Anaphoric dependencies <strong>in</strong> real time: Process<strong>in</strong>g of Russian numerical constructionsMaria Pol<strong>in</strong>sky and Eric PotsdamHarvard University and <strong>the</strong> University of FloridaNatural language has numerous ways to encode anaphoric dependencies, <strong>in</strong>clud<strong>in</strong>g filler-gap(movement) constructions, antecedent-anaphor relations, control, variable b<strong>in</strong>d<strong>in</strong>g, andcoreference. Such relations can be created <strong>in</strong> <strong>the</strong> syntax (e.g. movement constructions), <strong>in</strong> <strong>the</strong>semantics (e.g. variable b<strong>in</strong>d<strong>in</strong>g), or <strong>in</strong> <strong>the</strong> discourse (e.g. coreference). Reuland 2011,build<strong>in</strong>g on Re<strong>in</strong>hart 1983 and o<strong>the</strong>rs, proposes <strong>the</strong> follow<strong>in</strong>g hierarchy <strong>in</strong> <strong>the</strong> economy of <strong>the</strong>encod<strong>in</strong>g of anaphoric dependencies.(1) syntax < semantics < discourseThe hierarchy translates <strong>in</strong>to process<strong>in</strong>g preferences; <strong>the</strong> process<strong>in</strong>g of dependencies far<strong>the</strong>rto <strong>the</strong> left should be easier than <strong>the</strong> process<strong>in</strong>g of those to <strong>the</strong> right. A specific prediction isthat syntactic dependencies require less process<strong>in</strong>g effort than discourse-deriveddependencies (Koornneef 2008). In o<strong>the</strong>r words, movement is, perhaps surpris<strong>in</strong>gly, lessburdensome for process<strong>in</strong>g than pronom<strong>in</strong>alization. To test this prediction, this paperanalyzes two constructions from Russian which have not been previously fully analyzed.They m<strong>in</strong>imally differ on <strong>the</strong> surface but we show that <strong>the</strong>y <strong>in</strong>volve dist<strong>in</strong>ct k<strong>in</strong>ds ofanaphoric dependencies. An experimental study confirms that <strong>the</strong> syntactic dependencyrequires less effort than <strong>the</strong> discourse dependency.In <strong>the</strong> Russian examples <strong>in</strong> (2), a nom<strong>in</strong>al can be fronted out of a numericalexpression, strand<strong>in</strong>g a modify<strong>in</strong>g numeral. When <strong>the</strong> stranded numeral is a so-called paucalnumber (1.5, 2, 3, 4, and <strong>the</strong> expression ‘both’), <strong>the</strong> fronted nom<strong>in</strong>al can appear <strong>in</strong> a form thatmatches <strong>in</strong> number with <strong>the</strong> numeral, (2a), or <strong>in</strong> a non-match<strong>in</strong>g plural form, (2b).(2) a. A’ movement dependency: match<strong>in</strong>g morphologySobor-a v gorodke bylo tri sobor-aca<strong>the</strong>dral-PAUCAL <strong>in</strong> town was three.PAUCALb. pronom<strong>in</strong>al coreference dependency: non-match<strong>in</strong>g morphologySobor-ov v gorodke bylo tri proca<strong>the</strong>dral-PL <strong>in</strong> town was three.PAUCAL‘As for ca<strong>the</strong>drals, <strong>the</strong>re were three <strong>in</strong> that town.’We argue that <strong>the</strong> construction with match<strong>in</strong>g between <strong>the</strong> fronted element and <strong>the</strong> numeral(2a) <strong>in</strong>volves A'-movement of <strong>the</strong> fronted element but (2b) without match<strong>in</strong>g <strong>in</strong>volves co<strong>in</strong>dexationbetween <strong>the</strong> fronted element and a null pronoun, as shown. Evidence comes fromisland sensitivity, number connectivity, b<strong>in</strong>d<strong>in</strong>g reconstruction, parasitic gaps, word order,and resumption with a pronoun or epi<strong>the</strong>t. For example, only <strong>the</strong> match<strong>in</strong>g form is sensitiveto islands (<strong>in</strong> blue), (3), and only <strong>the</strong> match<strong>in</strong>g form reconstructs to yield a B<strong>in</strong>d<strong>in</strong>g Pr<strong>in</strong>cipleC violation, (4).(3) a. *Ženixa ja pomnju [vremja [kogda u nee bylo tri ženixa]]suitor.PAUC 1SG remember time when by her was threeb. Ženixov ja pomnju [vremja [kogda u nee bylo tri pro]]suitor.PL 1SG remember time when by her was three‘As for suitors, I remember <strong>the</strong> time when she had three.’
(4) a. [Raza kogda Mašu i xvalili] ona k,*i nasčitala tri [ … ]time.PAUC when Masha.ACC praised she counted three‘She k,*i found three times when Masha i got praised.’b. [Raz kogda Mašu i xvalili] ona k,i nasčitala tri protime.PL when Masha.ACC praised she counted three‘As for times when Masha i got praised, she k,i counted three.’In contrast, only <strong>the</strong> base-generated non-match<strong>in</strong>g form allows an expletive (<strong>in</strong> red) at <strong>the</strong>post-numeral position because pronouns but not traces alternate with overt elements, (5).(5) a. Želanija u Peti bylo tri (*štuki)wish.PAUC by Petya was three piece.PAUCb. Želanij u Peti bylo tri (štuki)wish.PL by Petya was three piece.PAUC‘Wishes, Petya had three’.We conclude that <strong>the</strong> match<strong>in</strong>g form (2a) <strong>in</strong>volves A'-movement, a syntactic dependency, and<strong>the</strong> non-match<strong>in</strong>g form (2b) <strong>in</strong>volves coreference, a discourse dependency. The contrastreplicates <strong>the</strong> well-known dist<strong>in</strong>ction <strong>in</strong> Romance between topicalization and Hang<strong>in</strong>g TopicLeft Dislocation. The m<strong>in</strong>imal pair is an excellent candidate for a process<strong>in</strong>g study.The hierarchy <strong>in</strong> (1) predicts that (2a) should be processed more easily than (2b). Wetest this prediction <strong>in</strong> a read<strong>in</strong>g time experiment. The results show a strong effect of <strong>the</strong>number difference (p=0.0085), with a statistically significant slowdown <strong>in</strong> read<strong>in</strong>g time <strong>in</strong> <strong>the</strong>region after <strong>the</strong> numeral <strong>in</strong> <strong>the</strong> non-match<strong>in</strong>g case (discourse dependency) compared to <strong>the</strong>match<strong>in</strong>g case (syntactic dependency). The result supports Reuland’s hierarchy <strong>in</strong> (1) and isparticularly strik<strong>in</strong>g s<strong>in</strong>ce match<strong>in</strong>g topics are less frequent than non-match<strong>in</strong>g ones (5 tokensof (2a) vs. 12 tokens of (2b) over 1000 randomly selected sentences from <strong>the</strong> RussianNational Corpus, http://www.ruscorpora.ru/en/<strong>in</strong>dex.html).We reject an alternative explanation of <strong>the</strong> data <strong>in</strong> which <strong>the</strong> read<strong>in</strong>g time slow downis due to a simple morphological mismatch, which has been noted by several studies (e.gMol<strong>in</strong>aro et al. 2011). First, agreement mismatches <strong>in</strong> Russian numeral expressions do noto<strong>the</strong>rwise result <strong>in</strong> read<strong>in</strong>g time slow down (Xiang et al. 2011). Second, native speakers rated<strong>the</strong> two constructions comparably; <strong>in</strong> o<strong>the</strong>r studies, agreement mismatch has yielded loweracceptability rat<strong>in</strong>gs (Fanselow and Frisch 2006).In conclusion, our analysis of a syntactic m<strong>in</strong>imal pair <strong>in</strong> Russian forms <strong>the</strong> basis of aprocess<strong>in</strong>g study compar<strong>in</strong>g a syntactic dependency to a discourse dependency. Ourexperimental <strong>in</strong>vestigation of <strong>the</strong> two constructions shows that syntactic dependencies areprocessed more quickly than discourse dependencies, provid<strong>in</strong>g novel support for <strong>the</strong>hierarchy <strong>in</strong> (1). From a process<strong>in</strong>g perspective, movement is less burdensome thanpronom<strong>in</strong>alization.ReferencesFanselow, G., and S. Frisch. 2006. Effects of process<strong>in</strong>g difficulty on judgments of acceptability. InG. Fanselow, et al. Gradience <strong>in</strong> grammar, 291-316. Oxford: Oxford University Press.Koornneef, A. 2008. Eye-catch<strong>in</strong>g anaphora. Utrecht: LOT International Dissertation Series.Mol<strong>in</strong>aro, N. et al. 2011. Grammatical agreement process<strong>in</strong>g <strong>in</strong> read<strong>in</strong>g: ERP f<strong>in</strong>d<strong>in</strong>gs and futuredirections. Cortex 47, 908-930.Re<strong>in</strong>hart, T. 1983. Anaphora and semantic <strong>in</strong>terpretation. London: Croom Helm.Reuland, E. 2011. Anaphora and language design. Cambridge, Ma.: MIT Press.Xiang, M. et al. 2011. Process<strong>in</strong>g morphological ambiguity: An experimental <strong>in</strong>vestigation of Russiannumerical phrases. L<strong>in</strong>gua 121, 548-560.
- Page 1 and 2:
GLOW Newsletter #70, Spring 2013Edi
- Page 3 and 4:
INTRODUCTIONWelcome to the 70 th GL
- Page 5:
Welcome to GLOW 36, Lund!The 36th G
- Page 8 and 9:
REIMBURSEMENT AND WAIVERSThe regist
- Page 10 and 11:
STATISTICS BY COUNTRYCountry Author
- Page 12 and 13:
15:45-16:00 Coffee break16:00-17:00
- Page 14 and 15:
14:00-15:00 Adam Albright (MIT) and
- Page 16 and 17:
17:00-17:30 Anna Maria Di Sciullo (
- Page 18 and 19:
16.10-16.50 Peter Svenonius (Univer
- Page 20 and 21:
GLOW 36 WORKSHOP PROGRAM IV:Acquisi
- Page 22 and 23:
The impossible chaos: When the mind
- Page 24 and 25:
17. Friederici, A. D., Trends Cogn.
- Page 26 and 27:
Second, tests replicated from Bruen
- Page 28 and 29:
clusters is reported to be preferre
- Page 30 and 31:
occur (cf. figure 1). Similar perfo
- Page 32 and 33:
argument that raises to pre-verbal
- Page 34 and 35:
Timothy Bazalgette University of
- Page 36 and 37:
. I hurt not this knee now (Emma 2;
- Page 38 and 39:
Rajesh Bhatt & Stefan Keine(Univers
- Page 40 and 41:
SIZE MATTERS: ON DIACHRONIC STABILI
- Page 42 and 43:
ON THE ‘MAFIOSO EFFECT’ IN GRAM
- Page 44 and 45:
The absence of coreferential subjec
- Page 46 and 47:
PROSPECTS FOR A COMPARATIVE BIOLING
- Page 48 and 49:
A multi-step algorithm for serial o
- Page 50 and 51:
Velar/coronal asymmetry in phonemic
- Page 52 and 53:
On the bilingual acquisition of Ita
- Page 54 and 55:
Hierarchy and Recursion in the Brai
- Page 56 and 57:
Colorful spleeny ideas speak furiou
- Page 58 and 59:
A neoparametric approach to variati
- Page 60 and 61:
Lexical items merged in functional
- Page 62 and 63:
Setting the elements of syntactic v
- Page 64 and 65:
Language Faculty, Complexity Reduct
- Page 66 and 67:
Don’t scope your universal quanti
- Page 68 and 69:
Restricting language change through
- Page 70 and 71:
4. Conclusion This micro-comparativ
- Page 72 and 73:
2. Central Algonquian feature hiera
- Page 74 and 75:
availability of the SR reading in (
- Page 76 and 77:
Repairing Final-Over-Final Constrai
- Page 78 and 79:
a PF interface phenomenon as propos
- Page 80 and 81:
(b) Once the learner has determined
- Page 82 and 83:
cognitive recursion (including Merg
- Page 84 and 85:
can be null, or lexically realized,
- Page 86 and 87:
feature on C and applies after Agre
- Page 88 and 89: Nobu Goto (Mie University)Deletion
- Page 90 and 91: Structural Asymmetries - The View f
- Page 92 and 93: FROM INFANT POINTING TO THE PHASE:
- Page 94 and 95: Some Maladaptive Traits of Natural
- Page 96 and 97: Constraints on Concept FormationDan
- Page 98 and 99: More on strategies of relativizatio
- Page 100 and 101: ReferencesBayer, J. 1984. COMP in B
- Page 102 and 103: Improper movement and improper agre
- Page 104 and 105: Importantly, while there are plausi
- Page 106 and 107: This hypothesis makes two predictio
- Page 108 and 109: (3) a. Það finnst alltaf þremur
- Page 110 and 111: (2) Watashi-wa hudan hougaku -wa /*
- Page 112 and 113: However when the VP (or IP) is elid
- Page 114 and 115: More specifically, this work reflec
- Page 116 and 117: modality, or ii) see phonology as m
- Page 118 and 119: (I) FWHA The wh-word shenme ‘what
- Page 120 and 121: 1The historical reality of biolingu
- Page 122 and 123: Rita Manzini, FirenzeVariation and
- Page 124 and 125: Non-counterfactual past subjunctive
- Page 126 and 127: THE GRAMMAR OF THE ESSENTIAL INDEXI
- Page 128 and 129: Motivating head movement: The case
- Page 130 and 131: Limits on Noun-suppletionBeata Mosk
- Page 132 and 133: Unbounded Successive-Cyclic Rightwa
- Page 134 and 135: Same, different, other, and the his
- Page 136 and 137: Selectivity in L3 transfer: effects
- Page 140 and 141: Constraining Local Dislocation dial
- Page 142 and 143: A Dual-Source Analysis of GappingDa
- Page 144 and 145: [9] S. Repp. ¬ (A& B). Gapping, ne
- Page 146 and 147: of Paths into P path and P place is
- Page 148 and 149: Deriving the Functional HierarchyGi
- Page 150 and 151: Reflexivity without reflexivesEric
- Page 152 and 153: Reuland, E. (2001). Primitives of b
- Page 154 and 155: on v, one associated with uϕ and t
- Page 156 and 157: Merge when applied to the SM interf
- Page 158 and 159: 1 SachsThe Semantics of Hindi Multi
- Page 160 and 161: Covert without overt: QR for moveme
- Page 162 and 163: Morpho-syntactic transfer in L3 acq
- Page 164 and 165: one where goals receive a theta-rel
- Page 166 and 167: 51525354555657585960616263646566676
- Page 168 and 169: follow Harris in assuming a ranked
- Page 170 and 171: changing instances of nodes 7 and 8
- Page 172 and 173: Sam Steddy, steddy@mit.eduMore irre
- Page 174 and 175: Fleshing out this model further, I
- Page 176 and 177: (5) Raman i [ CP taan {i,∗j}Raman
- Page 178 and 179: properties with Appl (introduces an
- Page 180 and 181: econstruct to position A then we ca
- Page 182 and 183: (5) Kutik=i ez guret-a.dog=OBL.M 1S
- Page 184 and 185: sults summarized in (2) suggest tha
- Page 186 and 187: Building on Bhatt’s (2005) analys
- Page 188 and 189:
Underlying (derived from ON) /pp, t
- Page 190 and 191:
out, as shown in (3) (that the DP i
- Page 192 and 193:
Word order and definiteness in the
- Page 194 and 195:
Visser’s Generalization and the c
- Page 196 and 197:
the key factors. The combination of
- Page 198 and 199:
Parasitic Gaps Licensed by Elided S
- Page 200 and 201:
Stages of grammaticalization of the