<strong>in</strong>ds ano<strong>the</strong>r PG (aga<strong>in</strong> Fanselow 1993).<br />

Consequences: Opacity <strong>in</strong> phonology has traditionally served as an argument for derivational <strong>the</strong>ories<br />

(Kenstowicz & Kisseberth 1979). However, with <strong>the</strong> advent of conta<strong>in</strong>ment <strong>the</strong>ory (Pr<strong>in</strong>ce & Smolensky<br />

1993) th<strong>in</strong>gs changed: now it was possible to analyze opacity representationally. Not surpris<strong>in</strong>gly,<br />

<strong>the</strong>n, <strong>the</strong> <strong>in</strong>stances of counter-feed<strong>in</strong>g and counter-bleed<strong>in</strong>g discussed above can be reanalyzed by mak<strong>in</strong>g<br />

reference to traces, too. However, <strong>the</strong>re is a reason why one should refra<strong>in</strong> from do<strong>in</strong>g so. On <strong>the</strong><br />

one hand, Brody (2001) argues that a <strong>the</strong>ory that is <strong>in</strong>herently derivationally to beg<strong>in</strong> with any resort to<br />

representational concepts should be m<strong>in</strong>imized. To <strong>the</strong> extend that this is conv<strong>in</strong>c<strong>in</strong>g, opacity effects <strong>in</strong><br />

a derivational syntax should be analyzed without reference to traces/copies, as was proposed above. On<br />

<strong>the</strong> o<strong>the</strong>r hand, it can be argued that streng<strong>the</strong>n<strong>in</strong>g <strong>the</strong> representational character of <strong>the</strong> <strong>the</strong>ory as a whole<br />

by deriv<strong>in</strong>g scrambl<strong>in</strong>g via base-generation is ceteris paribus, not able to derive <strong>the</strong> facts about opaque<br />

<strong>in</strong>tervention <strong>in</strong> <strong>the</strong> context of PG-b<strong>in</strong>d<strong>in</strong>g and FQ-association discussed above.<br />

Selected references: •Assmann, Anke (2010): Parasitic Gaps <strong>in</strong> Derivational Grammar. Master’s <strong>the</strong>sis, Universität<br />

Leipzig. •Brody, Michael (2001): Some Aspects of Elegant Syntax. Ms., University College London. •Cho<strong>ms</strong>ky, Noam<br />

(2001): Derivation by Phase. In: Kenstowicz, ed., Ken Hale. A Life <strong>in</strong> Language. MIT Press, Cambridge, Massachusetts,<br />

pp. 1–52. •Cho<strong>ms</strong>ky, Noam (2007): Approach<strong>in</strong>g UG from Below. In: Sauerland & Gärtner, eds, Interfaces + Recursion =<br />

Language? Mouton de Gruyter, Berl<strong>in</strong>, pp. 1–30. •Dikken, Marcel den and René Mulder (1991): ‘Double Object Scrambl<strong>in</strong>g’,<br />

MIT Work<strong>in</strong>g Papers <strong>in</strong> <strong>L<strong>in</strong>guistics</strong> 14, 67–82. •Fanselow, Gisbert (1993): ‘Die Rückkehr der Basisgenerierer’, Gron<strong>in</strong>ger<br />

Arbeiten zur Germanistischen L<strong>in</strong>guistik 36, 1–74. •Fanselow, Gisbert (2001): ‘Features, θ-Roles, and Free Constituent Order’,<br />

L<strong>in</strong>guistic Inquiry 32, 405–437. •Müller, Gereon (2010): ‘On Deriv<strong>in</strong>g CED Effects from <strong>the</strong> PIC’, L<strong>in</strong>guistic Inquiry 41, 35–<br />

82. •Müller, Gereon (2011): Constra<strong>in</strong>ts on Displacement. Benjam<strong>in</strong>s, A<strong>ms</strong>terdam. •Reis, Marga (1992): The Category of<br />

Invariant alles <strong>in</strong> Wh-Clauses. In: Tracy, ed., Who Climbs <strong>the</strong> Grammar Tree? Niemeyer, Tüb<strong>in</strong>gen, pp. 465–492. •Richards,<br />

Marc (2004): Object Shift and Scrambl<strong>in</strong>g <strong>in</strong> North and West Germanic: A Case Study <strong>in</strong> Symmetrical Syntax. PhD <strong>the</strong>sis,<br />

University of Cambridge. •Stroik, Thomas (2009): Locality <strong>in</strong> M<strong>in</strong>imalist Syntax. MIT Press, Cambridge, Massachusetts.


Rebekah Bagl<strong>in</strong>i (Chicago)<br />

Reduced clausal structure <strong>in</strong> comparatives: evidence from Wolof<br />

(Keywords: syntax, phrasal comparatives, clausal comparatives, subject/non-subject asymmetry)<br />

Introduction: A central topic <strong>in</strong> <strong>the</strong> study of comparative constructions cross-l<strong>in</strong>guistically is <strong>the</strong> syntactic<br />

complexity of <strong>the</strong> standard constituent: whe<strong>the</strong>r a simple nom<strong>in</strong>al standard is base generated as a phrase or<br />

derived from a full clause. Clausal standards are assumed to be typologically rarer and restricted to certa<strong>in</strong><br />

grammatical strategies. Although underly<strong>in</strong>g clausal structure has been demonstrated <strong>in</strong> particle comparatives<br />

<strong>in</strong> English, German (Lechner 2001), and Greek (Merchant 2009), Stassen (1985) clai<strong>ms</strong> that exceed<br />

comparatives (a cross-l<strong>in</strong>guistically widespread strategy <strong>in</strong> which comparison is expressed via a transitive<br />

verb mean<strong>in</strong>g ‘exceed’) are categorically restricted to phrasal standards.<br />

This paper represents <strong>the</strong> first formal challenge to Stassen’s prediction, and draws on novel evidence<br />

from Wolof to demonstrate that some exceed comparatives must <strong>in</strong>volve underly<strong>in</strong>g clausal structure <strong>in</strong> <strong>the</strong><br />

standard.<br />

Evidence: The expression of unequal quality comparison <strong>in</strong> Wolof <strong>in</strong>volves <strong>the</strong> comparative ma<strong>in</strong> verb<br />

gen-a, ‘exceed’ or ‘surpass’, comb<strong>in</strong><strong>in</strong>g with a gradable predicate (GP) (1):<br />

(1) Mu-a gën-a-gatt B<strong>in</strong>ta.<br />

3sg-clf surpass-short B<strong>in</strong>ta.<br />

‘S/he is shorter than B<strong>in</strong>ta.’<br />

Stassen’s (1985) classification of gen-a as a phrasal exceed comparative <strong>in</strong>itially see<strong>ms</strong> plausible: no prepositions<br />

or special morphology <strong>in</strong>troduce <strong>the</strong> standard <strong>in</strong> (1) and <strong>the</strong> nom<strong>in</strong>al follow<strong>in</strong>g gen-a+GP appears<br />

to be a simple <strong>in</strong>dividual denot<strong>in</strong>g-expression, a DP. Thus, <strong>the</strong> surface structure of a Wolof comparative<br />

construction like (1) see<strong>ms</strong> consistent with a phrasal analysis, <strong>in</strong> which both <strong>the</strong> target and standard are<br />

arguments of <strong>the</strong> transitive gen-a+GP compound verb.<br />

I argue, however, that <strong>the</strong> apparent surface simplicity of comparative sentences like (1) is mislead<strong>in</strong>g.<br />

First, it is crucial to note that <strong>the</strong> target is obligatorily clefted <strong>in</strong> Wolof comparatives. This is curious<br />

as cleft<strong>in</strong>g is not widespread <strong>in</strong> Wolof, but restricted to <strong>the</strong> expression of focus and <strong>the</strong> formation of whquestions<br />

(Torrence 2005). Clefted arguments with<strong>in</strong> an embedded clause receive no special mark<strong>in</strong>g, but<br />

cleft<strong>in</strong>g with<strong>in</strong> <strong>the</strong> matrix clause is <strong>in</strong>dicated by dist<strong>in</strong>ctive morphology. The attachment site of <strong>the</strong> copular<br />

cleft marker -a- dist<strong>in</strong>guishes subject cleft (SC) sentences like (1)-(2) from non-subject cleft (NSC) sentences<br />

like (3): -a- follows a subject (pro)nom<strong>in</strong>al <strong>in</strong> SC, and an expletive l- <strong>in</strong> NSC.<br />

(2) subject cleft<br />

Ma-a gën-a-bëgg djënn (*ci) B<strong>in</strong>ta.<br />

1sg-clf surpass-like fish (*ci) B<strong>in</strong>ta.<br />

‘I like fish more than B<strong>in</strong>ta (does).’<br />

(3) non-subject cleft<br />

G<strong>in</strong>aar l-a-a gën-a-bëgg ci djënn.<br />

Chicken l-clf -1sg surpass-like ci fish<br />

‘I like chicken more than fish.’<br />

Besides <strong>the</strong> different morphological mark<strong>in</strong>g of SC and NSC comparative constructions, <strong>the</strong>re is a crucial<br />

asymmetry exhibited by (2) and (3): <strong>the</strong> standard of comparison may be <strong>in</strong>troduced by <strong>the</strong> word ci <strong>in</strong> NSC<br />

constructions, but not <strong>in</strong> SC constructions. Ci most commonly occurs as a locative preposition <strong>in</strong> Wolof,<br />

but it is also known to function as a (prepositional) complementizer <strong>in</strong>troduc<strong>in</strong>g a CP (Torrence 2005), as<br />

illustrated <strong>in</strong> (4)-(5):<br />

(4) Da-ma soon ci tóx póón<br />

do-1sg tired c smoke tobacco<br />

‘I’m tired of smok<strong>in</strong>g tobacco.’<br />

(5) Da-ma dogu ci jënd-kó.<br />

do-1sg decide c buy-3s.<br />

‘I decided to buy it.’<br />

If ci <strong>in</strong> sentences (2) and (3) were a simple preposition, <strong>the</strong>re would be no obvious explanation for <strong>the</strong><br />

contrast <strong>in</strong> grammaticality. I argue <strong>in</strong>stead that ci is a prepositional complementizer <strong>in</strong> <strong>the</strong>se sentences.<br />

Follow<strong>in</strong>g Merchant (2009)’s analysis of Greek reduced clausal comparatives, I propose that <strong>the</strong> standard<br />

<strong>in</strong> both SC and NSC comparatives <strong>in</strong> Wolof is generated with<strong>in</strong> a full clause and clefted to clause-<strong>in</strong>itial<br />

1


position, trigger<strong>in</strong>g clausal ellipsis. Accord<strong>in</strong>g to <strong>the</strong> standard account of cleft<strong>in</strong>g <strong>in</strong> Wolof, clefted subjects<br />

move to SpecTP, non-subjects to SpecCP (Torrence 2005). Thus, <strong>the</strong> observed asymmetry between (2)-(3)<br />

with <strong>the</strong> complementizer ci is entirely predicted: while <strong>the</strong> standard <strong>in</strong> both (2) and (3) is generated <strong>in</strong>side<br />

a clause, only <strong>the</strong> latter projects a full CP structure and thus co-occurs with ci.<br />

(2’) sc standard (3’) nsc standard<br />

TP<br />

PP<br />

B<strong>in</strong>tai<br />

T vP<br />

ti<br />

bëgg<br />

likes<br />

v’<br />

djënn<br />

fish<br />

ci CP<br />

djënnk<br />

fish C TP<br />

tk<br />

T vP<br />

1sg v’<br />

Besides expla<strong>in</strong><strong>in</strong>g <strong>the</strong> asymmetry observed <strong>in</strong> (2)-(3), an additional positive consequence of <strong>the</strong> clausal<br />

analysis proposed here is that it accounts for <strong>the</strong> o<strong>the</strong>rwise mysterious cleft<strong>in</strong>g of <strong>the</strong> target <strong>in</strong> Wolof comparatives.<br />

Accord<strong>in</strong>g to my analysis, Wolof comparatives embed a full-fledged clausal doma<strong>in</strong> <strong>in</strong> which <strong>the</strong><br />

standard undergoes movement to a clause-<strong>in</strong>itial position, just as has been proposed for clausal comparative<br />

constructions <strong>in</strong> English, German, Greek, and o<strong>the</strong>r languages. The cleft<strong>in</strong>g of <strong>the</strong> target is thus readily<br />

expla<strong>in</strong>ed as obligatory parallel movement to license ellipsis under identity with <strong>the</strong> standard. The trees <strong>in</strong><br />

(2’) and (3’) illustrate <strong>the</strong> proposed underly<strong>in</strong>g clausal structure for <strong>the</strong> standard <strong>in</strong> sentences (2)-(3).<br />

Conclusion: This paper shows that <strong>the</strong> Wolof gen-a construction is a clausal comparative, <strong>in</strong> which both<br />

target and standard are moved to a clause peripheral position to license ellipsis. This f<strong>in</strong>d<strong>in</strong>g has significant<br />

implications for <strong>the</strong> study of comparatives cross-l<strong>in</strong>guistically, <strong>in</strong>dicat<strong>in</strong>g that clausal comparatives are not<br />

as typologically rare as is often assumed, nor <strong>in</strong>compatible with exceed strategies of comparison.<br />

References<br />

Lechner, W. 2001. Reduced and phrasal comparatives. Natural Language & L<strong>in</strong>guistic Theory 19:683–735.<br />

Merchant, J. 2009. Phrasal and clausal comparatives <strong>in</strong> Greek and <strong>the</strong>abstractnessofsyntax. Journal of<br />

Greek <strong>L<strong>in</strong>guistics</strong> 9:134–164.<br />

Stassen, L. 1985. Comparison and universal grammar. Oxford: Basil Blackwell.<br />

Torrence, W.H. 2005. On <strong>the</strong> Distribution of Complementizers <strong>in</strong> Wolof. Doctoral Dissertation, UCLA.<br />

2<br />

bëgg<br />

like<br />

tk


Stefan Baumann (Köln)<br />

Types of Secondary Prosodic Prom<strong>in</strong>ence and <strong>the</strong>ir Relation to Information<br />

Structure<br />

A l<strong>in</strong>guistic element is prosodically prom<strong>in</strong>ent if it stands out from its neighbour<strong>in</strong>g<br />

elements by virtue of its suprasegmental properties. The prosodic differences between<br />

<strong>the</strong> elements can vary considerably, though. Thus, prosodic prom<strong>in</strong>ence is both gradual<br />

and relational <strong>in</strong> nature. These characteristics lead to proble<strong>ms</strong> <strong>in</strong> <strong>the</strong> annotation of<br />

naturally spoken language, s<strong>in</strong>ce <strong>the</strong> annotator is forced to classify <strong>in</strong>tonational events<br />

as categories.<br />

In order to circumvent this general problem of gradience/relativity versus categoriality,<br />

several types of 'secondary' prom<strong>in</strong>ences have been proposed <strong>in</strong> <strong>the</strong> literature, e.g.<br />

duration accent (Kohler 2005), secondary accent (Chafe 1994), ornamental accent<br />

(Bür<strong>in</strong>g 2007), subord<strong>in</strong>ate/secondary nucleus (Crystal 1969), or phrase accent (Grice et<br />

al. 2000). However, <strong>the</strong> various concepts refer to quite different phenomena or levels,<br />

namely<br />

(a) presence or absence of a pitch movement, i.e. tonal vs. non-­‐tonal prom<strong>in</strong>ence<br />

(b) accent type, i.e. <strong>the</strong> form of a pitch movement on a metrically strong syllable<br />

(c) accent position or <strong>the</strong> status of a prom<strong>in</strong>ence <strong>in</strong> <strong>the</strong> prosodic hierarchy<br />

The current paper presents experimental <strong>in</strong>vestigations on <strong>the</strong>se three types of<br />

secondary prosodic prom<strong>in</strong>ences and <strong>the</strong>ir relation to various aspects of <strong>in</strong>formation<br />

structure. First, <strong>in</strong> several acoustic and articulatory production studies on Second<br />

Occurrence Focus (SOF), i.e. words which are both focussed and contextually given, it<br />

could be shown that SOF ite<strong>ms</strong> receive an <strong>in</strong>termediate degree of prom<strong>in</strong>ence (e.g. Féry<br />

& Ishihara (2009) and Baumann et al. (2010) for German) which is not triggered by<br />

tonal movement but by <strong>in</strong>creased duration and <strong>in</strong>tensity (compared to background<br />

elements). The same pattern can be observed <strong>in</strong> some cases of bridg<strong>in</strong>g or 'implicit<br />

<strong>in</strong>ferrability' (Partee 1999), which do not display segmental copies of first occurrence<br />

expressions (cf. Riester & Baumann 2011).<br />

Second, a number of production studies confirmed that different accent types are<br />

related to different degrees of (perceived) givenness or <strong>in</strong>formation status of discourse<br />

referents (Pierrehumbert & Hirschberg 1990, Baumann & Grice 2006). Latest results of<br />

a perception study on German (Röhr & Baumann 2011) revealed a stepwise decrease <strong>in</strong><br />

<strong>the</strong> degree of perceived givenness from deaccentuation and prenuclear accents through<br />

low and early peak nuclear accents to high and ris<strong>in</strong>g nuclear accents. Fur<strong>the</strong>r evidence<br />

for <strong>the</strong> relevance of different accent types <strong>in</strong> <strong>the</strong> process<strong>in</strong>g of an item's <strong>in</strong>formation<br />

status comes from a neurol<strong>in</strong>guistic perception study on <strong>the</strong> prosody of <strong>in</strong>ferentially<br />

accessible referents (Schumacher & Baumann 2010): Three different accent types on a<br />

meronymic expression revealed significant differences <strong>in</strong> process<strong>in</strong>g effort<br />

(deaccentuation > H* > H+L*) <strong>in</strong>dicated by variations <strong>in</strong> event-­‐related bra<strong>in</strong> potentials<br />

(<strong>in</strong> particular N400). Crucially, <strong>the</strong> 'secondarily prom<strong>in</strong>ent' accent type H+L* proved to<br />

be <strong>the</strong> most appropriate marker of <strong>the</strong> 'secondarily given' referent.<br />

Third, <strong>the</strong> syntagmatic relation between prosodic prom<strong>in</strong>ences <strong>in</strong>fluences <strong>the</strong>ir<br />

perceived strength degree. In a recent perception study on German (Jagdfeld &<br />

Baumann 2011) subjects clearly dist<strong>in</strong>guished between strongly accented and<br />

unaccented words <strong>in</strong> two different positions with<strong>in</strong> an utterance. However, an order<br />

effect was found for weak pitch accents, which were not perceived as accents <strong>in</strong><br />

prenuclear but <strong>in</strong> nuclear position. In general, listeners proved to be less sensitive to<br />

categorical prom<strong>in</strong>ence judgements <strong>in</strong> prenuclear position (also reflected <strong>in</strong> longer


eaction times) than <strong>in</strong> nuclear position, which <strong>in</strong>dicates <strong>the</strong> special functional status of<br />

nuclear accents.<br />

The <strong>in</strong>sights of <strong>the</strong>se studies lead to <strong>the</strong> proposal of a model of prosodic prom<strong>in</strong>ence<br />

<strong>in</strong>tegrat<strong>in</strong>g <strong>the</strong> three levels discussed (see (1)). The model ai<strong>ms</strong> at be<strong>in</strong>g 'truly<br />

autosegmental-­‐metrical' <strong>in</strong> that it does not equate prom<strong>in</strong>ence with pitch movement but<br />

that it also takes non-­‐tonal criteria such as duration and <strong>in</strong>tensity <strong>in</strong>to account. It<br />

fur<strong>the</strong>rmore considers <strong>the</strong> relative position of prom<strong>in</strong>ences with<strong>in</strong> an utterance.<br />

(1) + phonological prom<strong>in</strong>ence -­‐<br />

References<br />

nuclear non-­‐nuclear<br />

prenuclear postnuclear<br />

metrical level<br />

-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐-­‐<br />

pitch accents non-­‐tonal prom<strong>in</strong>ences tonal level<br />

high ... low duration accent, phrase accent<br />

ris<strong>in</strong>g ... fall<strong>in</strong>g<br />

Baumann, Stefan & Mart<strong>in</strong>e Grice (2006). The Intonation of Accessibility. Journal of Pragmatics 38 (10).<br />

1636-­‐1657.<br />

Baumann, Stefan, Doris Mücke & Johannes Becker (2010). Expression of Second Occurrence Focus <strong>in</strong><br />

German. L<strong>in</strong>guistische Berichte 221. 61-­‐78.<br />

Bür<strong>in</strong>g, D. (2007). Intonation, Semantics and Information Structure. In: Ramchand, G. Reiss, C. (Eds.), The<br />

Oxford Handbook of L<strong>in</strong>guistic Interfaces.<br />

Chafe, W. (1994). Discourse, Consciousness, and Time. University of Chicago Press, Chicago/London.<br />

Crystal, D. (1969). Prosodic Syste<strong>ms</strong> and Intonation <strong>in</strong> English. Cambridge: Cambridge University Press.<br />

Féry, C. & Ishihara, S. (2009). The phonology of second occurrence focus. Journal of <strong>L<strong>in</strong>guistics</strong> 45, 285–<br />

313.<br />

Grice, M., Ladd, D.R., Arvaniti, A., (2000). On <strong>the</strong> Place of Phrase Accents <strong>in</strong> Intonational Phonology.<br />

Phonology 17 (2), 143-­‐185.<br />

Jagdfeld, Nils & Stefan Baumann (2011). Order Effects on <strong>the</strong> Perception of Relative Prom<strong>in</strong>ence.<br />

Proceed<strong>in</strong>gs 17th ICPhS, Hongkong, Ch<strong>in</strong>a. 958-­‐961.<br />

Kohler, K.J. (2005). Form and Function of Non-­‐Pitch Accents. AIPUK 35a. Prosodic Patterns of German<br />

Spontaneous Speech. 97-­‐123.<br />

Partee, B. (1999). Focus, Quantification, and Semantics-­‐Pragmatics Issues. In: P. Bosch & F. van der Sandt<br />

(eds.), Focus: L<strong>in</strong>guistic, Cognitive, and Computational Perspectives (Studies <strong>in</strong> Natural Language<br />

Process<strong>in</strong>g), Cambridge, Mass.: Cambridge University Press.<br />

Pierrehumbert, J.B. & Hirschberg, J. (1990). The Mean<strong>in</strong>g of Intonational Contours <strong>in</strong> <strong>the</strong> Interpretation of<br />

Discourse. In: P.R. Cohen, J. Morgan, M.E. Pollack, (eds.), Intentions <strong>in</strong> Communication. Cambridge: MIT<br />

Press. 271-­‐311.<br />

Riester, Arndt & Stefan Baumann (2011). Information Structure Annotation and Secondary Accents. In:<br />

Dipper, Stefanie & Heike Z<strong>in</strong>smeister (eds.): Beyond Semantics: Corpus-­‐based Investigations of Pragmatic<br />

and Discourse Phenomena. Bochumer L<strong>in</strong>guistische Arbeitsberichte 3. 111-­‐127.<br />

Röhr, Christ<strong>in</strong>e & Stefan Baumann (2011). Decod<strong>in</strong>g Information Status by Type and Position of Accent <strong>in</strong><br />

German. Proceed<strong>in</strong>gs 17th ICPhS, Hongkong, Ch<strong>in</strong>a. 1706-­‐1709.<br />

Schumacher, Petra & Stefan Baumann (2010). Pitch accent type affects <strong>the</strong> N400 dur<strong>in</strong>g referential<br />

process<strong>in</strong>g. NeuroReport 21 (9). 618-­‐622.


Krist<strong>in</strong>e Bentzen (Tro<strong>ms</strong>ø/CASTL) & Merete Anderssen (Tro<strong>ms</strong>ø)<br />

Discourse effects on <strong>the</strong> availability of Object Shift<br />

In Norwegian, like <strong>in</strong> <strong>the</strong> o<strong>the</strong>r Ma<strong>in</strong>land Scand<strong>in</strong>avian language, weak/unstressed pronom<strong>in</strong>al<br />

objects typically undergo Object Shift (OS) across negation, while full DP objects rema<strong>in</strong> <strong>in</strong> situ:<br />

(1) a. Jeg fant {*huset} ikke {huset}. b. Jeg fant {det} ikke {*det}.<br />

I found house.<strong>the</strong> not house.<strong>the</strong> I found it not it<br />

‘I didn’t f<strong>in</strong>d <strong>the</strong> house.’ ‘I didn’t f<strong>in</strong>d it.’<br />

However, as shown by Andréasson (2008) and Anderssen et al. (to appear), <strong>the</strong> pronom<strong>in</strong>al<br />

object det ‘it’ has special properties with respect to OS. When this pronoun, which is <strong>the</strong> neuter<br />

form of <strong>the</strong> third person s<strong>in</strong>gular personal pronoun, refers back to an identifiable noun <strong>in</strong> <strong>the</strong><br />

neuter gender, it undergoes OS, as <strong>in</strong> (1). However, det ‘it’ may also have a clausal or VP<br />

referent, or a DP referent with a type <strong>in</strong>terpretation (cf. Houser et al. 2007, Josefsson 2010,<br />

Bor<strong>the</strong>n 2003, Lødrup to appear), and <strong>in</strong> such contexts, it resists OS, despite be<strong>in</strong>g unstressed:<br />

(2) A: Har hun gått hjem? B: Jeg tror {*det CLAUSE} ikke {det CLAUSE}.<br />

has she gone home I th<strong>in</strong>k it. NEUT not it. NEUT<br />

‘Has she gone home?’ ‘I don’t th<strong>in</strong>k so.’ (det = ‘that she has gone home’)<br />

(3) A: Spiste du noe frukt? B: Nei, jeg gjorde {*det VP} ikke {det VP}.<br />

ate you any fruit. MASC no I did it. NEUT not it. NEUT<br />

‘Did you eat any fruit?’ ‘No, I didn’t.’ (det = ‘eat any fruit’)<br />

(4) A: Hva med fisk til middag? B: Nei, Per spiser {#det} ikke {det}.<br />

what with fish. MASC to d<strong>in</strong>ner no Per eats it. NEUT not it. NEUT<br />

‘How about fish for d<strong>in</strong>ner?’ ‘No, Per doesn’t eat that.’ (det=fish as a type of food)<br />

Influenced by Andréasson (2008), Anderssen et al. (to appear) propose an analysis of OS <strong>in</strong><br />

ter<strong>ms</strong> of IP-<strong>in</strong>ternal topicalization, and fur<strong>the</strong>rmore argue that such topicalization only applies to<br />

pronom<strong>in</strong>al objects with an <strong>in</strong>dividuated referent, as <strong>in</strong> (1), while objects with a non-<strong>in</strong>dividuated<br />

referent, such as those <strong>in</strong> (2)-(4) refra<strong>in</strong> from OS. We adopt this approach and develop it fur<strong>the</strong>r.<br />

1. Type of topichood matters<br />

First of all, we connect <strong>the</strong> availability of OS with pronom<strong>in</strong>al det ‘it’ to what type of topic this<br />

pronoun constitutes. We argue that det ‘it’ <strong>in</strong> (1) has <strong>the</strong> typical characteristics of familiar topics<br />

(Frascarelli 2007, Frascarelli & H<strong>in</strong>terhölzl 2007); it is distressed, <strong>in</strong>formationally given, and dl<strong>in</strong>ked<br />

<strong>in</strong> <strong>the</strong> sense that it has an accessible referent <strong>in</strong> <strong>the</strong> discourse. In contrast, det ‘it’ <strong>in</strong> (2)-(4)<br />

refers to what <strong>the</strong> sentence is about, and is hence analysed as an aboutness topic. We thus<br />

propose that OS only applies to familiar topics <strong>in</strong> Norwegian, while aboutness topics typically<br />

rema<strong>in</strong> <strong>in</strong> situ. Fur<strong>the</strong>rmore, although <strong>the</strong> pattern <strong>in</strong> (2)-(4) out of context is <strong>the</strong> default word<br />

order when det ‘it’ has a non-<strong>in</strong>dividuated reference, <strong>in</strong> certa<strong>in</strong> discourse situations OS of even<br />

such <strong>in</strong>stances of det is preferred. The follow<strong>in</strong>g dialogue between speakers ‘Karmøy03_gm’ and<br />

‘Karmøy04_gk’ is taken from spontaneous speech <strong>in</strong> <strong>the</strong> Nordic Dialect Corpus 1 :<br />

(5) Discourse topic: The problem of tourists hik<strong>in</strong>g on foot <strong>in</strong> <strong>the</strong> ski tracks:<br />

Karmøy03_gm: Men problemet var at når da alle fotturistene kom så fant jo ut de at det var<br />

f<strong>in</strong>est å gå i skisporene for da slapp de å vasse.<br />

‘But <strong>the</strong> problem was that when all <strong>the</strong> hik<strong>in</strong>g tourists came, <strong>the</strong>y found out<br />

that it was best to walk <strong>in</strong> <strong>the</strong> ski track ‘cause <strong>the</strong>n <strong>the</strong>y didn’t have to walk<br />

through snow.’<br />

1<br />

http://www.tekstlab.uio.no/nota/scandiasyn/Search_Facilities.html, cf. Johannessen, Priestley,<br />

Hagen, Åfarli & Vangsnes 2009.


Karmøy04_gk: mm. Var akkurat det så skjedde nå i Bjørgene og sant # at # så alle for og<br />

trødde og så # ødela de.<br />

‘Right. Exactly what happened now <strong>in</strong> Bjørgene, right, # that #, so everyone<br />

stepped around and messed it up’<br />

Karmøy03_gm: mm. Ja # for der er jo ikke kultur her veit du for… Så de skjønner det ikke.<br />

‘Right. Yes # ‘cause <strong>the</strong>re <strong>the</strong>y have no tradition for… So <strong>the</strong>y don’t get it.’<br />

[det = that <strong>the</strong>y shouldn’t hike on foot <strong>in</strong> <strong>the</strong> ski tracks]<br />

In this dialogue, we see that <strong>in</strong> ‘Karmøy03_gm’s f<strong>in</strong>al utterance, he uses <strong>the</strong> pronoun det ‘it’ to<br />

refer to a whole clause. However, <strong>in</strong> contrast to (2), it is here perfectly natural to place this<br />

pronoun <strong>in</strong> <strong>the</strong> OS position. We argue that <strong>the</strong> reason for this is that <strong>the</strong> pronoun here refers to <strong>the</strong><br />

topic of <strong>the</strong> whole discourse. Hence it functions as a cont<strong>in</strong>u<strong>in</strong>g topic. Accord<strong>in</strong>g to Frascarelli &<br />

H<strong>in</strong>terhölzl (2007), cont<strong>in</strong>u<strong>in</strong>g topics syntactically/phonologically behave like familiar topics.<br />

Thus, it is expected that it will undergo OS. We <strong>the</strong>refore propose that OS applies to pronom<strong>in</strong>al<br />

objects that constitute ei<strong>the</strong>r familiar or cont<strong>in</strong>u<strong>in</strong>g topics, but not to aboutness topics.<br />

2. Parallels between ±OS and it vs. that <strong>in</strong> English<br />

Accord<strong>in</strong>g to Gundel et al. (1993) it and that differ <strong>in</strong> cognitive status, it be<strong>in</strong>g IN FOCUS (i.e.<br />

topic) and that be<strong>in</strong>g merely ACTIVATED. Andréasson (2008) has argued that OS <strong>in</strong> Swedish and<br />

Danish applies to IN FOCUS elements. We adopt that proposal, and argue IN FOCUS elements are <strong>in</strong><br />

fact familiar topics. Moreover, compar<strong>in</strong>g <strong>the</strong> distribution of it and that <strong>in</strong> English (to <strong>the</strong><br />

distribution of shifted and non-shifted det ‘it’ <strong>in</strong> Norwegian, we demonstrate <strong>the</strong>re is a strik<strong>in</strong>g<br />

correlation between: In contexts where det ‘it’ typically does not undergo OS <strong>in</strong> Norwegian,<br />

Enlgish uses that. In contrast, when det ‘it’ <strong>in</strong> Norwegian undergoes OS, this corresponds to <strong>the</strong><br />

use of it <strong>in</strong> English. Two examples are provided here:<br />

(6) a. John liker å svømme. Maria liker ikke det. b. John liker å svømme. Maria liker det ikke.<br />

John likes to swim Maria likes not it John likes to swim Maria likes it not<br />

‘John likes swimm<strong>in</strong>g. Maria doesn’t like that.’ ‘John likes swimm<strong>in</strong>g. Maria doesn’t like it.’<br />

(det/that = ‘that John likes swimm<strong>in</strong>g’) (det/it = ‘swimm<strong>in</strong>g’)<br />

(7) John gikk til jobben. Maria forventet ikke det 1 . Susanne forventet det 2 heller ikke.<br />

John walked to work.<strong>the</strong> Maria expected not it Susanne expected it ei<strong>the</strong>r not<br />

‘John walked to work. Maria didn’t expect that. Susanne didn’t expect it ei<strong>the</strong>r.’<br />

(both det/that/it = ‘that John walked to work’)<br />

In both <strong>the</strong>se examples, we argue that <strong>the</strong> shifted det and English it are cont<strong>in</strong>u<strong>in</strong>g topics,<br />

whereas <strong>the</strong> non-shifted det and English that are aboutness topics. This correlation <strong>the</strong>refore<br />

provides support for our proposal that OS applies to both familiar and cont<strong>in</strong>u<strong>in</strong>g topics.<br />

Selected references:<br />

Anderssen et al. To appear. The acquisition of Norwegian Object Shift. Language Acquisition.<br />

Andréasson 2008. Not all objects are born alike. In Proceed<strong>in</strong>gs of <strong>the</strong> LFG08 Conference. Bor<strong>the</strong>n<br />

2003. Norwegian Bare S<strong>in</strong>gulars. Ph.D. <strong>the</strong>sis, NTNU. Frascarelli 2007. Subjects, topics and <strong>the</strong><br />

<strong>in</strong>terpretation of referential pro. NLLT 25. Frascarelli & H<strong>in</strong>terhölzl 2007. Types of topics <strong>in</strong> German<br />

and Italian. In On Information Structure, Mean<strong>in</strong>g and Form. Gundel, Hedberg & Zacharski. 1993.<br />

Cognitive status and <strong>the</strong> form of referr<strong>in</strong>g expressions <strong>in</strong> discourse. Language. Houser, Mikkelsen &<br />

Toosarvandani 2007. Verb phrase pronom<strong>in</strong>alization <strong>in</strong> Danish: Deep or surface anaphora? Proceed<strong>in</strong>gs<br />

of <strong>the</strong> Thirty-Fourth Western Conference on <strong>L<strong>in</strong>guistics</strong>. Josefsson 2010. “Disagree<strong>in</strong>g” pronom<strong>in</strong>al<br />

reference <strong>in</strong> Swedish and <strong>the</strong> <strong>in</strong>terplay between formal and semantic gender. L<strong>in</strong>gua 120. Lødrup to<br />

appear. Are Norwegian ‘type anaphora’ really surface anaphora? Journal of Germanic <strong>L<strong>in</strong>guistics</strong>.


Valent<strong>in</strong>a Bianchi (Siena) & Mara Frascarelli (Roma III)<br />

ON HOW TO BE ROOTED IN A CONTEXT<br />

1. Introduction. The impact of <strong>in</strong>formation-structural categories on <strong>the</strong> formal properties of<br />

<strong>the</strong> sentence has led some authors to assume dedicated discourse/contextual projections <strong>in</strong> <strong>the</strong><br />

C-doma<strong>in</strong> (Rizzi 1997 and related work). However, this conflicts with a strongly modular<br />

view accord<strong>in</strong>g to which only features that are truth-conditionally relevant, but not purely<br />

contextual features, belong <strong>in</strong> <strong>the</strong> syntactic computation (Horvath 2010, Fanselow–Lenertova<br />

2010). The arguments are often focussed on a s<strong>in</strong>gle phenomenon; we contend <strong>in</strong>stead that an<br />

approach compar<strong>in</strong>g different phenomena allows a more comprehensive view of <strong>the</strong> <strong>in</strong>terplay<br />

between syntax and context. This will be shown by consider<strong>in</strong>g a set of ‘root phenomena’<br />

which can apply <strong>in</strong> dist<strong>in</strong>ct types of embedded clauses, requir<strong>in</strong>g a f<strong>in</strong>e-gra<strong>in</strong>ed dist<strong>in</strong>ction <strong>in</strong><br />

<strong>the</strong>ir distributional properties. Our discussion will focus exclusively on declarative clauses.<br />

2. Data. We show that it is necessary to dist<strong>in</strong>guish at least two types of ‘root phenomena’:<br />

– Type I phenomena occur <strong>in</strong> clauses embedded under bridge verb (e.g. <strong>the</strong> question tag <strong>in</strong><br />

(1)), and <strong>the</strong>ir <strong>in</strong>terpretation does not <strong>in</strong>teract with elements <strong>in</strong> <strong>the</strong> matrix clause: (a) In Italian,<br />

a fronted focus cannot associate with a matrix clause negation (2); (b) In English, a Left<br />

Dislocated QP <strong>in</strong> <strong>the</strong> embedded clause does not scopally <strong>in</strong>teract with a matrix QP (3):<br />

(1) I {th<strong>in</strong>k/?*am glad} you like it, don’t you?<br />

(2) ?* Leo non ha detto che [DI SARA] è <strong>in</strong>namorato.<br />

Leo NEG has said that WITH SARA (he) is <strong>in</strong>-love<br />

(3) ? Every student tells me that one exam, I can prepare it <strong>in</strong> two weeks (*∀>∃)<br />

– Type II phenomena (German V2, English Topicalization) also occur <strong>in</strong> clauses embedded<br />

under non-bridge verbs (4), and <strong>the</strong>y do not require an <strong>in</strong>dependent <strong>in</strong>terpretation of <strong>the</strong><br />

embedded clause w.r.t. <strong>the</strong> matrix clause: cp. (5) to (2) and (6) to (3):<br />

(4) I am glad that this unreward<strong>in</strong>g job, she has f<strong>in</strong>ally decided to give up.<br />

(5) (?) Hans glaubt NICHT, Peter hat GEWONNEN... (Truckenbrodt 2006, (67))<br />

Hans believes not, Peter has won...<br />

(6) a. Jeder me<strong>in</strong>er Kollegen me<strong>in</strong>te auf e<strong>in</strong> Examen kann ich mich <strong>in</strong>nerhalb von 2 Wochen<br />

vorbereiten. b. Every student tells me that one exam, I can prepare <strong>in</strong> two weeks ( √ ∀>∃)<br />

Draw<strong>in</strong>g a first dist<strong>in</strong>ction, type I phenomena occur <strong>in</strong> discourse active embedded clauses (<strong>in</strong><br />

<strong>the</strong> sense of Dayal–Gri<strong>ms</strong>haw 2009): <strong>the</strong> proposition expressed bears a non-vacuous update<br />

potential w.r.t. <strong>the</strong> discourse context, because <strong>the</strong> modal base <strong>in</strong>troduced by <strong>the</strong> matrix verb<br />

is not realistic. Type II phenomena also occur <strong>in</strong> embedded clauses whose update potential<br />

gets ‘absorbed’ <strong>in</strong>to <strong>the</strong> derived context <strong>in</strong>troduced by <strong>the</strong> matrix attitude verb (Gärtner 2002,<br />

Truckenbrodt 2006; cf. also Krifka 2011, §3.6), with no impact on <strong>the</strong> discourse context. Our<br />

conclusion is that root phenomena are not a unitary class; <strong>the</strong> debate about whe<strong>the</strong>r or not<br />

‘assertive potential’ is a necessary condition for root phenomena (Bentzen et al. 2007) is due<br />

to <strong>the</strong> failure to dist<strong>in</strong>guish <strong>the</strong> two subclasses.<br />

3. Analysis. Step 1: How to characterize <strong>the</strong> two types of root-like embedded clause?<br />

In Stalnaker/Heim update semantics, <strong>the</strong> Context Change Potential (CCP) of an assertion<br />

updates <strong>the</strong> set of shared beliefs of <strong>the</strong> conversational community (common ground), while <strong>in</strong><br />

Gunlogson (2003) it updates <strong>the</strong> set of public commitments of a s<strong>in</strong>gle participant. Farkas–<br />

Bruce (2010) propose a componential view of <strong>the</strong> context which subsumes both <strong>the</strong>se aspects:<br />

an assertion has <strong>the</strong> effect of updat<strong>in</strong>g <strong>the</strong> speaker’s commitment set and also of project<strong>in</strong>g an<br />

updated common ground. Suppose, however, that <strong>the</strong>se two updates can be separated; we <strong>the</strong>n<br />

have <strong>the</strong> follow<strong>in</strong>g parameters w.r.t. CCP: (i) <strong>in</strong>dividual commitment / (proposed) shared<br />

commitment; (ii) discourse context / derived context <strong>in</strong>troduced by a matrix attitude verb.We


assume that derived contexts do not comprise a shared common ground, but only <strong>in</strong>dividual<br />

commitment sets (cf. Truckenbrodt 2006, 281). We obta<strong>in</strong> <strong>the</strong> follow<strong>in</strong>g comb<strong>in</strong>ations:<br />

1. Simple assertion 2. Non-discourse- 3. Discourse-active<br />

active subord<strong>in</strong>ate subord<strong>in</strong>ate<br />

discourse derived discourse derived discourse derived<br />

context context context context context context<br />

Individual commitment + (nil) - + - +<br />

Shared commitment + (nil) - (nil) + (nil)<br />

This view of discourse-active subord<strong>in</strong>ates agrees with Dayal–Gri<strong>ms</strong>haw’s claim that <strong>the</strong><br />

speaker does not commit herself to <strong>the</strong> truth of <strong>the</strong> proposition expressed by <strong>the</strong> subord<strong>in</strong>ate<br />

(pace Hooper–Thompson 1973), yet <strong>the</strong> latter has a proper CCP w.r.t. <strong>the</strong> discourse context.<br />

In o<strong>the</strong>r ter<strong>ms</strong>, <strong>the</strong> speaker places <strong>the</strong> embedded proposition ‘on <strong>the</strong> table’ as a potential<br />

relevant update of <strong>the</strong> shared common ground, without tak<strong>in</strong>g responsibility for its truth.<br />

Step 2: The permissible contexts for root phenomena depend on <strong>the</strong>ir <strong>in</strong>terpretive impact.<br />

– Type I phenomena are l<strong>in</strong>ked to <strong>the</strong> conversational dynamics. Specifically, we show that:<br />

a) Question tags implement a request for <strong>the</strong> addressee to confirm <strong>the</strong> relevant proposition;<br />

b) Italian focus front<strong>in</strong>g implements a correction (partial reversal of a previous assertion);<br />

c) English LD implements a topic shift, steer<strong>in</strong>g <strong>the</strong> conversation toward a new topic.<br />

These conversational moves update <strong>the</strong> shared commitments of <strong>the</strong> conversational<br />

community: this is why type I phenomena are only allowed <strong>in</strong> <strong>the</strong> first and third clause type.<br />

– Type II phenomena, <strong>in</strong>stead, have no impact on <strong>the</strong> conversational dynamics, but <strong>the</strong>y at<br />

most implement <strong>the</strong> structur<strong>in</strong>g of <strong>in</strong>formation with<strong>in</strong> a s<strong>in</strong>gle assertion: hence, <strong>the</strong>y only<br />

require <strong>the</strong> presence of an <strong>in</strong>dividual commitment set, ei<strong>the</strong>r <strong>in</strong> <strong>the</strong> discourse context or <strong>in</strong> a<br />

derived context. This is why <strong>the</strong>y are licensed <strong>in</strong> all three clause types.<br />

– Type I phenomena are only licensed if <strong>the</strong> proposition expressed by <strong>the</strong> embedded clause<br />

has autonomous CCP w.r.t. <strong>the</strong> matrix clause: this is why <strong>the</strong>re cannot be any scopal<br />

<strong>in</strong>teraction (or association with focus) between a matrix element and an element <strong>in</strong>cluded <strong>in</strong><br />

<strong>the</strong> embedded clause (cf. (2)-(3)), contrary to type II phenomena (cf. (5)-(7)).<br />

4. Consequences. The data <strong>in</strong> (2)-(3) show that <strong>the</strong> discourse active status of <strong>the</strong> subord<strong>in</strong>ate<br />

clause rules out a syntactically and semantically permissible relation between a matrix and an<br />

embedded element. The two compet<strong>in</strong>g views of <strong>the</strong> syntax–context <strong>in</strong>terface <strong>the</strong>n reduce to<br />

<strong>the</strong> opposition between failure–proof grammar vs. «free generation + filter<strong>in</strong>g» (Boeckx 2010,<br />

a.o.). If syntax and semantics are bl<strong>in</strong>d to <strong>the</strong> discourse active status of <strong>the</strong> embedded clause,<br />

(2)-(3) must be filtered out at <strong>the</strong> <strong>in</strong>terface with <strong>the</strong> context. (CP-extraposition (Me<strong>in</strong>unger<br />

2004) may account for (2)-(3), but, if it is <strong>in</strong>terface-driven, it still requires reference set<br />

computation and filter<strong>in</strong>g of <strong>the</strong> non-<strong>in</strong>terface-comply<strong>in</strong>g outputs.) If <strong>in</strong>stead <strong>the</strong> syntax<br />

encodes <strong>the</strong> CCP status of a clause <strong>in</strong> its C head, <strong>the</strong> relevant scope <strong>in</strong>teractions will be<br />

blocked <strong>in</strong> <strong>the</strong> course of <strong>the</strong> derivation. We believe that, by considerations of computational<br />

efficiency, failure-proof computation should be regarded as <strong>the</strong> null hypo<strong>the</strong>sis.<br />

[1] Bentzen et al 2007. Work<strong>in</strong>g Papers <strong>in</strong> Scand<strong>in</strong>avian Syntax 79, 93–118. [2] Boeckx<br />

2010. In Putnam (ed.), Explor<strong>in</strong>g Crash-Proof Grammars, 105–124. Benjam<strong>in</strong>s. [3] Dayal–<br />

Gri<strong>ms</strong>haw 2009. Subord<strong>in</strong>ation at <strong>the</strong> <strong>in</strong>terface. Ms., Rutgers Univ. [4] Fanselow–Lenertova<br />

2011. NLLT 29, 169-209. [5] Farkas–Bruce 2010. J. Semantics 27, 81–118. [6] Gärtner 2002.<br />

Theoretical <strong>L<strong>in</strong>guistics</strong> 28, 33–42. [7] Gunlogson 2003. True to Form. Routledge. [7] Horvath<br />

2010. L<strong>in</strong>gua 120, 1346–1369. [8] Hooper–Thompson 1973. LI 4, 465–497. [9] Krifka 2011.<br />

Embedd<strong>in</strong>g speech acts. Ms., Humboldt Univ. zu Berl<strong>in</strong>. [10] Me<strong>in</strong>unger 2004. In Lohnste<strong>in</strong>–<br />

Trissler (eds.), The Syntax and Semantics of <strong>the</strong> Left Periphery, 313–341. Mouton de Gruyter.<br />

[11] Rizzi 1997. In Haegeman (ed.), Elements of Grammar, 281–337. Kluwer. [12]<br />

Truckenbrodt 2006. Theoretical <strong>L<strong>in</strong>guistics</strong> 32, 257–306.


Jason Bishop (UCLA)<br />

Information structure, prosodic prom<strong>in</strong>ence and <strong>in</strong>dividual differences:<br />

evidence from on-l<strong>in</strong>e process<strong>in</strong>g<br />

In <strong>the</strong> present study we explored listeners’ knowledge about how prosody can express<br />

<strong>in</strong>formation structure <strong>in</strong> English. In particular, we probed listeners for expectations regard<strong>in</strong>g<br />

how prosodic prom<strong>in</strong>ence relates to <strong>the</strong> size of <strong>the</strong> focus constituent (broad vs. narrow) <strong>in</strong><br />

SVO constructions. While it is well-known that <strong>the</strong> prosodic realizations of contrasts along<br />

this dimension are ambiguous <strong>in</strong> ter<strong>ms</strong> of nuclear pitch accent location, recent studies have<br />

shown that speakers can dist<strong>in</strong>guish <strong>the</strong>m us<strong>in</strong>g gradient phonetic cues. In particular, it has<br />

been reported that (a) narrow focus on an object is associated with phonetically lower<br />

prenuclear prom<strong>in</strong>ence and/or phonetically higher nuclear prom<strong>in</strong>ence, and (b) broader VP<br />

and Sentence focus show <strong>the</strong> opposite pattern (e.g., Breen et al. 2010 and references <strong>the</strong>re<strong>in</strong>).<br />

Fur<strong>the</strong>r, it has also been shown that, <strong>in</strong> perception, listeners exhibit expectations for exactly<br />

this asymmetry (e.g., Bishop 2010). In <strong>the</strong> present study, we made use of <strong>the</strong> cross-modal<br />

associative prim<strong>in</strong>g paradigm to probe on-l<strong>in</strong>e process<strong>in</strong>g mechanis<strong>ms</strong> for a similar<br />

sensitivity.<br />

In <strong>the</strong> experiment presented, 174 English-speak<strong>in</strong>g listeners made lexical decisions<br />

about 32 visual targets (e.g., BRUNETTE) which followed auditorily-presented SVO sentences<br />

<strong>in</strong> which <strong>the</strong> object was <strong>the</strong> prime (e.g., He kissed a blonde).<br />

To test for <strong>the</strong> effects of<br />

prosody, sentences were heard by listeners with two different patterns of prom<strong>in</strong>ence: high<br />

prenuclear prom<strong>in</strong>ence/low nuclear prom<strong>in</strong>ence (ToBI annotations with H* on <strong>the</strong> verb and<br />

!H* on <strong>the</strong> object) and low prenuclear/high nuclear prom<strong>in</strong>ence (unaccented verb and (L+)H*<br />

on <strong>the</strong> object). To hold <strong>the</strong>ir acoustic properties constant, <strong>the</strong> primes <strong>in</strong> <strong>the</strong> two conditions<br />

were identical productions, spliced <strong>in</strong>to different SV productions (see Fig 1). These two<br />

prosodic conditions were <strong>the</strong>n crossed with two focus conditions. The first represented our<br />

primary <strong>in</strong>formation structural <strong>in</strong>terest, namely focus size (broad VP vs. narrow object focus);<br />

however, we also explored <strong>the</strong> possible effect of focus type (non-contrastive/WH focus vs.<br />

contrastive focus). Focus size and type were manipulated us<strong>in</strong>g lead-<strong>in</strong> questions (e.g., What<br />

did John do? for broad VP focus and Who did John kiss? for narrow object focus), with<br />

explicit alternatives offered (Did John kiss Mary?) <strong>in</strong> <strong>the</strong> case of contrastive focus.<br />

Mixed-effects l<strong>in</strong>ear regression was used to model listeners’ reaction times <strong>in</strong> <strong>the</strong><br />

lexical decision task as a function of our manipulations, and also several stimuli and listener<br />

variables. Among <strong>the</strong>se were scores on <strong>the</strong> Autism Spectrum Quotient (AQ; Baron-Cohen et<br />

al. 2001), recently shown to be a predictor of performance <strong>in</strong> speech perception and<br />

pragmatic tasks by “normal-function<strong>in</strong>g” adults (Yu 2010; Xiang, Grove and Giannakidou<br />

2011). Results were as follows. The most significant f<strong>in</strong>d<strong>in</strong>g was that lexical decision times<br />

were slowest follow<strong>in</strong>g a prime that was <strong>in</strong>terpreted as a narrow contrastive focus, but bore<br />

an accent that was low <strong>in</strong> relative prom<strong>in</strong>ence (Fig 2). This disruption to prim<strong>in</strong>g is<br />

<strong>in</strong>terpreted as <strong>the</strong> result of a focus/ prosody mis-match, and is consistent with our prediction<br />

regard<strong>in</strong>g listeners’ expectations. Also consistent with those predictions was <strong>the</strong> fact that<br />

response times were fastest when <strong>the</strong> prime was narrowly focused and high <strong>in</strong> relative<br />

prom<strong>in</strong>ence. Interest<strong>in</strong>gly, however, it was found that <strong>the</strong> advantage from this focusprom<strong>in</strong>ence<br />

match depended considerably on listeners’ AQ scores, as listeners with more<br />

“autistic” traits benefited less (Fig 3). Taken toge<strong>the</strong>r, <strong>the</strong> results provide strong evidence that<br />

English-speak<strong>in</strong>g listeners have conventionalized knowledge about <strong>the</strong> relationship between<br />

focus size and prosodic prom<strong>in</strong>ence, as this knowledge was found to have psychologically<br />

real consequences. We discuss <strong>the</strong>se results and <strong>the</strong>ir fur<strong>the</strong>r implications for <strong>the</strong> relationship<br />

between prosody and <strong>in</strong>formation structure, sentence and lexical process<strong>in</strong>g, and <strong>the</strong><br />

relevance of <strong>in</strong>dividual differences <strong>in</strong> l<strong>in</strong>guistic process<strong>in</strong>g.


Fig1. Example of <strong>the</strong> two prosodic conditions. Sentences were presented with ei<strong>the</strong>r high<br />

prenuclear/low nuclear prom<strong>in</strong>ence (left), or low prenuclear/high nuclear prom<strong>in</strong>ence (right).<br />

Primes (always <strong>the</strong> nuclear accented object) were thus relatively high or relatively lower <strong>in</strong><br />

prom<strong>in</strong>ence.<br />

Fig 2. Mean reaction times to targets as a<br />

function of a preced<strong>in</strong>g prime word’s focus<br />

size and relative prosodic prom<strong>in</strong>ence.<br />

Fig 3. Benefit from a focus-prom<strong>in</strong>ence<br />

match: mean RTs to targets follow<strong>in</strong>g a<br />

prom<strong>in</strong>ent prime that was narrowly<br />

focused and highly prom<strong>in</strong>ent. Three AQ<br />

groups of listeners are shown (higher AQ<br />

scores <strong>in</strong>dicate more prom<strong>in</strong>ent “autistic”<br />

traits).<br />

References<br />

Bishop, J. (2010). Information structural expectations <strong>in</strong> <strong>the</strong> perception of prosodic<br />

prom<strong>in</strong>ence, UCLA Work<strong>in</strong>g Papers <strong>in</strong> Phonetics, 108, 223-225.<br />

Baron-Cohen, S., Wheelwright, S., Sk<strong>in</strong>ner, R., Mart<strong>in</strong>, J., Clubley, E. (2001). The Autism-<br />

Spectrum Quotient (AQ): evidence from Asperger Syndrome/high-function<strong>in</strong>g autism,<br />

males and females, scientists and ma<strong>the</strong>maticians. Journal of Autism and Developmental<br />

Disorders 31, 5-17.<br />

Breen, M., Fedorenko E., Wagner., M, & Gibson, E. (2010). Acoustic correlates of<br />

<strong>in</strong>formation structure. Language and Cognitive Processes 25, 1044-1098.<br />

Yu, A.C.L. (2010). Perceptual compensation is correlated with <strong>in</strong>dividuals' “autistic” traits:<br />

implications for models of sound change. PLoS ONE, 5, e11950.<br />

doi:10.1371/journal.pone.0011950<br />

Xiang, M., Grove, J., Giannakidou, A. (2011). Interference “licens<strong>in</strong>g” of NPIs: pragmatic<br />

reason<strong>in</strong>g and <strong>in</strong>dividual differences. Poster presented at <strong>the</strong> 2011 CUNY Conference on<br />

Sentence Process<strong>in</strong>g.


In search of (im)perfection: <strong>the</strong> illusion of counterfactual aspect<br />

Puzzle: It has long been noticed that <strong>in</strong> certa<strong>in</strong> grammatical contexts, morphemes that o<strong>the</strong>rwise<br />

contribute temporal mean<strong>in</strong>gs suddenly fail to do so. One such context is counterfactual<br />

conditionals (CFs), which <strong>in</strong> many languages are marked by “fake” tense and aspect (Iatridou,<br />

2000).Example (1) from Greek illustrates both past and imperfective morphology used to mark<br />

afuture-less-vividconditional:<br />

(1) [An<br />

if<br />

Bronwyn M. Bjorkman (Nor<strong>the</strong>astern) & Claire Halpert (MIT)<br />

peTene<br />

die.PST.IMP<br />

o<br />

<strong>the</strong><br />

arXiGos]<br />

chief<br />

Ta<br />

FUT<br />

ton Tavame st<strong>in</strong> korifi tu vunu<br />

him bury.PST.IMP on.<strong>the</strong> top <strong>the</strong> mounta<strong>in</strong><br />

‘If <strong>the</strong> chief died, we would bury him on <strong>the</strong> top of <strong>the</strong> mounta<strong>in</strong>.’ (Iatridou, 2000)<br />

The use of “fake” past morphology associated with CF <strong>in</strong>terpretations has been widely<br />

documented (Steele, 1975; James, 1982; Iatridou, 2000, a.o.). Several proposals analyze fake<br />

past as <strong>the</strong> locus of CF semantics, ei<strong>the</strong>r by constru<strong>in</strong>g “past”asamarkerofmodal,ra<strong>the</strong>rthan<br />

temporal, remoteness (Steele, 1975; Iatridou, 2000; RitterandWiltschko,2010)orbyderiv<strong>in</strong>g<br />

CF mean<strong>in</strong>g from a purely temporal past (Ippolito, 2002; Arregui, 2009). It has been claimed<br />

that fake imperfective is also <strong>in</strong>volved <strong>in</strong> CF mark<strong>in</strong>g, though itsuseislesswell-understood:<br />

for Iatridou (2000, 2009) imperfective <strong>in</strong> CFs is a default aspect; Arregui (2004) clai<strong>ms</strong> that<br />

it reflects <strong>in</strong>compatibility between perfective and CFs; while Ippolito (2004) proposes that a<br />

“modal imperfective” reflects a speaker’s <strong>in</strong>direct evidence foraproposition. Iatridou(2009)<br />

proposes that imperfective-marked CFs occur <strong>in</strong> a subset of <strong>the</strong> languages with past-marked<br />

CFs, a generalization she based on <strong>the</strong> fact that Slavic languages have “fake” past but “real”<br />

aspect <strong>in</strong> CFs. In this paper we show that a full typology <strong>in</strong>cludes languages with “fake”<br />

perfective aspect <strong>in</strong> CFs as well, contradict<strong>in</strong>g <strong>the</strong>se previous approaches.<br />

Proposal: We argue that <strong>the</strong> apparent requirement for imperfective <strong>in</strong> CFs <strong>in</strong> some languages<br />

is illusory, amorphologicalreflexof<strong>the</strong>needtorealizeatruePAST feature. “Past<br />

imperfective” morphology <strong>in</strong> <strong>the</strong>se languages actually expresses only PAST, weclaim;itreceives<br />

an imperfective <strong>in</strong>terpretation due to contrast with atruePERFECTIVE. In CFs, this<br />

“past imperfective” morphology reflects only CF “past”; it does not reflect syntactic IMPER-<br />

FECTIVE features. We illustrate this proposal with <strong>the</strong> morphological paradigm of three types<br />

of languages: (a) Greek, Romance, and Zulu, where imperfective unspecified and see<strong>ms</strong> to occur<br />

<strong>in</strong> CFs; (b) Arabic, where perfective is unspecified and occurs <strong>in</strong> CFs; and (c) Slavic, where<br />

PAST is specified <strong>in</strong>dependently of aspect, and CFs preserve full aspectual contrasts.<br />

Pattern A: In languages like Greek and Romance, where <strong>the</strong> puzzle of fake CF imperfective<br />

<strong>in</strong>itially arose, CFs are always marked with past-imperfective morphology, while “real” tense<br />

and aspect is suppressed, as illustrated above <strong>in</strong> (1).<br />

We argue based on <strong>in</strong>dependent evidence that this pattern arises because <strong>the</strong> “past imperfective”<br />

encodes only PAST features and is unspecified for aspect; “past perfective”, by contrast,<br />

expresses only PERFECTIVE aspect, and receives a past <strong>in</strong>terpretation by default, due to<strong>in</strong>compatibility<br />

between perfective and present tense (Dahl, 1985). The imperfective <strong>in</strong>terpretation of<br />

<strong>the</strong> “past imperfective” arises due to <strong>the</strong> absence of a privative PERFECTIVE feature.<br />

The absence of “real” aspect mark<strong>in</strong>g <strong>in</strong> CFs <strong>in</strong> <strong>the</strong>se languages arisesbecausePAST and<br />

PERFECTIVE morphemes compete for realization <strong>in</strong> a s<strong>in</strong>gle position. In Zulu, which we argue<br />

has <strong>the</strong> same morphological specifications as Greek and Romance and thus appears to require<br />

“past imperfective” mark<strong>in</strong>g <strong>in</strong> CFs, a perfective suffix can co-occur with this “fake imperfective”<br />

<strong>in</strong> perfective CFs, though <strong>the</strong> two are normally <strong>in</strong>compatible (due to <strong>the</strong> redundancy of<br />

mark<strong>in</strong>g both PAST and PERFECTIVE when both are <strong>in</strong>terpreted temporally):<br />

(2) [ukuba be- ngi-thimulile ]be-ngi-zo-d<strong>in</strong>gaithishi<br />

if PST.IMP1SGsneeze PFV IMP-1SG-FUT-need 5tissue<br />

‘If I had sneezed, I would have needed a tissue.’<br />

1


What dist<strong>in</strong>guishes Zulu from languages like Greek and Romance is that PAST and PERFEC-<br />

TIVE do not compete for a s<strong>in</strong>gle morphological “slot”; as a result, realPERFECTIVE features<br />

and fake PAST features can both be realized on a s<strong>in</strong>gle verb.<br />

Pattern B: Fur<strong>the</strong>r evidence for our approach comes from CF mark<strong>in</strong>g <strong>in</strong> Arabic, a language<br />

<strong>in</strong> which perfective, ra<strong>the</strong>rthanimperfective,isrequired<strong>in</strong>CFs. Wearguethat<strong>the</strong><br />

morphological specification <strong>in</strong> this language is <strong>the</strong> reverse of Pattern A languages: here “past<br />

perfective” is simply specified for syntactic PAST features, while “imperfective” is underspecified<br />

for tense (see Benmamoun 2000, Karawani & Zeijlstra 2010, Bjorkman 2011, a.o. for<br />

<strong>in</strong>dependent evidence of this feature specification).<br />

(3) [iza úileQ halaP,] kaan b-iwsal Qal waPt la l-muèaadara<br />

if leave.PST.PFV now, be.PST B-arrive.IMP on <strong>the</strong>-time for <strong>the</strong>-lecture<br />

‘If he left now, he would arrive on time for <strong>the</strong> lecture.’ (Halpert and Karawani, 2011)<br />

Like Zulu, Arabic expresses real tense and aspect <strong>in</strong> CFs via an auxiliary<strong>in</strong>aseparate<br />

position from fake past. Auxiliary kaan, whichisa“perfective”formof<strong>the</strong>verb,markspast<br />

tense alone, while real temporal morphology occurs on <strong>the</strong> ma<strong>in</strong> verb:<br />

(4) [iza kanno b-yitlaQ bakkeer kul yom,] kaan b-iwsal Qa l-waPt<br />

if be.PST B-leave.IMP early every day, be.PST B-arrive.IMP on <strong>the</strong>-time<br />

‘If he were <strong>in</strong> <strong>the</strong> habit of leav<strong>in</strong>g early, he would arrive on time.’ (H&K)<br />

This ability of <strong>the</strong> past perfective auxiliary kaan alone to mark CFs supports <strong>the</strong> view that<br />

<strong>the</strong> “past perfective” CF mark<strong>in</strong>g <strong>in</strong> (3) is <strong>the</strong> exponent of PAST features only.<br />

Pattern C: F<strong>in</strong>ally, Slavic allows full aspectual contrasts <strong>in</strong> CFs, as <strong>in</strong>(5)fromRussian:<br />

(5) a. Esli by Dˇzon umer, my poxoroni-l-i by ego na gor-e.<br />

if SUBJ J. die.PFV.PST we bury.PFV-PST-PL SUBJ he.ACC on mounta<strong>in</strong>-LOC<br />

‘If John died, we would bury him on <strong>the</strong> mounta<strong>in</strong>.’<br />

b. Esli by Dˇzon umira-l, s nim by-l by doktor.<br />

if SUBJ J. die.IMPF-PST with he.INSTR be-PST SUBJ doctor<br />

‘If John were dy<strong>in</strong>g, <strong>the</strong> doctor would be with him.’<br />

We propose that <strong>in</strong> <strong>the</strong>se languages, unlike <strong>in</strong> Patterns A and B, PAST, PERFECTIVE, and<br />

IMPERFECTIVE all have separate morphological exponents. Evidence for this can be found <strong>in</strong><br />

<strong>the</strong> existence of morphological past and present for<strong>ms</strong> for imperfective and perfective verbs<br />

<strong>in</strong> Slavic (though a morphological “present” perfective receives a future <strong>in</strong>terpretation). As a<br />

result, real aspectual features PERFECTIVE and IMPERFECTIVE can both be realized <strong>in</strong> addition<br />

to <strong>the</strong> fake PAST mark<strong>in</strong>g required by CFs and <strong>the</strong> illusion of counterfactual aspect disappears.<br />

Implications: The claim that “past imperfective” morphology expresses only PAST features,<br />

and that “imperfective” <strong>in</strong> past-marked CFs is illusory, br<strong>in</strong>gs us closer to understand<strong>in</strong>g<br />

<strong>the</strong> broader contextual requirement for PAST <strong>in</strong> CFs. We argue that perfective morphology<br />

cannot mark CFs <strong>in</strong> Pattern A because “past perfective” does not express PAST features <strong>in</strong><br />

<strong>the</strong>se languages. Given that past perfectives have past <strong>in</strong>terpretations, we argue that it is past<br />

tense features, ra<strong>the</strong>rthanpasttense<strong>in</strong>terpretations, thatresult<strong>in</strong>CF<strong>in</strong>terpretations. Iftrue,<br />

this proposal is an argument <strong>in</strong> favor of analyses <strong>in</strong> which PAST features represent a broader<br />

temporal or modal “exclusion” (Iatridou, 2000) or [-co<strong>in</strong>cidence] (Ritter and Wiltschko, 2010)<br />

feature, and aga<strong>in</strong>st <strong>the</strong> view that <strong>the</strong> past mark<strong>in</strong>g of CFs reflects <strong>the</strong> contribution of past tense<br />

<strong>in</strong>terpretation (Arregui, 2009; Ippolito, 2002).<br />

Selected References: Arregui, A. 2009. On similarity <strong>in</strong> counterfactuals. L&P 32.; Halpert, C., & H. Karawani.<br />

2011. Aspect <strong>in</strong> counterfactuals from A(rabic) to Z(ulu). WCCFL 29.; Iatridou, S. 2000. The grammatical <strong>in</strong>gredients<br />

of counterfactuality. LI 31.; — 2010. Some thoughts about <strong>the</strong> imperfective <strong>in</strong> counterfactuals. Handout.;<br />

Ippolito, M. 2002. On <strong>the</strong> Semantic Composition of Subjunctive Conditionals. Ms. MIT/Tüb<strong>in</strong>gen.; — 2004.<br />

Imperfect modality. In The syntax of time. MITPress.;Ritter, E., & M. Wiltschko. 2010. The composition of<br />

INFL. L<strong>in</strong>gBuzz/001078.<br />

2


Andreas Blümel (Goe<strong>the</strong>-Universität Frankfurt am Ma<strong>in</strong>)<br />

Successive-cyclic Movement as <strong>in</strong>termediate Labell<strong>in</strong>g Indeterm<strong>in</strong>acies<br />

Background: Cho<strong>ms</strong>ky (2008) proposes (1-a) as an algorithm to determ<strong>in</strong>e <strong>the</strong> label when two<br />

syntactic objects undergo external merge (EM, a symmetric process of set-formation) and (1-b)<br />

for <strong>in</strong>ternal merge (IM). Both statements express <strong>the</strong> fact that <strong>the</strong> set result<strong>in</strong>g from IM/EM<br />

must be identified to participate <strong>in</strong> fur<strong>the</strong>r operations. But especially <strong>the</strong> stipulation (1-b) is<br />

dubious on conceptual grounds and failures of <strong>the</strong> application of (1-b) have been claimed to be<br />

empirically desireable to derive free relatives among o<strong>the</strong>r constructions (Cho<strong>ms</strong>ky 2008, Citko<br />

2008, Ott 2011a). Thus (1-b) is better derived from <strong>in</strong>dependent factors. There is no shortage<br />

of proposals accord<strong>in</strong>g to which ‘reprojection’ of <strong>the</strong> mov<strong>in</strong>g XP may take place (Citko 2008,<br />

Georgi/Müller 2010, Hornste<strong>in</strong>/Uriagereka 2002), but <strong>in</strong> <strong>the</strong>se works violations of (1-b) serve<br />

as descriptive devices, not as sources of (failures to apply) (1-b).<br />

Already for EM Cho<strong>ms</strong>ky (2008:fn.34) addresses an exception to <strong>the</strong> application of (1-a),<br />

namely EM of a subject-DP to v*P as <strong>in</strong> (2): “[T]hese structures lack a label and have an<br />

<strong>in</strong>herent <strong>in</strong>stability, so that one of <strong>the</strong> two members [. . . ] must raise.” An attractive side-effect<br />

of this idea is that it gives a partial explanation for <strong>the</strong> EPP: <strong>the</strong> symmetry created by EM<br />

is broken by IM of <strong>the</strong> subject to Spec-TP. Recently, a number of works have capitalized on<br />

this idea and suggested that <strong>in</strong>stances of IM are ‘triggered’ by <strong>the</strong> <strong>in</strong>terfaces’ avoidance and<br />

<strong>in</strong>tolerance of symmetry and ambiguity created by narrow syntax (Boecks 2008:116-118, Moro<br />

2008, Ott 2011b). When one of <strong>the</strong> two members <strong>in</strong> (2) moves at <strong>the</strong> phase level, an asymmetry<br />

and concomitantly <strong>the</strong> label (3) is created: IM of DP yields a discont<strong>in</strong>uous object, a set of<br />

occurrences of DP, while v* is identified as <strong>the</strong> label (boldfaced on <strong>the</strong> right <strong>in</strong> (3)); just as only<br />

<strong>the</strong> head of a movement cha<strong>in</strong> can move and cause <strong>in</strong>tervention for Agree, only <strong>the</strong> head of a<br />

movement cha<strong>in</strong> is elegible for labell<strong>in</strong>g.<br />

Proposal: Adopt<strong>in</strong>g phase <strong>the</strong>ory, I propose that such ‘<strong>in</strong>stable’ structures may not only result<br />

from EM but also from IM (cf. also Cho<strong>ms</strong>ky 2010): symmetry-break<strong>in</strong>g movement effectively<br />

leaves <strong>the</strong> <strong>in</strong>-situ category as <strong>the</strong> label. But IM results aga<strong>in</strong> <strong>in</strong> a symmetric, unlabellable<br />

structure <strong>in</strong> need of desymmetrization once we partially dispense with (1-b). Thus <strong>the</strong> process<br />

solves <strong>the</strong> labell<strong>in</strong>g problem downstairs, but only to create a new one <strong>in</strong> <strong>the</strong> target. I claim that<br />

this pr<strong>in</strong>cipally <strong>in</strong>f<strong>in</strong>ite symmetry-destroy<strong>in</strong>g/symmetry-creat<strong>in</strong>g character of derivations is <strong>the</strong><br />

source, i.e. <strong>the</strong> ‘trigger’ of <strong>in</strong>termediate steps of successive-cyclic A’-movement, <strong>the</strong> empirical<br />

doma<strong>in</strong> to which I conf<strong>in</strong>e myself here. The symmetry-avoid<strong>in</strong>g character at Transfer (i.e. at<br />

<strong>the</strong> next higher phase) is thus what derives (1-b): <strong>the</strong> target category becomes <strong>the</strong> label due to<br />

<strong>the</strong> fact that <strong>the</strong> mov<strong>in</strong>g XP must move fur<strong>the</strong>r to break <strong>the</strong> newly created symmetry. Of course,<br />

this powerful mechanism needs tam<strong>in</strong>g. I propose that Agree between a phase-head and <strong>the</strong><br />

mov<strong>in</strong>g WH-element does just that: at <strong>the</strong> po<strong>in</strong>t when a C-head with an unvalued Q-feature<br />

is merged, it probes for (<strong>the</strong> <strong>in</strong>terpretable Q-feature on) <strong>the</strong> WH-element. It is <strong>the</strong> successful<br />

prob<strong>in</strong>g relationship which ultimately and effectively br<strong>in</strong>gs <strong>the</strong> WH-element to a halt, captured<br />

<strong>in</strong> (4) (taken from Boeckx 2008:91-98): as <strong>the</strong> probe becomes <strong>the</strong> label by (4), movement of<br />

XP to Spec-probe creates a symmetry which is unproblematic, because (4) renders <strong>the</strong> structure<br />

sufficiently asymmetric for label determ<strong>in</strong>ation. The probe functions similar to what Boeckx<br />

(2003) calls ‘Strong Occurrences’ and Rizzi (2006) ‘Criterial Positions.’<br />

Consider (5), an abstract partial representation of <strong>the</strong> pervasive, structurally unbounded phenomenon<br />

of successive-cyclic WH-movement: an object WH-element ends up <strong>in</strong> <strong>the</strong> specifier<br />

of <strong>the</strong> phase head v* (for <strong>in</strong>dependent reasons). At this derivational stage v* or WH may determ<strong>in</strong>e<br />

<strong>the</strong> label. At <strong>the</strong> po<strong>in</strong>t when C merges, IM of <strong>the</strong> WH-phrase to Spec-CP asymmetrizes<br />

<strong>the</strong> structure, solv<strong>in</strong>g <strong>the</strong> labell<strong>in</strong>g problem as far as <strong>the</strong> TP is concerned: TP-<strong>in</strong>ternally, v* determ<strong>in</strong>es<br />

<strong>the</strong> label because IM applies to WH, stabiliz<strong>in</strong>g <strong>the</strong> structure (6). But <strong>in</strong> <strong>the</strong> result<strong>in</strong>g<br />

1


workspace <strong>the</strong> label <strong>in</strong>determ<strong>in</strong>acy problem once aga<strong>in</strong> arises, i.e. <strong>the</strong> edge{WH{C{. . .}}} is<br />

aga<strong>in</strong> ‘too symmetric’ and unstable. In this sense <strong>the</strong> edge configurations <strong>in</strong> (2)/(5) and (6) are<br />

parallel. Aga<strong>in</strong> <strong>the</strong> asymmetry is restored once WH moves on (on <strong>the</strong> right <strong>in</strong> (6)).<br />

In (7-a) <strong>the</strong> matrix verb doesn’t select an <strong>in</strong>direct question. A spurious [WH]/[EPP]-feature<br />

on <strong>the</strong> embedded C head to derive <strong>in</strong>termediate movement steps is thus dubious and ad hoc.<br />

Under <strong>the</strong> present analysis, <strong>the</strong> successive-cyclic character of long WH-movement follows automatically<br />

from <strong>the</strong> symmetry-avoid<strong>in</strong>g character <strong>the</strong> <strong>in</strong>terfaces impose on each <strong>in</strong>termediate<br />

configuration. What blocks movement to <strong>the</strong> matrix Spec-C <strong>in</strong> cases like (8-a), i.e. why don’t<br />

we have to get (8-b)? I propose that <strong>in</strong>terrogative C bears a prob<strong>in</strong>g Q-feature which Agrees<br />

with <strong>the</strong> WH-word. Such a feature appears plausible as <strong>the</strong> CP is selected by wonder. By contrast,<br />

such an Agree relation between non-<strong>in</strong>terrogative C (=that) and which book is crucially<br />

miss<strong>in</strong>g <strong>in</strong> (7-a), which is why (7-b) is out: <strong>the</strong> WH-word is stuck <strong>in</strong> <strong>the</strong> symmetric structure{WH{C=that<br />

. . .}}, where an <strong>in</strong>termediate labell<strong>in</strong>g <strong>in</strong>determ<strong>in</strong>acy arises which rema<strong>in</strong>s<br />

unresolved. In (8-a)/(9) on <strong>the</strong> o<strong>the</strong>r hand, Agree(uQ,WH) determ<strong>in</strong>es C as <strong>the</strong> label by (4).<br />

Whe<strong>the</strong>r or not IM of WH takes place at this po<strong>in</strong>t is irrelevant for labell<strong>in</strong>g: if WH moves, no<br />

problematic symmetry arises, as (4) renders C <strong>the</strong> label. But notice that WH has to move, because<br />

<strong>the</strong> labell<strong>in</strong>g question between v and WH is still not decided. Once <strong>the</strong> labell<strong>in</strong>g question<br />

is settled for v/WH (i.e. by mov<strong>in</strong>g WH), no fur<strong>the</strong>r movement is enforced despite <strong>the</strong> fact that<br />

<strong>the</strong> WH-element ends up be<strong>in</strong>g a sister of C. As no fur<strong>the</strong>r movement is enforced by symmetry,<br />

I take it that fur<strong>the</strong>r movement must not take place. In this sense, <strong>the</strong> function of (4) <strong>in</strong> <strong>the</strong><br />

current system is to stop fur<strong>the</strong>r movement, i.e. to prevent fur<strong>the</strong>r symmetries. Notice also that<br />

after Agree(uQ,WH) Q is valued, which means that for SEM it becomes <strong>in</strong>dist<strong>in</strong>guishable from<br />

<strong>in</strong>terpretable Q-features. As such it must be selected by <strong>in</strong>terrogative-embedd<strong>in</strong>g predicates,<br />

and no fur<strong>the</strong>r dist<strong>in</strong>ctions (as <strong>in</strong> Pesetsky/Torrego 2007) are needed.<br />

Consequences: The current analysis reconciles tensions between ‘late trigger’ <strong>the</strong>ories of successivecyclic<br />

movement that rely on (variants of) stipulations like Shortes Steps or <strong>the</strong> M<strong>in</strong>imize Cha<strong>in</strong><br />

L<strong>in</strong>ks Condition (cf. Boeckx 2003 and references <strong>the</strong>re<strong>in</strong>) on <strong>the</strong> one hand, and ‘early trigger’<br />

<strong>the</strong>ories that rely on spurious <strong>in</strong>termediate WH/EPP/Edge-features on <strong>the</strong> o<strong>the</strong>r (cf. Pesetsky/Torrego<br />

2007 among o<strong>the</strong>rs). The former are problematic <strong>in</strong> that <strong>the</strong>y violate <strong>the</strong> Extension<br />

Condition and are <strong>in</strong>compatible with phases. The latter rely on stipulated <strong>in</strong>termediate features<br />

to trigger movement. Also, <strong>the</strong> current analysis is less costly than Boˇsković’s (2007) <strong>in</strong><br />

requir<strong>in</strong>g less features: movement is a by-product of labell<strong>in</strong>g, which <strong>in</strong> his work is implicitly<br />

presupposed (and stipulated) to work without justification. But goal-driven movement is<br />

superfluous under <strong>the</strong> current assumptions.<br />

(1) a. if EM of XP and simplex H yields{H, XP}, <strong>the</strong>n H is <strong>the</strong> label<br />

b. ifαundergoes IM toβ, form<strong>in</strong>g{α,β} <strong>the</strong>n <strong>the</strong> label ofβis <strong>the</strong> label of{α,β}<br />

(2) {DP{v* . . .}}<br />

(3) C{DP{T{〈DP〉{v* . . .}}}}→C{DP{T{〈DP〉{v* . . .}}}}<br />

(4) Probe-Label Correspondence Axiom: The label of{α,β} is whichever ofαorβprobes<br />

<strong>the</strong> o<strong>the</strong>r, where <strong>the</strong> Probe=Lexical Item whose uF gets valued<br />

(5) {WH v*{. . .}}<br />

(6) {WH C{. . . T{〈WH〉 v*{. . .}}}}→WH . . .{〈WH〉 C{. . . T{〈WH〉 v*{. . .}}}}<br />

(7) a. [Which book]i do you [vP ti th<strong>in</strong>k/believe [CP ti C=that Mary [vP ti wrote ti]]]<br />

b. *You th<strong>in</strong>k/believe which book (that) Mary wrote?<br />

(8) a. John wonders [what C Mary will eat]<br />

b. *What does John wonder Mary will eat?<br />

(9) [C [uQ] [. . . [v∗P what[Q] [v*=eat . . . ]]]]<br />

2


Lucas Champollion (Tüb<strong>in</strong>gen)<br />

Temporal dependencies: Anaphora vs. movement<br />

Temporal dependencies provide a useful testbed for syntactic and semantic <strong>the</strong>ories of<br />

mean<strong>in</strong>g/context <strong>in</strong>teraction. In this talk, I consider two k<strong>in</strong>ds of temporal dependencies:<br />

those that hold between temporal adjuncts as <strong>in</strong> (1a) and (1b), and those that hold<br />

between before and its complement, as <strong>in</strong> (2a).<br />

(1) a. Last year, it ra<strong>in</strong>ed every day.<br />

b. Last year, it ra<strong>in</strong>ed <strong>in</strong> <strong>the</strong> afternoon.<br />

(2) a. John left before Mary claimed she would arrive.<br />

b. John left before Mary made <strong>the</strong> claim that she would arrive.<br />

These dependencies are rem<strong>in</strong>iscent of well-known phenomena such as quantifier doma<strong>in</strong><br />

restriction, pronoun b<strong>in</strong>d<strong>in</strong>g, and movement. For example, it looks like Last year restricts<br />

<strong>the</strong> doma<strong>in</strong> of every day <strong>in</strong> (1a) and provides an antecedent to <strong>in</strong> <strong>the</strong> afternoon <strong>in</strong> (1b).<br />

The dependency <strong>in</strong> (2a) is like movement <strong>in</strong> that it can be ei<strong>the</strong>r local (John left before<br />

Mary’s claim) or long-distance (John left before Mary’s claimed time of arrival), and <strong>in</strong><br />

that its long-distance <strong>in</strong>terpretation is blocked by island constra<strong>in</strong>ts (2b). Despite <strong>the</strong>se<br />

similarities, <strong>the</strong>re is currently no consensus on <strong>the</strong> status of temporal dependencies. For<br />

example, it is not agreed whe<strong>the</strong>r <strong>the</strong> process that allows last year to restrict <strong>the</strong> doma<strong>in</strong> of<br />

<strong>the</strong> quantifier every day <strong>in</strong> (1a) is due to movement (von Stechow, 2002) or variable b<strong>in</strong>d<strong>in</strong>g<br />

(Pratt and Francez, 2001, henceforth P&F), presupposition accommodation (Onea,<br />

2011) or contextual processes as <strong>in</strong> Recanati (2002). Determ<strong>in</strong><strong>in</strong>g <strong>the</strong> nature of temporal<br />

dependencies is <strong>the</strong>refore crucial <strong>in</strong> elucidat<strong>in</strong>g <strong>the</strong> role of context <strong>in</strong> <strong>in</strong>terpretation.<br />

The Pratt&Francez/von Stechow debate. P&F analyze temporal dependencies<br />

between adjuncts semantically. Technically, <strong>the</strong>y add an extra temporal argument λI to<br />

<strong>the</strong> mean<strong>in</strong>g of nouns, as <strong>in</strong> (3a), and transfer this up <strong>the</strong> tree by some ad hoc rules<br />

result<strong>in</strong>g <strong>in</strong> contextualized constituent mean<strong>in</strong>gs as <strong>in</strong> (3b)-(3d).<br />

(3) a. [day] = λIλx[day(x) ∧ τ(x) ⊆ I]<br />

b. [every day] = λP λI∀x[day(x) ∧ τ(x) ⊆ I → P (x)]<br />

c. [it ra<strong>in</strong>s every day] = λI∀x[day(x) ∧ τ(x) ⊆ I → it-ra<strong>in</strong>s(τ(x))]<br />

d. [(1a)] = λI∀y[year(y)∧τ(y) ⊆ I → ∀x[day(x)∧τ(x) ⊆ τ(y) → it-ra<strong>in</strong>s(τ(x))]]<br />

This solution is criticized by Stechow as “too local” because it does not allow for long<br />

distance dependencies as <strong>in</strong> (2a). Stechow favors a syntactic approach, where only one<br />

temporal adjunct attaches to <strong>the</strong> VP, and where <strong>the</strong> o<strong>the</strong>rs are embedded <strong>in</strong> it and undergo<br />

quantifier rais<strong>in</strong>g at LF. For example, Stechow’s LF of (1a) would be this:<br />

(4) [Last year] λi [every day (with<strong>in</strong>) ti] λj [it ra<strong>in</strong>ed (on) tj].<br />

This proposal. I argue that nei<strong>the</strong>r <strong>the</strong> semantic account of P&F nor <strong>the</strong> movementbased<br />

approach of Stechow captures <strong>the</strong> whole picture, s<strong>in</strong>ce <strong>the</strong>re is a previously unnoticed<br />

dichotomy with<strong>in</strong> <strong>the</strong> field of temporal dependencies: Attitude-verb dependencies,<br />

like (2a), are subject to syntactic movement constra<strong>in</strong>ts and should <strong>the</strong>refore be handled<br />

by a Stechow-style movement-based account. By contrast, dependencies between<br />

adjuncts, like (1a), systematically violate constra<strong>in</strong>ts on syntactic movement, as shown<br />

by <strong>the</strong> novel data below <strong>in</strong> (5) and (6). Specifically, <strong>the</strong>y can span because-clauses (5a),<br />

relative clauses (5b), if -clauses (5c), sentential subjects (5d), etc. Moreover, <strong>the</strong>y show<br />

<strong>the</strong> same range of <strong>in</strong>terpretations as pronouns and implicit variables: donkey read<strong>in</strong>gs<br />

(6a), quantificational (6b) and modal subord<strong>in</strong>ation (6c).


(5) a. Every year, John got anxious because he needed to file taxes <strong>in</strong> April.<br />

b. Every year, some guy who needs to file taxes <strong>in</strong> April gets anxious <strong>in</strong> March.<br />

c. Last year, I wonder if John went to France <strong>in</strong> August or <strong>in</strong> September.<br />

d. On most days, that it ra<strong>in</strong>s <strong>in</strong> <strong>the</strong> afternoon is a good possibility.<br />

(6) a. Whenever John spends a week <strong>in</strong> <strong>the</strong> Alps, he hikes every day.<br />

b. Every year, John spends a week <strong>in</strong> <strong>the</strong> Alps. Some years, he hikes every day.<br />

c. John might spend <strong>the</strong> next summer <strong>in</strong> <strong>the</strong> Alps. He would hike every day.<br />

Formal implementation. The parallel between temporal adjunct dependencies and<br />

anaphora has not been previously noticed and can shed new light on exist<strong>in</strong>g formal accounts<br />

of anaphoric dependencies. Follow<strong>in</strong>g Partee (1989), it is common to avoid mak<strong>in</strong>g<br />

a pr<strong>in</strong>cipled dist<strong>in</strong>ction between implicit dependencies and overt phenomena such as pronouns.<br />

We can <strong>the</strong>refore take several routes <strong>in</strong> model<strong>in</strong>g temporal adjunct dependencies.<br />

The textbook account of pronoun b<strong>in</strong>d<strong>in</strong>g (Heim and Kratzer, 1998) requires quantificational<br />

antecedents to move (possibly str<strong>in</strong>g-vacuously) – o<strong>the</strong>rwise <strong>the</strong>y cannot create<br />

an <strong>in</strong>dex, which <strong>in</strong> that system is required for b<strong>in</strong>d<strong>in</strong>g a pronoun. This movement appears<br />

unmotivated <strong>in</strong> <strong>the</strong> case of adjunct dependencies. I will present and discuss two<br />

syste<strong>ms</strong> which allow b<strong>in</strong>ders to rema<strong>in</strong> <strong>in</strong> situ: dynamic semantics and variable-free semantics<br />

(VFS). Each of <strong>the</strong>se syste<strong>ms</strong> has been successfully applied to implicit arguments<br />

(Condoravdi and Gawron, 1996; Pedersen, 2011). VFS is especially appeal<strong>in</strong>g because it<br />

allows us to free P&F of its drawbacks. It represents anaphoric dependencies as additional<br />

lambda slots, <strong>in</strong> <strong>the</strong> same way as P&F represent temporal dependencies. For example,<br />

<strong>the</strong> λx <strong>in</strong> (7) is directly comparable to <strong>the</strong> λI <strong>in</strong> (3a)-(3d). Through its g (Geach) rule,<br />

VFS provides a pr<strong>in</strong>cipled way of pass<strong>in</strong>g up <strong>the</strong>se slots, <strong>in</strong>clud<strong>in</strong>g at long distances if<br />

necessary. We <strong>the</strong>refore no longer need to rely on <strong>the</strong> ad hoc rules <strong>in</strong> P&F.<br />

(7) a. [him] = λx.x b. [Mary loves him] = λx[love(mary, x)]<br />

References<br />

Condoravdi, C. and Gawron, J. M. (1996). The context-dependency of implicit arguments.<br />

In Kanazawa, M., Piñón, C., and de Swart, H., editors, Quantifiers, deduction and<br />

context, pages 1–32. CSLI Publications, Stanford, CA.<br />

Heim, I. and Kratzer, A. (1998). Semantics <strong>in</strong> <strong>Generative</strong> Grammar. Blackwell Publish<strong>in</strong>g,<br />

Oxford, UK.<br />

Onea, E. (2011). On temporal quantification. In Reich, I., Horch, E., and Pauly, D.,<br />

editors, Proceed<strong>in</strong>gs of S<strong>in</strong>n und Bedeutung 15, pages 451–465. Universaar – Saarland<br />

University Press.<br />

Partee, B. H. (1989). B<strong>in</strong>d<strong>in</strong>g implicit variables <strong>in</strong> quantified contexts. In Wiltshire,<br />

C., Music, B., and Graczyk, R., editors, Papers from CLS 25, pages 342–356. Chicago<br />

<strong>L<strong>in</strong>guistics</strong> Society, Chicago, IL.<br />

Pedersen, W. A. (2011). Implicit arguments, paychecks and variable-free semantics. In<br />

Ashton, N., Chereches, A., and Lutz, D., editors, Proceed<strong>in</strong>gs of SALT 21, pages 155–<br />

175. Rutgers University.<br />

Pratt, I. and Francez, N. (2001). Temporal prepositions and temporal generalized quantifiers.<br />

<strong>L<strong>in</strong>guistics</strong> and Philosophy, 24(2):187–255.<br />

Recanati, F. (2002). Unarticulated constituents. <strong>L<strong>in</strong>guistics</strong> and Philosophy, 25:299–345.<br />

von Stechow, A. (2002). Temporal prepositional phrases with quantifiers: Some additions<br />

to Pratt and Francez (2001). <strong>L<strong>in</strong>guistics</strong> and Philosophy, 25:755–800.


Hsu-Te Cheng (UConn)<br />

Ellipsis: its Correlates with Phase and Movement<br />

Synopsis The aim of <strong>the</strong> paper is to argue for two generalizations regard<strong>in</strong>g ellipsis: (1) Phases cannot be<br />

elided. (2) The Movement Ellipsis Generalization (MEG): For a certa<strong>in</strong>ly category XP, if XP can be<br />

elided, it cannot be moved and vice versa. It will be argued that <strong>the</strong>se two generalizations can capture a<br />

wider range of data regard<strong>in</strong>g Argument Ellipsis (AE), and provide some <strong>in</strong>sight on how to detect phases.<br />

Introduction It is well known that languages differ <strong>in</strong> whe<strong>the</strong>r <strong>the</strong>y allow free occurrence of null<br />

arguments (Huang (1984)). One special type of null arguments, AE, where arguments, but not adjuncts,<br />

may be elided (cf. Oku (1998), Saito (2007), Takahashi (2008), among o<strong>the</strong>rs), has received close<br />

<strong>in</strong>spections recently. Japanese, for example, allows objects and subjects to be elided, as evidenced by <strong>the</strong><br />

availability of <strong>the</strong> sloppy read<strong>in</strong>g <strong>in</strong> (1)-(4), irrespective of <strong>the</strong>ir categories (DP, PP, CP, etc). AE,<br />

however, is fairly constra<strong>in</strong>ed. Cheng (2011) argues that <strong>the</strong> paradigm and <strong>the</strong> distribution of AE may be<br />

represented <strong>in</strong> (5). Only languages listed <strong>in</strong> (5a), but not those <strong>in</strong> (5b), are claimed to allow AE.<br />

(1) Taroo-ga [NP zibun-no hahaoya-o ] sonkeisiteiru Ziroo-mo [NP e ] sonkeisiteiru<br />

Taroo-nom self-gen mo<strong>the</strong>r-acc respect Ziroo-also respect<br />

‘Taroo respects his own mo<strong>the</strong>r. ‘Ziroo also respects e.’( OK strict, OK sloppy)<br />

(2) a. Taroo-wa [NP zibun-no teian-ga ] Hanako-o odorokasu to omotteiru<br />

Taroo-top self-gen proposal-nom Hanako-acc surprise that th<strong>in</strong>k<br />

‘Taroo th<strong>in</strong>ks that his proposal will surprise Hanako.’ (Takahashi (2008), p.404)<br />

b. Ken-wa [NP e ] Yumiko-o odorokasu to omotteiru ( OK strict, OK sloppy)<br />

Ken-top Yumiko-o surprise that th<strong>in</strong>k ‘lit. Ken th<strong>in</strong>ks that e will surprise Yumiko.’<br />

(3) a. Taroo to Hanako-ga [PP otagai kara ] meeru-o uketotta<br />

Taroo and Hanako-nom each.o<strong>the</strong>r from e-mail-acc received<br />

‘Taroo and Hanako received e-mail from each o<strong>the</strong>r.’<br />

b. Ken to Yumiko-wa [PP e ] tegami-o uketotta<br />

Ken and Yumiko-top letter-acc received ‘Ken and Yumiko received letters.’<br />

(4) a. Taroo-wa [CP zibun-no teian-ga Hanako-o odorokasu to] omotteiru<br />

Taroo-top self-gen proposal-nom Hanako-acc surprise that th<strong>in</strong>k<br />

‘Taroo th<strong>in</strong>ks that his proposal will surprise Hanako.’<br />

b. Ziroo-mo [CP e ] omotteiru<br />

Ziroo-also th<strong>in</strong>k ‘Ziroo also th<strong>in</strong>ks (that his proposal will surprise Hanako).’<br />

(5) a. Languages that allow AE (and sloppy <strong>in</strong>terpretation): Japanese Korean Ch<strong>in</strong>ese, Turkish…<br />

b. Languages that DO NOT allow AE (and sloppy <strong>in</strong>terpretation): English, French, German, Dutch…<br />

The Proposal I propose that <strong>the</strong>re is a general ban on ellipsis that phases cannot be elided, as <strong>in</strong> (6).<br />

English provides supports for this generalization. Arguments <strong>in</strong> English have been argued to be DPs (cf.<br />

Abney (1987)), and it has been argued that DPs are phases (Svenonius (2004), among o<strong>the</strong>rs). Therefore,<br />

it is <strong>the</strong>n expected that AE is not available <strong>in</strong> English, given (6), s<strong>in</strong>ce arguments are phases (DPs) <strong>in</strong><br />

English, as <strong>in</strong> (7). Moreover, it has been argued that vPs and CPs are strong phases <strong>in</strong> English. (8b) shows<br />

that CPs cannot be elided, conform<strong>in</strong>g with (6). As for vP ellipsis, Merchant (2008) argues that it is really<br />

VP ellipsis, based on <strong>the</strong> facts <strong>in</strong> (9) that ellipsis tolerates voice mismatches. Under <strong>the</strong> assumption that<br />

voice is hosted <strong>in</strong> v, it follows that what is elided is VP, not vP, conform<strong>in</strong>g with (6). For Japanese and<br />

Ch<strong>in</strong>ese, it has been argued that DP is miss<strong>in</strong>g <strong>in</strong> <strong>the</strong>se languages (Bošković (2008)). The existence of AE<br />

<strong>in</strong> <strong>the</strong>se languages <strong>the</strong>n follows, given that NP is not a phase (Despić (2011)) and thus can be elided.<br />

(6) The Non-Elidability Condition of Phases: If XP is a phase, XP cannot be elided <strong>in</strong> <strong>the</strong> PF component.<br />

(7) a. *John likes his mo<strong>the</strong>r, and Bill also likes [DP e ].<br />

b. *John th<strong>in</strong>ks that his son is smart. Bill also th<strong>in</strong>ks that [DP e ] is smart.<br />

(8) a. John suspects [CP whe<strong>the</strong>r Mary will pass <strong>the</strong> exam]. b. *Bill also suspects [CP e ].<br />

(9) a. This problem was to have been looked <strong>in</strong>to, but obviously nobody did [VP e ].<br />

b. Actually, I have implemented it with a manager, but it doesn’t have to be [VP e ].<br />

Analysis Hav<strong>in</strong>g substantiated <strong>the</strong> validity of (6), I argue that it may be derived from o<strong>the</strong>r <strong>in</strong>dependently<br />

motivated assumptions about phases and Spell-Out. I adopt <strong>the</strong> three commonly assumed proposals <strong>in</strong><br />

(10a-c) plus <strong>the</strong> condition <strong>in</strong> (10d). Given <strong>the</strong>se assumptions, <strong>the</strong> ban <strong>in</strong> (6) may be derived. As shown <strong>in</strong>


(11a), when <strong>the</strong> sentence is built to a phase (e.g. vP), <strong>the</strong> complement (VP) will be sent to Spell-Out. The<br />

PF component can choose whe<strong>the</strong>r to realize it properly (with full phonological realization) or not, as <strong>in</strong><br />

(11b). Later, when <strong>the</strong> structure is built to <strong>the</strong> next higher phase, as <strong>in</strong> (11c), <strong>the</strong> shaded element <strong>in</strong> (11c),<br />

namely v to IP, will be sent to Spell-Out, given <strong>the</strong> cyclic (phase-by-phase) nature of Spell-Out. The<br />

whole result will be ei<strong>the</strong>r sluic<strong>in</strong>g, if <strong>the</strong>se elements are spelled out as null, as <strong>in</strong> (11d), or VP ellipsis, if<br />

<strong>the</strong>y are properly realized, as <strong>in</strong> (11e). The condition <strong>in</strong> (6) is thus derived. The complement and <strong>the</strong> edge<br />

of <strong>the</strong> phase are always <strong>in</strong> two different spell<strong>in</strong>g out cycles. The only way to elide <strong>the</strong> whole phase is, <strong>in</strong><br />

<strong>the</strong> second Spell-Out, to realize <strong>the</strong> edge of <strong>the</strong> phase as null but to assign <strong>the</strong> rest of <strong>the</strong> spell<strong>in</strong>g out cycle<br />

proper realization. This option, however, is impossible, given that PF cannot handle elements with<br />

<strong>in</strong>consistent phonological features <strong>in</strong> one spell<strong>in</strong>g out cycle, as <strong>in</strong> (10d). In o<strong>the</strong>r words, <strong>in</strong> (11c) PF<br />

cannot realize v and vP as null but realize I and IP properly. The ban on elid<strong>in</strong>g phases is thus derived.<br />

One consequence of this condition is that <strong>the</strong> category that may be elided is restricted to complements of<br />

phase heads, given that only <strong>the</strong> complement of phase head HP may be sent to Spell-Out, as <strong>in</strong> (12).<br />

(10) a. Derivations proceed successive cyclically and cycle is def<strong>in</strong>ed by phase (Cho<strong>ms</strong>ky (2000, 2001)).<br />

b. Transfer: In phase P with head HP, Transfer applies to <strong>the</strong> complement doma<strong>in</strong> of HP as<br />

soon as <strong>the</strong> edge of P is extended. (Hiraiwa (2003))<br />

c. At each Spell-Out cyle, <strong>the</strong> PF component must decide whe<strong>the</strong>r to spell out <strong>the</strong> elements properly<br />

or not (with null/zero phonological realization). (cf. Holmberg (2001))<br />

d. PF Uniformity: In each cycle, PF can only handle elements with consistent phonological features.<br />

(11) a. [vP [ v [VP eat a cake ]]] b. [vP [ v [VP Δ ] ]] c. [CP C [IP I [vP v [VP ] ] ] ]<br />

� sent to Spell-Out � VP realized as null � sent to Spell-Out<br />

d. [CP C [IP Δ ] ] � sluic<strong>in</strong>g (IP-ellipsis) e. [CP C [IP I [vP v [VP Δ ] ] ] ] � VP ellipsis<br />

(12) a. [Phase H [ …… [Phase Specifier H [XP complement] ]]]<br />

|_____Spell-Out______| |__Spell-Out__|<br />

The Movement Ellipsis Generalization (MEG) MEG, as <strong>in</strong> (13), can <strong>the</strong>n be derived from <strong>the</strong> condition<br />

<strong>in</strong> (6). For a phrase to undergo ellipsis, it must be <strong>the</strong> complement of phase heads, which has been argued<br />

to be immobile, given <strong>the</strong> <strong>in</strong>terplay of PIC and Anti-Locality (cf. Abels (2003) and o<strong>the</strong>rs), as <strong>in</strong> (14). On<br />

<strong>the</strong> o<strong>the</strong>r hand, for a phrase to undergo movement, it must NOT be <strong>the</strong> complement of phase heads so that<br />

movement will not be blocked by PIC and Anti-Locality. However, not be<strong>in</strong>g <strong>the</strong> complement of phase<br />

will <strong>in</strong> turn exclude it to be eligible for ellipsis. MEG thus predicts that movement and ellipsis are<br />

mutually-exclusive for a given phrase, and this is empirically substantiated. (15-17) show that MEG holds<br />

for CP, DP, and IP. For VP, while it may be elided (18a), it cannot be moved (18b), evidence com<strong>in</strong>g<br />

from Huang’s (1993) b<strong>in</strong>d<strong>in</strong>g examples show<strong>in</strong>g what’s fronted is vP, which conta<strong>in</strong>s <strong>the</strong> subject trace.<br />

(13) MEG: For a certa<strong>in</strong>ly category XP, if XP can be elided, it cannot be moved and vice versa.<br />

(14) a. [Phase Spec [ HP XP ] ] b. [Phase Spec [ H XP ] ]<br />

(banned by Anti-Locality) (banned by PIC)<br />

(15) a. [CP that John will marry a supermodel]1, I will never believe t1 .<br />

b. *John believes that Mary is smart, and Bill also believes [CP e ].<br />

(16) a. [DP Banana]1, John really likes t1 . b. *John likes this book, but Bill does not like [DP e ].<br />

(17) a. *[IP Sam likes Sue]1, Joe doesn’t believe that t1 . b. Joe saw someone, but I don’t know who [IP e ].<br />

(18) a. Sam will come, but Sue will not [VP e ]. b. Sami said that [wash hi<strong>ms</strong>elf*i/j] Joej certa<strong>in</strong>ly would tvP.<br />

Discussion I fur<strong>the</strong>r argue that <strong>the</strong> existence of scrambl<strong>in</strong>g and AE <strong>in</strong> Japanese does not pose a problem<br />

to MEG. Follow<strong>in</strong>g Ohtaki (2011), I will argue that what is elided is not a full DP/NP <strong>in</strong> Japanese, but a<br />

sub-part of it, licensed by (null) KP, a phase. I will argue that <strong>the</strong> same holds for Mandar<strong>in</strong> Ch<strong>in</strong>ese as<br />

well. MEG thus provides us a tool to detect what a phase is, which may be moved but not elided.<br />

Selected References Abels, K. (2003). Successive cyclicity, anti-locality, and adposition strand<strong>in</strong>g. UConn<br />

dissertation. Abney, S. (1987). The English noun phrase <strong>in</strong> its sentential aspect. Doctoral Dissertation, MIT.<br />

Bošković, Ž. (2008). What will you have, DP or NP? In Proceed<strong>in</strong>gs of <strong>the</strong> 37 th North NELS. Cho<strong>ms</strong>ky, N. (2000).<br />

M<strong>in</strong>imalist Inquires. In Step by step, 89-155. Huang, J. (1984). On <strong>the</strong> distribution and reference of empty<br />

pronouns. LI 15: 531-574. Oku, S. (1998). A <strong>the</strong>ory of selection and reconstruction <strong>in</strong> <strong>the</strong> m<strong>in</strong>imalist perspective.<br />

Doctoral Dissertation, UConn. Saito, M. (2007). Notes on East Asian argument ellipsis. Language Research 43:<br />

203-227. Takahashi, D. (2008). Noun Phrase Ellipsis. In The Oxford Handbook of Japanese <strong>L<strong>in</strong>guistics</strong>, 394-422.


Sofiana Chiriacescu (Stuttgart)<br />

Focus<strong>in</strong>g on <strong>in</strong>def<strong>in</strong>ite noun phrases <strong>in</strong> German and English: consequences of reference<br />

form on <strong>the</strong> subsequent discourse.<br />

One of <strong>the</strong> fundamental questions underly<strong>in</strong>g <strong>the</strong>ories of language production concerns<br />

referent-track<strong>in</strong>g, <strong>in</strong>clud<strong>in</strong>g what referents are preferred to be picked up <strong>in</strong> <strong>the</strong> subsequent<br />

discourse and what types of referr<strong>in</strong>g expressions are used for this purpose. A body of<br />

l<strong>in</strong>guistic and psychol<strong>in</strong>guistic studies found out that several factors and criteria <strong>in</strong>fluence <strong>the</strong><br />

frequency of re-mention of referents, such as prom<strong>in</strong>ent syntactic positions (e.g. subjects,<br />

focus of clefts) and different <strong>the</strong>matic roles (e.g. Stimulus role <strong>in</strong> a transitive event with<br />

Stimulus and Experiencer roles). Given <strong>the</strong>ir high accessibility or prom<strong>in</strong>ence <strong>in</strong> ter<strong>ms</strong> of<br />

syntactic and semantic factors, <strong>the</strong>se referents are fur<strong>the</strong>rmore likely to be picked up <strong>in</strong> <strong>the</strong><br />

follow<strong>in</strong>g discourse by means of a more reduced type of referr<strong>in</strong>g expression (typically a<br />

pronoun) compared to <strong>the</strong>ir less-prom<strong>in</strong>ent counterparts (Ariel 2001, Grosz, Joshi and<br />

We<strong>in</strong>ste<strong>in</strong> 1995). In this paper, we focus on referents mentioned <strong>in</strong> non-prom<strong>in</strong>ent positions<br />

<strong>in</strong> English and German, i.e. as direct objects realized as <strong>in</strong>def<strong>in</strong>ite noun phrases, and argue<br />

that <strong>the</strong>y differ <strong>in</strong> ter<strong>ms</strong> of frequency of subsequent mention and likelihood of<br />

pronom<strong>in</strong>alization. The results are discussed <strong>in</strong> ter<strong>ms</strong> of how different types of <strong>in</strong>def<strong>in</strong>ite<br />

noun phrases affect <strong>the</strong> discourse structur<strong>in</strong>g potential of <strong>the</strong>ir referents dur<strong>in</strong>g reference<br />

production.<br />

Study1: The English data (Indef<strong>in</strong>ite-this vs. <strong>in</strong>def<strong>in</strong>ite-a)<br />

Accord<strong>in</strong>g to several studies (Pr<strong>in</strong>ce 1981, Ion<strong>in</strong> 2006), English this can be used as an<br />

<strong>in</strong>def<strong>in</strong>ite determ<strong>in</strong>er alongside <strong>the</strong> simple <strong>in</strong>def<strong>in</strong>ite article a(n). The Experiment (Exp1)<br />

<strong>in</strong>vestigates whe<strong>the</strong>r referents <strong>in</strong>troduced by this and a(n) differ <strong>in</strong> ter<strong>ms</strong> of likelihood of<br />

subsequent mention and likelihood pronom<strong>in</strong>alization. Design. We used a sentencecont<strong>in</strong>uation<br />

task with no pronoun-prompt. Participants (n=20) read story fragments (e.g. (1))<br />

and were asked to add five logical and natural-sound<strong>in</strong>g sentence cont<strong>in</strong>uations to each of <strong>the</strong><br />

stories. All critical referents were constructed <strong>in</strong> direct object position and were realized as<br />

<strong>in</strong>def<strong>in</strong>ite noun phrases. We only manipulated <strong>the</strong> morphological realization of <strong>the</strong> direct<br />

objects (2 conditions: this-condition and a(n)-condition). Note that <strong>in</strong>def<strong>in</strong>ite this occurs<br />

ma<strong>in</strong>ly <strong>in</strong> conversational English, thus, <strong>the</strong> test ite<strong>ms</strong> were kept <strong>in</strong> a colloquial tone.<br />

(1) Sample experimental item from Exp1<br />

this-condition a(n)-condition<br />

Yesterday even<strong>in</strong>g was so warm that Yesterday even<strong>in</strong>g was so warm that James<br />

James decided to hang out with friends decided to hang out with friends at <strong>the</strong> local<br />

at <strong>the</strong> local coffee shop. On his way coffee shop. On his way downtown, he saw<br />

downtown, he saw this kid com<strong>in</strong>g down<br />

<strong>the</strong> street.<br />

a kid com<strong>in</strong>g down <strong>the</strong> street.<br />

Each target item conta<strong>in</strong>ed <strong>in</strong>dividual references to two characters. In (1), for example, <strong>the</strong><br />

first referent (James) is <strong>the</strong> clearly established topic constituent of <strong>the</strong> story fragment, as it is<br />

mentioned twice (with a proper name and pronoun) <strong>in</strong> grammatical subject position. The<br />

critical item <strong>in</strong> (1), this kid, is <strong>in</strong>troduced as an <strong>in</strong>def<strong>in</strong>ite noun phrase <strong>in</strong> direct object position<br />

<strong>in</strong> <strong>the</strong> last clause of <strong>the</strong> story fragment. In light of previous studies on <strong>in</strong>def<strong>in</strong>ite-this<br />

(Gernsbacher & Shroyer 1989), we predict that this-referents will be: (i) more frequently<br />

picked up and (ii) more likely to be mentioned with a pronoun <strong>in</strong> <strong>the</strong> subsequent discourse,<br />

compared to a-referents. Results. The first part of our prediction was confirmed, as thisreferents<br />

were picked up <strong>in</strong> <strong>the</strong> subsequent discourse more often than a-referents (<strong>in</strong> 85% vs.<br />

15% of <strong>the</strong> cases). The second part of our prediction was not confirmed, as <strong>the</strong> anaphoric


expressions used for both <strong>in</strong>def<strong>in</strong>ite types were def<strong>in</strong>ite noun phrases (Fig1).<br />

Fig. 1. Type of referr<strong>in</strong>g expression used for <strong>the</strong> first re-mention of <strong>the</strong> referents <strong>in</strong> <strong>the</strong> this-condition and a(n)-condition<br />

Study2: The German data (Indef<strong>in</strong>ite-so’n vs. <strong>in</strong>def<strong>in</strong>ite-e<strong>in</strong>)<br />

The German determ<strong>in</strong>er so’n can be used <strong>in</strong> a similar way as English <strong>in</strong>def<strong>in</strong>ite this (von<br />

Heus<strong>in</strong>ger 2011). Experiment 2 (Exp2) had <strong>the</strong> same design, but tested <strong>the</strong> discourse behavior<br />

of <strong>in</strong>def<strong>in</strong>te-so’n compared to that of <strong>the</strong> simple <strong>in</strong>def<strong>in</strong>ite headed by e<strong>in</strong>(e) (‘a(n)’). Aga<strong>in</strong>,<br />

we manipulated only <strong>the</strong> type of <strong>in</strong>def<strong>in</strong>ite noun phrase, which resulted <strong>in</strong> 2 conditions: so’ncondition<br />

and e<strong>in</strong>(e)-condition, as <strong>in</strong> (2). Our prediction is, that if <strong>the</strong> accessibility of so’nreferents<br />

is comparable to that of referents preceded by <strong>in</strong>def<strong>in</strong>ite-this, <strong>the</strong>n <strong>the</strong> results of <strong>the</strong><br />

two experiments should be similar. Results: Similar to <strong>the</strong> f<strong>in</strong>d<strong>in</strong>gs of Exp 1, so’n-referents<br />

were picked up more often <strong>in</strong> <strong>the</strong> ensu<strong>in</strong>g discourse than <strong>the</strong> e<strong>in</strong>(e)-referents (<strong>in</strong> 80% vs. 17%<br />

of <strong>the</strong> cases), but did not show a preference for pronom<strong>in</strong>alization (Fig.2).<br />

Fig. 2. Type of referr<strong>in</strong>g expression used for <strong>the</strong> first re-mention of <strong>the</strong> referents <strong>in</strong> <strong>the</strong> so’n-condition and e<strong>in</strong>(e)-condition<br />

Conclusions: First, both <strong>in</strong>def<strong>in</strong>ite this and <strong>in</strong>def<strong>in</strong>ite so’n signal <strong>the</strong> likelihood of subsequent<br />

mention of <strong>the</strong>ir referents. Second, <strong>the</strong> f<strong>in</strong>d<strong>in</strong>gs of both Exp1 and Exp2 underl<strong>in</strong>e <strong>the</strong><br />

necessity to dissociate between likelihood of subsequent mention and likelihood of<br />

pronom<strong>in</strong>alization (confirmation of recent studies on language production, e.g. Kehler, Kertz,<br />

Rohde & Elman 2008), as <strong>the</strong>y do not po<strong>in</strong>t to <strong>the</strong> same type of accessibility of a referent.<br />

Third, we argue that <strong>the</strong> different markers of <strong>in</strong>def<strong>in</strong>ite noun phrases (i.e. this <strong>in</strong> English and<br />

so’n <strong>in</strong> German) were developed to dist<strong>in</strong>guish between accessible and non-accessible<br />

referents when realized as direct objects, as such referents are better competitors for <strong>the</strong><br />

subject referents, at least <strong>in</strong> ter<strong>ms</strong> of likelihood of subsequent mention.<br />

References:<br />

*Ariel, M.: Accessibility <strong>the</strong>ory: An overview. In Ted J.M. Sanders, Joost Schilperoord & Wilbert Spooren<br />

(eds.), Text representation: L<strong>in</strong>guistic and psychol<strong>in</strong>guistic aspects, 29–87. A<strong>ms</strong>terdam: John Benjam<strong>in</strong>s (2001).<br />

*Gernsbacher, M & Shroyer, A.: The cataphoric use of <strong>the</strong> <strong>in</strong>def<strong>in</strong>ite this <strong>in</strong> spoken narratives. Memory &<br />

Cognition 1989, 17 (5), 536-540. *Grosz, A. Joshi, A. & We<strong>in</strong>ste<strong>in</strong>, S.: Center<strong>in</strong>g: a framework for modell<strong>in</strong>g<br />

<strong>the</strong> local coherence of discourse. Computational <strong>L<strong>in</strong>guistics</strong> 21: 203 - 225 (1995). *Ion<strong>in</strong>, T.: This is def<strong>in</strong>itely<br />

specific: specificity and def<strong>in</strong>iteness <strong>in</strong> article syste<strong>ms</strong>. Natural Language Semantics. Spr<strong>in</strong>ger. 14. 175-234<br />

(2006). *von Hes<strong>in</strong>ger, K.: Specificity, referentiality and discourse prom<strong>in</strong>ence: German <strong>in</strong>def<strong>in</strong>ite<br />

demonstratives. In Reich, Ingo et al. (eds.), Proceed<strong>in</strong>gs of S<strong>in</strong>n & Bedeutung 15, Saarland University Press:<br />

Saarbrücken, Germany (2011). *Kehler, A., Kertz, L., Rohde, H., & Elman, J. L.: Coherence and coreference<br />

revisited. Journal of Semantics, 25, 1–44 (2008).


Kaily Clackson (Essex), Vera Heyer (Potsdam) & Harald Clahsen (Potsdam)<br />

Onl<strong>in</strong>e application of B<strong>in</strong>d<strong>in</strong>g Pr<strong>in</strong>ciples A and B:<br />

Evidence from eye movements dur<strong>in</strong>g listen<strong>in</strong>g<br />

Previous experiments exam<strong>in</strong><strong>in</strong>g <strong>the</strong> time course of <strong>the</strong> application of b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts<br />

dur<strong>in</strong>g onl<strong>in</strong>e process<strong>in</strong>g <strong>in</strong> adults have yielded mixed results. Early cross-modal prim<strong>in</strong>g<br />

experiments suggested that only grammatically accessible antecedents were considered as<br />

potential antecedents for a pronoun or reflexive, lead<strong>in</strong>g to <strong>the</strong> proposal that b<strong>in</strong>d<strong>in</strong>g<br />

pr<strong>in</strong>ciples act as an ‘<strong>in</strong>itial filter’, restrict<strong>in</strong>g <strong>the</strong> search for an antecedent to syntactically<br />

appropriate positions (Nicol & Sw<strong>in</strong>ney 1989). More recent studies have shown that o<strong>the</strong>r<br />

sources of <strong>in</strong>formation such as gender, discourse salience and recency also contribute to<br />

determ<strong>in</strong>e <strong>the</strong> referent, although <strong>the</strong> question of whe<strong>the</strong>r <strong>the</strong>se factors operate <strong>in</strong> parallel<br />

with syntactic constra<strong>in</strong>ts or somewhat later is still controversial (see e.g. Badecker and<br />

Straub 2002, Sturt, 2003).<br />

To <strong>in</strong>vestigate <strong>the</strong> moment-to-moment process<strong>in</strong>g of reflexives and pronouns, <strong>the</strong> present<br />

study employed <strong>the</strong> visual world paradigm, monitor<strong>in</strong>g <strong>the</strong> eye-movements of 42 adult native<br />

speakers to ‘task-relevant visual contexts’ (Trueswell, 2008: 145) while <strong>the</strong>y listened to<br />

stories such as (1) and (2).<br />

(1) Reflexives<br />

Peter/Susan was wait<strong>in</strong>g outside <strong>the</strong> corner shop. He/She watched as Mr Jones<br />

bought a large box of popcorn for hi<strong>ms</strong>elf over <strong>the</strong> counter.<br />

(2) Pronouns<br />

Peter/Susan was wait<strong>in</strong>g outside <strong>the</strong> corner shop. He/She watched as Mr Jones<br />

bought a large box of popcorn for him/her over <strong>the</strong> counter.<br />

Each story <strong>in</strong>volved two characters and an object (e.g. popcorn <strong>in</strong> (1)). One character was<br />

structurally accessible as an antecedent for <strong>the</strong> reflexive/pronoun (e.g., Mr. Jones <strong>in</strong> (1)),<br />

and one character was structurally <strong>in</strong>accessible (e.g., Peter/Susan <strong>in</strong> (1)). The gender of <strong>the</strong><br />

<strong>in</strong>accessible antecedent ei<strong>the</strong>r matched or mismatched <strong>the</strong> gender of <strong>the</strong> reflexive/pronoun.<br />

Participants first saw a picture of <strong>the</strong> object, <strong>the</strong>n were asked to listen to <strong>the</strong> story, <strong>in</strong>dicat<strong>in</strong>g<br />

as quickly as possible ‘who is it for?’ by press<strong>in</strong>g a button correspond<strong>in</strong>g to one of 4 pictures:<br />

<strong>the</strong> two characters mentioned, <strong>the</strong> object, and a distracter character (whose gender did not<br />

match <strong>the</strong> reflexive/pronoun). A video camera record<strong>in</strong>g <strong>the</strong> participant’s face provided an<br />

onl<strong>in</strong>e measure of gaze direction at <strong>the</strong> po<strong>in</strong>t of <strong>the</strong> pronoun/reflexive, and button presses<br />

provided onl<strong>in</strong>e reaction times and accuracy scores reflect<strong>in</strong>g participants’ f<strong>in</strong>al offl<strong>in</strong>e<br />

<strong>in</strong>terpretation of <strong>the</strong> pronoun/reflexive.<br />

Offl<strong>in</strong>e accuracy scores showed that for reflexive sentences <strong>the</strong> experimental manipulation<br />

had little effect on <strong>the</strong> f<strong>in</strong>al <strong>in</strong>terpretation (Match: 95%; Mismatch: 98%), but for <strong>the</strong> pronoun<br />

sentences accuracy was significantly lower when <strong>the</strong> gender of <strong>the</strong> <strong>in</strong>accessible antecedent<br />

matched that of <strong>the</strong> pronoun (Match: 86%; Mismatch: 96%). In Match conditions almost all<br />

erroneous responses selected <strong>the</strong> <strong>in</strong>accessible antecedent. In both reflexive and pronoun<br />

sentences, reaction times were significantly slower when <strong>the</strong> <strong>in</strong>accessible antecedent<br />

matched <strong>in</strong> gender with <strong>the</strong> pronoun/reflexive (108<strong>ms</strong> difference for reflexive sentences,<br />

243<strong>ms</strong> for pronoun sentences).


For both reflexives and pronouns, gaze direction was affected by <strong>the</strong> gender of <strong>the</strong><br />

<strong>in</strong>accessible antecedent, with participants’ gaze mov<strong>in</strong>g more quickly away from <strong>the</strong> picture<br />

of <strong>the</strong> <strong>in</strong>accessible antecedent when it mismatched <strong>in</strong> gender with <strong>the</strong> reflexive/pronoun.<br />

Statistical analysis showed significant 3-way <strong>in</strong>teractions between Time, Antecedent<br />

(accessible/<strong>in</strong>accessible) and Gender of <strong>in</strong>accessible antecedent (match/mismatch) from <strong>the</strong><br />

earliest po<strong>in</strong>t at which eye-movements reflect process<strong>in</strong>g of <strong>the</strong> pronoun/reflexive (200<strong>ms</strong><br />

after <strong>the</strong> onset of <strong>the</strong> reflexive/pronoun).<br />

Results show that although <strong>the</strong> f<strong>in</strong>al <strong>in</strong>terpretation of a reflexive is constra<strong>in</strong>ed by pr<strong>in</strong>ciple A,<br />

a discourse salient <strong>in</strong>accessible antecedent is also considered as a potential referent from<br />

<strong>the</strong> earliest measurable time. In <strong>the</strong> case of pronouns (where <strong>the</strong> structural constra<strong>in</strong>t does<br />

not pick out a particular referent), <strong>the</strong> f<strong>in</strong>al <strong>in</strong>terpretation is affected by a recently mentioned<br />

gender match<strong>in</strong>g <strong>in</strong>accessible antecedent, and this competition is also reflected <strong>in</strong> onl<strong>in</strong>e<br />

measures.<br />

Our results provide support for <strong>the</strong> view (e.g., Badecker & Straub, 2002) that from <strong>the</strong><br />

earliest measurable po<strong>in</strong>t <strong>in</strong> time b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts work <strong>in</strong> tandem with o<strong>the</strong>r sources of<br />

<strong>in</strong>formation to determ<strong>in</strong>e <strong>the</strong> referent for a reflexive or pronoun.<br />

References<br />

Badecker, W., Straub, K. (2002) The process<strong>in</strong>g role of structural constra<strong>in</strong>ts on <strong>the</strong><br />

<strong>in</strong>terpretation of pronouns and anaphors. Journal of Experimental Psychology:<br />

Learn<strong>in</strong>g, Memory and Cognition, 28 748-769<br />

Nicol, J., Sw<strong>in</strong>ney, D. (1989) The role of structure <strong>in</strong> coreference assignment dur<strong>in</strong>g<br />

sentence comprehension. Journal of Psychol<strong>in</strong>guistic research, 18 5-19.<br />

Sturt, P. (2003) The time-course of <strong>the</strong> application of b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts <strong>in</strong> reference<br />

resolution. Journal of Memory and Language, 48 542-562.<br />

Trueswell, J. (2008). Us<strong>in</strong>g eye movements as a developmental measure with<strong>in</strong><br />

psychol<strong>in</strong>guistics. Developmental Psychol<strong>in</strong>guistics: On-l<strong>in</strong>e methods <strong>in</strong> children‘s<br />

language process<strong>in</strong>g, 73-96.


����� ����������� �� ��� ������ ��� ������ �������������<br />

� ������������ ��������� ��� ����������� �����<br />

�� ���� ������ � ����� ��� � ��� ��������� ��� ����������� ����� ���� ������������� ���� ������<br />

�� � ����� ����������� �������� �� �������� ������ ����������� ����������� �������� ��<br />

�������� �������� ���� ������ ����� �� �������� �� ������������ �� ���� �� ������� �������<br />

�� � ��� ���� �� ��������� ���� �������� ��������� ��� ���������� ��� ��������� ����������<br />

���������� �� ��� �� ������������ ���� �������� ��������������� ������ ��� ��������� ��<br />

���������� ��� ������� �������� ����� �� �� ��������� ��� �������� �� ���� ��� �����������<br />

�� ��� �������� ���������<br />

��� ���� ���� ���� �������� �� �� ��� �� ������� ��� �� ������������ ������ �����������<br />

���� ��� ��������� �� ��� ��� ��� ������ �� �� ��� ���� ������ �� ��� ����� �� ��� ��� �����<br />

��� ���������� �� ��� ������� � �������� �������� ����� ������ �� ���������� ���� �������<br />

���� ��� ���� ���������� ��� �� ������� �� ����� �� ���� ����� ��� ����������� ������ ����<br />

��� ����� ������� ���� ���� �� ��� ��� �� ���� ���<br />

��� �� ���� ��� ���� ��� ���� ����<br />

�� ���� ��� ��� ������<br />

���� ���� �� ����<br />

��� �� ��� ��� ��� ����� ��� ��� ������<br />

�� ���� ��� ��� ������<br />

�� �� ���� ����<br />

� ������� �� ������� ��� ����� ������� ������� ����� � �������� ������ ��������� �� �����������<br />

������� �������� ���� ��� ����� �������� ������ ��� ��� ����� �������� �� ���� ����� �����������<br />

���� � ����� �τ �������� �� ��� ����� ����� � �������� ���� �τ �� ����������� ����� ���� ��������<br />

��� �������� �� ���� ������ ��� ����� ��� ������ ������ �� � ��������� ��� ���������<br />

������������ ��� ��������� ��� ������ ���� �� ��������� ����������� �������������� �������������<br />

��� ���� ������� �� ���� ���� �������� �� ����������� ����� ���� ���� ������ ����<br />

��� �� [[���i α]] o�g � λx.[[α]] o�gx/i<br />

�������� �������� �����<br />

�� [[���i α]] f�g � {λx.[[α]] f�gx/i } ����� �������� �����<br />

������������ ������������ ������� ���� ���� �� ��� ����� �� � ����� �������� �������� �� ���<br />

������� � ������ �� �� ���� ��������� ������������ �� ����� ������ ������� �� � ������ �����<br />

����� ���������� ���� �� ������������� ������ �� ������ ���� �� ��������� �� ��� ����������<br />

������ ������ ������� ��� � ���� ������� ����������<br />

���<br />

∼<br />

�����<br />

!<br />

����<br />

"<br />

��<br />

Noah Constant (UMass Amherst)<br />

[[!]] f�g �<br />

#<br />

���<br />

�<br />

{���� ��� ��� ������ ���� ��� ��� ������ �}�<br />

{���� ��� ��� ������ ���� ��� ��� ������ �}�<br />

� � �<br />

�<br />

[["]] f�g � {λx.{x ��� ������ x ��� ������ �}}<br />

[[#]] f�g � {g��� ��� ������ g��� ��� ������ �}<br />

� ��� ����� ��<br />

�� ���� ��������� �� ��� �������� ���� ��� ����� ����� �� � �������� ���� ���������� ��<br />

������� �� ��� ��������� ��������� �� ����� �� � ����������� ������� �� ������� ������ ��������<br />

�������� �∼�� �� ���� ��������������� �������������� �� �������� ∼ ��������� �� � ������ φ<br />

���� ���������� ���� ��� ������� �������� �� ������������ �������� �� ���� ��������� �� �<br />

������ �� [[φ]] f ���������� �� ����� ��� ��������� ��� �� ���������� ���������� [[φ]] o ���������<br />

������ ��� ��� ���� ���� ���� ������� ���� ��� ������� ������� � ������� �������� ��� ���������<br />

��������� ���� �� �� �������� ��������� �� ��� ���� ���� ��� � ���� ��� ��������� �� ����������<br />


��� ������������ ���� ��� ���� ���� ��������� �������� ������� ��� ��������� ��� ����<br />

����������������� ���������� �� � �������� ��� �������� ���� ��� ��� ����� �� ����� ����� ���<br />

������ �� ����� ������������ �� �������� ������ ����������� �������� �� �������� ��� ���������<br />

������������� ���� ��� ��� ������ �� ���������� �� ������<br />

��� ��������� ��������� �� ��� ����� ����������� ������� �� ���� �� ������� �� ����� ����� ��<br />

���������� �� ���������� ����� ������� ���� ��� ��������� ��� ��� ����� ������� �� ����������<br />

������� ��� �������� �������� �� ��������� ���� ��� ��� ������ �� �������� �������� ������<br />

�������� ��������� ������ ��� ����� ���������� ��� �� ������ �� ���� ��������� �� ���<br />

���� �������� ���� �� ���� �� ������������� ��� ��� ������ �� �� ���� � �������� ���� �������<br />

�������� �� ��� ���������� ����� ��� ����� ��������<br />

��� � $%&& !"#$%'(& )*+*', -.&'/*0<br />

���� ��������� ����� ������<br />

��� ���� ����� ����� ��� ����� ����<br />

������� ����� �<br />

��������� ���� �� ������� ���� � ������� ��� �������� ������ ������ �������� ���� ���<br />

���� ���� ��� ��������� ��� �������� �� ����� �� ������������� �� ��� �������� �������� ������<br />

����� ��������� ������� ������ ����� ����� ��������� ���� ������� �� �������� ����� ����� ��<br />

���������� ��� ����� �� �������� ��� ��� �������� �� ��������� ����������� �� ��� �������<br />

�������� �� ��� ������� ���������� ������� � ��������� �������� ��� � ������� ��������<br />

�� ��������� ������ �� ����������� ��� ����� ����� ��� �������� ��������� �� ��� ��������<br />

�� [[� ���� ��� ����]] o ������� ��� �������� {���� ��� ��� ������ ���� ��� ��� ������ �}� ����<br />

��� ����� ����� [[����� ���� � �� ��� ����]] f ���� �� ���� ��� ������� �������� �� ��� �� ����<br />

������ ��� ���� ������� ���� ��� ���� ����<br />

�������� ����� ����������� ����� ��� ��������� ��� ������� �� ��� ����� ���� ������� �� ��������<br />

� ����� ���� ���� �� ��������� �� ������� ��� ����� �� �������� ����� ���� ���� ����� ��� ���<br />

����� �� ���� ��� ����� ������ �� ��� ������ ��� �������� ����� ����� ��������� ������ �������<br />

�� � ����� �������� ���������� �� ������� ��������� ����� ���� ������� ���� ������ ������ ���<br />

��������� ���� ������� ��� �������� ��� ���� ���� ���� ���� ���� ������ �� �� ���� ���� ���<br />

����� ������� ���������� ���� ���� �� ������ ������ ���� ���� ��� ����� ����� �� ������� �������<br />

�� ����� �������� ������ �� ������� ����� �������� ��������� ������ ����� �� ���� � ������<br />

����������� ������� ��� ��� ��������� ��� ����� ����� ��������� �� ��� ����������� ���������<br />

��� �� �� ��������<br />

����<br />

���� ���� ���� ����<br />

����� �����<br />

�� ����<br />

�� �����<br />

�� ��������<br />

�� ����� �����<br />

���� ���� ���� ����<br />

�� ����<br />

�<br />

��� � ���� ��� ���������<br />

!!��� �� �������� �� ������<br />

�<br />

�<br />

��� �� �������� �� ������<br />

!!��� � ���� ��� ���������<br />

�������� ���������� � ������� �� ����� �� �������� ������ ��� ���������� ����������� ���<br />

���������� ��� ��������� �� ����� �� ��� ������������ �� �������� ��������� ��� �����������<br />

�������������� ��� �� ����� ����� ��� ��������� �� ����� ����������� ���������� ���<br />

��� ������� ����������� �� ����� �������� �������������� �� ���������� �������� ��� ������<br />

�������� �� ����� ����������� ��������� �� ���������� ������� �� ���������� ������ ������ ��<br />

����������� ���� ����� �� ��� ��� ������ �� ����� ����������� ���� ������ ����� �������<br />

��� ������� ����� �� ����� ������� ��������� ������� ������������ ����� ��� �������������<br />

���������� �� ����� ���� ��� ������� �� ������ �������� �� ����� ����������� ������ ������� ��<br />

������ ����� �� ����������� ���������� ������������ ������������ ��� ������������ �������������<br />

���� �� ����������� �� ����� ������ �� ������<br />

�<br />


NPI licens<strong>in</strong>g by exclusives: Just scope could ever expla<strong>in</strong> it<br />

We offer an analysis of <strong>the</strong> NPI licens<strong>in</strong>g properties of a range of exclusives, <strong>in</strong>clud<strong>in</strong>g adverbial<br />

and adjectival only, mere, exclusive, and just. Our proposal is that all exclusives license NPIs<br />

with<strong>in</strong> <strong>the</strong>ir semantic scope, and differences <strong>in</strong> behavior arise from differences of scope, which <strong>in</strong><br />

turn derive from differences of semantic type: some exclusives correspond to one application of <strong>the</strong><br />

Geach type-shift<strong>in</strong>g rule to Beaver and Clark’s (2008) lexical entry for only, and o<strong>the</strong>rs correspond<br />

to two.<br />

Data. Despite <strong>the</strong> synonymy between He is only a child and He is a mere child, <strong>the</strong> two exclusives<br />

cannot both license NPIs <strong>in</strong> <strong>the</strong> VP of a clause whose subject <strong>the</strong>y modify:<br />

(1) Only a child said anyth<strong>in</strong>g.<br />

Elizabeth Coppock (Lund) & David Beaver (U. Texas, Aust<strong>in</strong>)<br />

(2) *A mere child said anyth<strong>in</strong>g.<br />

The adjectival exclusives only and exclusive do not license NPIs <strong>in</strong> <strong>the</strong> VP ei<strong>the</strong>r, although <strong>the</strong>y do<br />

license NPIs <strong>in</strong> <strong>the</strong>ir syntactic scope:<br />

(3) a. *The only author got any royalties.<br />

b. The *(only) student who asked any questions got an A.<br />

(4) a. *The exclusive supplier of gas energy got any new contracts.<br />

b. As part of our agreement with your group, <strong>the</strong> Club is <strong>the</strong> *(exclusive) supplier<br />

of any beverage served on our property.<br />

Unlike only and exclusive, mere does not license NPIs <strong>in</strong> <strong>the</strong> modified noun phrase ((5)), but it<br />

does license NPIs outside its syntactic scope <strong>in</strong> certa<strong>in</strong> constructions ((6)):<br />

(5) *He is a mere author of any children’s books<br />

(6) I toiled for decades on a Wiscons<strong>in</strong> campus on which *(a mere) 18 percent of <strong>the</strong> enter<strong>in</strong>g<br />

freshmen [VP ever graduate ].<br />

Just sometimes behaves like mere, and sometimes like only. (7) can be paraphrased as (8).<br />

(7) Just <strong>the</strong> thought of him sends shivers down my sp<strong>in</strong>e.<br />

(8) The mere thought of him sends shivers down my sp<strong>in</strong>e.<br />

On this read<strong>in</strong>g (<strong>the</strong> scalar read<strong>in</strong>g), his touch would send shivers too, if noth<strong>in</strong>g worse. On <strong>the</strong><br />

non-scalar read<strong>in</strong>g, noth<strong>in</strong>g o<strong>the</strong>r than <strong>the</strong> thought of him, <strong>in</strong>clud<strong>in</strong>g his touch, would send <strong>the</strong><br />

shivers. This is <strong>the</strong> most prom<strong>in</strong>ent read<strong>in</strong>g with only:<br />

(9) Only <strong>the</strong> thought of him sends shivers down my sp<strong>in</strong>e.<br />

It is only on <strong>the</strong> non-scalar read<strong>in</strong>g that just licenses NPIs. (10) is acceptable when it means (11),<br />

but not on <strong>the</strong> most prom<strong>in</strong>ent read<strong>in</strong>g of <strong>the</strong> subject NP (≈ a mere smile from him).<br />

(10) ?Just a smile from him would make any difference.<br />

(11) Only a smile from him would make any difference.


Scalar particles and competition<br />

Synopsis. We provide an account for <strong>the</strong> variation <strong>in</strong> <strong>the</strong> distribution of scalar particles that is based<br />

on three assumptions: (i) scalar particles may move at LF (cf. Karttunen & Peters 1979), (ii) some<br />

scalar particles are morphologically complex (cf. Guerzoni 2003, Lahiri 2010), and (iii) scalar<br />

particles form scales and compete for <strong>in</strong>sertion.<br />

1. Cross-l<strong>in</strong>guistic variation. At first sight <strong>the</strong>re appear to be many distributional differences<br />

between scalar particles <strong>in</strong> different languages. However, a more careful exam<strong>in</strong>ation reveals that<br />

<strong>the</strong>ir distribution varies along only two dimensions: <strong>the</strong> pragmatic strength of <strong>the</strong> associate <strong>in</strong> <strong>the</strong><br />

immediate surface scope of <strong>the</strong> scalar particle and, roughly, whe<strong>the</strong>r <strong>the</strong> scalar particle can be<br />

characterized as an n-word (cf. Gast & van der Auwera 2011).<br />

1.1. First dimension of variation. Scalar particles can be classified with respect to whe<strong>the</strong>r <strong>the</strong>y may<br />

associate with pragmatically weak or strong elements <strong>in</strong> <strong>the</strong>ir immediate surface scope – we call<br />

<strong>the</strong>se particles weak and strong scalar particles, respectively. With respect to this criterion, <strong>the</strong>re are<br />

three ma<strong>in</strong> groups of scalar particles: (i) scalar particles that may be weak or strong (e.g. even <strong>in</strong><br />

English, même <strong>in</strong> French), (ii) scalar particles that may only be strong (e.g. sogar <strong>in</strong> German, hasta <strong>in</strong><br />

Spanish), and (iii) scalar particles that may only be weak (e.g. so much as <strong>in</strong> English, auch nur,<br />

e<strong>in</strong>mal <strong>in</strong> German). This idiosyncratic distribution is schematized <strong>in</strong> (1), while <strong>the</strong> relevant implicational<br />

generalization is <strong>in</strong> (2).<br />

(1) a. Hans read { even / sogar / *auch nur } SYNTACTIC STRUCTURES<br />

b. If Hans read { even / *sogar / auch nur } ONE book, he passed <strong>the</strong> exam<br />

(2) There is a scalar particle that is only strong <strong>in</strong> <strong>the</strong> language<br />

⇒ There is a scalar particles that is only weak <strong>in</strong> <strong>the</strong> language<br />

1.2. Second dimension of variation. Scalar particles that may only be weak can be split <strong>in</strong>to three<br />

subclasses: (iii.a) scalar particles that may occur both <strong>in</strong> <strong>the</strong> immediate scope of negation and <strong>in</strong> o<strong>the</strong>r<br />

DE environments (so much as <strong>in</strong> English), (iii.b) scalar particles that occur only <strong>in</strong> <strong>the</strong> immediate<br />

scope of negation (e.g. e<strong>in</strong>mal <strong>in</strong> German, niti <strong>in</strong> Slovenian), and (iii.c) scalar particles that never<br />

occur <strong>in</strong> <strong>the</strong> immediate scope of negation (e.g. auch nur <strong>in</strong> German, tudi <strong>in</strong> Slovenian). This<br />

distribution is schematized <strong>in</strong> (3), while <strong>the</strong> relevant implicational generalization is <strong>in</strong> (4).<br />

(3) a. Hans didn't read { so much as / e<strong>in</strong>mal / *auch nur } ONE book<br />

b. If Hans read { so much as / *e<strong>in</strong>mal / auch nur } ONE book, he passed <strong>the</strong> exam<br />

(4) There is a scalar particle that may only be weak and only occurs <strong>in</strong> <strong>the</strong> immediate scope of<br />

negation <strong>in</strong> <strong>the</strong> language ⇒ No o<strong>the</strong>r weak scalar particle that may only be weak occurs <strong>in</strong><br />

<strong>the</strong> immediate scope of negation <strong>in</strong> <strong>the</strong> language<br />

2. Derivation. We derive <strong>the</strong> variation described above from two parameters (whe<strong>the</strong>r a particle<br />

spells out one or two scalar components, whe<strong>the</strong>r a particle bears a negative feature) and <strong>in</strong>dependent<br />

pr<strong>in</strong>ciples <strong>in</strong> grammar (Maximize Presupposition, Relativized M<strong>in</strong>imality).<br />

2.1. Morphology. All scalar particles share <strong>the</strong> scalar component that requires its sister to denote a<br />

proposition that is less likely than a relevant alternative (Bennett 1982, Kay 1990). We represent this<br />

component with EVEN (5a). There is ano<strong>the</strong>r component to scalar particles, which is however not<br />

shared by all scalar particles: a scalar component that requires its sister to denote a proposition that is<br />

most likely among <strong>the</strong> alternatives (cf. Guerzoni 2003, Lahiri 2010). We represent this component<br />

with NUR (5b). These two components may move at LF (cf. Karttunen & Peters 1979) and <strong>the</strong>y<br />

associate with <strong>the</strong> same focused element (cf. Krifka 1991). F<strong>in</strong>ally, <strong>in</strong> some languages, <strong>the</strong> NUR<br />

component may bear an un<strong>in</strong>terpretable negative feature, mak<strong>in</strong>g <strong>the</strong> scalar particle comparable to nwords<br />

or n-<strong>in</strong>def<strong>in</strong>ites of <strong>the</strong> language.<br />

(5) a. [[ EVEN ]] g,c (C, p, w) is def<strong>in</strong>ed only if ∃q∈C [p < c q].<br />

If def<strong>in</strong>ed, [[ EVEN ]] g,c (C, p, w) = 1 iff p(w) = 1<br />

Luka Crnic (Hebrew University)


. [[ NUR ]] g,c (C, p, w) is def<strong>in</strong>ed only if ∀q∈C [q < c p].<br />

If def<strong>in</strong>ed, [[ NUR ]] g,c (C, p, w) = 1 iff p(w) = 1<br />

In a language like German, sogar spells out (6a), auch nur spells out (6b), while (nicht) e<strong>in</strong>mal spells<br />

out (6c). In English, even is ambiguous between (6a) and (6b), while so much as spells out (6b); <strong>the</strong>re<br />

is no scalar particle <strong>in</strong> English that bears a negative feature.<br />

(6) a. [EVEN] ↔ even; sogar b. [EVEN] [NUR] ↔ even; so much as; auch nur<br />

c. [EVEN] [NUR] [uNEG] ↔ (nicht) e<strong>in</strong>mal<br />

2.2. Expla<strong>in</strong><strong>in</strong>g distribution. The characterization <strong>in</strong> (6) does not suffice to account for <strong>the</strong> peculiar<br />

restrictions on <strong>the</strong> associates of sogar and auch nur. For example, sogar could occur with a weak<br />

associate where EVEN would move at LF. We block this by assum<strong>in</strong>g that sogar, auch nur and (nicht)<br />

e<strong>in</strong>mal form a scale and compete for <strong>in</strong>sertion (7). The same holds for its k<strong>in</strong> <strong>in</strong> o<strong>the</strong>r languages.<br />

(7) < [EVEN], [EVEN][NUR], [EVEN][NUR] [uNeg] ><br />

The competition of <strong>the</strong> particles <strong>in</strong> (7) is governed by Maximize Presupposition, which requires one<br />

to use among contextually equivalent alternatives <strong>the</strong> one with stronger presuppositions (Heim 1991).<br />

On <strong>the</strong> one hand, this necessitates sogar to be base-generated adjo<strong>in</strong>ed to clauses denot<strong>in</strong>g strong<br />

propositions: if <strong>the</strong> clause denoted a weak proposition, auch nur would have to be <strong>in</strong>serted s<strong>in</strong>ce this<br />

would lead to stronger presuppositions of <strong>the</strong> sentence (8).<br />

(8) a. ✗ [EVEN C 3] [I doubt [[EVEN C 3] you read one F book]] (spell out: sogar)<br />

b. ✔ [EVEN C 3] [I doubt [[EVEN C 3] [NUR C 0] you read one F book]] (spell out: auch nur)<br />

On <strong>the</strong> o<strong>the</strong>r hand, auch nur may only be adjo<strong>in</strong>ed to clauses denot<strong>in</strong>g weak propositions: if it were<br />

adjo<strong>in</strong>ed to a strong clause, <strong>the</strong>n NUR would ei<strong>the</strong>r trigger an <strong>in</strong>correct presupposition (9b) or it<br />

would have to move above EVEN which is ruled out by Relativized M<strong>in</strong>imality (9c). Fur<strong>the</strong>rmore,<br />

[EVEN][NUR] must be embedded under an appropriate non-upward entail<strong>in</strong>g operator s<strong>in</strong>ce <strong>the</strong> two<br />

scalar components would o<strong>the</strong>rwise trigger clash<strong>in</strong>g presuppositions (cf. Guerzoni 2003).<br />

(9) a. ✔ [I doubt [[EVEN C 3] you read (all) twelve F books]] (spell out: sogar)<br />

b. ✗ [EVEN C 3][I doubt [[EVEN C 3][NUR C 0] you read twelve F books]] (spell out: auch nur)<br />

c. ✗ [NUR C 3][I doubt [[EVEN C 3][NUR C 0] you read twelve F books]] (spell out: auch nur)<br />

[EVEN][NUR] [uNEG] may be spelled out only <strong>in</strong> <strong>the</strong> immediate scope of negation where its negative<br />

feature, which has no semantic reflex, can be checked. An Elsewhere Condition <strong>the</strong>n dictates that <strong>in</strong> a<br />

language that has both weak scalar particles, [EVEN][NUR] may not be spelled out <strong>in</strong> <strong>the</strong> immediate<br />

scope of negation but only <strong>in</strong> o<strong>the</strong>r non-upward-entail<strong>in</strong>g environments. Thus, (nicht) e<strong>in</strong>mal and<br />

auch nur are treated analogously to <strong>in</strong>def<strong>in</strong>ites nie and je, respectively (e.g. Penka & Zeijlstra 2005).<br />

As expected, <strong>the</strong>y exhibit a parallel distribution, e.g. both <strong>the</strong> <strong>in</strong>def<strong>in</strong>ite je and auch nur may occur<br />

under (covert) negation <strong>in</strong> <strong>the</strong> presence of an n-<strong>in</strong>def<strong>in</strong>ite (10). This is different <strong>in</strong> Slavic languages<br />

where all <strong>in</strong>def<strong>in</strong>ites and weak scalar particles under clausemate negation must be n-marked.<br />

(10) a. Niemand hat { je, *nie } etwas gegessen<br />

n-<strong>in</strong>def-nobody has { <strong>in</strong>def, *n-<strong>in</strong>def } someth<strong>in</strong>g eaten<br />

b. Ich habe nie { auch nur, *nicht e<strong>in</strong>mal, *e<strong>in</strong>mal } EIN Buch gelesen<br />

I have n-<strong>in</strong>def-never { auch nur, *nicht e<strong>in</strong>mal, *e<strong>in</strong>mal } one book read<br />

3. Conclusion. We have derived <strong>the</strong> five classes of scalar particles – (i)-(iii.c) above – <strong>in</strong> a<br />

framework that, first, allows for movement of EVEN and, second, assumes morphological variation<br />

among scalar particles. We left aside <strong>the</strong> issue of additivity of scalar particles; we propose that this is<br />

derived along <strong>the</strong> l<strong>in</strong>es put forward by Rullmann (1997). Ano<strong>the</strong>r issue left aside is <strong>the</strong> noncanonicity<br />

of movement of EVEN. A fur<strong>the</strong>r <strong>in</strong>quiry <strong>in</strong>to this is mandated.<br />

Selected references. Gast, V. & J. van der Auwera (2011) "Scalar additive operators <strong>in</strong> languages of Europe." Guerzoni,<br />

E. (2003) Why even ask? Lahiri, U. (2010) "Some evens are even ifs." Rullmann, H. (1997) "Even, polarity and scope."


Ian Cunn<strong>in</strong>gs & Patrick Sturt (Ed<strong>in</strong>burgh)<br />

The Time-Course of Reference Resolution <strong>in</strong> Picture Noun Phrases:<br />

Evidence from Eye-Movements Dur<strong>in</strong>g <strong>Read<strong>in</strong>g</strong><br />

Although b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory (Cho<strong>ms</strong>ky, 1981) as orig<strong>in</strong>ally formulated predicts that reflexives<br />

and pronouns should be <strong>in</strong> complementary distribution (e.g. ‘John1 <strong>in</strong>jured hi<strong>ms</strong>elf1/him*1’), a<br />

number of researchers have noted that complementarity breaks down <strong>in</strong> certa<strong>in</strong> contexts (e.g.<br />

Pollard & Sag, 1992; Re<strong>in</strong>hart & Reuland, 1993). One construction <strong>in</strong> particular where<br />

complementarity appears to break down is <strong>the</strong> picture noun phrase (PNP), as <strong>in</strong> ‘John1 saw a<br />

picture of hi<strong>ms</strong>elf1/him1’, and <strong>the</strong> question of whe<strong>the</strong>r or not possessed picture noun phrases<br />

(PPNPs; e.g. ‘John’s1 picture of hi<strong>ms</strong>elf1/him?1’) should also be exempt from b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory<br />

has been <strong>the</strong> subject of some debate (see e.g. Runner et al. 2006).<br />

Psychol<strong>in</strong>guistic research has exam<strong>in</strong>ed to what extent b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts are violable dur<strong>in</strong>g<br />

onl<strong>in</strong>e sentence process<strong>in</strong>g. While it has been argued that, at least for reflexives, b<strong>in</strong>d<strong>in</strong>g<br />

<strong>the</strong>ory applies early to help guide <strong>the</strong> antecedent search for b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory accessible<br />

antecedents (e.g. ‘John’ <strong>in</strong> ‘Steven knew that John had <strong>in</strong>jured hi<strong>ms</strong>elf’) ra<strong>the</strong>r than<br />

<strong>in</strong>accessible ones (e.g. ‘Steven’; see e.g. Sturt 2003) dur<strong>in</strong>g onl<strong>in</strong>e sentence comprehension, it<br />

has been claimed that <strong>in</strong>accessible antecedents can have early effects on process<strong>in</strong>g <strong>in</strong> (P)PNP<br />

contexts (Kaiser et al. 2009; Runner et al. 2006).<br />

We exam<strong>in</strong>ed <strong>the</strong> process<strong>in</strong>g of reflexives and pronouns <strong>in</strong> PNP and PPNP contexts <strong>in</strong> four<br />

eye-track<strong>in</strong>g experiments. In each experiment, 28 different native English speakers read 32<br />

critical and 64 filler texts while <strong>the</strong>ir eye-movements were monitored. Experiments 1 and 2<br />

exam<strong>in</strong>ed reflexives. Critical texts conta<strong>in</strong>ed one accessible and one <strong>in</strong>accessible antecedent,<br />

and gender congruence (match vs. mismatch) between each antecedent and <strong>the</strong> reflexive was<br />

manipulated <strong>in</strong> a 2x2 design. Congruence between <strong>the</strong> reflexive and <strong>in</strong>accessible antecedent<br />

was manipulated us<strong>in</strong>g proper names (Jonathan/Jennifer), while pre-tested gender stereotypes<br />

were used for <strong>the</strong> accessible antecedent (<strong>the</strong> soldier… hi<strong>ms</strong>elf/herself). PNP and PPNP<br />

contexts were tested <strong>in</strong> Experiments 1 and 2 respectively:<br />

Jonathan/Jennifer was walk<strong>in</strong>g through <strong>the</strong> military barracks. He/she heard...<br />

(1a) ... that <strong>the</strong> soldier had a picture of hi<strong>ms</strong>elf/herself <strong>in</strong> <strong>the</strong> middle of <strong>the</strong> mess hall.<br />

(1b) ... about <strong>the</strong> soldier’s picture of hi<strong>ms</strong>elf/herself <strong>in</strong> <strong>the</strong> middle of <strong>the</strong> mess hall.<br />

Experiments 3 and 4 exam<strong>in</strong>ed pronouns <strong>in</strong> PNP and PPNP contexts. Similar manipulations<br />

between <strong>the</strong> accessible and <strong>in</strong>accessible antecedents and a pronoun were used, except that this<br />

time we avoided use of <strong>the</strong> female pronoun as a result of its temporary ambiguity as a full<br />

noun phrase or specifier (see Clifton et al. 1997):<br />

The medical staff had a meet<strong>in</strong>g <strong>in</strong> <strong>the</strong> office. The surgeon/nurse recalled...<br />

(2a) ... that Jonathan/Jennifer noticed a portrait of him over at <strong>the</strong> back of <strong>the</strong> room.<br />

(2b) ... about Jonathan’s/Jennifer’s portrait of him over at <strong>the</strong> back of <strong>the</strong> room.<br />

The results of Experiments 1 and 2 <strong>in</strong>dicated <strong>the</strong> same relative time-course of antecedent<br />

effects for reflexives <strong>in</strong> both PNP and PPNP contexts. In both experiments, comparatively<br />

earlier read<strong>in</strong>g time measures were longer when <strong>the</strong> accessible antecedent mismatched <strong>the</strong><br />

stereotypical gender of <strong>the</strong> reflexive compared to when <strong>the</strong>re was a gender match, while<br />

effects of <strong>the</strong> <strong>in</strong>accessible antecedent were <strong>in</strong> comparison delayed. For example, <strong>in</strong> (1b),<br />

stereotypical gender mismatches between <strong>the</strong> reflexive and accessible antecedent <strong>in</strong>curred<br />

longer read<strong>in</strong>g times dur<strong>in</strong>g <strong>the</strong> <strong>in</strong>itial <strong>in</strong>spection of both <strong>the</strong> reflexive and a spillover region<br />

consist<strong>in</strong>g of <strong>the</strong> two words after <strong>the</strong> reflexive, whereas effects of <strong>the</strong> <strong>in</strong>accessible antecedent<br />

were only observed <strong>in</strong> second pass times of <strong>the</strong> spillover region. In Experiments 3 and 4, we


observed reliable effects of <strong>the</strong> accessible antecedent only. For example, <strong>in</strong> both experiments,<br />

second pass times of <strong>the</strong> pronoun were found to be reliably longer follow<strong>in</strong>g a stereotypical<br />

gender mismatch between <strong>the</strong> pronoun and accessible antecedent <strong>in</strong> comparison to when <strong>the</strong>re<br />

was a stereotypical gender match.<br />

The results of Experiments 1 and 2 extend previous f<strong>in</strong>d<strong>in</strong>gs that b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory applies early<br />

to help guide <strong>the</strong> resolution of reflexives (Sturt, 2003). Although previous offl<strong>in</strong>e studies have<br />

shown that comprehenders will accept a non-local ‘<strong>in</strong>accessible’ antecedent as a potential<br />

antecedent for a reflexive <strong>in</strong>side a (P)PNP (e.g. Asudeh & Keller, 2001; Runner et al. 2003<br />

Experiment 1), our data suggest that, contra <strong>the</strong> results of Runner et al. (2006) who did not<br />

<strong>in</strong>clude gender (mis)match<strong>in</strong>g control conditions <strong>in</strong> <strong>the</strong>ir visual world experiments, accessible<br />

antecedents only are <strong>in</strong>itially considered as a potential antecedent for a reflexive dur<strong>in</strong>g earlier<br />

stages of process<strong>in</strong>g dur<strong>in</strong>g read<strong>in</strong>g. The antecedent search for pronouns <strong>in</strong> (P)PNP contexts<br />

appears to be similarly constra<strong>in</strong>ed.<br />

References.<br />

Asudeh, A. & Keller, F. (2001). Experimental evidence for a predication-based b<strong>in</strong>d<strong>in</strong>g<br />

<strong>the</strong>ory. Papers from <strong>the</strong> 37 th meet<strong>in</strong>g of <strong>the</strong> Chicago L<strong>in</strong>guistic Society, 1-14.<br />

Cho<strong>ms</strong>ky, N. (1981). Lectures on Government and B<strong>in</strong>d<strong>in</strong>g. Dordrecht, Foris.<br />

Clifton, C., Kennison, S., & Albrecht, J. (1997). <strong>Read<strong>in</strong>g</strong> <strong>the</strong> words her, his, him: Implications<br />

for pars<strong>in</strong>g pr<strong>in</strong>ciples based on frequency and on structure. Journal of Memory and<br />

Language, 36, 276-292.<br />

Kaiser, E., Runner, J., Sussman, R., & Tanenhaus, M. (2009). Structural and semantic<br />

constra<strong>in</strong>ts on <strong>the</strong> resolution of pronouns and reflexives. Cognition, 112, 55-80.<br />

Pollard, C. & Sag, I. (1992). Anaphors <strong>in</strong> English and <strong>the</strong> scope of B<strong>in</strong>d<strong>in</strong>g Theory. L<strong>in</strong>guistic<br />

Inquiry, 23, 261-303.<br />

Runner, J., Sussman, R. & Tanenhaus, M. (2003). Assignment of reference to reflexives and<br />

pronouns <strong>in</strong> picture noun phrases: evidence from eye-movements. Cognition, 2003,<br />

B1-B13.<br />

- (2006). Process<strong>in</strong>g reflexives and pronouns <strong>in</strong> picture noun phrases. Cognitive<br />

Science, 30, 193-241.<br />

Re<strong>in</strong>hart, T. & Reuland, E. (1993). Reflexivity. L<strong>in</strong>guistic Inquiry, 24, 657-720.<br />

Sturt, P. (2003). The time-course of <strong>the</strong> application of b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts <strong>in</strong> reference<br />

resolution. Journal of Memory and Language, 48, 542-562.


Roberta D’Alessandro & Marc van Oostendorp (Leiden)<br />

Cyclic syntax mirrors cyclic phonology. Sou<strong>the</strong>rn Italian Vocatives <strong>in</strong> context.<br />

1. The issue. Vocatives pose a number of proble<strong>ms</strong> for various modules of l<strong>in</strong>guistic <strong>the</strong>ory. For<br />

syntax, <strong>the</strong>y raise <strong>the</strong> issue where exactly such 'case', as vocative is traditionally def<strong>in</strong>ed, is<br />

supposed to be assigned, given that it does not fit <strong>the</strong> traditional taxonomy of <strong>in</strong>herent vs. structural<br />

case: vocative is nei<strong>the</strong>r l<strong>in</strong>ked to any obvious <strong>the</strong>matic role, nor to any classical case position. Yet,<br />

it displays a dedicated end<strong>in</strong>g (or special phonological form; pace Schaden 2010). Syntactically,<br />

vocatives have been identified with topics (Lambrecht 1969, Portner 2004) or with isolated<br />

elements (Zwicky 2004), and <strong>the</strong>ir function has been classified <strong>in</strong> various ways (Schlegoff 1968,<br />

Zwicky 1974, Portner 2004, Schaden 2010) accord<strong>in</strong>g to <strong>the</strong>ir semantics. None of <strong>the</strong> syntaxsemantics<br />

based analyses, however, offers a solution for <strong>the</strong> follow<strong>in</strong>g phonological puzzle:<br />

vocatives are expressed <strong>in</strong> many languages as truncation patterns (e.g. Yapese, Jensen 1977;<br />

Indonesian, Cohn 2005; Arabic, Russian, Yadroff 1996), even though truncation is o<strong>the</strong>rwise very<br />

marked, while <strong>the</strong>re are no conv<strong>in</strong>c<strong>in</strong>g examples of e.g. Nom<strong>in</strong>ative or Dative be<strong>in</strong>g expressed <strong>in</strong><br />

this way <strong>in</strong> any language. Observe that it has been noted that <strong>in</strong> <strong>the</strong> verbal doma<strong>in</strong>, imperatives<br />

show a very similar behavior to that of vocatives (e.g. <strong>in</strong> Hebrew, Bat-El 2002, and much of <strong>the</strong><br />

work by Zanutt<strong>in</strong>i & Portner on imperatives and exclamatives).<br />

2. Proposal. In this paper, we suggest that <strong>the</strong> syntactic and <strong>the</strong> phonological puzzles should be<br />

solved <strong>in</strong> tandem, as <strong>the</strong>y are two sides of <strong>the</strong> same issue: <strong>the</strong> exocentric phonological behaviour is<br />

a result of <strong>the</strong> peripheral/external syntactic position and of <strong>the</strong> fact that discourse-related features<br />

(which we will <strong>in</strong>clude among <strong>the</strong> edge features) are read by a different phonological cycle than φ-<br />

and case features. Follow<strong>in</strong>g Zanutt<strong>in</strong>i (2008), we claim that vocatives are different from imperative<br />

subjects. Yet, we ma<strong>in</strong>ta<strong>in</strong> that <strong>the</strong>y are l<strong>in</strong>ked to an [addressee] position, though be<strong>in</strong>g external to<br />

<strong>the</strong> sp<strong>in</strong>e (much like Cho<strong>ms</strong>ky’s adjuncts). Their position is with<strong>in</strong> <strong>the</strong> left periphery of <strong>the</strong> clause,<br />

and it is precisely such position which allows for prosodic exponence (<strong>in</strong> <strong>the</strong> sense of <strong>in</strong>tonation).<br />

We argue that truncation is – at least <strong>in</strong> <strong>the</strong> syste<strong>ms</strong> we are study<strong>in</strong>g here - Sou<strong>the</strong>rn Italian dialects<br />

– actually a response to <strong>the</strong> demands of an <strong>in</strong>tonational pattern imposed by <strong>the</strong> [addressee] head.<br />

There is converg<strong>in</strong>g evidence that <strong>the</strong>re are two “submodules” <strong>in</strong> syntax, although authors do not<br />

concur on <strong>the</strong>ir actual implementation: one type is 'core syntax' deal<strong>in</strong>g with <strong>the</strong>matic relations etc.<br />

and correspond<strong>in</strong>g roughly to propositional semantics. The o<strong>the</strong>r is ‘peripheral syntax’, and deals<br />

with 'edge features' (Cho<strong>ms</strong>ky 2000 ff.), i.e. features <strong>in</strong>duc<strong>in</strong>g movement which impacts semantic<br />

<strong>in</strong>terpretation, and corresponds to discourse semantics (topic-focus structure, question formation,<br />

etc.). The fact that <strong>the</strong>se two types of syntax also roughly correlate with different phonological<br />

implementation is less frequently observed, although it see<strong>ms</strong> quite evident: <strong>in</strong> typical non-tonal<br />

languages, (<strong>in</strong>tonational) tone can be used to mark e.g. questions, or focus, but it is never used to<br />

mark e.g. passive formation or number agreement.<br />

The syntactic split has received several types of formalization. Most prom<strong>in</strong>ently, <strong>the</strong>re have been<br />

those who argue that <strong>the</strong> split is somewhere <strong>in</strong> <strong>the</strong> syntax, so that <strong>the</strong>re are at least two k<strong>in</strong>ds of<br />

syntactic features (φ –and case- features, ma<strong>in</strong>ly active <strong>in</strong> <strong>the</strong> A-doma<strong>in</strong>, and “edge” and discourse<br />

features like Speaker and Addressee, ma<strong>in</strong>ly active <strong>in</strong> <strong>the</strong> left periphery, Speas 2000, 2004;<br />

Sigurðsson 2000, 2001 ff.) correspond<strong>in</strong>g to different semantics (and, we would say, different<br />

phonologies), vs. those who claim that <strong>in</strong>formation structure is only extrasyntactic. We argue that<br />

vocatives shed light on this issue, especially if looked at from a cross-l<strong>in</strong>guistic po<strong>in</strong>t of view. On<br />

<strong>the</strong> one hand, <strong>the</strong>y have many of <strong>the</strong> properties of 'external' syntax, but on <strong>the</strong> o<strong>the</strong>r hand, it is<br />

difficult to l<strong>in</strong>k <strong>the</strong>m precisely to <strong>in</strong>formation structure, and more importantly, <strong>the</strong>y can also be<br />

expressed <strong>in</strong> some languages by non-<strong>in</strong>tonational morphological means (e.g. <strong>the</strong> special case<br />

end<strong>in</strong>gs of Lat<strong>in</strong> for <strong>the</strong> vocative). We claim that this argues for a syntax-phonology implementation<br />

of <strong>the</strong> split.<br />

3. Data. As observed above, truncation has been attested as <strong>the</strong> phonological exponence of <strong>the</strong><br />

vocative for a number of languages. Here we concentrate on <strong>the</strong> case of Sou<strong>the</strong>rn Italian dialects. In<br />

<strong>the</strong>se dialects, a vocative is formed by a prefix (a or o) and a truncated form of <strong>the</strong> DP, which<br />

consists of all <strong>the</strong> phonological material up until and <strong>in</strong>clud<strong>in</strong>g <strong>the</strong> stressed vowel:


(1) a. (A) Mariagiova’! c. (A) Sje’! [Abruzzese]<br />

Mariagiovanna (VOC) Sjef (VOC)<br />

b. (A) Robbe’! d. (A) surelle de lu padre di Giuwa'!<br />

Roberto (VOC) Sister of Giovanni's fa<strong>the</strong>r (VOC)<br />

These examples show that <strong>the</strong> orig<strong>in</strong>al size of <strong>the</strong> DP does not matter: everyth<strong>in</strong>g until <strong>the</strong> stressed<br />

vowel is <strong>in</strong>cluded, and everyth<strong>in</strong>g follow<strong>in</strong>g it (<strong>in</strong>clud<strong>in</strong>g coda consonants) is deleted. An analysis<br />

of <strong>the</strong> phenomenon necessarily calls for a syntax-phonology <strong>in</strong>terface analysis.<br />

4. Phonological analysis. Phonologically, <strong>the</strong>re is one major issue to be resolved: <strong>the</strong> ra<strong>the</strong>r curious<br />

shape of <strong>the</strong> truncation. Alber (2010) argues that <strong>the</strong>re are two (OT) Alignment constra<strong>in</strong>ts at work:<br />

one ensur<strong>in</strong>g that <strong>the</strong> left edge of <strong>the</strong> constituent is preserved, and one mak<strong>in</strong>g sure that <strong>the</strong><br />

truncation ends <strong>in</strong> a stressed vowel. (Yet ano<strong>the</strong>r constra<strong>in</strong>t ensures that everyth<strong>in</strong>g between two<br />

preserved segments is also preserved.) Although this is def<strong>in</strong>itely descriptively adequate, it should<br />

be noted that hav<strong>in</strong>g a constra<strong>in</strong>t align<strong>in</strong>g a stressed vowel with <strong>the</strong> edge of a phrase is somewhat<br />

suspicious for a template. Fur<strong>the</strong>rmore, phonological accounts so far do not take <strong>in</strong>to account why it<br />

is exactly vocatives that show this behavior. We <strong>the</strong>refore propose a different analysis: <strong>the</strong> exponent<br />

of <strong>the</strong> vocative is (at least) a pitch accent, which is specified for be<strong>in</strong>g at <strong>the</strong> same time a boundary<br />

tone (i.e. H*%). The fact that it is specified as a pitch accent makes it want to be on <strong>the</strong> stressed<br />

syllable; <strong>the</strong> fact that it is a boundary tone makes it want to be at <strong>the</strong> edge of <strong>the</strong> constituent. The<br />

paradox is resolved by mak<strong>in</strong>g <strong>the</strong> stressed vowel be exactly on <strong>the</strong> edge of <strong>the</strong> constituent – i.e. by<br />

truncation of everyth<strong>in</strong>g follow<strong>in</strong>g it.<br />

5. Syntactic analysis. Although we believe that this phonological account describes <strong>the</strong> truncation<br />

<strong>in</strong> a more elegant way (as well as one which makes more precise predictions about <strong>the</strong> <strong>in</strong>tonational<br />

pattern), it does not yet expla<strong>in</strong> why truncation patterns are so typical for vocatives (as well as for<br />

imperatives) and why <strong>the</strong>y do not occur for o<strong>the</strong>r morphological cases or verbal tenses. As a po<strong>in</strong>t<br />

of departure for a formalization of <strong>the</strong> 'two syntaxes' mentioned above, we take Cho<strong>ms</strong>ky's (2000)<br />

dist<strong>in</strong>ction between 'edge features' and 'φ-features', where <strong>the</strong> latter seem typical for '<strong>in</strong>ternal'<br />

syntax, and <strong>the</strong> former for 'external syntax'. We take ‘edge features’ to <strong>in</strong>clude all “semantic” and<br />

discourse-related features (i.e. not only topic-focus or wh, but also Speaker and Addressee). We<br />

propose that <strong>the</strong>se two types of features are derivationally dist<strong>in</strong>ct, and syntax has, as it were, two<br />

<strong>in</strong>dependent cycles: first, one <strong>in</strong> which φ-features operate, and afterwards one <strong>in</strong> which edge<br />

features are operative. Each of <strong>the</strong>se two cycles has its own spell-out: <strong>the</strong> former to segmental<br />

phonology (and propositional semantics), <strong>the</strong> second to <strong>in</strong>tonational and o<strong>the</strong>r types of prosodic<br />

phonology (and discourse semantics). The reason why <strong>in</strong>tonation etc. can only play a role <strong>in</strong> <strong>the</strong><br />

second cycle is obvious: syntactic heads which phonologically consist of only suprasegmental<br />

material can only be realized if segmental material has already been provided on an earlier cycle. A<br />

model such as this has as its advantage, fur<strong>the</strong>rmore, that both modules are syntactic (thus stay<strong>in</strong>g<br />

close to <strong>the</strong> conservative assumption that syntax is <strong>the</strong> sole mediator between phonology and<br />

semantics) while <strong>the</strong>y are also ordered and have different functions (thus captur<strong>in</strong>g some of <strong>the</strong><br />

<strong>in</strong>sights of <strong>the</strong> <strong>in</strong>formation structure based accounts).<br />

Vocative 'case' is obviously not assigned <strong>in</strong> <strong>the</strong> earlier cycle, s<strong>in</strong>ce it has no connection to <strong>the</strong>matic<br />

structure and is not a structural case <strong>in</strong> any way. We argue that it is assigned by an [addressee] head.<br />

Vocative is <strong>the</strong>refore different from o<strong>the</strong>r morphological cases, s<strong>in</strong>ce it is assigned <strong>in</strong> <strong>the</strong> 'outer<br />

cycle' of syntax. This opens <strong>the</strong> possibility for it to be realized by <strong>in</strong>tonational tones.<br />

6. Possible extensions. We show how an <strong>in</strong>tonational analysis might also work for some of <strong>the</strong><br />

o<strong>the</strong>r truncated vocatives of languages of <strong>the</strong> world, as well as possibly to imperatives (which<br />

similarly may be l<strong>in</strong>ked to an [adressee] head).<br />

Selected References<br />

Alber, Birgit. (2010) An exploration of truncation <strong>in</strong> Italian. Rutgers Work<strong>in</strong>g Papers <strong>in</strong> <strong>L<strong>in</strong>guistics</strong> vol. 3: 1-30. Bat-El,<br />

Outi. (2002). True truncation <strong>in</strong> Colloquial Hebrew Imperatives. Language 78.4: 651-665. Cabré, Teresa & Maria del<br />

Mar Vanrell. (2010) Non-templatic truncation: <strong>the</strong> case of vocatives. Cohn, Abby. (2005). Truncation <strong>in</strong> Indonesian.<br />

Evidence for violable m<strong>in</strong>imal words and AnchorRight. Proceed<strong>in</strong>gs of NELS 34, 175-189. Howell, Mortimer Sloper.<br />

1986. A grammar of <strong>the</strong> Classical Arabic language. Delhi: Gian Publish<strong>in</strong>g House. Jensen, John Thayer. (1977). Yapese<br />

Reference Grammar. Honolulu: University Press of Hawaii. Michael Yadroff (1996). Modern Russian Vocatives: A Case<br />

of Subtractive Morphology. Journal of Slavic <strong>L<strong>in</strong>guistics</strong>, pp. 133-153.


Alexandre Delf<strong>in</strong>o, Maria Luiza Cunha Lima & Pablo Arantes (Universidade Federal de M<strong>in</strong>as Gerais)<br />

Prosodic mark<strong>in</strong>g of referential status <strong>in</strong> Brazilian Portuguese<br />

A ris<strong>in</strong>g trend <strong>in</strong> <strong>the</strong> area of reference process<strong>in</strong>g clai<strong>ms</strong> that prosodic and contextual<br />

<strong>in</strong>formation are processed along with grammatical <strong>in</strong>formation as <strong>the</strong> hearer receives<br />

<strong>the</strong> <strong>in</strong>put. A view widely accepted (Gundel, 1993, <strong>in</strong>ter alia) posits referential<br />

expressions as tak<strong>in</strong>g a whole range of referential statuses. One central question on<br />

<strong>the</strong> production and perception of prosodically-encoded <strong>in</strong>formation structure relates<br />

to determ<strong>in</strong><strong>in</strong>g which of <strong>the</strong>se different statuses can receive a specific prosodic<br />

counterpart. Different methods have been used to tap <strong>in</strong>to <strong>the</strong>se differences, rang<strong>in</strong>g<br />

from phonological descriptions to ERP experiments.<br />

In west Germanic languages as English and German, a three-way dist<strong>in</strong>ction - new,<br />

given and accessible - is said to be dist<strong>in</strong>ct <strong>in</strong> ter<strong>ms</strong> of prosodic encod<strong>in</strong>g (Baumann,<br />

2006; Baumann & Grice, 2006). Us<strong>in</strong>g ToBI representation, new referents tend<br />

to be marked with a phrasal accent (H*), given referents tend to be deaccented<br />

and accessible referents tend to be marked with an <strong>in</strong>termediate phrase accent<br />

(H+L*). Us<strong>in</strong>g an ERP experiment, Schumacher & Baumann (2010) <strong>in</strong>vestigated<br />

how <strong>the</strong> prosodic <strong>in</strong>formation can affect <strong>the</strong> process<strong>in</strong>g of referents with three<br />

different <strong>in</strong>formational statuses: given, new and accessible. The N400 and latepositivity<br />

results lead <strong>the</strong> authors to conclude that (i) <strong>the</strong> three-way classification<br />

of <strong>the</strong> <strong>in</strong>formational status makes difference not only for production, but also for<br />

perception, and (ii) <strong>the</strong> prosodic <strong>in</strong>formation is processed very early, along with o<strong>the</strong>r<br />

grammatical and context <strong>in</strong>formation.<br />

We propose that <strong>the</strong>se methods can be fruitfully aided by more concrete acoustical<br />

analyses. We also <strong>in</strong>tend to <strong>in</strong>vestigate which acoustic parameters are more heavily<br />

<strong>in</strong>volved with this prosodic mark<strong>in</strong>g. Ano<strong>the</strong>r goal is <strong>the</strong> description of <strong>the</strong> prosody<strong>in</strong>formational<br />

status relationship <strong>in</strong> Brazilian Portuguese, a language scarcely<br />

studied, expand<strong>in</strong>g <strong>the</strong> <strong>in</strong>vestigation about prosody and referential status relationship<br />

to languages different from <strong>the</strong> West Germanic languages. More specifically,<br />

we tried to see if <strong>the</strong> proposed three-folded statuses can be found <strong>in</strong> Brazilian<br />

Portuguese as well.<br />

For this study, we designed a corpus of approximately 30 groups of sentences,<br />

distributed <strong>in</strong>to three conditions: given, new and accessible. For each group<br />

of sentences, we set one target word, which is embedded <strong>in</strong> a control phrase.<br />

Preced<strong>in</strong>g text determ<strong>in</strong>ed if <strong>the</strong> target NP was given, new or accessible. The<br />

sentences below illustrate <strong>the</strong> construction of <strong>the</strong> <strong>in</strong>formational status from <strong>the</strong><br />

context (target words <strong>in</strong> bold):<br />

New<br />

Um terremoto causou destruição em boa parte da costa leste. Várias cidades não<br />

t<strong>in</strong>ham um programa de evacuação, o que deu trabalho para as equipes de resgate.<br />

(An earthquake caused destruction <strong>in</strong> a huge part of <strong>the</strong> East coast. Several cities<br />

did not have an evacuation program, which caused proble<strong>ms</strong> to <strong>the</strong> rescue tea<strong>ms</strong>)<br />

Given<br />

O governo decidiu fechar a us<strong>in</strong>a nuclear após o terremoto ocorrido no mês<br />

passado. O terremoto causou destruição no núcleo do reator, aumentando o risco<br />

de contam<strong>in</strong>ação. (The government decided to shut down <strong>the</strong> nuclear plant after <strong>the</strong><br />

earthquake occurred last month. The earthquake caused destruction to <strong>the</strong> reactor<br />

nucleus, <strong>in</strong>creas<strong>in</strong>g <strong>the</strong> risk of contam<strong>in</strong>ation.)


Acessible<br />

Estudiosos da Sismologia têm procurado analisar os dados de tremores para prever<br />

novas ocorrências. O terremoto causou destruição sem que n<strong>in</strong>guem pudesse se<br />

prevenir.<br />

(Seismology experts have tried to analyse <strong>the</strong> tremors data to predict new<br />

occurences. The earthquake caused destruction without any one be<strong>in</strong>g able to<br />

prevent it.)<br />

Four participants read a list conta<strong>in</strong><strong>in</strong>g n<strong>in</strong>ety randomized groups of sentences,<br />

presented one by one <strong>in</strong> a computer screen. In order to ascerta<strong>in</strong> that <strong>the</strong> three<br />

referential statuses have dist<strong>in</strong>ct prosodic patterns, a number of acoustic features<br />

(traditionally seen as correlates of prosodic prom<strong>in</strong>ence) were analyzed: (a) target<br />

word duration, (b) F0 mean, standard deviation (SD) and range and (c) timenormalized<br />

F0 contours of <strong>the</strong> DP conta<strong>in</strong><strong>in</strong>g <strong>the</strong> target word. Data from <strong>the</strong> four<br />

subjects were analyzed separately. Analysis of variance was used to determ<strong>in</strong>e if<br />

differences <strong>in</strong> mean values of <strong>the</strong> acoustic parameters were statistically different.<br />

The statistical analysis showed that referential status does not significantly affect<br />

F0 SD and range for all speakers. Mean F0 and word duration were affected by<br />

referential status for three of <strong>the</strong> subjects (new referents have higher mean F0 and<br />

are longer than given and/or accessible ones). Analysis of <strong>the</strong> time-nomalized F0<br />

contours revealed that F0 contours of new referents are different from <strong>the</strong> given and<br />

accessible ones. Despite <strong>in</strong>dividual variability, new referents are characterized by <strong>the</strong><br />

presence of two major pitch peaks, one extend<strong>in</strong>g through <strong>the</strong> cha<strong>in</strong> of pre-stressed<br />

syllables and <strong>the</strong> o<strong>the</strong>r aligned to <strong>the</strong> stressed syllable. Given and accessible<br />

referents contours are very similar to each o<strong>the</strong>r and tend to: (a) have just one peak,<br />

ei<strong>the</strong>r aligned to <strong>the</strong> pre-stressed syllables or to <strong>the</strong> stressed or (b) be relatively<br />

flat. The current results lead us to conclude that from <strong>the</strong> speech production side<br />

prosody plays an important role signal<strong>in</strong>g referential status, ma<strong>in</strong>ly differentiat<strong>in</strong>g<br />

new referents from o<strong>the</strong>rs.<br />

References<br />

BAUMANN, S. 2006. The Intonation of Givenness - Evidence from German.<br />

L<strong>in</strong>guistische Arbeiten, n.508. Tüb<strong>in</strong>gen: Niemeyer.<br />

BAUMANN S, GRICE, M. 2006. The <strong>in</strong>tonation of accessibility. Journal of<br />

Pragmatics. v. 38, 1636–1657.<br />

GUNDEL, J. et al. 1993. Cognitive Status and <strong>the</strong> Form of Referr<strong>in</strong>g Expressions. In:<br />

Discourse. Language, v. 69, n. 2, 274-307.<br />

SCHUMACHER, P. & BAUMANN, S. 2010. Pitch accent type affects <strong>the</strong> N400<br />

dur<strong>in</strong>g referencial process<strong>in</strong>g. NeuroReport, vol. 21, n. 9, 618-622.


Ewan Dunbar (Maryland), Brian Dillon (UMass Amherst) & William Idsardi (Maryland)<br />

Learn<strong>in</strong>g phonetic categories by learn<strong>in</strong>g allophony & vice versa: a computational model<br />

We consider <strong>the</strong> <strong>in</strong>fant’s early problem of group<strong>in</strong>g signals <strong>in</strong>to phonetic categories, (Werker and<br />

Tees 1984), as well as <strong>the</strong> problem of learn<strong>in</strong>g allophonic processes, and demonstrate how <strong>the</strong>se<br />

relate to Tesar et al.’s (2003) problem of mutual dependence <strong>in</strong> language acquisition: learn<strong>in</strong>g<br />

one component of a full grammatical system affects what <strong>the</strong> (best) hypo<strong>the</strong>ses about some<br />

o<strong>the</strong>r component will be, and vice versa. We present <strong>the</strong> results of an unsupervised Bayesian<br />

statistical model of phonetic category and process learn<strong>in</strong>g which learns categories and grammar<br />

jo<strong>in</strong>tly, allow<strong>in</strong>g for <strong>in</strong>teraction between <strong>the</strong> two learn<strong>in</strong>g proble<strong>ms</strong>. We compare our model to<br />

modular learners, which separate phonetic category learn<strong>in</strong>g from grammatical learn<strong>in</strong>g, and<br />

demonstrate that <strong>the</strong> phonetic category learn<strong>in</strong>g problem and <strong>the</strong> grammar learn<strong>in</strong>g problem are<br />

mutually dependent. This implies that standard phonological learn<strong>in</strong>g models cannot be taken<br />

at face value, because, as <strong>the</strong>y are conventionally formulated, <strong>the</strong>y take <strong>in</strong>puts pre-categorized<br />

<strong>in</strong>to phones or features, and thus build <strong>in</strong> an implicit modularity assumption. We show fur<strong>the</strong>r<br />

that our model succeeds <strong>in</strong> learn<strong>in</strong>g contextual rules without observ<strong>in</strong>g <strong>the</strong> contexts <strong>the</strong><strong>ms</strong>elves.<br />

In particular, we consider data from Inuktitut (Eskimo-Aleut, Nor<strong>the</strong>rn Canada), a language<br />

with a three-vowel system. Uvular consonants trigger allophony for all three vowels: before<br />

[q] and [K], /i u a/ → [e o A] respectively (see Figure 1). We use phonetic data elicited from a<br />

native speaker (F1 × F2, segmented; Denis and Pollard 2008) and fit a Bayesian model of phonetic<br />

category syste<strong>ms</strong> as simple dependent Dirichlet processes (Ferguson 1973; MacEachern<br />

1999). Each category is a multivariate Gaussian distribution which is shifted by some amount<br />

<strong>in</strong> <strong>the</strong> context of a uvular. (The shift is a real vector, i.e., we model <strong>the</strong> allophonic process as<br />

sub-categorical; see Port and O’Dell 1986; Cohn 1990.) The output of learn<strong>in</strong>g is a system of<br />

categories and processes learned jo<strong>in</strong>tly. The model is unsupervised: it learns <strong>the</strong> number of<br />

categories, <strong>the</strong>ir locations and shapes, and <strong>the</strong> degree of pre-uvular retraction per category, entirely<br />

on <strong>the</strong> basis of <strong>the</strong> statistics of <strong>the</strong> <strong>in</strong>put. In Model 1, <strong>the</strong> learner is told whe<strong>the</strong>r each token<br />

preceded a uvular. The learner f<strong>in</strong>ds <strong>the</strong> three vowel phonemes of Inuktitut (Figure 2) with results<br />

more accurate than a standard phonetic category model (Dirichlet process mixture model,<br />

e.g. <strong>the</strong> distributional learn<strong>in</strong>g algorithm of Feldman et al. 2009, F statistic for classification<br />

0.75 for our model vs 0.69 for <strong>the</strong> standard model, 12 runs; t = 4.26, p < 0.001). Crucially, if<br />

we encourage a distributional learner to f<strong>in</strong>d more categories, lead<strong>in</strong>g to <strong>the</strong> discovery of all six<br />

allophones, and <strong>the</strong>n use <strong>the</strong>se categories’ context distributions as <strong>in</strong>put to <strong>the</strong> GLA (Boersma<br />

1997) or <strong>the</strong> statistical allophone learn<strong>in</strong>g model of Peperkamp et al. (2006), we fail to learn <strong>the</strong><br />

relevant categorical rule, as <strong>the</strong> learned allophones are too poorly aligned with <strong>the</strong> true ones.<br />

In Model 2, <strong>the</strong> learner is not told whe<strong>the</strong>r a uvular followed, but only <strong>the</strong> approximate frequency<br />

of follow<strong>in</strong>g uvulars. This simulates a learner who has not fully learned to categorize<br />

<strong>the</strong> relevant features of <strong>the</strong> context, or one who is learn<strong>in</strong>g a language <strong>in</strong> which a trigger<strong>in</strong>g<br />

environment is obscured by a later process <strong>in</strong> a counterbleed<strong>in</strong>g relation. This model succeeds<br />

<strong>in</strong> learn<strong>in</strong>g <strong>the</strong> vowels of Inuktitut and simultaneously discover<strong>in</strong>g <strong>the</strong> uvular/non-uvular dist<strong>in</strong>ction<br />

only by observ<strong>in</strong>g vowel tokens (Figure 3). As before, <strong>the</strong> model f<strong>in</strong>ds three categories<br />

and three rules (although it cannot reliably determ<strong>in</strong>e which phones are <strong>the</strong> output and which<br />

<strong>the</strong> <strong>in</strong>put of <strong>the</strong> rules). We show that this is not an accident of Inuktitut uvulars, and that <strong>the</strong><br />

model also learns a subset of English vowel categories along with systematic sex differences <strong>in</strong><br />

<strong>the</strong> pronunciations of those vowels, without observ<strong>in</strong>g <strong>the</strong> sex of <strong>the</strong> speaker.<br />

Our results demonstrate that (1) allophony and o<strong>the</strong>r contextual effects are crucial <strong>in</strong> learn<strong>in</strong>g<br />

phonetic categories; (2) <strong>the</strong> shape of phonetic categories <strong>the</strong><strong>ms</strong>elves can be helpful <strong>in</strong> detect<strong>in</strong>g<br />

<strong>the</strong> effects of context. This mutual dependence is consistent with <strong>the</strong> strongly <strong>in</strong>teractive<br />

character of category identification <strong>in</strong> speech perception (Liberman et al., 1953; Whalen et al.,<br />

1997). We conclude that this lack of strong modularity extends to learn<strong>in</strong>g.


Figure 1 Figure 2 Figure 3<br />

Figure 1: Phones of Inuktitut; [e], [A] and [o] are allophonically related to [i], [a], and [u].<br />

Figure 2: Output of Model 1. Dotted l<strong>in</strong>es: categories shifted by learned contextual rules.<br />

Figure 3: Output of Model 2 (contextual rules learned without knowledge of context).<br />

References<br />

de Boer, B., and P. Kuhl. 2003. Investigat<strong>in</strong>g <strong>the</strong> role of <strong>in</strong>fant-directed speech with a<br />

computer model. Acoustics Research Letters Onl<strong>in</strong>e 4.129–134.<br />

Boersma, P. How we learn variation, optionality, and probability. IFA Proceed<strong>in</strong>gs 21.37–42.<br />

Cohn, A. 1990. Phonetic and phonological rules of nasalization. UCLA dissertation.<br />

Denis, D., and M. Pollard. 2008. An Acoustic Analysis of The Vowel Space of Inuktitut.<br />

Presented at Inuktitut <strong>L<strong>in</strong>guistics</strong> Workshop, University of Toronto.<br />

Feldman, N., T. Griffiths, and J. Morgan. 2009. Learn<strong>in</strong>g Phonetic Categories by Learn<strong>in</strong>g a<br />

Lexicon. Proceed<strong>in</strong>gs of 31st Annual Conference of Cognitive Science Society 2208–2213.<br />

Ferguson, T. 1973. Bayesian analysis of nonparametric proble<strong>ms</strong>. Ann. of Stats 1.209–230.<br />

Liberman, M., and J. Pierrehumbert. 1984. Intonational <strong>in</strong>variance under changes <strong>in</strong> pitch<br />

range and length. In Language sound and structure. Cambridge: MIT Press.<br />

MacEachearn, S. 1999. Dependent nonparametric processes. ASA Proceed<strong>in</strong>gs of <strong>the</strong> Section<br />

on Bayesian Statistical Science 50–55.<br />

McMurray, B., R. Asl<strong>in</strong>, and J. Toscano. 2009. Statistical learn<strong>in</strong>g of phonetic categories:<br />

Insights from a computational approach. Developmental Science 12.369–78.<br />

Peperkamp, S., R. Le Calvez, J-P. Nadal, and E. Dupoux. 2006. The acquisition of allophonic<br />

rules: statistical learn<strong>in</strong>g with l<strong>in</strong>guistic constra<strong>in</strong>ts. Cognition 101.B31–B41.<br />

Port, R., and M. O’Dell. 1986. Neutralization of syllable-f<strong>in</strong>al voic<strong>in</strong>g <strong>in</strong> German. J. of<br />

Phonetics 13.455–471.<br />

Teh, YW, M. Jordan, M. Beal, and D. Blei. 2006. Hierarchical Dirichlet processes. Journal of<br />

<strong>the</strong> American Statistical Association 101.1566–1581.<br />

Tesar, B., J. Alderete, G. Horwood, N. Merchant, K. Nishitani, and A. Pr<strong>in</strong>ce. 2003. Surgery <strong>in</strong><br />

language learn<strong>in</strong>g. In Proceed<strong>in</strong>gs of <strong>the</strong> 22nd West Coast Conference on Formal <strong>L<strong>in</strong>guistics</strong>.<br />

Vallabha, G., J. McClelland, F. Pons, J. Werker, and S. Amano. 2007. Unsupervised learn<strong>in</strong>g<br />

of vowel categories from <strong>in</strong>fant-directed speech. Proceed<strong>in</strong>gs of <strong>the</strong> National Academy of<br />

Sciences of <strong>the</strong> United States of America 104.13273–8.<br />

Werker, J., and R. Tees. 1984. Cross-language speech perception: Evidence for perceptual<br />

reorganization dur<strong>in</strong>g <strong>the</strong> first year of life. Infant Behavior and Development 7.49–63.


Francesca Foppolo (Milano-Bicocca), Marco Marelli (Milano-Bicocca), Luisa Meroni (Utrecht) & Andrea Gualm<strong>in</strong>i (Utrecht)<br />

Pars<strong>in</strong>g Semantic Ambiguity: strategies and commitments<br />

Consider sentence (1) <strong>in</strong> <strong>the</strong> scenario <strong>in</strong> Fig. 1:<br />

(1) The small square is <strong>the</strong> only one that is a policeman<br />

Sentence (1) is ambiguous depend<strong>in</strong>g on what referent is<br />

considered for one: ei<strong>the</strong>r <strong>the</strong> set of all geometrical<br />

figures/th<strong>in</strong>gs <strong>in</strong> <strong>the</strong> context (exhophoric <strong>in</strong>terpretation)<br />

or <strong>the</strong> set of squares (anaphoric <strong>in</strong>terpretation). The way<br />

we <strong>in</strong>terpret one has an effect on <strong>the</strong> truth value of <strong>the</strong><br />

whole sentence: (1) is True if we <strong>in</strong>terpret one<br />

anaphorically (amongst <strong>the</strong> squares, <strong>the</strong> small one is<br />

<strong>in</strong>deed <strong>the</strong> only one to be a policeman) but it is False if<br />

we <strong>in</strong>terpret one exophorically (<strong>in</strong> fact, amongst <strong>the</strong><br />

th<strong>in</strong>gs <strong>in</strong> <strong>the</strong> context, <strong>the</strong> small square is not <strong>the</strong> only one<br />

to be a policeman).<br />

Fig.1<br />

Test<strong>in</strong>g children and adults with sentences similar to (1), Cra<strong>in</strong> et.al. (1994) concluded that<br />

����������� ������� �������� �� ��������� ������������ - ��<br />

exhophoric <strong>in</strong>terpretation) while �������� ������� ��� ��������� �������<br />

choose <strong>the</strong> weaker-anaphoric <strong>in</strong>terpretation). We <strong>in</strong>tend to argue aga<strong>in</strong>st this conclusion on <strong>the</strong> basis<br />

of two experimental studies <strong>in</strong> which we show that adults do not conform to a M<strong>in</strong>imal<br />

Commitment strategy (Exp.1) or to a Maximal Commitment strategy (Exp. 2).<br />

We conclude by propos<strong>in</strong>g a more general pr<strong>in</strong>ciple of process<strong>in</strong>g and ambiguity resolution that<br />

<strong>in</strong>structs <strong>the</strong> parser to consider <strong>the</strong> highest number of available referents .<br />

Our experimental studies. We tested two groups of Italian speak<strong>in</strong>g adults by means of two<br />

studies employ<strong>in</strong>g a Visual <strong>World</strong> Paradigm <strong>in</strong> which subjects were asked to judge a series of<br />

sentences (as True or False) relatively to a scenario while <strong>the</strong>ir eye-movements were recorded. The<br />

rationale beh<strong>in</strong>d <strong>the</strong> two experiments was <strong>the</strong> same: sentences were presented auditorily aga<strong>in</strong>st a<br />

visual scenario similar to <strong>the</strong> one <strong>in</strong> <strong>in</strong> Fig. 1 display<strong>in</strong>g 4 sets of geometrical figures, letters or<br />

digits. We tested 30 were critical statements like (1) that conta<strong>in</strong>ed <strong>the</strong> ambiguous expression è<br />

�����������<br />

(is <strong>the</strong> only one) that needed to be <strong>in</strong>terpreted. The scenario used <strong>in</strong> both experiments<br />

was identical, but <strong>the</strong> two experiments differed <strong>in</strong> <strong>the</strong> presence/absence of negation <strong>in</strong> <strong>the</strong> sentence.<br />

For example, (1) was given <strong>in</strong> Exp.1 aga<strong>in</strong>st <strong>the</strong> scenario <strong>in</strong> Fig. 1 and (5) was given <strong>in</strong> Exp. 2:<br />

(5) The small square is not <strong>the</strong> only one that is a policeman<br />

We also tested 36 unambiguous true and false control sentences �������������<br />

All/Three/None/Some<br />

of <strong>the</strong> N-(Adj) is do<strong>in</strong>g P������ varied with respect to <strong>the</strong> type of exploration strategy required to<br />

evaluate <strong>the</strong>m: (a) controls �������������������<br />

required <strong>the</strong> exploration of <strong>the</strong> whole scenario to be<br />

judged true or false: e.g., to evaluate (2) All <strong>the</strong> yellow numbers are k<strong>in</strong>gs one needs to check all <strong>the</strong><br />

yellow ite<strong>ms</strong> <strong>in</strong> <strong>the</strong> scenario to make sure that all of <strong>the</strong>m are k<strong>in</strong>gs; (b) controls ���������������<br />

required <strong>the</strong> exploration of only one of <strong>the</strong> quadrants <strong>in</strong> order to come up with a decision: e.g., to<br />

evaluate (3) Three crosses are play<strong>in</strong>g football, it suffices to check only <strong>the</strong> quadrant that conta<strong>in</strong>s<br />

crosses; (c) controls �������� ������ an exploration ��������� at random <strong>in</strong> order to f<strong>in</strong>d a<br />

counterexample and tell if <strong>the</strong> sentence was true or false: e.g., to evaluate (4) None of <strong>the</strong> blue<br />

letters are happy ��������������<br />

suffices to f<strong>in</strong>d a letter that is blue and happy and <strong>the</strong>n stop explor<strong>in</strong>g<br />

Crucially, <strong>the</strong> presence of <strong>the</strong> negation <strong>in</strong> <strong>the</strong> critical sentences presented <strong>in</strong> Experiment 2 reverses<br />

<strong>the</strong> entailment pattern between <strong>the</strong> two <strong>in</strong>terpretations, so that <strong>the</strong> anaphoric <strong>in</strong>terpretation becomes<br />

<strong>the</strong> strongest: if ambiguity resolution were based on <strong>the</strong> entailment relationship between <strong>the</strong>


alternative read<strong>in</strong>gs, <strong>the</strong>n an opposite pattern of exploration should be expected <strong>in</strong> <strong>the</strong> two<br />

experiments. In particular, if adults followed a M<strong>in</strong>imal Commitment strategy, as Cra<strong>in</strong> et al.<br />

suggest, <strong>the</strong>y should always select <strong>the</strong> weakest <strong>in</strong>terpretation, <strong>the</strong> one that makes <strong>the</strong> sentence True.<br />

This corresponds to <strong>the</strong> anaphoric <strong>in</strong>terpretation <strong>in</strong> Exp. 1 (it is true that <strong>the</strong> small square is <strong>the</strong> only<br />

square that is a policeman) but to <strong>the</strong> exophoric <strong>in</strong>terpretation <strong>in</strong> Exp. 2 (it is true that <strong>the</strong> small<br />

square is not <strong>the</strong> only th<strong>in</strong>g that is a policeman).<br />

Results. ������� acceptance rate of <strong>the</strong> critical sentences was below 10% <strong>in</strong> Exp. 1 and above 90%<br />

<strong>in</strong> Exp. 2. This shows that <strong>the</strong>y always <strong>in</strong>terpreted one exhophorically, <strong>in</strong>dependently of <strong>the</strong><br />

entailment pattern between <strong>in</strong>terpretations: i.e., <strong>the</strong>y do not conform to a M<strong>in</strong>imal or to a Maximal<br />

Commitment strategy. The pattern of exploration revelaed by <strong>the</strong> eye-movement record<strong>in</strong>gs is<br />

consistent with this hypo<strong>the</strong>sis: <strong>in</strong> both studies, towards <strong>the</strong> end of <strong>the</strong> sentence, subjects followed<br />

an exploration pattern that ���������� ���� ��������� <strong>in</strong> which <strong>the</strong>y had �������<br />

to look for a<br />

counterexample and all quadrants had equal probability to be looked at, and differed significantly<br />

����� ���� ���������� <strong>in</strong> which fixations ����������<br />

were concentrated on a s<strong>in</strong>gle quadrant (Fig. 2,<br />

based on Exp. 1 as an example).<br />

critical trials ����������������� ������������������������������<br />

Fig. 2. The proportion to looks to target/quadrant 1 (arbitrary chosen) are plotted as a function of <strong>the</strong> sentence timecourse:<br />

IP1 (from 0 to red l<strong>in</strong>e) correspond to <strong>the</strong> (Q)N+Adj segment (e.g. The small square); IP2 (<strong>in</strong> between <strong>the</strong> l<strong>in</strong>es)<br />

corresponds to <strong>the</strong> critical segment is <strong>the</strong> only one that <strong>in</strong> <strong>the</strong> critical trials and to a short pause <strong>in</strong> <strong>the</strong> controls; IP3:<br />

(from green l<strong>in</strong>e to <strong>the</strong> end) corresponds to <strong>the</strong> whole VP (e.g. is a policeman).<br />

Instead of focus<strong>in</strong>g on <strong>the</strong> quadrant of <strong>the</strong> squares (anaphoric <strong>in</strong>terpretation), adults moved around<br />

to check for <strong>the</strong> truth/falsity of <strong>the</strong> exhophoric <strong>in</strong>terpretation, as if <strong>the</strong>y were scann<strong>in</strong>g <strong>the</strong> whole<br />

scenario <strong>in</strong> search of a counterexample. This result was corroborated by a series of logistic<br />

regression models <strong>in</strong> which <strong>the</strong> likelihood of look<strong>in</strong>g at <strong>the</strong> target quadrant was modeled as a<br />

function of <strong>the</strong> strategy of exploration (cf. Table 1, based on Exp. 1 as an example): <strong>in</strong> <strong>the</strong> decisionmak<strong>in</strong>g<br />

period (IP4), after <strong>the</strong> end of <strong>the</strong> sentence, <strong>the</strong> proportion of looks to target <strong>in</strong> <strong>the</strong> critical<br />

��������������������������������������������������<br />

=.182).<br />

IP 1 IP 2 IP 3 IP 4<br />

Table 1 Estimate p Estimate p Estimate p Estimate p<br />

critical trials -.37 .0397 .71 .0001 -.34 .0037 -.58 .0001<br />

controls ����� -.97 .0269 -2.08 .0001 -1.08 .0001 -.35 .1820<br />

controls ��������� -.69 .0351 -1.79 .0001 -0.71 .0001 -.47 .0154<br />

controls ���������-1.04<br />

.0005 -.71 .0011 1.72 .0001 1.21 .0001<br />

Conclusions. We found no evidence of a bias that would be based on <strong>the</strong> truth value of <strong>the</strong> read<strong>in</strong>gs<br />

under consideration or <strong>the</strong> entailment relationship between <strong>the</strong>m: <strong>in</strong> both studies, adults <strong>in</strong>terpreted<br />

one exhophorically, always consider<strong>in</strong>g <strong>the</strong> whole scenario. This f<strong>in</strong>d<strong>in</strong>g, <strong>in</strong> conjunction with<br />

�������������������������������������������������������<br />

<strong>the</strong> need to postulate a preference for<br />

<strong>the</strong> strong read<strong>in</strong>g of <strong>the</strong> same sentences <strong>in</strong> children, <strong>in</strong>vites us to consider whe<strong>the</strong>r such truth<br />

conditional properties play any role <strong>in</strong> ambiguity resolution. Instead, our f<strong>in</strong>d<strong>in</strong>gs suggest that<br />

adults follow a more general pars<strong>in</strong>g strategy <strong>in</strong> solv<strong>in</strong>g semantic ambiguity: when a (visual/rich)<br />

context is given, <strong>the</strong> parser maximally exploits it.


Michael Freedman (Yale)<br />

Contextual Disambiguation of Have-Sentences<br />

Background: Have-sentences can encode relations that express k<strong>in</strong>ship (1a), part-whole (1b),<br />

possessor-possessee (1c), and conta<strong>in</strong>er-conta<strong>in</strong>ee (1d) relationships. They also seem to be able<br />

to encode completely context dependent relations (1e).<br />

(1) (a) KINSHIP: John has a sister. (Partee, 1999)<br />

(b) PART-WHOLE/INALIENABLE POSSESSION: John has a hand.<br />

(c) TRUE POSSESSION/CONTROL: John has a pen.<br />

(d) CONTAINER: That glass has w<strong>in</strong>e. (Gutierrez-Rexach, 2006)<br />

(e) ASSIGNMENT (BY CONTEXT): Eliza has <strong>the</strong> mirror (to wash). (Tham, 2004)<br />

Problem: A subset of have-sentences are ambiguous <strong>in</strong> what relation <strong>the</strong>y express (table 1). But<br />

it is clear that one read<strong>in</strong>g is available out-of-<strong>the</strong>-blue, which I will refer to as <strong>in</strong>herently salient.<br />

Because <strong>the</strong> o<strong>the</strong>r read<strong>in</strong>g(s) need context to become salient; I will call <strong>the</strong>se contextually licensed.<br />

The questions that I aim to address are (1) what mechanism expla<strong>in</strong>s why certa<strong>in</strong> read<strong>in</strong>gs are <strong>in</strong>herently<br />

salient and (2) what is <strong>the</strong> <strong>in</strong>teraction with context that expla<strong>in</strong>s when <strong>the</strong> o<strong>the</strong>r (contextually<br />

licensed) read<strong>in</strong>gs become available.<br />

Sentence <strong>in</strong>herently salient contextually licensed<br />

John has a sister k<strong>in</strong>ship control, assignment(?)<br />

John has a d<strong>in</strong>osaur tail part-whole/control<br />

Eliza has a mirror. control assignment<br />

The dumpster has a hand part-whole conta<strong>in</strong>er<br />

Table 1: Have-sentences and <strong>the</strong>ir possible read<strong>in</strong>gs.<br />

In this paper, I argue that a pragmatic filter can disambiguate <strong>the</strong> different read<strong>in</strong>gs of havesentences<br />

and that <strong>the</strong> amount of ambiguity present comes not only from <strong>the</strong> underspecification<br />

of <strong>the</strong> relation encoded by have but by <strong>the</strong> context dependence of some verbal complements.<br />

Analysis: Assume that have encodes a relation between <strong>in</strong>dividuals (cf. Beavers et al. (2008),<br />

Partee (1999)). Relational nouns <strong>the</strong>n need to be treated as properties (type 〈e, t〉) <strong>in</strong> order for <strong>the</strong><br />

ord<strong>in</strong>ary derivation of sentences with transitive verbs to take place. I treat relational nouns as this<br />

type (〈e, t〉) but with a free contextual argument, follow<strong>in</strong>g work on o<strong>the</strong>r relational predicates like<br />

local and w<strong>in</strong> (Condoravdi & Gawron (1996)). Concretely, have is treated as an underspecified<br />

relation fixed by an assignment function (as <strong>in</strong> (2d)) and a relational noun is represented as a relation<br />

that has a free variable as an argument (as <strong>in</strong> (2a)). A full compositional treatment of <strong>the</strong> sentence<br />

John has a sister is provided <strong>in</strong> (2):<br />

(2) (a) [sister]= λx.sister(x,y), where <strong>the</strong> value of y is fixed by an assignment function.<br />

(b) [a]= λPλQ∃z[P(z) ∧ Q(z)]<br />

(c) [a sister]= λQ∃z[sister(c,z) ∧ Q(z) ]<br />

(d) [have](type-raised) = λPλw[P(λv[π(w,v)])]<br />

where π is an underspecified relation fixed by an assignment function.<br />

(e) [have a sister]= λPλw[P(λv[π(w,v)])](λQ∃z[sister(c,z) ∧ Q(z) ])<br />

= λw[∃z[sister(c,z) ∧ π(w,z)]]<br />

(f) [John]= j<br />

(g) [John has a sister]= λw[∃z[sister(c,z) ∧ π(w,z)]](j)<br />

= ∃z[sister(c,z) ∧ π(j,z)]<br />

A pragmatic mechanism is necessary because simply posit<strong>in</strong>g a context variable makes no<br />

predictions on what relation <strong>the</strong> variable will express. I will assume a question-based model of<br />

discourse (QUD) (Roberts, 2006) <strong>in</strong> order to resolve <strong>the</strong> variables <strong>in</strong> have and <strong>in</strong> relational nouns;


<strong>in</strong> addition, I will have <strong>the</strong> discourse model track what elements (<strong>in</strong>dividuals, relations) are salient<br />

(similar to Grosz & Sidner, 1986). Salient <strong>in</strong>dividuals are <strong>the</strong> set of <strong>in</strong>dividuals and salient relations<br />

are <strong>the</strong> set of relations that have been <strong>in</strong>troduced <strong>in</strong>to <strong>the</strong> discourse ei<strong>the</strong>r explicitly or through<br />

<strong>the</strong> use of discourse questions. The QUD is <strong>the</strong> current question that needs to be addressed <strong>in</strong><br />

some way as <strong>the</strong> next “move” <strong>in</strong> <strong>the</strong> discourse. In addition, I assume that <strong>the</strong>re is a default relation<br />

that have expresses when <strong>the</strong>re is no context; CONTROL is <strong>the</strong> default when <strong>the</strong> subject is animate.<br />

This relation is a reasonable default because it is closest to <strong>the</strong> core notion of possession. I will go<br />

through three cases to illustrate how <strong>the</strong> system works:<br />

Case 1 (In context): In a standard discourse <strong>the</strong> resolution of have and relational nouns is determ<strong>in</strong>ed<br />

jo<strong>in</strong>tly by what is salient and what <strong>the</strong> current question is. The variable associated with have<br />

resolves to a salient relation, if <strong>the</strong>re is a relational noun, <strong>the</strong> variable associated with it resolves<br />

to a salient <strong>in</strong>dividual, and <strong>the</strong> sentence as a whole addresses <strong>the</strong> current question under discussion.<br />

In (3), <strong>the</strong> QUD is “where are <strong>the</strong> victim’s ar<strong>ms</strong>”, <strong>the</strong> salient relations are <strong>the</strong> hand-relation (from<br />

<strong>the</strong> utterance) and <strong>the</strong> conta<strong>in</strong>er relation (based on <strong>the</strong> use of <strong>the</strong> locative <strong>in</strong> <strong>the</strong> QUD), and <strong>the</strong><br />

only salient <strong>in</strong>dividual is <strong>the</strong> victim. π is resolved to <strong>the</strong> conta<strong>in</strong>er-relation because <strong>the</strong> relation is<br />

salient and because with that relation <strong>the</strong> QUD is partially answered (<strong>the</strong> subquestion “where is<br />

one of <strong>the</strong> victim’s hands” is answered); c resolves to <strong>the</strong> victim because <strong>the</strong> victim is salient and<br />

because it is coherent as part of a response that partially answers <strong>the</strong> QUD. This example shows<br />

how a CONTAINER read<strong>in</strong>g for have can be contexutally licensed.<br />

(3) Scenario: Two police detectives John and Mary are at a crime scene<br />

<strong>in</strong> an alley. A murder victim has been dismembered and <strong>the</strong> detectives are look<strong>in</strong>g for <strong>the</strong><br />

victim’s rema<strong>in</strong>s. John asks “where are his ar<strong>ms</strong>?” Mary replies “that dumpster has a hand”.<br />

(4)<br />

QUD: Where are <strong>the</strong> victim’s ar<strong>ms</strong>?<br />

Salient sets: hand(x,y), conta<strong>in</strong>s(x,y)<br />

Salient <strong>in</strong>dividuals: <strong>the</strong> victim<br />

QUD relevant = conta<strong>in</strong>s<br />

⇒ c = <strong>the</strong> victim, π = conta<strong>in</strong>s(x,y)<br />

Case 2 (Out-of-<strong>the</strong>-blue): In out-of-<strong>the</strong>-blue scenarios, <strong>the</strong>re is no QUD and no salient <strong>in</strong>dividuals<br />

or relations prior to <strong>the</strong> first utterance of <strong>the</strong> discourse. In a sentence without a relational noun,<br />

<strong>the</strong> mean<strong>in</strong>g of have will assume a default value of CONTROL (for animates) or CONTAINER (for<br />

<strong>in</strong>animates). For <strong>in</strong>stance, if person A says to person B out-of-<strong>the</strong>-blue “I have a pen” <strong>the</strong>re is<br />

no QUD, and no salient relations. Person B can give mean<strong>in</strong>g to person B’s utterance by giv<strong>in</strong>g<br />

a value to π which comes from what rema<strong>in</strong>s of <strong>the</strong> “bleached” default lexical mean<strong>in</strong>g of have<br />

(CONTROL). This example shows why have expresses a CONTROL relation when <strong>the</strong>re is no prior<br />

discourse context and <strong>the</strong> subject is animate.<br />

Case 3 (Inherently salient w/ relational noun complement): If <strong>the</strong>re is a relational noun <strong>in</strong> object<br />

position, <strong>the</strong>n <strong>the</strong> relation associated with <strong>the</strong> relational noun gets added to <strong>the</strong> list of salient<br />

relations and becomes a suitable relation for have to express and so π is resolved to that relation.<br />

The resolution of <strong>the</strong> variable associated with <strong>the</strong> relational noun will be to an <strong>in</strong>dividual that is<br />

salient <strong>in</strong> that utterance (i.e. <strong>the</strong> subject). So, if out-of-<strong>the</strong>-blue Speaker A says to Speaker B “I have<br />

a sister”, <strong>the</strong> mean<strong>in</strong>g of have resolves to <strong>the</strong> sister relation. The c variable resolves to <strong>the</strong> subject<br />

because<strong>the</strong> subject is <strong>the</strong> only salient <strong>in</strong>dividual. This example shows why have takes <strong>the</strong> relation<br />

of its complement if <strong>the</strong>re is no discourse context and <strong>the</strong> complement is a relational noun.<br />

Selected References: Beavers, J., S. Wechsler, and E. Ponvert. (2008) Possession of a controlled substantive: Light<br />

have and verbs of possession. Condoravdi, C. & Gawron, J.M. (1996). The context-dependency of implicit arguments.<br />

Grosz, B. & Sidner, L. (1986) Attentions, Inentions, and <strong>the</strong> Structure of Discourse. Gutierrez-Rexach, J. (2006).<br />

Beyond <strong>the</strong> (<strong>in</strong>)def<strong>in</strong>iteness restriction: A unified semantics for have. Partee, B. (1999). Weak NP’s <strong>in</strong> HAVE Sentences.<br />

Roberts, C. (1996) Information Structure <strong>in</strong> Discourse. Tham, S. (2004) The Def<strong>in</strong>iteness Effect <strong>in</strong> English Have<br />

Sentences.


Long­distance
Anaphora
<strong>in</strong>
Mandar<strong>in</strong>,
<strong>the</strong>
PCC,
and
Cyclic
Agree
<br />

Mandar<strong>in</strong>
conta<strong>in</strong>s
a
reflexive
form,
ziji,
that
can
function
as
a
local
reflexive.
However,
<br />

ziji
can
also
take
an
antecedent
outside
its
local
doma<strong>in</strong>.
As
shown
<strong>in</strong>
(1a‐e),
ziji
can
be
<br />

bound
 by
 <strong>the</strong>
 closest
 subject
 or
 <strong>the</strong>
 matrix
 subject
 (Cole,
 et
 al.,
 2006
 and
 references
<br />

<strong>the</strong>re<strong>in</strong>).
Such
long‐distance
b<strong>in</strong>d<strong>in</strong>g
is
constra<strong>in</strong>ed
by
a
block<strong>in</strong>g
effect,
such
that
subject
<br />

DPs
that
differ
<strong>in</strong>
person
features
appear
to
block
higher
subject
DPs
from
b<strong>in</strong>d<strong>in</strong>g
ziji.
<br />

However,
a
difference
<strong>in</strong>
person
features
is
not
a
sufficient
condition
for
<strong>the</strong>
block<strong>in</strong>g
<br />

effect.
 Instead,
 we
 see
 that
 subjects
 <strong>in</strong>
 a
 1>3
 configuration
 allow
 long‐distance
<br />

antecedents
but
subjects
<strong>in</strong>
a
3>1
configuration
block
long‐distance
antecedents.
Thus,
<br />

<strong>the</strong>
 <strong>in</strong>terference
 pattern
 that
 emerges
 <strong>in</strong>
 <strong>the</strong>
 block<strong>in</strong>g
 effect
 is
 not
 symmetrical
 (see
<br />

examples
(1a‐g)).

<br />


<br />

1)
<br />

a)

 Zhangsani
zhidao
Lisij
bu


xihuan
zijii/j
 e)
 Nii





zhidao
Lisij

bu


xihuan
zijii/j
<br />


 Zhagshan

know


Lisi

not

like






self
 
 You



know


Lisi


not

like






self
<br />

‘Zhangsan
 knew
 that
 Lisi
 did
 not
 like
 ‘You
 knew
 that
 Lisi
 did
 not
 like


<br />

me/hi<strong>ms</strong>elf’
<br />

you/hi<strong>ms</strong>elf
<br />

b)
 Woi
zhidao
Lisij
bu


xihuan
zijii/j
 f)
 Lisii



zhidao

Woj



bu


xihuan
ziji*i/j
<br />


 I






know


Lisi

not

like






self
<br />


 Lisi




know



I









not

like






self
<br />

‘I
 knew
 that
 Lisi
 did
 not
 like
 ‘Lisi
 knew
 that
 I
 did
 not
 like
<br />

me/hi<strong>ms</strong>elf
<br />

*him/myself’
<br />

c)
 Woi





zhidao
Nij



bu


xihuan
zijii/j
 g)
 Lisii





zhidao
Nij



bu


xihuan
ziji*i/j
<br />


 I











know


you

not

like






self
 
 Lisi






know


you

not

like






self
<br />

‘I
 knew
 that
 you
 did
 not
 like
 ‘Lisi
 knew
 that
 you
 did
 not
 like
<br />

me/yourself’
<br />

*him/yourself’
<br />

d)
 Nii




zhidao
Woj
bu


xihuan
zijii/j
 
<br />


 You


know


I






not

like






self
<br />

‘You
 knew
 that
 I
 did
 not
 like
<br />

you/myself’
<br />


<br />

Strik<strong>in</strong>gly,
this
<strong>in</strong>terference
pattern
replicates
a
pattern
of
<strong>in</strong>tervention
that
is
known
as
<br />

<strong>the</strong>
 person­case
 constra<strong>in</strong>t.
 Anagnostopoulou
 characterizes
 <strong>the</strong>
 PCC
 such
 that
 “[i]n
 a
<br />

comb<strong>in</strong>ation
of
a
weak
direct
object
and
an
<strong>in</strong>direct
object
[clitic,
agreement
marker,
or
<br />

weak
pronoun],
if
<strong>the</strong>re
is
a
third
person
it
has
to
be
<strong>the</strong>
direct
object”
(2005,
p.
203).
<br />

Compare
Tables
1
and
2.
<br />


<br />

INDIRECT

 DIRECT

 PCC
 
 HIGHEST
SUBJECT
 LOWEST
SUBJECT
 ZIJI
<br />

OBJECT
<br />

1
<br />

OBJECT
<br />

3
 �
<br />


<br />

PERSON
<br />

1
<br />

PERSON
<br />

3
<br />

LDR
<br />

�
<br />

1
<br />

2
<br />

2
<br />

1
<br />

�
<br />

�
<br />


 1
<br />

2
<br />

2
<br />

1
<br />

�
<br />

�
<br />

2
 3
 �
 
 2
 3
 �
<br />

3
 1
 �
<br />

3
 1
 �
<br />

3
 2
 �
 
 3
 2
 �
<br />

Table
1
–
Interference
pattern
for
PCC
 
 







Table
2
–
Interference
pattern
for
ziji
<br />

Follow<strong>in</strong>g
an
analysis
<strong>in</strong>
Cole,
et
al.
(2006),
I
propose
that
long‐distance
b<strong>in</strong>d<strong>in</strong>g
of
ziji
is
<br />

<strong>the</strong>
result
of
covert
head
movement
and
that
ziji
is
sensitive
to
person
hierarchies
as
<br />

conceived

<strong>in</strong>
Bejar
and
Rezac
(2009).
These
two
facts
restrict
<strong>the</strong>
configurations
that
<br />

license
<strong>the</strong>
long‐distance
b<strong>in</strong>d<strong>in</strong>g
of
ziji.
<br />


<br />

Ia<strong>in</strong> Gibl<strong>in</strong> (MIT)


2)
 [IP
Zhangsan
[I
zijii
]
[VP
yiwei
[IP
Lisi
[I
t’i
]
[VP
pip<strong>in</strong>g‐le
ti
]
]
]
]
<br />


 




Zhangsan



self









th<strong>in</strong>k






Lisi

















criticize‐Perf
<br />

In
 (2)
 above,
 each
 I0 
 agrees
 with
 its
 specifier
 and
 ziji
 must
 <strong>the</strong>refore
 agree
 with
 <strong>the</strong>
<br />

subject
 upon
 adjunction
 to
 I0 .
 
 However,
 if
 ziji’s
 ϕ‐features
 are
 structured
 so
 <strong>the</strong>y
 are
<br />

sensitive
to
<strong>the</strong>
person
hierarchies
of
Bejar
and
Rezac
(2009,
see
Table
3
below)
we
can
<br />

generate
<strong>the</strong>
block<strong>in</strong>g
effect
and
its
asymmetrical
structure.
<br />


<br />

Table 2<br />

Person specifications<br />

C Y C L I C A G R E E 43<br />

A: Person specifications B: Shorthand 1�2�3 C: Shorthand 2�1�3<br />

3rd 2nd 1st 3rd 2nd 1st 3rd 2nd 1st<br />


<br />


<br />

Table
3
–
Bejar
and
Rezac
person
hierarchy
/
articulated
probe
<br />

Let
us
assume
that
ziji
is
a
partially
articulated
probe
that
searches
for
[participant]
and
<br />

<strong>the</strong>refore
seeks
to
check
[3][2].
In
1/2
>
3
configurations
movement
of
ziji
to
<strong>the</strong>
lowest
<br />

I0 
 checks
 <strong>the</strong>
 [3]
 segment
 of
 <strong>the</strong>
 anaphor
 <strong>the</strong>refore
 allow<strong>in</strong>g
 ziji
 to
 be
 valued
 by
 I0 .
<br />

However,
this
leaves
an
unchecked
[2].
Thus,
<strong>the</strong>
unchecked
[2]
on
<strong>the</strong>
anaphor
licenses
<br />

fur<strong>the</strong>r
movement
of
ziji
to
<strong>the</strong>
higher
I0 
<strong>in</strong>
search
of
a
[+participant]
argument.
If
<strong>the</strong>
<br />

higher
I0 
has
obta<strong>in</strong>ed
[+participant]
features
through
agreement
with
<strong>the</strong>
subject,
<strong>the</strong>n
<br />

ziji
can
adjo<strong>in</strong>
to
<strong>the</strong>
higher
I0 
and
valuation
can
take
place.
Thus,
a
higher
argument
can
<br />

be
 [+participant]
 and
 a
 lower
 argument
 can
 be
 [‐participant].
 However,
 <strong>the</strong>
 converse
<br />

does
 not
 hold.
 If
 ziji
 first
 adjo<strong>in</strong>s
 to
 an
 I0 
 that
 is
 [+participant]
 both
 of
 ziji’s
 person
<br />

features
 ([3]
 and
 [2])
 will
 be
 checked
 leav<strong>in</strong>g
 no
 residue
 that
 would
 license
 fur<strong>the</strong>r
<br />

movement
(unless
<strong>the</strong>
higher
I0 
agrees
for
[+participant]).
Thus,
a
higher
[3]
argument
<br />

is
<strong>in</strong>accessible
because
<strong>the</strong>
anaphor
has
been
marked
as
[+participant]
and
<strong>the</strong>
higher
I0 [�] [�] [�] [3] [3] [3] [3] [3] [3]<br />

[participant] [participant] [2] [2] [1] [1]<br />

[speaker] [1] [2]<br />

(structure) [F] entails a feature (structure) [F′] if and only if [F′] is a subset (<strong>in</strong>clud<strong>in</strong>g identity)<br />

of <strong>the</strong> least set conta<strong>in</strong><strong>in</strong>g [F]. For example, be<strong>in</strong>g specified as [speaker] entails be<strong>in</strong>g specified as<br />

[participant] and as [�].<br />


<br />

is
[‐participant].

<br />

The
proposed
analysis
has
some
welcome
consequences.
Firstly,
it
expla<strong>in</strong>s
<strong>the</strong>
block<strong>in</strong>g
<br />

effect.
Secondly,
it
provides
a
pr<strong>in</strong>cipled
explanation
of
<strong>the</strong>
asymmetry
<strong>in</strong>
<strong>the</strong>
block<strong>in</strong>g
<br />

effect
 that
 is
 not
 expla<strong>in</strong>ed
 by
 <strong>the</strong>
 head
 movement
 approach
 alone.
 Thirdly,
 it
<br />

demonstrates
 that
 AGREE
 based
 approaches
 to
 b<strong>in</strong>d<strong>in</strong>g
 offer
 pr<strong>in</strong>cipled
 empirical
<br />

coverage
 and
 can
 help
 expla<strong>in</strong>
 some
 of
 <strong>the</strong>
 more
 recalcitrant
 phenomena
 of
 b<strong>in</strong>d<strong>in</strong>g
<br />

<strong>the</strong>ory.
<br />

References
<br />

Anagnostopoulou,
E.
(2005).
Strong
and
weak
person
restrictions:
A
feature
check<strong>in</strong>g
analysis.
<br />

In
 L.
 Heggie
 and
 F.
 Ordonez
 (Eds.),
 Clitics
 and
 Affixation
 (pp.
 199‐235).
 A<strong>ms</strong>terdam:
 John
<br />

Benjam<strong>in</strong>s.
<br />

5 These entailments translate <strong>in</strong>to degrees of privative feature specification<br />

through a heuristic of logical underspecification, where �-values are differentiated only by <strong>the</strong><br />

presence versus absence of features, as <strong>in</strong> table 2 (A). This requires specify<strong>in</strong>g default <strong>in</strong>terpretations<br />

for underspecified representations: for example, [�] is common to all persons, but a bare<br />

[�] feature is <strong>in</strong>terpreted as 3rd person.<br />

(6) Entailment: [speaker] N [participant] N [�]<br />

We adopt <strong>the</strong>se feature specifications, but for convenience we employ a shorthand from here on:<br />

we write [�] as [3], [participant] as [2], and [speaker] as [1], and we refer to each of [3], [2],<br />

and [1] as a segment, mean<strong>in</strong>g ‘feature <strong>in</strong> a hierarchical feature structure’. The representations<br />

correspond<strong>in</strong>g to table 2 (A) <strong>in</strong> this abbreviated system are given <strong>in</strong> table 2 (B). The notation is<br />

convenient because <strong>the</strong> <strong>in</strong>terpretation can be transparently read by <strong>in</strong>spect<strong>in</strong>g <strong>the</strong> bottommost<br />

segment <strong>in</strong> <strong>the</strong> feature bundle. It is important, however, that <strong>the</strong>se segments not be read as person<br />

categories. For example, [1] <strong>in</strong> table 2 (B) does not refer to <strong>the</strong> category of 1st person; ra<strong>the</strong>r, it<br />

refers to [speaker]. It is only <strong>the</strong> feature structure as a whole that corresponds to a traditional<br />

category like 1st person.<br />

The system assumes limited variation <strong>in</strong> <strong>the</strong> selection of features (see Harley and Ritter<br />

2002). Of relevance below will be that some languages differentiate 1st and 2nd persons by<br />

specify<strong>in</strong>g <strong>the</strong> latter as [addressee] ra<strong>the</strong>r than by specify<strong>in</strong>g <strong>the</strong> former as [speaker], and by<br />

contrast <strong>in</strong>terpret<strong>in</strong>g a bare [participant] as 1st person. This is shown <strong>in</strong> table 2 (C).<br />

In light of this feature-<strong>the</strong>oretic approach to �-specification, match<strong>in</strong>g requirements can be<br />

relativized to specific �-structures by manipulat<strong>in</strong>g <strong>the</strong> specifications of a probe: <strong>the</strong> more highly<br />

articulated a probe is, <strong>the</strong> more highly specified a DP must be to match all of a probe’s features<br />

(cf. Béjar 2003). (7)–(9) show this for <strong>the</strong> three possible articulations of <strong>the</strong> probe: a flat probe<br />

that is just [u�] ([u3] <strong>in</strong> our notation) <strong>in</strong> (7), a partially articulated probe <strong>in</strong> (8), and a fully<br />

articulated probe <strong>in</strong> (9). For each probe, a DP as highly specified as (or more highly specified<br />

Bejar,
S.
and
Rezac,
M.
(2003).
Person
licens<strong>in</strong>g
and
<strong>the</strong>
derivation
of
PCC
effects.
In
Y.
Roberge
<br />

and
 A.
 T.
 Perez‐Leroux
 (Eds.),
 Romance
 l<strong>in</strong>guistics:
 Theory
 and
 acquisition
 (pp.
 49‐62).
<br />

Anderson’s (1992) [�me, �you].<br />

A<strong>ms</strong>terdam:
John
Benjam<strong>in</strong>s.
<br />

5 The entailment relation between feature segments is <strong>in</strong>tegral to our formalization of <strong>the</strong> operations Match and<br />

Value, as we will show directly. This excludes feature syste<strong>ms</strong> that do not encode <strong>in</strong>tr<strong>in</strong>sic entailment relations, like<br />

Cole,
P.,
Hermon,
G.,
and
Huang,
C.‐T.
J.
(2006).
Long‐Distance
B<strong>in</strong>d<strong>in</strong>g
<strong>in</strong>
Asian
Languages,
The
<br />

Blackwell
Companion
to
Syntax.
Vol.
3,
Ch.
39.
London:
Blackwell.
<br />


Yael Greenberg & Keren Khrizman (Bar Ilan)<br />

Bixlal: A general streng<strong>the</strong>n<strong>in</strong>g operator <strong>in</strong> Hebrew<br />

Basic observations: bixlal is <strong>the</strong> Hebrew translation of <strong>the</strong> NPI at all, as <strong>in</strong> (1). However,<br />

Migron (2003) shows that, unlike at all, bixlal can also appear <strong>in</strong> positive (and o<strong>the</strong>r UE)<br />

constructions (as <strong>in</strong> (2),(4),(5)). In addition, we get different <strong>in</strong>terpretations of bixlal depend<strong>in</strong>g<br />

on whe<strong>the</strong>r it is stressed ((1)-(2)) or not ((3)-(5)):<br />

(1) dani lo gavoha [bixlal]F (“Danny is not tall at all‟)<br />

(2) A: Yosi is tall! B: ve-dani [bixlal]F gavoha (“And Danny is clearly tall / even taller”)<br />

(3)A: Are <strong>the</strong>y married? (B: hem bixlal lo [makirim]F!) (“They don‟t even [know]F each o<strong>the</strong>r”)<br />

(4) A: Do <strong>the</strong>y know each o<strong>the</strong>r? B: hem bixlal [nesuim]F! (“They are even [married]F!”)<br />

(5) A:R<strong>in</strong>a is from Jerusalem B: lo,hi bixlal mi-[Tel Aviv]Contrastive F (“No, she is actually from<br />

[Tel Aviv]Contrastive F")<br />

Analysis Follow<strong>in</strong>g a prelim<strong>in</strong>ary suggestion <strong>in</strong> Migron (2003), we suggest that bixlal p<br />

<strong>in</strong>dicates that p is stronger than its contextually salient alternatives, where its particular<br />

effects depend on <strong>the</strong> <strong>in</strong>teraction between (a) <strong>the</strong> nature of <strong>the</strong> streng<strong>the</strong>n<strong>in</strong>g relation (b) <strong>the</strong><br />

nature and placement of focus and (c) <strong>the</strong> polarity of <strong>the</strong> sentence.<br />

We start with (1)-(4), <strong>in</strong> which p is stronger than q iff p entails q and q does not entail p.<br />

When focus is on <strong>the</strong> predicate (as <strong>in</strong> (3)-(4)) we get standard “Roothian” alternatives to p. With<br />

positive and negative cases (as <strong>in</strong> (4), (3) respectively) <strong>the</strong> context has to conta<strong>in</strong> a salient<br />

alternative which is weaker or stronger than p, respectively. In contrast, when bixlal itself is<br />

focused <strong>the</strong> alternatives are different <strong>in</strong>terpretations of p. For example, <strong>in</strong> <strong>the</strong> negative (or DE)<br />

(1) p is “Danny is not tall under <strong>the</strong> most liberal precision standard”, and <strong>the</strong> alternatives are of<br />

<strong>the</strong> form “Danny is not tall under a stricter precision standard” (as <strong>in</strong> Krifka 1995)).<br />

Unlike Migron‟s view, we show that bixlal is also felicitous with non-entailment,<br />

evaluative scales, where p is stronger than q s<strong>in</strong>ce it is considered more significant than q, as <strong>in</strong><br />

(6) (cf. Beaver & Clark 2008 on only and Amaral & del Prete (2010) on almost):<br />

(6) moshe zaxa be dekel ha zahav ve yosi bixlal zaxa [ba oskar]F<br />

„Moshe won <strong>the</strong> Golden Palm and Yosi even won <strong>the</strong> [Oscar]F.‟<br />

We discuss two potential mechanis<strong>ms</strong> <strong>in</strong> <strong>the</strong> case of stressed bixlal. (a) <strong>the</strong> stress on <strong>the</strong><br />

operator <strong>in</strong>duces an alternative operator (cf. Beck (2006), Fery (to appear)). This direction is<br />

supported by <strong>the</strong> existence of sentences like (7):<br />

(7) efSar liknot Sam dagim [bixlal]F ve-dgey yam [bifratF]<br />

"You can buy <strong>the</strong>re fish <strong>in</strong> general, and sea fish <strong>in</strong> particular"<br />

But s<strong>in</strong>ce (a) cannot account for <strong>the</strong> range of read<strong>in</strong>gs expressed by bixlal we suggest (b) The<br />

stress <strong>the</strong> operator <strong>in</strong>dicates no stress on any o<strong>the</strong>r element <strong>in</strong> p (cf. Egg & Zimmerman 2011)<br />

and hence no “Roothian” alternatives. But s<strong>in</strong>ce <strong>the</strong> streng<strong>the</strong>n<strong>in</strong>g semantics of bixlal makes<br />

reference to alternatives, we end up with different alternative versions of p. We support (b) with<br />

e.g. <strong>the</strong> behavior of <strong>the</strong> Hebrew stressed stam (Orenste<strong>in</strong> & Greenberg (2010)).<br />

Turn<strong>in</strong>g now to (5), which <strong>in</strong>volves contrastive focus, and <strong>in</strong> which p is nei<strong>the</strong>r stronger<br />

nor weaker than q, we show that <strong>the</strong> presence of bixlal <strong>in</strong> such cases <strong>in</strong>dicates „significant<br />

contrast” between p and its q. E.g. bixlal <strong>in</strong> (5) <strong>in</strong>dicates that be<strong>in</strong>g from Tel Aviv is<br />

significantly <strong>in</strong>compatible with be<strong>in</strong>g from Jerusalem. Thus, for example, <strong>in</strong> a context where this<br />

<strong>in</strong>compatibility is not significant (e.g. <strong>in</strong> an <strong>in</strong>ternational forum where each participant has to say<br />

where he or she is from) <strong>the</strong> use of bixlal <strong>in</strong> (5) is odd.


To capture this 'stronger contrast' use we rely on Umbach‟s (2007) idea that contrast as<br />

<strong>in</strong>volves both similarity and dissimilarity, and make use of Morzycki‟s (2011) alternative-based<br />

model for captur<strong>in</strong>g degrees of (im)precision based on similarity, (orig<strong>in</strong>ally developed to<br />

analyze metal<strong>in</strong>guistic comparatives). In this system, <strong>the</strong> <strong>in</strong>terpretation function is relativized to<br />

contexts and degrees of precision (which are based on degrees of similarity). E.g. [[tel aviv]] d,C<br />

= {fl: f �d,C tel aviv} (l is location}, i.e. <strong>the</strong> set of locations sufficiently similar (similar to a<br />

degree d) <strong>in</strong> C to Tel Aviv, Where d is a real number <strong>in</strong> <strong>the</strong> <strong>in</strong>terval [0,1].<br />

Us<strong>in</strong>g this system, we propose that us<strong>in</strong>g standard contrastive focus on an element <strong>in</strong> p <strong>in</strong><br />

<strong>the</strong> presence of an alternative q <strong>in</strong> a context c, <strong>the</strong> speaker <strong>in</strong>dicates that <strong>the</strong> degree of precision<br />

w.r.t p is true is not necessarily 1 (i.e. <strong>the</strong> speaker can consider "He is from [Aviv]Contrastive F even<br />

if Danny lives close to Tel Aviv), but that this degree of precision is higher than <strong>the</strong> degree w.r.t.<br />

q is true (e.g. <strong>in</strong> (5) „Tel Aviv‟ is taken as not „similar enough‟ to „Jerusalem‟), thus mak<strong>in</strong>g q<br />

false <strong>in</strong> c. Crucially, <strong>the</strong> addition of bixlal makes <strong>the</strong> contrast significant. This is captured by<br />

tak<strong>in</strong>g <strong>the</strong> degree of precision w.r.t p is true to be much higher than <strong>the</strong> degree of precision w.r.t.<br />

q is true (so p is considered highly dissimilar from q), Thus, bixlal leads aga<strong>in</strong> to streng<strong>the</strong>n<strong>in</strong>g,<br />

but not of p (relative to q), but of <strong>the</strong> contrast created by us<strong>in</strong>g p+contrastive focus <strong>in</strong> C. It may<br />

thus be seen here as a streng<strong>the</strong>n<strong>in</strong>g operator on a speech act (cf. Cohen & Krifka (to appear)).<br />

We compare our analysis of bixlal to Anderssen‟s (2006) analysis of <strong>the</strong> similar German<br />

uberhaupt <strong>in</strong> ter<strong>ms</strong> of widen<strong>in</strong>g, where, e.g. <strong>the</strong> German counterpart of (8) <strong>in</strong>volves widen<strong>in</strong>g of<br />

<strong>the</strong> comparative class needed for fix<strong>in</strong>g <strong>the</strong> standard associated with <strong>the</strong> adjective:<br />

(8) A: Danny is tall for his age B: hu [bixlal]F gavoha („He is tall <strong>in</strong> general‟).<br />

We show, however that bixlal is felicitous also <strong>in</strong> cases where no doma<strong>in</strong> is relevant (e.g. (3-5)),<br />

as well as <strong>in</strong> cases where <strong>the</strong> doma<strong>in</strong> doesn‟t change (9), or even gets narrower (10):<br />

(9) biSvil yalda bat 10, r<strong>in</strong>a lo gvoha [bixlal]F (‘For a ten years old girl, R<strong>in</strong>a is not tall at all’)<br />

(10) ba-balSanut hu tov, ve-be-semantika hu [bixlal]F tov!(‘He is good at l<strong>in</strong>guistics, and he is<br />

even better / very good at semantics’)<br />

This, as well as similar data with multidimensional adjectives (Sassoon 2010), <strong>in</strong>dicates<br />

that widen<strong>in</strong>g is just one of <strong>the</strong> ways <strong>in</strong> which streng<strong>the</strong>n<strong>in</strong>g can be satisfied with bixlal.<br />

Moreover, unlike e.g. any (under Kadmon & Landman 1993, Chierchia (2006), streng<strong>the</strong>n<strong>in</strong>g<br />

see<strong>ms</strong> lexically associated with bixlal and not triggered by <strong>in</strong>formaitvity-based considerations.<br />

This conclusion is fur<strong>the</strong>r supported by <strong>the</strong> compatibility of bixlal with nonentailment scales,<br />

which do not <strong>in</strong>volve <strong>in</strong>formativity. Thus, bixlal can be seen as a lexicalization of an emphatic<br />

operator (perhaps similar to Krifka‟s EMPHATIC ASSERT operator).<br />

F<strong>in</strong>ally, given Beaver & Clark's (2008) (B&C), model nonstressed bixlal is predicted to<br />

conventionally associate with focus, s<strong>in</strong>ce it can be taken to impose an order<strong>in</strong>g on <strong>the</strong> set of<br />

alternatives <strong>in</strong> <strong>the</strong> CQ and to <strong>in</strong>dicate that p is stronger than a salient alternative <strong>in</strong> this set.<br />

Empirically, however, unlike rak („only‟) and tamid („always‟), which behave as predicted <strong>in</strong><br />

B&C‟s tests (<strong>the</strong>y pattern as conventionalized and free, respectively), <strong>the</strong> behavior of bixlal is<br />

not consistent with e.g. extraction and reduced pronouns. We exam<strong>in</strong>e potential explanations for<br />

this behavior, and connect it to <strong>the</strong> fact that, unlike rak, bixlal operates on a completely different<br />

scale of alternatives when stressed.


Daniel Gutzmann (Frankfurt) & Kathar<strong>in</strong>a Hartmann (Humboldt)<br />

Dissociat<strong>in</strong>g verum from focus<br />

The term verum focus as it is commonly used refers to a special k<strong>in</strong>d of H*L accent that <strong>in</strong>stead of<br />

focus<strong>in</strong>g <strong>the</strong> accent bear<strong>in</strong>g expression, is used to focus a covert operator called verum that provides<br />

<strong>the</strong> truth value/polarity of <strong>the</strong> propositional content of a sentence (Höhle 1992). While this focus<br />

accent <strong>the</strong>sis, or fat as we call this approach henceforth, may be plausible for languages like German<br />

or English, we argue that it is not so if typological more diverse languages are taken <strong>in</strong>to consideration<br />

as well. Instead, we suggest an alternative <strong>the</strong>sis, which we call <strong>the</strong> lexical operator <strong>the</strong>sis or lot.<br />

Accord<strong>in</strong>g to <strong>the</strong> fat, verum focus is what <strong>the</strong> term suggests: it is an ord<strong>in</strong>ary focus accent on<br />

<strong>the</strong> covert verum operator (Bür<strong>in</strong>g 2006; Höhle 1992; Zimmermann & Hole 2008). In order for <strong>the</strong><br />

fat to work properly, verum must be present <strong>in</strong> every sentence and <strong>the</strong> accent focusses it. In order<br />

to ensure that, <strong>in</strong> <strong>the</strong> absence of a verum accent, verum has no efficient contribution to <strong>the</strong> mean<strong>in</strong>g<br />

of a sentence, it has to be rendered as an identity function on truth values/propositions, <strong>the</strong>refore<br />

be<strong>in</strong>g <strong>the</strong> reverse of negation (Zimmermann & Hole 2008). The fat can be summarized as follows.<br />

(fat) verum focus ∶= covert propositional identity function + focus mark<strong>in</strong>g<br />

In contrast, <strong>the</strong> lot does not assume an omnipresent identity function. Instead, velum is rendered as<br />

a conversational operator (cf. Gutzmann & Castroviejo Miró 2011; Romero & Han 2004) for different<br />

formalizations of this idea) that, by convention, is realized by an accent.<br />

(lot) verum focus ∶= conversational operator realized by accent<br />

In languages like German and English, this happens to be realized by a pitch accent just like focus<br />

is. However, this ra<strong>the</strong>r is a k<strong>in</strong>d of homonymie. Accord<strong>in</strong>g to <strong>the</strong> lot, verum focus is no focus at<br />

all, despite <strong>the</strong> traditional term.<br />

Both <strong>the</strong>ories make different predictions. (i) Means of focus/verum mark<strong>in</strong>g. S<strong>in</strong>ce verum focus<br />

is focus, <strong>the</strong> fat predicts that verum focus is marked by <strong>the</strong> same means as focus, whereas <strong>the</strong> lot<br />

does not predict that <strong>the</strong>re is a necessary overlap between verum and focus mark<strong>in</strong>g strategy. (ii) Cooccurrences<br />

of focus and verum. If a language exhibits multiple foci, <strong>the</strong> FAT predicts that verum and<br />

focus can also co-occur. Fur<strong>the</strong>rmore, <strong>the</strong> co-occurrence of verum and focus is not to be expected<br />

if a languages prohibits multiple foci. In contrast, <strong>the</strong> LOT does not predict such a correlation and,<br />

ceteris paribus, all four comb<strong>in</strong>ations should be possible (Of course, <strong>the</strong>re could be o<strong>the</strong>r reasons<br />

why one or <strong>the</strong> o<strong>the</strong>r comb<strong>in</strong>ation is not possible or widespread). (iii) Association with focus. S<strong>in</strong>ce,<br />

accord<strong>in</strong>g to <strong>the</strong> fat, <strong>the</strong> verum operator is an ord<strong>in</strong>ary focus, it should be possible to have focus<br />

sensitive operators to be associated with verum focus. The lot excludes this possibility.<br />

While it is hard to differentiate between <strong>the</strong> two <strong>the</strong>ses on <strong>the</strong> basis of German or English alone,<br />

tak<strong>in</strong>g <strong>in</strong>to account languages that differ from <strong>the</strong> well-studied languages can help to evaluate <strong>the</strong>m<br />

properly. We will present arguments for Chadic languages that favor <strong>the</strong> lot over <strong>the</strong> fat.<br />

Ad (i). In contrast to <strong>in</strong>tonational languages like German or English, that do not exhibit formal<br />

differences <strong>in</strong> <strong>the</strong> mark<strong>in</strong>g of focus and verum, <strong>the</strong> realization of verum differs considerably from <strong>the</strong><br />

mark<strong>in</strong>g of constituent focus <strong>in</strong> <strong>the</strong> Chadic languages. In Bura, for <strong>in</strong>stance, subject focus is marked<br />

by <strong>the</strong> subsequent focus marker án (Hartmann & Zimmermann t.a.), cf. (1a). In contrast, verum is<br />

marked by <strong>the</strong> particle ku, as <strong>in</strong> (1b).<br />

(1) a. subject focus: án<br />

[P<strong>in</strong>dar]F án sá mbal.<br />

P. FM dr<strong>in</strong>k beer<br />

“P<strong>in</strong>dar drank beer.”<br />

b. verum: ku<br />

P<strong>in</strong>dár ku sá mbal.<br />

P. VERUM dr<strong>in</strong>k beer<br />

“P<strong>in</strong>dar did dr<strong>in</strong>k beer.”<br />

Chadic languages exhibit quite a variety <strong>in</strong> focus mark<strong>in</strong>g. Crucially, focus mark<strong>in</strong>g formally differs<br />

1


from verum mark<strong>in</strong>g <strong>in</strong> all languages we <strong>in</strong>vestigated so far (South Marghi, Bura (Central Chadic),<br />

Hausa, Bole, Ngizim (West Chadic)).<br />

Ad (ii). Like English or German, Bura allows for multiple focus and/or question mark<strong>in</strong>g:<br />

(2) a. Ga masta mi ama ri?<br />

2SG.S buy what where Q<br />

b.<br />

“What did you buy where?”<br />

Iya masta [kwara ni]F [akwa kwasuku]F, iya masta …<br />

1SG.S buy donkey DEF at market 1SG.S buy<br />

“I bought <strong>the</strong> donkey at <strong>the</strong> market, I bought …”<br />

However, contrary to what <strong>the</strong> fat predicts, verum cannot co-occur with wh-questions and focus <strong>in</strong><br />

<strong>the</strong> Chadic languages, see Bura aga<strong>in</strong> for illustration.<br />

(3) a. Wán (*ku) sá mbal? *subj wh + verum<br />

b. [P<strong>in</strong>dár]F án (*ku) sá mbal. *subj foc + verum<br />

<strong>in</strong>tended: “P<strong>in</strong>dar did dr<strong>in</strong>k <strong>the</strong> beer.”<br />

While <strong>the</strong> English and German pattern does not conflict with <strong>the</strong> fat, <strong>the</strong> Chadic data is at odds<br />

with it and favors <strong>the</strong> lot. We leave it for fur<strong>the</strong>r <strong>in</strong>vestigation <strong>the</strong> availability of <strong>the</strong> rema<strong>in</strong><strong>in</strong>g two<br />

configurations.<br />

Ad (iii). Focus sensitive particles are generally impossible <strong>in</strong> <strong>the</strong> presence of <strong>the</strong> verum operator<br />

ku <strong>in</strong> Bura as shown <strong>in</strong> (4) (Hartmann, Jacob & Zimmermann 2008: 35). The focus sensitive particle<br />

daci, which may associate at a distance (Hartmann & Zimmermann 2008), nei<strong>the</strong>r associates with<br />

<strong>the</strong> contrastively focused verb, nor with <strong>the</strong> verum operator itself. This would be unexpected given<br />

<strong>the</strong> fat.<br />

(4) Mwala ní adí tsá ní wá ama tsá (*ku) buhá ní daci.<br />

woman DEF EXIST hit 3SG NEG but 3SG VERUM push 3SG only<br />

“The woman didn’t hit him, but she only pushed him”.<br />

We conclude that both analyses of verum are <strong>in</strong> pr<strong>in</strong>ciple able to account for <strong>the</strong> observed facts <strong>in</strong> <strong>the</strong><br />

<strong>in</strong>tonational languages. The Chadic languages, however, do not show evidence <strong>in</strong> favour of a focus<br />

accent analysis of verum. Instead <strong>the</strong>y support a lexical analysis of verum as an epistemic conversational<br />

operator, an approach that disassociates verum from <strong>the</strong> notion of focus. If we generalize<br />

<strong>the</strong>se f<strong>in</strong>d<strong>in</strong>g to languages like English or German – which gave rise to <strong>the</strong> focus analysis <strong>in</strong> first place<br />

– <strong>the</strong> accent pattern traditionally associated with verum focus is not licensed by <strong>the</strong> focus status of<br />

<strong>the</strong> covert verum operator. Instead, <strong>the</strong> accent itself is <strong>the</strong> realization of <strong>the</strong> lexical operator. That is,<br />

verum focus is no focus at all.<br />

Bür<strong>in</strong>g, Daniel. 2006. Intonation und Informationsstruktur.<br />

Grammatik und darüber h<strong>in</strong>aus. In Hardarik Blühdorn, Eva<br />

Bre<strong>in</strong>dl & Ulrich Hermann Waßner eds., Text — Verstehen,<br />

144–163. Berl<strong>in</strong> and New York: Walter de Gruyter.<br />

Gutzmann, Daniel & Elena Castroviejo Miró. 2011. The dimensions<br />

of verum. In Olivier Bonami & Patricia Cabredo-<br />

Hofherr eds., Empirical Issues <strong>in</strong> Syntax and Semantics 8,<br />

143–165.<br />

Hartmann, Kathar<strong>in</strong>a, Peggy Jacob & Malte Zimmermann. 2008.<br />

Focus asymmetries <strong>in</strong> Bura. In Sh<strong>in</strong>shiro Ishihara, S. Petrova<br />

& Anne Schwarz eds., Interdiscipl<strong>in</strong>ary Studies on Information<br />

Structure 10, 45–92. Potsdam: Universitätsverlag Potsdam.<br />

2<br />

Hartmann, Kathar<strong>in</strong>a & Malte Zimmermann. 2008. Not only<br />

“only” but “too” too. Alternative sensitive particles <strong>in</strong> Bura.<br />

Proceed<strong>in</strong>gs of S<strong>in</strong>n und Bedeutung 12. 196–211.<br />

Hartmann, Kathar<strong>in</strong>a & Malte Zimmermann. t.a. Focus mark<strong>in</strong>g<br />

<strong>in</strong> Bura. Semantic uniformity matches syntactic heterogeneity.<br />

Natural Language and L<strong>in</strong>guistic Theory.<br />

Höhle, Tilman N. 1992. Über Verum-Fokus im Deutschen. In<br />

Joachim Jacobs ed., Informationsstruktur und Grammatik,<br />

112–141. Opladen: Westdeutscher Verlag.<br />

Romero, Maribel & Chung-hye Han. 2004. On negative yes/no<br />

questions. <strong>L<strong>in</strong>guistics</strong> and Philosophy 27(5). 609–658.<br />

Zimmermann, Malte & Daniel Hole. 2008. “Predicate focus,<br />

verum focus, verb focus. Similarities and difference”. Talk at<br />

<strong>the</strong> Potsdam-London IS Meet<strong>in</strong>g. 12. 12. 2008.


When Maria is considered to be he. Gender mismatch effects dur<strong>in</strong>g pronoun resolution<br />

Pronoun resolution is an <strong>in</strong>tegral part of sentence comprehension. To accomplish this task, <strong>the</strong><br />

parser uses various sources of <strong>in</strong>formation, <strong>in</strong>clud<strong>in</strong>g syntactic, semantic and pragmatic <strong>in</strong>formation<br />

(for an overview see Garnham, 2001). With regard to syntactic <strong>in</strong>formation, we can<br />

dist<strong>in</strong>guish two k<strong>in</strong>ds of constra<strong>in</strong>ts: phrase-structural constra<strong>in</strong>ts which restrict <strong>the</strong> set of candidate<br />

antecedents <strong>in</strong> ter<strong>ms</strong> of phrase-structure geometry (e.g., B<strong>in</strong>d<strong>in</strong>g Pr<strong>in</strong>ciples, Cho<strong>ms</strong>ky,<br />

1981) and morpho-syntactic constra<strong>in</strong>ts which restrict <strong>the</strong> set of candidate antecedents <strong>in</strong> ter<strong>ms</strong><br />

of feature compatibility. The onl<strong>in</strong>e-application of both types of constra<strong>in</strong>ts has been attested <strong>in</strong><br />

prior studies (e.g. Badecker & Straub, 2002; Kennison, 2003; Sturt, 2003; van Gompel & Liversedge,<br />

2003; Kazan<strong>in</strong>a, Lau, Lieberman, Yoshida & Phillips, 2007). The exact tim<strong>in</strong>g, however,<br />

rema<strong>in</strong>s an open issue, and <strong>in</strong> particular <strong>the</strong> question whe<strong>the</strong>r syntactic constra<strong>in</strong>ts constra<strong>in</strong> <strong>the</strong><br />

search for an antecedent from <strong>the</strong> onset on or only later <strong>in</strong> a second step evaluat<strong>in</strong>g candidates.<br />

The present study, which focuses on person, gender and number agreement, addresses two ma<strong>in</strong><br />

questions: (i) Do agreement requirements exclude featurally <strong>in</strong>appropriate NPs immediately or<br />

only at a later stage? (ii) Are <strong>the</strong>re differential effects of person versus gender agreement? The<br />

study makes use of a mismatch paradigm: <strong>the</strong> pronoun is preceded by an NP which matches<br />

or mismatches <strong>the</strong> pronoun <strong>in</strong> agreement features. In contrast to previous studies (e.g., Kennison,<br />

2003), <strong>the</strong> mismatch sentences do not represent <strong>in</strong>coherent discourses lack<strong>in</strong>g a proper<br />

antecedent for <strong>the</strong> pronoun. As shown <strong>in</strong> (1), <strong>the</strong> pronoun is sandwiched between two potential<br />

antecedents – one, <strong>the</strong> ‘distractor’, preced<strong>in</strong>g <strong>the</strong> pronoun and ano<strong>the</strong>r one, <strong>the</strong> actual antecedent,<br />

follow<strong>in</strong>g it.<br />

(1)<br />

Als<br />

When<br />

Distractor<br />

Maria<br />

M.<br />

Jana Häussler (Potsdam) & Markus Bader (Frankfurt)<br />

hörte,<br />

heard<br />

dass<br />

that<br />

er<br />

he<br />

abreist,<br />

departs<br />

war<br />

was<br />

der Mann<br />

<strong>the</strong> man<br />

Pronoun Antecedent<br />

schon<br />

already<br />

weg<br />

away<br />

The cataphoric pronoun is conta<strong>in</strong>ed <strong>in</strong> a fronted adjunct clause, for which Gordon & Hendrick<br />

(1997) have shown that cataphoric reference is highly acceptable. In <strong>the</strong> experiments <strong>the</strong> fronted<br />

adjunct clause is complex itself consist<strong>in</strong>g of a matrix clause conta<strong>in</strong><strong>in</strong>g <strong>the</strong> distractor and<br />

an embedded clause conta<strong>in</strong><strong>in</strong>g <strong>the</strong> pronoun. The actual antecedent occurs <strong>in</strong> <strong>the</strong> ma<strong>in</strong> clause<br />

<strong>the</strong>reafter.<br />

We present two self-paced read<strong>in</strong>g experiments that <strong>in</strong>vestigate sentences follow<strong>in</strong>g <strong>the</strong> schema<br />

<strong>in</strong> (1). Both experiments vary <strong>the</strong> feature specifications of <strong>the</strong> distractor (1st person pronoun, fem<strong>in</strong><strong>in</strong>e<br />

proper name or mascul<strong>in</strong>e proper name). Experiment 1 fur<strong>the</strong>rmore varies <strong>the</strong> pronoun,<br />

which was ei<strong>the</strong>r er (‘he’) or sie (‘she’). Experiment 2 concentrates on <strong>the</strong> number ambiguous<br />

pronoun sie (‘she/<strong>the</strong>y’) and varies its number specification signaled by subject-verb agreement<br />

on <strong>the</strong> correspond<strong>in</strong>g verb, cf. (2).<br />

(2) Als Maria erfuhr, . . . (‘When Maria heard’)<br />

a. dass sie abreisen muss, hatte die Tante die Stadt längst verlassen.<br />

that she depart must.SG had.SG <strong>the</strong> aunt <strong>the</strong> town already left<br />

‘When Maria heard that she had to depart <strong>the</strong> aunt had already left <strong>the</strong> town.’<br />

b. dass sie abreisen müssen, hatten die Tanten die Stadt längst verlassen.<br />

that <strong>the</strong>y depart must.PL had.PL <strong>the</strong> aunts <strong>the</strong> town already left<br />

‘When Maria heard that <strong>the</strong>y had to depart <strong>the</strong> aunts had already left <strong>the</strong> town.’<br />

1


<strong>Read<strong>in</strong>g</strong> times <strong>in</strong> <strong>the</strong> pronoun region are prolonged when pronoun and distratcor mismatch <strong>in</strong><br />

gender. In Experiment 1, <strong>the</strong> effect starts on <strong>the</strong> pronoun itself and lasts through <strong>the</strong> whole<br />

embedded clause; <strong>in</strong> Experiment 2, <strong>the</strong> effect is slightly delayed and starts only immediately<br />

after <strong>the</strong> pronoun. We <strong>in</strong>terpret this mismatch penalty as an <strong>in</strong>dication for <strong>the</strong> parser’s attempt to<br />

l<strong>in</strong>k <strong>the</strong> pronoun to <strong>the</strong> <strong>in</strong>appropriate distractor. Despite <strong>the</strong> feature <strong>in</strong>compatibility and although<br />

<strong>the</strong> pronoun occurs <strong>in</strong> a configuration where it could well be (and <strong>in</strong> fact is) a cataphor, <strong>the</strong> parser<br />

yields to <strong>the</strong> pressure of establish<strong>in</strong>g an antecedent as soon as possible. In a way, this pressure<br />

overrules <strong>the</strong> agreement constra<strong>in</strong>t and lets <strong>the</strong> parser ignore <strong>the</strong> alternative offered by b<strong>in</strong>d<strong>in</strong>g<br />

constra<strong>in</strong>ts, namely establish<strong>in</strong>g a cataphoric b<strong>in</strong>d<strong>in</strong>g relation. At <strong>the</strong> same time, <strong>the</strong> agreement<br />

violation is noticed as reflected <strong>in</strong> <strong>the</strong> prolonged read<strong>in</strong>g times. Hence, <strong>the</strong> agreement constra<strong>in</strong>t<br />

is active but it does not guide <strong>the</strong> parser’s first choice. Taken toge<strong>the</strong>r, this argues for a twostep<br />

model: first a referential l<strong>in</strong>k is established and only <strong>in</strong> a second step its appropriateness<br />

is evaluated. However, gender and person features are treated differently. Notably, a mismatch<br />

effect occurs only <strong>in</strong> case of a gender mismatch but not with a person conflict. Apparently,<br />

<strong>the</strong> first-person pronoun distractor is not considered a potential antecedent for <strong>the</strong> third-person<br />

pronoun. We conclude that person features restrict <strong>the</strong> set of candidate antecedents immediately.<br />

We will discuss our f<strong>in</strong>d<strong>in</strong>gs <strong>in</strong> <strong>the</strong> context of <strong>the</strong> feature delay hypo<strong>the</strong>sis (van Gompel &<br />

Liversedge, 2003) and <strong>the</strong> active search hypo<strong>the</strong>sis (Kazan<strong>in</strong>a et al., 2007).<br />

Fur<strong>the</strong>rmore, Experiment 2 provides evidence for <strong>the</strong> hypo<strong>the</strong>sis that <strong>the</strong> parser’s association<br />

with a once established antecedent is really strong. Remember that <strong>the</strong> pronoun sie is number<br />

ambiguous. Disambiguation is achieved by subject-verb agreement with <strong>the</strong> correspond<strong>in</strong>g verb<br />

(muss/müssen <strong>in</strong> (2)). While read<strong>in</strong>g times for this verb do not show any difference between<br />

s<strong>in</strong>gular and plural <strong>in</strong> sentences with a 1st-person distractor, disambiguation towards plural<br />

causes a process<strong>in</strong>g disruption <strong>in</strong> sentences with a fem<strong>in</strong><strong>in</strong>e distractor. To our view this suggests<br />

that just adjust<strong>in</strong>g <strong>the</strong> pronoun’s number feature is basically costless whereas break<strong>in</strong>g up <strong>the</strong><br />

l<strong>in</strong>k to a seem<strong>in</strong>g antecedent is particularly hard.<br />

Literatur<br />

Badecker, W. & Straub, K. (2002). The process<strong>in</strong>g role of structural constra<strong>in</strong>ts on <strong>the</strong> <strong>in</strong>terpretation<br />

of pronouns and anaphors. Journal of Experimental Psychology 28, 748–769.<br />

Cho<strong>ms</strong>ky, N. (1981). Lectures on government and b<strong>in</strong>d<strong>in</strong>g. Dordrecht: Foris.<br />

Garnham, A. (2001). Mental models and <strong>the</strong> <strong>in</strong>terpretation of anaphora. Hove: Psychology<br />

Press.<br />

Gordon, P. C. & Hendrick, R. (1997). Intuitive knowledge of l<strong>in</strong>guistic co-reference. Cognition<br />

62, 325–370.<br />

Kazan<strong>in</strong>a, N., Lau, E. F., Lieberman, M., Yoshida, M. & Phillips, C. (2007). The effect of<br />

syntactic constra<strong>in</strong>ts on <strong>the</strong> process<strong>in</strong>g of backwards anaphora. Journal of Memory and Language<br />

56, 384–409.<br />

Kennison, S. M. (2003). Comprehend<strong>in</strong>g <strong>the</strong> pronouns her, him, and his: Implications for <strong>the</strong>ories<br />

of referential process<strong>in</strong>g. Journal of Memory and Language 49, 335–352.<br />

Sturt, P. (2003). The time-course of <strong>the</strong> application of b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts <strong>in</strong> reference resolution.<br />

Journal of Memory and Language 48, 542–562.<br />

van Gompel, R. P. & Liversedge, S. P. (2003). The <strong>in</strong>fluence of morphological <strong>in</strong>formation on<br />

cataphoric pronoun assignment. Journal of Experimental Psychology 29, 128–139.<br />

2


Claire Halpert (MIT)<br />

Structural case and <strong>the</strong> nature of vP <strong>in</strong> Zulu<br />

Recent work has demonstrated that <strong>in</strong> Icelandic and Faroese, a nom<strong>in</strong>al can move around a head<br />

before <strong>the</strong> head probes (Holmberg and Hroársdóttir 2004; Sigur�sson and Holmberg 2008, Asar<strong>in</strong>a<br />

2011). As a result of this order of movement and prob<strong>in</strong>g, <strong>the</strong> trace of <strong>the</strong> moved nom<strong>in</strong>al does<br />

not act as an <strong>in</strong>tervener. I show that we f<strong>in</strong>d <strong>in</strong>dependent evidence for this type of order<strong>in</strong>g <strong>in</strong> two<br />

syntactic puzzles <strong>in</strong> Zulu: (1) <strong>the</strong> ‘conjo<strong>in</strong>t/disjo<strong>in</strong>t’ alternation (van der Spuy 1993; Buell 2005)<br />

and (2) <strong>the</strong> distribution of augmentless nom<strong>in</strong>als (Buell 2011). I argue for a unified account of <strong>the</strong>se<br />

two puzzles, draw<strong>in</strong>g on recent work that argues for <strong>the</strong> ability of syntactic operations to be countercyclic<br />

and asymmetrical without yield<strong>in</strong>g a crash (Ndayiragije 1999, Legate 2005, Prem<strong>in</strong>ger 2009,<br />

2010, 2011). This analysis also suggests that nom<strong>in</strong>als <strong>in</strong> Zulu are subject to <strong>the</strong> Case Filter (contra<br />

Harford Perez 1985, Diercks to appear, a. o. on <strong>the</strong> lack of case effects <strong>in</strong> Bantu).<br />

Puzzle 1: Conjo<strong>in</strong>t/disjo<strong>in</strong>t. In certa<strong>in</strong> tenses, Zulu verbs alternate between a ‘conjo<strong>in</strong>t’ form<br />

(marked by Ø <strong>in</strong> present) and a ‘disjo<strong>in</strong>t’ form (marked by -ya- <strong>in</strong> present). This alternation is<br />

syntactically conditioned: <strong>the</strong> conjo<strong>in</strong>t form is used if vP conta<strong>in</strong>s overt material o<strong>the</strong>r than <strong>the</strong><br />

verb and <strong>the</strong> disjo<strong>in</strong>t form is used if not (van der Spuy 1993; Buell 2005, 2006):<br />

(1) uSipho u- *(ya)- dla (2) ku- (*ya)- dla uSipho (3) uSipho u- (*ya)- dla iqanda<br />

1Sipho 1s- *(YA)- eat 17s- (*YA)- eat 1Sipho 1Sipho 1s- (*YA)- eat 5egg<br />

‘Sipho is eat<strong>in</strong>g.’<br />

‘SIPHO’s eat<strong>in</strong>g.’<br />

‘Sipho is eat<strong>in</strong>g an egg.’<br />

Puzzle 2: Augmentless nom<strong>in</strong>al distribution. Nom<strong>in</strong>als <strong>in</strong> Zulu are typically marked with an<br />

<strong>in</strong>itial vowel (<strong>the</strong> ‘augment’) that reflects noun class. In a subset of environments that permit<br />

augmented nom<strong>in</strong>als, this vowel may be omitted. The omission of <strong>the</strong> augment is subject to various<br />

conditions (see Buell 2011), but <strong>the</strong> ones I focus on here are structural: augmentless nom<strong>in</strong>als must<br />

appear <strong>in</strong>side vP, and are limited to certa<strong>in</strong> configurations with<strong>in</strong> vP.<br />

(4) a. SVO with augmentless object<br />

c. �VSO +augment -augment<br />

umuntu a- ka-dli qanda<br />

a- ku-dli muntu iqanda<br />

1person NEG1Seat NEG egg<br />

NEG17Seat NEG 1person 5egg<br />

‘A/<strong>the</strong> person didn’t eat any egg.’<br />

‘NOBODY is eat<strong>in</strong>g <strong>the</strong>/an/any egg.’<br />

b. *VSO -augment -augment<br />

d. *VSO +augment -augment<br />

*a-ku-dl-i muntu qanda<br />

a- ku-dli umuntu qanda<br />

NEG-17S-eat-NEG 1person egg<br />

NEG17Seat NEG 1person 5egg<br />

Solution. The analysis depends on three components–failure to agree, asymmetrical case relationships,<br />

and <strong>the</strong> ability of an argument to move around a head before that head probes.<br />

Recent work by Prem<strong>in</strong>ger (2009, 2010, 2011) suggests that it is not obligatory for a prob<strong>in</strong>g<br />

head to undergo Agreement <strong>in</strong> order for a derivation to converge. Prem<strong>in</strong>ger argues, however, that<br />

Agree itself is not optional: heads probe obligatorily and so will always Agree if a goal is present.<br />

If a probe fails to f<strong>in</strong>d a goal, <strong>the</strong> derivation will still converge as long as prob<strong>in</strong>g was attempted.<br />

We can understand <strong>the</strong> Zulu conjo<strong>in</strong>t/disjo<strong>in</strong>t pattern <strong>in</strong> ter<strong>ms</strong> of Prem<strong>in</strong>ger’s proposal: <strong>the</strong> head<br />

responsible for <strong>the</strong> alternation is an (unrelativized) probe that searches <strong>the</strong> vP for an XP to agree<br />

with. When <strong>the</strong> vP is empty, and thus lacks a goal, <strong>the</strong> derivation still converges, as predicted by<br />

Prem<strong>in</strong>ger. In Zulu we see a morphological marker of this failure: where Agree does not occur, <strong>the</strong><br />

prob<strong>in</strong>g head spells out as -ya-, ra<strong>the</strong>r than Ø.<br />

I propose that augmented nom<strong>in</strong>als, which exhibit no structural restrictions, are <strong>in</strong>herently cased<br />

<strong>in</strong> Zulu. Augmentless nom<strong>in</strong>als, on <strong>the</strong> o<strong>the</strong>r hand, do display structural restrictions <strong>in</strong> l<strong>in</strong>e with<br />

classic case effects. I argue that this structurally-restricted distribution is evidence for structural<br />

case assignment with<strong>in</strong> vP. In l<strong>in</strong>e with Ndayiragije (1999), <strong>the</strong> case probe can check even DPs that<br />

1


don’t need case, such as augmented nom<strong>in</strong>als. However, while for Ndayiragije Bantu nom<strong>in</strong>als<br />

generally don’t require case, and thus <strong>the</strong> case prob<strong>in</strong>g is always driven by <strong>the</strong> prob<strong>in</strong>g head <strong>in</strong><br />

Bantu languages, I argue that augmentless nom<strong>in</strong>als do require case <strong>in</strong> Zulu and must be checked<br />

to prevent a crash. The result of <strong>the</strong> possibility for asymmetrical probe-goal relationships is <strong>the</strong><br />

follow<strong>in</strong>g: <strong>the</strong> case-licen<strong>in</strong>g head does not specifically probe for an augmentless nom<strong>in</strong>al and will<br />

not cause a crash regardless of whe<strong>the</strong>r it f<strong>in</strong>ds any XP to check. Augmentless nom<strong>in</strong>als that are<br />

not checked, however, do cause a crash and thus only surface <strong>in</strong> particular configurations.<br />

With <strong>the</strong>se concepts <strong>in</strong> place, we can understand <strong>the</strong> mechanism driv<strong>in</strong>g conjo<strong>in</strong>t/disjo<strong>in</strong>t alternations<br />

and augmentless nom<strong>in</strong>al distribution. Both phenomena depend on a head, L, that is<br />

immediately above vP (because vP-<strong>in</strong>ternal subjects are <strong>in</strong> its prob<strong>in</strong>g doma<strong>in</strong>) and that probes for<br />

<strong>the</strong> highest XP. We can understand <strong>the</strong> patterns <strong>in</strong> (4) <strong>in</strong> ter<strong>ms</strong> of this structure: <strong>the</strong> licens<strong>in</strong>g head<br />

probes vP and checks <strong>the</strong> highest XP, which is <strong>the</strong> object <strong>in</strong> (4a) and <strong>the</strong> subject <strong>in</strong> (4b,c,d). In<br />

(4a), <strong>the</strong> object is augmentless but is licensed through Agree with <strong>the</strong> functional head. Similarly,<br />

<strong>the</strong> augmentless subjects <strong>in</strong> (4b,c) Agree with and are licensed by <strong>the</strong> head. In (4c,d) <strong>the</strong> construction<br />

is ungrammatical because of <strong>the</strong> presence of a lower augmentless nom<strong>in</strong>al: <strong>the</strong> object. Here,<br />

<strong>the</strong> subject <strong>in</strong>tervenes and prevents <strong>the</strong> object from be<strong>in</strong>g licensed, caus<strong>in</strong>g a crash. As discussed<br />

above, all of <strong>the</strong> for<strong>ms</strong> <strong>in</strong> (4) use conjo<strong>in</strong>t morphology because <strong>the</strong> licens<strong>in</strong>g head is able to Agree.<br />

In cases where noth<strong>in</strong>g appears <strong>in</strong> vP, <strong>the</strong> licens<strong>in</strong>g head fails to Agree, but <strong>the</strong> derivation still<br />

converges, with <strong>the</strong> head spell<strong>in</strong>g out as -ya-. The proposed structure is schematized below <strong>in</strong> (5):<br />

(5) LP<br />

L(icenser) vP<br />

S<br />

augment<br />

optional<br />

v o VP<br />

V O<br />

augment<br />

necessary<br />

(6) LP<br />

L<br />

ya<br />

tS<br />

vP<br />

v o VP<br />

V<br />

prob<strong>in</strong>g<br />

fails!<br />

The tim<strong>in</strong>g of operations. Both puzzles described here are sensitive to syntactic movement. The<br />

head L o treats constructions <strong>in</strong> which elements move out of vP dur<strong>in</strong>g <strong>the</strong> derivation <strong>in</strong> <strong>the</strong> same<br />

way as constructions <strong>in</strong> which <strong>the</strong> vP is empty throughout. This is a familiar pattern from Icelandic<br />

and Faroese, where dative experiencers are able to move around a number probe before it probes<br />

(Holmberg & Hróarsdóttir 2004, a.o.). In Icelandic, however, we f<strong>in</strong>d optionality <strong>in</strong> agreement,<br />

<strong>in</strong>dicative of both orders of movement and prob<strong>in</strong>g. If operations are freely ordered with<strong>in</strong> <strong>the</strong><br />

phase (Cho<strong>ms</strong>ky 2005), this is an expected result. In Zulu, however, we only see <strong>the</strong> counter-cyclic<br />

movement-before-prob<strong>in</strong>g order. I propose that this pattern is due to <strong>the</strong> Activity Condition: nom<strong>in</strong>als<br />

<strong>in</strong> Zulu cannot undergo A movement after enter<strong>in</strong>g <strong>in</strong>to a case-check<strong>in</strong>g relationship with<br />

L (Cho<strong>ms</strong>ky 2000, 2001). While recent work has suggested that Bantu languages lack both case<br />

and Activity effects <strong>in</strong> <strong>the</strong> TP doma<strong>in</strong> (Carstens 2011, Carstens and Diercks to appear, Diercks to<br />

appear), I argue that we f<strong>in</strong>d both case and Activity at work <strong>in</strong> <strong>the</strong> lower part of <strong>the</strong> clause <strong>in</strong> Zulu.<br />

Selected References: Buell, L. 2005. Issues <strong>in</strong> Zulu Morphosyntax. PhD diss. Carstens, V. 2011.<br />

Hyperactivity and Hyperagreement <strong>in</strong> Bantu. L<strong>in</strong>gua. Diercks, M. to appear. Parameteriz<strong>in</strong>g case<br />

<strong>the</strong>ory: evidence from Bantu. Syntax. Holmberg, A. & T. Hroársdóttir 2004. Agreement and<br />

movement <strong>in</strong> Icelandic rais<strong>in</strong>g constructions. L<strong>in</strong>gua. Ndayiragije, J. 1999. Check<strong>in</strong>g economy.<br />

LI. Prem<strong>in</strong>ger, O. 2009. Break<strong>in</strong>g agreements: dist<strong>in</strong>guish<strong>in</strong>g agreement and clitic doubl<strong>in</strong>g by<br />

<strong>the</strong>ir failures. LI. Van der Spuy, A. 1993. Dislocated noun phrases <strong>in</strong> Nguni. L<strong>in</strong>gua.<br />

2


Bart Hollebrandse, Petra Hendriks & Jacolien van Rij (Gron<strong>in</strong>gen)<br />

Eye gaze patterns reveal subtle discourse effects on object pronoun resolution<br />

Discourse coherence plays a major role <strong>in</strong> <strong>in</strong>terpret<strong>in</strong>g sentences. This paper shows<br />

that adults are sensitive to very subtle <strong>in</strong>formation from discourse. To do this, we use<br />

<strong>the</strong> well-known phenomenon of Delayed Pr<strong>in</strong>ciple B Effect (Chien and Wexler, 1990;<br />

Koster, 1993) and show that children are capable of circumvent<strong>in</strong>g <strong>the</strong> DPBE on <strong>the</strong><br />

basis of subtle discourse <strong>in</strong>formation.<br />

Issue Spenader, Smits & Hendriks (2009) found that coherent discourse helps <strong>the</strong><br />

DPBE-child to f<strong>in</strong>d <strong>the</strong> correct referent for a pronoun. The discourse-<strong>in</strong>troduction of a<br />

s<strong>in</strong>gle referent made <strong>the</strong> effect disappear. By add<strong>in</strong>g a condition <strong>in</strong> which we varied<br />

<strong>the</strong> order of referents as potentials antecedents for object pronouns we are zoom<strong>in</strong>g<br />

<strong>in</strong>to subtle discourse prom<strong>in</strong>ence effects such as first-mention bias (cf subject<br />

pronouns <strong>in</strong> Arnold, Eisenband, Brown-Schmidt & Trueswell (2000)).<br />

Our Study We have not only collected accuracy as a measure, but also eye-gaze.<br />

It is especially eye-gaze which can give us subtle onl<strong>in</strong>e <strong>in</strong>sights <strong>in</strong>to <strong>the</strong> resolution of<br />

pronouns and <strong>the</strong> discourse effects on it. We tested <strong>the</strong> Agent–Patient order, <strong>the</strong> order<br />

Patient-Agent and <strong>the</strong> <strong>in</strong>troduction of a s<strong>in</strong>gle referent (1)-(3).<br />

(1) Double Topic Agent-Patient Condition<br />

Hier zie je een olifant en een d<strong>in</strong>o. De olifant heeft hem met een hamer geslagen.<br />

here see you an elephant and a d<strong>in</strong>osaur. The elephant has him with a hammer hit<br />

“Here you see an elephant and a d<strong>in</strong>osaur. The elephant has hit him with a<br />

hammer.”<br />

(2) S<strong>in</strong>gle Topic Condition<br />

Hier zie je een d<strong>in</strong>o. De olifant heeft hem met een hamer geslagen.<br />

here see you a d<strong>in</strong>osaur. The elephant has him with a hammer hit<br />

“Here you see a d<strong>in</strong>osaur. The elephant has hit him with a hammer.”<br />

(3) Double Topic Patient-Agent Condition<br />

Hier zie je een d<strong>in</strong>o en een olifant. De olifant heeft hem met een hamer geslagen.<br />

here see you a d<strong>in</strong>oaur and an elephant. The elephant has him with a hammer hit<br />

“Here you see a d<strong>in</strong>osaur and an elephant. The elephant has hit him with a<br />

hammer.”<br />

Participants and Design 33 Dutch children between <strong>the</strong> ages 4;2 and 6;5 (mean=<br />

5;4) were tested on <strong>the</strong> conditions (1) - (3), as well as 37 adults, us<strong>in</strong>g a Picture-<br />

Verification task. Each participant saw 48 ite<strong>ms</strong> (16 ite<strong>ms</strong> <strong>in</strong>troduc<strong>in</strong>g two referents,<br />

divided equally <strong>in</strong> <strong>the</strong> two orders and test<strong>in</strong>g pronouns and reflexives; 16 ite<strong>ms</strong><br />

<strong>in</strong>troduc<strong>in</strong>g s<strong>in</strong>gle referents, test<strong>in</strong>g pronouns and reflexives; 16 fillers). All ite<strong>ms</strong><br />

were tested <strong>in</strong> matched and mismatched cases, but participants saw only one variant<br />

of each item.<br />

Results The data was analyzed by compar<strong>in</strong>g l<strong>in</strong>ear mixed-effects models<br />

(<strong>in</strong>clud<strong>in</strong>g <strong>the</strong> maximum random effects structure supported by <strong>the</strong> data) to test <strong>the</strong><br />

effects of Context, Match and Group. In addition to a significant ma<strong>in</strong> effect of Match<br />

(χ2()=), show<strong>in</strong>g that children are more likely to say yes than no, we found a marg<strong>in</strong>al<br />

effect of <strong>in</strong>troduction of reference (χ2(1)= 18.621; p < .001) on children’s off-l<strong>in</strong>e<br />

performance (yes/no-answers) on <strong>the</strong> task: children perform more adult-like <strong>in</strong> <strong>the</strong><br />

S<strong>in</strong>gle Topic condition than <strong>the</strong> Agent-Patient Condition (β=0.860, SE=0.376, zvalue=2.288,<br />

p=.022), and perform better <strong>in</strong> <strong>the</strong> Patient-Agent Condition than <strong>in</strong> <strong>the</strong><br />

Agent-Patient Condition (β=-0831, SE=0.476, z-value=-1.747, p=.081). This shows<br />

that children are <strong>in</strong>deed capable of resolv<strong>in</strong>g <strong>the</strong> DPBE by use of discourse clues.


Adults show ceil<strong>in</strong>g performance <strong>in</strong> <strong>the</strong> offl<strong>in</strong>e answers. However, <strong>in</strong> <strong>the</strong>ir gazebehavior,<br />

adults show an effect of Context (see Figure 1). Adults look more at <strong>the</strong><br />

patient <strong>in</strong> <strong>the</strong> Agent-Patient order (1) than <strong>in</strong> <strong>the</strong> o<strong>the</strong>r two referent <strong>in</strong>troductions,<br />

whereas children do not show differences <strong>in</strong> gaze behavior between <strong>the</strong> different<br />

conditions. In <strong>the</strong> second 500 <strong>ms</strong> b<strong>in</strong> (500-100 <strong>ms</strong> from <strong>the</strong> onset of <strong>the</strong> anaphor) of<br />

<strong>the</strong> mismatch-ite<strong>ms</strong> we found a marg<strong>in</strong>al significant <strong>in</strong>teraction of Context and<br />

AgeGroup (χ2(2)=5.40; p=.067).<br />

Conclusions Our f<strong>in</strong>d<strong>in</strong>gs support two important po<strong>in</strong>ts: 1. Children are sensitive<br />

to very subtle effects of discourse prom<strong>in</strong>ence such as topicality and <strong>the</strong> first mention<br />

bias. This is shown <strong>in</strong> <strong>the</strong>ir offl<strong>in</strong>e data, but data of younger children alludes to onl<strong>in</strong>e<br />

differences as well. 2. The adult onl<strong>in</strong>e data clearly shows a different gaze pattern for<br />

<strong>the</strong> Agent-Patient <strong>in</strong>troduction than for <strong>the</strong> o<strong>the</strong>r ones. This onl<strong>in</strong>e f<strong>in</strong>d<strong>in</strong>g supports<br />

<strong>the</strong> idea that different referent <strong>in</strong>troductions, <strong>the</strong> Agent-Patient order, have an effect<br />

on pronoun resolution.<br />

patient<br />

equal<br />

agent<br />

patient<br />

equal<br />

agent<br />

Adults<br />

0 500 1000 1500<br />

Children<br />

0 500 1000 1500<br />

Figure 1: Eye gaze for pronoun resolution <strong>in</strong> <strong>the</strong> mismatched case (0 <strong>in</strong>dicates onset<br />

of <strong>the</strong> pronoun).<br />

Arnold, J. E., Eisenband, J. G., Brown-Schmidt, S. & Trueswell, J. C. (2000). The<br />

rapid use of gender <strong>in</strong>formation: evidence of <strong>the</strong> time course of pronoun<br />

resolution from eyetrack<strong>in</strong>g. Cognition, 76, B13–B26.<br />

Chien, Y. C., & Wexler, K. (1990). Children's Knowledge of Locality Conditions <strong>in</strong><br />

B<strong>in</strong>d<strong>in</strong>g as Evidence for <strong>the</strong> Modularity of Syntax and Pragmatics. Language<br />

Acquisition, 1(3), 225-295.<br />

Koster, C. (1993). Errors <strong>in</strong> anaphora acquisition. Doctoral dissertation, Utrecht<br />

University, Utrecht, The Ne<strong>the</strong>rlands.<br />

Spenader, J., Smits, E. J., & Hendriks, P. (2009). Coherent discourse solves <strong>the</strong><br />

Pronoun Interpretation Problem. Journal of Child Language, 36(1), 23-52.<br />

CAP<br />

CPA<br />

S<br />

CAP<br />

CPA<br />

S


Elim<strong>in</strong>at<strong>in</strong>g rightward movement: Extraposition as flexible l<strong>in</strong>earisation of adjuncts<br />

Traces were <strong>in</strong>troduced <strong>in</strong>to syntactic <strong>the</strong>ory to serve two purposes: <strong>the</strong>y allowed displaced<br />

elements to be construed as semantically related to <strong>the</strong>ir pre-movement positions, and allowed<br />

constra<strong>in</strong>ts on <strong>the</strong> locality of movement to be stated. These concerns perta<strong>in</strong> not only to <strong>the</strong><br />

more canonical <strong>in</strong>stances of leftward movement, but also to apparent <strong>in</strong>stances of rightward<br />

movement: <strong>the</strong> relative clause must be <strong>in</strong>terpreted as a modifier of ‘book’ <strong>in</strong> (1), and must be<br />

ruled to have violated some constra<strong>in</strong>t on locality <strong>in</strong> (2).<br />

(1) A book appeared [which was written by Cho<strong>ms</strong>ky]<br />

(2) * It was believed [that a book appeared] by everybody [which was written by Cho<strong>ms</strong>ky]<br />

The locality constra<strong>in</strong>ts on rightward movement have long been known to differ from those on<br />

leftward movement, rais<strong>in</strong>g significant <strong>the</strong>oretical issues: (i) why should <strong>the</strong>re be two k<strong>in</strong>ds of<br />

movement? and (ii) why should <strong>the</strong> two k<strong>in</strong>ds correlate so precisely with <strong>the</strong> l<strong>in</strong>ear direction<br />

of <strong>the</strong> displacement? We propose an account of <strong>the</strong> <strong>in</strong>terpretive and locality properties of this<br />

extraposition phenomenon that does not <strong>in</strong>volve movement or traces. We <strong>in</strong>stead rely only on<br />

<strong>the</strong> <strong>in</strong>dependently-motivated derivational flexibility <strong>in</strong> <strong>the</strong> <strong>in</strong>troduction of adjuncts proposed by<br />

Hunter (2010), <strong>the</strong>reby unify<strong>in</strong>g extraposition of adjuncts with <strong>the</strong> well-known flexibility <strong>in</strong> (3)<br />

and with <strong>the</strong> anti-reconstruction properties of adjuncts discussed by Lebeaux (1988) <strong>in</strong> (4).<br />

(3) a. Read a book quietly (is what) John did (‘quietly’ <strong>in</strong>terpreted and l<strong>in</strong>earised as part of VP)<br />

b. Read a book (is what) John did quietly (‘quietly’ <strong>in</strong>terpreted but not l<strong>in</strong>earised as part of VP)<br />

(4) a. * Which argument that Johni is a genius did hei believe?<br />

b. Which argument that Johni made did hei believe?<br />

A central assumption of Hunter’s analysis is that each maximal projection is a spellout doma<strong>in</strong>.<br />

Hence a derivation is naturally partitioned <strong>in</strong>to “chunks”. Dur<strong>in</strong>g each chunk C, <strong>the</strong>re is<br />

a s<strong>in</strong>gle syntactic head X such that every merge step <strong>in</strong> C establishes ei<strong>the</strong>r a complement or<br />

a specifier of X. At <strong>the</strong> end of this chunk, spellout applies and produces an atomic word-like<br />

object, lack<strong>in</strong>g <strong>in</strong>ternal syntax but with <strong>the</strong> semantics, phonology, and formal features of <strong>the</strong> derivedXP,<br />

which can participate <strong>in</strong> a subsequent chunk of <strong>the</strong> derivation. Under this conception,<br />

“adjunction” of an elementAtakes place whenAis <strong>in</strong>troduced <strong>in</strong>to <strong>the</strong> derivational workspace<br />

but never merged. As before, semantic <strong>in</strong>terpretation and l<strong>in</strong>earization of A take place at <strong>the</strong><br />

po<strong>in</strong>t of spellout, but, because of <strong>the</strong> absence of merger, <strong>the</strong>se processes proceed without <strong>the</strong><br />

benefit of <strong>the</strong> configurational guidance that governs <strong>the</strong> <strong>in</strong>terpretation and l<strong>in</strong>earization of arguments,<br />

e.g. <strong>the</strong> UTAH and <strong>the</strong> LCA. Hunter (2010) argues that this derivational dichotomy<br />

is supported by <strong>the</strong> <strong>in</strong>terpretive and order<strong>in</strong>g flexibility found <strong>in</strong> adjuncts but not arguments.<br />

Typically, adjunction ofAtoXP takes place dur<strong>in</strong>g <strong>the</strong> chunk where its semantic associateXP<br />

is constructed. This is what takes place <strong>in</strong> (3a): ‘quietly’ is <strong>in</strong>troduced dur<strong>in</strong>g <strong>the</strong> derivational<br />

chunk of <strong>the</strong> VP it modifies and is spelled out at <strong>the</strong> conclusion of this chunk, concatenat<strong>in</strong>g<br />

‘quietly’ at <strong>the</strong> end of <strong>the</strong> VP. This allows front<strong>in</strong>g of ‘read a book quietly’.<br />

Note though that because adjunction<br />

of A to XP does not require<br />

reference to <strong>the</strong> <strong>in</strong>ternal syntax of<br />

XP, <strong>the</strong>re is an alternative: A can<br />

be <strong>in</strong>troduced and spelled out dur<strong>in</strong>g<br />

<strong>the</strong> next higher chunk, say with head<br />

Y . Because spellout has access not<br />

Tim Hunter & Robert Frank (Yale)<br />

John<br />

VP<br />

read [a book]DP<br />

TP<br />

quietly<br />

quietly<br />

T [read a book]VP<br />

spellout<br />

−−−−−−→ read a book quietly<br />

spellout<br />

−−−−−−→ John read a book quietly<br />

only to Y , but also to <strong>the</strong> units merged as Y ’s complement and specifier, spellout can <strong>in</strong>terpret<br />

A as semantically modify<strong>in</strong>g one of <strong>the</strong>se units, say XP; with regard to l<strong>in</strong>earisation, we assume<br />

that <strong>the</strong>re are two options: ei<strong>the</strong>r (i) <strong>the</strong> adjunct A may be concatenated to <strong>the</strong> edge of<br />

its hostXP, which reproduces <strong>the</strong> effects of Lebeaux’s (1988) counter-cyclic adjunction, or (ii)<br />

<strong>the</strong> adjunct A may be concatenated to <strong>the</strong> edge of <strong>the</strong> entire Y P chunk’s output, an operation<br />

we will call “outer adjunction”. Outer adjunction is responsible for (3b): ‘quietly’ is <strong>in</strong>troduced<br />

dur<strong>in</strong>g T’s chunk, <strong>in</strong>terpreted at spellout as semantically modify<strong>in</strong>g T’s complement, namely


VP, but concatenated to <strong>the</strong> edge of TP. Because spellout applied to VP dur<strong>in</strong>g a prior chunk, it<br />

functions as a unit and can be subsequently fronted without <strong>the</strong> adjunct ‘quietly’.<br />

On this account, extraposition amounts to <strong>the</strong> outer<br />

adjunction of a DP modifier at <strong>the</strong> phrase immediately<br />

TP who knew his bro<strong>the</strong>r<br />

conta<strong>in</strong><strong>in</strong>g <strong>the</strong> host, i.e. VP for objects, or TP for subjects.<br />

This expla<strong>in</strong>s <strong>the</strong> strict locality properties observed<br />

by Balt<strong>in</strong> (1981): material extraposed from sub-<br />

some<br />

would [ride with Fred]VP<br />

ject position can (and must) be stranded under VP front<strong>in</strong>g/ellipsis, as shown <strong>in</strong> (5), whereas<br />

material extraposed from object position cannot be, as shown <strong>in</strong> (6).<br />

(5) [Some ] would ride with Fred [who knew his bro<strong>the</strong>r].<br />

a. Ride with Fred, some would who knew his bro<strong>the</strong>r. (RC stranded by front<strong>in</strong>g)<br />

b. * Ride with Fred who knew his bro<strong>the</strong>r, some would. (RC brought along by front<strong>in</strong>g)<br />

(6) John said that he would call [people ] up [who are from Boston], and . . .<br />

a. * . . . call people up he did who are from Boston. (RC stranded by front<strong>in</strong>g)<br />

b. . . . call people up who are from Boston he did. (RC brought along by front<strong>in</strong>g)<br />

As a result, elements extraposed from objects<br />

will <strong>in</strong> general precede those extraposed from sub-<br />

CP which were on <strong>the</strong> table<br />

jects; but if <strong>the</strong> object undergoes movement to a<br />

position higher than <strong>the</strong> subject, this gives rise to<br />

an additional derivational option for late adjunc-<br />

[which books]DP<br />

did [someone pick up . . . ]TP<br />

tion. The adjunct can be <strong>in</strong>troduced at <strong>the</strong> po<strong>in</strong>t where <strong>the</strong> object merges <strong>in</strong>to a higher specifier,<br />

giv<strong>in</strong>g rise to <strong>the</strong> possibility of <strong>the</strong> object modifier appear<strong>in</strong>g at <strong>the</strong> far right.<br />

(7) a. ? Which book did someone pick up [which was on <strong>the</strong> table] [who didn’t really<br />

want to]?<br />

b. Which book did someone pick up [who didn’t really want to] [which was on <strong>the</strong><br />

table]?<br />

Fox and Nissenbaum (1999) note that extraposition obviates Condition C effects, but strik<strong>in</strong>gly,<br />

this obviation is restricted to adjuncts; see (8). This recalls <strong>the</strong> pattern <strong>in</strong> (4) discussed<br />

by Lebeaux. S<strong>in</strong>ce our outer adjunction and counter-cyclic adjunction are <strong>in</strong>stances of <strong>the</strong> same<br />

mechanism (differ<strong>in</strong>g only <strong>in</strong> questions of l<strong>in</strong>earisation), this is as we would expect.<br />

(8) a. * I gave himi an argument [that supports Johni’s <strong>the</strong>ory] yesterday<br />

b. I gave himi an argument yesterday [that supports Johni’s <strong>the</strong>ory]<br />

c. * I gave him an argument yesterday [that this sentence supports John’s <strong>the</strong>ory]<br />

This <strong>in</strong>teracts <strong>in</strong> just <strong>the</strong> expected way with possibility of outer adjunction to non-base positions<br />

as <strong>in</strong> (7): extraposition from moved objects can obviate Condition C effects even <strong>in</strong> cases<br />

<strong>in</strong>duced by a subject b<strong>in</strong>der, as shown <strong>in</strong> (9) (from Culicover and Rochemont (1990)).<br />

(9) a. * Hei <strong>in</strong>vited several girls to <strong>the</strong> party that Johni dated <strong>in</strong> high school.<br />

b. How many girls did hei <strong>in</strong>vite to <strong>the</strong> party that Johni dated <strong>in</strong> high school.<br />

However, our implementation of counter-cyclic adjunction differs from Lebeaux’s <strong>in</strong> a crucial<br />

respect: it correctly predicts that anti-reconstruction should be sensitive to <strong>the</strong> depth of<br />

embedd<strong>in</strong>g <strong>in</strong> a manner similar to extraposition, as <strong>in</strong> (10) (Landau, 2007).<br />

(10) a. Food that Maryi cooks, shei knows I would never eat.<br />

b. * Eat food that Maryi cooks, shei knows I never would. (cf. (6a))<br />

Note f<strong>in</strong>ally that we have considered only right-adjuncts; <strong>the</strong>re should <strong>in</strong> pr<strong>in</strong>ciple be similar<br />

phenomena with left-adjuncts. We will suggest that floated quantifiers constitute such a case.<br />

Balt<strong>in</strong>, M. (1981). Strict bound<strong>in</strong>g. In Baker, C. L. and McCarthy, J., editors, The Logical Problem of Language<br />

Acquisition. MIT Press.<br />

Culicover, P. W. and Rochemont, M. S. (1990). Extraposition and <strong>the</strong> complement pr<strong>in</strong>ciple. L<strong>in</strong>guistic Inquiry.<br />

Fox, D. and Nissenbaum, J. (1999). Extraposition and Scope: A case for overt QR. In Proceed<strong>in</strong>gs of WCCFL 18.<br />

Hunter, T. (2010). Relat<strong>in</strong>g Movement and Adjunction <strong>in</strong> Syntax and Semantics. PhD <strong>the</strong>sis, U. of Maryland.<br />

Landau, I. (2007). Constra<strong>in</strong>ts on Partial VP-front<strong>in</strong>g. Syntax, 10(2):127–164.<br />

Lebeaux, D. (1988). Language acquisition and <strong>the</strong> form of <strong>the</strong> grammar. PhD <strong>the</strong>sis, UMass Amherst.


Laura Kal<strong>in</strong> (UCLA) & Coppe van Urk (MIT)<br />

A novel aspect split <strong>in</strong> Senaya<br />

Summary: This paper adds to <strong>the</strong> typology of aspect-based case/agreement splits by present<strong>in</strong>g<br />

<strong>the</strong> unusual system of <strong>the</strong> Neo-Aramaic dialect of Senaya. In Senaya, a nom<strong>in</strong>ativeaccusative<br />

system <strong>in</strong> <strong>the</strong> imperfective alternates with what resembles an antipassive <strong>in</strong> <strong>the</strong><br />

perfective. We argue that this system can be derived from <strong>the</strong> assumption that imperfective<br />

aspect <strong>in</strong>troduces additional structure, developed <strong>in</strong> recent work on split ergativity (Laka<br />

2006; Coon 2010; Coon & Prem<strong>in</strong>ger 2011). Specifically, we argue that <strong>in</strong> Senaya imperfective<br />

aspect <strong>in</strong>troduces an additional case assigner not present <strong>in</strong> <strong>the</strong> perfective. We extend<br />

our analysis to o<strong>the</strong>r dialects of Neo-Aramaic <strong>in</strong> which <strong>the</strong> addition of <strong>the</strong> same φ-probe <strong>in</strong><br />

<strong>the</strong> imperfective results <strong>in</strong> a complete reversal of agreement.<br />

The data: Most nor<strong>the</strong>astern dialects of Neo-Aramaic display aspect-based split ergativity,<br />

with an ergative agreement alignment <strong>in</strong> <strong>the</strong> perfective but an accusative one <strong>in</strong> <strong>the</strong><br />

imperfective (Doron & Khan 2010). In Senaya, however, this aspect-based split surfaces<br />

<strong>in</strong> an unusual way. As <strong>in</strong> o<strong>the</strong>r Neo-Aramaic languages, <strong>the</strong>re are two sets of agreement<br />

morphemes — <strong>the</strong> so-called L-suffixes and S-suffixes. In <strong>the</strong> imperfective, an S-suffix on <strong>the</strong><br />

verb references <strong>the</strong> subject (1a-b), while def<strong>in</strong>ite objects are marked by an L-suffix (1b):<br />

(1)a. Axnii damx-ox. b. Aanii xazy-ii-lan<br />

we sleep.impf-1pl.S <strong>the</strong>y see.impf-3pl.S-1pl.L<br />

‘We sleep.’<br />

‘They see us.’<br />

In <strong>the</strong> perfective, however, someth<strong>in</strong>g closer to an antipassive is found, as <strong>the</strong> examples <strong>in</strong><br />

(2a-b) illustrate. The subject is marked on <strong>the</strong> verb with an L-suffix (<strong>the</strong> object marker of<br />

<strong>the</strong> imperfective), while objects cannot be marked and are <strong>the</strong>refore obligatorily <strong>in</strong>def<strong>in</strong>ite:<br />

(2)a. Axnii dmex-lan.<br />

b. Axnii xa yaala xzee-lan.<br />

we sleep.perf-1pl.L we one child see.perf-1pl.L<br />

‘We slept.’<br />

‘We saw a child.’<br />

Imperfective aspect <strong>in</strong>volves a locative head: Dixon (1994) observes that if a system<br />

has an aspect-based split, it is <strong>the</strong> perfective side of <strong>the</strong> split that displays an ergative<br />

pattern. To account for this generalization, it has been proposed that imperfective and/or<br />

progressive aspect may <strong>in</strong>troduce complex structure that disturbs <strong>the</strong> usual mechanis<strong>ms</strong> of<br />

case assignment, ei<strong>the</strong>r <strong>in</strong> <strong>the</strong> form of a biclausal structure (Laka 2006; Coon 2010) or an<br />

additional phase boundary (Coon & Prem<strong>in</strong>ger 2011). Coon (2010) relates this to work<br />

by Demirdache & Uribe-Etxebarria (2000, 2007), who propose that aspect is encoded us<strong>in</strong>g<br />

spatiotemporal relations. They argue that nonperfective aspects make use of prepositional,<br />

locative relations, such as WITHIN, BEFORE, and AFTER; perfective aspect, on <strong>the</strong> o<strong>the</strong>r<br />

hand, cannot be expressed this way, aris<strong>in</strong>g <strong>in</strong>stead as a default when Asp is empty. As such,<br />

nonperfective aspects <strong>in</strong>volve a locative predicate absent from perfective.<br />

The proposal: We assume <strong>the</strong> <strong>the</strong>ory of aspect-based splits outl<strong>in</strong>ed above and propose<br />

that, <strong>in</strong> Senaya, <strong>the</strong> prepositional locative head selected by imperfective Asp <strong>in</strong>troduces<br />

a φ-probe, whose reflex <strong>in</strong> <strong>the</strong> morphology is an S-suffix. This P head agrees<br />

with and assigns case to <strong>the</strong> subject. As a result, <strong>the</strong> ord<strong>in</strong>ary case/agreement pattern is<br />

disturbed, s<strong>in</strong>ce it enables <strong>the</strong> φ-probe on T (a morphological L-suffix) to license a def<strong>in</strong>ite<br />

object. In addition, we assume that, aside from T and P under Asp, <strong>the</strong>re is no locus of<br />

agreement or case assigner <strong>in</strong> <strong>the</strong> verbal sp<strong>in</strong>e <strong>in</strong> Senaya. As such, <strong>in</strong> <strong>the</strong> perfective, only<br />

<strong>in</strong>def<strong>in</strong>ite objects are licit, s<strong>in</strong>ce <strong>in</strong>def<strong>in</strong>ite objects may pseudo-<strong>in</strong>corporate <strong>in</strong>to <strong>the</strong> verb as<br />

NPs and hence do not require case (Dayal 2011, Massam 2001).


Derivations: Perfective (<strong>in</strong>)transitive(3): NolocativePhead; Tprobes<strong>the</strong>subject, valu<strong>in</strong>g<br />

case/agr., seen as an L-suffix. No case/agr. locus is available to license a def<strong>in</strong>ite object.<br />

Imperfective <strong>in</strong>transitive (4): The P head under Asp probes and f<strong>in</strong>ds <strong>the</strong> subject, valu<strong>in</strong>g<br />

case/agr., seen as an S-suffix. T probes but f<strong>in</strong>ds no active argument. Imperfective transitive<br />

(4): Same as above, but T probes <strong>the</strong> object; agreement with T spells out as an L-suffix.<br />

(3)<br />

(4)<br />

T<br />

[nom]<br />

L-suffix<br />

DP subj<br />

v<br />

V (NP <strong>in</strong>def obj )<br />

T<br />

[nom]<br />

L-suffix Asp<br />

P<br />

[loc]<br />

S-suffix<br />

DP subj<br />

v V (DPobj )<br />

Extension to agreement reversal dialects: The analysis outl<strong>in</strong>ed here can be extended<br />

straightforwardly to so-called “extended ergative” dialects of Neo-Aramaic (Doron & Khan<br />

2010), e.g., Christian Barwar. In <strong>the</strong>se dialects, <strong>the</strong> perfective base can host agreement with<br />

an object, an S-suffix, result<strong>in</strong>g <strong>in</strong> a complete reversal of agreement from <strong>the</strong> imperfective:<br />

(5)a. Tpít-le<br />

sneeze.perf-3<strong>ms</strong>.L<br />

‘He sneezed.’<br />

b. Qtil-á-le.<br />

kill.perf-3fs.S-3<strong>ms</strong>.L<br />

‘He killed her.’<br />

c. Mtăm@zz-a.<br />

clean.impf-3fs.S<br />

‘She cleans.’<br />

d. Qatl-á-le.<br />

kill.impf-3fs.S-3<strong>ms</strong>.L<br />

‘She kills him.’ (Khan 2008)<br />

The morpheme -le (3<strong>ms</strong>) marks <strong>the</strong> nom<strong>in</strong>ative argument <strong>in</strong> <strong>the</strong> perfective, (5a,b), but<br />

<strong>the</strong> accusative one <strong>in</strong> <strong>the</strong> imperfective, (5d). The morpheme -a (3fs) marks <strong>the</strong> accusative<br />

argument <strong>in</strong> <strong>the</strong> perfective, (5b), but <strong>the</strong> nom<strong>in</strong>ative one <strong>in</strong> <strong>the</strong> imperfective, (5c,d).<br />

The difference between <strong>the</strong>se dialects and Senaya, we argue, is <strong>the</strong> mechanism underly<strong>in</strong>g<br />

L-suffixes. In Senaya, L-suffixes result from φ-agreement, evidenced by <strong>the</strong> fact that <strong>the</strong>y<br />

trigger stress shift, as true affixes do <strong>in</strong> Senaya. In C. Barwar and similar dialects, however,<br />

L-suffixes are clitics (Doron & Khan 2010). Follow<strong>in</strong>g Anagnostopoulou (2003) and Rezac<br />

(2011) among o<strong>the</strong>rs, we assume: (i) φ-probes split <strong>in</strong>to number and person, (ii) person<br />

probes before number, and (iii) clitic-doubled arguments are ignored by subsequent probes.<br />

Ifweposit, <strong>the</strong>n, that<strong>the</strong>personprobeonT<strong>in</strong>extendedergativedialects<strong>in</strong>discrim<strong>in</strong>ately<br />

triggers clitic-doubl<strong>in</strong>g with <strong>the</strong> first argument encountered, <strong>the</strong>n <strong>the</strong> rest of <strong>the</strong> differences<br />

between Senaya and <strong>the</strong>se dialects fall out naturally. In <strong>the</strong> perfective, T licenses both<br />

<strong>the</strong> subject, by means of its person probe (⇒ clitic-doubl<strong>in</strong>g), and <strong>the</strong> object, by means<br />

of its number probe (⇒ agreement). In <strong>the</strong> imperfective, <strong>the</strong> φ-probe on P licenses <strong>the</strong><br />

subject, such that it is <strong>the</strong> object that gets clitic-doubled by person on T. Importantly,<br />

this account expla<strong>in</strong>s <strong>the</strong> fact that <strong>the</strong> object is limited to third person <strong>in</strong> <strong>the</strong> perfective:<br />

this is a standard Person Case Constra<strong>in</strong>t (PCC) effect (Bonet 1991), and <strong>the</strong> above is<br />

straightforwardlycompatiblewi<strong>the</strong>xist<strong>in</strong>g accountsof<strong>the</strong>PCC(e.g., BéjarandRezac2003).<br />

Selected references: Coon, J.2010. Complementation <strong>in</strong> Chol: A <strong>the</strong>ory of split ergativity.<br />

Doctoral dissertation, MIT. – Doron, E., & G. Khan. 2010. The typology of morphological<br />

ergativity <strong>in</strong> Neo-Aramaic. <strong>ms</strong>. – Rezac, M. 2011. Phi-features and <strong>the</strong> modular architecture<br />

of language. Dordrecht: Spr<strong>in</strong>ger.


Laura Kertz (Brown)<br />

Referential Process<strong>in</strong>g Influences <strong>the</strong> Resolution of Verbal Anaphors<br />

Interest <strong>in</strong> verb phrase ellipsis has rebounded <strong>in</strong> recent years, as various proposals have<br />

sought to ref<strong>in</strong>e predictions from l<strong>in</strong>guistic <strong>the</strong>ory by <strong>in</strong>troduc<strong>in</strong>g process<strong>in</strong>g-based assumptions.<br />

The long-stand<strong>in</strong>g Surface/Deep Hypo<strong>the</strong>sis of Hankamer & Sag (1976) holds that verb phrase<br />

ellipsis (like o<strong>the</strong>r surface anaphors) is governed by a parallelism constra<strong>in</strong>t which does not<br />

apply to deep/proform anaphora. Under this approach, parallelism is implicated <strong>in</strong> both <strong>the</strong><br />

licens<strong>in</strong>g and <strong>in</strong>terpretation of elliptical anaphors. It has been demonstrated that non-parallel verb<br />

phrase ellipses, for example, take longer to read and are rated less acceptable than comparable<br />

parallel structures. It is well-known, however, that <strong>the</strong> parallelism effect under ellipsis is<br />

unreliable and that parallelism effects are observed even for deep anaphors.<br />

Moreover, <strong>the</strong>re is scant evidence to demonstrate that parallelism actually constra<strong>in</strong>s possible<br />

<strong>in</strong>terpretations for an ellipsis. Ra<strong>the</strong>r, it is <strong>the</strong> case that <strong>the</strong> auxiliary verb govern<strong>in</strong>g <strong>the</strong> ellipsis<br />

site can force ei<strong>the</strong>r a parallel or a non-parallel <strong>in</strong>terpretation, as <strong>in</strong> (1).<br />

(1) The accident was <strong>in</strong>vestigated by <strong>the</strong> police, and <strong>the</strong> <strong>in</strong>surance company did/was too.<br />

Indeed Garnham & Oakhill (1987) show that when <strong>the</strong> auxiliary is ambiguous, <strong>the</strong> preference for<br />

a parallel <strong>in</strong>terpretation competes with a preference for choos<strong>in</strong>g <strong>the</strong> most plausible antecedent.<br />

If it is not <strong>the</strong> case that parallelism constra<strong>in</strong>s possible <strong>in</strong>terpretations, <strong>the</strong> question emerges:<br />

What role does parallelism play <strong>in</strong> process<strong>in</strong>g ellipsis?<br />

A recent proposal (Arregui et al 2006) holds that non-parallel antecedents <strong>in</strong>duce extra<br />

process<strong>in</strong>g costs associated with syntactic repair at <strong>the</strong> ellipsis site. Similarly, Kim et al (2010)<br />

argue that parallelism effects arise as a function of pars<strong>in</strong>g heuristics that favor canonical<br />

(parallel) VP antecedents. Both of <strong>the</strong>se proposals seek to reconcile a categorical licens<strong>in</strong>g<br />

mechanism with an <strong>in</strong>herently gradient performance model. An alternative view holds that<br />

parallelism effects under ellipsis are epiphenomenal. Kehler (2000, 2002), for example, argues<br />

that parallelism effects arise as an <strong>in</strong>teraction between <strong>the</strong> establishment of discourse coherence<br />

relations and processes support<strong>in</strong>g ellipsis resolution, while Kertz (2010) argues that parallelism<br />

effects follow from general <strong>in</strong>formation structural constra<strong>in</strong>ts on <strong>the</strong> alignment of contrastive<br />

arguments. Each of <strong>the</strong>se latter two proposals makes <strong>the</strong> specific prediction that parallelism<br />

effects arise <strong>in</strong> ellipses where <strong>the</strong> subject of <strong>the</strong> target clause is contrastive with a syntactically<br />

non-parallel argument <strong>in</strong> <strong>the</strong> antecedent clause.<br />

The current study tested this prediction by compar<strong>in</strong>g effects of <strong>the</strong> referential status of <strong>the</strong><br />

subject noun phrase preced<strong>in</strong>g an ellipsis on <strong>the</strong> process<strong>in</strong>g of that ellipsis. Stimuli used for <strong>the</strong><br />

study featured ei<strong>the</strong>r a lexical NP (<strong>the</strong> firefighters) which contrasted with an argument <strong>in</strong> <strong>the</strong><br />

preced<strong>in</strong>g clause or a pronoun (it) which was co-referent with an argument <strong>in</strong> <strong>the</strong> preced<strong>in</strong>g<br />

clause. The referential status of <strong>the</strong> pronoun was crossed with <strong>the</strong> form of <strong>the</strong> antecedent<br />

(parallel vs. non-parallel). A sample stimulus set is shown <strong>in</strong> (2)-(3).<br />

(2) The rangers didn’t <strong>in</strong>spect <strong>the</strong> campsite as thoroughly as<br />

<strong>the</strong> firefighters did. [lexical NP, parallel] / it could have been. [pronoun, non-parallel]<br />

(3) The campsite wasn’t <strong>in</strong>spected by <strong>the</strong> rangers as thoroughly as<br />

<strong>the</strong> firefighters did. [lexical NP, non-parallel] / it could have been. [pronoun, parallel]<br />

A parallel discourse relationship was made explicit by <strong>the</strong> use of an equative adverbial<br />

construction. Thus participants were not faced with <strong>the</strong> task of determ<strong>in</strong><strong>in</strong>g <strong>the</strong> coherence<br />

relation obta<strong>in</strong><strong>in</strong>g between <strong>the</strong> antecedent and <strong>the</strong> target. Interpretation of <strong>the</strong> ellipsis, however,


did require properly identify<strong>in</strong>g co-referent/contrastive arguments across <strong>the</strong> two clauses. For<br />

pronouns, coreference was unambiguously signaled by number mark<strong>in</strong>g; for lexical NPs, we<br />

predicted that identification of <strong>the</strong> <strong>in</strong>tended contrast between <strong>the</strong> firefighters and <strong>the</strong> rangers<br />

would be dependent on whe<strong>the</strong>r <strong>the</strong> rangers <strong>in</strong> <strong>the</strong> antecedent occurred <strong>in</strong> a parallel (subject)<br />

position. We fur<strong>the</strong>r predicted that successful identification of <strong>the</strong> <strong>in</strong>tended contrast <strong>in</strong> <strong>the</strong><br />

parallel condition would lead to facilitation at <strong>the</strong> auxiliary verb region, where ellipsis resolution<br />

is <strong>in</strong>itiated.<br />

Sixteen stimulus sets like (2)-(3) were tested <strong>in</strong> an offl<strong>in</strong>e magnitude estimation task (n=36).<br />

Stimuli were adapted to <strong>in</strong>clude a spill-over region follow<strong>in</strong>g <strong>the</strong> ellipsis for a self-paced read<strong>in</strong>g<br />

task (n=48). Statistical analysis was conducted on log-transformed ratios of stimulus to modulus<br />

<strong>in</strong> <strong>the</strong> ME task and on both raw and residual read<strong>in</strong>g times trimmed to 3sd by region for <strong>the</strong> SPR<br />

task. (Results for raw and residual analyses were identical.) For all analyses, a mixed effects<br />

analysis with forward model selection for <strong>in</strong>clusion of random effects was conducted.<br />

Offl<strong>in</strong>e results showed an <strong>in</strong>teraction where non-parallel structures were dispreferred<br />

compared to parallel structures, but that difference was greater with lexical NPs. Onl<strong>in</strong>e, no<br />

statistically reliable effects of parallelism were observed. At <strong>the</strong> subject NP region, a crossover<br />

<strong>in</strong>teraction was observed where lexical NPs were read more slowly <strong>in</strong> <strong>the</strong> parallel condition<br />

while pronouns were read more quickly. At <strong>the</strong> auxiliary verb region an effect of subject type<br />

only was observed. No reliable effects were observed <strong>in</strong> <strong>the</strong> post-ellipsis spill-over region.<br />

0.2<br />

0<br />

-0.2<br />

-0.4<br />

Offl<strong>in</strong>e<br />

Acceptability<br />

parallel non-parallel<br />

lexical NP pronoun<br />

550<br />

500<br />

450<br />

400<br />

350<br />

300<br />

<strong>the</strong> firefighters<br />

it<br />

<strong>Raw</strong> <strong>Read<strong>in</strong>g</strong> <strong>Times</strong> (<strong>ms</strong>)<br />

did<br />

could have been<br />

after <strong>the</strong><br />

after <strong>the</strong><br />

lexical NP<br />

parallel<br />

lexical NP<br />

non-parallel<br />

pronoun<br />

parallel<br />

pronoun<br />

non-parallel<br />

Consistent with f<strong>in</strong>d<strong>in</strong>gs <strong>in</strong>, e.g. Birch & Rayner (1997), we <strong>in</strong>terpret <strong>the</strong> <strong>in</strong>creased read<strong>in</strong>g times<br />

for <strong>the</strong> lexical NPs <strong>in</strong> <strong>the</strong> parallel condition as evidence that readers treated <strong>the</strong>se NPs as<br />

l<strong>in</strong>guistically focused constituents, i.e. readers recognized <strong>the</strong> <strong>in</strong>tended contrast between <strong>the</strong> two<br />

arguments. Identification of focus <strong>in</strong> this region had a downstream effect of reduc<strong>in</strong>g process<strong>in</strong>g<br />

time at <strong>the</strong> ellipsis. In <strong>the</strong> non-parallel, lexical NP condition, this facilitative effect was not<br />

observed, and <strong>in</strong> <strong>the</strong> pronoun conditions, no effects of parallelism were seen.<br />

These results thus confirm <strong>the</strong> prediction that parallelism <strong>in</strong>teracts with referential process<strong>in</strong>g<br />

for contrastive noun phrases, which can <strong>in</strong> turn <strong>in</strong>fluence verbal anaphor resolution. This<br />

<strong>in</strong>teraction was also seen to <strong>in</strong>fluence offl<strong>in</strong>e acceptability. While <strong>the</strong> results are not <strong>in</strong>compatible<br />

with models that would reta<strong>in</strong> a categorical licens<strong>in</strong>g mechanism for ellipsis, <strong>the</strong> results do show<br />

that parallelism constra<strong>in</strong>ts on process<strong>in</strong>g are weak/violable and crucially <strong>in</strong>teract <strong>in</strong> a<br />

predictable manner with o<strong>the</strong>r aspects of <strong>the</strong> discourse context.


Ezra Keshet (Michigan)<br />

Scopal Effects of Embedded Coherence Relations<br />

Researchers s<strong>in</strong>ce Hobbs (1979) have proposed that an important part of understand<strong>in</strong>g a mult-­‐<br />

sentence discourse is understand<strong>in</strong>g <strong>the</strong> (often unspoken) relationship between <strong>the</strong> various<br />

sentences (see also Kehler 2002 and Asher and Lascarides 2003). For <strong>in</strong>stance, <strong>the</strong> two<br />

sentences of (1) below are not understood as be<strong>in</strong>g <strong>in</strong>dependent of one ano<strong>the</strong>r. Instead,<br />

speakers tend to <strong>in</strong>fer a relationship between paraphraseable as <strong>in</strong> (2):<br />

(1) John ate <strong>the</strong> shellfish. He got horrible food poison<strong>in</strong>g.<br />

(2) John got horrible food poison<strong>in</strong>g because he ate <strong>the</strong> shellfish.<br />

Such connections between sentences <strong>in</strong> a discourse are known as coherence relations. In (1),<br />

<strong>the</strong> second sentence is understood to be <strong>the</strong> result of <strong>the</strong> first, and <strong>the</strong>refore <strong>the</strong> coherence<br />

relation <strong>in</strong> effect here is called <strong>the</strong> Result relation.<br />

This phenomenon immediately raises <strong>in</strong>terest<strong>in</strong>g questions about <strong>the</strong> <strong>in</strong>teraction between<br />

grammar and context: for <strong>in</strong>stance, what is <strong>the</strong> grammatical status of such coherence relations?<br />

Are <strong>the</strong>y due to purely pragmatic reason<strong>in</strong>g, or is <strong>the</strong>re some syntactic component responsible<br />

for such <strong>in</strong>ferences? Do <strong>the</strong>y only hold at <strong>the</strong> sentence level, or can <strong>the</strong>y affect sentence-­‐<br />

<strong>in</strong>ternal mean<strong>in</strong>g? Rohde, Levy, and Kehler (2011) take a stab at this last question, exam<strong>in</strong><strong>in</strong>g<br />

cases like (3), a s<strong>in</strong>gle sentence whose <strong>in</strong>ternal clauses seem to exhibit <strong>the</strong> same potential for<br />

coherence relations as <strong>the</strong> separate sentences of (1). Example (3) is usually understood as <strong>in</strong><br />

(4), thus exhibit<strong>in</strong>g an Explanation relation between <strong>the</strong> matrix clause and <strong>the</strong> relative clause.<br />

(3) John detests <strong>the</strong> coworkers who are arrogant and rude.<br />

(4) John detests <strong>the</strong> coworkers who are arrogant and rude because <strong>the</strong>y are arrogant and<br />

rude.<br />

Rohde et al (2011) show experimentally that such sentence-­‐<strong>in</strong>ternal coherence relations can<br />

<strong>in</strong>fluence <strong>the</strong> process<strong>in</strong>g of a purely syntactic phenomenon, namely where to attach a<br />

prepositional phrase. This suggests that whe<strong>the</strong>r coherence is pragmatic or <strong>in</strong> <strong>the</strong> grammar, it<br />

plays a role <strong>in</strong> sentence-­‐<strong>in</strong>ternal mean<strong>in</strong>g.<br />

This paper takes a step fur<strong>the</strong>r and suggests that coherence relations arise due to a covert<br />

syntactic operator, which can <strong>in</strong>teract with scopal elements such as quantifiers. Take (5), for<br />

<strong>in</strong>stance, which see<strong>ms</strong> to have a sentence-­‐<strong>in</strong>ternal Result relation between <strong>the</strong> relative clause<br />

<strong>in</strong> <strong>the</strong> subject and <strong>the</strong> matrix sentence:<br />

(5) Everyone who ate <strong>the</strong> shellfish got horrible food poison<strong>in</strong>g.<br />

Imag<strong>in</strong>e that party guests A, B, and C all ate <strong>the</strong> shellfish and got food poison<strong>in</strong>g. The crucial<br />

observation is that <strong>the</strong> coherence <strong>in</strong>fe������������������������������<br />

������������������������������������������������<br />

shellfish <strong>in</strong>gestion r���������������������������������������<br />

<strong>in</strong> ���������������������������������������������<br />

�����������������������������������������������<br />

thought to lead to his or her own poison<strong>in</strong>g. Some similar examples:


(6) Everyone who John detested was arrogant and rude. [Explanation]<br />

(7) No one who drank pomegranate juice regularly got cancer. [Result]<br />

(8) No one who John detests is ever nice to him. [Explanation]<br />

I propose to capture <strong>the</strong>se connections with a set of covert syntactic operators that presuppose<br />

a relationship between <strong>the</strong> event represented by <strong>the</strong> sentence <strong>the</strong>y appear <strong>in</strong> and a<br />

contextually previous event (assum<strong>in</strong>g that events may be states or dynamic events). For<br />

<strong>in</strong>stance, <strong>the</strong> structure I propose for (1) is given <strong>in</strong> (9), and a few relevant operator def<strong>in</strong>itions<br />

are given below:<br />

(9) [[John ate <strong>the</strong> shellfish] =e1] [[Result(e1) [He got food poison<strong>in</strong>g]] =e2]<br />

(10) ������������������������������������������<br />

(11) ����������������������������������<br />

����������������������������<br />

�������������������������������<br />

(12) �������������������������������������������<br />

����������������������������������������<br />

An example like (5) arises when such operators scope below a quantifier, as shown below:<br />

(13) [Everyone whox [[x ate <strong>the</strong> shellfish] =e1]]x [[Result(e1) [x got food poison<strong>in</strong>g]] =e2]]<br />

Here, <strong>the</strong> variable e1 is an e-­‐type pronoun rang<strong>in</strong>g over events where different <strong>in</strong>dividuals x eat<br />

shellfish. The Result operator that scopes below <strong>the</strong> quantification DP relates each such event<br />

to <strong>the</strong> event where <strong>the</strong> same <strong>in</strong>dividual x gets food poison<strong>in</strong>g. (I named all <strong>the</strong> <strong>in</strong>dividual<br />

variables <strong>in</strong> (13) x s<strong>in</strong>ce <strong>the</strong>y all end up pick<strong>in</strong>g out <strong>the</strong> same <strong>in</strong>dividual.)<br />

If such operators exist, <strong>the</strong>y should <strong>in</strong> pr<strong>in</strong>ciple give rise to scope ambiguities, and this is exactly<br />

what I propose happens <strong>in</strong> (14):<br />

(14) John is arrogant and rude. Everyone he works with detests him.<br />

One read<strong>in</strong>g of (14) is that each <strong>in</strong>dividual coworker happens to detest John for <strong>the</strong> same<br />

reason � namely that he is arrogant and rude. Co-­‐worker A detests him for this reason, co-­‐<br />

worker B detests him for this reason, etc. I argue that this read<strong>in</strong>g arises when a Result<br />

operator scopes below <strong>the</strong> quantificational DP Everyone he works with analogously to structure<br />

(13). However, <strong>the</strong>re is ano<strong>the</strong>r read<strong>in</strong>g of (14) � one where <strong>the</strong> fact that John is arrogant and<br />

rude expla<strong>in</strong>s why every one of his co-­‐workers detests him. Perhaps co-­‐workers A and B detest<br />

arrogant people, and C and D detest rude people. It is because John is arrogant and rude that<br />

<strong>the</strong> whole office detests him. I argue that this read<strong>in</strong>g arises when a Result operator scopes<br />

above <strong>the</strong> quantificational DP.<br />

This new evidence suggests a close connection between coherence relations and grammar.<br />

After mak<strong>in</strong>g this argument, <strong>the</strong> paper concludes by exam<strong>in</strong><strong>in</strong>g <strong>the</strong> <strong>in</strong>teraction of such<br />

coherence operators with o<strong>the</strong>r scopal operators, such as modals, and by look<strong>in</strong>g at <strong>the</strong> effect<br />

this analysis might have on a <strong>the</strong>ory of anaphora.


Arnout Koornneef & Eric Reuland (Utrecht)<br />

Grammar and Process<strong>in</strong>g Economy<br />

One of <strong>the</strong> <strong>in</strong>trigu<strong>in</strong>g issues <strong>in</strong> <strong>the</strong> relation between grammar and <strong>the</strong> process<strong>in</strong>g<br />

system concerns <strong>the</strong> resolution of anaphoric dependencies. What is <strong>the</strong> role of <strong>the</strong><br />

various sub-components of <strong>the</strong> language system (syntax, logical form, discourse) and<br />

how can <strong>the</strong>ir roles be dist<strong>in</strong>guished as <strong>the</strong> <strong>in</strong>terpretation of an utterance develops<br />

through time? We know that at least two fundamentally different types of <strong>in</strong>terpretive<br />

processes must be dist<strong>in</strong>guished <strong>in</strong> anaphora resolution (leav<strong>in</strong>g aside lexical and<br />

strictly syntactic encod<strong>in</strong>g): Coreference (cont<strong>in</strong>gently assign<strong>in</strong>g <strong>the</strong> same value to<br />

two expressions) and Variable B<strong>in</strong>d<strong>in</strong>g (VB) (Heim 1982, Re<strong>in</strong>hart 1983, Reuland<br />

2001, 2011). VB is encoded <strong>in</strong> <strong>the</strong> grammatical system and sensitive to particular<br />

structural conditions (e.g. c-command), but <strong>the</strong> former is freely available, modulo a<br />

regulatory pr<strong>in</strong>ciple such as Re<strong>in</strong>hart and Grodz<strong>in</strong>sky (1993)’s Rule I, or Reuland<br />

(2011)’s Rejection is F<strong>in</strong>al. The operative pr<strong>in</strong>ciple here is an economy preference for<br />

b<strong>in</strong>d<strong>in</strong>g over co-reference. For details about <strong>the</strong> economy metric, based on <strong>the</strong> idea<br />

that cross-modular steps carry a cost, see Reuland (2001, 2011).<br />

This ties <strong>in</strong> with a debate <strong>in</strong> process<strong>in</strong>g studies. A consistent f<strong>in</strong>d<strong>in</strong>g <strong>in</strong> offl<strong>in</strong>e<br />

studies on <strong>the</strong> <strong>in</strong>terpretation of ambiguous VP-ellipses is that bound-variable (BV)<br />

based <strong>in</strong>terpretations are preferred over coreferential <strong>in</strong>terpretations, see Frazier and<br />

Clifton 2000 (F&C) for an overview.<br />

On <strong>the</strong> basis of such results – see also Avrut<strong>in</strong> (1999), and Vasic´ et al. (2006)<br />

for a more recent result <strong>in</strong> agrammatic aphasics - F&C formulate and test, us<strong>in</strong>g selfpaced<br />

read<strong>in</strong>g and questionnaires, <strong>the</strong> hypo<strong>the</strong>sis <strong>in</strong> (1):<br />

(1) LF only/first hypo<strong>the</strong>sis: Bound-variable <strong>in</strong>terpretations are preferred because<br />

<strong>the</strong> perceiver need only consult <strong>the</strong> LF representation (not <strong>the</strong> discourse<br />

representation) <strong>in</strong> order to identify <strong>the</strong> bound-variable analysis of <strong>the</strong> sentence.<br />

F&C f<strong>in</strong>d two proble<strong>ms</strong> with (1): i. With some qualifications, a BV-preference also<br />

appears to obta<strong>in</strong> across sentence boundaries, as <strong>in</strong> (2), which accord<strong>in</strong>g to <strong>the</strong>m is<br />

<strong>in</strong>compatible with <strong>the</strong> conception of LF. ii. They f<strong>in</strong>d a preference for a co-referential<br />

<strong>in</strong>terpretation <strong>in</strong> only-sentences as <strong>in</strong> (3b).<br />

(2) The clown loves his cat. The acrobat does too.<br />

(3) a. Only Alfred th<strong>in</strong>ks (that) he is a good cook.<br />

b. Only Alfred th<strong>in</strong>ks that Alfred is a good cook (coreference)<br />

c. The only person who th<strong>in</strong>ks of hi<strong>ms</strong>elf as a good cook is Alfred. (BV).<br />

As we will briefly show, problem i. is resolved by us<strong>in</strong>g an explicit <strong>the</strong>ory of ellipsis,<br />

as <strong>in</strong> Elbourne (2008). Our paper <strong>the</strong>refore focuses on ii. We will consider not only<br />

<strong>the</strong> <strong>in</strong>terpretation <strong>the</strong> processor assigns, but also <strong>the</strong> process lead<strong>in</strong>g to this<br />

<strong>in</strong>terpretation. For <strong>the</strong> latter one needs on-l<strong>in</strong>e evidence, which so far is almost absent.<br />

In a comb<strong>in</strong>ation of two experiments (i. a questionnaire; and ii. an eyetrack<strong>in</strong>g<br />

experiment) we show that <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g preference is not solely an offl<strong>in</strong>e<br />

phenomenon, but can be detected <strong>in</strong> real-time measures of language comprehension.<br />

Contra F&C we found that <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g preference of <strong>the</strong> language processor shows up<br />

<strong>in</strong> ambiguous structures with <strong>the</strong> only-operator as well.<br />

In both experiments short stories were presented conta<strong>in</strong><strong>in</strong>g a critical sentence that<br />

was ambiguous between a sloppy and a strict read<strong>in</strong>g. There were 4 categories:<br />

[Sloppy-bias, only-operator], [Strict-bias, only-operator], [Sloppy-bias, ellipsis],<br />

[Strict-bias, ellipsis]. For space reasons we only illustrate <strong>the</strong> type of story we used:<br />

(4) Lisa and Anouk love <strong>the</strong> music channel MTV. They were very happy when <strong>the</strong>y<br />

were selected for <strong>the</strong> show ‘‘Pimp My Room,’’ <strong>in</strong> which <strong>the</strong>ir roo<strong>ms</strong> were<br />

redecorated. Only Lisa th<strong>in</strong>ks that her pimped room has a touch of class. Oh<br />

well, to each <strong>the</strong>ir owntaste. (sloppy bias)


The aim of experiment i. was two-fold: a. to pretest materials for experiment ii; and b.<br />

to provide a prelim<strong>in</strong>ary check on <strong>the</strong> solidity of F&C’s questionnaire f<strong>in</strong>d<strong>in</strong>gs.<br />

There were 20 participants. The materials were 36 short stories. The stories were<br />

directly followed by two <strong>in</strong>ferences that were consistent with ei<strong>the</strong>r a sloppy or a<br />

strict <strong>in</strong>terpretation. The participants had to rate both <strong>in</strong>ferences, and <strong>in</strong>dicate which<br />

one <strong>the</strong>y preferred. F<strong>in</strong>ally, participants rated <strong>the</strong> stories on difficulty and plausibility.<br />

Overall, <strong>the</strong> results revealed that it was easier to create a bias towards a sloppy<br />

read<strong>in</strong>g, which is consistent with <strong>the</strong> hypo<strong>the</strong>sis that <strong>in</strong> <strong>the</strong> case of ambiguity <strong>the</strong><br />

parser prefers a bound-variable dependency over a coreferential dependency.<br />

Fur<strong>the</strong>rmore, planned pairwise comparisons showed that <strong>the</strong> proper <strong>in</strong>ference of<br />

strict-biased stories <strong>in</strong> <strong>the</strong> only-condition was relatively difficult to obta<strong>in</strong>. In addition,<br />

<strong>the</strong>se stories were also rated more difficult and less plausible than strict biased stories<br />

<strong>in</strong> <strong>the</strong> ellipsis condition. In all, <strong>the</strong>se rat<strong>in</strong>gs suggest that <strong>in</strong> <strong>the</strong> stories with <strong>the</strong> onlyoperator,<br />

<strong>the</strong> strict read<strong>in</strong>g is less accessible, which is <strong>in</strong>consistent with an account <strong>in</strong><br />

which <strong>the</strong> preference for a bound-variable dependency is stronger <strong>in</strong> ambiguous<br />

ellipses than <strong>in</strong> ambiguous only-structures.<br />

In <strong>the</strong> eye-track<strong>in</strong>g experiment <strong>the</strong> pretested stimuli were presented to 32<br />

healthy adults who did not participate <strong>in</strong> experiment i. Our model predicts that readers<br />

prefer a BV <strong>in</strong>terpretation regardless of <strong>the</strong> content of <strong>the</strong> preced<strong>in</strong>g discourse. That<br />

is, <strong>the</strong>y <strong>in</strong>itially assign a sloppy <strong>in</strong>terpretation, which requires back-track<strong>in</strong>g once all<br />

discourse <strong>in</strong>formation is processed.<br />

In <strong>the</strong> ellipsis region (note, though, that this region is absent <strong>in</strong> <strong>the</strong> onlyconditions)<br />

<strong>the</strong> regression-path measure revealed a clear advantage for sloppy<br />

<strong>in</strong>terpretations, show<strong>in</strong>g reliably that <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g preference is not solely an offl<strong>in</strong>e<br />

phenomenon, but is also observable <strong>in</strong> onl<strong>in</strong>e measures of language comprehension.<br />

The results are consistent with <strong>the</strong> claim that <strong>in</strong> strict-biased stories readers<br />

experience proble<strong>ms</strong>, because <strong>the</strong>y <strong>in</strong>itially try to <strong>in</strong>tegrate a sloppy <strong>in</strong>terpretation <strong>in</strong>to<br />

a strict context. Before mov<strong>in</strong>g on, <strong>the</strong>y have to change <strong>the</strong>ir <strong>in</strong>terpretation to resolve<br />

this <strong>in</strong>consistency and longer regressions-path durations are <strong>the</strong>refore expected.<br />

The second sentence region revealed a clear contrast between first-pass and<br />

second-pass eye movement measures. The second-pass durations were almost twice<br />

as long for <strong>the</strong> strict-biased stories. Apparently, readers need to reexam<strong>in</strong>e <strong>the</strong> second<br />

sentence (conta<strong>in</strong><strong>in</strong>g <strong>the</strong> critical manipulation) longer if <strong>the</strong>y process a strict-biased<br />

story. This was true for both ellipses and only-structures. Hence, <strong>the</strong> preference for<br />

sloppy <strong>in</strong>terpretations reflects a more general property of <strong>the</strong> language processor.<br />

These results give a clearer picture of what happens while readers encounter<br />

<strong>the</strong> sloppy-strict ambiguity, and support our economy based model. There were some<br />

o<strong>the</strong>r <strong>in</strong>trigu<strong>in</strong>g f<strong>in</strong>d<strong>in</strong>gs, shedd<strong>in</strong>g light on <strong>the</strong> details of <strong>the</strong> <strong>in</strong>terpretation process<br />

which we will discuss if time permits.<br />

Selected References<br />

Avrut<strong>in</strong>, S. 1999. Development of <strong>the</strong> syntax-discourse <strong>in</strong>terface. Dordrecht: Kluwer<br />

Elbourne, P. 2008. Ellipsis Sites as Def<strong>in</strong>ite Descriptions. L<strong>in</strong>guistic Inquiry 39, 191-<br />

220<br />

Frazier, L., & Clifton, C. 2000. On bound variable <strong>in</strong>terpretations: The LF-only<br />

hypo<strong>the</strong>sis. Journal of Psychol<strong>in</strong>guistic Research, 29, 125–139<br />

Koornneef, A.W. 2008. Eye-catch<strong>in</strong>g Anaphora. Utrecht: LOT dissertation<br />

Reuland, E. 2011. Anaphora and Language Design. MIT Press<br />

Vasic´, N., Avrut<strong>in</strong>, S., & Ruigendijk, E. 2006. Interpretation of pronouns <strong>in</strong> VPellipsis<br />

constructions <strong>in</strong> Dutch Broca’s and Wernicke’s aphasia. Bra<strong>in</strong> and Language,<br />

96, 191–206.


Hadas Kotek (MIT)<br />

Wh-Front<strong>in</strong>g <strong>in</strong> a Two-Probe System<br />

The study of wh-movement has dist<strong>in</strong>guished among several types of wh-front<strong>in</strong>g languages that<br />

permit dist<strong>in</strong>ct patterns of overt and covert movement, <strong>in</strong>stantiated for example by <strong>the</strong> Slavic<br />

languages, English and German (Beck 2006). In this talk I extend <strong>the</strong> cross-l<strong>in</strong>guistic typology of<br />

multiple questions by argu<strong>in</strong>g that Hebrew <strong>in</strong>stantiates a new k<strong>in</strong>d of wh-front<strong>in</strong>g language, unlike<br />

any that are presently discussed <strong>in</strong> <strong>the</strong> literature. I will show that Hebrew dist<strong>in</strong>guishes between two<br />

k<strong>in</strong>ds of wh-phrases—those that are headed by a wh-word (wh-headed phrases: what, who, [ DP which X],<br />

where ...) and those that conta<strong>in</strong> a wh-word but are headed by some o<strong>the</strong>r element (wh-conta<strong>in</strong><strong>in</strong>g<br />

phrases: [ NP N of wh], [ PP P wh] …). Wh-headed phrases are privileged <strong>in</strong> that <strong>the</strong>y can be targeted by<br />

Agree/Attract operations that ignore mere wh-conta<strong>in</strong><strong>in</strong>g phrases <strong>in</strong> <strong>the</strong> same structure.<br />

To establish <strong>the</strong> privileged status of wh-headed phrases, I exam<strong>in</strong>e <strong>the</strong> behavior of multiple questions<br />

<strong>in</strong> Hebrew. I first observe that, as with D-l<strong>in</strong>ked wh-phrases <strong>in</strong> English, <strong>the</strong>re is a correlation <strong>in</strong><br />

Hebrew between apparent superiority violations and <strong>in</strong>tervention effects (Pesetsky 2000). I add a<br />

third correlate to this f<strong>in</strong>d<strong>in</strong>g, concern<strong>in</strong>g possible read<strong>in</strong>gs of <strong>the</strong> questions:<br />

(1) a. Superiority-Obey<strong>in</strong>g Questions (SOQs) never exhibit <strong>in</strong>tervention effects; <strong>the</strong>y may have<br />

s<strong>in</strong>gle-pair answers as well as pair-list answers.<br />

b. Superiority-Violat<strong>in</strong>g Questions (SVQs) are grammatical but <strong>the</strong>y are sensitive to Beck<br />

<strong>in</strong>tervention effects; <strong>the</strong>y may only have pair-list answers but not s<strong>in</strong>gle-pair answers.<br />

I will show that <strong>the</strong>se facts can be captured with<strong>in</strong> <strong>the</strong> <strong>the</strong>ory of Q-particles under standard<br />

assumptions about superiority and <strong>in</strong>tervention effects (Pesetsky 2000, Beck 2006, Cable 2010). In<br />

<strong>the</strong>se <strong>the</strong>ories, <strong>in</strong> <strong>the</strong> derivation of SOQs, all wh-phrases move to Spec,CP by LF. In <strong>the</strong> derivation<br />

of SVQs <strong>the</strong> higher wh-phrase is left <strong>in</strong> situ, allow<strong>in</strong>g <strong>the</strong> lower wh-phrase to be <strong>the</strong> first element<br />

attracted to Spec,CP. Intervention effects arise when a focus-sensitive element occurs between <strong>the</strong><br />

wh-<strong>in</strong>-situ and <strong>the</strong> head with which it Agrees. I adopt an economy pr<strong>in</strong>ciple on LFs based on <strong>the</strong><br />

work of Fox (2000) which, toge<strong>the</strong>r with <strong>the</strong> work of Dayal (2002) on <strong>the</strong> presuppositions of s<strong>in</strong>glepair<br />

and pair-list answers, can expla<strong>in</strong> <strong>the</strong> distribution of read<strong>in</strong>gs.<br />

The evidence for <strong>the</strong> existence of two k<strong>in</strong>ds of wh-phrases <strong>in</strong> Hebrew comes from <strong>the</strong> same three<br />

sources: superiority, <strong>in</strong>tervention effects and read<strong>in</strong>gs of <strong>the</strong> questions. We observe <strong>the</strong> special status<br />

of wh-headed phrases when a wh-headed phrase occurs structurally lower <strong>in</strong> a question than a whconta<strong>in</strong><strong>in</strong>g<br />

phrase. I argue that <strong>in</strong> that case, certa<strong>in</strong> operations target <strong>the</strong> lower wh-headed phrase and<br />

entirely overlook <strong>the</strong> presence of <strong>the</strong> higher wh-conta<strong>in</strong><strong>in</strong>g phrase. This allows <strong>the</strong> question to be<br />

derived from a structure <strong>in</strong> which no wh-phrase rema<strong>in</strong>s <strong>in</strong> situ at LF; <strong>the</strong> first operation <strong>in</strong> <strong>the</strong><br />

derivation is one that only targets wh-headed phrases and a subsequent operation is one that can<br />

target <strong>the</strong> (higher) wh-conta<strong>in</strong><strong>in</strong>g phrase. This derivation correctly predicts that such questions are<br />

not sensitive to Beck <strong>in</strong>tervention effects and have s<strong>in</strong>gle-pair (as well as pair-list) answers.<br />

One example of <strong>the</strong> privileged status of wh-headed phrases is illustrated by <strong>the</strong> m<strong>in</strong>imal pairs <strong>in</strong> (2ac).<br />

These example conta<strong>in</strong> <strong>the</strong> optional object marker et—analyzed here as a preposition—on <strong>the</strong><br />

direct object of read. When et is present, <strong>the</strong> phrase et ma is headed by et and is hence a wh-conta<strong>in</strong><strong>in</strong>g<br />

phrase; when et is absent, bare ma is a wh-headed phrase. Questions with <strong>the</strong> (based-generated)<br />

configuration [wh-conta<strong>in</strong><strong>in</strong>g phrase 1] ≫ [wh-conta<strong>in</strong><strong>in</strong>g phrase 2] behave as expected: <strong>the</strong> SOQ is<br />

grammatical, (2a), but <strong>the</strong> SVQ is ungrammatical when an <strong>in</strong>tervener (here: negation) is present,<br />

(2b). In <strong>the</strong> configuration [wh-conta<strong>in</strong><strong>in</strong>g phrase 1] ≫ [wh-headed phrase 2] we f<strong>in</strong>d that <strong>the</strong> SVQ is<br />

unexpectedly not sensitive to <strong>the</strong> presence of an <strong>in</strong>tervener, (2c), and that it can have a s<strong>in</strong>gle-pair<br />

answer as well as a pair-list answer, just like <strong>the</strong> superiority-obey<strong>in</strong>g question.<br />

(2) Unexpected lack of <strong>in</strong>tervention effect <strong>in</strong> superiority-violat<strong>in</strong>g question<br />

a. [et mi] ha-mora lo šixne’a [ likro (et) [ma] ]?<br />

OM who <strong>the</strong>-teacher neg persuaded to.read OM what


‘Who did <strong>the</strong> teacher not persuade to read what?’<br />

b. *[et ma] ha-mora lo šixne’a [et mi] [ likro ___ ]?<br />

OM what <strong>the</strong>-teacher neg persuaded OM who to.read<br />

c. ? [ma] ha-mora lo šixne’a [et mi] [ likro ___ ]?<br />

what <strong>the</strong>-teacher neg persuaded OM who to.read<br />

‘What did <strong>the</strong> teacher not persuade whom to read?’<br />

The sensitivity of <strong>in</strong>terrogative prob<strong>in</strong>g operations to material <strong>in</strong>side QP is unexpected <strong>in</strong> Cable's Q<strong>the</strong>ory<br />

framework, given <strong>the</strong> structure <strong>in</strong> (3a). I will propose two m<strong>in</strong>imal, yet significant, additions<br />

to <strong>the</strong> <strong>the</strong>ory. First, I adopt a f<strong>in</strong>e-gra<strong>in</strong>ed structure of QP <strong>in</strong> Hebrew-type wh-front<strong>in</strong>g languages: I<br />

propose that when Q is merged with XP, it immediately attracts <strong>the</strong> next lower head, X. After X<br />

head-moves to Q, <strong>the</strong>y both project and become co-heads of <strong>the</strong> result<strong>in</strong>g phrase. The derivation of<br />

wh-<strong>in</strong>-situ languages and of English/German-type wh-front<strong>in</strong>g languages proceeds as <strong>in</strong> Cable (2010):<br />

if X projects we derive a wh-<strong>in</strong>-situ language like Japanese or Korean. If Q projects we derive <strong>the</strong><br />

familiar English/German-type wh-front<strong>in</strong>g language <strong>in</strong> which <strong>the</strong> <strong>in</strong>ternal structure of QP is opaque<br />

to <strong>the</strong> outside derivation. If both X and Q project we derive a Hebrew-type wh-front<strong>in</strong>g language, <strong>in</strong><br />

which not only Q but also X is visible to outside prob<strong>in</strong>g. Crucially, we predict that a wh-word<br />

projects to <strong>the</strong> QP level when it is <strong>the</strong> head of XP but not when it is buried somewhere with<strong>in</strong> XP.<br />

We can thus dist<strong>in</strong>guish between wh-headed phrases (3bi) and wh-conta<strong>in</strong><strong>in</strong>g phrases (3bii).<br />

(3) a. proposal <strong>in</strong> Cable (2010) b. proposed new QP structure for Hebrew-type wh-front<strong>in</strong>g languages:<br />

for all wh-front<strong>in</strong>g languages (i) wh-headed phrases (ii) wh-conta<strong>in</strong><strong>in</strong>g phrases<br />

QwhP<br />

QP<br />

X+QP<br />

XP<br />

XP<br />

XP<br />

Q<br />

Q<br />

Q<br />

wh<br />

…<br />

X<br />

… wh …<br />

… wh …<br />

Moreover, I propose that our <strong>the</strong>ory must have a wh-probe <strong>in</strong> addition to <strong>the</strong> Q-probe. The Q-probe<br />

operates <strong>in</strong> <strong>the</strong> familiar way, target<strong>in</strong>g any k<strong>in</strong>d of wh-phrase. The wh-probe, on <strong>the</strong> o<strong>the</strong>r hand, only<br />

targets wh-headed phrases and cannot detect <strong>the</strong> presence of wh-conta<strong>in</strong><strong>in</strong>g phrases. This proposal, I<br />

argue, <strong>in</strong>corporates all of <strong>the</strong> advantages of Cable’s orig<strong>in</strong>al <strong>the</strong>ory but is empirically superior <strong>in</strong><br />

provid<strong>in</strong>g a natural explanation for an <strong>in</strong>tricate set of data that is o<strong>the</strong>rwise unexpla<strong>in</strong>ed. In addition,<br />

I suggest that this proposal is conceptually more appeal<strong>in</strong>g than <strong>the</strong> orig<strong>in</strong>al <strong>the</strong>ory s<strong>in</strong>ce it allows for<br />

all <strong>in</strong>terrogative phrases <strong>in</strong> a question to undergo syntactic Agreement, whereas <strong>in</strong> Cable’s <strong>the</strong>ory<br />

some phrases trigger <strong>in</strong>terrogative semantics but are <strong>in</strong>visible to <strong>the</strong> o<strong>the</strong>rwise correspond<strong>in</strong>g syntax.<br />

F<strong>in</strong>ally, I show that <strong>the</strong> complicated system of <strong>in</strong>terrogative prob<strong>in</strong>g that emerges from <strong>the</strong> addition<br />

of a second probe <strong>in</strong>teracts <strong>in</strong> <strong>in</strong>terest<strong>in</strong>g ways with locality restrictions: Superiority reemerges when a<br />

f<strong>in</strong>ite clause-boundary is <strong>in</strong>troduced between <strong>the</strong> two wh-phrases. In this configuration, superiority<br />

can be violated just <strong>in</strong> case a lower wh-headed phrase is moved over a wh-conta<strong>in</strong><strong>in</strong>g phrase, see (4ab).<br />

Surpris<strong>in</strong>gly, we <strong>the</strong>n observe <strong>in</strong>tervention effects for <strong>the</strong> question <strong>in</strong> (4b), and <strong>the</strong> only possible<br />

answer is pair-list. I will show how my proposal can be extended to account for <strong>the</strong>se facts.<br />

(4) Superiority effect reemerges when wh’s separated by clause boundary, and <strong>the</strong> exception to <strong>the</strong> rule<br />

a. *[et ma] ha-mora šixne’a [et mi] [ CP še-Yosi kara ___ ]?<br />

OM what <strong>the</strong>-teacher persuaded OM who that-Yosi read<br />

b. ? [ma] ha-mora šixne’a [et mi] [ CP še-Yosi kara ___ ]?<br />

what <strong>the</strong>-teacher persuaded OM who that-Yosi read<br />

‘What did <strong>the</strong> teacher persuade whom that Yosi read?’<br />

Selected references: Beck, S. 2006. Intervention effects follow from focus <strong>in</strong>terpretation. NALS � Cable,<br />

S. 2010. The grammar of Q: Q-particles, wh-movement and pied-pip<strong>in</strong>g. Oxford Uni. Press. � Dayal, V. 2002.<br />

S<strong>in</strong>gle-pair vs. multiple-pair answers: Wh <strong>in</strong>-situ and scope. LI � Fox, D. 2000. Economy and semantic<br />

<strong>in</strong>terpretation. MIT Press. � Pesetsky, D. 2000. Phrasal movement and its k<strong>in</strong>. MIT Press.


Ivona Kučerová (McMaster)<br />

Case Independence and Split Ergativity: Toward a Unified Theory of Case Assignment<br />

Two recent proposals, Coon and Prem<strong>in</strong>ger 2011 (henceforth, C&P) and Kučerová 2011 (henceforth, K),<br />

argued that case assignment is sensitive to <strong>the</strong> size of <strong>the</strong> syntactic structure. While C&P, follow<strong>in</strong>g Coon<br />

2010, concentrated on <strong>the</strong> Ergative/Absolutive (E/A) case syste<strong>ms</strong>, more precisely, on <strong>the</strong> syntactic source<br />

of split ergativity, K <strong>in</strong>vestigated emergence of Accusative(Acc)assignment<strong>in</strong>syntacticenvironmentslack<strong>in</strong>g<br />

an external argument and Nom<strong>in</strong>ative (Nom) case. The goal of this paper is to unify <strong>the</strong> two proposals<br />

and argue that cross-l<strong>in</strong>guistically case assignment is sensitive to <strong>the</strong> size of <strong>the</strong> syntactic structure. More<br />

precisely, <strong>the</strong> actual case assignment reflects which heads are strong phase heads and as such constitute<br />

Spell-out doma<strong>in</strong>s <strong>in</strong> <strong>the</strong> given syntactic structure (Cho<strong>ms</strong>ky 2001,Cho<strong>ms</strong>ky2005,Cho<strong>ms</strong>ky2008).Consequently,<br />

case assignment splits are predicted not to be restricted to E/A syste<strong>ms</strong> but should be <strong>in</strong> pr<strong>in</strong>ciple<br />

available <strong>in</strong> N/A syste<strong>ms</strong> as well. Data from Arabic copular clauses confirm this prediction.<br />

C&P: C&P observed that tense and person related <strong>in</strong>stances of split ergativityhavetwoproperties<strong>in</strong>common:<br />

(i) Structures that exhibit <strong>the</strong> split case system are syntactically larger than <strong>the</strong> structures that exhibit<br />

<strong>the</strong> regular E/A pattern. (ii) The split pattern is best characterized as a lack of morphological mark<strong>in</strong>g.<br />

Thus, <strong>the</strong>re are no dist<strong>in</strong>ct morphological case end<strong>in</strong>gs. Instead, <strong>the</strong> split pattern exhibits a lack of overt<br />

case end<strong>in</strong>gs. Based on <strong>the</strong>se two empirical generalizations, <strong>the</strong>yproposedthatsplitergativityisalackof<br />

case assignment due to a non-local configuration, i.e., <strong>the</strong> two relevant DPs end up be<strong>in</strong>g <strong>in</strong> two separate<br />

case doma<strong>in</strong>s, one of which is not local with respect to <strong>the</strong> Erg case assigner. The emerg<strong>in</strong>g morphological<br />

pattern is best characterized as default case result<strong>in</strong>g from lackoffeaturecheck<strong>in</strong>g.<br />

K: Karguedthatview<strong>in</strong>gAccasacasedependentonNomoron<strong>the</strong>presence of an external argument is<br />

empirically <strong>in</strong>adequate s<strong>in</strong>ce we can f<strong>in</strong>d Acc <strong>in</strong> <strong>the</strong> absence of an external argument or an argument receiv<strong>in</strong>g<br />

Nom. Accord<strong>in</strong>g to her, Acc is assigned only if <strong>the</strong> <strong>in</strong>ternal argument is spelled-out with<strong>in</strong> <strong>the</strong> vP<br />

phase. If v does not trigger Spell-out, Acc is not available. KadoptsRichard’sDist<strong>in</strong>ctnessConditionon<br />

L<strong>in</strong>earization (Richards 2003, 2006) accord<strong>in</strong>g to which vP can be spelled out only if it can be l<strong>in</strong>earized.<br />

For <strong>the</strong> condition to be satisfied <strong>the</strong> merge of v and its complement cannot be <strong>the</strong> maximal projection of v<br />

but <strong>the</strong> vP must be fur<strong>the</strong>r extended. The required extension is usuallyachievedbymergeofanexternal<br />

argument but <strong>in</strong> pr<strong>in</strong>ciple is <strong>in</strong>dependent of it. K discuses data from Polish, Ukra<strong>in</strong>ian, and North Russian<br />

<strong>in</strong> which <strong>the</strong> required extension is achieved <strong>in</strong> certa<strong>in</strong> <strong>in</strong>stances of have-Perfect. Even though <strong>the</strong>re is no<br />

external argument, <strong>the</strong> ‘transitivity’-like extension is achieved by add<strong>in</strong>g a structure semantically associated<br />

with sub<strong>in</strong>terval properties and <strong>in</strong> some languages morphologically realized as have. Ascanbeseen<strong>in</strong>(1)<br />

from North Russian dialects, an <strong>in</strong>ternal argument can get ei<strong>the</strong>r Nom or Acc; crucially, <strong>the</strong> case assignment<br />

is <strong>in</strong>dependent of agreement: <strong>in</strong> both cases, <strong>the</strong> verbal agreement shows morphological default. However,<br />

once <strong>the</strong> have-Perfect <strong>in</strong>terpretation is enforced, <strong>the</strong> <strong>in</strong>ternal argument must be realized as Acc, (2).<br />

(1) North Russian (Danylenko, 2006, p. 255–256, (18), orig<strong>in</strong>ally from Kuz’m<strong>in</strong>a 1993, 135–137):<br />

a. (u njego) syn<br />

(bylo)<br />

otpravleno<br />

at him son.NOM.SG.M. be.N.SG.AUX.PRET. send-away.N.SG.PPP<br />

‘His son has been sent away (by him).’<br />

b. (u njego) parnja<br />

(bylo)<br />

uvedeno<br />

at him fellow.ACC.SG.M. be.N.SG.AUX.PRET. take-away.N.SG.PPP<br />

(2)<br />

‘The guy has been taken away (by him).’<br />

North Russian (Kucerova 2011)<br />

a. *Vot uˇze tre goda kak u nego syn v amerku uvezeno.<br />

here already three years how by him son.NOM to America taken away<br />

b. Vot uˇze tre goda kak u nego syna v amerku uvezeno.<br />

here already three years how by him son.ACC=GEN to america taken away<br />

‘It has been three years s<strong>in</strong>ce his son has been taken away to America.’<br />

1


Proposal: What <strong>the</strong>se two proposals have <strong>in</strong> common is <strong>the</strong> observation that creat<strong>in</strong>g a locality boundary<br />

changes <strong>the</strong> case assignment properties. In <strong>the</strong> N/A system creat<strong>in</strong>g a locality boundary yields emergence of<br />

aspecialmorphologicalmark<strong>in</strong>g(Acc<strong>in</strong>steadof<strong>the</strong>expectedNom).In<strong>the</strong>E/Asystem,creat<strong>in</strong>galocality<br />

boundary yields loss of a special morphological mark<strong>in</strong>g (default or Abs <strong>in</strong>stead of <strong>the</strong>expectedErg).The<br />

question is whe<strong>the</strong>r we can unify <strong>the</strong>se two observations. Crucially, it has been argued that E/A and N/A<br />

languages are not syntactically identical. In particular, <strong>the</strong> v <strong>in</strong> <strong>the</strong> E/A pattern is ei<strong>the</strong>r defective or entirely<br />

miss<strong>in</strong>g (e.g., Marantz 1984, Nash 1995, 1995, Alexiadou 2001). We argue that consequently, vP <strong>in</strong> <strong>the</strong><br />

regular E/A pattern is not a spell-out doma<strong>in</strong>. If we assume, follow<strong>in</strong>g most of <strong>the</strong> current literature that Erg<br />

is assigned by a T head (Infl) while Acc is assigned by v <strong>the</strong> follow<strong>in</strong>g empirical generalization emerges:<br />

(3) A special morphological mark<strong>in</strong>g (Erg, Acc) arises only if <strong>the</strong>m<strong>in</strong>imalspell-outdoma<strong>in</strong>thatconta<strong>in</strong>s<br />

an <strong>in</strong>ternal argument is <strong>the</strong> spell-out doma<strong>in</strong> that conta<strong>in</strong>s <strong>the</strong> head responsible for <strong>the</strong> special<br />

morphological mark<strong>in</strong>g (T for Erg and v for Acc).<br />

In o<strong>the</strong>r words, Acc is possible only if <strong>the</strong> <strong>in</strong>ternal argument is spelled-out with<strong>in</strong> <strong>the</strong> vP spell-out doma<strong>in</strong><br />

and Erg is possible only if <strong>the</strong> <strong>in</strong>ternal argument is not spelled out with<strong>in</strong> vP but <strong>in</strong>stead it is spelled-out <strong>in</strong><br />

<strong>the</strong> next phase. Crucially, we have to ensure that Erg would be assign only if <strong>the</strong>re is more than one DP<br />

be<strong>in</strong>g spelled out <strong>in</strong> <strong>the</strong> same doma<strong>in</strong>. Let’s assume that someth<strong>in</strong>g like (4) holds. (A parallel condition<br />

might be needed for enforc<strong>in</strong>g Acc/Dat dist<strong>in</strong>ction <strong>in</strong> double-object constructions but we leave <strong>the</strong> issue of<br />

Dat aside for now.)<br />

(4) Case Dist<strong>in</strong>ctness Condition: Erg is assigned only if <strong>the</strong>re is more than one DP spelled-out <strong>in</strong> <strong>the</strong><br />

same Spell-out doma<strong>in</strong> as <strong>the</strong> head assign<strong>in</strong>g Erg.<br />

Predictions: Under <strong>the</strong> current proposal, splits <strong>in</strong> <strong>the</strong> sense of loos<strong>in</strong>g a special mark<strong>in</strong>g depend on <strong>the</strong> size<br />

of <strong>the</strong> syntactic doma<strong>in</strong> (as <strong>in</strong> C&P and K). In contrast to C&P, wepredictthatthistypeofcasesplitshould<br />

not be restricted to E/A but should be equally possible <strong>in</strong> N/A syste<strong>ms</strong>. In particular, we predict that if <strong>the</strong>re<br />

is only one DP <strong>in</strong> a clause, this DP will be assigned Nom if vP is a weak phase (or not a phase at all) or it is<br />

go<strong>in</strong>g to be assigned Acc if it is spelled out with<strong>in</strong> <strong>the</strong> vP phase. This prediction see<strong>ms</strong> to be borne out, for<br />

example, <strong>in</strong> Nor<strong>the</strong>rn Russian dialects, as we saw <strong>in</strong> (2).<br />

If <strong>the</strong>re is more than one DP <strong>in</strong> a clause, <strong>the</strong> prediction is that none of <strong>the</strong> DPs will get Acc if both DPs will<br />

be spelled out <strong>in</strong> <strong>the</strong> same phase, more precisely with<strong>in</strong> CP phase. If such a structure gets <strong>in</strong>dependently<br />

extended and <strong>the</strong> two DPs get spelled out <strong>in</strong> two separate phases, <strong>the</strong> prediction is that <strong>the</strong> higher one should<br />

be assigned Nom, while <strong>the</strong> lower one should be assigned Acc. Interest<strong>in</strong>gly, this prediction is borne out <strong>in</strong><br />

Arabic nom<strong>in</strong>al copular clauses. In particular, nom<strong>in</strong>al copular clauses <strong>in</strong> Arabic show dist<strong>in</strong>ct case pattern<br />

depend<strong>in</strong>g on <strong>the</strong> tense of <strong>the</strong> copular clause. As can be seen <strong>in</strong> (5-a), copula-less small clauses show no Acc<br />

assignment. Crucially, as <strong>in</strong> <strong>the</strong> ergative split pattern what weseehereisnotaspecialmark<strong>in</strong>gbutalackof<br />

mark<strong>in</strong>g. If <strong>the</strong> structure gets extended, which is presumably <strong>the</strong>caseofPastandFuture<strong>in</strong>(5-b)–(5-c),<strong>the</strong><br />

lower DP gets spelled out <strong>in</strong> <strong>the</strong> vP phase, result<strong>in</strong>g <strong>in</strong> a dist<strong>in</strong>ct case assignment of <strong>the</strong> two DPs.<br />

(5) a. Hassan Tabiib<br />

Hassan.NOM doctor.NOM<br />

‘Hassan is a doctor.’<br />

b. kaan Hassan Tabiib-an<br />

was Hassan.NOM doctor.ACC.INDEF<br />

‘Hassan was a doctor.’<br />

c. Sayakuunu Hassan<br />

be.FUT Hassan.NOM<br />

Tabiib-an<br />

doctor.ACC.INDEF<br />

‘Hassan will be a doctor.’<br />

Conclusion: Case assignment reflects Spell-out properties of a given syntactic structure. Crucially, creat<strong>in</strong>g<br />

aphaseboundaryalternatescasepatternsnotonly<strong>in</strong>E/Alanguages but <strong>in</strong> <strong>the</strong> N/A pattern as well.<br />

2


Dave Kush (Maryland)<br />

On-l<strong>in</strong>e use of relational structural <strong>in</strong>formation <strong>in</strong> process<strong>in</strong>g anaphora: evidence from English<br />

and H<strong>in</strong>di<br />

We present results from three experiments that show <strong>the</strong> parser makes rapid use of relational structural<br />

constra<strong>in</strong>ts <strong>in</strong> memory retrieval.<br />

F<strong>in</strong>d<strong>in</strong>gs on <strong>the</strong> on-l<strong>in</strong>e implementation of grammatical constra<strong>in</strong>ts are an important tool for<br />

understand<strong>in</strong>g how speakers encode and navigate hierarchical syntactic representations <strong>in</strong> memory.<br />

On-l<strong>in</strong>e sensitivity to a structural relation implies that this relation is mentally encoded, and that it is<br />

employed by <strong>the</strong> parser’s dependency-build<strong>in</strong>g mechanis<strong>ms</strong>. Here we <strong>in</strong>vestigate an <strong>in</strong>terest<strong>in</strong>g tension<br />

that arises when compar<strong>in</strong>g different recent f<strong>in</strong>d<strong>in</strong>gs <strong>in</strong> psychol<strong>in</strong>guistics. On <strong>the</strong> one hand, a grow<strong>in</strong>g<br />

body of evidence motivates a parser that relies on parallel access mechanis<strong>ms</strong> <strong>in</strong> content addressable<br />

memory [1]. Evidence comes from effects of <strong>in</strong>terference from grammatically illicit licensers [2,3],<br />

and from non-effects of syntactic dependency length [4,5]. Importantly, relational notions such as ccommand<br />

are difficult to exploit <strong>in</strong> such memory architectures, as <strong>the</strong>y are properties of configurations<br />

ra<strong>the</strong>r than <strong>in</strong>herent properties of <strong>in</strong>dividual nodes (‘content’). On <strong>the</strong> o<strong>the</strong>r hand, f<strong>in</strong>d<strong>in</strong>gs on <strong>the</strong><br />

apparent immunity-to-<strong>in</strong>terference of local anaphoric b<strong>in</strong>d<strong>in</strong>g suggest that dependencies that obey ccommand<br />

constra<strong>in</strong>ts can be accurately implemented on-l<strong>in</strong>e [6,7,8,9,10]. However, f<strong>in</strong>d<strong>in</strong>gs on English<br />

local b<strong>in</strong>d<strong>in</strong>g may be mislead<strong>in</strong>g. English reflexive anaphors require a clause-mate antecedent, and <strong>the</strong>y<br />

follow <strong>the</strong> verb, and so grammatically accurate retrieval might be achieved simply by retriev<strong>in</strong>g a coargument<br />

of <strong>the</strong> immediately preced<strong>in</strong>g verb. Here we present results from three experiments <strong>in</strong> English<br />

and H<strong>in</strong>di that avoid this confound by test<strong>in</strong>g (i) non-local b<strong>in</strong>d<strong>in</strong>g relations that unambiguously <strong>in</strong>volve<br />

c-command constra<strong>in</strong>ts, and (ii) local b<strong>in</strong>d<strong>in</strong>g relations <strong>in</strong> a verb-f<strong>in</strong>al language that must be established<br />

before verb <strong>in</strong>formation is encountered.<br />

Experiments 1 & 2 <strong>in</strong>vestigated <strong>the</strong> parser’s sensitivity to <strong>the</strong> c-command constra<strong>in</strong>t on boundvariable<br />

pronouns. Unlike local reflexive licens<strong>in</strong>g, this constra<strong>in</strong>t on bound-variable pronouns<br />

applies across an unbounded distance, and hence provides a strong test of <strong>the</strong> parser’s use of relational<br />

<strong>in</strong>formation <strong>in</strong> retrieval. We focused on <strong>the</strong> contrast<strong>in</strong>g ability of clauses headed by but and when to host<br />

bound variable pronouns <strong>in</strong> (1-2). The contrast reflects <strong>the</strong> lower attachment site of when-clauses, which<br />

allows any janitor to c-command <strong>the</strong> underl<strong>in</strong>ed pronoun. Importantly, coreference does not require ccommand,<br />

so replac<strong>in</strong>g <strong>the</strong> quantified NP with a referential NP (<strong>the</strong> janitor) elim<strong>in</strong>ates <strong>the</strong> contrast.<br />

(1) Kathy didn’t th<strong>in</strong>k that any janitor liked his job, when he had to clean up after <strong>the</strong> prom.<br />

(2) *Kathy didn’t th<strong>in</strong>k that any janitor liked his job, but he had to clean up after <strong>the</strong> prom.<br />

The experiments manipulated <strong>the</strong> type of antecedent (quantificational vs. referential) and <strong>the</strong> structural<br />

relation between <strong>the</strong> antecedent and <strong>the</strong> critical pronoun (WHEN/b<strong>in</strong>d<strong>in</strong>g vs. BUT/no-b<strong>in</strong>d<strong>in</strong>g). 24<br />

sets of ite<strong>ms</strong> were distributed across 4 lists <strong>in</strong> a Lat<strong>in</strong> Square design and comb<strong>in</strong>ed with 64 fillers.<br />

In Experiment 1 (self-paced read<strong>in</strong>g, n=24) immediate sensitivity to <strong>the</strong> c-command constra<strong>in</strong>t was<br />

demonstrated <strong>in</strong> an <strong>in</strong>teraction of antecedent-type and structural relation at <strong>the</strong> pronoun (p < .05) 1 , due<br />

to slower read<strong>in</strong>g times <strong>in</strong> <strong>the</strong> quantifier/no-b<strong>in</strong>d<strong>in</strong>g condition. In Experiment 2 (eye-track<strong>in</strong>g, n = 24)<br />

<strong>the</strong> same <strong>in</strong>teraction was found at <strong>the</strong> post-pronoun region (p < .05) <strong>in</strong> re-read, second-pass and total<br />

time measures. This shows that <strong>the</strong> parser is able to rapidly exploit c-command <strong>in</strong>formation <strong>in</strong> on-l<strong>in</strong>e<br />

dependency formation.<br />

Experiment 3 turned to H<strong>in</strong>di for an additional test of whe<strong>the</strong>r structure-sensitive anaphor<br />

1 All data were fit to a l<strong>in</strong>ear mixed-effects model with subject and item as random effects. P-values were<br />

estimated us<strong>in</strong>g <strong>the</strong> pvals.fnc() function <strong>in</strong> <strong>the</strong> LanguageR library [11].


process<strong>in</strong>g is cont<strong>in</strong>gent on a verb-based retrieval heuristic, or whe<strong>the</strong>r it is <strong>the</strong> product of general<br />

availability of relational <strong>in</strong>formation. H<strong>in</strong>di reciprocals must be locally licensed [12], like <strong>the</strong>ir English<br />

counterparts, but s<strong>in</strong>ce H<strong>in</strong>di is an SOV language <strong>the</strong>y appear pre-verbally, thus mak<strong>in</strong>g antecedentretrieval<br />

through mediation by <strong>the</strong> verb’s argument structure impossible.<br />

The design of <strong>the</strong> experiment used <strong>the</strong> <strong>in</strong>terference logic of previous reflexive studies.<br />

Reciprocals must be bound by a plural-marked, c-command<strong>in</strong>g NP, thus <strong>the</strong> relevant cue for retrieval<br />

is [+plural]. The experiment manipulated plural-mark<strong>in</strong>g on <strong>the</strong> matrix subject and on a potential<br />

<strong>in</strong>terfer<strong>in</strong>g NP <strong>in</strong> a pre-nom<strong>in</strong>al relative clause that l<strong>in</strong>early preceded <strong>the</strong> reciprocal. The structure of a<br />

test-sentence is given <strong>in</strong> (3) below. When NP1 (<strong>the</strong> matrix subject) is plural it can b<strong>in</strong>d <strong>the</strong> reciprocal.<br />

NP2, embedded <strong>in</strong>side an RC (boundaries marked with brackets), cannot grammatically b<strong>in</strong>d <strong>the</strong><br />

reciprocal, regardless of its number.If relational <strong>in</strong>formation is used <strong>in</strong> local licens<strong>in</strong>g, we predict no<br />

effect of plural-mark<strong>in</strong>g on NP2. If verb-mediated retrieval is required to block <strong>in</strong>terference, we expect<br />

<strong>in</strong>terference <strong>in</strong> conditions <strong>in</strong> which NP2 is plural.<br />

(3) NP1{sg/pl} [ ... NP2{sg/pl} ...] ... Reciprocal... {AdvP} V.<br />

An example sentence is given <strong>in</strong> (4). The matrix subject doctor(s) (underl<strong>in</strong>ed below) can b<strong>in</strong>d <strong>the</strong><br />

reciprocal ek-dusre when plural. The potential <strong>in</strong>terferer NP patient(s) (italicized) is embedded <strong>in</strong>side<br />

a pre-verbal RC <strong>the</strong> nurse who took care of <strong>the</strong> patients, thus remov<strong>in</strong>g its ability to c-command (and<br />

<strong>the</strong>refore b<strong>in</strong>d) <strong>the</strong> reciprocal.<br />

(4) Us/un doctor(on)-ne [us/un mariz(on)-ko dekhbal karne wali nurse] ke station me<br />

That/those doctor(s)-ERG that/those patients(s)-ACC care do<strong>in</strong>g RP nurse GEN station <strong>in</strong><br />

ek-dusre ke-saath gupt-ruup-se bat kii.<br />

one-ano<strong>the</strong>r with secretly chat did.<br />

`That/those doctor(s) talked secretly with one ano<strong>the</strong>r at <strong>the</strong> station of <strong>the</strong> nurse who was look<strong>in</strong>g after<br />

<strong>the</strong> patient(s).'<br />

24 sets of ite<strong>ms</strong> were distributed across 4 lists <strong>in</strong> a Lat<strong>in</strong> Square design and comb<strong>in</strong>ed with 50 filler<br />

ite<strong>ms</strong>. The experiment (n=30, native-speakers of H<strong>in</strong>di from Northwestern India, ages 18-26) revealed<br />

immediate sensitivity to <strong>the</strong> constra<strong>in</strong>t on reciprocal licens<strong>in</strong>g. A ma<strong>in</strong> effect of Subject-Number was<br />

observed <strong>in</strong> <strong>the</strong> region immediately follow<strong>in</strong>g <strong>the</strong> reciprocal phrase (p


Locality <strong>in</strong> Agreement: A New Approach<br />

Aim. The goal of this talk is to propose a new approach to syntactic structure-build<strong>in</strong>g<br />

that is based upon <strong>the</strong> idea that syntactic derivations are driven by a specificity pr<strong>in</strong>ciple<br />

on merge and agreement. The approach has <strong>the</strong> <strong>in</strong>terest<strong>in</strong>g property that it correctly<br />

derives relative locality effects (i.e., locality effects <strong>in</strong> Merge and Agree that are due to<br />

<strong>in</strong>tervention) without actually <strong>in</strong>vok<strong>in</strong>g closeness-based pr<strong>in</strong>ciples such as <strong>the</strong> M<strong>in</strong>imal<br />

L<strong>in</strong>k Condition. This talk first presents <strong>the</strong> new <strong>the</strong>ory and some of its assets and applications,<br />

but<strong>the</strong>n focusesonaparticular empirical doma<strong>in</strong>: <strong>in</strong>tervention-driven agreement<br />

alternations such as complementarity effects (e.g. complementarity <strong>in</strong> Breton agreement)<br />

and cases of (seem<strong>in</strong>g) optional alternation (e.g. defective <strong>in</strong>tervention <strong>in</strong> Icelandic).<br />

Theoretical Background. Specificity is arguably one of <strong>the</strong> ma<strong>in</strong> basic concepts of<br />

morphological <strong>the</strong>ory, where is used to resolve competitions between <strong>the</strong> markers of a<br />

language which arise due to underspecification of <strong>in</strong>flectional markers (Subset Pr<strong>in</strong>ciple).<br />

I would like to propose that structure build<strong>in</strong>g <strong>in</strong> syntactic derivations, too, is driven by<br />

specificity. The syntactic constra<strong>in</strong>t relevant to <strong>the</strong> local decisions made is <strong>the</strong> General<br />

Specificity Pr<strong>in</strong>ciple, an extension of Maximize Match<strong>in</strong>g Effects (Cho<strong>ms</strong>ky 2001:15):<br />

(1) General Specificity Pr<strong>in</strong>ciple: A Probe undergoes a syntactic operation with <strong>the</strong><br />

most specific match<strong>in</strong>g goal. Specificity is determ<strong>in</strong>ed by card<strong>in</strong>ality of morphosyntactic<br />

features: a set Q is more specific than a set H iff |Q| > |H|.<br />

The GSP has <strong>the</strong> effect that with more than one potential Goal be<strong>in</strong>g available <strong>in</strong> <strong>the</strong><br />

search space of a Probe, <strong>the</strong> Probe always agrees first with <strong>the</strong> Goal with which it can<br />

satisfy <strong>the</strong> highest number of features, even if it is not <strong>the</strong> closest available goal.<br />

Analysis. Let me briefly sketch a specificity-driven derivation for two different Agreement<br />

phenomena: complementarity <strong>in</strong> Breton agreement, and defective <strong>in</strong>tervention <strong>in</strong><br />

Icelandic. Breton shows a complementarity effect <strong>in</strong> that <strong>the</strong> φ-features of a phonologically<br />

null NP are coded by φ-agreement morphology on <strong>the</strong> verb (=‘rich agreement’),<br />

whereas <strong>the</strong> φ-features of a phonologically overt NP are not coded by φ-agreement morphology<br />

on <strong>the</strong> target (‘<strong>in</strong>variant agreement’ [=frozen 3sg agreement or bare stem]):<br />

(2) a. Gant<br />

with<br />

o mamm e karf-ent /*karf-e pro bez-añ<br />

<strong>the</strong>ir mo<strong>the</strong>r r would.love-3pl /*would.love-3sg 3pl be-<strong>in</strong>f<br />

‘They would like to be with <strong>the</strong>ir mo<strong>the</strong>r’<br />

b. Gant<br />

with<br />

Antje Lahne (Konstanz)<br />

o mamm e *karf-ent /karf-e Azenor ha Iona bez-añ<br />

<strong>the</strong>ir mo<strong>the</strong>r r *would.love-3pl /would.love-3sg Azenor and Iona be-<strong>in</strong>f<br />

‘Azenor and Iona would like to be with <strong>the</strong>ir mo<strong>the</strong>r’<br />

Jouitteau & Rezac (2006) analyse <strong>the</strong> complementarity as a locality effect. The start<strong>in</strong>g<br />

po<strong>in</strong>t of <strong>the</strong> analysis is <strong>the</strong> observation that v <strong>in</strong> Breton has nom<strong>in</strong>al properties. It is thus<br />

assumed to bear <strong>in</strong>terpretable 3sg φ-features. Consequently, when I probes for φ-features<br />

<strong>in</strong> its search space, <strong>the</strong>n v <strong>in</strong>tervenes between I and <strong>the</strong> external argument, which is conta<strong>in</strong>ed<br />

<strong>in</strong> <strong>the</strong> vP. I must <strong>the</strong>refore value its unvalued features with φ-features of v, which<br />

results <strong>in</strong> 3sg (‘frozen’) agreement on I. If, on <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> external argument<br />

is an affixal pro, <strong>the</strong>n it <strong>in</strong>corporates <strong>in</strong>to T and thus contributes its φ-features to <strong>the</strong><br />

feature set of T (i.e., it becomes a bound pronoun), which surfaces as rich agreement.<br />

While fully agree<strong>in</strong>g with <strong>the</strong> <strong>in</strong>corporation analysis, I argue that <strong>the</strong> alleged A-over-A<br />

effect is actually a Specificity effect: The GSP has <strong>the</strong> effect that <strong>the</strong> current probe I has<br />

to handle its selectional feature [•cat:v•] by merg<strong>in</strong>g with a v-type element; due to <strong>the</strong><br />

1


GSP, Agree between I and v must <strong>in</strong>volve handl<strong>in</strong>g of <strong>the</strong> maximal number of match<strong>in</strong>g<br />

features. Thus, if I has more features (φ-features: [*pers*], [*num*]) that it can value<br />

with a feature of v, <strong>the</strong>n it must value it with v. This is illustrated <strong>in</strong> (3).<br />

(3) a. b.<br />

I ′<br />

I o<br />

[•v•, *φ*]<br />

x<br />

vP<br />

DP v ′<br />

v o ...<br />

I o<br />

[•v•, *φ*]<br />

I ′<br />

vP<br />

DP v ′<br />

v o ...<br />

A second type of data that can be expla<strong>in</strong>ed by means of specificity are cases of apparant<br />

<strong>in</strong>decisiveness, namely when <strong>the</strong>re aretwo goalsthat <strong>the</strong>probecanagreewith <strong>in</strong><strong>the</strong>same<br />

number of features. I argue that this constellation leads to alternation. One example is<br />

defective <strong>in</strong>tervention <strong>in</strong> Icelandic (Holmberg & Hróarsdóttir 2004):<br />

(4) Mér f<strong>in</strong>nst/f<strong>in</strong>nast tölvurnar ljótar<br />

me.dat f<strong>in</strong>d.sg/f<strong>in</strong>d.pl computers:def.nom.pl ugly<br />

‘I consider <strong>the</strong> computers ugly’<br />

In brief, <strong>the</strong> approach to <strong>the</strong>se data is that I o has three k<strong>in</strong>ds of prob<strong>in</strong>g features: <strong>the</strong><br />

EPP feature, <strong>the</strong> set of phi-features, and <strong>the</strong> case feature [∗nom∗]. However, <strong>the</strong>re is no<br />

constituent that I o can handle all three features with – it cannot use <strong>the</strong> EPP feature on<br />

<strong>the</strong> lower DP tölvurnar (as it is an argument of <strong>the</strong> embedded predicate), and nei<strong>the</strong>r<br />

can it assign nom<strong>in</strong>ative case to <strong>the</strong> dative subject mér – but it is possible to use two<br />

of <strong>the</strong> three features on one goal, and <strong>the</strong> rema<strong>in</strong><strong>in</strong>g feature on <strong>the</strong> o<strong>the</strong>r goal. There<br />

are two possible orders <strong>in</strong> which this can be done. I would like to propose that <strong>the</strong><br />

order of operations is optional, and that both orders are attested: one order leads to <strong>the</strong><br />

s<strong>in</strong>gular agreement option, <strong>the</strong> o<strong>the</strong>r leads to plural agreement. The features [•D•] and<br />

[*nom*] cannot be handled toge<strong>the</strong>r, as <strong>the</strong> EPP feature cannot be used on <strong>the</strong> lower<br />

DP. This leaves two possible feature comb<strong>in</strong>ations: One possibility is to handle two of <strong>the</strong><br />

three features, [•D•] and [n:�], with <strong>the</strong> dative argument, and subsequently us<strong>in</strong>g <strong>the</strong><br />

rema<strong>in</strong><strong>in</strong>g feature [*case:nom*] on <strong>the</strong> lower DP. As a result, I shows s<strong>in</strong>gular agreement.<br />

The second possibility is to use [*case:nom*] and [n:�] on <strong>the</strong> lower DP, and <strong>the</strong>n handle<br />

[•D•] with <strong>the</strong> quirky subject. The verb consequently shows plural agreement.<br />

Consequences. The new approach yields new <strong>in</strong>sights <strong>in</strong>to <strong>the</strong> work<strong>in</strong>gs of locality <strong>in</strong><br />

Agreement: Agreement alternations arise due to extremely local decisions about feature<br />

handl<strong>in</strong>g made dur<strong>in</strong>g <strong>the</strong> derivation; <strong>the</strong>se decisions are driven by <strong>the</strong> need of probes to<br />

match with a goal as fully as possible, disregard<strong>in</strong>g structural distances. Optionalities<br />

arise when this need can be resolved <strong>in</strong> more than one way. L<strong>in</strong>k<strong>in</strong>g this <strong>in</strong>sight to <strong>the</strong><br />

f<strong>in</strong>d<strong>in</strong>gs presented at <strong>the</strong> beg<strong>in</strong>n<strong>in</strong>g of <strong>the</strong> talk, I would like to go one step fur<strong>the</strong>r and<br />

propose that Specificity, not Closeness, is <strong>the</strong> core pr<strong>in</strong>ciple that drives derivations and<br />

yields relative locality <strong>in</strong> Merge and Agree.<br />

References. Cho<strong>ms</strong>ky, Noam (2001): Derivation by Phase. In: M. Kenstowicz, ed.,<br />

Ken Hale: A Life <strong>in</strong> Language. MIT Press, Cambridge, Mass., pp. 1-52. Holmberg,<br />

Anders & Thorbjörg Hróarsdóttir (2004): Agreement and movement <strong>in</strong> Icelandic rais<strong>in</strong>g<br />

constructions, L<strong>in</strong>gua 114(5), 651-673. Jouitteau, Mélanie & Milan Rezac (2006): Deriv<strong>in</strong>g<br />

<strong>the</strong> Complementarity Effect: Relativized M<strong>in</strong>imality <strong>in</strong> Breton Agreement, L<strong>in</strong>gua<br />

116(11), 1915-1945.<br />

2


Mohamed Lahrouchi (CNRS/Paris 8)<br />

Phasal Spellout and <strong>the</strong> glide – high vowel alternation <strong>in</strong> Berber<br />

Berber languages present a wealth of <strong>in</strong>tricate phonological alternations <strong>in</strong>volv<strong>in</strong>g<br />

glides and high vowels, some of which still resist standard phonological analyses. These<br />

alternations shed light on <strong>the</strong> k<strong>in</strong>d of <strong>in</strong>ter-modular communication, which many current<br />

<strong>in</strong>terface <strong>the</strong>ories address.<br />

Glides typically appear <strong>in</strong> <strong>the</strong> immediate vic<strong>in</strong>ity of a vowel, <strong>in</strong> complementary<br />

distribution with <strong>the</strong> correspond<strong>in</strong>g high vowels. Based on this k<strong>in</strong>d of observations, standard<br />

<strong>the</strong>ories analyse glides and high vowels as phonetic reflexes of <strong>the</strong> same underly<strong>in</strong>g segments.<br />

The examples <strong>in</strong> (1) illustrate <strong>the</strong> situation <strong>in</strong> Tashlhiyt Berber:<br />

(1)<br />

gru ‘pick up’ agraw ‘assembly’<br />

bri ‘crush, pound’ abraj ‘crushed seeds’<br />

The problem arises with <strong>the</strong> correspond<strong>in</strong>g dative for<strong>ms</strong> grujas ‘pick to him’ and brijas<br />

‘crush seeds to him’. Followed by a vowel-<strong>in</strong>itial morpheme –as, U and I should normally<br />

surface as glides, lead<strong>in</strong>g to *grwas and *brjas. Faced with similar for<strong>ms</strong> <strong>in</strong> Tamazight Berber<br />

Guerssel (1986: 3) posits “a phonemic dist<strong>in</strong>ction between glides and high vowels”.<br />

The key to understand<strong>in</strong>g this paradox lies, we argue, <strong>in</strong> <strong>the</strong> morpho-syntactic<br />

structure of dative formations. In l<strong>in</strong>e with recent works at <strong>the</strong> syntax – phonology <strong>in</strong>terface<br />

(Marv<strong>in</strong> 2002, Marantz 2007, Pigott and Newell 2006, Samuels 2010), we argue that <strong>the</strong><br />

behaviour of I and U <strong>in</strong> <strong>the</strong> for<strong>ms</strong> just discussed is <strong>the</strong> result of <strong>the</strong> application of spell-out<br />

and phase impenetrability condition (Cho<strong>ms</strong>ky 2001) at different levels <strong>in</strong> <strong>the</strong> syntactic<br />

structure. In particular, we show that vP corresponds to a phase where U and I are spelled-out<br />

as high vowels before <strong>the</strong> enclitic –as is added. The result<strong>in</strong>g hiatus is <strong>the</strong>n resolved by means<br />

of j epen<strong>the</strong>sis, lead<strong>in</strong>g to grujas and brijas. In agraw and abraj, U and I surface as glides s<strong>in</strong>ce<br />

<strong>the</strong>y belong to <strong>the</strong> same phase (nP) as <strong>the</strong> neighbour<strong>in</strong>g a. The for<strong>ms</strong> represented <strong>in</strong> (2)<br />

illustrate <strong>the</strong> proposal:<br />

(2) a. grujas "pick to him" b. agraw ‘assembly’<br />

Spell-out-as-you-merge and Phase Impenetrability thus allow expla<strong>in</strong><strong>in</strong>g <strong>the</strong><br />

paradoxical behaviour of I and U <strong>in</strong> Tashlhiyt Berber (see also Pigott and Newell 2006 about<br />

Ojibwa). The question that arises <strong>the</strong>n is <strong>the</strong> follow<strong>in</strong>g: under <strong>the</strong> assumption that phonology<br />

is sensitive to external morph-syntactic <strong>in</strong>formation, what k<strong>in</strong>d of devices allows direct <strong>in</strong>termodular<br />

communication? Do morpho-syntactic phases leave any phonological traces?<br />

Follow<strong>in</strong>g Scheer (2011), we will assume that empty CV units, purely phonological objects


(Lowenstamm 1996, 1999), carry morpho-syntactic <strong>in</strong>formation <strong>in</strong>to phonology. They mark<br />

phase boundaries. How does phonology use <strong>the</strong>se empty CVs is an issue we will discuss.<br />

References<br />

Cho<strong>ms</strong>ky, Noam. 2001. Derivation by phase. In Ken Hale: A Life <strong>in</strong> Language, edited by<br />

Michael Kenstowicz, 1-52. Cambridge, Mass.: MIT Press.<br />

Dell, François & Mohamed Elmedlaoui. 2002. Syllables <strong>in</strong> Tashlhiyt Berber and <strong>in</strong> Moroccan<br />

Arabic. Dordrecht: Kluwer Academic Publisher.<br />

Guerssel, Mohand. 1986. Glides <strong>in</strong> Berber and Syllabicity. L<strong>in</strong>guistic Inquiry 17-1: 1-12.<br />

Lowenstamm, Jean. 1996. CV as <strong>the</strong> only syllable type. In Current trends <strong>in</strong> Phonology.<br />

Models and Methods, edited by Jacques Durand & Bernard Laks, 419-441.<br />

Salford, Manchester: ESRI.<br />

Lowenstamm, Jean. 1999. The beg<strong>in</strong>n<strong>in</strong>g of <strong>the</strong> word. In Phonologica 1996, edited by John<br />

Rennison & Klaus Kühnhammer, 153-166. The Hague: Holland Academic Graphics.<br />

Marantz, Alec. 2007. Phases and Words. In Phases <strong>in</strong> <strong>the</strong> Theory of Gramma, edited by Sook-<br />

Hee Choe, 191-222. Seoul: Dong In.<br />

Marv<strong>in</strong>, Tatjana. 2002. Topics <strong>in</strong> <strong>the</strong> Stress and Syntax of Words. Ph.D dissertation, MIT.<br />

Piggott, Glyne & Hea<strong>the</strong>r Newell. 2006. Syllabification and <strong>the</strong> Spell-Out of Phases <strong>in</strong><br />

Ojibwa Words. McGill Work<strong>in</strong>g Papers <strong>in</strong> <strong>L<strong>in</strong>guistics</strong> 20.2: 39-64.<br />

Samuels, Bridget. 2010. Phonological Derivation by Phase: Evidence from Basque.<br />

University of Pennsylvania Work<strong>in</strong>g Papers <strong>in</strong> <strong>L<strong>in</strong>guistics</strong> 16: 166-175.<br />

Scheer, Tobias. 2011. A Guide to Morphosyntax – Phonology Interface Theories. Berl<strong>in</strong>:<br />

Mouton de Gruyter.<br />

Uriagereka, Juan 1999. Multiple spell-out. In Work<strong>in</strong>g M<strong>in</strong>imalism, edited by Samuel Epste<strong>in</strong><br />

& Norbert Hornste<strong>in</strong>, 251-282. Cambridge, Mass.: MIT Press.


Chris Laterza (Maryland)<br />

Gaps with<strong>in</strong> Silence<br />

This talk provides a novel account of <strong>the</strong> role of traces <strong>in</strong> <strong>the</strong> licens<strong>in</strong>g conditions for VP ellipsis<br />

(VPE). As I will show, this approach unifies <strong>the</strong> licens<strong>in</strong>g conditions for canonical VPE with<br />

related phenomena such as pseudogapp<strong>in</strong>g and ACD, a task that has been attempted <strong>in</strong> several<br />

different ways s<strong>in</strong>ce at least as early as Lapp<strong>in</strong> (1992). Assum<strong>in</strong>g that VPE is PF deletion<br />

licensed by LF identity (an approach pursued, for example, by Sag (1976) and Kennedy (2003)),<br />

my ma<strong>in</strong> claim (<strong>in</strong>formally put) is that traces <strong>in</strong> <strong>the</strong> ellipsis clause (EC) act as “wildcard” element<br />

for <strong>the</strong> determ<strong>in</strong>ation of identity; just so long as <strong>the</strong>re is a DP/PP <strong>in</strong> <strong>the</strong> antecedent clause (AC)<br />

VP occupy<strong>in</strong>g <strong>the</strong> same type of argument/adjunct position as <strong>the</strong> trace <strong>in</strong> <strong>the</strong> EC VP, <strong>the</strong>n <strong>the</strong> that<br />

AC DP/PP and <strong>the</strong> EC trace count as identical for <strong>the</strong> purposes of licens<strong>in</strong>g VPE, even if <strong>the</strong> LF<br />

content of <strong>the</strong> AC DP/PP is not identical to that of <strong>the</strong> DP/PP which moved from <strong>the</strong> position of<br />

<strong>the</strong> trace <strong>in</strong> <strong>the</strong> EC. To illustrate, take <strong>the</strong> follow<strong>in</strong>g <strong>in</strong>stance of VPE:<br />

(1) John [VP met Cho<strong>ms</strong>ky at WCCFL]. Guess who1 Bill did [VP meet t1 at WCCFL].<br />

A more str<strong>in</strong>gent approach to ellipsis licens<strong>in</strong>g would have a hard time expla<strong>in</strong><strong>in</strong>g (without ad<br />

hoc str<strong>in</strong>g vacuous movement) how VPE is licensed <strong>in</strong> (1), s<strong>in</strong>ce [Cho<strong>ms</strong>ky] is not identical to<br />

[t1] nor [who1] <strong>in</strong> any approach to VPE (previous versions of PF deletion, LF copy<strong>in</strong>g, pro-form<br />

VPs, etc.). However, <strong>the</strong> relaxed approach to identity offered here treats (1) as a licit case of<br />

VPE: almost everyth<strong>in</strong>g <strong>in</strong> <strong>the</strong> AC and EC VPs are LF-identical, <strong>the</strong> sole exception be<strong>in</strong>g <strong>the</strong><br />

direct objects, and s<strong>in</strong>ce <strong>the</strong> object <strong>in</strong> <strong>the</strong> EC is a trace, it counts as identical with [Cho<strong>ms</strong>ky] <strong>in</strong><br />

<strong>the</strong> AC under this approach and thus ellipsis is licensed. The rema<strong>in</strong>der of <strong>the</strong> abstract provides<br />

some o<strong>the</strong>r examples of this approach at work.<br />

PSEUDOGAPPING: The present approach fits well with a popular treatment of pseudogapp<strong>in</strong>g<br />

(as argued for by Jayaseelan (1990) and Lasnik (1999)) which states that pseudogapp<strong>in</strong>g is<br />

extraction of <strong>the</strong> "survivor" from <strong>the</strong> VP followed by VPE, as shown <strong>in</strong> (2). This creates a<br />

configuration that is short of full LF-identity between <strong>the</strong> AC and EC VPs, though it is one that<br />

falls with<strong>in</strong> <strong>the</strong> purview of <strong>the</strong> current approach and counts as a licit case of VPE, s<strong>in</strong>ce t1 will<br />

match [on <strong>the</strong> porch] under <strong>the</strong> current approach, even though t1 is a trace of [<strong>in</strong> <strong>the</strong> gazebo].<br />

(2) John [VP read a book on <strong>the</strong> porch this morn<strong>in</strong>g], while Bill did (so) [VP read a<br />

book t1 this morn<strong>in</strong>g] [PP <strong>in</strong> <strong>the</strong> gazebo]1<br />

ACD: Virtually all analyses of relative clause formation <strong>in</strong>volve some sort of movement from<br />

<strong>the</strong> relativization site to some A'-position. Therefore, it is typical to treat cases of ACD as<br />

<strong>in</strong>volv<strong>in</strong>g a trace <strong>in</strong> <strong>the</strong> EC VP (whenever <strong>the</strong> relativization site is with<strong>in</strong> <strong>the</strong> VP). Under <strong>the</strong><br />

present approach, (3) is licensed s<strong>in</strong>ce <strong>the</strong> AC VP = [read [DP few books that Bill did]] and <strong>the</strong><br />

EC VP = [read t1], and t1 will match any direct object DP <strong>in</strong> <strong>the</strong> AC clause, even if it is complex<br />

and conta<strong>in</strong>s <strong>the</strong> trace itself.<br />

(3) John [VP read few (of <strong>the</strong>) books Op1 that Bill did [VP read t1 ] ]<br />

While some argue that <strong>the</strong>re is QR <strong>in</strong> (3), <strong>the</strong>reby mak<strong>in</strong>g <strong>the</strong> current approach unnecessary for<br />

cases of ACD, <strong>the</strong>re are some cases where <strong>the</strong> content of <strong>the</strong> determ<strong>in</strong>er matters for VPE<br />

licens<strong>in</strong>g, and thus it must somehow be represented with<strong>in</strong> <strong>the</strong> VP. We can see this <strong>in</strong> <strong>the</strong> non-<br />

ACD (4), as <strong>the</strong> content of <strong>the</strong> EC is clearly that Bill read few books.<br />

(4) John read few (of <strong>the</strong>) books, and Bill did [VP read few (of <strong>the</strong>) books] too<br />

If QR is present <strong>in</strong> (4), leav<strong>in</strong>g a trace <strong>in</strong> <strong>the</strong> VP that is devoid of any content (an assumption<br />

made by May (1985) for QR to deal with <strong>in</strong>f<strong>in</strong>ite regress), <strong>the</strong>n it is surpris<strong>in</strong>g that <strong>the</strong><br />

determ<strong>in</strong>er’s mean<strong>in</strong>g, <strong>the</strong> aspect of “fewness”, is relevant for VPE. (3) differs from (4) <strong>in</strong> that<br />

<strong>the</strong>re is no DP movement out of <strong>the</strong> VP <strong>in</strong> (4), for if <strong>the</strong>re were, we are left without an<br />

explanation for why <strong>the</strong> DP’s mean<strong>in</strong>g is recovered <strong>in</strong> <strong>the</strong> EC VP <strong>in</strong> (4). Thus, (4) has stricter


identity conditions than <strong>in</strong> <strong>the</strong> case of (3) which <strong>in</strong>volves a trace <strong>in</strong> <strong>the</strong> relativization site, which<br />

lightens <strong>the</strong> identity conditions under <strong>the</strong> present approach; i.e., <strong>the</strong> semantic content of <strong>the</strong> direct<br />

object’s determ<strong>in</strong>er matters for VPE licens<strong>in</strong>g <strong>in</strong> (4) but not <strong>in</strong> (3). From what I can tell, <strong>the</strong><br />

observation of <strong>the</strong> asymmetry above between (3) and (4) (whe<strong>the</strong>r or not determ<strong>in</strong>er mean<strong>in</strong>gs<br />

matter for VPE) brought about by non-universal determ<strong>in</strong>ers is novel, and is expla<strong>in</strong>ed by <strong>the</strong><br />

present account of traces <strong>in</strong> VPE identity conditions. It is also noteworthy that <strong>the</strong> <strong>in</strong>f<strong>in</strong>ite regress<br />

problem is no problem at all for <strong>the</strong> current approach, s<strong>in</strong>ce VPE licens<strong>in</strong>g does not require <strong>the</strong><br />

content of <strong>the</strong> AC direct object <strong>in</strong> (3) to be represented <strong>in</strong> <strong>the</strong> position of <strong>the</strong> trace <strong>in</strong> <strong>the</strong> EC.<br />

Fur<strong>the</strong>r support for <strong>the</strong> AC direct object <strong>in</strong> (3) not be<strong>in</strong>g represented <strong>in</strong> <strong>the</strong> EC aga<strong>in</strong> comes from<br />

<strong>the</strong> paradigm above <strong>in</strong>volv<strong>in</strong>g a proportional determ<strong>in</strong>er; if <strong>the</strong> AC direct object were represented<br />

<strong>in</strong> <strong>the</strong> EC <strong>in</strong> (3) (mak<strong>in</strong>g it relevant for VPE licens<strong>in</strong>g), <strong>the</strong>n why is <strong>the</strong> aspect of "fewness"<br />

miss<strong>in</strong>g from this clause, <strong>in</strong> a way that it is clearly present <strong>in</strong> <strong>the</strong> non-ACD (4)? This account is<br />

conceptually similar to Wyngaerd and Zwart (1991) <strong>in</strong> that both try to account for ellipsis<br />

licens<strong>in</strong>g <strong>in</strong> ACD constructions without be<strong>in</strong>g dependent on QR.<br />

SPROUTING: Sprout<strong>in</strong>g (Chung et al. (1995)) is when, <strong>in</strong> a case of sluic<strong>in</strong>g, a DP/PP that was<br />

not present <strong>in</strong> <strong>the</strong> AC appears overtly <strong>in</strong> <strong>the</strong> ellipsis clause (typically as a wh-expression).<br />

(5) John saw The Lion K<strong>in</strong>g, but I don't know where<br />

(6) There navy seals were fir<strong>in</strong>g, but at what is unclear<br />

(7) Mary f<strong>in</strong>ished <strong>the</strong> exam on time, but with whose help is unclear<br />

However, <strong>the</strong> possibility of sprout<strong>in</strong>g see<strong>ms</strong> to be limited to cases of sluic<strong>in</strong>g. (8-10) below<br />

show that <strong>the</strong> acceptability of sprout<strong>in</strong>g <strong>in</strong> cases of VPE is severely degraded. Unacceptability of<br />

sprout<strong>in</strong>g <strong>in</strong> cases of VPE is, to <strong>the</strong> best of my knowledge, a novel empirical observation.<br />

(8) *John saw The Lion K<strong>in</strong>g. I wonder where Bill did<br />

(9) *The navy seals were fir<strong>in</strong>g, but at what <strong>the</strong> green berets were is unclear<br />

(10) *Mary f<strong>in</strong>ished <strong>the</strong> exam on time, but with whose help did Susan<br />

(Notice that <strong>the</strong> subject DPs <strong>in</strong> <strong>the</strong> ECs are different from those <strong>in</strong> <strong>the</strong>ir respective ACs <strong>in</strong> order<br />

to rule out MaxElide as be<strong>in</strong>g responsible for <strong>the</strong> unacceptability; see Hartman (2011).) The<br />

present approach can easily account for unacceptability of (8-10). Consider a case like (8): <strong>the</strong><br />

AC VP = [saw The Lion K<strong>in</strong>g], and <strong>the</strong> EC VP = [saw The Lion K<strong>in</strong>g t]; <strong>the</strong>re is no<br />

correspond<strong>in</strong>g locative PP <strong>in</strong> <strong>the</strong> AC for <strong>the</strong> t <strong>in</strong> <strong>the</strong> EC to match, hence (8) is not a licit case of<br />

VPE under <strong>the</strong> present approach s<strong>in</strong>ce <strong>the</strong> LF-identity conditions are not met. It is unclear to me<br />

why sprout<strong>in</strong>g would be acceptable <strong>in</strong> cases of sluic<strong>in</strong>g and not VPE, and it is worth po<strong>in</strong>t<strong>in</strong>g out<br />

that <strong>the</strong> fact that sprout<strong>in</strong>g is allowed <strong>in</strong> cases of sluic<strong>in</strong>g is precisely why this account is limited<br />

to VPE and cannot be extended to IP ellipsis. Perhaps sprout<strong>in</strong>g is just ano<strong>the</strong>r way <strong>in</strong> which<br />

sluic<strong>in</strong>g and VPE differ, on par with as Chung et al.’s (1995) observation that while sluic<strong>in</strong>g can<br />

repair island violations, VPE cannot.<br />

References: Chung, S. et al. 1995. Sluic<strong>in</strong>g and logical form. Natural Language Semantics 3:<br />

239-282. Jayaseelan. K.A. 1990. Incomplete VP deletion and gapp<strong>in</strong>g. L<strong>in</strong>guistic Analysis 20:<br />

64-81. Kennedy, C. 2003 Ellipsis and syntactic representation. In The Interfaces: deriv<strong>in</strong>g and<br />

<strong>in</strong>terpret<strong>in</strong>g omitted structures. Hartman, J. 2011. The semantic uniformity of traces: evidence<br />

from ellipsis parallelism. L<strong>in</strong>guistic Inquiry 42.3: 367-388. Lapp<strong>in</strong>, S. 1992. The syntactic basis<br />

of ellipsis resolution. In Proceed<strong>in</strong>gs of <strong>the</strong> Stuttgart Workshop on Ellipsis. Lasnik, H. 1999.<br />

Pseudogapp<strong>in</strong>g puzzles. In Fragments: Studies <strong>in</strong> ellipsis and gapp<strong>in</strong>g, 141-174. May, R. 1985.<br />

Logical Form. Cambridge: MIT Press. Sag, I. 1976. Deletion and Logical Form. PhD diss., MIT.<br />

Wyngaerd, G. V., and J.-W. Zwart. 1991. Reconstruction and Vehicle Change. In <strong>L<strong>in</strong>guistics</strong><br />

<strong>in</strong> <strong>the</strong> Ne<strong>the</strong>rlands, 151-160.


Timothy Leffel (NYU)<br />

Nonrestrictive adjectives and <strong>the</strong> <strong>the</strong>ory of scalar implicature<br />

Summary Nonrestrictive adjectives (NAs) attribute a general property to <strong>in</strong>dividuals <strong>in</strong> <strong>the</strong><br />

denotation of <strong>the</strong> adjacent noun (Larson & Maruˇsič 2004; Morzycki 2008; etc.).<br />

(1) a. Cigarettes conta<strong>in</strong> harmful tox<strong>in</strong>s. b. ∴ Tox<strong>in</strong>s are (generally) harmful.<br />

Nom<strong>in</strong>al modifiers typically give rise to implicatures <strong>in</strong> downward entail<strong>in</strong>g contexts ((2)). Correspond<strong>in</strong>g<br />

implicatures are not triggered by NAs ((3)).<br />

(2) a. Every harmful chemical will be elim<strong>in</strong>ated by this product.<br />

b. � Not every chemical will be elim<strong>in</strong>ated by this product.<br />

(3) a. Every harmful tox<strong>in</strong> will be elim<strong>in</strong>ated by this product.<br />

b. �� Not every tox<strong>in</strong> will be elim<strong>in</strong>ated by this product.<br />

Here I argue that <strong>the</strong> contrast between (2) and (3) is not due to <strong>the</strong> contextual knowledge that<br />

tox<strong>in</strong>s are generally harmful. First I argue that scalar alternatives should be computed structurally<br />

(Katzir 2007) and excludability should be computed without access to contextual <strong>in</strong>formation<br />

(Magri 2009). I <strong>the</strong>n suggest that NAs <strong>in</strong> non-def<strong>in</strong>ite DPs comb<strong>in</strong>e with nouns via<br />

a presupposition-<strong>in</strong>troduc<strong>in</strong>g functional head <strong>in</strong> C<strong>in</strong>que’s (2010) “direct modification” doma<strong>in</strong><br />

(build<strong>in</strong>g on <strong>in</strong>sights of Larson 1998;2000). This correctly predicts <strong>the</strong> <strong>in</strong>ference <strong>in</strong> (3) to be<br />

<strong>in</strong>valid because (3a) has a presupposition that <strong>the</strong> negation of (3b) lacks (details below).<br />

Background on SI The computation of a sentence’s scalar implicatures (SIs) depends on what<br />

<strong>the</strong> alternatives to that sentence are. In “Horn scale”-based approaches (e.g. Gazdar 1979;<br />

Sauerland 2004), alternatives are obta<strong>in</strong>ed by replac<strong>in</strong>g scalar ter<strong>ms</strong> with “scalemates,” so that<br />

(4a) and (4b) are alternatives to one ano<strong>the</strong>r (s<strong>in</strong>ce 〈some,all〉 is a Horn scale).<br />

(4) a. John ate some of <strong>the</strong> beans. b. John ate all of <strong>the</strong> beans.<br />

SIs are <strong>the</strong>n computed on <strong>the</strong> basis of whe<strong>the</strong>r an alternative is excludable <strong>in</strong> <strong>the</strong> sense that<br />

its negation asymmetrically entails what is asserted. If an alternative is excludable, <strong>the</strong>n its<br />

negation is a predicted SI (details differ from <strong>the</strong>ory to <strong>the</strong>ory). Scale-based approaches do not<br />

predict implicatures to arise from <strong>in</strong>tersective modifiers as <strong>in</strong> (2), s<strong>in</strong>ce nouns like chemical are<br />

not on a lexical scale with harmful chemical.<br />

Katzir’s (2007) structural <strong>the</strong>ory of alternatives provides a natural explanation for (2) while<br />

ma<strong>in</strong>ta<strong>in</strong><strong>in</strong>g <strong>in</strong>sights of scale-based approaches. Alternatives are determ<strong>in</strong>ed syntactically: given<br />

a sentence ϕ, ψ ∈ Alt(ϕ) iff ψ can be obta<strong>in</strong>ed from ϕ by a f<strong>in</strong>ite series of (i) deletions of constituents,<br />

(ii) contractions (remove tree edge and identify end nodes), and (iii) substitutions of<br />

term<strong>in</strong>al elements for o<strong>the</strong>r elements (which are ei<strong>the</strong>r lexical or are subtrees of ϕ) of <strong>the</strong> same<br />

(syntactic) category. In this framework excludability is def<strong>in</strong>ed <strong>in</strong> ter<strong>ms</strong> of a better-than relation<br />

≺. ϕ is at least as good (�) as ψ iff ϕ is an alternative to ψ and ϕ entails ψ. If ϕ � ψ but<br />

ψ �� ϕ, <strong>the</strong>n ϕ is strictly better (≺) than ψ. Equivalently, ϕ ≺ ψ iff<br />

(5) a. ϕ is an alternative to ψ and entails ψ; and ei<strong>the</strong>r<br />

b. ψ is not an alternative to ϕ, or ψ does not entail ϕ.<br />

Determ<strong>in</strong><strong>in</strong>g excludability The better-than relation <strong>in</strong> (5) makes crucial reference to “entailment.”<br />

Two candidate def<strong>in</strong>itions of “entailment” are: (from Magri 2009)<br />

(6) a. ϕ logically entails ψ iff �ϕ� ⊆ �ψ�<br />

b. ϕ contextually entails ψ iff (�ϕ� ∩C) ⊆ �ψ�, where C is <strong>the</strong> Context Set.


Magri (2009;2011) argues that (6a) is <strong>the</strong> notion of entailment relevant for comput<strong>in</strong>g SIs—this<br />

is called <strong>the</strong> “bl<strong>in</strong>dness hypo<strong>the</strong>sis” (BH) because it says that SIs are computed without access<br />

to world knowledge. BH is crucial <strong>in</strong>, e.g. captur<strong>in</strong>g <strong>the</strong> oddness of #Some Italians come from a<br />

warm country: it triggers <strong>the</strong> bl<strong>in</strong>d implicature that not all Italians come from a warm country.<br />

I propose that BH is required for <strong>the</strong> <strong>the</strong>ory of structural alternatives to work. The argument<br />

is based on <strong>the</strong> existence of lexical ite<strong>ms</strong> whose sole function is to <strong>in</strong>troduce a presupposition:<br />

(7) Context: It is mutually known that several people o<strong>the</strong>r than Mary came to <strong>the</strong> party.<br />

a. Mary came to <strong>the</strong> party too. b. Mary came to <strong>the</strong> party.<br />

Given <strong>the</strong> contextual <strong>in</strong>formation, (7a) and (7b) are equivalent. Therefore if entailment is def<strong>in</strong>ed<br />

as <strong>in</strong> (6b), it follows that (7b)≺(7a), because (7b) is an alternative to (7a), but (7a) is not<br />

an alternative to (7b) (see (5)). Hence an utterance of (7a) is <strong>in</strong>correctly predicted to give rise to<br />

<strong>the</strong> implicature that (7b) is not assertable. This argument can be replicated for also, even, etc.<br />

If <strong>in</strong>stead logical entailment (=(6a)) is used <strong>in</strong> (5), <strong>the</strong>n (7b) is not predicted to be strictly<br />

better than (7a), because:<br />

(8) a. (7b) is an alternative to (7a) (delete too and contract its branch), but<br />

b. (7b) does not logically entail (7a), s<strong>in</strong>ce <strong>the</strong>re are logically consistent worlds <strong>in</strong><br />

which Mary came to <strong>the</strong> party and no one else did.<br />

and hence an utterance of (7a) is correctly predicted to not give rise to <strong>the</strong> implicature that (7b)<br />

is not assertable. Fur<strong>the</strong>r, logical entailment must be computed relative to local contexts (see<br />

Schlenker 2009), s<strong>in</strong>ce if entailment were computed globally, <strong>the</strong>n (9a) should be dispreferred<br />

to (9b), s<strong>in</strong>ce (9a) has no global presupposition.<br />

(9) a. John came to <strong>the</strong> party, and Mary came to <strong>the</strong> party too.<br />

b. John came to <strong>the</strong> party, and Mary came to <strong>the</strong> party.<br />

Nonrestrictive read<strong>in</strong>gs NR read<strong>in</strong>gs of attributive adjectives are licensed only <strong>in</strong> certa<strong>in</strong> syntactic<br />

positions, which may vary depend<strong>in</strong>g on <strong>the</strong> language (Bol<strong>in</strong>ger 1967; C<strong>in</strong>que 2010).<br />

(10) a. Every unsuitable word was deleted. ✓Rest. ✓NR<br />

b. Every word unsuitable was deleted. ✓Rest. ✗NR<br />

C<strong>in</strong>que (2010) has argued that NAs are <strong>in</strong>troduced as <strong>the</strong> specifiers of a DP-<strong>in</strong>ternal functional<br />

head <strong>in</strong> <strong>the</strong> “direct modification” doma<strong>in</strong>, which <strong>in</strong> English is immediately above NP but below<br />

a doma<strong>in</strong> of reduced relative clause adjectives with <strong>in</strong>tersective semantics. Larson (1998;2000)<br />

notes that many adjectives <strong>in</strong> <strong>the</strong> direct modification doma<strong>in</strong> are associated with generic <strong>in</strong>terpretation.<br />

I propose that NAs are also associated with genericity. The functional head that<br />

<strong>in</strong>troduces NAs (GENNR) is assigned <strong>the</strong> follow<strong>in</strong>g semantics, which encodes <strong>the</strong> <strong>in</strong>ference <strong>in</strong><br />

(1b) as a presupposition (“:” <strong>in</strong>dicates def<strong>in</strong>edness condition, Γ is a generic quantifier).<br />

(11) GENNR : λPλQλy : Γx[P(x)][Q(x)].P(y)<br />

If harmful and tox<strong>in</strong> <strong>in</strong> (3a) comb<strong>in</strong>e via (11), <strong>the</strong>n (3a) presupposes that tox<strong>in</strong>s are generically<br />

harmful. The structural alternative Every tox<strong>in</strong> will be... to (3a) does not have this presupposition,<br />

and so does not logically entail (3a) (by reason<strong>in</strong>g along <strong>the</strong> l<strong>in</strong>es of (8)). Therefore <strong>the</strong><br />

<strong>in</strong>ference from (3a) to (3b) is correctly predicted to be <strong>in</strong>valid. This explanation is not available<br />

to any pragmatic <strong>the</strong>ory of NR read<strong>in</strong>gs.<br />

Selected references C<strong>in</strong>que, G. 2010. The syntax of adjectives • Katzir, R. 2007. Structurallydef<strong>in</strong>ed<br />

alternatives. L&P. • Larson, R. 1998. Events and modification <strong>in</strong> nom<strong>in</strong>als. SALT<br />

8. • Magri, G. 2009. A <strong>the</strong>ory of <strong>in</strong>dividual-level predicates based on bl<strong>in</strong>d mandatory scalar<br />

implicatures. NaLS. • Sauerland, U. 2004. Scalar implicatures <strong>in</strong> complex sentences. L&P.


Shevaun Lewis (Maryland) & W<strong>in</strong>g Yee Chow (Maryland)<br />

!<br />

" # $<br />

' $ " $<br />

% &<br />

" ( )<br />

"! * ! ! !<br />

" " $ !<br />

!<br />

! $ "<br />

" % )<br />

$ $ !<br />

+ $<br />

! !<br />

" (<br />

, ! " "<br />

! " $<br />

" !<br />

) !<br />

$ - .! !<br />

$ ! -<br />

$<br />

"! . . /0 $ $ $<br />

1 " 2 3<br />

$ ! $ # #<br />

! $ " " " $<br />

$ $<br />

" + $<br />

/ !3 $ " "<br />

+ " " $ $<br />

+ $ " $ $<br />

4 5 $ "<br />

" ! "<br />

$ ! , ! "<br />

"! ! 5 $ $<br />

" $<br />

6<br />

" ( ! " "!<br />

' $ !<br />

+ ! "<br />

! $<br />

$<br />

$ " - $ -<br />

$ " + "


! " 1 2<br />

" $ !<br />

+ / 3 $<br />

" $ /7 83 , " - / "9 3 + " -<br />

/ "9 3 $ ! " "9 / 3<br />

/ 3 5 :; < = :* = "<br />

$<br />

/ "3 :; < = : < = "<br />

$<br />

/ 3 5 :; < = : < = $<br />

$ $ "<br />

$<br />

) " "9 ! $ "<br />

, " > ! $ + ; $<br />

" "9 $ $ + " "9<br />

- > ! !<br />

" $ ! $<br />

* # ! ! $<br />

? ! $ " ! + "<br />

, " /<br />

3<br />

+ $ /<br />

; " 3 )<br />

" , " > $ "<br />

+ " > $<br />

0 # ! " !<br />

$ "!<br />

+ ! $<br />

$ " $ )<br />

! + " ! @<br />

) "!<br />

" " ( " ; ! !<br />

!<br />

" " ! ! /<br />

3 " " ! !<br />

" " /<br />

; 3 " $ !<br />

/; " 3 6 $ !<br />

"<br />

. # A " / BB 3 C &4C &%D<br />

E / BB 3 4D<br />

- A $ ! / DCD3 C D<br />

4 ? E A , " / DD43 % &% D<br />

F A ) / BB43 G "<br />

% / BB 3 4C 4 %<br />

& H I A / BBD3 BC 4B


Charles L<strong>in</strong> (Indiana)<br />

Typological Perspectives on Relative Clause Process<strong>in</strong>g: Thematic Mapp<strong>in</strong>g, Case<br />

Markedness, Filler-Gap Integrations, and Their Relative Tim<strong>in</strong>g<br />

It is well-known that <strong>in</strong> head-<strong>in</strong>itial languages, relative clauses <strong>in</strong>volv<strong>in</strong>g subject<br />

extractions are processed with greater ease than those <strong>in</strong>volv<strong>in</strong>g object extractions (Brazilian<br />

Portuguese: Gouvea, 2003; Dutch: Frazier, 1987b; English: Ford, 1983, Gibson, Desmet,<br />

Grodner, Watson, & Ko, 2005, K<strong>in</strong>g & Just, 1991, K<strong>in</strong>g & Kutas, 1995, Traxler, Morris, &<br />

Seely, 2002; French: Cohen & Mehler, 1996, Frauenfelder, Segui, & Mehler, 1980, Holmes<br />

& O’Regan, 1981; German: Meckl<strong>in</strong>ger, Schriefers, Ste<strong>in</strong>hauer, & Friederici, 1995,<br />

Schriefers, Friederici, & Kuhn, 1995). In <strong>the</strong> process<strong>in</strong>g of relative clauses that are headf<strong>in</strong>al,<br />

however, <strong>the</strong> effect has been <strong>in</strong>consistent. Easier comprehension of subject relative<br />

clauses has been reported <strong>in</strong> Japanese (Miyamoto & Nakamura, 2003; Ueno & Garnsey,<br />

2008), Korean (Kwon, Lee, Gordon, Kluender, & Pol<strong>in</strong>sky, 2010), Mandar<strong>in</strong> Ch<strong>in</strong>ese (L<strong>in</strong> &<br />

Bever, 2006) and Turkish (Kahraman, Sato, Ono & Sakai, 2010), while easier process<strong>in</strong>g of<br />

object relative clauses has been reported <strong>in</strong> Basque (Carreiras, Duñabeitia, Vergara, de la<br />

Cruz-Pavía, & Laka, 2010), Japanese (Ishizuka, Nakatani, & Gibson, 2006) and Mandar<strong>in</strong><br />

Ch<strong>in</strong>ese (Hsiao & Gibson, 2003; Gibson & Wu, 2011; Packard, Ye, & Zhou, 2011).<br />

Research so far mostly assumed that <strong>in</strong>tegrat<strong>in</strong>g <strong>the</strong> relativized gap and <strong>the</strong> head noun<br />

(i.e., <strong>the</strong> filler) is critical to relative clause process<strong>in</strong>g. A distance-based account has been<br />

adopted to account for <strong>the</strong> subject-object asymmetries <strong>in</strong> relative clause process<strong>in</strong>g across<br />

languages. The distance between <strong>the</strong> filler and <strong>the</strong> gap has been counted l<strong>in</strong>early based on <strong>the</strong><br />

number of <strong>in</strong>terven<strong>in</strong>g discourse entities (Gibson, 1998) and hierarchically based on <strong>the</strong><br />

number of <strong>in</strong>terven<strong>in</strong>g structural nodes (O’Grady, 1997). No <strong>the</strong>ory, however, has been able<br />

to account for <strong>the</strong> variations of <strong>the</strong> effect <strong>in</strong> both head-<strong>in</strong>itial and head-f<strong>in</strong>al relativization.<br />

In this talk, we review previous research (<strong>in</strong>clud<strong>in</strong>g ours) and propose a typological<br />

<strong>the</strong>ory of relative clause process<strong>in</strong>g, pay<strong>in</strong>g special attention to head positions, case mark<strong>in</strong>g,<br />

and <strong>the</strong>matic orders (Figure 1). Languages are first classified based on <strong>the</strong> head positions<br />

<strong>in</strong>side NPs. Filler-gap dependencies (<strong>in</strong> head-<strong>in</strong>itial relatives) and gap-filler dependencies (<strong>in</strong><br />

head-f<strong>in</strong>al relatives) are taken to <strong>in</strong>volve dist<strong>in</strong>ctive process<strong>in</strong>g effects. The word-order<br />

variations with<strong>in</strong> languages with head-<strong>in</strong>itial relatives suggest that when <strong>the</strong> filler precedes<br />

<strong>the</strong> gap, <strong>the</strong> processor adopts an Active Filler Strategy (Frazier, 1987): upon encounter<strong>in</strong>g a<br />

filler, <strong>the</strong> parser <strong>in</strong>itiates <strong>the</strong> search for a gap to m<strong>in</strong>imize <strong>the</strong> work<strong>in</strong>g memory load.<br />

If <strong>the</strong> gap precedes <strong>the</strong> filler as <strong>in</strong> a head-f<strong>in</strong>al type of dependency, <strong>the</strong> construction of<br />

a gap-filler dependency would be complicated by issues of structural garden path regard<strong>in</strong>g<br />

where <strong>the</strong> relative clauses starts and where <strong>the</strong> gap is located (Hirose, 2006; L<strong>in</strong> & Bever,<br />

2011). Strategies that do not focus on construct<strong>in</strong>g a filler-gap relation would be used <strong>in</strong>stead.<br />

In <strong>the</strong> literature, it has been suggested that <strong>in</strong> case-prom<strong>in</strong>ent languages like Basque,<br />

Japanese, Korean, and Turkish, whe<strong>the</strong>r <strong>the</strong> grammatical case is marked determ<strong>in</strong>es<br />

process<strong>in</strong>g difficulty (Carreiras et al., 2010). In a nom<strong>in</strong>ative-accusative language like<br />

Japanese, Korean, and Turkish, an object relative is more difficult than a subject relative<br />

because <strong>the</strong> object of a transitive verb receives an accusative case, which is more marked<br />

than <strong>the</strong> nom<strong>in</strong>ative case of a subject NP. In an ergative language like Basque, <strong>the</strong> subject of<br />

a transitive verb receives <strong>the</strong> ergative case and is more marked than an object NP. Therefore,<br />

a subject relative is more difficult than an object relative.<br />

In a word order prom<strong>in</strong>ent language like Mandar<strong>in</strong> Ch<strong>in</strong>ese, <strong>the</strong> dom<strong>in</strong>ant <strong>the</strong>matic<br />

order functions as a template to be mapped with <strong>in</strong>com<strong>in</strong>g nom<strong>in</strong>al arguments and verbs for<br />

“quick and dirty” <strong>the</strong>matic <strong>in</strong>terpretations (similar to <strong>the</strong> pseudosyntax of Townsend and


Bever, 2001 and <strong>the</strong> good-enough process<strong>in</strong>g of Ferreira, 2003). Sequences of nom<strong>in</strong>al<br />

arguments and verbs that follow <strong>the</strong> dom<strong>in</strong>ant <strong>the</strong>matic orders <strong>in</strong> <strong>the</strong> language (e.g., Agentaction-Patient<br />

<strong>in</strong> Ch<strong>in</strong>ese object relatives) are easier to process than those that do not.<br />

Therefore, a subject relative is easier to process than an object relative (as was found by<br />

Gibson and Wu, 2011).<br />

The proposed typology demonstrates effects and strategies of structural <strong>in</strong>tegration,<br />

<strong>the</strong>matic mapp<strong>in</strong>gs as well as sensitivity to case markedness that h<strong>in</strong>ge on <strong>the</strong> typological<br />

properties of specific languages. We will fur<strong>the</strong>r exam<strong>in</strong>e previous experimental results of L2<br />

relative clause process<strong>in</strong>g and discuss <strong>the</strong> implications that such a typology has for second<br />

language studies of relative clause process<strong>in</strong>g, <strong>in</strong> particular, whe<strong>the</strong>r <strong>the</strong> process<strong>in</strong>g of<br />

relative clauses <strong>in</strong> L1s and L2s of different types would be affected by <strong>the</strong> dist<strong>in</strong>ctive<br />

process<strong>in</strong>g mechanis<strong>ms</strong> <strong>in</strong> ei<strong>the</strong>r language (cf., Juffs, 2005; Clahsen & Felser, 2006, 2008).<br />

Head-<strong>in</strong>itial<br />

Filler-gap <strong>in</strong>tegration<br />

Relative Clauses<br />

+ case-mark<strong>in</strong>g<br />

Morphological<br />

unmarkedness<br />

Figure 1. Typology of relative clause process<strong>in</strong>g<br />

Head-f<strong>in</strong>al<br />

- case-mark<strong>in</strong>g<br />

Thematic template<br />

match<strong>in</strong>g


Bethany Lochbihler (McGill)<br />

F<strong>in</strong>al and non-f<strong>in</strong>al phase status<br />

This paper discusses <strong>in</strong>equality between syntactic phases as f<strong>in</strong>al that show stronger<br />

boundaries versus non-f<strong>in</strong>al phases that <strong>in</strong>stitute a more permeable boundary. F<strong>in</strong>al phases<br />

correspond to top level functional projections <strong>in</strong> <strong>the</strong> spirit of Gri<strong>ms</strong>haw (2000), <strong>in</strong>clud<strong>in</strong>g CP,<br />

DP, PP and AP, and act as absolute boundaries on head movement (Li 1990, 2005) and<br />

restrict scrambl<strong>in</strong>g (Fowlie 2010). Non-f<strong>in</strong>al phases, like category def<strong>in</strong><strong>in</strong>g heads v and n<br />

(Marantz 2001) and <strong>the</strong>ta-complete heads v* (Cho<strong>ms</strong>ky 2004), allow head movement and<br />

scrambl<strong>in</strong>g past <strong>the</strong> phase boundary. I propose that non-f<strong>in</strong>al phases occur with<strong>in</strong> extended<br />

projections, while f<strong>in</strong>al phases mark <strong>the</strong> edge of an extended projection. I <strong>in</strong>vestigate <strong>the</strong><br />

phonological predictions made implicit by <strong>the</strong> syntactic literature, namely that <strong>the</strong> difference<br />

<strong>in</strong> type of phase boundary (as more or less permeable) should affect <strong>the</strong> phonological form<br />

that takes syntactic phase spell-out as <strong>in</strong>put. I test <strong>the</strong> phonological effect of f<strong>in</strong>al phases <strong>in</strong><br />

Ojibwe, a Central Algonquian language spoken <strong>in</strong> Canada and parts of <strong>the</strong> USA, which has<br />

syntactically complex prosodic words exhibit<strong>in</strong>g word-<strong>in</strong>ternal phases. I show that <strong>the</strong> status<br />

of a phase as f<strong>in</strong>al or non-f<strong>in</strong>al bears on prosodic word boundaries, stress assignment and <strong>the</strong><br />

order of application of phonological processes.<br />

At <strong>the</strong> heart of <strong>the</strong> dist<strong>in</strong>ction between f<strong>in</strong>al and non-f<strong>in</strong>al phases is <strong>the</strong><br />

lexical/functional specification of categories. First, Gri<strong>ms</strong>haw (2000) puts forth <strong>the</strong> notion of<br />

extended projection where a lexical head resides with<strong>in</strong> a functional shell (e.g. V with<strong>in</strong> InflP<br />

and CP) but shares its categorical properties with <strong>the</strong>se functional projections. Edges of<br />

extended projections are marked by <strong>the</strong> maximal projection of a top level functional head<br />

(e.g. C for a V head). I want to claim that <strong>the</strong>se top level functional projections <strong>in</strong> fact def<strong>in</strong>e<br />

f<strong>in</strong>al phases because <strong>the</strong>y create a firm boundary between <strong>the</strong> extended doma<strong>in</strong>s of lexical<br />

heads (e.g. between verbs and <strong>the</strong>ir nom<strong>in</strong>al arguments) and show unique properties with<br />

respect to syntactic movement, discussed below.<br />

Second, f<strong>in</strong>al phases <strong>in</strong>teract with head movement. Li (1990, 2005) looks at verb<br />

<strong>in</strong>corporation where a lexical verb comb<strong>in</strong>es with a causative morpheme via head movement,<br />

form<strong>in</strong>g one prosodic word. Li argues that <strong>in</strong>corporation of a lower verb <strong>in</strong>to a higher<br />

causative Vº cannot pass through Inflº and Cº (as was proposed by Baker 1988) s<strong>in</strong>ce<br />

embedded <strong>in</strong>flection is never realized and verbs tak<strong>in</strong>g full clausal complements cannot<br />

undergo verb <strong>in</strong>corporation. Li proposes that movement of a lexical head cannot go through a<br />

functional head back to a lexical one, parallel to improper movement of XPs from A-bar to<br />

A-positions (Cho<strong>ms</strong>ky 1986). Although <strong>the</strong> specifier of a f<strong>in</strong>al phase can act as an escape<br />

hatch for A-bar movement (e.g. long distance wh-front<strong>in</strong>g), Li’s analysis implies that f<strong>in</strong>al<br />

phase heads, such as Cº, are an absolute boundary for head movement (Nºs also cannot move<br />

past Dº) while o<strong>the</strong>r phase heads, such as v*º, do not restrict head movement.<br />

A third set of data show<strong>in</strong>g <strong>the</strong> syntactic uniqueness of f<strong>in</strong>al phases comes from <strong>the</strong><br />

k<strong>in</strong>ds of scrambl<strong>in</strong>g possible cross-l<strong>in</strong>guistically. Fowlie (2010) discusses data from different<br />

languages (e.g. Tagalog, Walpiri and Tohono O’odham), <strong>in</strong>dicat<strong>in</strong>g that scrambl<strong>in</strong>g can occur<br />

with<strong>in</strong> DPs (also APs, PPs) and with<strong>in</strong> CPs, but cannot occur past <strong>the</strong> edge of a DP or CP.<br />

Aga<strong>in</strong> we f<strong>in</strong>d <strong>the</strong> same set of f<strong>in</strong>al phases creat<strong>in</strong>g a stronger boundary, now with respect to<br />

scrambl<strong>in</strong>g, while non-f<strong>in</strong>al phases (xPs, v*P) do not show such strong restrictions.<br />

F<strong>in</strong>al phases (CP, DP, AP, PP) have syntactic status as <strong>the</strong> top level functional<br />

projection <strong>in</strong> <strong>the</strong> extended projection of a lexical head, and non-f<strong>in</strong>al phases (xPs, v*P) are<br />

found <strong>in</strong>ternally to extended projections of lexical heads. However, phases conceptually exist<br />

so that <strong>the</strong> derivation is divided <strong>in</strong>to packets to be sent to <strong>the</strong> phonological and semantic<br />

<strong>in</strong>terfaces of <strong>the</strong> grammar (Cho<strong>ms</strong>ky 2001). On <strong>the</strong> semantic side, f<strong>in</strong>al phases as whole<br />

extended projections are semantically saturated and <strong>the</strong>ta-marked. Next I will discuss some<br />

effects of <strong>the</strong> different phase types beyond <strong>the</strong> syntax at <strong>the</strong> phonological <strong>in</strong>terface.<br />

Ojibwe phonology shows <strong>the</strong> effects of f<strong>in</strong>al phases (versus non-f<strong>in</strong>al) <strong>in</strong> several ways.<br />

First, this polysyn<strong>the</strong>tic language tends to def<strong>in</strong>e <strong>the</strong> edges of prosodic words by f<strong>in</strong>al phases<br />

1


F<strong>in</strong>al and non-f<strong>in</strong>al phase status<br />

(potentially conta<strong>in</strong><strong>in</strong>g <strong>in</strong>termediate phases), such that <strong>in</strong>dependent words correspond to CP,<br />

DP, AP, PP or are elements outside of <strong>the</strong>se doma<strong>in</strong>s, such as quantifiers or discourse<br />

particles <strong>in</strong> <strong>the</strong> specifiers of DP and CP respectively. The spell-out of a f<strong>in</strong>al phase places a<br />

f<strong>in</strong>al prosodic word boundary as well as a syntactic boundary.<br />

Second, stress assignment <strong>in</strong> Ojibwe is dependent on phase doma<strong>in</strong>, where secondary<br />

stress is computed at <strong>the</strong> spell-out of each phase, but primary stress is only computed at f<strong>in</strong>al<br />

phases. Newell (2008) shows <strong>the</strong> sensitivity of secondary stress to word-<strong>in</strong>ternal (non-f<strong>in</strong>al)<br />

phases where degenerate feet are only allowed at <strong>the</strong> right edge of a phase doma<strong>in</strong> (1).<br />

Conversely, primary stress marks <strong>the</strong> antepenultimate foot of <strong>the</strong> entire prosodic word, which<br />

is <strong>the</strong> f<strong>in</strong>al phase CP <strong>in</strong> (1) (bolded).<br />

(1) (bòo)(ní)|(m<strong>in</strong>ì)(kwèe) (*(bóo)(nimì)(nikwèe))<br />

[CP[InflP[aP booni]-[vP m<strong>in</strong>ik-iwee]]]<br />

quit-dr<strong>in</strong>k<strong>in</strong>g-VAI ‘he quit dr<strong>in</strong>k<strong>in</strong>g’ (Newell 2008:214)<br />

F<strong>in</strong>ally, f<strong>in</strong>al phases affect <strong>the</strong> distribution of T-Pal(atalization) (i.e. /t, /→[č,]/__/i/), which<br />

can only occur between segments spelled-out <strong>in</strong> <strong>the</strong> same phase and is blocked across phase<br />

edges (see Slav<strong>in</strong> 2007; Mathieu 2009). T-Pal <strong>in</strong>teracts with vowel Apocope <strong>in</strong> Ojibwe,<br />

which can delete <strong>the</strong> environment for T-Pal, namely /i/ (Piggott & Kaye 1973). (2a) <strong>in</strong>dicates<br />

a T-Pal>Apocope order (or a rank<strong>in</strong>g <strong>in</strong> OT) s<strong>in</strong>ce /t/ becomes [č] despite <strong>the</strong> deletion of /i/,<br />

and (2b) is <strong>the</strong> opposite Apocope>T-Pal order where vowel deletion bleeds palatalization. I<br />

account for this order<strong>in</strong>g paradox by appeal<strong>in</strong>g to f<strong>in</strong>al phases and <strong>in</strong>dicat<strong>in</strong>g that certa<strong>in</strong><br />

phonological processes can only occur at <strong>the</strong> spell-out of a specific type of phase. I propose<br />

that <strong>the</strong> process of T-Pal applies at <strong>the</strong> spell-out of every phase but Apocope only applies at a<br />

f<strong>in</strong>al phase spell-out (account<strong>in</strong>g for Apocope exclusively delet<strong>in</strong>g actual word f<strong>in</strong>al<br />

segments, and not <strong>in</strong>termediate f<strong>in</strong>al segments cross-l<strong>in</strong>guistically, e.g. Hale 1973).<br />

Therefore, <strong>in</strong> (2b) both T-Pal and Apocope are triggered at a f<strong>in</strong>al phase, and so Apocope can<br />

bleed T-Pal, but <strong>in</strong> (2a) T-Pal is triggered <strong>in</strong> an earlier v*P phase and Apocope must wait<br />

until <strong>the</strong> spell-out of <strong>the</strong> f<strong>in</strong>al CP phase, giv<strong>in</strong>g <strong>the</strong> apparent T-Pal>Apocope order<strong>in</strong>g.<br />

(2) a. nimiškawač b. peemaatisit<br />

[CP ni-[v*P miškaw-a-t-i]] [[peem-aat-isi]-t-i]<br />

1-freeze/hard-VII-<strong>in</strong>trans-VAI along-live-VAI -3conj-PART<br />

‘I am frozen.’ ‘he who lives’ (Kaye & Piggott 1973:356)<br />

The idea that CP and DP (as well as AP and PP) constitute special doma<strong>in</strong>s is implicit <strong>in</strong> <strong>the</strong><br />

syntactic literature, a property I aim to make explicit with <strong>the</strong> notion of f<strong>in</strong>al phases def<strong>in</strong>ed<br />

by extended projection of a lexical category. The effects of phases are seen <strong>in</strong> <strong>the</strong> syntax as<br />

well as <strong>the</strong> phonology, and <strong>the</strong> dist<strong>in</strong>ction between f<strong>in</strong>al and non-f<strong>in</strong>al phases likewise affects<br />

both parts of <strong>the</strong> grammar.<br />

References: Baker, M. 1988. Incorporation. UChicago Press. | Cho<strong>ms</strong>ky, N. 1986. Barriers.<br />

MIT Press. | Cho<strong>ms</strong>ky, N. 2004. Beyond explanatory adequacy. In Belletti et al (eds), The<br />

cartography of syntactic structures. OUP. | Fowlie, M. 2010. More Multiple Multiple Spellout.<br />

Ms. UCLA. | Gri<strong>ms</strong>haw, J. 2000. Locality and Extended Projection. In Coopmans et al (eds),<br />

Lexical Specification and Insertion. Benjam<strong>in</strong>s. | Hale, K. 1973. Deep and surface canonical<br />

disparities <strong>in</strong> relation to analysis and change. Current Trends <strong>in</strong> <strong>L<strong>in</strong>guistics</strong> 11: 408-458. | Kaye,<br />

J. D., and Piggott, G. L. 1973. On <strong>the</strong> Cyclical Nature of Ojibwa T-Palatalization. LI 4:345-362. |<br />

Li, Y. 1990. X 0 -b<strong>in</strong>d<strong>in</strong>g and Verb Incorporation. LI 21:399-426. | Li, Y. 2005. Xº: A Theory of <strong>the</strong><br />

Morphology-Syntax Interface. MIT Press. | Marantz, Alec. 2001. Words. Ms. MIT. | Mathieu, E.<br />

2009. Noun Incorporation and Word Formation via Phrasal Movement. Ms. UOttawa. | Newell,<br />

H. 2008. Aspects of <strong>the</strong> morphology and phonology of phases. PhD Dissertation, McGill. |<br />

Slav<strong>in</strong>, T. 2007. T-Palatalization <strong>in</strong> Oji-Cree as a w<strong>in</strong>dow <strong>in</strong>to <strong>the</strong> morphosyntactic structure of<br />

<strong>the</strong> verbal complex. Generals Paper, UToronto.<br />

2


Terje Lohndal (Maryland) & Bridget Samuels (CalTech)<br />

On how null elements and unpronounced copies are different<br />

Syntax creates sound-mean<strong>in</strong>g pairs. With<strong>in</strong> Cho<strong>ms</strong>kyan generative grammar, this is<br />

typically described <strong>in</strong> ter<strong>ms</strong> of transferr<strong>in</strong>g (portions of) syntactic structures to <strong>the</strong><br />

phonological/phonetic and semantic/conceptual <strong>in</strong>terfaces through PF and LF, respectively.<br />

However, to paraphrase Cho<strong>ms</strong>ky & Halle (1968), whe<strong>the</strong>r <strong>the</strong> output of <strong>the</strong> syntactic<br />

component and <strong>the</strong> <strong>in</strong>put of <strong>the</strong> phonological component are <strong>the</strong> same th<strong>in</strong>g is an empirical<br />

issue. Indeed, s<strong>in</strong>ce <strong>the</strong> 1970’s, much work on <strong>the</strong> syntax-phonology <strong>in</strong>terface has focused on<br />

how to derive phonological doma<strong>in</strong>s from syntactic structures under <strong>the</strong> assumption that<br />

<strong>the</strong>se two representations are related but non-isomorphic. This issue of doma<strong>in</strong>s is, however,<br />

but one piece of <strong>the</strong> much larger puzzle of how to characterize <strong>the</strong> transformations at PF<br />

which turn hierarchical, phonology-free morphosyntactic structures <strong>in</strong>to l<strong>in</strong>ear phonological<br />

representations. The purpose of this talk is to shed light on <strong>the</strong> order<strong>in</strong>g of operations at PF,<br />

and crucially argue that empty categories <strong>in</strong> <strong>the</strong> syntax can sometimes create a configuration<br />

that is not l<strong>in</strong>earizable. Our goal <strong>in</strong> this paper is to scrut<strong>in</strong>ize <strong>the</strong> conditions under which such<br />

configurations arise, and <strong>in</strong>tegrate <strong>the</strong>se with more general observations about <strong>the</strong> effects of<br />

and differences between unpronounced copies and null elements <strong>in</strong> morphophonology.<br />

The specific case study that we will focus on relates to ‘empty edges’: phonologically<br />

contentless edges of certa<strong>in</strong> doma<strong>in</strong>s. Whe<strong>the</strong>r to characterize <strong>the</strong>se doma<strong>in</strong>s <strong>in</strong> syntactic or<br />

phonological ter<strong>ms</strong> will be one of our primary concerns. Towards this goal, we provide a new<br />

analysis of facts traditionally attributed to <strong>the</strong> ECP (e.g., Stowell 1981; cf. Pesetsky &<br />

Torrego 2001, Bošković & Lasnik 2003). An (2007a) suggests that <strong>the</strong>se data are captured by<br />

<strong>the</strong> Intonational Phrase Edge Generalization (IPEG), which states that “<strong>the</strong> edge of an<br />

Intonational phrase cannot be empty (where <strong>the</strong> edge encompasses <strong>the</strong> specifier and <strong>the</strong> head<br />

of <strong>the</strong> relevant syntactic constituent).” This expla<strong>in</strong>s why sentences which are unacceptable<br />

with an empty CP edge can be ameliorated by overt content, be it <strong>in</strong> SpecCP or <strong>in</strong> C (1).<br />

(1) a. *I saw <strong>the</strong> child yesterday [Øspec ØC Mary was wait<strong>in</strong>g for]<br />

b. I saw <strong>the</strong> child yesterday [who ØC Mary was wait<strong>in</strong>g for]<br />

c. I saw <strong>the</strong> child yesterday [Øspec that Mary was wait<strong>in</strong>g for]<br />

In order to allow for <strong>the</strong> omission of C <strong>in</strong> (2b), An (2007a) needs to assume that <strong>the</strong><br />

formation of a separate I-phrase for clausal complements of verbs (and restrictive relative<br />

clauses) is optional. The IPEG also extends to o<strong>the</strong>r categories: vP (3a,b), DP (3c,d), and AP<br />

(3e,f).<br />

(2) a. I believe [CP that [IP John liked l<strong>in</strong>guistics]].<br />

b. I believe [CP ØC [IP John liked l<strong>in</strong>guistics.]]<br />

(3) a. Eat <strong>the</strong> cake John did and eat <strong>the</strong> cookie Mary did<br />

b. *[vP Eat <strong>the</strong> cake] John did and [Øspec Øv <strong>the</strong> cookie] Mary did<br />

c. John likes this book of l<strong>in</strong>guistics and Mary, that book of physics.<br />

d. *John likes this book of l<strong>in</strong>guistics and Mary likes [Øspec ØD book of physics.<br />

e. Eager to w<strong>in</strong> <strong>the</strong> Pulitzer Prize, John is, and eager to w<strong>in</strong> <strong>the</strong> Nobel Prize, Mary is.<br />

f. *[AP Eager to w<strong>in</strong> <strong>the</strong> Pulitzer Prize], John is, and [Øspec ØA to w<strong>in</strong> <strong>the</strong> Nobel Prize,<br />

Mary is].<br />

However, <strong>the</strong> justification for <strong>the</strong> IPEG is unclear and does not follow from <strong>in</strong>dependent facts<br />

about <strong>the</strong> architecture of grammar. Moreover, it requires syntactic analyses which violate<br />

Bare Phrase Structure because An is forced to stipulate empty specifiers, which Bare Phrase<br />

structure rejects (cf. Cho<strong>ms</strong>ky 1995, Alexiadou & Anagnostopoulou 1998). We suggest a<br />

new analysis which accounts for <strong>the</strong> above data by referr<strong>in</strong>g to <strong>in</strong>dependently needed<br />

constructs, which we argue should be <strong>the</strong> null hypo<strong>the</strong>sis given M<strong>in</strong>imalist concerns about<br />

modular architecture (cf. Idsardi and Raimy <strong>in</strong> press).<br />

We argue for a syntactic—or at least, ‘pre-phonological’—account of empty-edge<br />

phenomena which h<strong>in</strong>ges on exactly how we understand <strong>the</strong> PF <strong>in</strong>terface. We will make a


specific proposal concern<strong>in</strong>g this architecture, show how this new account can capture all <strong>the</strong><br />

data that have been discussed <strong>in</strong> <strong>the</strong> literature, and extend it to cover additional data which<br />

have not previously been addressed <strong>in</strong> treatments of <strong>the</strong> ECP. We argue that (1a) and (3b,d,f)<br />

are unacceptable because <strong>the</strong>y cannot be l<strong>in</strong>earized. Specifically, two consecutive syntactic<br />

objects cannot be null at <strong>the</strong> stage when l<strong>in</strong>earization applies, or else <strong>the</strong> algorithm will be<br />

unable to return a l<strong>in</strong>earization statement, cf. (4).<br />

(4) L<strong>in</strong>earization algorithm: When encounter<strong>in</strong>g <strong>the</strong> merged {α, β}, α and β ccommand<strong>in</strong>g<br />

each o<strong>the</strong>r, upon Spell-Out, return an ordered set or .<br />

We follow Epste<strong>in</strong> et al. (1998) and Richards (2004) <strong>in</strong> assum<strong>in</strong>g that mutual ccommand<br />

‘overdeterm<strong>in</strong>es’ l<strong>in</strong>earization; <strong>the</strong> Precedence Resolution Pr<strong>in</strong>ciple demands that<br />

<strong>in</strong> such a configuration, one object’s c-command relations over <strong>the</strong> o<strong>the</strong>r must be ignored<br />

(which object be<strong>in</strong>g subject to parametric variation). The key component to our analysis of<br />

(1)-(3) is that <strong>the</strong> algorithm (4) only succeeds <strong>in</strong> return<strong>in</strong>g an ordered pair when <strong>the</strong> elements<br />

to be l<strong>in</strong>earized are featurally dist<strong>in</strong>ct. They have to be dist<strong>in</strong>ct because of <strong>the</strong> irreflexivity<br />

condition on l<strong>in</strong>earization, as Nunes (1995, 2004) po<strong>in</strong>ts out. If <strong>the</strong> two elements are not<br />

featurally dist<strong>in</strong>ct, this results <strong>in</strong> a crash when a merged pair of elements has only Edge<br />

Features (EF; Cho<strong>ms</strong>ky 2008), as a result of non-<strong>in</strong>sertion of phonological content (which is<br />

crucially dist<strong>in</strong>ct from copy deletion, contra An 2007b.) Two elements with only EF are nondist<strong>in</strong>ct,<br />

which will result <strong>in</strong> a crash at l<strong>in</strong>earization. EFs are visible to <strong>the</strong> l<strong>in</strong>earization<br />

algorithm s<strong>in</strong>ce <strong>the</strong>se features are never deleted, as Cho<strong>ms</strong>ky (2008) argues, but <strong>the</strong>y are also<br />

never phonologically realized. Fur<strong>the</strong>rmore, we provide evidence that copies are marked for<br />

deletion prior to l<strong>in</strong>earization but actually deleted after l<strong>in</strong>earization, which exempts<br />

unpronounced copies from caus<strong>in</strong>g empty-edge crashes and resolves a tension between <strong>the</strong><br />

empty-edge bans <strong>in</strong> An (2007a) and (2007b), one of which is asymmetrical (left edge only)<br />

but <strong>in</strong>cludes copies, and <strong>the</strong> o<strong>the</strong>r which is symmetrical but excludes copies.<br />

Assum<strong>in</strong>g <strong>in</strong> accordance with Bare Phrase Structure that C does not project a specifier<br />

<strong>in</strong> declarative sentences such as (5a) [or <strong>in</strong> (2b)] (cf. Starke 2004), our account<br />

straightforwardly predicts <strong>the</strong> sentences <strong>in</strong> (5) to be grammatical, while such cases constitute<br />

prima facie exceptions to <strong>the</strong> IPEG.<br />

(5) a. [CP ØC [TP Mary was wait<strong>in</strong>g for <strong>the</strong> child]]<br />

b. [CP Who did Mary [vP see [VP tv tobj ]]]<br />

The same holds for adjuncts to matrix clauses, where <strong>the</strong>re is no null element <strong>in</strong>terven<strong>in</strong>g:<br />

(6) a. [CP [AP True to herself], [TP she planned to rema<strong>in</strong> <strong>the</strong>re ]]. (Selkirk 2005)<br />

b. [CP [PP On <strong>the</strong> fourth of July], [TP we’ll have a parade and fireworks]].<br />

We fur<strong>the</strong>r show that <strong>the</strong> above analysis extends to account for <strong>the</strong> unacceptable<br />

sentences <strong>in</strong> (3) as well as those <strong>in</strong> (7), <strong>in</strong>volv<strong>in</strong>g a sentential subject (7b) and a topicalized<br />

CP (7d). In both <strong>the</strong> unacceptable cases, <strong>the</strong>re will be two objects that only have EFs that are<br />

adjacent, thus <strong>the</strong> l<strong>in</strong>earization algorithm crashes.<br />

(7) a. [CP ØC [CP That John is a genius] was believed by many people.<br />

b. *[CP ØC [CP ØC John is a genius] was believed by many people.<br />

c. [CP ØC [CP That John is a genius], Mary believed.<br />

d. *[CP ØC [CP ØC John is a genius], Mary believed.<br />

We will address fur<strong>the</strong>r cases as well, and <strong>the</strong>n turn to o<strong>the</strong>r data concern<strong>in</strong>g how<br />

morphophonology deals with null elements. We argue that morphophonology ‘knows about’<br />

null elements s<strong>in</strong>ce <strong>the</strong>y trigger phonological cycles and produce cyclic block<strong>in</strong>g effects, but<br />

that unpronounced copies are phonologically <strong>in</strong>ert (Samuels 2011). This streng<strong>the</strong>ns our<br />

claim that unpronounced copies and null elements must be dist<strong>in</strong>guished, and we argue that<br />

<strong>the</strong> former are elim<strong>in</strong>ated earlier than <strong>the</strong> latter – null elements actually persist until <strong>the</strong><br />

handoff to phonology proper.


Sophia Manika, Sergey Avrut<strong>in</strong> & Eric Reuland (Utrecht)<br />

The bits of dependencies<br />

Background<br />

Consider a m<strong>in</strong>imally different pair:<br />

1. Maria blames, <strong>in</strong> most of <strong>the</strong> cases, herself for <strong>the</strong> bad grades<br />

2. Maria blames, <strong>in</strong> most of <strong>the</strong> case, <strong>the</strong> teacher for <strong>the</strong> bad grades<br />

Interpretation of <strong>the</strong> first sentence requires establish<strong>in</strong>g a syntactic dependency between an<br />

anaphoric element and its antecedent. Interpretation of <strong>the</strong> second sentence requires<br />

<strong>in</strong>tegration of def<strong>in</strong>ite NP <strong>the</strong> teacher <strong>in</strong>to discourse as part of <strong>the</strong> blam<strong>in</strong>g event.<br />

Accord<strong>in</strong>g to Reuland's model of <strong>the</strong> language - relevant architecture of work<strong>in</strong>g memory<br />

(Reuland 2009, 2010), <strong>the</strong>se two operations nicely reflect two dist<strong>in</strong>ct memory networks<br />

<strong>in</strong>volved <strong>in</strong> language process<strong>in</strong>g. In Reuland's term<strong>in</strong>ology one computes form/syntax (socalled<br />

Declarative / Procedural Interface, or DPI) and one is responsible for <strong>the</strong> <strong>in</strong>tegration of<br />

<strong>the</strong> content/discourse <strong>in</strong>formation.<br />

The current study provides experimental evidence for <strong>the</strong> existence of <strong>the</strong>se two dist<strong>in</strong>ct<br />

syste<strong>ms</strong> and proposes a novel measure of process<strong>in</strong>g complexity. Specifically, we<br />

demonstrate that it is <strong>the</strong> entropy of <strong>the</strong> verbal paradigm (an <strong>in</strong>formation - <strong>the</strong>oretic<br />

notion <strong>in</strong>troduced by Shannon 1948) that <strong>in</strong>fluences <strong>the</strong> operation of <strong>the</strong> relevant<br />

memory syste<strong>ms</strong>. In o<strong>the</strong>r words, we demonstrate that changes <strong>in</strong> <strong>the</strong> entropy value of<br />

<strong>the</strong> verb family <strong>in</strong>fluence (crucially, <strong>in</strong> a different way!) <strong>the</strong> process<strong>in</strong>g cost of <strong>the</strong><br />

reflexive object NP and a def<strong>in</strong>ite object NP.<br />

Entropy calculation<br />

Inflectional entropy of a verb’s paradigm <strong>in</strong>dicates <strong>the</strong> level of similarity 1 among its ite<strong>ms</strong>. It<br />

is a function of <strong>the</strong> number of <strong>in</strong>flected for<strong>ms</strong> of <strong>the</strong> verb, <strong>the</strong> frequency of each verb form<br />

and <strong>the</strong> number of possible l<strong>in</strong>guistic <strong>in</strong>terpretations <strong>the</strong> form can be found <strong>in</strong> (i.e. <strong>the</strong> verb<br />

form “work” <strong>in</strong> English can be used as 1 st sg, 2 nd sg, <strong>in</strong>f<strong>in</strong>itive etc). In this sense, higher<br />

entropy reflects a more “uniform” distribution of memory traces between for<strong>ms</strong> (hence,<br />

higher uncerta<strong>in</strong>ty <strong>in</strong> lexical retrieval), while a more “diverse” distribution is represented by a<br />

lower entropy (and thus by a lower uncerta<strong>in</strong>ty <strong>in</strong> retrieval). Van Ewijk & Avrut<strong>in</strong> (2011,<br />

among o<strong>the</strong>rs) showed that <strong>in</strong>flectional entropy of a verb modulated retrieval of a target form<br />

and suggest that entropy is an <strong>in</strong>dex of complexity.<br />

Method<br />

Thirty-four Dutch native students, aged 21-29 years old, were tested <strong>in</strong> a self-paced read<strong>in</strong>g<br />

task. <strong>Read<strong>in</strong>g</strong> times <strong>in</strong> each word were measured.<br />

Twenty-four Dutch verbs were chosen based on <strong>the</strong>ir entropy values (taken from van Ewijk &<br />

Avrut<strong>in</strong>, 2011) and were classified <strong>in</strong> two groups: low and high entropy. Each verb was used<br />

<strong>in</strong> two conditions; with a syntactic dependency (a reflexive pronoun) and without (proper<br />

name), result<strong>in</strong>g to forty-eight experimental ite<strong>ms</strong>. Although <strong>the</strong> critical regions are <strong>the</strong> po<strong>in</strong>t<br />

after <strong>the</strong> verb (rg 2) and <strong>the</strong> po<strong>in</strong>t of object <strong>in</strong>tegration (rg 7), we <strong>in</strong>cluded subsequent regions<br />

(rg 8- 11) to account for spillover effects. All words <strong>in</strong> regions 3-11, (except for <strong>the</strong> proper<br />

name), were identical ensur<strong>in</strong>g weighed <strong>in</strong>formation load across ite<strong>ms</strong> (see Table below).<br />

1 Shannon’s notion of Entropy was proposed to apply to a random variable X to describe <strong>the</strong> degree of<br />

uncerta<strong>in</strong>ty related to <strong>the</strong> distribution of probabilities of <strong>the</strong> variable. For simplicity reasons we refer to Entropy<br />

as degree of similarity because we are study<strong>in</strong>g <strong>in</strong>flectional families and when entropy is applied it mirrors how<br />

different <strong>the</strong> verb-for<strong>ms</strong> with<strong>in</strong> <strong>the</strong> paradig<strong>ms</strong> are.


He/She Verb, <strong>in</strong> <strong>the</strong> most cases, HIM(HER)SELF/ Loes and not <strong>the</strong> o<strong>the</strong>r (noun)<br />

Reflexive<br />

NP<br />

HIJ/ZIJ verb <strong>in</strong> de meeste gevallen<br />

ZICHZELF<br />

Loes<br />

en niet de andere<br />

vary<strong>in</strong>g entropy constant entropy object spill over<br />

region 1 2 3 4 5 6 7 8 9 10 11<br />

Results<br />

At <strong>the</strong> po<strong>in</strong>t of verb retrieval (rg 3), high entropy verbs were read significantly faster than low<br />

entropy ones (p< .05, r=.35).<br />

At <strong>the</strong> po<strong>in</strong>t of <strong>the</strong> object (rg7), <strong>the</strong>re was a significant <strong>in</strong>teraction between type of object and<br />

verb entropy (p< 0.5, r=.43). Importantly, contrasts revealed that <strong>the</strong> <strong>in</strong>teraction was due to<br />

<strong>the</strong> reflexives; higher entropy verbs delayed establish<strong>in</strong>g <strong>the</strong> dependency between <strong>the</strong><br />

reflexive and its referent.<br />

Moreover, <strong>the</strong>re was a ma<strong>in</strong> effect of object type <strong>in</strong> rg8. Reflexives were read significantly<br />

faster across conditions than referential NPs (p< .001, r=.56).<br />

Discussion<br />

These data show how entropy can <strong>in</strong>fluence memory processes. Fur<strong>the</strong>rmore, <strong>the</strong>y show that<br />

<strong>the</strong> establishment of a dependency between a reflexive and its antecedent is not only<br />

grammatically constra<strong>in</strong>ed, but it is <strong>in</strong>fluenced by non-grammatical constra<strong>in</strong>ts such as verbal<br />

entropy as well.<br />

Activation of a verb form automatically activates its paradigm network. The more “uniform”<br />

<strong>the</strong> probability distribution of <strong>the</strong> verb for<strong>ms</strong> with<strong>in</strong> <strong>the</strong> paradigm is, <strong>the</strong> higher is <strong>the</strong><br />

<strong>in</strong>flectional entropy. Verbs with high entropy lose less energy than low entropy verbs dur<strong>in</strong>g<br />

<strong>the</strong> spread of activation, and retrieval is faster (as <strong>in</strong> van Ewijk & Avrut<strong>in</strong>, replicated here <strong>in</strong><br />

rg3). The reflexive cases signal <strong>the</strong> need for an <strong>in</strong>tra-sentential referent and <strong>the</strong> verb is readdressed.<br />

High <strong>in</strong>flectional entropy <strong>in</strong>duces competition between <strong>the</strong> target verb form and<br />

<strong>the</strong> “yet-not-decayed” for<strong>ms</strong> of <strong>the</strong> paradigm that have similar probability. This<br />

“<strong>in</strong>terference”, present only <strong>in</strong> <strong>in</strong>flectional paradig<strong>ms</strong> with high entropy, delays <strong>the</strong><br />

establish<strong>in</strong>g of <strong>the</strong> dependency between <strong>the</strong> reflexive and its antecedent (<strong>in</strong>teraction <strong>in</strong> rg 7).<br />

Additionally, <strong>the</strong> effect at region 8 <strong>in</strong>dicates that grammatically-constra<strong>in</strong>ed dependencies<br />

(reflexives) are established faster than non-grammatically constra<strong>in</strong>ed dependencies (noun).<br />

That implies <strong>the</strong> existence of a dist<strong>in</strong>ct network that computes, very fast, syntactic<br />

requirements, such as assignment of an <strong>in</strong>tra-sentential referent. Delay of <strong>the</strong> nouns results<br />

from <strong>the</strong> fact that <strong>in</strong>formation from discourse is needed for a referent to be found.<br />

References:<br />

Reuland, E. (2009). Language, Symbolization and beyond. In R. B. Knight. (Ed.), The<br />

Prehistory of Language. Oxford: Oxford University Press.<br />

Reuland, E. (2010). Imag<strong>in</strong>ation, plann<strong>in</strong>g and work<strong>in</strong>g memory: <strong>the</strong> emergence of language.<br />

Current Anthropology, 51, 99-110.<br />

Shannon, C. E. (1948). A ma<strong>the</strong>matical <strong>the</strong>ory of communication. Bell System Technical<br />

Journal , 27, 379–423.<br />

van Ewijk, L., & Avrut<strong>in</strong>, S. (2011). Auditory lexical decision <strong>in</strong> healthy elderly and young<br />

subjects.The effect of <strong>in</strong>formation load and <strong>in</strong>flectional entropy. Procedia- Social and<br />

Behavioural Sciences, 23, 104-105.


Rita Manz<strong>in</strong>i (Firenze) & Anna Roussou (Patras)<br />

Empty categories: empty operators and variables at <strong>the</strong> LF <strong>in</strong>terface<br />

We argue for a restricted <strong>in</strong>ventory of <strong>in</strong>terpretable non-pronounced elements of grammar,<br />

<strong>in</strong>clud<strong>in</strong>g only empty operators and variables. Our model observes m<strong>in</strong>imalist postulates<br />

(Inclusiveness, one level of syntactic representation, etc.), but is representational, s<strong>in</strong>ce we ma<strong>in</strong>ta<strong>in</strong><br />

that operators and variables are <strong>in</strong>troduced by <strong>in</strong>terpretive pr<strong>in</strong>ciples at <strong>the</strong> LF <strong>in</strong>terface.<br />

1. Phrasal ‘trace’ and ‘PRO’ are (just) variables at <strong>the</strong> LF <strong>in</strong>terface. In standard m<strong>in</strong>imalism,<br />

movement is Internal Merge (IM): a Probe α with an EPP feature targets a Goal β already merged <strong>in</strong><br />

<strong>the</strong> derivation, as <strong>in</strong> (1). IM is construed as a mechanism that <strong>in</strong>troduces a second copy of β. If all<br />

<strong>in</strong>stances of ‘movement’ are overt (Cho<strong>ms</strong>ky 2000), <strong>the</strong>n <strong>the</strong> EPP can be viewed as an <strong>in</strong>struction<br />

for lexicalization.<br />

(1) [αP β αEPP [… [ γP γ.. β]]]<br />

Alternatively, <strong>in</strong> multi-dom<strong>in</strong>ance (MD) approaches (Starke 2001; Frampton 2004; Johnson 2009;<br />

de Vries 2009; a.o.) β merges once (externally, EM), and ‘re-merge’ amounts to immediate<br />

dom<strong>in</strong>ance by a higher node, so that β has more than one mo<strong>the</strong>r node, as <strong>in</strong> (2). (2) allows for<br />

spell-out of β <strong>in</strong> <strong>the</strong> higher or <strong>the</strong> lower position, subject to some PF-algorithm.<br />

(2) [αP α(EPP) [… [γP γ.. β]]]<br />

(1) and (2) share <strong>the</strong> requirement that β merge <strong>in</strong> <strong>the</strong> lower position for argument structure<br />

satisfaction, at least if β is an argument, i.e. EM expresses <strong>the</strong> core properties of D-structure.<br />

A different logical possibility is that Merge is always external, with β directly merg<strong>in</strong>g <strong>in</strong> <strong>the</strong><br />

EPP position; <strong>the</strong>re is a s<strong>in</strong>gle copy/occurrence and no re-merge, as <strong>in</strong> (3). S<strong>in</strong>ce <strong>the</strong> position of<br />

Merge <strong>in</strong> (3) is also that of spell-out, <strong>the</strong>re is no need for a PF-deletion operation, as <strong>in</strong> (1), or<br />

l<strong>in</strong>earization algorithm as <strong>in</strong> (2). In ter<strong>ms</strong> of <strong>in</strong>terpretation, though, β associates with some lower<br />

part of <strong>the</strong> configuration, here <strong>in</strong>dicated as e. Therefore we need to consider what e amounts to –<br />

and what ensures cha<strong>in</strong> formation between β and e.<br />

(3) [αP β αEPP [… [ γP γ .. (e) ]]]<br />

In standard generative <strong>the</strong>ory, if β is an argument, <strong>the</strong>n e corresponds to its <strong>the</strong>ta-position – whe<strong>the</strong>r<br />

e is a copy or its GB counterpart, i.e. <strong>the</strong> identity element e of <strong>the</strong> str<strong>in</strong>g. But suppose <strong>the</strong> <strong>the</strong>ta-role<br />

has no structural correlate. This means that <strong>the</strong> cha<strong>in</strong> should be computed at <strong>the</strong> <strong>in</strong>terface by some<br />

<strong>in</strong>terpretive algorithm.<br />

For <strong>the</strong> sake of concreteness consider <strong>the</strong> wh-construction <strong>in</strong> (4). In <strong>the</strong> copy-<strong>the</strong>ory of<br />

movement, <strong>the</strong> second gap e2 <strong>in</strong> (4) is a copy of what which deletes at PF, while LF converts it to a<br />

variable (cf. Fox 2002). The first gap e1 is a copy of you, which aga<strong>in</strong> deletes, while <strong>the</strong> PFrealization<br />

co<strong>in</strong>cides with that of <strong>the</strong> EPP argument (syntactic subject). If e, now understood as a<br />

variable, is not structurally represented (as a copy or some o<strong>the</strong>r empty category), as we propose<br />

here, it must be <strong>in</strong>troduced directly at <strong>the</strong> LF <strong>in</strong>terface.<br />

(4) [C What did [I you [V (e1) see (e2) ]]]<br />

Indeed see, be<strong>in</strong>g a two-place predicate, <strong>in</strong>troduces two variables (x, y) which are bound by <strong>the</strong><br />

external and <strong>the</strong> <strong>in</strong>ternal arguments respectively (by <strong>the</strong> λ-b<strong>in</strong>d<strong>in</strong>g mechanism of Heim & Kratzer<br />

1997, cf. Adger & Ramchand’s (2005) Λ feature, etc.). This (operator, variable) relation is fully<br />

equivalent to <strong>the</strong> movement operation – and it is redundant when added to it (Brody’s (2003)<br />

argument for representationalism). In an A’-dependency, <strong>the</strong> wh-operator is <strong>in</strong> turn construed as a<br />

λ-abstractor, and so can be construed <strong>the</strong> EPP <strong>in</strong> relation to <strong>the</strong> subject, follow<strong>in</strong>g Butler (2004).<br />

In short, a ‘movement’ relation is def<strong>in</strong>ed by a s<strong>in</strong>gle argument satisfy<strong>in</strong>g two <strong>in</strong>stances of<br />

λs. Though control will not be considered here <strong>in</strong> any detail, we notice that <strong>the</strong> same also holds of<br />

(obligatory) control, which m<strong>in</strong>imally <strong>in</strong>volves two <strong>the</strong>ta-roles (λ-abstractors) bound by <strong>the</strong> same<br />

argument. In ‘arbitrary’ control, one variable is simply bound by a generic closure operator. Hence<br />

PRO, like trace, <strong>in</strong> present ter<strong>ms</strong> translates to a variable at <strong>the</strong> LF <strong>in</strong>terface.<br />

2. There are no head ‘trace’ and pro. The empirical relevance of our discussion becomes more<br />

evident when we move from <strong>the</strong> dislocation of phrases (specifically DP’s) to that of heads<br />

(specifically verbal heads). Cho<strong>ms</strong>ky (2001) excludes that IM applies to heads. If on <strong>the</strong> o<strong>the</strong>r hand,


head movement is a PF operation, as Cho<strong>ms</strong>ky proposes, <strong>the</strong>n <strong>in</strong> syntax heads always sit <strong>in</strong> <strong>the</strong>ir<br />

EM position, conceived of as <strong>the</strong>ir D-structure position. At <strong>the</strong> same time it is not clear what PF<br />

movement amounts to. A possible construal for it is Morphological Merger (Marantz 1988, Embick<br />

and Noyer 2001). However, to <strong>the</strong> extent that similar Merge operations occur <strong>in</strong> different<br />

components, obvious notions of economy (e.g. non redundancy) appear to be violated.<br />

The discussion <strong>in</strong> part 1 suggests ano<strong>the</strong>r logical possibility, namely that <strong>in</strong> syntax heads<br />

<strong>in</strong>deed occupy <strong>the</strong>ir EM position which is now identified with <strong>the</strong>ir surface position. Consider for<br />

example V-to-C ‘movement’ (or V-<strong>in</strong>-C position<strong>in</strong>g) as <strong>in</strong> German (5). The Merge position of <strong>the</strong><br />

verb <strong>in</strong> (5) will be determ<strong>in</strong>ed by <strong>the</strong> same k<strong>in</strong>d of considerations as <strong>the</strong> Merge of <strong>the</strong> wh-phrase <strong>in</strong><br />

(4), i.e. <strong>the</strong> (‘EPP’) <strong>in</strong>struction to lexicalize certa<strong>in</strong> properties (<strong>the</strong> wh-property or <strong>the</strong> syntactic<br />

subject EPP <strong>in</strong> (4), <strong>the</strong> ‘f<strong>in</strong>iteness’ property <strong>in</strong> (5)).<br />

(5) [CP Gestern sahst [IP du den Peter ]] (lit: ‘yesterday saw you Peter’)<br />

At <strong>the</strong> same time, <strong>the</strong>re is no evidence that <strong>the</strong>re are head variables <strong>in</strong> <strong>the</strong> putative Dstructure<br />

position. The <strong>in</strong>terpretive question <strong>the</strong>n is: how can predicate-argument relations be<br />

established? When <strong>the</strong> argument is <strong>in</strong> a position to b<strong>in</strong>d <strong>the</strong> variable <strong>in</strong>troduced by <strong>the</strong> predicate, <strong>the</strong><br />

discussion goes through as <strong>in</strong> (4). So <strong>the</strong> question is what happens when <strong>the</strong> reverse configuration<br />

holds, as <strong>in</strong> (5) (or <strong>in</strong> any declarative sentence <strong>in</strong> languages with V <strong>in</strong> I, etc.). Given <strong>the</strong> availability<br />

of overt agreement between predicates and <strong>the</strong>ir arguments, we have evidence that functional heads<br />

must have ‘probes’ reach<strong>in</strong>g <strong>in</strong>to argument positions <strong>the</strong>ir ‘goals’), yield<strong>in</strong>g standard m<strong>in</strong>imalist<br />

Agree – which we could use to this end. Yet Cho<strong>ms</strong>ky (2000) argues that <strong>the</strong>ta-roles are not<br />

features and <strong>the</strong>refore <strong>the</strong>ta-relations cannot be established by feature-check<strong>in</strong>g under Agree (contra<br />

Hornste<strong>in</strong> 1999, Manz<strong>in</strong>i & Roussou 2000).<br />

We take Cho<strong>ms</strong>ky’s (2000) conclusion to be correct, but we also argue that agreement is not<br />

a feature check<strong>in</strong>g relation. Consider <strong>the</strong> classical example of subject-f<strong>in</strong>ite verb agreement, say<br />

between du and –st <strong>in</strong> (5). It is a traditional idea (see <strong>the</strong> null subject parameter) that <strong>the</strong> f<strong>in</strong>ite verb<br />

<strong>in</strong>flection is pronom<strong>in</strong>al(-like), hav<strong>in</strong>g all of <strong>the</strong> crucial properties of a pronoun, for <strong>in</strong>stance<br />

reference to ‘hearer’ <strong>in</strong> <strong>the</strong> case of –st ‘2 nd sg’. If so, agreement need not <strong>in</strong>volve <strong>the</strong> deletion of<br />

un<strong>in</strong>terpretable features. Ra<strong>the</strong>r, <strong>the</strong> ‘hearer’ specifications of du and –st must be attributed to <strong>the</strong><br />

same argument slot. The non-dist<strong>in</strong>ctness (i.e. agreement) of <strong>the</strong>ir referentially relevant properties<br />

(features) will simply be a consequence of this. Therefore, feature check<strong>in</strong>g is not what probe-goal<br />

relations are about. Instead, what <strong>the</strong>y are about is precisely <strong>the</strong> predicate-argument calculus, of<br />

which agreement is but a reflex.<br />

In o<strong>the</strong>r words, saturation of <strong>the</strong>ta-roles corresponds to <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g of variables (<strong>the</strong> <strong>the</strong>taroles<br />

accord<strong>in</strong>g to <strong>the</strong> discussion <strong>in</strong> section 1, i.e. <strong>the</strong> ‘probes’) by arguments (i.e. <strong>the</strong> ‘goals’). Phifeature<br />

bundles, notably f<strong>in</strong>ite verb <strong>in</strong>flections, are <strong>in</strong>terpretable as <strong>the</strong> most elementary possible<br />

satisfaction of <strong>the</strong>ta-roles. In turn, <strong>the</strong> agreement relation (for <strong>in</strong>stance between a verb <strong>in</strong>flection and<br />

a DP) is part of <strong>the</strong> predicate-argument calculus, and is essentially <strong>the</strong> reverse of <strong>the</strong> movement<br />

relation. ‘Movement’ has a s<strong>in</strong>gle argument and two λ abstractions, while ‘agreement’ has a s<strong>in</strong>gle λ<br />

abstraction (i.e. <strong>the</strong>ta-role or argument slot) and two arguments (<strong>in</strong>clud<strong>in</strong>g now <strong>in</strong>flections).<br />

A consequence of <strong>the</strong> construal of phi-feature bundles as pronom<strong>in</strong>al elements is of course<br />

<strong>the</strong> elim<strong>in</strong>ation of pro (both expletive and referential). Null subject languages (or person for<strong>ms</strong><br />

with<strong>in</strong> a language) are those where <strong>the</strong> word-level lexicalization provided by <strong>the</strong> verb <strong>in</strong>flection<br />

suffices for <strong>the</strong> EPP; non-null subject languages (or for<strong>ms</strong>) are those where it does not.<br />

3. Empty operators. Our proposal so far admits only of variables at <strong>the</strong> LF <strong>in</strong>terface, elim<strong>in</strong>at<strong>in</strong>g<br />

purely syntactic (or ‘PF’) elements such as head trace and pro. In such a <strong>the</strong>ory <strong>the</strong>re will of course<br />

be empty operators as well. As time allows, we will argue that some ‘functional’ categories are<br />

better understood as operators (e.g. ‘number’, cf. <strong>the</strong> discussion with<strong>in</strong> Distributed Morphology<br />

from Noyer (1992) to Harbour (2011)). Therefore ‘silent’ categories, i.e. phonologically empty<br />

lexical item <strong>in</strong> <strong>the</strong> sense of Kayne (2010) may prove unnecessary (<strong>the</strong> converse of <strong>the</strong> conclusion <strong>in</strong><br />

section 2 that <strong>the</strong>re are no ‘un<strong>in</strong>terpretable’ lexical entries ei<strong>the</strong>r), once <strong>the</strong> more restrictive category<br />

of empty operator is adequately def<strong>in</strong>ed.


Tania Leal Méndez & Christ<strong>in</strong>e Shea (Iowa)<br />

L1 and L2 Mexican Spanish and <strong>in</strong>formation structure: P-movement or <strong>in</strong>-situ<br />

prosody?<br />

The present experiments focus on P-movement, a Spanish structure that lies at <strong>the</strong> syntaxdiscourse<br />

<strong>in</strong>terface. Accord<strong>in</strong>g to Zubizarreta (1998), <strong>the</strong> syntactic reflex of P-movement<br />

results <strong>in</strong> two word orders: VOS and [S]VPPO: (underl<strong>in</strong><strong>in</strong>g represents new <strong>in</strong>formation)<br />

Context: Who ate <strong>the</strong> apple? P-movement type 1<br />

Comió la manzana Juan. (VOS)<br />

Ate <strong>the</strong> apple Juan<br />

#Juan comió la manzana. (#SVO)<br />

Context: What did Juan put on top of <strong>the</strong> table? P-movement type 2<br />

Juan puso sobre la mesa un libro. (SVPPO)<br />

Juan put on top of <strong>the</strong> table a book.<br />

#Juan puso un libro sobre la mesa (#SVOPP)<br />

Juan put a book on top of <strong>the</strong> table<br />

Given that p-movement is a result of phonological alignment requirements, it has been<br />

argued to be prosodically motivated (Zubizarreta, 1998; but see López, 2009 for a<br />

syntactically-motivated explanation).<br />

Experiment 1 (Judgments)<br />

To test whe<strong>the</strong>r native Spanish speakers and L2 learners would <strong>the</strong> pert<strong>in</strong>ent word orders to<br />

<strong>the</strong> <strong>in</strong>formational structure (context), we used a used a bi-modal (text-audio) contextualized<br />

acceptability task, conta<strong>in</strong><strong>in</strong>g <strong>the</strong>se context-test sentence comb<strong>in</strong>ations (all <strong>in</strong> Spanish): 12<br />

P-movement (VOS and [S]VPPO), 6—Rheme (VSO), plus 18 fillers. SVO was excluded as<br />

an option because it is a „default‟ (biased) order <strong>in</strong> Spanish. Us<strong>in</strong>g a scale (1-4 or “I don‟t<br />

know”), participants judged (<strong>in</strong>)felicity <strong>in</strong> context. Participants also completed a proficiency<br />

test (two multiple-choice sections of a standardized test). Participants (N=137) <strong>in</strong>cluded 49<br />

native speakers, 25 advanced learners, 29 <strong>in</strong>termediate, and 46 low-<strong>in</strong>termediate. All learners<br />

had English as <strong>the</strong>ir L1. Group and <strong>in</strong>dividual results <strong>in</strong>dicate <strong>the</strong>se properties can be<br />

successfully acquired, although not all to <strong>the</strong> same extent (see Figures for results).<br />

Additionally, <strong>the</strong> results of <strong>the</strong> native speakers <strong>in</strong>dicate that <strong>in</strong> <strong>the</strong> case of P-movement 1, <strong>the</strong><br />

dist<strong>in</strong>ction between <strong>the</strong> two word orders <strong>in</strong> context is clear, while this was not <strong>the</strong> case for Pmovement<br />

2.<br />

Experiment 2 (Production)<br />

To test whe<strong>the</strong>r native Mexican Spanish speakers (from central Mexico, Puebla State)<br />

produced this orders <strong>in</strong> context and to analyze <strong>the</strong>ir prosodic encod<strong>in</strong>g of <strong>in</strong>formation<br />

structure, we used a contextualized forced choice production task. In a video, participants<br />

were presented with a context (a question) (8 seconds). After each question, participants<br />

viewed a short segment of a silent film (10-15 seconds). Then, participants were presented<br />

with three options to answer <strong>the</strong> question. These options differed only <strong>in</strong> <strong>the</strong> word order.<br />

Responses were audio-recorded (10 seconds). Participants <strong>in</strong>cluded a monol<strong>in</strong>gual Spanish<br />

group (n=53) all born and resid<strong>in</strong>g <strong>in</strong> Mexico and a bil<strong>in</strong>gual group (n=11) Bil<strong>in</strong>gual NS of<br />

Spanish who had completed a Master‟s degree <strong>in</strong> <strong>the</strong> States and had lived <strong>in</strong> an Englishspeak<strong>in</strong>g<br />

country at least 7 years (TOEFL m<strong>in</strong>: 500). Group and <strong>in</strong>dividual results show that<br />

<strong>the</strong> preferential manner for encod<strong>in</strong>g <strong>in</strong>formation structure (focus) is not through syntactic<br />

means (word order) but through prosody. Spanish ToBI analyses reveal that native Mexican<br />

Spanish speakers do mark <strong>in</strong>formation focus, but not <strong>in</strong> <strong>the</strong> expected fashion. Instead, <strong>the</strong>se


speakers prefer to mark narrow focus by means of pitch accents on <strong>the</strong> narrow-focus words<br />

ra<strong>the</strong>r than <strong>the</strong> expected syntactic movement. While Zubizarreta (1998) does not necessarily<br />

rule out this option, it is not hypo<strong>the</strong>sized to be <strong>the</strong> overwhelm<strong>in</strong>gly preferred one. While<br />

dialect-specific differences <strong>in</strong> ter<strong>ms</strong> of how prosody and <strong>in</strong>formation structure <strong>in</strong>teract have<br />

been well-documented, <strong>the</strong> evidence presented here suggests that not only can prosody be<br />

dialect specific (Prieto & Roseano, 2010), but <strong>the</strong> syntax-prosody <strong>in</strong>terface may also manifest<br />

dialect-specific effects <strong>in</strong> ter<strong>ms</strong> of production.<br />

Experiment 3 (Perception)<br />

To fur<strong>the</strong>r test <strong>the</strong> hypo<strong>the</strong>sis that Mexican Spanish speakers prefer <strong>the</strong> prosodic option <strong>in</strong><br />

ter<strong>ms</strong> of <strong>in</strong>formation structure realization (as perta<strong>in</strong><strong>in</strong>g to P-movement), we used a forced<br />

choice listen<strong>in</strong>g identification task. Participants heard a sentence with ei<strong>the</strong>r narrow focus<br />

on <strong>the</strong> subject (p-movement 1, VOS, 6 sentences) or narrow focus on <strong>the</strong> object (p-movement<br />

2, [S]VPPO, 6 sentences), plus 10 fillers and had to select which question best corresponded<br />

to <strong>the</strong> sentence heard, whe<strong>the</strong>r a question focused on <strong>the</strong> subject (e.g., ¿Quién comió? Who<br />

ate?, or ¿Qué pasó? What happened?). Based upon <strong>the</strong> data from <strong>the</strong> production study cited<br />

above, we hypo<strong>the</strong>size that Mexican Spanish speakers will prefer sentences with SVO order,<br />

accompanied by prosodically-realized narrow focus on <strong>the</strong> subject <strong>in</strong> response to „Who ate?‟<br />

type questions over VOS word-order. The same preference is predicted for p-movement 2.<br />

We are currently analyz<strong>in</strong>g <strong>the</strong> data for this experiment.<br />

Discussion<br />

We present prelim<strong>in</strong>ary perception and production data suggest<strong>in</strong>g that p-movement may be a<br />

dialect-dependent manifestation of <strong>the</strong> syntax-prosody <strong>in</strong>terface, given that Mexican Spanish<br />

speakers do not seem to use it to express <strong>in</strong>formation-structure shifts. This implies that <strong>the</strong><br />

realization of <strong>the</strong> prosody-syntax <strong>in</strong>terface <strong>in</strong> ter<strong>ms</strong> of <strong>in</strong>formation structure may need to be<br />

re-conceptualized to <strong>in</strong>corporate optional, less categorical preferences <strong>in</strong>to <strong>the</strong> grammar of<br />

native speakers.<br />

Figures<br />

Experiment 1: Acceptability Judgment averages for P-movement 1, P-movement 2<br />

Experiment 2:<br />

P-movement 1, Bil<strong>in</strong>guals P-movement 1, Monol<strong>in</strong>guals<br />

References<br />

Zubizarreta, M. L. (1998). Prosody, focus, and word order. Cambridge: MIT Press.<br />

Prieto, P., Roseano, P. (coords.) (2010). Transcription of Intonation of <strong>the</strong> Spanish Language.<br />

L<strong>in</strong>com Europa: München.


Luisa Meroni & Andrea Gualm<strong>in</strong>i (Utrecht)<br />

Do you know all SI? I know some<br />

Context –dependence of children’s computation of SIs<br />

Many experimental studies have shown that children don't compute scalar implicatures (SIs)<br />

as much as adults, despite master<strong>in</strong>g <strong>the</strong> prerequisites to <strong>the</strong>ir computation (Chierchia et al.,<br />

2000). In addition, different tasks (truth value judgment, picture‐selection or act‐out) have<br />

been shown to affect children's SIs computation, lead<strong>in</strong>g to <strong>the</strong> claim that <strong>the</strong> complexity of<br />

judgment‐tasks is beyond children's limited cognitive resources (Pouscoulous et al. 2007;<br />

Katsos et el., <strong>in</strong> press). This paper presents experimental data show<strong>in</strong>g that 1) children can <strong>in</strong><br />

fact compute SIs to <strong>the</strong> same extent as adults when this is <strong>the</strong> only contextually available<br />

option (Gualm<strong>in</strong>i et al., 2008) and 2) <strong>the</strong>y do so <strong>in</strong> a typical Truth Value Judgment task (Cra<strong>in</strong><br />

& Thornton, 1998).<br />

The role of contextual <strong>in</strong>formation for SIs computation was discussed by Welker<br />

(1994) and Carston (1998), through examples (1) and (2).<br />

(1) A: I'm hav<strong>in</strong>g a d<strong>in</strong>ner party and I need six more chairs.<br />

B: John has four chairs.<br />

(2) A: I'm hav<strong>in</strong>g a d<strong>in</strong>ner party and I need four more chairs.<br />

B: John has four chairs.<br />

The contrast lies <strong>in</strong> <strong>the</strong> fact that <strong>the</strong> SI associated with <strong>the</strong> numeral four (exactly 4) is<br />

computed <strong>in</strong> <strong>the</strong> first dialogue, but not <strong>in</strong> <strong>the</strong> second one (see Zondervan et al., 2009). In light<br />

of <strong>the</strong>oretical and experimental evidence suggest<strong>in</strong>g that implicatures are costly (Re<strong>in</strong>hart<br />

1998, Chierchia et al. 2000, Noveck 2001 among o<strong>the</strong>rs) a plausible view is that <strong>the</strong>ir cost<br />

needs to be justified. Our proposal is that contextual relevance provides children with a reason<br />

to comply with this cost. In particular, we draw upon <strong>the</strong> question that is raised <strong>in</strong> <strong>the</strong> context,<br />

and that is usually referred to as <strong>the</strong> Question Under Discussion (QUD, Roberts 1996) that has<br />

been already shown to exert an effect on resolv<strong>in</strong>g scope ambiguities for both children and<br />

adults (see Gualm<strong>in</strong>i et al. 2008).<br />

To <strong>in</strong>vestigate whe<strong>the</strong>r children compute <strong>the</strong> SI associated with some (some but not<br />

all), when presented with <strong>the</strong> appropriate QUD, we tested children with a Truth Value<br />

Judgment task. In a typical Truth Value Judgment task experiment, children listen to stories<br />

with a puppet and have to evaluate what <strong>the</strong> puppet says relatively to <strong>the</strong> story <strong>the</strong>y just heard.<br />

S<strong>in</strong>ce we wanted to study <strong>the</strong> effect of an explicit QUD on children's computation of SIs,<br />

follow<strong>in</strong>g <strong>the</strong> design by Zondervan et al. (2009), we elicited <strong>the</strong> target question from <strong>the</strong><br />

puppet by overtly ask<strong>in</strong>g <strong>the</strong> relevant QUD. The target question <strong>the</strong>n had to be evaluated by<br />

<strong>the</strong> subjects relatively to <strong>the</strong> story <strong>the</strong>y had just heard.<br />

To illustrate, children heard sentences like 3) I th<strong>in</strong>k some hotdogs were delivered as a<br />

description of a story <strong>in</strong> which <strong>the</strong> relevant character delivered all <strong>the</strong> hot‐dogs. As <strong>in</strong><br />

Zondervan et al. (2009), <strong>the</strong> target sentence (3) was presented as an answer to two different<br />

questions (4a or 4b), depend<strong>in</strong>g on which condition children had been assigned to:<br />

4) a. Were some hot‐dogs delivered? Condition1<br />

b. Were all <strong>the</strong> hot‐dogs delivered? Condition2<br />

Thirty English-speak<strong>in</strong>g children participated <strong>in</strong> this experiment, rang<strong>in</strong>g <strong>in</strong> age from 3;8 to<br />

6;5 (Mean = 4;10). We used a between-subject design where 15 children were presented with<br />

<strong>the</strong> question under discussion <strong>in</strong> 4a (Mean = 5;1) and <strong>the</strong> o<strong>the</strong>r 15 children with <strong>the</strong> question<br />

<strong>in</strong> 4b (Mean = 4;8). Each child saw four target trials <strong>in</strong>terspersed with fillers to balance <strong>the</strong><br />

expected number of yes and no answers.


We predicted that <strong>the</strong> implicature calculation would be affected by <strong>the</strong> contextually<br />

available question. Thus, subjects should calculate <strong>the</strong> SI ‐ and reject (3) ‐ when asked to<br />

evaluate (3) as an answer to <strong>the</strong> QUD <strong>in</strong> (4b), but not when (3) was presented as an answer to<br />

(4a). The reason is that both read<strong>in</strong>gs of (3), with or without SI, answer <strong>the</strong> QUD; thus a<br />

different criterion needs to be <strong>in</strong>voked to select one read<strong>in</strong>g (for <strong>in</strong>stance, computational<br />

complexity or <strong>the</strong> Pr<strong>in</strong>ciple of Charity, both lead<strong>in</strong>g to an affirmative answer). Differently,<br />

without <strong>the</strong> relevant implicature, (3) is not a good answer to (4b) (Gualm<strong>in</strong>i et al., 2008).<br />

Subjects should thus be led to consider whe<strong>the</strong>r an alternative <strong>in</strong>terpretation can be found that<br />

answers <strong>the</strong> QUD, and <strong>the</strong> <strong>in</strong>terpretation conta<strong>in</strong><strong>in</strong>g <strong>the</strong> relevant SI fits <strong>the</strong> bill. The results<br />

confirmed our prediction. Children accepted <strong>the</strong> sentence <strong>in</strong> (3) 85 % of <strong>the</strong> time, when this<br />

was preceded by question (4a), but <strong>the</strong>ir acceptance dropped to 13% when (3) followed (4b),<br />

exactly like <strong>the</strong> English speak<strong>in</strong>g adults tested by Zondervan et al. (2009). Children were also<br />

asked to motivate <strong>the</strong>ir negative answers to ensure <strong>the</strong>y were <strong>in</strong>deed comput<strong>in</strong>g <strong>the</strong> SI when<br />

reject<strong>in</strong>g <strong>the</strong> puppets' statement.<br />

To evaluate if children's computation of SI <strong>in</strong> Condition 2 was due to <strong>the</strong> mere<br />

presence of <strong>the</strong> quantifier all <strong>in</strong> <strong>the</strong> QUD, a control experiment was conducted. To this end,<br />

we slightly changed <strong>the</strong> QUD, as exemplified <strong>in</strong> Error! Reference source not found.:<br />

(5) Control Condition<br />

a. QUD: Were all <strong>the</strong> hot-dogs delivered?<br />

b. Target Sentence: I th<strong>in</strong>k some pizza were delivered<br />

As one can see, <strong>the</strong> QUD still conta<strong>in</strong>s <strong>the</strong> quantifier all but this time it revolves around hotdogs<br />

ra<strong>the</strong>r than <strong>the</strong> pizzas. To make <strong>the</strong> question contextually felicitous, <strong>the</strong> story also<br />

<strong>in</strong>cluded some hot-dogs. A third group of 15 English-speak<strong>in</strong>g children participated <strong>in</strong> this<br />

experiment. Age was from 4;3 to 6;0 (Mean = 5;1). Each child was presented with 4 target<br />

trials <strong>in</strong>terspersed with fillers. Turn<strong>in</strong>g to <strong>the</strong> results, <strong>the</strong> fifteen children <strong>in</strong>terviewed<br />

computed <strong>the</strong> SI associated with (5b) only 10% of <strong>the</strong> time. Children did not compute <strong>the</strong> SI<br />

associated with <strong>the</strong> quantifier some despite <strong>the</strong> occurrence of <strong>the</strong> stronger term all <strong>in</strong> <strong>the</strong> QUD<br />

allow<strong>in</strong>g us to ma<strong>in</strong>ta<strong>in</strong> <strong>the</strong> view that <strong>the</strong> computation of <strong>the</strong> SI <strong>in</strong> Experiment 1 can be<br />

attributed to <strong>the</strong> QUD ra<strong>the</strong>r than to some prim<strong>in</strong>g effect. In particular, children were not<br />

concerned with <strong>the</strong> change of focus implied by <strong>the</strong> target sentence and accepted <strong>the</strong> target<br />

sentence.<br />

Our f<strong>in</strong>d<strong>in</strong>gs show that once <strong>the</strong> question under discussion is explicitly given nei<strong>the</strong>r<br />

developmental nor methodological differences emerge. Taken toge<strong>the</strong>r with <strong>the</strong> exist<strong>in</strong>g<br />

literature, our results show that one contextual property, that of <strong>the</strong> QUD, can account for<br />

participants’ preferences on two apparently dist<strong>in</strong>ct phenomena: scope ambiguities and scalar<br />

implicatures. This result is an important step <strong>in</strong> <strong>the</strong> process of understand<strong>in</strong>g how context<br />

drives <strong>in</strong>terpretation


The L<strong>in</strong>guistic Expression of Causation<br />

Ad Neeleman & Hans van de Koot (UCL)<br />

1. The mental representation of causal relations The most common def<strong>in</strong>ition of causation<br />

has three components (Lewis 1973): (i) it is a relation between a caus<strong>in</strong>g and a caused event,<br />

such that (ii) <strong>the</strong> caus<strong>in</strong>g event may not follow <strong>the</strong> caused event and (iii) if <strong>the</strong> caus<strong>in</strong>g event<br />

had not occurred <strong>the</strong> caus<strong>in</strong>g event would not have occurred ei<strong>the</strong>r. There is ample evidence<br />

that causation is primarily a tool humans use for structur<strong>in</strong>g <strong>the</strong>ir mental model of <strong>the</strong> world<br />

(Johnson-Laird 1983, 2006). Causation cannot be l<strong>in</strong>guistic <strong>in</strong> nature, as it can be present <strong>in</strong><br />

<strong>the</strong> absence of language (Call 2004, Varley & Siegel 2002, Leslie 1984). The central question<br />

for l<strong>in</strong>guists, <strong>the</strong>n, is how <strong>the</strong> psychological notion of causation is expressed l<strong>in</strong>guistically.<br />

We propose that causative verbs do not encode a caus<strong>in</strong>g event. Ra<strong>the</strong>r, <strong>the</strong>y encode a<br />

macro-event, a resultant state, and <strong>the</strong> Crucial Contribut<strong>in</strong>g Factor (CCF) <strong>in</strong> <strong>the</strong> com<strong>in</strong>g about<br />

of <strong>the</strong> matrix event. The CCF is realized as <strong>the</strong> external argument (Re<strong>in</strong>hart 2002) and<br />

identifies which factor, out of a range of potential factors contribut<strong>in</strong>g to an outcome, is<br />

deemed by <strong>the</strong> speaker not to fall <strong>in</strong> <strong>the</strong> ceteris paribus category. As causal cha<strong>in</strong>s can be<br />

quite complex, <strong>the</strong>re are few <strong>in</strong>herent restrictions on what can function as a CCF, expla<strong>in</strong><strong>in</strong>g<br />

<strong>the</strong> variation of x <strong>in</strong> x eventually killed Leo (e.g. whisky, Bill, or <strong>the</strong> distance to <strong>the</strong> nearest<br />

oasis). Our proposal does not imply that language cannot express caus<strong>in</strong>g events. But when<br />

<strong>the</strong>se are expressed, this is not done through <strong>the</strong> verb, but ra<strong>the</strong>r through an adjunct, as <strong>in</strong><br />

Little Orson grew <strong>in</strong>to a big man [by eat<strong>in</strong>g Irish Oatmeal], where grow is an anticausative.<br />

2. The paradox of direct causation The standard view is that <strong>the</strong> lexical semantics of a<br />

causative verb conta<strong>in</strong>s a caus<strong>in</strong>g event. Whisky eventually killed Leo is <strong>the</strong>n taken to mean<br />

that <strong>the</strong>re is an event <strong>in</strong>volv<strong>in</strong>g whisky, such that it caused Leo’s death. The well-known<br />

Restriction to Direct Causation <strong>in</strong> simplex causatives (RDC) can be construed as evidence for<br />

this view, as formulat<strong>in</strong>g it requires reference to caus<strong>in</strong>g events. It follows that our proposal is<br />

falsified if <strong>the</strong> RDC is correct. However, evidence for <strong>the</strong> RDC is surpris<strong>in</strong>gly weak.<br />

The ma<strong>in</strong> source of evidence for <strong>the</strong> RDC consists of what we will call Katz effects (Katz<br />

1970): <strong>in</strong> <strong>the</strong> context of (1), (2a) is accepted by all speakers, but (2b) is not.<br />

(1) A sheriff’s six-shooter is faultily repaired by <strong>the</strong> local gunsmith. As a result, his<br />

weapon ja<strong>ms</strong> at a critical moment and <strong>the</strong> sheriff is gunned down.<br />

(2) a. The gunsmith caused <strong>the</strong> sheriff to die.<br />

b. The gunsmith killed <strong>the</strong> sheriff.<br />

Although <strong>the</strong> effect <strong>in</strong> (1) is undoubtedly real, <strong>the</strong>re are many examples <strong>in</strong> which <strong>the</strong> RDC is<br />

violated. For <strong>in</strong>stance, a 1943 Duke Ell<strong>in</strong>gton song entitled A slip of <strong>the</strong> lip can s<strong>in</strong>k a ship<br />

describes a situation <strong>in</strong> which loose talk by sailors may allow a spy to obta<strong>in</strong> and transmit<br />

<strong>in</strong>formation that may <strong>in</strong> turn allow a foreign navy to <strong>in</strong>struct its submar<strong>in</strong>es to torpedo and<br />

s<strong>in</strong>k ships <strong>in</strong> a convoy. We conclude from examples of this type that <strong>the</strong> RDC is <strong>in</strong>correct and<br />

must <strong>the</strong>refore be attributed to <strong>in</strong>dependent factors. We discuss <strong>the</strong>se factors below.<br />

Adverbial modification is sometimes claimed to provide fur<strong>the</strong>r evidence for <strong>the</strong> RDC (see<br />

Fodor 1970 for related discussion). We will show that this claim is <strong>in</strong>correct.<br />

3. Expla<strong>in</strong><strong>in</strong>g Katz effects The basis for our analysis of Katz effects is <strong>the</strong> hypo<strong>the</strong>sis that<br />

language encodes accountability for outcome through <strong>the</strong> rule <strong>in</strong> (3), where [+m] is a feature<br />

borrowed from Re<strong>in</strong>hart 2002 that expresses that <strong>the</strong> referent of an argument has a m<strong>in</strong>d.<br />

(3) Accountability: The referent of a DP specified as [+m] is held accountable for <strong>the</strong><br />

action expressed by <strong>the</strong> verb if and only if it is <strong>the</strong> CCF argument of that verb.<br />

Accountability only plays a role when a causative verb takes a [+m] subject and is <strong>the</strong>refore<br />

<strong>in</strong>applicable to <strong>the</strong> slip of <strong>the</strong> lip example, expla<strong>in</strong><strong>in</strong>g <strong>the</strong> absence of a Katz effect. The<br />

awkwardness of (2b) <strong>in</strong> <strong>the</strong> context of (1) also follows, because (1) does not <strong>in</strong>vite <strong>the</strong> reader<br />

to hold <strong>the</strong> gunsmith responsible for <strong>the</strong> sheriff’s death. But tweaks of <strong>the</strong> scenario improve<br />

examples like (2b). Suppose that follow<strong>in</strong>g <strong>the</strong> demise of <strong>the</strong> sheriff, <strong>the</strong> gunsmith visits a


psychoanalyst because he is struggl<strong>in</strong>g with feel<strong>in</strong>gs of guilt. He might <strong>the</strong>n say I killed <strong>the</strong><br />

sheriff, mean<strong>in</strong>g that he holds hi<strong>ms</strong>elf responsible. Notice that <strong>the</strong> RDC makes <strong>the</strong> <strong>in</strong>correct<br />

prediction that <strong>the</strong> gunsmith would under such circu<strong>ms</strong>tances know<strong>in</strong>gly utter a falsity.<br />

Katz effects also disappear if <strong>the</strong> gunsmith sabotages <strong>the</strong> sheriff’s gun because he has a<br />

long-stand<strong>in</strong>g grudge aga<strong>in</strong>st him. In such a scenario, (2b) is unobjectionable (Wolff 2003).<br />

In fact it can be shown that accountability assignment is subject to a locality condition<br />

stated over causal relations <strong>in</strong> <strong>the</strong> mental model. In particular, ascription of accountability<br />

across participants of an <strong>in</strong>terven<strong>in</strong>g event <strong>in</strong> a causal cha<strong>in</strong> is forbidden, unless <strong>the</strong>ir actions<br />

fall <strong>in</strong> <strong>the</strong> ceteris paribus category. Thus, (2b) becomes unacceptable <strong>in</strong> <strong>the</strong> extended sabotage<br />

scenario if we <strong>in</strong>troduce an <strong>in</strong>terven<strong>in</strong>g agent whose decisions are presented as crucial to <strong>the</strong><br />

outcome. This locality effect confir<strong>ms</strong> that accountability assignment is real.<br />

Why is cause to die different from kill? In <strong>the</strong> case of kill, a [+m] CCF is accountable for<br />

<strong>the</strong> entire macro-event encoded by <strong>the</strong> verb, which <strong>in</strong>cludes <strong>the</strong> resultant state. In <strong>the</strong> case of<br />

cause to die, <strong>the</strong> accountability of a [+m] CCF is limited to <strong>the</strong> causation event (<strong>the</strong><br />

circu<strong>ms</strong>tances lead<strong>in</strong>g to <strong>the</strong> sheriff’s death) – it excludes <strong>the</strong> resultant state (<strong>the</strong> death itself)<br />

because that state is <strong>in</strong>troduced by a separate predicate.<br />

The above is sufficient to expla<strong>in</strong> <strong>the</strong> effects that <strong>in</strong>itially motivated <strong>the</strong> RDC. However, we<br />

discuss additional factors <strong>in</strong>duc<strong>in</strong>g Katz effects, <strong>in</strong>clud<strong>in</strong>g <strong>the</strong> lexical encod<strong>in</strong>g of <strong>in</strong>struments<br />

(The gunsmith shot <strong>the</strong> sheriff is <strong>in</strong>felicitous <strong>in</strong> all contexts mentioned above).<br />

4. Test<strong>in</strong>g for <strong>the</strong> presence of a caus<strong>in</strong>g event We close our presentation with a discussion<br />

of proposals that assume that <strong>the</strong> lexical semantics of causative verbs conta<strong>in</strong>s a caus<strong>in</strong>g<br />

event. Strik<strong>in</strong>gly, even <strong>in</strong> generative semantics <strong>the</strong>re was no syntactic position for such an<br />

event. Instead, <strong>the</strong> standard assumption was (and is) that <strong>the</strong> caus<strong>in</strong>g event is encoded as part<br />

of <strong>the</strong> semantics of <strong>the</strong> CAUSE predicate (compare Dowty 1979). Similar ideas can be found <strong>in</strong><br />

Hale & Keyser 1993, Ramchand 2008, and Pylkkänen 2008. What is <strong>the</strong> evidence for this<br />

representation of <strong>the</strong> caus<strong>in</strong>g event?<br />

Consider first <strong>the</strong> claim that eventive subjects of causative verbs denote <strong>the</strong> caus<strong>in</strong>g event.<br />

This claim must be <strong>in</strong>correct, <strong>in</strong> view of examples like (4), where <strong>the</strong> relevant event follows<br />

<strong>the</strong> caused event. This is unproblematic if <strong>the</strong> eventive subject is a CCF, because <strong>the</strong> caus<strong>in</strong>g<br />

event <strong>in</strong> <strong>the</strong> mental model can <strong>the</strong>n be people act<strong>in</strong>g on knowledge of a future event.<br />

(4) Tomorrow’s strike by London Underground staff caused mayhem on <strong>the</strong> North<br />

Circular Road dur<strong>in</strong>g this even<strong>in</strong>g’s rush hour.<br />

A second potential argument could be based on semantic restrictions on <strong>the</strong> subject. These<br />

could result from lexical specification of <strong>the</strong> putative caus<strong>in</strong>g event. However, although all<br />

lexical causatives express restrictions on <strong>the</strong> caused event (kill means ‘cause to die’), it is<br />

certa<strong>in</strong>ly not true <strong>in</strong> <strong>the</strong> general case that <strong>the</strong>y impose restrictions on <strong>the</strong> caus<strong>in</strong>g event. At<br />

least, <strong>the</strong>re is no necessary restriction on <strong>the</strong> choice of subject. As already mentioned, a very<br />

wide variety of elements that can replace x <strong>in</strong> x eventually killed Leo.<br />

Two fur<strong>the</strong>r arguments for caus<strong>in</strong>g events <strong>in</strong> lexical semantics are presented by Pylkkänen<br />

(2008), who argues, on <strong>the</strong> basis Japanese and F<strong>in</strong>nish, that a caus<strong>in</strong>g event can be present,<br />

even if no associated external argument is <strong>in</strong>troduced. In <strong>the</strong> Japanese case, <strong>the</strong> claim is that<br />

certa<strong>in</strong> adjuncts are excluded <strong>in</strong> anticausative contexts but permitted <strong>in</strong> adversity<br />

constructions, where <strong>the</strong>y are licensed by a hidden caus<strong>in</strong>g event. We show that this argument<br />

is based on an <strong>in</strong>correct premise, as <strong>the</strong> relevant type of modifier does <strong>in</strong> fact show up with<br />

anticausatives. In <strong>the</strong> F<strong>in</strong>nish case, sluic<strong>in</strong>g is claimed to identify a hidden caus<strong>in</strong>g event <strong>in</strong><br />

desiderative constructions. This analysis makes <strong>the</strong> <strong>in</strong>correct prediction that <strong>the</strong> same k<strong>in</strong>d of<br />

sluic<strong>in</strong>g should also be possible <strong>in</strong> uncontroversially causative constructions, contrary to fact.<br />

5. Conclusions Causative verbs do not encode a caus<strong>in</strong>g event. This is because subjects of<br />

causative verbs express a CCF (which is associated with accountability when [+m]). If<br />

expressed at all, caus<strong>in</strong>g events surface as adjuncts.


��������� ���������� ������� �� ������� �� ������<br />

����� ������ ������ ���� �� ������ ����������� ����� ���� ������� ��� ������ ���<br />

��������� �� ������ ��������� �� ������ ����� �� ��������� ����������� �� ��������<br />

�������� �� ��� ���� ������� ���� ��� ��������������� �� ������� ������ ���������� ���<br />

�� ��������� �� ���������������� �������� ���� �� ���������� ���������� ����������� ���<br />

���� ����� ��� �������� ��� ��������� �� ���� �������� �� ���� ��� ��������������� ��<br />

��������� �� ������� ���� ����� ������������� �� ��� ������ ����������<br />

��� ��� �� ���� ���� �� �� ����������� ������� ���������� ������� ����� �� ����� ������<br />

��� ������������ ��� ������������ ��� ������� ��� ��� ��������������� �� ������� ��<br />

������ ���������� �� ������� ��������� ���� ����������� ���� ���������� ������� �� ���<br />

������� ��� ��� �������� ������� �� ��� ������� ����<br />

��� ����� �� ���������� � ������������ ���� ��� ���������� ���� �� �������������<br />

���������� ���� ��� ���������� �� ��������� ������������� �� �������� �� � ������ �� ������<br />

������� �������� �������� ������� ��� ����������� �� ��������� �� � ������� ��� �� ���������<br />

���� ��� ������� ���� ����������� ���������� ����� ��� ���� ���� ����� �� �������� ���<br />

����������� ��� ������ ������� ���� �� ��� ��������� ���� �� ����� ��� ������������ ���������<br />

����� ��� �� ���� ����� ��� �������� ��� �������������������� ����������� ������� ��� ���<br />

���� �� ��������� �� ����� ��� ����� ����������� ��� � ���������� ������� ����������<br />

����������� ��� ������� �� ��� ������������� ����� �� ������<br />

���������� � ����� ���� ��������� �������� ��� ������������� ��������� �� ���������<br />

���� ����� ����������� �������� ���� ������������ ���� ��� ���� ��� ��������� �����<br />

����������� � ������� ���� ��������� ���� ����� ���������� ��� �������� ����������<br />

������� ������ ��������� ������� ��� �������� ���������� ��� ���� ������ ���������� ����<br />

��� �������� ��������������� �������� ����� ������� ������ ���� ��� ���� ���� ��� �����<br />

����������� �� ����� ���������� �� ���� ��� �� ����� ��� ���������� ������������<br />

��� ��� � �� ����� ���� �� ������� ����� ���� � ���������<br />

���� ����� �������� ��� ��� ���� ���� �� ��� � �� ����<br />

��������<br />

���� �����<br />

���� ��� ����� ������ ��� ���� ���� �� ��� � ���� ����� ������ �<br />

������������ ����������<br />

��� �� ���� ����� ����� ���� �� ������� ����� ���� �<br />

����� ����� ����� �������� ��� ��� ���� ���� �� ��� �<br />

��������� ��������<br />

�� ���� ���� �����<br />

����� ����� ��� ����� ������ ��� ���� ���� �� ��� � ���� ����� ������ �<br />

�������� ����������<br />

���������� � ����� �� ���������� ����� ��� ��������� �� ����� �������� ���������� ����<br />

����� ���������� ��� ��� ����������� �� ���������� ��������<br />

��� ��� � ��<br />

����<br />

������<br />

������<br />

�����<br />

���<br />

��<br />

���<br />

�����<br />

��� ���<br />

��������<br />

���� ����<br />

�<br />

�<br />

������<br />

������<br />

���� ��� ������ ���� ��� ��� ���� ���� � ���� ����� ������ �<br />

������������ ����������<br />

��� � ����<br />

�����<br />

�������<br />

���<br />

Bruno Nicemboim (Potsdam)<br />

������<br />

������<br />

�����<br />

���<br />

�<br />

��<br />

���<br />

�����<br />

��� ���<br />

��������<br />

���� ����<br />

�������<br />

�����<br />

�<br />

�<br />

������<br />

������


�������<br />

�����<br />

����� ��� ��� ������ ���� ��� ��� ���� ���� � ���� ����� ������ �<br />

�������� ����������<br />

���������� � ������� ���� ��������� ����������� ��� ���������� �� ��������� ������������ ��<br />

��������� ���� ����� ����������� �������� �� ����������� ������� ������������� �����<br />

����� ��� ������� ����� ��� ������ ���� ������ ������� �� ���������� � ���� ����<br />

��������� ���� ��� ������� ��� ������������� �� �������� ��������� ���� ����� ������<br />

����� �� ��� ����� �� ��� ����������� ��������� ��� ��� ������� �� ��� �������� ��������<br />

��� ��������� ���� ������ ��� ��� �������� ��������� ���� ����� ���������� ���� ��� ���<br />

��������� ���� �������� ���� ��� ���������� �� ��������� ������������ �� ����������� ��<br />

����� ������� ����� �� ��� ���������� ���� ��� ���� ���� ����������� �� ��� ���������<br />

��������� ���������� �� ��� ��������� ������������ ������� ��� ����������� ��� �����<br />

��������� �������<br />

��������� ��� ������� ���� ����������� � ��� � �������� ����� ���� ����� ���<br />

����� ��� ��� ���� �������� ��������� ��� � ����� ���� ��� ��������������� �� �����<br />

���������� �� ��� �� ����� �����������������<br />

��������� ����� ����������� ����� �� � ����� �������� �� ���������� �� ��� ���� ��<br />

��� ������ ���������� ���� ����� ���������� ��� ������� �� ���������� ������� �������� ����<br />

������ ��������� ������������ ������ � ������ �� ��� ������ ��� ����� ��� �������������<br />

�� ��������� ���� ����� ����������� ���� ���� ���� ��� ����������� ��������� ����������<br />

������ ��������� ��� ������ ���������� �� ��������� ������������ �� ����� ���������� ���<br />

�� ���� ������ ���� �� ����������� ��������� ����� ������� ����������� ���� �������� ����<br />

��� ���������� �� ��������� ������������ �� ��������� �� ��� ��������������� �� �����<br />

�����������<br />

����������<br />

����������� �� ��� ���� �� �� ������� ��������� ����������� ��� ������ ������� ���������<br />

���<br />

��������� �� ������� �� ��� ����������� ������� ������ ��� ���� �������� � ����������<br />

������������ ������ ��� ���������� ����� ��������<br />

��������� �� ������� ����������� �������� ��� ����������� �������� ��� ��������������<br />

�� ���� ������������ ����� �������<br />

����� �� ������� ����������� �� ��������� �� ������� ��� ������� ����<br />

���� �� ��� ����������� ��� ��� ������� �� ������� ���������� ���������� �� ����������<br />

����������� ��� ������� ���� ������ ����������� �� ����������� ���� ��� ������ �������<br />

�� ��� ������� ���������� �������� ������ ��� ����� �������� ���� ���� ��������<br />

�������� �� ������� � ������� ��� ������������ ������� ������� ����� ��� ���������� ��<br />

�������� �������������<br />


Leticia Pablos, Bobby Ruijgrok, Jenny Doetjes & Lisa Cheng (Leiden)<br />

Process<strong>in</strong>g cataphoric pronouns <strong>in</strong> Dutch: an ERP study<br />

The process<strong>in</strong>g of cataphoric pronouns has been shown to follow <strong>the</strong> same mechanis<strong>ms</strong> as<br />

<strong>the</strong> process<strong>in</strong>g of wh-dependencies <strong>in</strong> that <strong>the</strong> parser actively searches [1] for an antecedent to<br />

<strong>in</strong>terpret <strong>the</strong> pronoun with<strong>in</strong> <strong>the</strong> sentence, except <strong>in</strong> those cases where <strong>the</strong> pronoun must obey<br />

pr<strong>in</strong>ciple C of <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory and cannot be c-commanded by <strong>the</strong> antecedent [2]. This is <strong>the</strong><br />

case for <strong>the</strong> Gender Mismatch (GMM) effect, named after a slowdown effect that shows that <strong>the</strong><br />

parser tries to l<strong>in</strong>k an antecedent to a preced<strong>in</strong>g pronoun only when a pronoun can be bound by it<br />

[3,4]. Most of <strong>the</strong> studies that tested whe<strong>the</strong>r <strong>the</strong>re is an active search triggered for an antecedent<br />

after encounter<strong>in</strong>g cataphoric pronouns used behavioral techniques such as self-paced read<strong>in</strong>g or<br />

eye-track<strong>in</strong>g [5,6] and only a few used <strong>the</strong> electroencephalography (EEG) technique [7].<br />

The current study on Dutch uses Event Related Potentials (ERP) to exam<strong>in</strong>e <strong>the</strong><br />

process<strong>in</strong>g of sentences with cataphoric pronouns as <strong>the</strong> parser looks for an antecedent <strong>in</strong> <strong>the</strong><br />

upcom<strong>in</strong>g <strong>in</strong>put. If <strong>the</strong> parser attempts to b<strong>in</strong>d <strong>the</strong> pronouns zijn and haar <strong>in</strong> (1a) and (1b) to <strong>the</strong><br />

mascul<strong>in</strong>e antecedent Lodewijk, we expect a GMM effect at <strong>the</strong> position of <strong>the</strong> potential<br />

antecedent <strong>in</strong> (1b). On <strong>the</strong> o<strong>the</strong>r hand, if <strong>the</strong> parser respects Pr<strong>in</strong>ciple C of <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory <strong>in</strong><br />

(1c) and (1d), we do not expect to f<strong>in</strong>d any ERP difference at <strong>the</strong> position of <strong>the</strong> potential<br />

antecedent Lodewijk show<strong>in</strong>g that no l<strong>in</strong>k has tried to be made between <strong>the</strong> pronouns hij and zij<br />

and <strong>the</strong> antecedent. Additional proper nouns such as Mirjam and Thomas <strong>in</strong> (1) were <strong>in</strong>cluded to<br />

guarantee that all pronouns had an antecedent <strong>in</strong> <strong>the</strong> sentence.<br />

(1) No constra<strong>in</strong>t match<br />

a. Zijnj assistenten kwamen erachter dat Lodewijkj Boer geen prijsw<strong>in</strong>naar<br />

His assistants realized that Lodewijkmasc Boer no prizew<strong>in</strong>ner<br />

Geselecteerd had, maar Mirjami had geen <strong>in</strong>teresse <strong>in</strong> de roddel.<br />

Selected had but Mirjamfem had no <strong>in</strong>terest<strong>in</strong> <strong>the</strong> gossip.<br />

No constra<strong>in</strong>t mismatch<br />

b. Haari assistenten kwamen erachter dat Lodewijkj Boer geen prijsw<strong>in</strong>naar<br />

Her assistants realized that Lodewijkmasc Boer no prizew<strong>in</strong>ner,<br />

geselecteerd had, maar Mirjami had geen <strong>in</strong>teresse <strong>in</strong> de roddel.<br />

selected had, but Mirjamfem had no <strong>in</strong>terest <strong>in</strong> <strong>the</strong> gossip.<br />

Pr<strong>in</strong>ciple C match<br />

c. Hiji kwam erachter dat Lodewijkj Boer geen prijsw<strong>in</strong>naar<br />

He realized that Lodewijkmasc Boer no prize w<strong>in</strong>ner<br />

geselecteerd had, maar Thomasi had geen <strong>in</strong>teresse <strong>in</strong> de roddel.<br />

selected had, but Thomasmasc had no <strong>in</strong>terest <strong>in</strong> <strong>the</strong> gossip.<br />

Pr<strong>in</strong>ciple C mismatch<br />

d. Ziji kwam erachter dat Lodewijkj Boer geen prijsw<strong>in</strong>naar<br />

She realized that Lodewijkmasc Boer no prize w<strong>in</strong>ner,<br />

geselecteerd had, maar Mirjami had geen <strong>in</strong>teresse <strong>in</strong> de roddel.<br />

selected had, but Mirjamfem had no <strong>in</strong>terest <strong>in</strong> <strong>the</strong> gossip.


We conducted an ERP experiment where EEG was cont<strong>in</strong>uously recorded while native<br />

speakers (n=24) of Dutch read silently 36 sentences such as (1a-d) <strong>in</strong>terspersed with 35 fillers<br />

and subsequently answered a comprehension question for every sentence. As illustrated <strong>in</strong> <strong>the</strong><br />

accompany<strong>in</strong>g leftmost figure, results show that <strong>the</strong>re is a central anterior negativity <strong>in</strong> <strong>the</strong> 200-<br />

600<strong>ms</strong> w<strong>in</strong>dow <strong>in</strong> (1b) condition with respect to (1a) at <strong>the</strong> po<strong>in</strong>t of <strong>the</strong> potential antecedent<br />

Lodewijk (significant 3-way <strong>in</strong>teraction between factors Condition, Hemisphere (left, Right,<br />

Central) and electrode position (Anterior, Middle, Posterior); F(12,276)=2,05, p=0.045)). On <strong>the</strong><br />

o<strong>the</strong>r hand, comparison between conditions (1c) and (1d) yielded no significant difference <strong>in</strong> <strong>the</strong><br />

ERP wavefor<strong>ms</strong> as shown <strong>in</strong> <strong>the</strong> rightmost figure.<br />

Amplitude (μV)<br />

-10<br />

-5<br />

0<br />

GMM Exp - Condition A vs. B at "Lodewijk" - Electrode Fz<br />

5<br />

10<br />

A<br />

B<br />

-100 0 100 200 300 400 500 600 700 800<br />

Time (<strong>ms</strong>)<br />

Amplitude (μV)<br />

-10<br />

-5<br />

0<br />

Time (<strong>ms</strong>)<br />

GMM Exp - Condition C vs. D at "Lodewijk" - Electrode Fz<br />

5<br />

10<br />

C<br />

D<br />

-100 0 100 200 300 400 500 600 700 800<br />

Time (<strong>ms</strong>)<br />

The long susta<strong>in</strong>ed negativity generated at <strong>the</strong> antecedent Lodewijk for (1b) condition<br />

suggests that <strong>the</strong> parser attempts to l<strong>in</strong>k <strong>the</strong> antecedent to <strong>the</strong> preced<strong>in</strong>g pronoun haar and fails to<br />

<strong>in</strong>terpret it at <strong>the</strong> antecedent due to <strong>the</strong> gender mismatch between <strong>the</strong> pronoun -marked for<br />

fem<strong>in</strong><strong>in</strong>e gender - and <strong>the</strong> antecedent Lodewijk – a name <strong>in</strong> mascul<strong>in</strong>e gender. The same effect is<br />

absent <strong>in</strong> <strong>the</strong> o<strong>the</strong>r conditions, which shows that <strong>the</strong> parser does not try to l<strong>in</strong>k <strong>the</strong> antecedent<br />

with <strong>the</strong> preced<strong>in</strong>g pronoun <strong>in</strong> (1c) and (1d) so that pr<strong>in</strong>ciple C is respected. Overall results<br />

confirm <strong>the</strong> active search mechanism for an antecedent started whenever <strong>the</strong>re is a pronoun that<br />

must be bound <strong>in</strong> <strong>the</strong> local context as <strong>in</strong> (1b). We discuss implications of <strong>the</strong>se results <strong>in</strong> light of<br />

<strong>the</strong> accumulated knowledge on long-distance dependency process<strong>in</strong>g.<br />

References:<br />

[1] Clifton, C, & Frazier, L. (1989). Comprehend<strong>in</strong>g sentences with long distance dependencies.<br />

In Carlson, G.N. & Tanenhaus, M. eds., L<strong>in</strong>guistic Structure <strong>in</strong> Language Process<strong>in</strong>g, 273-317.<br />

[2] Cho<strong>ms</strong>ky, N. 1981. Lectures on Government and B<strong>in</strong>d<strong>in</strong>g, Dordrecht: Foris Publications.<br />

[3]Van Gompel, R.P.G., & Liversedge, S.P. (2003). The <strong>in</strong>fluence of morphological <strong>in</strong>formation<br />

on cataphoric pronoun assignment. Journal of Experimental Psychology: Learn<strong>in</strong>g, Memory,<br />

and Cognition, 29, 128-139.<br />

[4] Kazan<strong>in</strong>a, N., Lau, E., Lieberman, M., Yoshida, M., & Phillips, C. (2007). The Effect of<br />

Syntactic Constra<strong>in</strong>ts on <strong>the</strong> Process<strong>in</strong>g of Backwards Anaphora. Journal of Memory and<br />

Language, 56, 384–409.<br />

[5] Sturt, P. (2003). The time-course of <strong>the</strong> application of b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts <strong>in</strong> reference<br />

resolution. Journal of Memory and Language, 48, 542-562.<br />

[6] Aoshima, S. , Yoshida, M. & C.Phillips (2009). Incremental Process<strong>in</strong>g of Coreference and<br />

B<strong>in</strong>d<strong>in</strong>g <strong>in</strong> Japanese. Syntax, 12, 93-134.<br />

[7] Kre<strong>in</strong>er, H. Mohr, S. Kessler, K. and S. Garrod. (2008) Can context affect gender process<strong>in</strong>g?<br />

ERP Evidence about differences between lexical and stereotypical gender. Poster presented at<br />

Bra<strong>in</strong> Talk, Lund.


Daniel Parker & Sol Lago (Maryland)<br />

Retrieval Interference <strong>in</strong> <strong>the</strong> Resolution of Anaphoric PRO<br />

Recent research on <strong>the</strong> memory operations used <strong>in</strong> real-time language comprehension<br />

has revealed a selective profile for retrieval <strong>in</strong>terference effects. Dependencies such as<br />

subject-verb agreement show strong facilitatory <strong>in</strong>terference effects, as predicted by misretrieval<br />

due to partial cue-match<strong>in</strong>g <strong>in</strong> a content-addressable memory architecture [1,2]. On<br />

<strong>the</strong> o<strong>the</strong>r hand, previous studies of reflexive anaphors have not found facilitation effects.<br />

They have ei<strong>the</strong>r found no effects of structurally <strong>in</strong>accessible antecedents [3,4,5,6], or <strong>the</strong>y<br />

have found <strong>in</strong>hibition effects [3,7,8], which are not directly predicted by <strong>the</strong> cue-based<br />

retrieval model. This profile has been taken to suggest <strong>the</strong> use of different retrieval<br />

mechanis<strong>ms</strong> <strong>in</strong> <strong>the</strong> two cases, despite superficially similar structural and morphological<br />

requirements [5]. The reasons for this contrast rema<strong>in</strong> unresolved. The contrast may reflect<br />

<strong>the</strong> fact that anaphoric dependencies are <strong>in</strong>terpreted whereas agreement is not. Or it may<br />

reflect differential use of non-structural features as retrieval cues. The licens<strong>in</strong>g of anaphoric<br />

PRO provides a good test of <strong>the</strong> candidate retrieval mechanis<strong>ms</strong> because it shares properties<br />

with both agreement and reflexive b<strong>in</strong>d<strong>in</strong>g. Results from three studies us<strong>in</strong>g off-l<strong>in</strong>e<br />

judgments, computational model<strong>in</strong>g and self-paced read<strong>in</strong>g confirm <strong>the</strong> structural constra<strong>in</strong>ts<br />

on licens<strong>in</strong>g PRO, but show an on-l<strong>in</strong>e <strong>in</strong>terference profile similar to agreement, suggest<strong>in</strong>g<br />

<strong>the</strong> use of non-structural cues for retriev<strong>in</strong>g <strong>the</strong> antecedent of PRO. These results provide <strong>the</strong><br />

first case of facilitatory <strong>in</strong>terference effects <strong>in</strong> <strong>the</strong> resolution of anaphoric dependencies.<br />

The licens<strong>in</strong>g of PRO <strong>in</strong> adjunct clauses is subject to structural constra<strong>in</strong>ts. The<br />

controller must be <strong>the</strong> subject of <strong>the</strong> next higher clause (1).<br />

1a. Johni read <strong>the</strong> report after PROi dr<strong>in</strong>k<strong>in</strong>g his coffee.<br />

1b. *The report confused Johni after PROi dr<strong>in</strong>k<strong>in</strong>g his coffee.<br />

This constra<strong>in</strong>t and <strong>the</strong> anaphoric nature of <strong>the</strong> dependency both suggest that retrieval might<br />

proceed <strong>in</strong> a structure-sensitive fashion, similar to reflexive licens<strong>in</strong>g. Conversely, <strong>the</strong> search<br />

for a controller shares at least two properties with subject-verb agreement. First, onl<strong>in</strong>e<br />

detection of PRO and subsequent retrieval of <strong>the</strong> controller is triggered by a verb ra<strong>the</strong>r than<br />

by an <strong>in</strong>dependent anaphoric element. Second, selectional restrictions from <strong>the</strong> gerundive<br />

verb <strong>in</strong> <strong>the</strong> control clause (e.g. it may show a bias for a [+animate] subject) may provide<br />

additional retrieval cues. Previous studies of agreement have shown strong facilitatory<br />

<strong>in</strong>terference effects, attributed to fallible cue-based retrieval mechanis<strong>ms</strong> (e.g. [2]). If PRO<br />

behaves similarly to agreement, <strong>the</strong>n we should f<strong>in</strong>d evidence of facilitatory <strong>in</strong>terference<br />

from structurally <strong>in</strong>accessible antecedents, due to partial cue-match<strong>in</strong>g.<br />

We compared <strong>the</strong> profiles of subject-verb agreement and adjunct control <strong>in</strong> a design<br />

with 8 conditions, us<strong>in</strong>g 48 sets of sentences like those <strong>in</strong> (2). Conditions (2a-d) provided a<br />

basic profile for agreement <strong>in</strong>terference. The ma<strong>in</strong> clause subject NP ei<strong>the</strong>r agreed (2a/b) or<br />

disagreed (2c/d) with <strong>the</strong> highlighted verb, and we manipulated <strong>the</strong> number of an <strong>in</strong>terven<strong>in</strong>g<br />

relative clause subject (‘attractor’) to ei<strong>the</strong>r match or mismatch <strong>the</strong> critical verb. This allowed<br />

us to test for <strong>the</strong> ‘illusions of grammaticality’ observed <strong>in</strong> previous studies. Conditions (2e-h)<br />

were designed to test <strong>the</strong> process<strong>in</strong>g of anaphoric PRO us<strong>in</strong>g maximally similar<br />

configurations to (2a-d). We <strong>in</strong>dependently manipulated <strong>the</strong> animacy of <strong>the</strong> ma<strong>in</strong> clause<br />

subject (correct controller) and <strong>the</strong> animacy of <strong>the</strong> relative clause subject (<strong>in</strong>correct<br />

controller). The <strong>in</strong>animate NPs were chosen such that <strong>the</strong>y could be <strong>the</strong> subject of <strong>the</strong> critical<br />

verb <strong>in</strong> <strong>the</strong> control clause, although <strong>the</strong>y might be dispreferred due to an animacy bias. An<br />

emphatic reflexive requir<strong>in</strong>g an animate, gender match<strong>in</strong>g NP antecedent as <strong>the</strong> local subject<br />

(PRO) served as a probe to determ<strong>in</strong>e whe<strong>the</strong>r an animate NP was retrieved as <strong>the</strong> subject of<br />

<strong>the</strong> adjunct clause.<br />

1


2a/b: The doctor that <strong>the</strong> {researcher/reports} evaluated extensively was commended [after<br />

PRO disprov<strong>in</strong>g <strong>the</strong> controversial <strong>the</strong>ory at <strong>the</strong> research <strong>in</strong>stitute <strong>in</strong> Europe].<br />

2c/d: The experiment that <strong>the</strong> {researchers/report} evaluated extensively were commended<br />

[after PRO disprov<strong>in</strong>g <strong>the</strong> controversial <strong>the</strong>ory at <strong>the</strong> research <strong>in</strong>stitute <strong>in</strong> Europe].<br />

2e/f: The doctor that <strong>the</strong> {researcher/report} evaluated extensively was commended [after<br />

PRO disprov<strong>in</strong>g <strong>the</strong> controversial <strong>the</strong>ory hi<strong>ms</strong>elf at <strong>the</strong> research <strong>in</strong>stitute <strong>in</strong> Europe].<br />

2g/h: The experiment that <strong>the</strong> {researcher/report} evaluated extensively was commended<br />

[after PRO disprov<strong>in</strong>g <strong>the</strong> controversial <strong>the</strong>ory hi<strong>ms</strong>elf at <strong>the</strong> research <strong>in</strong>stitute <strong>in</strong><br />

Europe].<br />

Experiment 1 (off-l<strong>in</strong>e acceptability judgment, n=24) confirmed that PRO <strong>in</strong> an<br />

adjunct clause must be controlled by <strong>the</strong> next higher subject. Experiment 2 (computational<br />

model<strong>in</strong>g) used <strong>the</strong> ACT-R parser to establish predictions from a cue-based retrieval <strong>the</strong>ory.<br />

The simulations predict facilitatory <strong>in</strong>terference effects for both <strong>the</strong> agreement conditions<br />

(2a-d) and <strong>the</strong> adjunct control conditions (2e-h). Interference <strong>in</strong> <strong>the</strong> agreement conditions is<br />

predicted <strong>in</strong> ter<strong>ms</strong> of process<strong>in</strong>g time at <strong>the</strong> verbal region and error rate <strong>in</strong> <strong>the</strong> retrieval of<br />

attractors. In <strong>the</strong> adjunct control conditions, simulations predict access to both <strong>the</strong> licit and<br />

illicit subject NPs at <strong>the</strong> adjunct verb. Facilitatory <strong>in</strong>terference effects similar to those shown<br />

for agreement are predicted <strong>in</strong> ter<strong>ms</strong> of process<strong>in</strong>g time at <strong>the</strong> reflexive, and error rate <strong>in</strong> <strong>the</strong><br />

retrieval of a structurally <strong>in</strong>accessible animate NP subject.<br />

Experiment 3 (self-paced read<strong>in</strong>g, n=32) tested <strong>the</strong> predictions of <strong>the</strong> model. In <strong>the</strong><br />

agreement conditions (2a-d), l<strong>in</strong>ear mixed-effects model<strong>in</strong>g revealed effects of both<br />

grammaticality and attractor number (grammaticality: t=-2.043; attractor number: t=2.434;<br />

ps


Umesh Patil (Potsdam), Shravan Vasishth (Potsdam) & Richard Lewis (Michigan)<br />

Early effect of retrieval <strong>in</strong>terference on reflexive b<strong>in</strong>d<strong>in</strong>g<br />

The onl<strong>in</strong>e application of Pr<strong>in</strong>ciple A of <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory is claimed to be <strong>in</strong>fallible<br />

to memory phenomena like retrieval <strong>in</strong>terference from antecedents that are <strong>in</strong>accessible <strong>in</strong><br />

ter<strong>ms</strong> of <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory (Phillips, Wagers, & Lau, 2009; Dillon, 2011). Sturt (2003)<br />

and Xiang, Dillon, and Phillips (2009) report a set of studies with English reflexives and<br />

conclude that if <strong>the</strong>re is any effect of retrieval <strong>in</strong>terference from grammatically <strong>in</strong>accessible<br />

antecedents, it appears only dur<strong>in</strong>g later stages of process<strong>in</strong>g. Based on <strong>the</strong>se results,<br />

Phillips et al. (2009) and Dillon (2011) propose that reflexive b<strong>in</strong>d<strong>in</strong>g is immune to<br />

<strong>in</strong>terference dur<strong>in</strong>g early stages of process<strong>in</strong>g, because <strong>the</strong> antecedent of a reflexive is<br />

retrieved from memory us<strong>in</strong>g strictly syntactic <strong>in</strong>formation, and that agreement features<br />

like gender and number are completely ignored <strong>in</strong> <strong>the</strong> antecedent search process.<br />

On <strong>the</strong> o<strong>the</strong>r hand, a large body of work <strong>in</strong> <strong>the</strong> doma<strong>in</strong> of dependency resolution <strong>in</strong><br />

sentence process<strong>in</strong>g has shown that <strong>the</strong> memory retrieval process utilizes non-syntactic<br />

<strong>in</strong>formation as well. Van Dyke and colleagues (Van Dyke & Lewis, 2003; Van Dyke<br />

& McElree, 2006) have shown that semantic properties of nouns (e.g. animacy feature)<br />

and selectional requirements of verbs are utilized <strong>in</strong> retrievals. Moreover, <strong>the</strong> process of<br />

b<strong>in</strong>d<strong>in</strong>g English reflexives <strong>in</strong>side picture noun phrases (Runner, Sussman, & Tanenhaus,<br />

2006) and Ch<strong>in</strong>ese reflexives (Chen, Jäger, & Vasishth, 2011) is shown to be <strong>in</strong>fluenced<br />

by <strong>the</strong> agreement features of <strong>the</strong> grammatically <strong>in</strong>accessible antecedent. In fact, recently<br />

Cunn<strong>in</strong>gs and Felser (2011) have shown that high memory span readers occasionally<br />

consider <strong>in</strong>accessible antecedents dur<strong>in</strong>g b<strong>in</strong>d<strong>in</strong>g argument reflexives. In <strong>the</strong> light of <strong>the</strong>se<br />

results, <strong>the</strong> strictly syntactic retrieval account see<strong>ms</strong> to be an exception, which calls for<br />

a specialized retrieval mechanism to expla<strong>in</strong> only a limited set of results.<br />

We formulated <strong>the</strong> question—what type of retrieval cues are used <strong>in</strong> <strong>the</strong> reflexive b<strong>in</strong>d<strong>in</strong>g<br />

process—<strong>in</strong> ter<strong>ms</strong> <strong>the</strong> cue-based retrieval (CBR) <strong>the</strong>ory (Lewis & Vasishth, 2005). The<br />

CBR <strong>the</strong>ory provides a computational architecture for model<strong>in</strong>g sentence process<strong>in</strong>g phenomena.<br />

The <strong>the</strong>ory is based on <strong>the</strong> memory and process<strong>in</strong>g pr<strong>in</strong>ciples of ACT-R, a cognitive<br />

architecture developed for model<strong>in</strong>g general cognitive processes. We implemented<br />

two CBR models of reflexive b<strong>in</strong>d<strong>in</strong>g <strong>in</strong> English—model-1 that uses strictly syntactic cues<br />

and, model-2 that uses syntactic cues as well as gender mark<strong>in</strong>g on <strong>the</strong> reflexive to identify<br />

its antecedent. We also ran an eye track<strong>in</strong>g study to evaluate <strong>the</strong> predictions of <strong>the</strong> two<br />

models.<br />

The models’ predictions were generated for <strong>the</strong> four conditions (2x2 design; factors:<br />

accessible NP match/mismatch for gender x <strong>in</strong>accessible NP match/mismatch for gender)<br />

listed <strong>in</strong> (1). The predictions of <strong>the</strong> models are <strong>in</strong> ter<strong>ms</strong> of: (i) antecedent retrieval<br />

time and (ii) accuracy <strong>in</strong> retriev<strong>in</strong>g <strong>the</strong> grammatical antecedent. Model-1 predicts no<br />

<strong>in</strong>terference effect, whereas, model-2 predicts an <strong>in</strong>terference effect <strong>in</strong> ter<strong>ms</strong> of retrieval<br />

times and retrieval accuracies.<br />

(1) a. Accessible-match/<strong>in</strong>accessible-match: The tough soldier that Fred treated <strong>in</strong><br />

<strong>the</strong> military hospital <strong>in</strong>troduced hi<strong>ms</strong>elf to all <strong>the</strong> nurses.<br />

b. Accessible-match/<strong>in</strong>accessible-mismatch: The tough soldier that Katie treated<br />

<strong>in</strong> <strong>the</strong> military hospital <strong>in</strong>troduced hi<strong>ms</strong>elf to all <strong>the</strong> nurses.<br />

c. Accessible-mismatch/<strong>in</strong>accessible-match: The tough soldier that Katie treated<br />

<strong>in</strong> <strong>the</strong> military hospital <strong>in</strong>troduced herself to all <strong>the</strong> nurses.<br />

d. Accessible-mismatch/<strong>in</strong>accessible-mismatch: The tough soldier that Fred<br />

treated <strong>in</strong> <strong>the</strong> military hospital <strong>in</strong>troduced herself to all <strong>the</strong> nurses.


We ran an eye track<strong>in</strong>g study (n=40) with <strong>the</strong> four conditions listed above, to evaluate<br />

<strong>the</strong> predictions of <strong>the</strong> two models, assum<strong>in</strong>g that early and late effects are dist<strong>in</strong>guishable<br />

<strong>in</strong> <strong>the</strong> eye track<strong>in</strong>g measures. As predicted by model-2, <strong>the</strong> study showed an early effect of<br />

<strong>in</strong>terference from <strong>the</strong> <strong>in</strong>accessible antecedent <strong>in</strong> ter<strong>ms</strong> of first-pass regression probability;<br />

i.e. a gender match between <strong>the</strong> reflexive and <strong>the</strong> <strong>in</strong>accessible NP (1a and 1c) <strong>in</strong>duced<br />

a significantly higher (p=0.038) proportion of first-pass regressions from <strong>the</strong> reflexive <strong>in</strong><br />

<strong>the</strong> sentence. Although o<strong>the</strong>r early eye movements measures did not show any significant<br />

effect, a regression cont<strong>in</strong>gent analysis of first-fixation durations showed a pattern of<br />

fixations that was consistent with <strong>the</strong> <strong>in</strong>terference predictions of model-2.<br />

In sum, <strong>the</strong> eye track<strong>in</strong>g results are consistent with <strong>the</strong> predictions of <strong>the</strong> model that<br />

utilizes both syntactic and gender <strong>in</strong>formation to identify <strong>the</strong> antecedent of a reflexive.<br />

Moreover, <strong>the</strong> early <strong>in</strong>terference effect found <strong>in</strong> <strong>the</strong> current study is not consistent with <strong>the</strong><br />

claim that <strong>in</strong>accessible antecedents are not considered dur<strong>in</strong>g earlier stages of process<strong>in</strong>g<br />

(Sturt, 2003; Phillips et al., 2009; Dillon, 2011). We conclude that a strictly syntactic<br />

search mechanism is overly selective and, hence, unable to account for <strong>the</strong> data reported<br />

here and <strong>in</strong> o<strong>the</strong>r studies like Cunn<strong>in</strong>gs and Felser (2011) and Badecker and Straub (2002).<br />

References<br />

Badecker, W., & Straub, K.(2002). The process<strong>in</strong>g role of structural constra<strong>in</strong>ts on <strong>the</strong> <strong>in</strong>terpretation<br />

of pronouns and anaphors. Journal of Experimental Psychology: Learn<strong>in</strong>g, Memory,<br />

and Cognition, 28 (4), 748–769.<br />

Chen, Z., Jäger, L., & Vasishth, S. (2011). How structure sensitive is <strong>the</strong> parser? Evidence<br />

from Mandar<strong>in</strong> Ch<strong>in</strong>ese. In B. Stolterfoht & S. Fea<strong>the</strong>rston (Eds.), Empirical approaches<br />

to l<strong>in</strong>guistic <strong>the</strong>ory: Studies of mean<strong>in</strong>g and structure. Berl<strong>in</strong>: Mouton de Gruyter. (<strong>in</strong><br />

press)<br />

Cunn<strong>in</strong>gs, I., & Felser, C. (2011). The role of work<strong>in</strong>g memory <strong>in</strong> <strong>the</strong> process<strong>in</strong>g of reflexives.<br />

Language and Cognitive Processes. (<strong>in</strong> press)<br />

Dillon, B. (2011). Structured access <strong>in</strong> sentence comprehension. Unpublished doctoral dissertation,<br />

Maryland.<br />

Lewis, R. L., & Vasishth, S.(2005). An activation-based model of sentence process<strong>in</strong>g as skilled<br />

memory retrieval. Cognitive Science: A Multidiscipl<strong>in</strong>ary Journal, 29 (3), 375–419.<br />

Phillips, C., Wagers, M. W., & Lau, E. F.(2009). Grammatical illusions and selective fallibility<br />

<strong>in</strong> real time language comprehension. Language and <strong>L<strong>in</strong>guistics</strong> Compass.<br />

Runner, J. T., Sussman, R. S., & Tanenhaus, M. K.(2006). Process<strong>in</strong>g reflexives and pronouns<br />

<strong>in</strong> picture noun phrase. Cognitive Science, 30 (2), 193–241.<br />

Sturt, P.(2003). The time-course of <strong>the</strong> application of b<strong>in</strong>d<strong>in</strong>g constra<strong>in</strong>ts <strong>in</strong> reference resolution.<br />

Journal of Memory and Language, 48 (3), 542–562.<br />

Van Dyke, J. A., & Lewis, R. L.(2003). Dist<strong>in</strong>guish<strong>in</strong>g effects of structure and decay on attachment<br />

and repair: A cue-based pars<strong>in</strong>g account of recovery from misanalyzed ambiguities.<br />

Journal of Memory and Language, 49, 285–316.<br />

Van Dyke, J. A., & McElree, B.(2006). Retrieval <strong>in</strong>terference <strong>in</strong> sentence comprehension. Journal<br />

of Memory and Language, 55, 157–166.<br />

Xiang, M., Dillon, B., & Phillips, C. (2009). Illusory licens<strong>in</strong>g effects across dependency types:<br />

Erp evidence. Bra<strong>in</strong> and Language, 108 (1), 40–55.


Syntactic constra<strong>in</strong>ts on <strong>the</strong> process<strong>in</strong>g of pronouns<br />

Syntactic constra<strong>in</strong>ts affect<strong>in</strong>g <strong>the</strong> <strong>in</strong>terpretation of pronouns <strong>in</strong>clude, among o<strong>the</strong>rs, <strong>the</strong> l<strong>in</strong>ear<br />

distance between a pronoun and a potential antecedent (e.g. Cunn<strong>in</strong>gs et al., 2011) and<br />

condition B of <strong>the</strong> b<strong>in</strong>d<strong>in</strong>g <strong>the</strong>ory (Cho<strong>ms</strong>ky, 1981). Little is known, however, about how<br />

<strong>the</strong>se potentially compet<strong>in</strong>g constra<strong>in</strong>ts <strong>in</strong>teract dur<strong>in</strong>g real-time sentence comprehension. We<br />

will report <strong>the</strong> results from an eye-movement monitor<strong>in</strong>g study <strong>in</strong>vestigat<strong>in</strong>g when dur<strong>in</strong>g<br />

process<strong>in</strong>g local and/or non-local antecedents for pronouns are considered <strong>in</strong> different types<br />

of syntactic environment.<br />

B<strong>in</strong>d<strong>in</strong>g condition B demands that (non-reflexive) pronouns must not be bound with<strong>in</strong><br />

<strong>the</strong>ir local b<strong>in</strong>d<strong>in</strong>g doma<strong>in</strong>, which should exclude David from <strong>the</strong> set of candidate antecedents<br />

for him <strong>in</strong> sentences like (1) below.<br />

(1) Nicki th<strong>in</strong>ks that Davidk likes himi/*k.<br />

Clare Patterson & Claudia Felser (Potsdam)<br />

Pronouns <strong>in</strong> certa<strong>in</strong> syntactic environments appear to be exempt from condition B, however.<br />

These <strong>in</strong>clude so-called ‘short distance pronouns’ (henceforth, SDPs) <strong>in</strong> sentences such as (2)<br />

below, where ei<strong>the</strong>r Nick or David can be <strong>in</strong>terpreted as coreferential with <strong>the</strong> pronoun him.<br />

(2) Nicki saw Davidk put <strong>the</strong> cat beside himi/k.<br />

Possible reasons as to why SDPs might be exempt from condition B <strong>in</strong>clude proposals to <strong>the</strong><br />

effect that prepositional phrases such as beside him <strong>in</strong> (2), or certa<strong>in</strong> k<strong>in</strong>ds of (VP-<strong>in</strong>ternal)<br />

aspectual phrases, can be b<strong>in</strong>d<strong>in</strong>g doma<strong>in</strong>s (Hestvik, 1991; Tenny, 2004). Experimental<br />

evidence for <strong>the</strong> referential ambiguity of SDPs has been reported by Seker<strong>in</strong>a et al. (2004).<br />

Previous research on <strong>the</strong> role of condition B <strong>in</strong> onl<strong>in</strong>e pronoun resolution has yielded a<br />

number of different hypo<strong>the</strong>ses. Based on results from cross-modal prim<strong>in</strong>g, Nicol and<br />

Sw<strong>in</strong>ney (1989) argued that b<strong>in</strong>d<strong>in</strong>g pr<strong>in</strong>ciples act as an <strong>in</strong>itial filter on <strong>the</strong> process<strong>in</strong>g of<br />

pronouns, such that only antecedents that are licensed by condition B are considered dur<strong>in</strong>g<br />

process<strong>in</strong>g. An alternative view was put forward by Badecker and Straub (2002), who<br />

suggested that syntactically ‘<strong>in</strong>accessible’ antecedents for pronouns are also <strong>in</strong>cluded <strong>in</strong> <strong>the</strong><br />

<strong>in</strong>itial candidate set. F<strong>in</strong>ally, if Sturt’s (2003) f<strong>in</strong>d<strong>in</strong>gs on <strong>the</strong> process<strong>in</strong>g of reflexives (which<br />

are subject to b<strong>in</strong>d<strong>in</strong>g condition A) carry over to pronouns, <strong>the</strong>n condition B should act as an<br />

early but defeasible filter, with <strong>in</strong>accessible antecedents potentially be<strong>in</strong>g considered at later<br />

process<strong>in</strong>g stages only.<br />

34 native speakers of English were presented with sentences which conta<strong>in</strong>ed an<br />

object pronoun (him or her) and two potential sentence-<strong>in</strong>ternal antecedents. Gender<br />

congruence between <strong>the</strong> pronoun and <strong>the</strong> proper names was manipulated (match vs.<br />

mismatch) to serve as a diagnostic for referential dependency formation. In Experiment 1,<br />

‘condition B’ type sentences were used (see examples 3a-c):<br />

(3) a. DOUBLE MATCH<br />

John remembered that Mark had taught him a new song on <strong>the</strong> guitar.<br />

b. LOCAL MISMATCH<br />

John remembered that Jane had taught him a new song on <strong>the</strong> guitar.<br />

c. NON-LOCAL MISMATCH<br />

Jane remembered that John had taught him a new song on <strong>the</strong> guitar.<br />

In Experiment 2, sentences conta<strong>in</strong><strong>in</strong>g SDPs were used (examples 4a-c):


(4) a. DOUBLE MATCH<br />

Barry saw Gav<strong>in</strong> place a gun near him on <strong>the</strong> ground with great care.<br />

b. LOCAL MISMATCH<br />

Barry saw Megan place a gun near him on <strong>the</strong> ground with great care.<br />

c. NON-LOCAL MISMATCH<br />

Megan saw Barry place a gun near him on <strong>the</strong> ground with great care.<br />

While a general preference for l<strong>in</strong>early closer antecedents favours <strong>the</strong> local antecedent <strong>in</strong> both<br />

(3) and (4), b<strong>in</strong>d<strong>in</strong>g condition B rules <strong>the</strong> local antecedent out <strong>in</strong> sentences of type (3). All<br />

experimental and filler sentences were embedded with<strong>in</strong> short neutral discourse contexts, and<br />

comprehension questions followed two thirds of <strong>the</strong> trials. Participants’ eye movements were<br />

recorded while <strong>the</strong>y were read<strong>in</strong>g <strong>the</strong> stimulus materials on a computer screen.<br />

In <strong>the</strong> condition B type sentences (Experiment 1), <strong>the</strong> ‘non-local mismatch’ condition<br />

(3c) yielded significantly longer reread<strong>in</strong>g and total view<strong>in</strong>g times <strong>in</strong> <strong>the</strong> pronoun and<br />

spillover region compared to <strong>the</strong> ‘local mismatch’ condition (3b). The ‘non-local mismatch’<br />

condition (3c) also yielded significantly <strong>in</strong>creased regression-path times <strong>in</strong> <strong>the</strong> f<strong>in</strong>al region<br />

compared to <strong>the</strong> ‘local mismatch’ condition (3b). This <strong>in</strong>dicates that participants experienced<br />

late but significant difficulty when <strong>the</strong> non-local antecedent mismatched <strong>the</strong> pronoun <strong>in</strong><br />

gender. Additionally, <strong>the</strong>re was a trend for <strong>the</strong> ‘double match’ condition (3a) to have<br />

<strong>in</strong>creased reread<strong>in</strong>g and total view<strong>in</strong>g times compared to <strong>the</strong> ‘local mismatch’ condition (3b),<br />

<strong>in</strong>dicat<strong>in</strong>g potential <strong>in</strong>terference from a gender-match<strong>in</strong>g local antecedent. For SDP sentences<br />

(Experiment 2) a different pattern was observed. Here <strong>the</strong> ‘local mismatch’ condition (4b)<br />

elicited <strong>the</strong> longest reread<strong>in</strong>g times <strong>in</strong> <strong>the</strong> pref<strong>in</strong>al and f<strong>in</strong>al regions, <strong>in</strong>dicat<strong>in</strong>g that<br />

participants experienced process<strong>in</strong>g difficulty later dur<strong>in</strong>g <strong>the</strong> sentence when <strong>the</strong> local<br />

antecedent mismatched <strong>in</strong> gender.<br />

Taken toge<strong>the</strong>r, <strong>the</strong>se results suggest that when a pronoun is first encountered <strong>the</strong>re is<br />

not an immediate default to ei<strong>the</strong>r <strong>the</strong> local (Cunn<strong>in</strong>gs et al., 2011) or <strong>the</strong> first-mentioned<br />

(Arnold et al., 2000) antecedent. Nor is <strong>the</strong>re an automatic application of condition B at an<br />

early ‘bond<strong>in</strong>g’ stage. Instead, candidate antecedents are evaluated fully at a later ‘resolution’<br />

stage, which <strong>in</strong>cludes consideration of <strong>the</strong> syntactic configuration and <strong>the</strong> application of<br />

condition B. For pronouns which are exempt from condition B (Experiment 2), a preference<br />

for <strong>the</strong> local antecedent is observed.<br />

References<br />

Arnold, J. E., Eisenband, J. G., Brown-Schmidt, S, and Trueswell, J. C. (2000). The<br />

immediate use of gender <strong>in</strong>formation: Eyetrack<strong>in</strong>g evidence of <strong>the</strong> time-course of pronoun<br />

resolution. Cognition 76, B13-B26.<br />

Badecker, W., & Straub, K. (2002). The process<strong>in</strong>g role of structural constra<strong>in</strong>ts on <strong>the</strong><br />

<strong>in</strong>terpretation of pronouns and anaphors. J Exp Psychol Learn 28, 748-769.<br />

Cho<strong>ms</strong>ky, N. (1981). Lectures <strong>in</strong> Government and B<strong>in</strong>d<strong>in</strong>g. Dordrecht: Foris.<br />

Cunn<strong>in</strong>gs, I., Patterson, C., & Felser, C. (2011). “Variable b<strong>in</strong>d<strong>in</strong>g and coreference <strong>in</strong><br />

sentence comprehension: Evidence from eye movements”. AMLaP 17, Paris.<br />

Hestvik, A. (1991). Subjectless b<strong>in</strong>d<strong>in</strong>g doma<strong>in</strong>s. Nat Lang L<strong>in</strong>guist Th 9, 455-496.<br />

Nicol, J., & Sw<strong>in</strong>ney, D. (1989). The role of structure <strong>in</strong> coreference assignment dur<strong>in</strong>g<br />

sentence comprehension. J Psychol<strong>in</strong>guist Res 18, 5-20.<br />

Seker<strong>in</strong>a, I., Stro<strong>ms</strong>wold, K., & Hestvik, A. (2004). How do adults and children process<br />

referentially ambiguous pronouns? J Child Lang 31, 123-152.<br />

Tenny, C. (2004). “Pronoun b<strong>in</strong>d<strong>in</strong>g and <strong>the</strong> grammar of sentience”. Workshop on Semantic<br />

Approaches to B<strong>in</strong>d<strong>in</strong>g Theory, ESSLLI 16, Université Henri Po<strong>in</strong>caré, Nancy.


Uli Sauerland (ZAS) & Jonathan Bobaljik (UConn)<br />

Syncretism Distribution Model<strong>in</strong>g and Person Paradig<strong>ms</strong><br />

<strong>Generative</strong> analyses of paradigm morphology relate two layers: <strong>the</strong> morpheme layer and <strong>the</strong> underly<strong>in</strong>g<br />

feature layer. <strong>Generative</strong> rules relate <strong>the</strong> two layers – for <strong>in</strong>stance, <strong>in</strong>sertion rules <strong>in</strong> <strong>the</strong> case<br />

of distributed morphology, which we adopt for concreteness <strong>in</strong> <strong>the</strong> follow<strong>in</strong>g. Such a generative<br />

analysis is restricted by two factors: a) <strong>the</strong> set of features, and b) <strong>the</strong> type of rule-order<strong>in</strong>g it allows.<br />

We dist<strong>in</strong>guish between extr<strong>in</strong>sic, weak <strong>in</strong>tr<strong>in</strong>sic and strong <strong>in</strong>tr<strong>in</strong>sic rule order<strong>in</strong>g. Consider <strong>the</strong><br />

follow<strong>in</strong>g illustration: If <strong>the</strong>re are two <strong>in</strong>dependent features F1 and F2, <strong>the</strong>re are <strong>the</strong> follow<strong>in</strong>g<br />

three possible rules:<br />

A: [F1 & F2] ↦→ /a/, B: [F1] ↦→ /b/, C: [F2] ↦→ /c/<br />

Rules can be ordered <strong>in</strong>tr<strong>in</strong>sically by <strong>the</strong>ir specificity: e.g. rule A must be ordered first <strong>in</strong> any<br />

language that uses it because if it was ordered after any of <strong>the</strong> o<strong>the</strong>r rules, it wouldn’t have any<br />

effect. Rule B and C, however, don’t stand <strong>in</strong> any <strong>in</strong>tr<strong>in</strong>sic order. The three order pr<strong>in</strong>ciples are<br />

dist<strong>in</strong>guished <strong>in</strong> <strong>the</strong>ir treatment of <strong>the</strong>se two. Extr<strong>in</strong>sic order allows languages to specify <strong>the</strong> order<br />

of B and C, a weak <strong>in</strong>tr<strong>in</strong>sic order allows languages to use rule B and C, but only if <strong>the</strong> language<br />

also uses rule A – <strong>in</strong> that context, <strong>the</strong> order of rules B and C is actually irrelevant. A strong <strong>in</strong>tr<strong>in</strong>sic<br />

order never allows languages to make use of more than one of B and C.<br />

The feature set and constra<strong>in</strong>ts on rule order predict which paradig<strong>ms</strong> can be generated. However,<br />

paradigm morphology differs from o<strong>the</strong>r doma<strong>in</strong>s of generative l<strong>in</strong>guistics <strong>in</strong> that paradig<strong>ms</strong> that<br />

cannot be generated <strong>in</strong> a systematic fashion, can still be generated via accidental homophony. For<br />

example, consider <strong>the</strong> paradigm [+F1, +F2] ↔ /b/, [+F1, -F2] ↔ /b/, [-F1, +F2] ↔ /c/. If extr<strong>in</strong>sic<br />

rule order is possible, <strong>the</strong> rules B followed by C generate <strong>the</strong> paradigm. But, if only weak <strong>in</strong>tr<strong>in</strong>sic<br />

orders are possible, <strong>the</strong> paradigm can still be generated: Namely, it requires all three rules A, B,<br />

and C and <strong>the</strong> assumption that morphemes /a/ and /b/ are accidentally homophonous.<br />

How do we f<strong>in</strong>d out which feature set and which rule order<strong>in</strong>g type are correct? Traditionally<br />

morphologist follow are rule of thumb, that Halle & Marantz (2008) state as follows: avoid accidental<br />

homophony and maximize generalizations. Despite its usefulness, <strong>the</strong> rule of thumb has<br />

no pr<strong>in</strong>cipled justification. In this talk, we <strong>in</strong>stead focus on <strong>the</strong> accidental, i.e. random, nature of<br />

accidental homophony: We view accidental homophony like random noise and it must <strong>the</strong>refore<br />

be randomly distributed. For a concrete generative analysis, we compute for a large amount of<br />

morphological data from several languages which syncretis<strong>ms</strong> can be systematic and which must<br />

be accidental. We <strong>the</strong>n test how ‘random’ <strong>the</strong> distribution of <strong>the</strong> accidental syncretism is us<strong>in</strong>g an<br />

exact, non-parametric statistical test. The method is an <strong>in</strong>stance of maximum likelihood model<strong>in</strong>g:<br />

The best morphological analysis is one where <strong>the</strong> actual distribution of accidental is highly likely<br />

to have arisen from random noise.<br />

In <strong>the</strong> talk, we present two results: 1) A general statistical framework for <strong>the</strong> analysis of typological<br />

paradigm frequencies. 2) An application of <strong>the</strong> framework to Cysouw’s (2003, OUP) data on<br />

person mark<strong>in</strong>g to argue that extr<strong>in</strong>sic rule order is necessary. In this abstract, we focus on a<br />

exemplary subcase – <strong>the</strong> case of first person morphology. For this case, we tested all possible<br />

generative analyses, and f<strong>in</strong>d that surpris<strong>in</strong>gly two can account for <strong>the</strong> data.<br />

The First Person Case: Now consider real data: <strong>the</strong> four cells of first person (first exclusive<br />

m<strong>in</strong>imal = 1, first <strong>in</strong>clusive m<strong>in</strong>imal = 1+2, first exclusive augmented = 1+3, and first <strong>in</strong>clusive


0 5 10 15<br />

augmented = 1+2+3). In this case, <strong>the</strong>re are 15 possible paradig<strong>ms</strong> and also 15 possible features.<br />

The m<strong>in</strong>imal number of universal features sufficient to allow an analysis of all 15 paradigm is 3.<br />

Our computational analysis shows that <strong>the</strong>re are 47 dist<strong>in</strong>ct generative analyses with 3 universal<br />

features (out of over 16000 total possible analyses).<br />

We <strong>the</strong>n tested whe<strong>the</strong>r any of <strong>the</strong> generative analyses predicts <strong>the</strong> actual distribution of languages<br />

across <strong>the</strong> 15 possible paradig<strong>ms</strong>. To limit <strong>the</strong> effect of large language families, we counted maximally<br />

15 languages per paradigm follow<strong>in</strong>g Cysouw (Post hoc, we confirmed our results to be<br />

<strong>in</strong>dependent of this assumption.) Contra Cysouw’s claim, we found that <strong>the</strong>re are generative analyses<br />

that predict <strong>the</strong> actual distribution to be highly likely. Contra generative expectations, <strong>the</strong>re<br />

are actually two generative models yield likelihood greater 80% for <strong>the</strong> actual distribution (all o<strong>the</strong>r<br />

models yield <strong>in</strong> likelihood smaller 0.1%): Model A and Model B illustrated below. On <strong>the</strong> x-axis,<br />

<strong>the</strong> two graphs show <strong>the</strong> 15 possible paradig<strong>ms</strong>, and on <strong>the</strong> y-axis <strong>the</strong> number of language (grey =<br />

actual, white = most likely expected). The top right corner shows <strong>the</strong> 3 basic features (bold frame)<br />

and possible derived features, <strong>the</strong> rate of accidental homophony, and <strong>the</strong> likelihood predicted for<br />

<strong>the</strong> actual distribution. Model A uses three semantically def<strong>in</strong>able, features 1, -2 and m<strong>in</strong>imal, and<br />

assumes a rate of accidental homophony of 5.8%. Model B uses also uses <strong>the</strong> features 1 and -2,<br />

but <strong>in</strong> addition <strong>the</strong> semantically unexpected feature [[1,-2,m<strong>in</strong>imal] or [1,2,augmented]], while <strong>the</strong><br />

accidental homophony rate is 10.3%. For <strong>the</strong> paradig<strong>ms</strong>, a bold frame <strong>in</strong>dicates that <strong>the</strong> paradigm<br />

can be generated without accidental homophony and <strong>the</strong> t<strong>in</strong>y numbers below <strong>in</strong>dicate <strong>the</strong> optimal<br />

underly<strong>in</strong>g distribution of languages.<br />

21.6 16.4 14.5 14.4 13 2.2 0.8<br />

aHomph rate: 5.8 %<br />

p > 9.9 * 10^ -1<br />

actual # of languages<br />

predicted # of languages<br />

0 5 10 15<br />

aHomph rate: 10.3 %<br />

p > 8.5 * 10^ -1<br />

actual # of languages<br />

predicted # of languages<br />

27.6 16.3 13.8 13.7 11.6 0 0<br />

Model A Model B<br />

Conclusion: In sum, we argue that generative morphology must use statistical models to get<br />

around <strong>the</strong> problem of accidental homophony and propose such a model, syncretism distribution<br />

model<strong>in</strong>g. Us<strong>in</strong>g this approach, we show that <strong>the</strong> morphological data from person paradig<strong>ms</strong><br />

support generative models, but require extr<strong>in</strong>sic rule order.<br />

References: Cysouw, M., 2003. The Paradigmatic Structure of Person Mark<strong>in</strong>g. Oxford University<br />

Press.


Patrick Sturt (University of Ed<strong>in</strong>burgh)<br />

The tim<strong>in</strong>g of dependency formation: State of <strong>the</strong> art and future challenges<br />

In <strong>the</strong> last two decades, on-l<strong>in</strong>e experimental techniques have yielded a wealth of evidence about<br />

<strong>the</strong> tim<strong>in</strong>g of dependency formation. But what does this evidence tell us about <strong>the</strong> process<strong>in</strong>g<br />

strategies and overall architecture of <strong>the</strong> language comprehension system? In this talk, I will attempt<br />

to build up a picture of <strong>the</strong> sentence comprehension system from <strong>the</strong> available evidence, focus<strong>in</strong>g<br />

on recent work conducted at Ed<strong>in</strong>burgh and elsewhere, cover<strong>in</strong>g a range of dependency types<br />

<strong>in</strong>clud<strong>in</strong>g control dependencies, reflexive b<strong>in</strong>d<strong>in</strong>g (<strong>in</strong> picture noun and co-argument contexts),<br />

subject-verb agreement. I will review <strong>the</strong> evidence with <strong>the</strong> follow<strong>in</strong>g questions <strong>in</strong> m<strong>in</strong>d: Is <strong>the</strong>re a<br />

systematic explanation for <strong>the</strong> variability <strong>in</strong> <strong>the</strong> tim<strong>in</strong>g and violability of grammatical constra<strong>in</strong>ts<br />

under different conditions? What is <strong>the</strong> relation between on-l<strong>in</strong>e structure-build<strong>in</strong>g and f<strong>in</strong>al<br />

<strong>in</strong>terpretation? What k<strong>in</strong>d of memory access might be <strong>in</strong>volved <strong>in</strong> dependency formation? What is<br />

<strong>the</strong> relation between <strong>the</strong> use of top-down and bottom-up structural <strong>in</strong>formation <strong>in</strong> pars<strong>in</strong>g? The talk<br />

will end with an outl<strong>in</strong>e of challenges for future research, and some suggestions on how we might<br />

tackle <strong>the</strong>m.


Yanyan Sui (NYU)<br />

Metrical Structural Prom<strong>in</strong>ence Versus Perceived Prom<strong>in</strong>ence <strong>in</strong> Standard Ch<strong>in</strong>ese<br />

Proposal: This study dist<strong>in</strong>guishes stress as metrical structural prom<strong>in</strong>ence from perceived<br />

prom<strong>in</strong>ence <strong>in</strong> <strong>the</strong> metrical stress analysis of Standard Ch<strong>in</strong>ese, a tone language whose<br />

syllables have fixed underly<strong>in</strong>g tones. It clai<strong>ms</strong> that perceived prom<strong>in</strong>ence may diverge from<br />

metrical structural prom<strong>in</strong>ence under <strong>the</strong> <strong>in</strong>fluence of duration and tones. The <strong>in</strong>tr<strong>in</strong>sic<br />

prom<strong>in</strong>ence of particular tones <strong>in</strong> <strong>the</strong> second syllable of a disyllabic word, as well as longer<br />

duration as a boundary leng<strong>the</strong>n<strong>in</strong>g effect may lead to iambic prom<strong>in</strong>ence judgment, whereas<br />

consistent F0 and durational patterns <strong>in</strong> addition to <strong>the</strong> distribution of neutral tones and<br />

segmental reduction suggest <strong>the</strong> basic foot structure is trochaic.<br />

Background: The stress pattern of disyllabic words whose second syllables are toneless,<br />

ei<strong>the</strong>r underly<strong>in</strong>gly or as a result of tone deletion, is agreed to be Strong-Weak. What rema<strong>in</strong>s<br />

controversial, both <strong>in</strong> native speakers‟ judgment and <strong>in</strong> previous phonological analyses, is <strong>the</strong><br />

stress pattern of words that do not conta<strong>in</strong> toneless syllables, which is claimed to be (1)<br />

iambic (Chao 1968, Xu 1982), (2) ei<strong>the</strong>r iambic or trochaic depend<strong>in</strong>g on lexical specification<br />

(Hoa 1983), or (3) trochaic (Duanmu 2000), but (4) becomes iambic <strong>in</strong> <strong>the</strong> surface by stress<br />

shift (Chang 1992) or by employment of a special type of foot (Duanmu 2007), or ra<strong>the</strong>r, (5)<br />

<strong>the</strong> stress pattern is determ<strong>in</strong>ed by tonal prom<strong>in</strong>ence (Meredith 1990). This paper argues that<br />

it is <strong>in</strong>correct to treat perceived prom<strong>in</strong>ence directly as metrical structural prom<strong>in</strong>ence.<br />

Prom<strong>in</strong>ence judgment is subject to <strong>the</strong> <strong>in</strong>fluence of duration and tones, among many o<strong>the</strong>r<br />

factors <strong>in</strong>volved <strong>in</strong> prom<strong>in</strong>ence perception.<br />

Perceived prom<strong>in</strong>ence ≠ metrical structural prom<strong>in</strong>ence: Perceived prom<strong>in</strong>ence may not<br />

be <strong>the</strong> faithful reflection of metrical structural prom<strong>in</strong>ence due to durational and tonal effects<br />

<strong>in</strong> prom<strong>in</strong>ence judgment. First, durational difference between syllables <strong>in</strong> a word <strong>in</strong>duces<br />

judgment of syllables with longer duration as more salient. In a corpus study of 2239<br />

disyllabic words <strong>in</strong> connected speech by Deng (2010), it is found that before no pause, <strong>the</strong><br />

first syllable is longer than <strong>the</strong> second syllable, but before little pause, <strong>the</strong> second syllable is<br />

longer, and <strong>the</strong> durational difference grows substantially before an obvious pause and <strong>in</strong><br />

utterance f<strong>in</strong>al position. Patterns <strong>in</strong> prom<strong>in</strong>ence judgment closely correlate with <strong>the</strong> durational<br />

differences. Before no pause and little pause, 64% and 52% of words respectively are judged<br />

stronger <strong>in</strong> <strong>the</strong> first syllable, but before an obvious pause, <strong>in</strong>clud<strong>in</strong>g utterance f<strong>in</strong>ally, only<br />

about 26% of words are judged stronger <strong>in</strong> <strong>the</strong> first syllable. And when grouped by tones,<br />

words that are judged stronger <strong>in</strong> <strong>the</strong> second syllable <strong>in</strong>variably have longer second syllables.<br />

Second, <strong>the</strong> <strong>in</strong>tr<strong>in</strong>sic prom<strong>in</strong>ence of tones <strong>in</strong>terferes with prom<strong>in</strong>ence judgment. Previous<br />

studies have noted that syllables with High Fall<strong>in</strong>g tone tend to be judged as <strong>the</strong> most<br />

prom<strong>in</strong>ent, whereas syllables with Low tone are <strong>the</strong> least likely to be perceived prom<strong>in</strong>ent<br />

(Hoa 1983). This observation is confirmed by <strong>the</strong> perception experiment by Deng (2010) and<br />

by <strong>the</strong> present study on embedded made-up disyllabic words, which vary <strong>in</strong> tonal<br />

comb<strong>in</strong>ations, but are devoid of morphosyntactic and semantic <strong>in</strong>formation. The results show<br />

that <strong>in</strong> a disyllabic word if <strong>the</strong> second syllable bears High Fall<strong>in</strong>g tone, or if <strong>the</strong> first syllable<br />

bears Low tone, <strong>the</strong> word tends to be judged more salient <strong>in</strong> <strong>the</strong> second syllable (at about 65%<br />

and 59% chances respectively), and <strong>the</strong> likelihood reaches 89% if <strong>the</strong> first syllable bears Low<br />

tone and <strong>the</strong> second syllable bears High Fall<strong>in</strong>g tone. In contrast, a second syllable with Low<br />

tone is judged prom<strong>in</strong>ent only 21% of <strong>the</strong> time. The correlation of tones and prom<strong>in</strong>ence<br />

judgment is salient and consistent across various speech registers <strong>in</strong>clud<strong>in</strong>g read isolated


words, broadcast news, and conversational speech. Therefore, perceived prom<strong>in</strong>ence cannot<br />

be simply equated with metrical structural prom<strong>in</strong>ence. Representations of metrical structure<br />

may only be built on evidence <strong>in</strong>dependent of <strong>the</strong> <strong>in</strong>tr<strong>in</strong>sic tonal prom<strong>in</strong>ence and durational<br />

effects due to boundary leng<strong>the</strong>n<strong>in</strong>g.<br />

Phonetic evidence for disyllabic trochee: (1) Strong-Weak F0 alternation: <strong>in</strong> a corpus study<br />

of prosody <strong>in</strong> broadcast news by Lai et al. (2010) which <strong>in</strong>cludes 56378 disyllabic tokens,<br />

4540 trisyllabic tokens and 727 quadrisyllabic tokens, a robust strong-weak F0 pattern is<br />

found <strong>in</strong> disyllabic words, and <strong>the</strong> strong-weak pattern repeats <strong>in</strong> quadrisyllabic words<br />

between <strong>the</strong> third and fourth syllables; Fur<strong>the</strong>rmore, when grouped by tones, a tone is realized<br />

more closely to its tonal template <strong>in</strong> <strong>the</strong> first syllable than <strong>in</strong> <strong>the</strong> second syllable <strong>in</strong> disyllabic<br />

words; Strong-Weak alternation <strong>in</strong> metrical strength is also found <strong>in</strong> an <strong>in</strong>dependent study of<br />

<strong>in</strong>tonation model<strong>in</strong>g by Kochanski et al. (2003); (2) Durational patterns: duration is not an<br />

immediate acoustic cue to metrical strength, it also <strong>in</strong>dicates prosodic constituent boundaries.<br />

However, consistent durational patterns emerge <strong>in</strong> sentential medial position before no pause,<br />

<strong>in</strong> which <strong>the</strong> first syllable is longer than <strong>the</strong> second syllable, and <strong>the</strong> third syllable longer than<br />

<strong>the</strong> second syllable <strong>in</strong> polysyllabic words (Wang & Wang 1993, Lai et al. 2010, Deng 2010).<br />

Phonological evidence for disyllabic trochee: (1) Distribution of neutral tone: underly<strong>in</strong>g<br />

toneless syllables may only occur <strong>in</strong> <strong>the</strong> second syllable of a disyllabic word, never <strong>in</strong> <strong>the</strong> first<br />

syllable, e.g., zhǔn tou „accuracy‟, jì de „remember‟; (2) Position of tone deletion: only <strong>the</strong><br />

tone of <strong>the</strong> second syllable may be deleted <strong>in</strong> disyllabic words, e.g., gōng jiā/ jia „<strong>the</strong> state‟, yì<br />

wù/ wu „obligation‟; (3) Position of segmental reduction: i) vowel reduction only occurs <strong>in</strong> <strong>the</strong><br />

second syllable, e.g., mián huā, /x w ɑ/ → [x w ə] „cotton‟, chuāng hù, /xu/ → [x w ə] „w<strong>in</strong>dow‟; ii)<br />

vowel devoic<strong>in</strong>g only <strong>in</strong> <strong>the</strong> second syllable, e.g., kuò qì /tɕ h i/ → [tɕ h i ] „lavish‟, sòng qǜ /tɕ h y/<br />

→ [tɕ h y ] „send‟. These phenomena are manifestations of weaker metrical strength <strong>in</strong> <strong>the</strong><br />

second syllable of disyllabic words, which support <strong>the</strong> hypo<strong>the</strong>sis of disyllabic trochee.<br />

Foot construction: Follow<strong>in</strong>g <strong>the</strong> Generalized Trochee Hypo<strong>the</strong>sis (Hayes 1991) this study<br />

proposes that <strong>the</strong> metrical foot <strong>in</strong>ventory of Standard Ch<strong>in</strong>ese is Construct (σ σ), else (σµµ).<br />

When a disyllabic trochee cannot be formed, a bimoraic trochee is constructed, e.g., Ft(yǐ zi)<br />

„chair‟, Ft(xīn zàng) „heart‟, [Ft(qì chē) Ft(kù)] „garage‟. Stress <strong>in</strong>teracts with tone <strong>in</strong> ter<strong>ms</strong><br />

that stressed positions must bear tone, while unstressed position is subject to tone deletion.<br />

Conclusion: This study has shown that perceived prom<strong>in</strong>ence may diverge from metrical<br />

structural prom<strong>in</strong>ence under <strong>the</strong> <strong>in</strong>fluence of tone and duration. It is <strong>in</strong>correct to identify<br />

metrical structural prom<strong>in</strong>ence with perceived prom<strong>in</strong>ence. Close exam<strong>in</strong>ation of F0 and<br />

durational patterns as well as <strong>the</strong> distribution of tone and segmental reduction suggests <strong>the</strong><br />

metrical structure is trochaic. The study contributes to <strong>the</strong> general discussion of how acoustic<br />

and perceptual evidence may shed light to phonological representations, and <strong>the</strong> discussion of<br />

tone and stress furnishes ano<strong>the</strong>r example of how tone <strong>in</strong>teracts with stress <strong>in</strong> a tone language.<br />

Selected References: Chang, Mei-Chih Laura. 1992. A prosodic account of tone, stress, and<br />

tone sandhi <strong>in</strong> Ch<strong>in</strong>ese languages. Doctoral dissertation. University of Hawaii. Deng, Dan.<br />

2010. Hanyu Yunluci Yanjiu. [The study of prosodic words <strong>in</strong> Ch<strong>in</strong>ese] Beij<strong>in</strong>g: Beij<strong>in</strong>g<br />

University. Duanmu, San. 2000. 2007. The Phonology of Standard Ch<strong>in</strong>ese. Oxford: Oxford<br />

University Press. Lai, C., Sui, Y. and Yuan, J. 2010. “A corpus study of <strong>the</strong> prosody of<br />

polysyllabic words <strong>in</strong> Mandar<strong>in</strong> Ch<strong>in</strong>ese,” Proceed<strong>in</strong>gs of Speech Prosody 2010, 100457:1-4.<br />

Meredith, Scott. 1990. Issues <strong>in</strong> <strong>the</strong> phonology of prom<strong>in</strong>ence. Doctoral dissertation. MIT.


Megan Sutton, Michael Fetters & Jeffrey Lidz (Maryland)<br />

Pars<strong>in</strong>g for Pr<strong>in</strong>ciple C at 30-months<br />

In study<strong>in</strong>g <strong>the</strong> acquisition of any constra<strong>in</strong>t, it is important to recognize <strong>the</strong> contribution of<br />

both children’s grammatical representations and <strong>the</strong> process<strong>in</strong>g mechanis<strong>ms</strong> required to deploy<br />

<strong>the</strong>se representations. In order to correctly understand sentences like (1), exhibit<strong>in</strong>g Pr<strong>in</strong>ciple C,<br />

children must m<strong>in</strong>imally (a) access lexical <strong>in</strong>formation, (b) build <strong>the</strong> phrase structure and (c)<br />

<strong>in</strong>tegrate <strong>the</strong>se with Pr<strong>in</strong>ciple C to constra<strong>in</strong> <strong>the</strong> mean<strong>in</strong>g.<br />

(1) She’s patt<strong>in</strong>g Katie!<br />

Importantly, even adult-like behavior <strong>in</strong> Pr<strong>in</strong>ciple C contexts does not directly reflect adultlike<br />

knowledge: <strong>the</strong>re are a number of non-adultlike strategies children could rely on to yield <strong>the</strong><br />

same behavior. However, <strong>the</strong>se non-adultlike strategies are <strong>in</strong>dependent of structure; only adultlike,<br />

structure-dependent grammatical knowledge of Pr<strong>in</strong>ciple C predicts that performance on a<br />

Pr<strong>in</strong>ciple C task would be dependent on variation <strong>in</strong> speed of structure-build<strong>in</strong>g. Build<strong>in</strong>g on<br />

previous research show<strong>in</strong>g that early understand<strong>in</strong>g of sentences like (1) is affected by<br />

vocabulary size but not efficiency of lexical access ([1]), we explored <strong>the</strong> effects of Vocabulary,<br />

Lexical Access Speed (LAS), and Phrase Structure Integration Speed (PSIS) on 30-month-olds’<br />

comprehension of Pr<strong>in</strong>ciple C. We show that all 30-month-olds appropriately represent Pr<strong>in</strong>ciple<br />

C, as efficiency of pars<strong>in</strong>g <strong>in</strong> Pr<strong>in</strong>ciple C contexts is dependent on speed of structure-build<strong>in</strong>g.<br />

We tested 32 30-month-olds (28;3-31;25,M=30;1) on three preferential look<strong>in</strong>g tasks. In<br />

experiment 1 (Figure 1), test<strong>in</strong>g Pr<strong>in</strong>ciple C, children saw 8 trials which simultaneously<br />

presented a reflexive and non-reflexive event with a non-reflexive sentence (1). We measured <strong>the</strong><br />

proportion look<strong>in</strong>g to <strong>the</strong> non-reflexive video. Experiments 2 and 3 tested <strong>the</strong> same children to<br />

obta<strong>in</strong> measures LAS and PSIS, respectively. In experiment 2 children saw 8 trials featur<strong>in</strong>g 2<br />

familiar objects with an audio prompt to f<strong>in</strong>d one of <strong>the</strong> objects (2). In experiment 3, children<br />

saw 3 objects of <strong>the</strong> same k<strong>in</strong>d, <strong>the</strong> largest of which differed <strong>in</strong> color. In 12 control trials,<br />

children were prompted to f<strong>in</strong>d <strong>the</strong> biggest object (3), and <strong>in</strong> 12 test trials <strong>the</strong>y were prompted to<br />

f<strong>in</strong>d <strong>the</strong> biggest of <strong>the</strong> two same-color objects (4). Children were divided <strong>in</strong>to ‘fast’ and ‘slow’<br />

groups for LAS and each condition of PSIS by <strong>the</strong> median response latency on distractor-<strong>in</strong>itial<br />

trials [2]. We found no significant correlation between any of <strong>the</strong>se four measures (Vocabulary,<br />

LAS, PSIS:superlative, and PSIS:superlative+adjective; all p>.1), allow<strong>in</strong>g us to use <strong>the</strong>se<br />

measures and <strong>the</strong>ir median-split groups as <strong>in</strong>dependent covariates <strong>in</strong> analyz<strong>in</strong>g Experiment 1.<br />

(2) Where’s <strong>the</strong> fish?<br />

(3) Where’s <strong>the</strong> biggest tra<strong>in</strong>?<br />

(4) Where’s <strong>the</strong> biggest red tra<strong>in</strong>?<br />

We analyzed distractor-<strong>in</strong>itial trials (where <strong>the</strong> child was look<strong>in</strong>g at <strong>the</strong> reflexive event at<br />

<strong>the</strong> onset of <strong>the</strong> word Katie <strong>in</strong> (1)) across <strong>the</strong> 300-3000<strong>ms</strong> post-onset w<strong>in</strong>dow. Both median split<br />

groups for all four covariate measures reach significantly above chance (all p


[1] Sutton, M., Lukyanenko, C. &<br />

Lidz, J. (2011). The Onset of<br />

Pr<strong>in</strong>ciple C at 30 Months: <strong>the</strong> role<br />

of vocabulary, syntactic<br />

development, and process<strong>in</strong>g<br />

efficiency. Proceed<strong>in</strong>gs of BUCLD<br />

35. Cascadilla Press: Cambridge.<br />

[2] Sw<strong>in</strong>gley, D., P<strong>in</strong>to, J., & Fernald,<br />

A. (1999). Cont<strong>in</strong>uous process<strong>in</strong>g<br />

<strong>in</strong> word recognition at 24 months.<br />

Cognition, 71, 73-108.!<br />

!


North Sámi Pronouns<br />

Peter Svenonius (Tro<strong>ms</strong>ø/CASTL)<br />

This paper uses a detailed analysis of North Sámi data to clarify some important<br />

issues <strong>in</strong> <strong>the</strong> analysis of pronouns crossl<strong>in</strong>guistically. A central phenomenon analyzed<br />

is shifted <strong>in</strong>dexicals, which are different <strong>in</strong> North Sámi from o<strong>the</strong>r languages previously<br />

described. The account provides support to those analyses which posit an operator <strong>in</strong> <strong>the</strong><br />

left periphery supply<strong>in</strong>g personal pronouns with <strong>the</strong>ir referents (e.g. Bianchi 2003 <strong>in</strong>ter<br />

alia), but at <strong>the</strong> same time it provides evidence for locat<strong>in</strong>g certa<strong>in</strong> semantic features <strong>in</strong><br />

<strong>the</strong> pronouns <strong>the</strong><strong>ms</strong>elves (as <strong>in</strong> Schlenker 2003). Thus it reconciles what have been cast<br />

as alternatives <strong>in</strong> <strong>the</strong> literature. In addition, <strong>the</strong> analysis of <strong>the</strong> pronouns <strong>the</strong><strong>ms</strong>elves<br />

supports a more f<strong>in</strong>e-gra<strong>in</strong>ed approach to pronom<strong>in</strong>al semantics than is usually assumed.<br />

The k<strong>in</strong>d of shifted <strong>in</strong>dexicals which are central to <strong>the</strong> analysis are illustrated <strong>in</strong> (1),<br />

where <strong>the</strong> embedded clauses conta<strong>in</strong> shifted (i.e. non<strong>in</strong>dexical) second-person pronouns<br />

(and agreement, <strong>in</strong> (1b)), as well as unshifted third person (‘logophoric’) bound pronouns.<br />

(1) a. INgá lohkai: Ii son váldde<br />

du.<br />

Inga said not.3sg s/helog have.pres.conneg you.sg.acc<br />

‘Ingai said shei wouldn’t have “you” [<strong>in</strong> marriage]’<br />

b. De bohtet cizáˇzat ja lohket: Ehpet<br />

<strong>the</strong>n came.past.3pl sparrow.pl.nom and said.past.3pl not.2pl<br />

dii nagot<br />

váldit dan, muhto sii<br />

you.pl.nom manage.pres.conneg take.<strong>in</strong>f it but <strong>the</strong>ylog<br />

nagodit, sii leat nu ollugat.<br />

manage.pres.3pl <strong>the</strong>y are.pres.3pl so many<br />

‘Then sparrowsi came and said “you” can’t carry it, but <strong>the</strong>yi can, <strong>the</strong>yi are<br />

so numerous’<br />

Splitt<strong>in</strong>g and Lump<strong>in</strong>g. Semantic analyses of pronouns posit various k<strong>in</strong>ds of dist<strong>in</strong>ctions<br />

<strong>in</strong> <strong>the</strong> ways that <strong>the</strong>y pick up <strong>the</strong>ir reference. For example, it is often assumed that<br />

bound anaphoric and freely referr<strong>in</strong>g pronouns are underly<strong>in</strong>gly dist<strong>in</strong>ct, even when <strong>the</strong>y<br />

are <strong>the</strong> same on <strong>the</strong> surface (e.g. Cho<strong>ms</strong>ky’s 1955:525 he and he*).<br />

At <strong>the</strong> same time, <strong>the</strong>re is a trend <strong>in</strong> many semantic analyses toward unification of<br />

different k<strong>in</strong>ds of pronouns, e.g. different bound elements such as bound pronouns, fake<br />

reflexives, anaphors, PRO, etc. (Kratzer 2008, where <strong>the</strong> real semantic action takes place<br />

<strong>in</strong> <strong>the</strong> verbal heads, and <strong>the</strong> spell-out of pronouns is mostly just morphology, or Elbourne<br />

2008 <strong>in</strong>ter alia, where various k<strong>in</strong>ds of b<strong>in</strong>d<strong>in</strong>g are unified with donkey anaphora under<br />

a D-type analysis).<br />

Fake <strong>in</strong>dexicals (as <strong>in</strong> Heim’s Only I did my homework) appear to show both that<br />

one morphological form can pick up reference <strong>in</strong> different ways, and that bound pronouns<br />

can surface with more than one morphological form.<br />

This state of affairs might be taken to suggest that semantics is autonomous from<br />

morphosyntax, and that <strong>the</strong> dist<strong>in</strong>ctions made <strong>in</strong> one system simply fail to match <strong>the</strong><br />

dist<strong>in</strong>ctions made <strong>in</strong> <strong>the</strong> o<strong>the</strong>r. However, such a conclusion would be premature. Ra<strong>the</strong>r,<br />

I suggest, pronom<strong>in</strong>al elements are complex underly<strong>in</strong>gly. Bound pronom<strong>in</strong>al elements<br />

share some feature patterns with one ano<strong>the</strong>r, which is why <strong>the</strong>y can behave similarly.<br />

The surface exponents which spell out pronom<strong>in</strong>al syntactic structures are underspecified,<br />

which allows <strong>the</strong>m to match a range of different underly<strong>in</strong>g structures, which is why one<br />

surface pronoun can behave <strong>in</strong> several different ways.<br />

Crossl<strong>in</strong>guistic evidence and Shifted <strong>in</strong>dexicals. In order to see this, it is necessary<br />

1


to consider cross-l<strong>in</strong>guistic evidence. There are many languages <strong>in</strong> which <strong>the</strong>re are multiple<br />

series of pronouns, overtly reflect<strong>in</strong>g dist<strong>in</strong>ctions which o<strong>the</strong>r languages syncretize<br />

systematically at <strong>the</strong> surface. The logophors of Ewe (Clements 1975) or <strong>the</strong> obviative<br />

markers <strong>in</strong> Algonquian languages (Goddard 1984) are examples. However, detailed<br />

descriptions of <strong>the</strong>se understudied languages are usually not semantically current, and<br />

cutt<strong>in</strong>g-edge semantics does not have access to <strong>the</strong> details of <strong>the</strong>se languages necessary<br />

for choos<strong>in</strong>g among hypo<strong>the</strong>ses.<br />

An exception is Anand & Nev<strong>in</strong>s’ (2004) <strong>in</strong>vestigation of shifted <strong>in</strong>dexicals <strong>in</strong> Zazaki.<br />

Observ<strong>in</strong>g that <strong>in</strong>dexicals shift toge<strong>the</strong>r, <strong>the</strong>y argue aga<strong>in</strong>st Schlenker’s (2003) and von<br />

Stechow’s (2002) lexical approaches. Instead, <strong>the</strong>y posit an operator at <strong>the</strong> periphery<br />

of <strong>the</strong> embedded clause which shifts <strong>the</strong> context for <strong>the</strong> entire clause. In support of<br />

this, observe that <strong>in</strong> <strong>the</strong> North Sámi examples presented above, <strong>the</strong>re is no overt b<strong>in</strong>der<br />

referr<strong>in</strong>g to <strong>the</strong> addressee for <strong>the</strong> shifted pronouns. Instead, <strong>the</strong> only bound pronouns<br />

are third person. I argue that <strong>the</strong> operator <strong>in</strong> <strong>the</strong> left periphery can b<strong>in</strong>d as well as shift<br />

context (mak<strong>in</strong>g it more like <strong>the</strong> operators posited by Baker, Biachi, or SigurDsson).<br />

However, <strong>the</strong>re is also evidence from North Sámi that some critical features are located<br />

<strong>in</strong> <strong>the</strong> pronoun itself. North Sámi makes a four-way dist<strong>in</strong>ction <strong>in</strong> ‘def<strong>in</strong>ite’ pronom<strong>in</strong>al<br />

elements (sett<strong>in</strong>g aside deictics, <strong>in</strong>def<strong>in</strong>ites, <strong>in</strong>terrogatives, and o<strong>the</strong>rs). First, it dist<strong>in</strong>guishes<br />

locally bound anaphoric arguments and adjuncts (ieˇs, with various case for<strong>ms</strong>)<br />

both from lexical reflexives (with -d-) and from o<strong>the</strong>r pronom<strong>in</strong>al elements. Second,<br />

it uses pro-drop to dist<strong>in</strong>guish two levels of what Gundel (1999, 2003) calls referential<br />

givenness. F<strong>in</strong>ally, it has two series of overt pronouns (dat, with various case for<strong>ms</strong>, and<br />

son, with various case for<strong>ms</strong>) to dist<strong>in</strong>guish two levels of relational givenness. These are<br />

illustrated <strong>in</strong> (2).<br />

(2) a. Máhtte muitalii, ahte das lea goahti.<br />

Matte.nom told that pn.loc is hut.nom<br />

‘Matte expla<strong>in</strong>ed that he (some else) had a hut’<br />

b. Máhtte muitalii, ahte sus lea goahti.<br />

Matte.nom told that log.loc is hut.nom<br />

‘Matte expla<strong>in</strong>ed that he (Matte) had a hut’<br />

For convenience, <strong>the</strong>se can be referred to as <strong>the</strong> deictic and logophoric pronouns, respectively,<br />

though I will show how <strong>the</strong> first is not strictly deictic and <strong>the</strong> second is different<br />

from <strong>the</strong> logophoric pronouns described <strong>in</strong> <strong>the</strong> previous literature. Ra<strong>the</strong>r, <strong>the</strong>se two<br />

pronoun series dist<strong>in</strong>guish b<strong>in</strong>d<strong>in</strong>g from non-b<strong>in</strong>d<strong>in</strong>g <strong>in</strong> someth<strong>in</strong>g like <strong>the</strong> way discussed<br />

above. The dist<strong>in</strong>ction, I suggest, is made <strong>in</strong> o<strong>the</strong>r languages but systematically syncretized<br />

<strong>the</strong>re.<br />

I show that second person (but not first person) pronouns <strong>in</strong> examples like (1) can<br />

pattern with <strong>the</strong> ‘logophoric’ son-series. This requires a four-way dist<strong>in</strong>ction <strong>in</strong> features,<br />

where most analyses posit at most a two-way dist<strong>in</strong>ction: There are strictly <strong>in</strong>dexical<br />

elements (first person), underspecified elements which can be <strong>in</strong>dexical or shifted/bound<br />

(second person), obligatory bound elements (third person ‘logophoric’ series), and obligatorily<br />

unbound non<strong>in</strong>dexicals (<strong>the</strong> ‘deictic’ third person series).<br />

2


Kriszta Szendrői (UCL)<br />

Quantifier Rais<strong>in</strong>g is blocked by passives<br />

1. Quantifier Rais<strong>in</strong>g (QR) is a covert syntactic movement operation, <strong>in</strong>volved <strong>in</strong> sentences with<br />

an <strong>in</strong>def<strong>in</strong>ite subject and a universal object:<br />

(1) A dog chased every cat. distributive read<strong>in</strong>g LF: [[every cat] i [a dog chased t i]]<br />

In psychol<strong>in</strong>guistic experiments, <strong>in</strong>verse scope is often found to be hard for native speakers to<br />

access (Ioup 1975, Van Lehn 1978, Catl<strong>in</strong> & Micham 1975; especially Kurtzmann & MacDonald<br />

1993, Tunstall 1998 and Anderson 2004). In this paper, I show that one of <strong>the</strong> reasons why <strong>the</strong><br />

distributive read<strong>in</strong>g is hard to access is <strong>the</strong> existence of an alternative derivation with (sufficiently)<br />

identical <strong>in</strong>terpretation,― <strong>the</strong> passive variant of (1):<br />

(2) Every cat was chased by a dog. distributive read<strong>in</strong>g LF: [[every cat] i [was chased e i by a dog]]<br />

2. Theoretical background: If this turns out to be correct, it has important <strong>the</strong>oretical<br />

implications. Most <strong>the</strong>oreticians agree that QR is restricted by <strong>in</strong>terface considerations: Fox (1995,<br />

2000), Re<strong>in</strong>hart (2006) and Bobaljik & Wurmbrand (2011), henceforth B&W. But <strong>the</strong>y disagree<br />

about <strong>the</strong> nature of <strong>the</strong> <strong>in</strong>terface mechanis<strong>ms</strong> at play. Fox (2000:26) argued that <strong>the</strong> restriction<br />

applies locally <strong>in</strong> <strong>the</strong> syntactic derivation, prevent<strong>in</strong>g scope-shift<strong>in</strong>g operations unless <strong>the</strong>y have an<br />

effect. So, <strong>in</strong> <strong>the</strong> case of QR, <strong>the</strong> movement step is disallowed if it does not lead to a change <strong>in</strong><br />

<strong>in</strong>terpretation. Re<strong>in</strong>hart (2006:28) claimed that a global comparison of compet<strong>in</strong>g derivations is<br />

necessary, her ‘reference set’ computation: ‘<strong>the</strong> reference set consists of pairs of derivation<br />

and <strong>in</strong>terpretation, where <strong>the</strong> <strong>in</strong>terpretation i is identical <strong>in</strong> all pairs.’ 1 So, QR is allowed if <strong>the</strong>re is no<br />

alternative derivation without QR with<strong>in</strong> <strong>the</strong> reference set that has <strong>the</strong> same (i.e. distributive)<br />

<strong>in</strong>terpretation. The crucial question, which Re<strong>in</strong>hart rema<strong>in</strong>ed implicit about, is what determ<strong>in</strong>es<br />

membership <strong>in</strong> <strong>the</strong> reference set. As B&W:31, cit<strong>in</strong>g Fox (2000), correctly note, if membership is<br />

restricted to derivations with a shared numeration, <strong>the</strong> global computation can be easily reduced to a<br />

local one. With respect to <strong>the</strong> relevance of examples like (1)-(2) for QR, B&W take an explicit stand:<br />

‘[…] we assume that economy conditions only evaluate compet<strong>in</strong>g derivations from <strong>the</strong> same<br />

numeration (<strong>in</strong>put), and thus that correspond<strong>in</strong>g active and passive sentences will simply not<br />

compete with one ano<strong>the</strong>r.’ (B&W:7)<br />

3. Proposal: So, <strong>in</strong> <strong>the</strong> spirit of Re<strong>in</strong>hart’s approach and <strong>in</strong> contrast to B&W’s position, I propose<br />

that membership <strong>in</strong> a particular reference set is restricted by semantic identity <strong>in</strong> <strong>the</strong> sense of truthconditional<br />

equivalence. In <strong>the</strong> specific case of active-passive pairs, like (1) and (2), truth-conditional<br />

equivalence is trivially true, except for marg<strong>in</strong>al cases such as utterances <strong>in</strong>volv<strong>in</strong>g NPs that fail to<br />

refer (such as <strong>the</strong> k<strong>in</strong>g of France), as <strong>the</strong>se may lead to truth-value gaps, while <strong>the</strong>ir passive variant may<br />

not. (We also need to put generic sentences of <strong>the</strong> type Beavers build da<strong>ms</strong> to one side.) My hypo<strong>the</strong>sis<br />

makes <strong>the</strong> follow<strong>in</strong>g testable prediction. If passive variants of active sentences <strong>in</strong>deed block <strong>the</strong><br />

distributive <strong>in</strong>terpretation of <strong>the</strong> active variant by QR, <strong>the</strong>n <strong>in</strong>verse scope (by QR) should be more<br />

easily accessible <strong>in</strong> sentences that <strong>in</strong>volve verbs that resist passivisation. This is because <strong>in</strong> such<br />

cases, no passive competitor can be present <strong>in</strong> <strong>the</strong> reference set, as such a passive competitor does<br />

not exist, and thus, <strong>the</strong> derivation with QR could not be blocked by it. In contrast, if, as Fox<br />

suggested, decisions are taken locally, at <strong>the</strong> po<strong>in</strong>t <strong>in</strong> <strong>the</strong> syntactic derivation when QR applies, <strong>the</strong>n<br />

<strong>the</strong> passive variant would simply not be considered as a viable alternative. The same applies to<br />

B&W’s proposal, who chose to base membership on shared numeration. Thus, both for Fox and<br />

B&W, it should be irrelevant for <strong>the</strong> availability of <strong>in</strong>verse scope whe<strong>the</strong>r a particular (active) sentence<br />

1 B&W’s (p33 l6-8) <strong>in</strong>terpretation of Re<strong>in</strong>hart’s proposal is <strong>the</strong> direct opposite: ‘Re<strong>in</strong>hart holds that <strong>the</strong><br />

calculation <strong>in</strong>volves <strong>the</strong> competition among <strong>the</strong> members of a ‘reference set’ consist<strong>in</strong>g of derivations (LF,PF<br />

pairs) with a common PF but different LFs.’ I have not managed to ascerta<strong>in</strong> <strong>the</strong> source of this <strong>in</strong>terpretation.


<strong>in</strong>volves a passivis<strong>in</strong>g or non-passivis<strong>in</strong>g verb.<br />

It is important to clarify that, I do not necessarily expect that passive variants always block active<br />

variants with QR. (That would mean that <strong>the</strong> distributive read<strong>in</strong>g is never accessed if <strong>the</strong> verb has a<br />

passive alternant.) They only do so if <strong>the</strong>y are actually listed <strong>in</strong> <strong>the</strong> reference set at <strong>the</strong> po<strong>in</strong>t when<br />

<strong>the</strong> hearer considers <strong>in</strong>verse scope. This is a performance issue, and should be determ<strong>in</strong>ed by an<br />

appropriate <strong>the</strong>ory of language use. While I cannot offer a specific proposal at this po<strong>in</strong>t, it see<strong>ms</strong><br />

clear to me that general discourse considerations such as topic-focus articulation (see Sæbø 1997),<br />

<strong>in</strong>dividual differences between speakers (for <strong>in</strong>stance an aptitude to consider richer mean<strong>in</strong>gs e.g.<br />

metaphors, irony), and attention <strong>in</strong> <strong>the</strong> psychological sense would be relevant factors. What matters<br />

here is that if any evidence is found that <strong>in</strong>verse scope is obta<strong>in</strong>ed more easily <strong>in</strong> sentences <strong>in</strong>volv<strong>in</strong>g<br />

verbs that are not passivisable, that would favour Re<strong>in</strong>hart’s global <strong>the</strong>ory over its competitors.<br />

4. Experiment: I determ<strong>in</strong>ed a set of 7 passivis<strong>in</strong>g-nonpassivis<strong>in</strong>g verb pairs by a grammaticality<br />

judgment pre-test. Then, 50 native speakers were presented with a forced-choice questionnaire<br />

(adapted from Anderson 2004) with sentences like (3)-(4). Their task was to circle <strong>the</strong> read<strong>in</strong>g that is<br />

closest to <strong>the</strong>ir mean<strong>in</strong>g of <strong>the</strong> sentence. The ‘same’-paraphrase corresponds to <strong>the</strong> overt scope<br />

read<strong>in</strong>g and <strong>the</strong> ‘different’-paraphrase corresponds to <strong>the</strong> <strong>in</strong>verse scope read<strong>in</strong>g.<br />

(3) Passivis<strong>in</strong>g condition: A road connects every village on <strong>the</strong> pen<strong>in</strong>sula.<br />

The same road connects <strong>the</strong> villages. Different roads connect <strong>the</strong> villages.<br />

(4) Non-passivis<strong>in</strong>g condition: A road leads to every village on <strong>the</strong> pen<strong>in</strong>sula.<br />

The same road leads to <strong>the</strong> villages. Different roads leads to <strong>the</strong> villages.<br />

All participants saw all ite<strong>ms</strong>, but <strong>in</strong> different orders. The ite<strong>ms</strong> were <strong>in</strong>terspersed with an equal<br />

number of fillers and controls, and pseudo-randomised. The paraphrases were presented <strong>in</strong> different<br />

orders, balanced across conditions. Crucially, each test item pair was identical except for <strong>the</strong> verbs.<br />

5. Results and discussion: Speakers obta<strong>in</strong>ed <strong>the</strong> <strong>in</strong>verse scope read<strong>in</strong>g 25.4% of <strong>the</strong> time <strong>in</strong> <strong>the</strong><br />

non-passivis<strong>in</strong>g condition compared with 14.3% <strong>in</strong> <strong>the</strong> passiviz<strong>in</strong>g verb condition: a statistically<br />

significant difference (participants t=3.963, df=40, two-tailed p


Barbara Tomaszewicz (USC)<br />

A family of exclusives <strong>in</strong> Polish<br />

Polish has four different focus associat<strong>in</strong>g adverbs that can be identified as exclusives, all of<br />

<strong>the</strong>m scalar, yet not always <strong>in</strong>terparaphrasable. Tylko, ledwie, dopiero and aż all evoke<br />

alternative propositions ordered on a scale and imply an exclusion of those alternatives that<br />

are higher or lower on <strong>the</strong> scale than <strong>the</strong> prejacent. What varies is (i) <strong>the</strong> nature of <strong>the</strong><br />

exclusive implication (truth-conditional or pragmatic), (ii) <strong>the</strong> nature of <strong>the</strong> scale (likelihood<br />

vs. time vs. highly context-dependent pragmatic scales), (iii) <strong>the</strong> place of <strong>the</strong> prejacent on <strong>the</strong><br />

scale. Their common mean<strong>in</strong>g components, scalarity and exclusivity, <strong>in</strong>teract with (i-iii)<br />

thus creat<strong>in</strong>g a range of <strong>in</strong>terpretations that cannot be subsumed under <strong>the</strong> notions of<br />

unexpectedness and/or likelihood. The Polish <strong>in</strong>ventory raises a <strong>the</strong>oretically important<br />

question whe<strong>the</strong>r scalarity and exclusivity are generally l<strong>in</strong>ked to unexpectedness/likelihood<br />

or whe<strong>the</strong>r <strong>the</strong> dimension of <strong>the</strong> (default) scale should ra<strong>the</strong>r be lexically specified for each<br />

adverb, e.g. even as always <strong>in</strong>terpreted wrt. a scale of likelihood (Karttunen & Peters 1979)<br />

or its dual, noteworth<strong>in</strong>ess (Herburger 2000, Rullman 2007).<br />

(i) The exclusive component can be targeted by explicit denial, thus for all four exclusives<br />

it is an at-issue entailment (part of <strong>the</strong> assertion). Tylko/ledwie, like English only, can have<br />

quantificational (1b) or scalar (1c) read<strong>in</strong>gs and, and <strong>the</strong>ir prejacents project. The prejacent of<br />

dopiero and aż is at-issue (2-3), and aż cannot have a quantificational read<strong>in</strong>g (3b).<br />

(1) a. Maria widziała tylko/ledwie menadżera.<br />

Maria saw manager<br />

b. No, that’s false. – #She didn’t see <strong>the</strong> manager./She saw somebody else too.<br />

c. No, that’s false. – #She didn’t see <strong>the</strong> manager./She saw somebody more special.<br />

(2) a. Maria widziała dopiero menadżera.<br />

Maria saw manager<br />

b. No, that’s false. – She didn’t see <strong>the</strong> manager (yet)./She saw somebody else too.<br />

c. No, that’s false. – She didn’t see <strong>the</strong> manager (yet)./She saw somebody more special.<br />

(3) a. Maria widziała aż menadżera.<br />

Maria saw manager<br />

b. No, that’s false. – She didn’t see <strong>the</strong> manager./#She saw somebody else too.<br />

c. No, that’s false. – She didn’t see <strong>the</strong> manager./She saw somebody less special.<br />

The family of sentences tests (Chierchia & McConnell-G<strong>in</strong>et 1990) yield <strong>the</strong> same results.<br />

(ii) The quantificational read<strong>in</strong>gs of exclusives can be seen as a subtype of scalar read<strong>in</strong>gs:<br />

<strong>the</strong> relevant scale is a entailment scale (Kle<strong>in</strong>dienst 2005, Riester 2006); <strong>the</strong> rank<strong>in</strong>g of<br />

answers to <strong>the</strong> Question Under Discussion (Roberts 1996) is a boolean lattice (Beaver and<br />

Clark’s 2008). A scale not organized accord<strong>in</strong>g to logical entailment will have its dimension<br />

set by <strong>the</strong> context, e.g. importance <strong>in</strong> (1c, 2c, 3c), or temporal order<strong>in</strong>g <strong>in</strong> (2b,c) (c.f. German<br />

erst, König 1979, von Stechov 2006). Thus, scalarity does not automatically lead to a sense<br />

of unexpectedness/surprise. For (4) <strong>the</strong>re is no sense <strong>in</strong> which <strong>the</strong> speaker or <strong>the</strong> hearer would<br />

expect <strong>the</strong> team to do any better; each adverb contributes <strong>the</strong> read<strong>in</strong>g that <strong>the</strong> 30 th place is<br />

very low (i.e. lower than a contextually set degree, but not necessarily unexpected).<br />

(4) Zespół był słaby i zajął tylko/ledwie/dopiero/aż 30 miejsce.<br />

team was poor and achieved 30 th place<br />

‘The team performed poorly and earned only 30 th place.’<br />

Scalarity with evaluativity but without likelihood is commonly found with gradable<br />

predicates that license <strong>the</strong> use of very (e.g. The baby's relatives are all more than 6ft tall. The<br />

baby is go<strong>in</strong>g to be very tall.). Evaluativity does not always relate to expectations.<br />

(iii) The place of <strong>the</strong> prejacent is low on <strong>the</strong> scale with tylko/ledwie/dopiero (cf. English<br />

only, Kl<strong>in</strong>edienst 2005, König 1991, von Rooy & Schultz 2005) and high with aż, which is<br />

shown by <strong>the</strong> contrast between (1c, 2c) and (3c). The alternatives excluded with aż are higher<br />

on <strong>the</strong> scale than <strong>the</strong> prejacent, thus aż is a scalar opposite of <strong>the</strong> o<strong>the</strong>rs (cf. beyond-operators<br />

<strong>in</strong> Gast & van der Auwera 2011). The cont<strong>in</strong>uation <strong>in</strong> (5a) and (b) sounds like a


contradiction, while <strong>in</strong> (5c) <strong>the</strong> quantity implicature is more easily cancelled (with even<br />

plac<strong>in</strong>g <strong>the</strong> prejacent high on <strong>the</strong> scale of noteworth<strong>in</strong>ess.)<br />

(5) a. Marek potrafił zjeść aż 5 bananów, #a właściwie 6.<br />

Marek was.able to.eat 5 bananas and actually 6<br />

‘Marek was able to eat as many as 5 bananas, #<strong>in</strong> fact, he could eat 6’.<br />

b. Marek potrafił zjeść tylko/ledwie/dopiero 5 bananów. #a właściwie 6.<br />

Marek was.able to.eat 5 bananas and actually 6<br />

‘Marek was able to eat only 5 bananas, #<strong>in</strong> fact, he could eat 6’.<br />

c. Marek potrafił zjeść nawet 5 bananów, ?a właściwie 6.<br />

Marek was.able to.eat even 5 bananas and actually 6<br />

‘Marek was able to eat even 5 bananas, ?<strong>in</strong> fact, he could eat 6’.<br />

The reason (4) works is that <strong>the</strong> 30 th place is low on <strong>the</strong> “scale of success”, and at <strong>the</strong> same<br />

time high on <strong>the</strong> scale evoked by <strong>the</strong> numeral. (6) shows that <strong>the</strong> scale aż refers to is not<br />

coercible <strong>in</strong>to a scale of success.<br />

(6) Susan Boyle nie wygrała. Zajęła dopiero/#aż drugie miejsce.<br />

‘Susan Boyle did not w<strong>in</strong>. She only took second place.’<br />

If <strong>the</strong> dimension of <strong>the</strong> scale is underspecified <strong>in</strong> <strong>the</strong> semantics of each adverb, we expect it<br />

to be easily manipulated by <strong>the</strong> context (<strong>the</strong> semantics of focused constituent <strong>the</strong> adverb<br />

associates with and <strong>the</strong> pragmatics). The four Polish exclusives, however, appear to be more<br />

contextually restricted. Tylko cannot always get a scalar read<strong>in</strong>g, e.g. <strong>in</strong> (7) it can only mean<br />

<strong>the</strong> Marek woke up once. Exclusively ledwie can receive approximate (8) and counterfactual<br />

(9) read<strong>in</strong>gs.<br />

(7) Martwi mnie, że Marek obudził się #tylko/ledwie/dopiero/aż o godz<strong>in</strong>ie 13.<br />

worries me that Marek woke self at hour 13<br />

‘I am disappo<strong>in</strong>ted that Marek woke up as late as at 13h.’<br />

(8) #Tylko/ledwie/#Dopiero/#Aż Marek otworzył drzwi, a zadzwonił telefon.<br />

Marek opened door and rang telephone<br />

‘Marek barely opened <strong>the</strong> door, when <strong>the</strong> telephone rang.’<br />

(9) Marek #tylko/ledwie/#dopiero/#aż otworzył drzwi.<br />

Marek opened door<br />

‘Marek barely opened <strong>the</strong> door. (He thought he was not go<strong>in</strong>g to make it.)’<br />

The four Polish exclusives raise <strong>the</strong> question how much underspecified can <strong>the</strong> lexical<br />

mean<strong>in</strong>g be. While all of <strong>the</strong>m share <strong>the</strong> scalar mean<strong>in</strong>g component (10a), <strong>the</strong> scale as<br />

def<strong>in</strong>ed <strong>in</strong> (10c) cannot be entirely context dependent, so perhaps more f<strong>in</strong>e gra<strong>in</strong>ed<br />

descriptions (as <strong>in</strong> (10a-d) for dopiero and ledwie) should be specified for <strong>in</strong>dividual<br />

exclusives cross-l<strong>in</strong>guistically? How much can mirativity be detached from evaluativity<br />

which is a general property of degree related expressions (Rett 2010)?<br />

(10) “tylkoC ϕ” “dopieroC ϕ” “ledwieC ϕ” “aż/čakC ϕ”<br />

a. λw.¬∃p [p ∈S & p(w) = 1 & ⟦ϕ⟧


Satoshi Tomioka (Delaware)<br />

Focus Matters <strong>in</strong> Neo-Hambl<strong>in</strong> Semantics<br />

The recent re-emergence of Hambl<strong>in</strong> semantics of <strong>in</strong>terrogatives (e.g., L<strong>in</strong> 1996, Shimoyama<br />

1999, Kratzer and Shimoyama 2005) has proved fruitful <strong>in</strong> many areas, <strong>in</strong>clud<strong>in</strong>g <strong>the</strong> analysis<br />

of wh-<strong>in</strong>def<strong>in</strong>ites. In Beck’s (2006) rendition of Hambl<strong>in</strong> semantics, two additional advantages<br />

are ga<strong>in</strong>ed. (i) A formal relation between a wh-phrase and a Q-morpheme is established:<br />

Wh-phrases only have focus values (= <strong>the</strong>ir Hambl<strong>in</strong> denotations), lack<strong>in</strong>g ord<strong>in</strong>ary values.<br />

Any constituent that conta<strong>in</strong>s a wh-phrase also has no ord<strong>in</strong>ary value until it meets a Q-<br />

Operator <strong>in</strong> <strong>the</strong> derivation. The role of <strong>the</strong> Q-Operator is to elevate <strong>the</strong> focus value of a<br />

wh-conta<strong>in</strong><strong>in</strong>g constituent to <strong>the</strong> ord<strong>in</strong>ary value. (ii) The prosodic similarities between a<br />

wh-phrase and focus is expla<strong>in</strong>ed: Sabel (2006) and Haida (2007) enlist many languages <strong>in</strong><br />

which wh-phrases act as though <strong>the</strong>y are focused morphosyntactically. Beck’s analysis bodes<br />

well with this wh-(semantic) focus correspondence (cf. Truckenbrodt, to appear).<br />

In this paper, I will fur<strong>the</strong>r exam<strong>in</strong>e <strong>the</strong> wh-focus correlation with special attention<br />

to ‘double-focus’ cases (Krifka 1991, Wold 1996). Both wh-phrases and focus-associates<br />

with operators like only are required to have non-s<strong>in</strong>gleton focus values by <strong>the</strong>ir semantics,<br />

<strong>in</strong>dependently of <strong>the</strong>ir <strong>in</strong>formation structural statuses. Hence, <strong>the</strong>y can be backgrounded<br />

(a.k.a second occurrence focus) or <strong>the</strong>y can be additionally focused, as shown <strong>in</strong> (1).<br />

(1) a.‘Ann asked what Sue likes, right?’ ‘No, she asked WHO she likes.’<br />

b. ‘Ann only greeted BELLA.’ ‘No, she only greeted CASEY.’<br />

The first question that arises is; how can one get <strong>the</strong> mean<strong>in</strong>g of a contrasted wh-phrase<br />

when a wh-phrase itself has no ord<strong>in</strong>ary semantic value? Beck’s semantics can actually deal<br />

with a case like (1a) with a m<strong>in</strong>imal modification: All wh-phrases, even morphologically<br />

simplex ones (e.g., who, what) are decomposed with an NP restriction and a wh-determ<strong>in</strong>er.<br />

Then, (i) The Hambl<strong>in</strong> denotation is derived by <strong>the</strong> choice function mean<strong>in</strong>g of a Detwh.<br />

(ii) A contrastive focus on a wh-phrase is associated with <strong>the</strong> NP restriction. (iii) Both<br />

<strong>the</strong> Detwh and <strong>the</strong> restriction get focus-<strong>in</strong>dices. The first is required by <strong>the</strong> semantics of a<br />

wh, and <strong>the</strong> second comes from <strong>the</strong> constrastiveness. (iv) The <strong>in</strong>dex on <strong>the</strong> Detwh is bound<br />

by <strong>the</strong> closest Q-morpheme, and <strong>the</strong> one on <strong>the</strong> NP ‘passes up’ to a higher focus b<strong>in</strong>der<br />

(i.e., ∼). [Note: A focus <strong>in</strong>dex corresponds to a designated variable that is <strong>in</strong>terpreted by a<br />

dist<strong>in</strong>guished assignment (Kratzer 1991). And importantly; (a) � αFi� g,h = h(i) if i is <strong>in</strong> <strong>the</strong><br />

doma<strong>in</strong> of h. O<strong>the</strong>rwise, � αFi� g,h = � α � g . (b) h beg<strong>in</strong>s with <strong>the</strong> empty doma<strong>in</strong> (= { }).<br />

(c) An F-<strong>in</strong>dex b<strong>in</strong>der <strong>in</strong>troduces its <strong>in</strong>dex <strong>in</strong>to <strong>the</strong> doma<strong>in</strong> of h.] (2) is <strong>the</strong> computation of<br />

<strong>the</strong> second sentence <strong>in</strong> (1a).<br />

(2) a. [who] contrastive focus ⇒ [ whF1 [person]F2]<br />

b. � Q1 [CP [ whF1 [ person]F2] Sue likes ] � g = {p: p = � CP � g,h{f/1}<br />

= {p: p = Sue likes f(person) | f∈D, e >}<br />

c. �∼C2 [IP Anna asked Q1 [ whF1 [ person]F2] Sue likes]] � g,h = {p: p = � IP � g,h{P/2} }<br />

= {p: p =Anna asked which P Sue likes | P∈D< e, t >}<br />

If a focus b<strong>in</strong>der ∼ is also selective, <strong>the</strong> story for (1b) is <strong>the</strong> same except that two <strong>in</strong>dices<br />

are assigned to <strong>the</strong> same constituent (cf. Wold 1996).<br />

(3) a. ∼C2 [ Anna onlyD [∼D1 greeted [[CASEY]2]1]<br />

b. � ∼D1 greeted [[CASEY]2]1� g,h{x/1} = {λy. y greeted x | x ∈ De}<br />

c. � ∼C2 [Anna only greeted [[CASEY]]1]2� g,h{x/2} = {Anna only greeted x | x ∈ De}<br />

1


However, this strategy cannot be extended to (1b) <strong>in</strong> Beck (2006). A focus operator ∼ is<br />

argued to be unselective, and <strong>the</strong> closer operator (∼D above) closes off <strong>the</strong> focus mean<strong>in</strong>g<br />

of Casey, after which no F-<strong>in</strong>dex is left for <strong>the</strong> higher operator (= ∼C) to b<strong>in</strong>d. In order to<br />

make <strong>the</strong> parallelism between (1a) and (1b) <strong>in</strong>tact, I propose that, while a novel ∼ operator<br />

is unselective, a second occurrence ∼ is selective. The <strong>in</strong>tuition beh<strong>in</strong>d this amendment to<br />

Beck’s <strong>the</strong>ory is as follows. Focus sensitive adverbs like only are considered ‘familiar’ not<br />

only because <strong>the</strong>y <strong>the</strong><strong>ms</strong>elves had appeared <strong>in</strong> <strong>the</strong> previous discourse but also because <strong>the</strong>ir<br />

associations with <strong>the</strong> designated foci are familiar. A familiar ∼ operator <strong>the</strong>refore does not<br />

freely extend its w<strong>in</strong>gs and b<strong>in</strong>d novel focus variables.<br />

The proposed amendment not only provides a unified analysis of contrasted wh-phrases<br />

and contrasted focus-associated phrases but also makes correct predictions <strong>in</strong> o<strong>the</strong>r areas.<br />

First, <strong>in</strong> a multiple foci structure where <strong>the</strong>re are two ∼ operators, one novel and <strong>the</strong> o<strong>the</strong>r<br />

familiar, it is expected that <strong>the</strong> novel one must have scope over <strong>the</strong> familiar one. In <strong>the</strong> o<strong>the</strong>r<br />

scope configuration, <strong>the</strong> novel ∼ operator accidentally b<strong>in</strong>ds both foci, leav<strong>in</strong>g noth<strong>in</strong>g for<br />

<strong>the</strong> old operator to b<strong>in</strong>d. This prediction is <strong>in</strong> accordance with von F<strong>in</strong>tel’s (1994, p 49, fn.<br />

44) observation on <strong>the</strong> order<strong>in</strong>g asymmetry of two focus operators. The o<strong>the</strong>r phenomenon<br />

that <strong>the</strong> amended analysis works well is what is called an <strong>in</strong>tervention effect, which Beck<br />

(2006) def<strong>in</strong>es as a focus <strong>in</strong>tervention shown <strong>in</strong> (4). In this configuration, <strong>the</strong> <strong>in</strong>terven<strong>in</strong>g ∼<br />

C (an unselective b<strong>in</strong>der) accidentally b<strong>in</strong>ds <strong>the</strong> wh-phrase.<br />

(4) * [Qi ..... [∼ C [ ...... whi ] ]<br />

There have been several facts reported, however, that suggest that, even <strong>in</strong> <strong>the</strong> structure (4),<br />

discourse-familiar ∼ operators do not <strong>in</strong>duce <strong>in</strong>tervention effects (or <strong>the</strong>ir effects significantly<br />

weakened). In Beck and Kim (2006, p.167), <strong>the</strong> lack of <strong>the</strong> alternative question mean<strong>in</strong>g <strong>in</strong><br />

(5) is attributed to <strong>the</strong> focus <strong>in</strong>tervention effect.<br />

(5) Does only John like Mary or Susan? a. #Mary. [*AltQ] b. Yes. [Yes/NoQ]<br />

Eilam (2011) reports, however, that native speakers more readily accept <strong>the</strong> alternative question<br />

<strong>in</strong>terpretation when only John is backgrounded. Second, <strong>in</strong> many wh-<strong>in</strong>-situ languages<br />

that exhibit <strong>in</strong>tervention effects (e.g., Korean, Japanese, H<strong>in</strong>di), <strong>the</strong> effects are noticeably<br />

weak (or non-existent) <strong>in</strong> why-questions (cf. Miyagawa 1998, Ko 2005 among o<strong>the</strong>rs).<br />

(6) Amwuto/?John-pakkey way ku chayk-ul ilk-ci-anh-ass-ni?<br />

Anyone/John-only why that book-Acc read-CI-Neg-Past-Q<br />

‘Why did nobody/only John read that book?’ (Korean, from Ko 2005, (8a))<br />

Tomioka (2009) attributes this effect to <strong>the</strong> presupposition of why-questions. In ‘why p?’, p is<br />

presupposed, and <strong>the</strong> materials <strong>in</strong> p, <strong>in</strong>clud<strong>in</strong>g an <strong>in</strong>tervener, belong to <strong>the</strong> background. In<br />

o<strong>the</strong>r constituent questions, comparable presuppositions are not guaranteed (Groenendijk<br />

and Stokhof 1984). Third, while many languages prefer plac<strong>in</strong>g phonologically reduced<br />

materials after focus, Amharic can put reduced focus expressions (e.g., only NP) before<br />

wh-phrases. Eilam (2011) po<strong>in</strong>ts out that Amharic shows no <strong>in</strong>tervention effects with those<br />

reduced focus expressions, as expected by <strong>the</strong> current proposal.<br />

Selected References: Beck, S. 2006: ‘Intervention effects follow from focus <strong>in</strong>terpretations,’ <strong>in</strong> NLS, 14:1-<br />

56. Eilam, A. 2011: Explorations <strong>in</strong> <strong>the</strong> Informational Component, Ph.D. <strong>the</strong>sis, UPenn. Ko, H.-J. 2005:<br />

Syntax of why-<strong>in</strong>-situ: Merge <strong>in</strong>to [Spec CP] <strong>in</strong> <strong>the</strong> overt yyntax, NLLT, 23, 867-916. Tomioka, S. 2009:<br />

‘Why-questions, presuppositions, and <strong>in</strong>tervention effects,’ JEAL 18: 253-271. Wold, D. 1996: Long distance<br />

selective b<strong>in</strong>d<strong>in</strong>g: The case of focus, <strong>in</strong> The Proceed<strong>in</strong>gs of SALT 6, 311-328.<br />

2


Jacopo Torregrossa (Verona)<br />

Encod<strong>in</strong>g contrast at PF<br />

Introduction. This contribution is based on <strong>the</strong> assumption that contrast is an autonomous<br />

<strong>in</strong>formational notion with its own semantic content (Vallduvì/Vilkuna 1998). Its ma<strong>in</strong> aim is<br />

to <strong>in</strong>vestigate how contrast is coded. On <strong>the</strong> one hand, some scholars work<strong>in</strong>g with<strong>in</strong> <strong>the</strong><br />

cartographic framework propose that contrast is licensed <strong>in</strong> an A-bar position (ContrP) with<strong>in</strong><br />

<strong>the</strong> C-doma<strong>in</strong> of <strong>the</strong> sentence (see Frascarelli/H<strong>in</strong>terhölz 2007). On <strong>the</strong> o<strong>the</strong>r hand, several<br />

works on lab speech show that <strong>the</strong> <strong>in</strong>tervention of contrast activates specific prosodic effects,<br />

whenever it appears <strong>in</strong> association with ei<strong>the</strong>r focus (Breen et al. 2010) or topic (Braun/Laid<br />

2003). The aim of <strong>the</strong> work is threefold: i) to verify <strong>the</strong> hypo<strong>the</strong>sis that contrast is encoded<br />

syntactically, by <strong>in</strong>vestigat<strong>in</strong>g if it triggers phras<strong>in</strong>g effects at <strong>the</strong> syntax/phonology <strong>in</strong>terface;<br />

ii) to understand which acoustic parameters play a role <strong>in</strong> express<strong>in</strong>g contrastive<br />

<strong>in</strong>terpretations (<strong>in</strong> both cases, <strong>the</strong> <strong>in</strong>formational notion of contrast will be studied <strong>in</strong> isolation<br />

from ei<strong>the</strong>r focus or topic, <strong>in</strong> whole compliance with <strong>the</strong> assumption about its autonomy); iii)<br />

to speculate on which model of <strong>the</strong> <strong>in</strong>teraction between <strong>the</strong> different components of language<br />

best fits with <strong>the</strong> outcome of <strong>the</strong> analysis.<br />

Theoretical framework. The <strong>in</strong>vestigation of i) is couched with<strong>in</strong> Zubizarreta’s (2010)<br />

<strong>the</strong>ory on <strong>the</strong> syntax/phonology <strong>in</strong>terface. Instead of assum<strong>in</strong>g a hierarchy of prosodic<br />

constituents which is <strong>in</strong>dependent of syntax (see Selkirk 1986), <strong>the</strong> author proposes that <strong>the</strong><br />

<strong>in</strong>ventory of prosodic categories could be reduced to boundaries which are dist<strong>in</strong>guished <strong>in</strong><br />

ter<strong>ms</strong> of strength. Crucially, <strong>the</strong> <strong>in</strong>sertion of <strong>the</strong>se boundaries is sensitive to <strong>the</strong> difference<br />

among syntactic projections. Therefore, I propose that weak prosodic boundaries are <strong>in</strong>serted<br />

to <strong>the</strong> right of constituents related to lexical projections (be<strong>in</strong>g <strong>the</strong>ir specifiers, complements<br />

or adjuncts), while strong prosodic boundaries are <strong>in</strong>serted to <strong>the</strong> right of constituents related<br />

to discourse projections (e.g., TopP, FocP). (1) shows how that weak prosodic boundaries are<br />

represented as s<strong>in</strong>gle brackets, while strong prosodic boundaries as sequences of right and left<br />

brackets.<br />

(1) (DELLA PASTA)(mangerò) stasera) per cena). [FocP DELLA PASTAi[IP pro mangerò stasera ti]]<br />

(SOME MEAT I will eat tonight for d<strong>in</strong>ner).<br />

ContrP might trigger <strong>the</strong> <strong>in</strong>sertion of a strong p-boundary, s<strong>in</strong>ce it is a discourse projection.<br />

Methods of <strong>the</strong> analysis. I set up a production experiment <strong>in</strong> which three Italian speakers<br />

were asked to read aloud some dialogues which were designed so that <strong>the</strong> same sentence<br />

appeared <strong>in</strong> five different discourse contexts. For example, <strong>the</strong> <strong>in</strong>terpretation of <strong>the</strong><br />

constituent Michelangelo <strong>in</strong> (2) varied so that it ended up to be: i) an <strong>in</strong>formation focus<br />

(+FOC); ii) a contrastive focus (+FOC,+CONTR); iii) an exhaustive focus<br />

(+FOC,+EXHAUST) – <strong>in</strong> l<strong>in</strong>e with Horvath (2010), I assume that FocP encodes exhaustivity;<br />

iv) a ‘barely’ contrastive constituent (+CONTR) followed by an <strong>in</strong>formation focus; v) nei<strong>the</strong>r<br />

focused nor contrastive (-FOC,-CONTR).<br />

(2) Ho ripassato Michelangelo con la mia amica. (I reviewed Michelangelo with my friend).<br />

Up to now, <strong>the</strong> analysis has been carried out over 70 sentences and it will be extended <strong>in</strong> future.<br />

In order to verify if contrast is encoded syntactically, I measured <strong>the</strong> duration of <strong>the</strong> poststressed<br />

str<strong>in</strong>g of <strong>the</strong> target words. The rationale beh<strong>in</strong>d this choice is that syllables preced<strong>in</strong>g<br />

strong prosodic boundaries are traditionally taken to undergo leng<strong>the</strong>n<strong>in</strong>g processes.<br />

Accord<strong>in</strong>gly, <strong>the</strong> duration of post-stressed material with<strong>in</strong> contrastive constituents might be<br />

greater than that with<strong>in</strong> non-contrastive ones, if contrast is actually encoded <strong>in</strong> a dedicated<br />

functional projection (ContrP). In order to verify if contrast triggers specific prosodic effects,<br />

I labeled <strong>the</strong> <strong>in</strong>tonation pattern of each target constituent (TOBI annotation system) and I<br />

measured both <strong>the</strong> duration and <strong>the</strong> range of its stressed syllable.<br />

Results. As for <strong>the</strong> duration of <strong>the</strong> prosodic boundaries, I found that <strong>the</strong>re is a statistically<br />

significant difference between <strong>the</strong> values relative to exhaustive foci and those relative to each<br />

of <strong>the</strong> o<strong>the</strong>r semantic <strong>in</strong>terpretations (see Figure1 as an example). This suggests that only


exhaustive foci trigger <strong>the</strong> <strong>in</strong>sertion of strong prosodic boundaries (as predicted by <strong>the</strong><br />

syntax/phonology model with<strong>in</strong> which my analysis is couched). The phonological analysis<br />

reveals that <strong>the</strong>re is no systematic correlation between <strong>the</strong> <strong>in</strong>tervention of contrast and <strong>the</strong> use<br />

of a particular type of accent (Figure2). On <strong>the</strong> contrary, <strong>the</strong> acoustical analysis shows that<br />

contrast always correlates with a greater duration and a higher pitch range of <strong>the</strong> stressed<br />

syllable with<strong>in</strong> <strong>the</strong> target constituent (Figure3).<br />

Conclusions and implications. My analysis suggests that contrast is not encoded <strong>in</strong> a<br />

dedicated functional projection (vs. Frascarelli/H<strong>in</strong>terhölz). The hypo<strong>the</strong>sis about <strong>the</strong><br />

existence of ContrP see<strong>ms</strong> to be challenged by o<strong>the</strong>r relevant data, like <strong>the</strong> fact that<br />

contrastive constituents can appear both <strong>in</strong>-situ and <strong>in</strong> <strong>the</strong> left-periphery and that <strong>the</strong>y can be<br />

freely recursive across <strong>the</strong> CP (3) and <strong>the</strong> IP, which challenges <strong>the</strong> hypo<strong>the</strong>sis of a one-to-one<br />

correspondence between syntactic projections and semantic <strong>in</strong>terpretations.<br />

(3) A: Il nonno ha lasciato l’orologio a suo fratello? B: Non lo so, ma la scacchiera a suo nipote l’ha<br />

lasciata. (A: Did <strong>the</strong> grandpa bequeath <strong>the</strong> clock to his bro<strong>the</strong>r? B: I don’t know, but it is to his<br />

grandson that he bequea<strong>the</strong>d <strong>the</strong> chessboard).<br />

More <strong>in</strong> general, <strong>the</strong> analysis suggests that on <strong>the</strong> one hand <strong>the</strong>re is no phonological correlate<br />

of contrast and on <strong>the</strong> o<strong>the</strong>r hand <strong>the</strong>re are several acoustic parameters that express contrastive<br />

<strong>in</strong>terpretations. If this l<strong>in</strong>e of analysis is correct, <strong>the</strong> <strong>in</strong>vestigation on <strong>the</strong> l<strong>in</strong>guistic encod<strong>in</strong>g of<br />

contrast provides empirical and conceptual evidence for <strong>the</strong> necessity to dist<strong>in</strong>guish<br />

phonology and phonetics as two different modules of grammar. Accord<strong>in</strong>g to <strong>the</strong><br />

syntax/phonology model assumed <strong>in</strong> this paper, phonology has a more direct connection with<br />

syntax than it is traditionally assumed <strong>in</strong> prosodic phonology. Assum<strong>in</strong>g C<strong>in</strong>que’s (1993)<br />

Stress Rule, both stress assignment and prosodic boundaries <strong>in</strong>sertion turn out to apply<br />

directly to <strong>the</strong> syntactic tree. On <strong>the</strong> contrary, phonetics has been shown to be syntax-bl<strong>in</strong>d<br />

and sensitive to <strong>in</strong>terpretive dist<strong>in</strong>ctions. This last claim has significant implications for <strong>the</strong><br />

general model of <strong>the</strong> architecture of grammar.<br />

Figure1: Duration (<strong>ms</strong>.) of post-stressed<br />

material <strong>in</strong> contrastive foci and exhaustive<br />

foci.<br />

200<br />

150<br />

100<br />

50<br />

40<br />

30<br />

20<br />

10<br />

50<br />

0<br />

(+CONTR,+F<br />

OC)<br />

(+FOC,+EXH<br />

AUST)<br />

(+CONTR)<br />

(-CONTR,-<br />

FOC)<br />

0<br />

Figure3: Values of <strong>the</strong> range (Hz) of <strong>the</strong><br />

accented syllables <strong>in</strong> contrastive and noncontrastive<br />

contexts.<br />

Figure2: Frequency of <strong>in</strong>formational functions<br />

expressed by each type of accent.<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0<br />

(+CONTR)<br />

(-FOC) (-CONTR)<br />

(+FOC)<br />

(+EXHAUST.)<br />

(+FOC)<br />

Essential bibliography: Breen M./E. Federenko/M.<br />

Wagner/E. Gibson, 2010, Acoustic correlates of <strong>in</strong>formation<br />

structure, Language and Cognitive Processes 25(7/9), 1044-<br />

1098; C<strong>in</strong>que G., 1993, A null <strong>the</strong>ory of phrase and compound<br />

stress, LI 24(2), 239-282; Frascarelli, M./R. H<strong>in</strong>terhölzl, 2007,<br />

Types of topics <strong>in</strong> German and Italian, <strong>in</strong> S. W<strong>in</strong>kler/K.<br />

Schwabe (eds.), On Information Structure Mean<strong>in</strong>g and Form,<br />

A<strong>ms</strong>terdam/Philadelphia: John Benjam<strong>in</strong>s, 87-116. Horvath J.,<br />

2010, "Discourse-Features", Syntactic Displacement and <strong>the</strong><br />

Status of Contrast, L<strong>in</strong>gua 120, 1346-1369; Selkirk E., 1986,<br />

On derived doma<strong>in</strong>s <strong>in</strong> sentence phonology, Phonology<br />

Yearbook3, 371-405; Zubizarreta M.L., 2010, The Syntax and<br />

Prosody of Focus: The Bantu-Italian Connection, Iberia 2(1),<br />

131-168.


Coppe van Urk (MIT)<br />

!<br />

" " "<br />

" #<br />

"<br />

$ %&'<br />

%&' ( $ ) * + * #*<br />

* , - "* ./ * * 0 * * / ,1<br />

2 " 2 % $3 4 5 6776869:'<br />

; % 4 5<br />

6776' / < % 677= 67&7'<br />

- 4 % 67&7' "<br />

" *<br />

; %6' %> 4<br />

"<br />

&?:&3 @ 6777'<br />

%6' " A A<br />

* %B'<br />

%B' C! A 4 A<br />

" " ; 8<br />

%9' ! C!<br />

> " "<br />

! " 4"<br />

" % 67&7'<br />

%D' C! A A<br />

%D' %@ 6777' " 4"<br />

" 8<br />

" 4"<br />

" " "<br />

! "#<br />

! "<br />

" ; 4" "<br />

;<br />

, % %= ''8<br />

%=' CE 2 A E 2<br />

" * "<br />

4 " * 8<br />

%:' CE 2 A<br />

,<br />

- ; $ 8<br />

%F' C, " A 4<br />

2 " A # 2<br />

"<br />

8


4 %677:' "<br />

< %< ' % &?=:' , ;<br />

4 ; 8<br />

%?' C< A A " . ! ( ! A " +<br />

< 8 ;<br />

; %G 5 E 4 677B3<br />

/ 677:' ><br />

* % " #$<br />

" C% '<br />

" # 4 ;<br />

%, H . I 677D'8<br />

%&7' CJ A


Kai von F<strong>in</strong>tel (MIT), Danny Fox (Hebrew U./MIT) & Sab<strong>in</strong>e Iatridou (MIT)<br />

Def<strong>in</strong>iteness as Maximal Informativeness<br />

The unified L<strong>in</strong>k-style semantics for <strong>the</strong> def<strong>in</strong>ite article relies on an <strong>in</strong>herent order<strong>in</strong>g of<br />

objects, a part-whole order<strong>in</strong>g <strong>in</strong> <strong>the</strong> case of <strong>in</strong>dividuals.<br />

(1) �<strong>the</strong>�(φ〈α,t〉) is def<strong>in</strong>ed only if <strong>the</strong>re is a maximal object x (based on <strong>the</strong> <strong>in</strong>herent<br />

order<strong>in</strong>g of elements <strong>in</strong> Dα), such that φ(x). When def<strong>in</strong>ed, �<strong>the</strong>�(φ) refers to<br />

that maximal object.<br />

The φ refers to <strong>the</strong> maximal φ-object: <strong>the</strong> unique φ <strong>in</strong> <strong>the</strong> s<strong>in</strong>gular case, <strong>the</strong> maximal<br />

plurality conta<strong>in</strong><strong>in</strong>g all φ-<strong>in</strong>dividuals <strong>in</strong> <strong>the</strong> plural case, and <strong>the</strong> maximal collection of<br />

φ-stuff <strong>in</strong> <strong>the</strong> mass case.<br />

We propose an alternative account: maximality is def<strong>in</strong>ed with respect to a different<br />

order<strong>in</strong>g, that of “relative <strong>in</strong>formativity” i.e. asymmetrical entailment. Strength relates<br />

propositions, but, when given a property, can be def<strong>in</strong>ed derivatively for <strong>in</strong>dividuals.<br />

(2) �<strong>the</strong>�(φ〈α,st〉) is def<strong>in</strong>ed only if <strong>the</strong>re is an object x such that φ(x) is <strong>the</strong> maximally<br />

<strong>in</strong>formative proposition among <strong>the</strong> true propositions of <strong>the</strong> form φ(. . .)<br />

(henceforth <strong>the</strong> most <strong>in</strong>formative object <strong>in</strong> φ). When def<strong>in</strong>ed, �<strong>the</strong>�(φ) refers to<br />

that most <strong>in</strong>formative φ-object.<br />

The most <strong>in</strong>formative <strong>in</strong>dividual with respect to property φ is that <strong>in</strong>dividual from<br />

whose be<strong>in</strong>g φ we can deduce <strong>the</strong> φ-ness of all o<strong>the</strong>r φ-<strong>in</strong>dividual.<br />

For <strong>the</strong> basic cases discussed <strong>in</strong> <strong>the</strong> literature, L<strong>in</strong>k’s def<strong>in</strong>ition and our alternative<br />

co<strong>in</strong>cide. Assume that John, Bill, and Sam are <strong>the</strong> only boys. Then <strong>the</strong> boys will denote<br />

<strong>the</strong> maximal plurality made up of those three <strong>in</strong>dividuals, both on L<strong>in</strong>k’s account (s<strong>in</strong>ce<br />

that is <strong>the</strong> maximal boy-plurality) and on ours: <strong>the</strong> plurality made up of those three<br />

<strong>in</strong>dividuals is exactly that plurality from whose be<strong>in</strong>g all boys we can deduce <strong>the</strong> boyness<br />

of any of its components. Any smaller plurality would be less <strong>in</strong>formative (from<br />

John and Bill be<strong>in</strong>g boys we can’t deduce that Sam is a boy).<br />

In fact, both L<strong>in</strong>k’s proposal and ours make <strong>the</strong> same prediction for any property<br />

that is upward monotone <strong>in</strong> a technical sense. Properties of degrees such as λd. John<br />

is d tall or λd. John has n many children are upward monotone (say<strong>in</strong>g that John has<br />

4 children is more <strong>in</strong>formative than say<strong>in</strong>g that he has 3). Hence under both accounts<br />

def<strong>in</strong>ite descriptions such as John’s height or <strong>the</strong> number of children that John has will<br />

refer (<strong>in</strong> any world w) to <strong>the</strong> maximal object <strong>in</strong> <strong>the</strong> extension of <strong>the</strong> property <strong>in</strong> w.<br />

Once we look at properties that are not upward monotone wrt <strong>in</strong>formativity, L<strong>in</strong>k’s<br />

proposal and ours make different predictions. First consider properties that are downward<br />

monotone. These are cases where <strong>the</strong> smallest amount/object is more <strong>in</strong>formative<br />

— unlike <strong>the</strong> earlier examples, which characterized cases where <strong>the</strong> largest object/amount<br />

is <strong>the</strong> most <strong>in</strong>formative. Here is such a case:<br />

(3) I have <strong>the</strong> amount of flour sufficient to bake a cake.<br />

Propositions of <strong>the</strong> form d-much flour is sufficient to bake a cake become more <strong>in</strong>formative<br />

<strong>the</strong> smaller d is. We thus correctly predict that <strong>the</strong> def<strong>in</strong>ite description <strong>in</strong> (3) should<br />

refer to <strong>the</strong> m<strong>in</strong>imum amount of flour that would yield a true proposition, i.e. to <strong>the</strong><br />

m<strong>in</strong>imum amount that would suffice for cake bak<strong>in</strong>g. On <strong>the</strong> o<strong>the</strong>r hand, accord<strong>in</strong>g to<br />

1


L<strong>in</strong>k’s account, <strong>the</strong> def<strong>in</strong>ite description <strong>in</strong> this sentence should be undef<strong>in</strong>ed, s<strong>in</strong>ce <strong>the</strong>re<br />

can be no maximal amount of flour that is sufficient to bake a cake; as noted by Beck<br />

and Rullmann (1999), <strong>in</strong> a slightly different context, if an amount of flour, f , suffices to<br />

bake a cake, so does any amount larger than f .<br />

So we see that a def<strong>in</strong>ite description of <strong>the</strong> form <strong>the</strong> φ alternates between referr<strong>in</strong>g to<br />

<strong>the</strong> m<strong>in</strong>imal or <strong>the</strong> maximal <strong>in</strong>dividual <strong>in</strong> <strong>the</strong> extension depend<strong>in</strong>g on <strong>the</strong> monotonicity<br />

of <strong>the</strong> property φ. We get a maximality effect when φ is upward monotone and a<br />

m<strong>in</strong>imality effect when φ is downward monotone as <strong>in</strong> (3).<br />

Once <strong>the</strong> pr<strong>in</strong>ciple is clear, it is easy to construct fur<strong>the</strong>r cases show<strong>in</strong>g a m<strong>in</strong>imality<br />

effect: consider, for example, <strong>the</strong> number of Greek soldiers who toge<strong>the</strong>r can destroy<br />

<strong>the</strong> Trojan army. For L<strong>in</strong>k, this would result <strong>in</strong> a presupposition failure, s<strong>in</strong>ce <strong>the</strong>re is<br />

no maximal number of Greek soldiers that toge<strong>the</strong>r can destroy <strong>the</strong> Trojan army (<strong>the</strong><br />

more <strong>the</strong> merrier). For us, on <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> description will pick out <strong>the</strong> m<strong>in</strong>imal<br />

number of soldiers that toge<strong>the</strong>r can destroy <strong>the</strong> Trojan army, because that is <strong>the</strong> most<br />

<strong>in</strong>formative such number (once we know that number we can deduce that all larger<br />

numbers would also do).<br />

We conclude that L<strong>in</strong>k’s <strong>the</strong>ory got <strong>the</strong> right results only because <strong>the</strong> focus was<br />

limited to upward monotone properties.<br />

F<strong>in</strong>ally, <strong>the</strong>re are properties that are non-monotone. For <strong>the</strong>se, we predict a presupposition<br />

failure when <strong>the</strong>re is no unique <strong>in</strong>dividual <strong>in</strong> <strong>the</strong> extension of <strong>the</strong> property.<br />

Consider <strong>the</strong> follow<strong>in</strong>g. You are try<strong>in</strong>g to fit books (x, y, z, w, v . . . ) on shelves of various<br />

size (a, b, c . . . ). Suppose that book x toge<strong>the</strong>r with book y fit perfectly on shelf a, and<br />

book x, y, and z toge<strong>the</strong>r fit perfectly on shelf b. Suppose also that no o<strong>the</strong>r comb<strong>in</strong>ation<br />

of books fits perfectly on a shelf. Under L<strong>in</strong>k’s proposal, <strong>the</strong> def<strong>in</strong>ite description <strong>in</strong><br />

#Pass me <strong>the</strong> books that toge<strong>the</strong>r fit perfectly on a shelf should be acceptable: it should<br />

refer to <strong>the</strong> larger collection of books that fit perfectly on a shelf, namely, x + y + z. Our<br />

analysis, <strong>in</strong> contrast, correctly predicts that <strong>the</strong> description suffers from presupposition<br />

failure.<br />

An alternation similar to <strong>the</strong> one between m<strong>in</strong>imality and maximality shows up <strong>in</strong><br />

<strong>the</strong> doma<strong>in</strong> of times as well. In (4) <strong>the</strong> def<strong>in</strong>ite description refers to <strong>the</strong> latest time, t,<br />

such that Bill lived <strong>in</strong> Boston until t, while <strong>in</strong> (5) <strong>the</strong> def<strong>in</strong>ite refers to <strong>the</strong> earliest time, t,<br />

such that Bill has lived <strong>in</strong> Boston s<strong>in</strong>ce t.<br />

(4) January 5th 1999 is <strong>the</strong> date until which Bill had lived <strong>in</strong> Boston.<br />

(5) January 5th 1999 is <strong>the</strong> date s<strong>in</strong>ce which Bill has lived <strong>in</strong> Paris.<br />

We expect this alternation and do not need to stipulate it <strong>in</strong> <strong>the</strong> semantics of <strong>the</strong><br />

temporal operators until or s<strong>in</strong>ce. In both cases, <strong>the</strong> def<strong>in</strong>ite description refers to <strong>the</strong><br />

most <strong>in</strong>formative time that satisfies <strong>the</strong> relevant property (λt. Bill had lived <strong>in</strong> Boston<br />

until t, <strong>in</strong> (6) and λt. Bill has lived <strong>in</strong> Paris s<strong>in</strong>ce t, <strong>in</strong> (7)). The difference, once aga<strong>in</strong>, has<br />

to do with <strong>the</strong> monotonicity of <strong>the</strong> property.<br />

2


Mat<strong>the</strong>w Wagers (UC Santa Cruz), Manuel Borja (Inetnon Amot yan Kutturan Natibu) & Sandra Chung (UC Santa Cruz)<br />

WH Agreement and <strong>the</strong> Tim<strong>in</strong>g of Unbounded Dependency Formation:<br />

a Chamorro perspective on <strong>in</strong>crementality and accuracy <strong>in</strong> language comprehension<br />

SUMMARY. The comprehension of unbounded 1iller-­‐gap dependencies proceeds actively<br />

and predictively: comprehenders posit gaps <strong>in</strong> particular syntactic positions before direct<br />

evidence that a constituent does not already occupy that position [1]. Evidence for active<br />

comprehension comes from a number of psychol<strong>in</strong>guistic paradig<strong>ms</strong> and varied (majority)<br />

languages [2]. An important question concerns <strong>the</strong> sources and timecourse availability of<br />

<strong>in</strong>formation that shapes a comprehender’s expectations about <strong>the</strong> unbounded dependency.<br />

Island constra<strong>in</strong>ts are known to immediately restrict comprehenders’ hypo<strong>the</strong>ses: gaps are<br />

not predicted <strong>in</strong> doma<strong>in</strong>s where <strong>the</strong>y could not be licensed [3]. Lexically-­‐projected<br />

<strong>in</strong>formation, like subcategorization and argument structure, can also provide clues to <strong>the</strong><br />

likely role of <strong>the</strong> extracted phrase. There is less empirical clarity about how those clues<br />

affect <strong>the</strong> comprehender’s predictions [4-­‐6], and, <strong>in</strong> particular, what happens when<br />

expectations are formed before lexical access of <strong>the</strong> verb [7].<br />

Here we <strong>in</strong>vestigate real-­‐time sentence comprehension <strong>in</strong> Chamorro, whose<br />

grammar provides a morphological cue to <strong>the</strong> extraction site via its system of WH<br />

Agreement [8]. We report two major 1<strong>in</strong>d<strong>in</strong>gs: 1irst, displaced objects are only actively<br />

<strong>in</strong>terpreted semantically when <strong>the</strong> verb is (optionally) <strong>in</strong>1lected for WH Agreement.<br />

Secondly, when <strong>the</strong> verb occurs <strong>in</strong> its un<strong>in</strong>1lected form, comprehenders none<strong>the</strong>less have<br />

dif1iculty re-­‐analyz<strong>in</strong>g an extraction which term<strong>in</strong>ates <strong>in</strong> a non-­‐object position -­‐ <strong>in</strong> our case,<br />

as a possessor. We argue this dif1iculty is a Chamorro analogue of <strong>the</strong> 1illed gap effect [9]. It<br />

suggests that comprehenders do actively posit a syntactic gap <strong>in</strong> <strong>the</strong> un<strong>in</strong>1lected for<strong>ms</strong> even<br />

if <strong>the</strong>y don’t synchronously <strong>in</strong>terpret it, a conjecture recently made for English data [10].<br />

DESIGN. Chamorro is a verb-­‐<strong>in</strong>itial Austronesian language with some 40-­‐50,000 speakers,<br />

ma<strong>in</strong>ly <strong>in</strong> <strong>the</strong> Mariana Islands [8]. Of particular <strong>in</strong>terest is its system of WH Agreement: a<br />

transitive verb whose argument has been A-­‐bar extracted can be specially <strong>in</strong>1lected to<br />

re1lect <strong>the</strong> grammatical role of <strong>the</strong> extractee (as well as verbs along <strong>the</strong> extraction path) [8].<br />

As <strong>in</strong> example (1a), a subject extraction from a realis, transitive verb obligatorily triggers<br />

<strong>in</strong>1ixation of -­‐um-­‐ to <strong>the</strong> verb (cf. (1b), <strong>the</strong> unextracted version). Likewise, extractions of<br />

obliques obligatorily trigger a special verb form.<br />

(1)! (a)!Håyi na såstri l umåksi ! i magågu-mu?<br />

! ! who !L sea<strong>ms</strong>tress WH[nom].sew D clo<strong>the</strong>s-AGR?<br />

! ! “Which sea<strong>ms</strong>tress sewed your clo<strong>the</strong>s?”<br />

! (b)! Ha-låksi i såstri i magågu-mu<br />

! ! AGR-sew D sea<strong>ms</strong>tress D clo<strong>the</strong>s-AGR<br />

“The sea<strong>ms</strong>tress sewed your clo<strong>the</strong>s”<br />

Crucially, however, WH <strong>in</strong>1lection for object extractions is optional. Thus <strong>in</strong> (2a), <strong>the</strong> verb<br />

penta (‘pa<strong>in</strong>t’) is realized with <strong>the</strong> <strong>in</strong>1ix -­‐<strong>in</strong>-­‐ and a 2nd Person agreement suf1ix -mu. But <strong>in</strong><br />

(2b), <strong>the</strong> agreement, un-, is <strong>the</strong> same as found <strong>in</strong> clauses with no extractions.<br />

(2)! (a)!Håfa na tråk p <strong>in</strong>entåm-mu! ta’lu un biåhi?<br />

! ! what!L truck WH[obj].pa<strong>in</strong>t-AGR aga<strong>in</strong> a time<br />

! (b)! Håfa na tråk un penta ta’lu un biåhi?<br />

! ! what!L truck AGR-pa<strong>in</strong>t aga<strong>in</strong> a time<br />

! ! BOTH: “Which truck did you pa<strong>in</strong>t over aga<strong>in</strong>?


The optionality of WH <strong>in</strong>1lection for object extractions for<strong>ms</strong> <strong>the</strong> basis for two real-­‐<br />

time sentence comprehension studies that might elucidate how predictions with different<br />

sources <strong>in</strong>teract with <strong>the</strong> time-­‐courses of active comprehension. The WH <strong>in</strong>1lection<br />

morpheme provides a clear cue <strong>in</strong> <strong>the</strong> <strong>in</strong>put for a particular analysis of <strong>the</strong> extractee,<br />

whereas an un<strong>in</strong>1lected form supplies a more <strong>in</strong>direct cue: <strong>the</strong> absence of obligatory subject<br />

or oblique extraction morphology is compatible with an object extraction. Moreover a non-­‐<br />

WH-­‐<strong>in</strong>1lected form is also used when <strong>the</strong> possessor of <strong>the</strong> object phrase is extracted (3).<br />

(3)! ! Håfa na tråk un penta ta’lu sanme’nåñ-ña?<br />

! ! what L truck! AGR pa<strong>in</strong>t aga<strong>in</strong> truck-AGR<br />

! ! “Which truck did you pa<strong>in</strong>t its front aga<strong>in</strong>?”<br />

Previous research suggests that comprehenders can immediately discard an object analysis<br />

if an extractee’s semantic features are <strong>in</strong>compatible with <strong>the</strong> verb and <strong>the</strong> verb’s argument<br />

structure signals o<strong>the</strong>r gap sites (e.g. object control verbs like persuade)[4]. Thus a non-­‐<br />

WH-­‐<strong>in</strong>1lected verb could forestall active comprehension ei<strong>the</strong>r by (a) signall<strong>in</strong>g <strong>the</strong> object<br />

analysis more slowly; or (b) clue<strong>in</strong>g <strong>the</strong> possibility of possessor extraction.<br />

To test whe<strong>the</strong>r active comprehension was dim<strong>in</strong>ished for un<strong>in</strong>1lected for<strong>ms</strong>, we<br />

crossed <strong>the</strong> semantic plausibility of an extractee as verb’s object [±PLAUS.OBJ] with <strong>the</strong><br />

presence of WH <strong>in</strong>1lection [±WH.INFLECT]: (4)-­‐(5). We probed for an effect of semantic<br />

anomaly on <strong>the</strong> verb <strong>in</strong> two paradig<strong>ms</strong>: <strong>in</strong> Self-paced Listen<strong>in</strong>g, 40 native Chamorro<br />

speakers listened <strong>in</strong> an auditory mov<strong>in</strong>g w<strong>in</strong>dow paradigm and judged whe<strong>the</strong>r sentence<br />

made sense. In Passive Listen<strong>in</strong>g+Eye track<strong>in</strong>g, 72 native speakers listened to <strong>the</strong><br />

sentences passively and made <strong>the</strong> same judgement. Large ‘yes’/’no’ buttons were depicted<br />

on screen; participants’ gaze was recorded via a laptop webcam and later coded bl<strong>in</strong>d by a<br />

team of RAs unaware of <strong>the</strong> experimental manipulation. In both experiments <strong>the</strong>re were 12<br />

target ite<strong>ms</strong> and 28 ite<strong>ms</strong> counterbalanc<strong>in</strong>g for o<strong>the</strong>r factors. Experiments took place on<br />

Saipan, T<strong>in</strong>ian & Luta and all <strong>in</strong>structions, audio and text were <strong>in</strong> Chamorro.<br />

(4) (a)! [+PLAUS.OBJ, +WH.INFLECT]:! ! ⇢(2a)<br />

! (b)! [-PLAUS.OBJ, +WH.INFLECT]:<br />

! ! Håyi na ma’estru p<strong>in</strong>entåm-mu! ta’lu un biåhi?<br />

! ! who! L teacher WH[obj].pa<strong>in</strong>t-AGR aga<strong>in</strong> a time<br />

! ! “Which teacher did you pa<strong>in</strong>t over aga<strong>in</strong>?”<br />

(5)! (a)! [+PLAUS.OBJ, -WH.INFLECT]:! ⇢(3)<br />

! (b)! [-PLAUS.OBJ, -WH.INFLECT]:<br />

! ! Håyi na ma’estru un penta ta’lu tråk-ña?<br />

! ! what L teacher AGR pa<strong>in</strong>t aga<strong>in</strong> truck-AGR<br />

! ! “Which teacher did you pa<strong>in</strong>t his truck aga<strong>in</strong>?”<br />

For WH <strong>in</strong>1lected for<strong>ms</strong>, we 1<strong>in</strong>d clear evidence of active <strong>in</strong>terpretation of <strong>the</strong><br />

extractee, as <strong>in</strong>dicated by <strong>in</strong>creased listen<strong>in</strong>g times on <strong>the</strong> verb or <strong>in</strong>creased likelihood of<br />

look<strong>in</strong>g to <strong>the</strong> ‘no’ response button when -PLAUS.OBJ. For un<strong>in</strong>1lected for<strong>ms</strong>, nei<strong>the</strong>r<br />

those effects are observed. However participants did systematically disprefer <strong>the</strong> possessor<br />

extractions if <strong>the</strong> extractee were -PLAUS.OBJ. We <strong>in</strong>terpret this as evidence that <strong>the</strong>y<br />

do beg<strong>in</strong> syntactic construction of <strong>the</strong> object analysis but have dif1iculty reanalyz<strong>in</strong>g<br />

[1] Frazier, L., & Flores D’Arcais, G. B. (1989). [2] Aoshima, S., Phillips, C., and We<strong>in</strong>berg, A. S. (2004).<br />

[3] Phillips, C. (2006). [4] Boland, J.E., et al. (1995) [5] Picker<strong>in</strong>g, M. & Traxler, M. (2003). [6] Staub<br />

2007. [7] Omaki et al. (2011) [8] Chung (1998) [9] Stowe (1986). [10] Wagers & Phillips (2009) [11]<br />

Ferreira et al. (1996).


Conjunct
Morphology
Marks
Property
Self­Ascription
<br />


<br />

Conjunct/Disjunct
(CJ/DJ)
languages
lack
person
mark<strong>in</strong>g
on
verbs,
employ<strong>in</strong>g
<strong>in</strong>stead
a
<br />

special
 CJ
 verb
 form
 when
 (i)
 <strong>the</strong>
 subject
 of
 a
 declarative
 is
 1st
 person
 (1),
 (ii)
 <strong>the</strong>
<br />

subject
of
a
question
is
2nd
person
(5),
or
(iii)
<strong>the</strong>
verb
appears
<strong>in</strong>
a
de
se
attitude
report
<br />

(7).

The
DJ
form
appears
elsewhere.

Exs.
from
Kathmandu
Newar
(Hargreaves
2005):
<br />


<br />

1. jī:
 a:pwa
 twan‐ā.
<br />


 1.ERG
 much
 dr<strong>in</strong>k‐PST.CJ
<br />


 ‘I
drank
a
lot/too
much.’
<br />


<br />

2. chā
 a:pwa
 twan‐a.
<br />


 2.ERG
 much
 dr<strong>in</strong>k‐PST.DJ
<br />


 ‘You
drank
a
lot/too
much.’
<br />


<br />

3. wā:
 a:pwa
 twan‐a.
<br />


 
 3.ERG
 much
 dr<strong>in</strong>k‐PST.DJ
<br />


 ‘S/he
drank
a
lot/too
much.’
<br />


<br />

Stephen Wechsler (Texas)<br />

4. jī:
 a:pwa
 twan‐a
 lā?
<br />


 
 1.ERG
 much
 dr<strong>in</strong>k‐PST.DJ
Q
<br />


 ‘Did
I
dr<strong>in</strong>k
a
lot/too
much?’
<br />


<br />

5. chā
 a:pwa
 twan‐ā
 lā?
<br />


 2.ERG
 much
 dr<strong>in</strong>k‐PST.CJ
Q
<br />


 ‘Did
you
dr<strong>in</strong>k
a
lot/too
much?’
<br />


<br />

6. wā:
 a:pwa
 twan‐a
 lā?
<br />


 
 3.ERG
 much
 dr<strong>in</strong>k‐PST.DJ
Q
<br />


 ‘Did
s/he
dr<strong>in</strong>k
a
lot/too
much?’
<br />

7. Syām‐ā
 a:pwa
 twan‐ā
 hā.
<br />


 Syam‐ERG
 much
 dr<strong>in</strong>k‐PST.CJ
 EVD
<br />


 ‘Syami
said
that
hei
drank
too
much.’
<br />


<br />

CJ/DJ
syste<strong>ms</strong>
have
been
observed
<strong>in</strong>
S<strong>in</strong>o‐Tibetan,
e.g.
Newar
(Hale
1980;
Hargreaves
<br />

2005);
 Nakh‐Daghestanian,
 e.g.
Akhvakh
 (Creissels
 2008)
 and
Mehwb
 Dargwa
(Bickel
<br />

2008);
 Tsafiki
 (Barbacoan;
 Dick<strong>in</strong>son
 2000);
 Trans
 New
 Gu<strong>in</strong>ea,
 e.g.
 Oksapm<strong>in</strong>
<br />

(Loughnane
 2009),
 Duna
 and
 Kaluli
 (San
 Roque
 2011);
 Guambiano
 (Norcliffe
 2011);
<br />

Cha’palaa
(Floyd
2011).


<br />


 Conjunct
mark<strong>in</strong>g
forces
a
hard
look
at
a
key
question
of
<strong>the</strong>
GLOW
colloquium:
<br />

<strong>the</strong>
pragmatics‐grammar
<strong>in</strong>terface.

We
put
forth
<strong>the</strong>
hypo<strong>the</strong>sis
that
conjunct
(CJ)
verb
<br />

morphology
 marks
 a
 verb
 phrase
 as
 denot<strong>in</strong>g
 a
 doxastic
 self­ascribed
 property.

<br />

Follow<strong>in</strong>g
recent
work
by
Hans
Kamp,
reference
de
se
is
modeled
under
<strong>the</strong>
assumption
<br />

that
‘self‐attribut<strong>in</strong>g
thoughts
take
<strong>the</strong>
form
of
predications
of
a
special
s<strong>in</strong>gular
term
<br />

that
‘represents
<strong>the</strong>
self’
to
<strong>the</strong>
agent
as
her
self.

We
use
<strong>the</strong>
symbol
i
to
play
this
part.’
<br />

(Kamp
2011,
7).

Crucially,
we
posit
a
special
type
for
i.

A
conjunct‐VP
is
a
function
from
<br />

i‐type
 ter<strong>ms</strong>
 to
 beliefs.
 We
 dist<strong>in</strong>guish
 <strong>in</strong>dividual
 <strong>in</strong>terlocutors’
 discourse
<br />

representations,
as
well
as
participant‐neutral
ones.
<br />


 This
hypo<strong>the</strong>sis
applies
to
uses
of
CJ
as
follows:

(i)
A
declarative
utterance
is
a
<br />

declaration
of
speaker
belief.

In
1st
person
subject
CJ
declaratives
<strong>the</strong>
speaker
declares
<br />

<strong>the</strong>
belief
that
results
from
self­ascrib<strong>in</strong>g
<strong>the</strong>
VP­denoted
property:
e.g.
<strong>in</strong>
1
that
belief
is
<br />

BEL(drank­too­much(i)).

(ii)
A
question
is
an
<strong>in</strong>ducement
to
<strong>the</strong>
addressee
to
declare
a
<br />

belief
iff
she
holds
it.

In
2nd
person
subject
CJ
questions
<strong>the</strong>
addressee
is
<strong>in</strong>duced
to
<br />

declare
<strong>the</strong>
belief
that
results
from
self‐ascrib<strong>in</strong>g
<strong>the</strong>
VP‐denoted
property
iff
she
holds
<br />

that
belief.

(iii)
Reported
speech
us<strong>in</strong>g
CJ
reports
someone’s
self‐ascription.



<br />


 Illustrat<strong>in</strong>g
 with
 1,4,
 7,
 [a:pwa
 twanā]VP
 translates
 roughly
 as
 BEL(λi.drank­toomuch(i)).
<br />

(More
 precisely,
 an
 attitude
 is
 modeled
 as
 a
 set
 with
 two
 elements,
 <strong>in</strong>
 this
<br />

case:
 an
 anchored
 entity
 representation
 designat<strong>in</strong>g
 i
 as
 speaker;
 and
 a
 pair<strong>in</strong>g
 of
 an
<br />

attitud<strong>in</strong>al
 mode,
 here
 BEL(ief),
 with
 a
 DRS
 whose
 universe
 conta<strong>in</strong>s
 a
 propositional
<br />

discourse
referent
s
and
a
condition
s:
drank­too­much’(i).)

In
1
<strong>the</strong>
speaker
declares



<strong>the</strong>
 belief
 result<strong>in</strong>g
 from
 self‐ascrib<strong>in</strong>g
 that
 property,
 namely
 β
 =
 BEL(drank­toomuch(i)).
<br />


 In
 4
 <strong>the</strong>
 addressee
 is
 <strong>in</strong>duced
 to
 declare
 β
 iff
 she
 hold
 it.
 
 In
 7
 Syam’s
<br />

declaration
of
β
is
reported.


<br />


 This
 hypo<strong>the</strong>sis
 leads
 to
 <strong>the</strong>
 follow<strong>in</strong>g
 explanation
 for
 why
 CJ/DJ
 languages
<br />

generally
lack
normal
person
mark<strong>in</strong>g
on
<strong>the</strong>
verb.

By
hypo<strong>the</strong>sis
CJ
is
for
property
self‐<br />

ascription.
 
In
 person‐mark<strong>in</strong>g,
 a
 subject
 pronoun
 is
 <strong>in</strong>corporated
 <strong>in</strong>to
 <strong>the</strong>
 verb.
 
 But
<br />

<strong>the</strong>n
it
saturates
<strong>the</strong>
subject
role,
produc<strong>in</strong>g
a
proposition,
so
<strong>the</strong>
property
is
no
longer
<br />

available
 and
 <strong>the</strong>
 form
 of
 <strong>the</strong>
 verb
 cannot
 specify
 <strong>the</strong>
 mode
 of
 ascription
 as
 self‐<br />

ascription.


<br />



 This
 also
 offers
 an
 approach
 to
 two
 cross‐l<strong>in</strong>guistic
 tendencies
 that
 have
 been
<br />

observed.
 
 (i)
 CJ
 is
 favored
 for
 properties
 for
 which
 self‐ascription
 is
 <strong>the</strong>
 most
 likely.

<br />

(Example:
Newar
CJ
is
restricted
to
verbs
of
<strong>in</strong>tentional
action
(Hargreaves
2005).)
(ii)
<br />

CJ
 is
 favored
 where
 non‐self‐ascription
 would
 violate
 social
 proscriptions
 aga<strong>in</strong>st
<br />

presum<strong>in</strong>g
 to
 know
 o<strong>the</strong>r
 people’s
 thoughts
 better
 than
 <strong>the</strong>y
 know
 <strong>the</strong><strong>ms</strong>elves.
 
 The
<br />

particular
 restrictions
 vary
 cross‐l<strong>in</strong>guistically
 and
 can
 be
 grammaticalized.
 
 E.g.
 <strong>the</strong>
<br />

Newar
 restriction
 just
 noted
 applies
 even
 <strong>in</strong>
 a
 third
 person
 de
 se
 report
 (Hargreaves
<br />

2005,
ex.
64).


<br />


 As
far
as
we
know
this
is
<strong>the</strong>
first
formal
semantic
analysis
of
conjunct/disjunct
<br />

syste<strong>ms</strong>
to
be
proposed.

It
will
be
compared
to
previous
<strong>in</strong>formal
analyses
that
treat
CJ
<br />

mark<strong>in</strong>g
 as
 specify<strong>in</strong>g
 that
 <strong>the</strong>
 subject
 of
 <strong>the</strong>
 clause
 is
 <strong>the</strong>
 epistemic
 authority
 or
<br />

<strong>in</strong>formant:
 speaker
 of
 a
 declarative,
 addressee
 of
 a
 question.
 
 Bickel
 2008
 treats
 such
<br />

mark<strong>in</strong>g
 as
 <strong>in</strong>dicat<strong>in</strong>g
 epistemic
 scope
 on
 an
 <strong>in</strong>dividual
 argument
 and
 reasons
 that
<br />

‘def<strong>in</strong>ite
 knowledge
 is
 possible
 only
 if
 <strong>the</strong>
 <strong>in</strong>formant
 is
 <strong>in</strong>
 fact
 identical
 with
 <strong>the</strong>
<br />

argument.’
(Bickel
2008,
6).

But
this
assumption
of
‘privileged
access’
is
controversial.

<br />

We
take
a
slightly
different
tack,
tak<strong>in</strong>g
de
se
attribution
as
<strong>the</strong>
basic
significance
of
CJ
<br />

and
o<strong>the</strong>r
effects
as
derivative.


<br />


 The
account
is
fairly
elegant,
but
requires
agent‐relative
DRS’s.


These
have
been
<br />

<strong>in</strong>dependently
 motivated
 by
 speaker‐addressee
 knowledge
 asymmetries
 <strong>in</strong>
 epistemic
<br />

specificity
(H.
Kamp
and
Bende‐Farkas
2006)
and
person‐number
paradigm
universals
<br />

(Wechsler
2010).



<br />


<br />

Bickel,
 Balthasar.
 2008.
 VerbAgreement
 and
 EpistemicMark<strong>in</strong>g:
 a
 Typological
 Journey
<br />

from
 <strong>the</strong>
 Himalayas
 to
 <strong>the</strong>
 Caucasus.
 In
 Festschrift
 für
 Roland
 Bielmeier,
 ed.
<br />

Chomolangma,
 Demawend,
 and
 Kasbek,
 1–14.
 •
 Creissels,
 Denis.
 2008.
 “Person
<br />

variations
 <strong>in</strong>
 Akhvakh
 verb
 morphology:
 functional
 motivation
 and
 orig<strong>in</strong>
 of
 an
<br />

uncommon
 pattern.”
 STUF­Language
 Typology
 and
 Universals
 61
 (4):
 309–325.
 •
<br />

Dick<strong>in</strong>son,
C.
2000.
“Mirativity
<strong>in</strong>
Tsafiki.”
Studies
<strong>in</strong>
Language
24
(2)
(September):
379‐<br />

422.

•
Floyd,
Simeon.
2011.
Conjunct/disjunct
mark<strong>in</strong>g
<strong>in</strong>
Cha’palaa
HANDOUT.
•
Hale,
<br />

A.
1980.
“Person
markers:
F<strong>in</strong>ite
conjunct
and
disjunct
verb
for<strong>ms</strong>
<strong>in</strong>
Newari.”
Papers
<strong>in</strong>
<br />

South­East
 Asian
 <strong>L<strong>in</strong>guistics</strong>
 7:
 95–106.
 •
 Hargreaves,
 David.
 2005.
 “Agency
 and
<br />

<strong>in</strong>tentional
action
<strong>in</strong>
Kathmandu
Newar.”
Himalayan
<strong>L<strong>in</strong>guistics</strong>
5:
1–48.
•
Kamp,
H.,
and
<br />

A.
 Bende‐Farkas.
 2006.
 “Epistemic
 specificity
 from
 a
 communication‐<strong>the</strong>oretic
<br />

perspective.”
unpublished
<strong>ms</strong>.,
University
of
Stuttgart.
•
Kamp,
Hans.
2011.
Represent<strong>in</strong>g
<br />

De
Se
Thoughts
and
<strong>the</strong>ir
Reports
presented
at
<strong>the</strong>
Workshop
on
De
Se
Attitudes,
March
<br />

16‐18,
 2011,
 Rio
 de
 Janeiro.
 •
 Loughnane,
 Robyn.
 2009.
 A
 grammar
 of
 Oksapm<strong>in</strong>.
 •
<br />

Norcliffe,
Elisabeth.
2011.
Conjunct/disjunct
patterns
<strong>in</strong>
Guambiano.
LSA
<strong>in</strong>stitute.
•
San
<br />

Roque,
 Lila.
 2011.
 An
 <strong>in</strong>troduction
 to
 conjunct/disjunct
 alignment
 <strong>in</strong>
 Duna
 and
 Kaluli
<br />

(Trans
New
Gu<strong>in</strong>ea).
•
Wechsler,
Stephen.
2010.
“What
‘you’
and
‘I’
mean
to
each
o<strong>the</strong>r:
<br />

Person
<strong>in</strong>dexicals,
self‐ascription,
and
<strong>the</strong>ory
of
m<strong>in</strong>d.”
Language
86
(2):
332‐365.



Masaya Yoshida (Northwestern), Chizuru Nakao (Daito Bunka) & Iván Ortega-Santos (Memphis)<br />

On ellipsis structures <strong>in</strong>volv<strong>in</strong>g a wh-remnant and a non-wh-remnant simultaneously<br />

Introduction: While both Sluic<strong>in</strong>g and Stripp<strong>in</strong>g have received detailed attention <strong>in</strong> <strong>the</strong> field<br />

of generative grammar, this is not <strong>the</strong> case for ellipsis structures that <strong>in</strong>volve both a whremnant<br />

and a non-wh-remnant simultaneously. Typical examples are found <strong>in</strong> (1)-(2).<br />

(1) A: He oído que Juan ama a María. Spanish<br />

I heard that Juan loves to María.<br />

B: Por qué [XP a María] (y no a Susana)?<br />

Why [XP to Mary] (but not to Susana)?<br />

B: Why [XP Mary] (but not Susan)?<br />

(2) A: Alguno de estos tíos estaba comiendo chorizo.<br />

One of <strong>the</strong>se guys was eat<strong>in</strong>g chorizo.<br />

B: ¿Y cuál de ellos PAELLA?<br />

and which of <strong>the</strong>m paella?<br />

In spite of <strong>the</strong> fact that <strong>the</strong>se constructions look superficially similar, one <strong>the</strong> ma<strong>in</strong> goals of<br />

this paper is to show that <strong>the</strong>y are two dist<strong>in</strong>ct construction types, namely, Why-Stripp<strong>in</strong>g,<br />

where <strong>the</strong> wh-element is restricted to why, and ano<strong>the</strong>r phrase, a remnant, which is identical<br />

to an element <strong>in</strong> <strong>the</strong> antecedent clause, (e.g. María), show up fragmentally, (1), as opposed to<br />

Wh-Stripp<strong>in</strong>g, that is to say, which <strong>in</strong>volves a wh-element o<strong>the</strong>r than why, and a remnant<br />

which contrasts with a phrase <strong>in</strong> <strong>the</strong> antecedent clause, (2). We show that remnants <strong>in</strong> <strong>the</strong>se<br />

two constructions go through different operations and are located <strong>in</strong> <strong>the</strong> different positions <strong>in</strong><br />

<strong>the</strong> syntactic structure. More specifically, we establish <strong>the</strong> follow<strong>in</strong>g clai<strong>ms</strong>: (a) Why-<br />

Stripp<strong>in</strong>g <strong>in</strong>volves a a base-generated why and leftward movement of a focused non-wh<br />

phrase followed by clausal ellipsis; (b) Wh-Stripp<strong>in</strong>g <strong>in</strong>volves movement of wh-remnant<br />

followed by rightward movement of a focused phrase and clausal ellipsis (Nev<strong>in</strong>s 2008). To<br />

<strong>the</strong> best of our knowledge it is <strong>the</strong> first time those two construction types are thoroughly<br />

compared.<br />

Analysis: Why-Stripp<strong>in</strong>g, <strong>in</strong> clear opposition to Wh-Stripp<strong>in</strong>g, allows for preposition<br />

strand<strong>in</strong>g (cf. (3) & (4)), shows effects of repair by ellipsis (cf. (5) & (6)), <strong>the</strong> antecedent<br />

clause may precede elided clause (cf. (7) & (8)), and is not clause-bound (cf. (5) & (9):<br />

(3) A: Juan habló con María en la fiesta. Why-Stripp<strong>in</strong>g<br />

Juan talked with María <strong>in</strong> <strong>the</strong> party<br />

B: Por qué (con) María?<br />

Why (with) María<br />

(4) A: Alguno de estos tíos habló con María. Wh-Stripp<strong>in</strong>g<br />

B: ¿(Y) cuál de ellos *(CON) SUSANA?<br />

(and) which of <strong>the</strong>m with Susana?<br />

“(and) Which of <strong>the</strong>se guys with Susana?”<br />

(5) A: Juan jura que va a conocer [a una chica que habla francés]. Why-Stripp<strong>in</strong>g<br />

Juan swears that he will meet to a girl who speaks French<br />

B: Por qué francés (y no alemán)?<br />

Why French (and not German)<br />

(6) A: Juan jura que va a conocer al hombre al que algunos de estos tíos le sirvió la<br />

Juan swears that he-will to meet <strong>the</strong> man to whom some of <strong>the</strong>se guys served <strong>the</strong><br />

paella. Wh-Stripp<strong>in</strong>g<br />

paella<br />

B: *Y cuál de ellos el chorizo?<br />

And which of <strong>the</strong>m <strong>the</strong> chorizo<br />

(7) No sé por qué a ti, pero María quiere verte. Why-Stripp<strong>in</strong>g<br />

Not know why to you, but María want to see-you<br />

(8) ?*No sé cuáles de ellos el chorizo, pero algunos de estos tíos se comieron la paella.


Not know which of <strong>the</strong>m <strong>the</strong> chorizo, but some of <strong>the</strong>se guys ate <strong>the</strong> paella Wh-Str.<br />

(9) A: Alguno de estos tíos negó que Juan hubiera hablado con María. Wh-Stripp<strong>in</strong>g<br />

Some of <strong>the</strong>se guys denied that Juan had talked with María<br />

B: *Y cuál de ellos con Susana?<br />

And which of <strong>the</strong>m with Susana<br />

Thus, Why-Stripp<strong>in</strong>g shows signature properties of leftward movement (e.g., Sluic<strong>in</strong>g; see<br />

Merchant 1999), whereas Wh-Stripp<strong>in</strong>g shows signature properties of rightwards movement.<br />

As is well-known, rightward movement such as Heavy NP Shift (HNPS) does nei<strong>the</strong>r allow<br />

for long-distance movement, e.g. (10), nor for P-strand<strong>in</strong>g (Jayaseelan 1990, a.o.); see <strong>the</strong><br />

unacceptability of (11)b, as opposed to (11)c. Needless to say, if we are deal<strong>in</strong>g with<br />

movement to <strong>the</strong> right, this would expla<strong>in</strong> (8).<br />

(10) *Mary said [that John broke tx] to her mo<strong>the</strong>r [<strong>the</strong> statue that he had spent <strong>the</strong><br />

whole summer carv<strong>in</strong>g]x.<br />

(11) a. John counted on a total stranger for support.<br />

b. *John counted on for support a total stranger.<br />

c. John counted for support on a total stranger. (Jayaseelan 1990: 66)<br />

Crucially, both Why and Wh-Stripp<strong>in</strong>g <strong>in</strong>volve a full-fledged syntactic structure at <strong>the</strong><br />

ellipsis site as shown by connectivity effects such as selectional restrictions and b<strong>in</strong>d<strong>in</strong>g<br />

connectivity. Fur<strong>the</strong>rmore, w.r.t. analysis of Why-Stripp<strong>in</strong>g, given <strong>the</strong> standard assumption<br />

that wh-phrases o<strong>the</strong>r than Why land <strong>in</strong> [Spec,FocP], we expect those o<strong>the</strong>r wh-phrases to be<br />

unable to co-occur with <strong>the</strong> (repeated) focused XP. The prediction is fulfilled, (12).<br />

(12) A: Juan besó a María.<br />

Juan kissed to María<br />

B: Por qué a María?/ *Dónde a María?/*Cuándo a María?/*Quién a María?<br />

B: Why to María?/*Where to María?/*When to María?/*Who to María?<br />

In turn, we assume that <strong>the</strong> remnants must escape <strong>the</strong> ellipsis <strong>in</strong> order to be pronounced, and<br />

o<strong>the</strong>rwise-covert focus movement is manifested overtly only under ellipsis (see Nakao 2008<br />

for a similar approach to Stripp<strong>in</strong>g). Under current <strong>the</strong>oretical assumption where <strong>the</strong>re is no<br />

overt/covert movement dist<strong>in</strong>ction, but ra<strong>the</strong>r a choice on which copy of <strong>the</strong> movement cha<strong>in</strong><br />

to pronounce, this means that PF constra<strong>in</strong>ts force <strong>the</strong> pronunciation of <strong>the</strong> upper copy. We<br />

claim that <strong>the</strong> situation is parallel to Multiple Sluic<strong>in</strong>g. Although <strong>the</strong> second wh-phrase<br />

usually undergoes covert wh-movement, (13), it can escape <strong>the</strong> TP only when <strong>the</strong>re is<br />

ellipsis, (14) (see Lasnik 2006 and references <strong>the</strong>re<strong>in</strong>, for discussions on Multiple Sluic<strong>in</strong>g):<br />

(13) a. Qué compré yo para quién?<br />

What bought I for who?<br />

b. *Qué para quién compré yo?<br />

What for who bought I<br />

(14) (?)Yo compré algo para alguien, pero no recuerdo [CP qué1 para quién2 [IP<br />

I bought someth<strong>in</strong>g for somebody, but not remember.I what for who<br />

yo compré t1 t2]].<br />

I bought t1 t2<br />

References<br />

Jayaseelan, Karattuparambil. A. 1990. Incomplete Vp Deletion and Gapp<strong>in</strong>g. L<strong>in</strong>guistic<br />

Analysis 20:64-81.<br />

Lasnik, Howard. 2006. Multiple Sluic<strong>in</strong>g <strong>in</strong> English. Ms. University of Maryland, College<br />

Park.<br />

Nev<strong>in</strong>s, Andrew. 2008. “Sluic<strong>in</strong>g ≠ Stripp<strong>in</strong>g: Evidence from P-Strand<strong>in</strong>g”. Paper presented<br />

at The 3rd Annual Moscow Student Conference on <strong>L<strong>in</strong>guistics</strong>, Moscow, April 2008.<br />

Merchant, J. 1999. The syntax of silence: Sluic<strong>in</strong>g, islands and identity <strong>in</strong> ellipsis. Ph.D. Diss,<br />

UCSC.


Association with traces and <strong>the</strong> copy <strong>the</strong>ory of movement<br />

In this talk I give a pr<strong>in</strong>cipled account for <strong>the</strong> observation that exclusive only must associate with a focus<br />

with<strong>in</strong> its complement (Tancredi’s (1990) Pr<strong>in</strong>ciple of Lexical Association; PLA), utiliz<strong>in</strong>g <strong>the</strong> copy <strong>the</strong>ory<br />

of movement and associated work on <strong>the</strong> <strong>in</strong>terpretation of traces. Previous explanations for this fact come<br />

from <strong>the</strong> idea that traces cannot be F-marked. I argue contra Beaver and Clark (2008) that traces (lower<br />

copies of movement cha<strong>in</strong>s) can <strong>in</strong> fact be F-marked and that this is exemplified <strong>in</strong> F-mark<strong>in</strong>g conta<strong>in</strong>ed <strong>in</strong><br />

quantifiers which undergo QR. Instead, PLA effects arise through <strong>the</strong> <strong>in</strong>terpretation of focus alternatives of<br />

<strong>the</strong> predicate <strong>in</strong> both <strong>the</strong> higher and lower copies of of <strong>the</strong> moved constituent. When <strong>the</strong> higher copy and<br />

lower copy are not both under <strong>the</strong> scope of <strong>the</strong> exclusive operator, we yield <strong>in</strong>compatible requirements on<br />

<strong>the</strong> variables be<strong>in</strong>g quantified over, caus<strong>in</strong>g ei<strong>the</strong>r presupposition failure or contradiction.<br />

PLA and <strong>the</strong> state of <strong>the</strong> art<br />

Michael Yoshitaka Erlew<strong>in</strong>e (MIT)<br />

Tancredi (1990) described <strong>the</strong> ungrammaticality of (1) as due to <strong>the</strong> PLA: a requirement that only associate<br />

with a constituent <strong>in</strong> its c-command doma<strong>in</strong> at S-structure. Beaver and Clark (2008) argue that this is an<br />

expected result if traces cannot be F-marked: without F-marked material <strong>in</strong> <strong>the</strong> complement of only <strong>the</strong><br />

focus alternatives computed will be a s<strong>in</strong>gleton set and we cannot make a mean<strong>in</strong>gful assertion. In support<br />

of this view, Beaver and Clark note that “by def<strong>in</strong>ition, extraction gaps cannot be prosodically prom<strong>in</strong>ent”<br />

(Beaver and Clark, 2008, 172).<br />

(1) * [Which boy]i does John only like ti? (Tancredi, 1990)<br />

Intended: ‘Which boy x is such that John only likes [x]F ?’<br />

(2) ✓ Which boy is such thati John only likes [himi]F ?<br />

Example (2) shows that a bound variable, however, can <strong>in</strong>deed carry F-mark<strong>in</strong>g and give us <strong>the</strong> <strong>in</strong>tended<br />

read<strong>in</strong>g. This contrast between unpronounced traces and bound variables lends credence to this hypo<strong>the</strong>sis.<br />

F-mark<strong>in</strong>g <strong>in</strong> <strong>the</strong> lower position: evidence from QR<br />

Aoun and Li (1993) noted that <strong>the</strong> PLA see<strong>ms</strong> to also affect <strong>the</strong> possible scopes of QR: (3) does not have an<br />

<strong>in</strong>verse-scope read<strong>in</strong>g due to <strong>the</strong> only and <strong>the</strong> F-mark<strong>in</strong>g on (part of) <strong>the</strong> QP. That <strong>the</strong> scope of focus-sensitive<br />

only <strong>in</strong>deed <strong>in</strong>terferes with QR can also be verified <strong>in</strong> Antecedent Conta<strong>in</strong>ed Deletion examples which must<br />

<strong>in</strong>voke QR <strong>in</strong> order to satisfy <strong>the</strong> identity requirements on ellipsis (Larson and May, 1990, a.o.). The two<br />

ellipsis resolutions <strong>in</strong> (4a) correspond to a choice of how high <strong>the</strong> quantifier is raised to—ei<strong>the</strong>r to <strong>the</strong> edge<br />

of vP1 or vP2. However, <strong>in</strong> (4b) <strong>the</strong> larger ellipsis resolution is unavailable, <strong>in</strong>dicat<strong>in</strong>g an <strong>in</strong>ability of <strong>the</strong> QP<br />

to move across only to <strong>the</strong> edge of vP1.<br />

(3) Someone only loves [QP every [boy]F <strong>in</strong> <strong>the</strong> room]. (Aoun and Li, 1993)<br />

✓ someone > every boy, *every boy > someone<br />

(4) a. John [vP1 wanted to [vP2 read [QP every book that Mary did . ✓ “want to read,” ✓ “read”<br />

b. John [vP1 wanted to only [vP2 read [QP every book that [M]F did . *“want to read,” ✓ “read”<br />

As a covert movement operation, QR has been analyzed as movement with pronunciation of <strong>the</strong> foot of <strong>the</strong><br />

cha<strong>in</strong>. Examples such as (3) and (4b) <strong>the</strong>n arguably realize F-mark<strong>in</strong>g on <strong>the</strong> lower position, correspond<strong>in</strong>g<br />

to a trace position of an overt movement cha<strong>in</strong>. When <strong>the</strong> full paradigm of <strong>the</strong> PLA <strong>in</strong>clud<strong>in</strong>g its effect on<br />

QR is considered, its explanation as an <strong>in</strong>ability to realize F-mark<strong>in</strong>g on traces becomes untenable.<br />

Solution: F-mark<strong>in</strong>g <strong>in</strong> both higher and lower copies via <strong>the</strong> copy <strong>the</strong>ory of movement<br />

Under <strong>the</strong> copy <strong>the</strong>ory of movement, movement does not leave “traces” but <strong>in</strong>stead leaves full, lower copies.<br />

The lower copy is <strong>the</strong>n converted <strong>in</strong>to a def<strong>in</strong>ite description (with <strong>the</strong> restriction that it be equal to <strong>the</strong> variable<br />

<strong>in</strong> question) through a process of Trace Conversion (TC) (Fox, 2002; Rullmann and Beck, 1998).<br />

(5) “John read many books.” QR: [many books] λxi John read [many books]i<br />

TC: [many books] λx John read [<strong>the</strong> book x]<br />

F-mark<strong>in</strong>g itself is simply a syntactic feature on constituents (Jackendoff, 1972), and thus when a constituent


conta<strong>in</strong><strong>in</strong>g F-mark<strong>in</strong>g moves, <strong>the</strong> F-mark<strong>in</strong>g is reta<strong>in</strong>ed on both copies. This triggers <strong>the</strong> generation of<br />

focus alternatives <strong>in</strong> both positions. If this movement happens with<strong>in</strong> <strong>the</strong> scope of only (Case I), we yield a<br />

mean<strong>in</strong>gful assertion by elim<strong>in</strong>ation of self-contradictory alternatives when evaluat<strong>in</strong>g only. However, this<br />

elim<strong>in</strong>ation of contradictory alternatives does not occur when <strong>the</strong> movement is out of <strong>the</strong> scope of only and<br />

we yield fatal presupposition failure (Case II).<br />

Case I (XP conta<strong>in</strong><strong>in</strong>g F-mark<strong>in</strong>g moves with<strong>in</strong> <strong>the</strong> scope of only):<br />

“Every boy” QRs to <strong>the</strong> edge of vP, with<strong>in</strong> <strong>the</strong> scope of only, and <strong>the</strong> lower copy undergoes TC (a). There are<br />

now two F-marked boys <strong>in</strong> <strong>the</strong> structure, and <strong>the</strong>y each <strong>in</strong>troduce <strong>the</strong> alternatives {boy, girl}. Compute �vP�f po<strong>in</strong>twise, result<strong>in</strong>g <strong>in</strong> four alternatives (b). (Box <strong>in</strong>dicates prejacent.) Elim<strong>in</strong>ate alternatives which produce<br />

logical contradictions (strikeouts <strong>in</strong> b). Compute only as normal: assert<strong>in</strong>g <strong>the</strong> conjunction of negations of<br />

alternatives which are not entailed by <strong>the</strong> prejacent. The result<strong>in</strong>g assertion (c) is well-formed.<br />

(3a) a. TC: Someone λy . only [vP [every [boy]F ] λx . y loves [<strong>the</strong> [boy]F x] ]<br />

b. �vP�f { }<br />

=<br />

[every boy] λx . y loves [<strong>the</strong> boy x] , [every boy] λx . y loves [<strong>the</strong> girl x],<br />

[every girl] λx . y loves [<strong>the</strong> boy x], [every girl] λx . y loves [<strong>the</strong> girl x]<br />

c. Assert: <strong>the</strong>re exists a person y for which it is not <strong>the</strong> case that y loves every girl. �<br />

Case II (XP conta<strong>in</strong><strong>in</strong>g F-mark<strong>in</strong>g moves out of <strong>the</strong> scope of only):<br />

“Every boy” QRs out above only, and <strong>the</strong> lower copy undergoes TC (a). Only <strong>the</strong> lower copy is with<strong>in</strong><br />

<strong>the</strong> vP, so <strong>the</strong>re are two alternatives computed at vP (b). The exclusive assertion <strong>the</strong>n makes reference to<br />

<strong>the</strong> def<strong>in</strong>ite description “<strong>the</strong> girl x” (c). The higher boy will trigger alternatives, but <strong>the</strong> assertion simply<br />

asserts <strong>the</strong> ord<strong>in</strong>ary semantic value of <strong>the</strong> root, mean<strong>in</strong>g we will quantify over “every boy x.” The complete<br />

computation (d) attempts to compute <strong>the</strong> def<strong>in</strong>ite description “<strong>the</strong> girl x” for “every boy x,” necessarily<br />

result<strong>in</strong>g <strong>in</strong> presupposition failure for every <strong>in</strong>dividual <strong>in</strong> <strong>the</strong> quantification!<br />

(3b) a. TC: [every [boy]F ] λx . someone λy . only [vP y loves [<strong>the</strong> [boy]F x] ]<br />

b. �vP� f = { y loves [<strong>the</strong> boy x] , y loves [<strong>the</strong> girl x] }<br />

c. �only vP� = ¬ y loves [<strong>the</strong> girl x]<br />

d. Assert: for every boy x, <strong>the</strong>re exists y, such that it is not <strong>the</strong> case that y loves [<strong>the</strong> girl x] �<br />

The contrast between traces and bound variables<br />

If <strong>the</strong> PLA is not due to a simple lack of F-mark<strong>in</strong>g with<strong>in</strong> <strong>the</strong> complement of only, we lose <strong>the</strong> trivial<br />

explanation for <strong>the</strong> contrast <strong>in</strong> (1–2). However, <strong>the</strong> contrast above (Case I–II) crucially arises due to <strong>the</strong><br />

presuppositions <strong>in</strong>troduced by <strong>the</strong> def<strong>in</strong>ite description <strong>in</strong>terpretation of lower copies via TC. Bound variables,<br />

however, are simple variables and thus can host F-mark<strong>in</strong>g without <strong>in</strong>troduc<strong>in</strong>g conflict<strong>in</strong>g presuppositions.<br />

(6) No boy is such thati John only likes [himi]F = [no boy](λx.only John likes [x]F ) �<br />

Prediction: <strong>the</strong> PLA does not affect F-marked quantificational determ<strong>in</strong>ers<br />

As TC reta<strong>in</strong>s all of <strong>the</strong> predicative restriction of a quantifier but replaces <strong>the</strong> determ<strong>in</strong>er with ι, F-mark<strong>in</strong>g<br />

on <strong>the</strong> determ<strong>in</strong>ers <strong>the</strong><strong>ms</strong>elves are predicted to be immune to <strong>the</strong> PLA. As alternatives are not <strong>in</strong>troduced <strong>in</strong><br />

<strong>the</strong> two copies of <strong>the</strong> predicate, we avoid <strong>the</strong> issues <strong>in</strong> Case II above. This prediction is borne out.<br />

(3 ′ ) Someone only loves [ [most]F boys]. ✓ someone > most boys, ✓ most boys > someone<br />

Summary and implications<br />

The novel proposal presented here gives a unified explanation for PLA effects on both overt and covert<br />

movement, <strong>in</strong> a manner compatible with our recent, <strong>in</strong>dependently-motivated understand<strong>in</strong>g of <strong>the</strong> copy <strong>the</strong>ory<br />

of movement. The dist<strong>in</strong>ction drawn here between bound variables and traces highlights <strong>the</strong> importance<br />

of our semantic denotations for “variables,” particularly as <strong>the</strong> lack of focus-islands has been used as an<br />

argument for non-movement analyses of wh-<strong>in</strong>-situ (Aoun and Li, 1993).


Peng Zhou, Stephen Cra<strong>in</strong> & Likan Zhan (Macquarie)<br />

Children’s pragmatic use of prosody <strong>in</strong> sentence process<strong>in</strong>g<br />

This study exam<strong>in</strong>ed 4-year-old Mandar<strong>in</strong>-speak<strong>in</strong>g children’s sensitivity to prosodic cues <strong>in</strong><br />

resolv<strong>in</strong>g speech act ambiguities, us<strong>in</strong>g eye-movement record<strong>in</strong>gs. Most previous on-l<strong>in</strong>e<br />

studies have focused on children’s use of prosody <strong>in</strong> resolv<strong>in</strong>g structural ambiguities and<br />

have found that, although children are able to use prosodic <strong>in</strong>formation <strong>in</strong> on-l<strong>in</strong>e sentence<br />

process<strong>in</strong>g, <strong>the</strong>y use such <strong>in</strong>formation less effectively than adults. The present study takes<br />

advantage of special properties of Mandar<strong>in</strong> Ch<strong>in</strong>ese to <strong>in</strong>vestigate <strong>the</strong> role of prosody <strong>in</strong><br />

children’s on-l<strong>in</strong>e process<strong>in</strong>g of ambiguities <strong>in</strong> which prosody serves to signal <strong>the</strong><br />

illocutionary mean<strong>in</strong>g of an utterance (i.e., whe<strong>the</strong>r <strong>the</strong> speaker is ask<strong>in</strong>g a question or<br />

mak<strong>in</strong>g a statement). Negative sentences with a wh-word like (1) are ambiguous <strong>in</strong> Mandar<strong>in</strong><br />

Ch<strong>in</strong>se. This sentence can ei<strong>the</strong>r be used to pose a question, as <strong>in</strong> (1a), or to make a statement,<br />

as <strong>in</strong> (1b). Prosodic cues are used to dist<strong>in</strong>guish between <strong>the</strong>se two speech acts. Ris<strong>in</strong>g<br />

<strong>in</strong>tonation on <strong>the</strong> wh-phrase shenme shuiguo ‘what fruit’ <strong>in</strong>dicates <strong>the</strong> question read<strong>in</strong>g,<br />

whereas level <strong>in</strong>tonation on <strong>the</strong> same wh-phrase signals <strong>the</strong> statement read<strong>in</strong>g.<br />

(1) Yuehan meiyou zhai shenme shuiguo<br />

John not pick what fruit<br />

a. What fruit did John not pick? (ris<strong>in</strong>g <strong>in</strong>tonation on <strong>the</strong> wh-phrase)<br />

b. John didn’t pick any fruit. (level <strong>in</strong>tonation on <strong>the</strong> wh-phrase)<br />

Thirty-four monol<strong>in</strong>gual Mandar<strong>in</strong>-speak<strong>in</strong>g children (mean age 4;8, range 4;1 to 5; 5) and 30<br />

Mandar<strong>in</strong>-speak<strong>in</strong>g adults (mean age 25, range 23 to 26) were tested us<strong>in</strong>g <strong>the</strong> visual world<br />

paradigm. Participants were presented with a spoken sentence (e.g., (1)) while view<strong>in</strong>g a<br />

picture (e.g., Fig.1.). Two prosodic versions of each target sentence were presented, one with<br />

ris<strong>in</strong>g <strong>in</strong>tonation on <strong>the</strong> wh-phrase (Question Prosody condition) and one with level<br />

<strong>in</strong>tonation on <strong>the</strong> same wh-phrase (Statement Prosody condition). Eye-movements were<br />

recorded us<strong>in</strong>g an EyeL<strong>in</strong>k 1000 eye tracker (remote function). Participants’ fixations were<br />

coded <strong>in</strong> three categories (see Fig.1.): <strong>the</strong> statement-compatible area (objects correspond<strong>in</strong>g<br />

to <strong>the</strong> statement <strong>in</strong>terpretation), <strong>the</strong> question-compatible area (objects correspond<strong>in</strong>g to <strong>the</strong><br />

question <strong>in</strong>terpretation) and <strong>the</strong> irrelevant area (objects correspond<strong>in</strong>g to nei<strong>the</strong>r<br />

<strong>in</strong>terpretation). The proportion of fixations follow<strong>in</strong>g <strong>the</strong> onset of <strong>the</strong> subject NP (e.g.,<br />

Yuehan ‘John’) was computed <strong>in</strong> a time w<strong>in</strong>dow of 6000 <strong>ms</strong> for <strong>the</strong> two critical categories:<br />

<strong>the</strong> statement-compatible area and <strong>the</strong> question-compatible area. This 6000 <strong>ms</strong> time w<strong>in</strong>dow<br />

was broken down <strong>in</strong>to 20 segments, each with a duration of 300 <strong>ms</strong>.<br />

Figure 2 shows <strong>the</strong> proportion of fixations of adults and children <strong>in</strong> <strong>the</strong> question-compatible<br />

area (III) across <strong>the</strong> two prosodic conditions. Figure 3 summarizes <strong>the</strong> fixation proportions <strong>in</strong><br />

<strong>the</strong> statement-compatible area (I). The two figures <strong>in</strong>dicate that adults and children exhibited<br />

similar eye gaze patterns <strong>in</strong> <strong>the</strong> two critical areas <strong>in</strong> <strong>the</strong> two prosodic conditions. Hear<strong>in</strong>g <strong>the</strong><br />

wh-word with ris<strong>in</strong>g <strong>in</strong>tonation triggered <strong>in</strong>creased looks to <strong>the</strong> question-compatible area<br />

(Fig.2.) and hear<strong>in</strong>g <strong>the</strong> wh-word with level <strong>in</strong>tonation triggered <strong>in</strong>creased looks to <strong>the</strong><br />

statement-compatible area (Fig.3.). This prosodic effect found <strong>in</strong> children was as robust as it<br />

was <strong>in</strong> adults. The patterns were supported by fur<strong>the</strong>r statistical modell<strong>in</strong>g (mixed-effects<br />

logistic regression models). This f<strong>in</strong>d<strong>in</strong>g suggests that children are as sensitive as adults <strong>in</strong><br />

us<strong>in</strong>g prosody <strong>in</strong> on-l<strong>in</strong>e sentence process<strong>in</strong>g, when prosody is used to resolve a pragmatic<br />

ambiguity. Children are as good as adults <strong>in</strong> <strong>the</strong> pragmatic use of prosody <strong>in</strong> on-l<strong>in</strong>e sentence<br />

process<strong>in</strong>g.


Fig.1. Example of <strong>in</strong>terest areas<br />

Fig.2. Average fixation proportions over time <strong>in</strong> <strong>the</strong> question-compatible area (III) <strong>in</strong><br />

<strong>the</strong> two prosodic conditions, adults (left) and children (right)<br />

Fig.3. Average fixation proportions over time <strong>in</strong> <strong>the</strong> statement-compatible area (I) <strong>in</strong><br />

<strong>the</strong> two prosodic conditions, adults (left) and children (right)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!