29.10.2014 Views

A Study of Word Order Variation in German, with Special ... - CiteSeerX

A Study of Word Order Variation in German, with Special ... - CiteSeerX

A Study of Word Order Variation in German, with Special ... - CiteSeerX

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

A STUDY OF WORD ORDER VARIATION IN GERMAN,<br />

WITH SPECIAL REFERENCE TO MODIFIER PLACEMENT<br />

January 1994<br />

Ralf Günter Wilhelm Ste<strong>in</strong>berger<br />

Ph.D. Thesis<br />

submitted to the University <strong>of</strong> Manchester <strong>in</strong> the Faculty <strong>of</strong> Technology<br />

Department <strong>of</strong> Language and L<strong>in</strong>guistics<br />

University <strong>of</strong> Manchester Institute <strong>of</strong> Science and Technology<br />

I


This work was carried out under the supervision <strong>of</strong> Dr. Paul Bennett.<br />

No portion <strong>of</strong> the work referred to <strong>in</strong> the thesis has been submitted <strong>in</strong> support <strong>of</strong> an<br />

application for another degree or qualification <strong>of</strong> this or any other university or<br />

other <strong>in</strong>stitution <strong>of</strong> learn<strong>in</strong>g.<br />

II


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994)<br />

Abstract<br />

This work discusses word order <strong>in</strong> written <strong>German</strong> at sentence level, and suggests how to<br />

deal <strong>with</strong> word order variation <strong>in</strong> Mach<strong>in</strong>e Translation. It specially refers to modifier<br />

placement, as modifiers are generally neglected <strong>in</strong> l<strong>in</strong>guistic (word order) description.<br />

The order <strong>of</strong> phrases <strong>in</strong> free word order languages is not entirely free, as some variations<br />

can be ungrammatical, and further variations are less natural than others. The <strong>in</strong>tuition <strong>of</strong> a<br />

<strong>German</strong> speaker on the adequate word order <strong>in</strong> a specific context is <strong>in</strong>fluenced by at least<br />

eleven factors. In this thesis, these are described <strong>in</strong>dependently, and their <strong>in</strong>teraction is<br />

shown. The context plays a major role for the natural word order <strong>in</strong> a sentence, so that one<br />

can say that sentences are embedded <strong>in</strong> their context.<br />

After the l<strong>in</strong>guistic description, the methods suggested <strong>in</strong> literature to model word order<br />

variation <strong>in</strong> Natural Language Process<strong>in</strong>g are discussed, and a suggestion is made to<br />

overcome the problems l<strong>in</strong>ked to word order variation. For analysis, means are provided to<br />

recognise theme, rheme and focus <strong>of</strong> a given sentence. For synthesis, it is proposed to use a<br />

flexible canonical form which <strong>in</strong>volves over eighty position classes, <strong>in</strong>clud<strong>in</strong>g places for<br />

the categories theme, rheme and focus. Depend<strong>in</strong>g on the analysis <strong>of</strong> thematic, rhematic<br />

and focused phrases <strong>in</strong> the source language <strong>of</strong> the translation, vary<strong>in</strong>g <strong>German</strong> sentences<br />

are generated to guarantee their appropriate embedd<strong>in</strong>g <strong>in</strong> the context. The appendix<br />

conta<strong>in</strong>s a list <strong>of</strong> adverbs <strong>in</strong>clud<strong>in</strong>g the syntactic features which are necessary for their<br />

automatic treatment.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

Acknowledgements<br />

I would like to express my gratitude to my supervisor Dr. Paul Bennett for his support,<br />

po<strong>in</strong>ters to relevant literature, his fast reaction, and useful comments.<br />

I would also like to thank my <strong>in</strong>ternational colleagues and friends who helped me to f<strong>in</strong>d<br />

out about tricky aspects <strong>of</strong> their languages. I am particularly grateful to T<strong>in</strong>a Burnley,<br />

Archana H<strong>in</strong>duja, and Chris Chambers, who put a lot <strong>of</strong> effort <strong>in</strong>to pro<strong>of</strong>-read<strong>in</strong>g the<br />

thesis. Their comments and suggestions were very helpful. Chris' rules-<strong>of</strong>-thumb on<br />

modifier placement f<strong>in</strong>ally helped me to avoid most errors concern<strong>in</strong>g adverb position<strong>in</strong>g<br />

<strong>in</strong> English, on which I failed to have a native speaker's <strong>in</strong>tuition.<br />

Many thanks go also to my former colleagues Dr. Nadia Mesli, Oliver Streiter and Randy<br />

Sharp at the Institute for Applied Information Science (IAI) <strong>in</strong> Saarbrücken. They<br />

supported me a lot when I implemented my ideas on word order <strong>in</strong> the Mach<strong>in</strong>e<br />

Translation system CAT2, by expla<strong>in</strong><strong>in</strong>g the formalism as well as the <strong>German</strong>, English and<br />

French grammars.<br />

And f<strong>in</strong>ally, I want to thank T<strong>in</strong>a Burnley for her excellent cook<strong>in</strong>g and personal support,<br />

especially dur<strong>in</strong>g the last tir<strong>in</strong>g months.<br />

II


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

Education and Background<br />

1994 • Centre for Computational L<strong>in</strong>guistics (CCL) at UMIST: TRADE Mach<strong>in</strong>e<br />

Translation project (English-Spanish-Italian)<br />

1993 • CCL, UMIST: Conception and Implementation <strong>of</strong> a <strong>German</strong>-English dictionary<br />

for Computer-Assisted Language Learn<strong>in</strong>g (CALL); plann<strong>in</strong>g <strong>of</strong><br />

new projects, fund-rais<strong>in</strong>g<br />

1991 - 1992 • CCL, UMIST: Mach<strong>in</strong>e Translation project EUROTRA (French-English)<br />

• Tra<strong>in</strong><strong>in</strong>g <strong>in</strong> the ET-6 Mach<strong>in</strong>e Translation formalism ALEP, Luxembourg<br />

• Summer School for Logic, L<strong>in</strong>guistics and Information (LLI), Colchester<br />

1991 • Institute for Applied Information Science (IAI), Saarbrücken (FRG):<br />

CAT2 Mach<strong>in</strong>e Translation project (<strong>German</strong>-English)<br />

• LLI Summer School, Saarbrücken<br />

1/1991 • Magister Artium ("<strong>with</strong> dist<strong>in</strong>ction"), München<br />

1986 - 1990 • PANA Schaumst<strong>of</strong>f GmbH, Geretsried (FRG): Public Relations, sales promotion,<br />

production management<br />

1984 - 1985 • Lycée Louis-Le-Grand, Paris: Teacher Assistant (PAD scholarship)<br />

1982 - 1/1991 • Studies <strong>of</strong> Theoretical, French and Spanish L<strong>in</strong>guistics at Ludwig-<br />

Maximilians Universität München (1986-1990) and Freie Universität Berl<strong>in</strong><br />

(1982-1986)<br />

1981 - 1982 • PANA Werk KG, Wolfratshausen (FRG): Executive Tra<strong>in</strong><strong>in</strong>g <strong>in</strong> the<br />

textiles field<br />

1980 • Abitur, Gymnasium Pullach (FRG), mathematical/scientific orientation<br />

III


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

CONTENTS<br />

1 DESCRIPTION OF THE PROBLEM ........................................................................1<br />

1.1 Scope and Limits <strong>of</strong> the Thesis ..............................................................................1<br />

1.2 <strong>Word</strong> <strong>Order</strong> <strong>in</strong> a Wider Context.............................................................................6<br />

1.3 Contents................................................................................................................13<br />

1.4. Problems <strong>of</strong> <strong>German</strong> <strong>Word</strong> <strong>Order</strong> Description ....................................................16<br />

1.5. Why Describe <strong>Word</strong> <strong>Order</strong>?.................................................................................19<br />

1.5.1 Analysis <strong>of</strong> <strong>German</strong> Sentences <strong>in</strong> NLP................................................................19<br />

1.5.1.1 Disambiguation <strong>of</strong> Homonyms ............................................................................20<br />

1.5.1.2 Resolution <strong>of</strong> PP-Attachment...............................................................................21<br />

1.5.1.3 Recognition <strong>of</strong> Emphasis .....................................................................................22<br />

1.5.1.4 Contextual Embedd<strong>in</strong>g <strong>of</strong> Sentences ...................................................................23<br />

1.5.1.5 Scope <strong>of</strong> Degree Modifiers ..................................................................................26<br />

1.5.2 Synthesis <strong>of</strong> <strong>German</strong> <strong>in</strong> NLP................................................................................27<br />

1.5.2.1 Basic <strong>Order</strong><strong>in</strong>g <strong>of</strong> Constituents ............................................................................30<br />

1.5.2.2 Cumulation <strong>of</strong> Modifiers......................................................................................30<br />

1.5.2.3 Correct Scope .......................................................................................................31<br />

1.5.2.4 Sentence Embedd<strong>in</strong>g............................................................................................31<br />

1.5.3 Foreign Language Teach<strong>in</strong>g.................................................................................32<br />

2. COMPLEMENT, MODIFIER, ADVERB AND ADVERB SUBTYPES...............34<br />

2.1. Modifiers (Angaben) vs. Complements (Ergänzungen) ......................................35<br />

2.2. Def<strong>in</strong>ition and Classification <strong>of</strong> Adverbs.............................................................39<br />

2.2.1. Different Def<strong>in</strong>itions <strong>of</strong> the Adverb.....................................................................40<br />

2.2.2. Semantic Classification........................................................................................43<br />

2.2.3. Adverbs and Related <strong>Word</strong> Classes .....................................................................46<br />

2.2.3.1. Adverbs and Particles...........................................................................................46<br />

2.2.3.2. Adverbs vs. Conjunctions and Prepositions.........................................................47<br />

2.2.3.3. Adverbs vs. Adjectives.........................................................................................49<br />

2.2.4 Conclusion, F<strong>in</strong>al Def<strong>in</strong>ition................................................................................55<br />

2.3. Modifier Types (Angabeklassen).........................................................................57<br />

2.4. Some Information on the Position <strong>of</strong> Modifiers ..................................................62<br />

IV


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

2.5. Some Statistical Facts about Adverbs..................................................................67<br />

3. FACTORS WHICH DETERMINE GERMAN WORD ORDER ..........................70<br />

3.1 Theme-Rheme Structure ......................................................................................72<br />

3.1.1 Some Def<strong>in</strong>itions <strong>of</strong> the Terms Theme and Rheme .............................................72<br />

3.1.2 The Realization <strong>of</strong> Thematic and Rhematic Elements.........................................74<br />

3.1.3 The <strong>Order</strong> <strong>of</strong> Thematic and Rhematic Complements...........................................76<br />

3.1.4 The Separation <strong>of</strong> Theme and Rheme by Modifiers ............................................80<br />

3.2 Behaghel's "Gesetz der wachsenden Glieder"......................................................81<br />

3.3 Functional Sentence Perspective..........................................................................82<br />

3.3.1 Thematisation and Rhematisation........................................................................84<br />

3.3.2 Further Means to Express Functional Sentence Perspective ...............................85<br />

3.4. Verbnähe ..............................................................................................................90<br />

3.4.1. Which Elements are Semantically Close to the Verb?.........................................93<br />

3.4.1.1. Arguments ............................................................................................................93<br />

3.4.1.2. Modifiers ..............................................................................................................95<br />

3.4.2. Limits <strong>of</strong> the Verbnähe Pr<strong>in</strong>ciple .........................................................................98<br />

3.5. The Animacy-First Pr<strong>in</strong>ciple................................................................................99<br />

3.6. Semantic Roles...................................................................................................101<br />

3.7. Scope ..................................................................................................................103<br />

3.7.1. Def<strong>in</strong>itions <strong>of</strong> Scope...........................................................................................104<br />

3.7.2. Problems <strong>with</strong> the Term Scope ..........................................................................105<br />

3.7.3. The Difference between Scope and Focalisation...............................................107<br />

3.8. Rhythm...............................................................................................................108<br />

3.9. Natural Gender ...................................................................................................109<br />

3.10. Grammaticalisation (Habit)................................................................................110<br />

3.11. Lenerz' "Satzklammerbed<strong>in</strong>gung"......................................................................111<br />

4. THE INTERACTION OF PREFERENCE RULES, AND SOME<br />

RESTRICTIONS.................................................................................................................113<br />

4.1. Interaction <strong>of</strong> the Pr<strong>in</strong>ciples ...............................................................................113<br />

4.2. Relative Weight <strong>of</strong> some Pr<strong>in</strong>ciples...................................................................116<br />

4.3. Calculation <strong>of</strong> Acceptability ..............................................................................120<br />

4.4. Restriction on the Interaction <strong>of</strong> Preference Rules ............................................123<br />

4.4.1. Syntactic Subord<strong>in</strong>ation .....................................................................................123<br />

V


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

4.4.2. Possessive Relations...........................................................................................125<br />

4.4.3. Quantificational Elements..................................................................................126<br />

4.4.4. The Pragmatic Need <strong>of</strong> a Sentence Focus..........................................................126<br />

4.5. Summary ............................................................................................................129<br />

5. HOW TO DESCRIBE GERMAN FREE WORD ORDER FORMALLY ..........131<br />

5.1. Acceptability Calculation and LP Rule Disjunction ..........................................131<br />

5.2. The Relevance <strong>of</strong> a Canonical Form for <strong>German</strong> ..............................................133<br />

5.3. Canonical Forms for <strong>German</strong> <strong>in</strong> Literature........................................................137<br />

5.3.1. Engel's Canonical Form .....................................................................................138<br />

5.3.2. Hoberg's Canonical Form...................................................................................139<br />

5.3.3. New Prelim<strong>in</strong>ary Canonical Form .....................................................................141<br />

5.4. Why Do some Sentences Differ from the Basic <strong>Word</strong> <strong>Order</strong> ............................145<br />

5.5. The Vorfeld Position..........................................................................................149<br />

5.6. The Importance <strong>of</strong> Theme, Rheme and Focus ...................................................154<br />

6. AIDS FOR COMPUTATIONAL ANALYSIS........................................................159<br />

6.1. Compulsory <strong>Order</strong>s ............................................................................................159<br />

6.2. Recognition <strong>of</strong> Focus .........................................................................................163<br />

6.3. Recognition or Theme and Rheme.....................................................................167<br />

6.4. Some More Details.............................................................................................171<br />

6.4.1. Permutation <strong>of</strong> Pragmatic Modifiers..................................................................171<br />

6.4.2. Permutation <strong>of</strong> Modal Modifiers........................................................................172<br />

6.4.3. Permutation <strong>of</strong> Pragmatic and Situative/Modal Modifiers ................................173<br />

6.4.4. Permutation <strong>of</strong> Situative Modifiers....................................................................174<br />

6.5. F<strong>in</strong>al Version <strong>of</strong> the Canonical Form.................................................................177<br />

6.5.1. Placement <strong>of</strong> the Theme.....................................................................................177<br />

6.5.2. Placement <strong>of</strong> the Rheme.....................................................................................180<br />

6.5.3. Placement <strong>of</strong> the Focus ......................................................................................184<br />

6.6. Preferential PP Attachment ................................................................................189<br />

6.7. Dictionary Entries for Adverbs ..........................................................................191<br />

6.7.1. Cod<strong>in</strong>g <strong>of</strong> Adverbs <strong>in</strong> the Dictionary.................................................................191<br />

6.7.2. Some Generalizations.........................................................................................201<br />

6.8. Summary <strong>of</strong> Chapter 6 .......................................................................................203<br />

VI


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

7. CONCLUDING REMARKS ....................................................................................205<br />

8. APPENDIX.................................................................................................................210<br />

8.1. Angabestellungsklassen (accord<strong>in</strong>g to Hoberg, 1981: 106-131) .......................210<br />

8.2. Alphabetical List<strong>in</strong>g <strong>of</strong> Modifiers......................................................................214<br />

8.3. List<strong>in</strong>g <strong>of</strong> Modifiers Accord<strong>in</strong>g to Position Classes ..........................................230<br />

8.4. Canonical Form (F<strong>in</strong>al Version, cf. 6.5.3) .........................................................248<br />

9. BIBLIOGRAPHY......................................................................................................249<br />

VII


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

ABBREVIATIONS<br />

a<br />

a i<br />

8.1)<br />

a mod<br />

a neg<br />

Angabe (modifier)<br />

<strong>in</strong>dex numbers (i) refer to Hoberg's position classes_a1-a44 (cf.<br />

modal modifiers (a42-a44)<br />

negational modifiers (a40)<br />

a pragm , a exist pragmatic (existimatorial) modifiers (a1 - a18)<br />

a sit<br />

situative modifiers (a19-a40)<br />

+/- a +/- animate<br />

A<br />

accusative case<br />

Adj<br />

predicate adjective<br />

adv<br />

adverb<br />

ap<br />

adjectival phrase<br />

card<br />

card<strong>in</strong>al numbers<br />

Comp<br />

Comparability (can an adverb be compared?)<br />

conj<br />

conjunction<br />

D<br />

dative case<br />

+/- d +/- def<strong>in</strong>ite<br />

DIR<br />

directional complement<br />

Dist<br />

Distance (can a degree modifier be separated from the modified<br />

phrase?)<br />

EN Ulrich Engel (1988)<br />

Exp<br />

expansive complement<br />

FSP<br />

functional sentence perspective<br />

G, GEN genitive case<br />

Grad<br />

gradability (can an adverb be modified by a degree modifier?)<br />

HK87 Mannheimer Handbuchkorpus 1987 (cf. 6.7)<br />

HO Ursula Hoberg (1981)<br />

man manner modifiers (a 43 )<br />

N<br />

nom<strong>in</strong>ative case<br />

Neg<br />

Negability (can an adverb be negated?)<br />

Nom<br />

predicate noun<br />

NP<br />

nom<strong>in</strong>al phrase<br />

npp<br />

NP or PP<br />

OF<br />

Obligatorische-Folge-Regel<br />

PO<br />

prepositional object<br />

PP<br />

prepositional phrase<br />

pragm pragmatic modifiers (a 1 -a 18 )<br />

Pred<br />

predicative use (can an adverb be used predicatively?)<br />

Pre/Post<br />

Does a degree modifier precede or follow modified phrases?<br />

pron<br />

pronom<strong>in</strong>al<br />

RS<br />

Ralf Ste<strong>in</strong>berger<br />

s<br />

sentence<br />

Sit<br />

situative complement<br />

sit situative modifiers (a 19 -a 40 )<br />

VIII


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

SVC<br />

support verb construction<br />

SVO<br />

subject-verb-object word order<br />

SOV<br />

subject-object-verb word order<br />

TRS<br />

theme-rheme structure<br />

VF<br />

Vorfeld<br />

VL<br />

verb-last, f<strong>in</strong>al position <strong>of</strong> the verb <strong>in</strong> the clause<br />

VSO<br />

verb-subject-object word order<br />

V2<br />

verb-second, second position <strong>of</strong> the verb <strong>in</strong> the clause<br />

XP phrase <strong>of</strong> whatever category (NP, PP, AP, ...)<br />

SYMBOLS<br />

CAPITALS represent semantic roles (AGENT etc)<br />

or <strong>in</strong>dicate (contrastive) stress<br />

A < B<br />

A precedes (tends to precede) B<br />

A / B<br />

complementary occurrence <strong>of</strong> A and B<br />

? sentence is less acceptable/natural than <strong>with</strong>out "?"<br />

?? sentence is less acceptable/natural than sentence <strong>with</strong> "?"<br />

* sentence is ungrammatical<br />

# ungrammatical; can be considered acceptable if very strongly<br />

stressed (contrast)<br />

! unacceptable for semantic reasons<br />

+/-* The judgement on the +/- value <strong>of</strong> a feature is based on occurrences<br />

<strong>in</strong> the corpus HK87<br />

-*&<br />

Although no positive evidence was found <strong>in</strong> HK87, our <strong>in</strong>tuition is<br />

that the feature value should be positive (+) (cf. 8.1 and 8.2)<br />

[...] square brackets <strong>in</strong> quotations <strong>in</strong>dicate omission or addition <strong>of</strong> text<br />

IX


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST 1994)<br />

Me<strong>in</strong>en Eltern<br />

X


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994)<br />

1 DESCRIPTION OF THE PROBLEM<br />

Before present<strong>in</strong>g the contents <strong>of</strong> this thesis <strong>in</strong> detail, by describ<strong>in</strong>g the chapters one by one<br />

(1.3), we want to po<strong>in</strong>t out the scope and the limits <strong>of</strong> our work (1.1), and set the<br />

<strong>in</strong>vestigation <strong>in</strong> a wider context (1.2). The other sections <strong>of</strong> chapter 1 are dedicated to the<br />

problems (1.4) and to the benefits <strong>of</strong> an appropriate word order description (1.5).<br />

1.1 SCOPE AND LIMITS OF THE THESIS<br />

The adequate treatment <strong>of</strong> word order is still an open problem <strong>in</strong> l<strong>in</strong>guistics.<br />

(Erbach, 1993: 177)<br />

The goal <strong>of</strong> this section is to describe both the scope and the limits <strong>of</strong> this thesis. We also<br />

feel the need to describe the work we carried out earlier as it is the basis for the further<br />

development presented here.<br />

This thesis conta<strong>in</strong>s a l<strong>in</strong>guistic word order description <strong>of</strong> written <strong>German</strong> at sentence level,<br />

<strong>in</strong>clud<strong>in</strong>g the variation <strong>of</strong> constituents, and the limits <strong>of</strong> their <strong>in</strong>terchangeability.<br />

Furthermore, it makes suggestions on how to use this theoretical knowledge <strong>in</strong> Natural<br />

Language Process<strong>in</strong>g (NLP) <strong>in</strong> general, and Mach<strong>in</strong>e Translation (MT) <strong>in</strong> particular. In<br />

Mach<strong>in</strong>e Translation, the challenges <strong>of</strong> language analysis, language synthesis, and languagecontrastive<br />

differences are comb<strong>in</strong>ed. As most other NLP applications have to solve one <strong>of</strong><br />

these tasks, they can make use <strong>of</strong> our suggestions for the treatment <strong>of</strong> word order <strong>in</strong> Mach<strong>in</strong>e<br />

Translation.<br />

A further use <strong>of</strong> the data which can be found here concerns foreign language teach<strong>in</strong>g. <strong>Word</strong><br />

order <strong>in</strong> general, and the order<strong>in</strong>g <strong>of</strong> modifiers <strong>in</strong> particular are widely neglected <strong>in</strong> grammar<br />

books (cf. 2.4). Although the <strong>in</strong>formation presented <strong>in</strong> this work is too detailed for language<br />

learners, it conta<strong>in</strong>s many explanations and order<strong>in</strong>g rules, which can be presented <strong>in</strong> a<br />

simple way to non-l<strong>in</strong>guists (1.5.3) who <strong>in</strong>tend to learn <strong>German</strong> as a foreign language.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 2<br />

We limit ourselves to the sentence level, as this is the most difficult part <strong>of</strong> word order<br />

description. With<strong>in</strong> nom<strong>in</strong>al (NP) or prepositional phrases (PP), sequences are either<br />

relatively fixed, or they are easy to describe. 1 shows how limited the permutation <strong>of</strong> noun<br />

arguments is:<br />

1a Caesars Verteidigung der Stadt gegen die Angreifer<br />

1b die Verteidigung der Stadt gegen die Angreifer durch Caesar<br />

1c * die Verteidigung der Stadt durch Caesar gegen die Angreifer<br />

1d * die Verteidigung gegen die Angreifer der Stadt durch Caesar 1<br />

1e * die Verteidigung gegen die Angreifer durch Caesar der Stadt<br />

1f * die Verteidigung durch Caesar gegen die Angreifer der Stadt<br />

1g * die Verteidigung durch Caesar der Stadt gegen die Angreifer<br />

Furthermore, the position <strong>of</strong> determ<strong>in</strong>ers, prepositions and adjectival modifiers relative to<br />

nouns does not allow for any variation at all:<br />

2 * Angreifer die<br />

3 * die Angreifer gegen<br />

4 * die Angreifer bösen<br />

Verb participles <strong>with</strong> the function <strong>of</strong> noun modifiers are the only elements which allow for<br />

some variation <strong>in</strong> the sequence <strong>of</strong> their modifiers (5). Although we did not <strong>in</strong>vestigate the<br />

order <strong>of</strong> these adjuncts, it seems that they follow the same regularities as modifiers at<br />

sentence level. At least the comparison <strong>of</strong> 5a <strong>with</strong> 6a and 5f <strong>with</strong> 6b gives this impression:<br />

5a<br />

5b<br />

5c<br />

5d<br />

5e<br />

5f<br />

6a<br />

6b<br />

das damals dort aus München verfrüht angekommene Flugzeug<br />

das dort damals aus München verfrüht angekommene Flugzeug<br />

das damals aus München verfrüht dort angekommene Flugzeug<br />

das damals dort verfrüht aus München angekommene Flugzeug<br />

das damals verfrüht aus München dort angekommene Flugzeug<br />

* das dort verfrüht aus München damals angekommene Flugzeug<br />

Das Flugzeug kam damals dort aus München verfrüht an.<br />

* Das Flugzeug kam dort verfrüht aus München damals an.<br />

Our research concentrated on the description <strong>of</strong> declarative clauses, which represent the<br />

major part <strong>of</strong> most written discourse. It turned out to be the case that ma<strong>in</strong> clauses <strong>in</strong>volv<strong>in</strong>g<br />

a Vorfeld 2 can best be treated as a particular <strong>in</strong>stance <strong>of</strong> subord<strong>in</strong>ate clauses, which do not<br />

1 The * <strong>in</strong> 1d refers to 1d as a reformulation <strong>of</strong> the mean<strong>in</strong>g <strong>in</strong> 1a.<br />

2 For a brief description <strong>of</strong> the terms Vorfeld, Mittelfeld and Nachfeld, see section 1.5.2.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 3<br />

have a Vorfeld. Hence our ma<strong>in</strong> concern is the Mittelfeld. The treatment <strong>of</strong> the Vorfeld does<br />

not require a lot <strong>of</strong> attention (see section 5.5).<br />

<strong>Word</strong> order regularities <strong>in</strong> imperative sentences and questions will probably differ slightly<br />

from the order <strong>in</strong> declarative clauses, <strong>in</strong> that they can <strong>in</strong>volve a different sentential focus.<br />

The order <strong>of</strong> the elements which are not affected by the sentential focus, however, should<br />

rema<strong>in</strong> the same.<br />

Although the order<strong>in</strong>g <strong>of</strong> verb arguments is dealt <strong>with</strong> <strong>in</strong> many parts <strong>of</strong> the thesis, our<br />

<strong>in</strong>tention was to focus on modifiers. The reason for this is that verb arguments have been<br />

widely discussed, and from different perspectives, whereas the description <strong>of</strong> modifier<br />

position is generally limited to small subsets. When discuss<strong>in</strong>g word order, l<strong>in</strong>guists<br />

generally choose to compare the two frequent groups <strong>of</strong> temporal and local modifiers (e.g.<br />

Lenerz 1977: 79ff, Vennemann 1982: 4ff and 19f, Gadler 1982: 159f,<br />

Whittemore/Ferrara/Brunner 1990: 29, Oliva 1992b: 11). Others specialise on further<br />

subgroups, such as the dist<strong>in</strong>ction between verb modifiers and sentential modifiers<br />

(Her<strong>in</strong>ger/Strecker/Wimmer, 1980: 278). Thurmair (1989) and Weydt (1977) describe toners<br />

(Abtönungspartikeln 3 ), Waltz<strong>in</strong>g (1986) concentrates on existimatorial modifiers, Altmann<br />

(1976 and 1978) on degree modifiers, and Jacobs (1982) on the negational particle.<br />

To our knowledge, Engel (1988, 1973, 1970, and others) and Hoberg (1981) are the only<br />

authors who did not avoid mention<strong>in</strong>g all the other smaller and bigger modify<strong>in</strong>g subgroups.<br />

Consequently, their work played a major role <strong>in</strong> our research. The fact that modifiers are less<br />

frequently discussed <strong>in</strong> l<strong>in</strong>guistics and computational l<strong>in</strong>guistics is reflected <strong>in</strong> our<br />

bibliography. Many <strong>of</strong> the books and articles quoted were written more than five or ten years<br />

ago. More recent literature on this neglected field is scarce. 4<br />

3 For def<strong>in</strong>itions <strong>of</strong> the different subclasses see section 2.2.<br />

4 We have used and quoted all relevant articles which appeared <strong>in</strong> the follow<strong>in</strong>g journals, reviews and<br />

proceed<strong>in</strong>gs (<strong>in</strong> alphabetical order):<br />

• Archiv für das Studium der neueren Sprachen, 1990 and 1991, Berl<strong>in</strong>


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 4<br />

With<strong>in</strong> the description <strong>of</strong> modifier position, we propose order<strong>in</strong>g rules, as well as a list <strong>of</strong><br />

features, which are necessary to handle modifiers <strong>in</strong> NLP. We believe that these means are<br />

sufficient to deal <strong>with</strong> modifiers realised as adverbs, PPs and NPs. On practical grounds,<br />

however, we focus on adverbials realised by s<strong>in</strong>gle words, such as adverbs, toners, and<br />

others (cf. 2.2). The reason is that one-word modifiers are a closed class which can be listed<br />

exhaustively. This gives the advantage that full dictionary entries, <strong>in</strong>corporat<strong>in</strong>g all details<br />

on their classification and encod<strong>in</strong>g, can be made available.<br />

PPs and NPs, however, are an open class. Its elements have to be analysed and classified<br />

before their handl<strong>in</strong>g by the procedure suggested here. We did not suggest solutions for their<br />

analysis, as this is a complex problem which should be <strong>in</strong>vestigated <strong>in</strong>dependently. We are<br />

conv<strong>in</strong>ced, however, that our classification is a first step for their proper analysis, as it<br />

provides the necessary categories and features.<br />

This thesis builds on theoretical and practical work which we have carried out earlier. In our<br />

Magisterarbeit (Ste<strong>in</strong>berger, 1990), we compared word order permutations <strong>in</strong> hundreds <strong>of</strong><br />

example pairs, <strong>in</strong> order to judge which variations are more natural. The sentences <strong>in</strong>volved<br />

• ACL: Proceed<strong>in</strong>gs <strong>of</strong> the Conference: Annual Meet<strong>in</strong>g <strong>of</strong> the Association for Computational<br />

L<strong>in</strong>guistics, 1990, 1991, 1992, 1993<br />

• CLAUS: Reports <strong>of</strong> the Computational L<strong>in</strong>guistics Department at the University <strong>of</strong> Saarland, 1990,<br />

1991, 1992, 1993, Saarbrücken<br />

• COLING: Proceed<strong>in</strong>gs <strong>of</strong> the International Conference on Computational L<strong>in</strong>guistics, 1990, 1992<br />

• Der Deutschunterricht. Beiträge zu se<strong>in</strong>er Praxis und wissenschaftlichen Grundlegung, 1990,<br />

1991, 1992, Seelze<br />

• Deutsch als Fremdsprache, 1991 and 1992, München/Berl<strong>in</strong><br />

• Deutsche Sprache. Zeitschrift für Theorie, Praxis, Dokumentation, 1990 and 1991, Berl<strong>in</strong><br />

• EACL: European Chapter <strong>of</strong> the Association for Computational L<strong>in</strong>guistics, Proceed<strong>in</strong>gs <strong>of</strong> the<br />

Conference, 1991, 1992, 1993<br />

• EUROTRA-D: Work<strong>in</strong>g Papers <strong>of</strong> the Institute <strong>of</strong> Applied Information Science (IAI), 1986, 1987,<br />

1988, 1989, 1990, 1991, 1992<br />

• Muttersprache, 1991 and 1992, Wiesbaden<br />

• Language. Journal <strong>of</strong> the L<strong>in</strong>guistic Society <strong>of</strong> America, 1990, 1991, 1992, Baltimore<br />

• L<strong>in</strong>gua. International Review <strong>of</strong> General L<strong>in</strong>guistics, 1991 and 1992, Amsterdam<br />

• L<strong>in</strong>guistic Analysis, 1990 and 1991, Seattle<br />

• L<strong>in</strong>guistic Inquiry, 1985, 1986, 1987, 1989, 1990, 1991, 1992<br />

• Natural Language & L<strong>in</strong>guistic Theory, 1991 and 1992, Dordrecht/Boston/London<br />

• Proceed<strong>in</strong>gs <strong>of</strong> the Third Conference on Applied Language Process<strong>in</strong>g, 1992<br />

• UMIST/CCL-Reports: Reports <strong>of</strong> the Centre for Computational L<strong>in</strong>guistics at UMIST, 1981-1993,<br />

Manchester


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 5<br />

all types <strong>of</strong> verb arguments, as well as elements <strong>of</strong> all major modifier classes (cf. 2.3). We<br />

also carried out tests on the order <strong>of</strong> dative and accusative complements, as well as on the<br />

relevance <strong>of</strong> the animacy and def<strong>in</strong>iteness features for these verb arguments. Both results are<br />

discussed <strong>in</strong> 4.2 below. On the basis <strong>of</strong> these results, we described a canonical form, which<br />

we adopt <strong>in</strong> 5.3.3, and which we develop further <strong>in</strong> chapter 6.<br />

In the same work, we listed translational difficulties which are not l<strong>in</strong>ked to position<strong>in</strong>g.<br />

These <strong>in</strong>clude the complex problem which is the scope <strong>of</strong> negation, and the fact that some<br />

modifiers are not translated by the same category. In this thesis, we shall assume the<br />

simplest case only, namely the translation <strong>of</strong> an adverb by an adverb (7). However,<br />

translation is <strong>of</strong>ten more complicated.<br />

In the case <strong>of</strong> transposition (Pelz, 1963: 8ff), adverbs can be translated by another word<br />

class (8). In a special case <strong>of</strong> transposition, the chassé-croisé, the function <strong>of</strong> two words is<br />

<strong>in</strong>terchanged (Barth, 1961, 80ff). In 9, for <strong>in</strong>stance, the mean<strong>in</strong>g <strong>of</strong> the <strong>German</strong> adverb gern<br />

is represented by the English ma<strong>in</strong> verb to like, and the <strong>German</strong> ma<strong>in</strong> verb lesen by the<br />

English subord<strong>in</strong>ate participle read<strong>in</strong>g. The modifier doch <strong>in</strong> 10 disappears completely <strong>in</strong> the<br />

target language. In 11, the <strong>German</strong> adverb causes the use <strong>of</strong> the cont<strong>in</strong>uous form <strong>in</strong> English.<br />

The difficulty <strong>in</strong> 12 is that the adverb ganz refers to a PP, whereas its English equivalent<br />

modifies the noun <strong>in</strong>side the PP 5 :<br />

7 Er kam gestern.<br />

He arrived yesterday.<br />

8 Er wäre be<strong>in</strong>ahe h<strong>in</strong>gefallen.<br />

Il a failli tomber.<br />

9 Ich lese gern.<br />

I like read<strong>in</strong>g.<br />

10 Ich kann doch nichts dafür.<br />

Je n'y peux rien.<br />

I cannot do anyth<strong>in</strong>g about it.<br />

5 For universally available means to express the mean<strong>in</strong>g <strong>of</strong> adverbs, see Schachter (1985: 22).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 6<br />

11 Er liest gerade.<br />

He is read<strong>in</strong>g.<br />

12 ganz am Ende<br />

at the very end 6<br />

We shall not discuss these problems any further. Complex translation as <strong>in</strong> 8 to 12 has to be<br />

treated by different means. For a discussion <strong>of</strong> such cases see Thurmair (1990), Dorr (1990),<br />

and L<strong>in</strong>dop/Tsujii (1991). We shall not comment on the scope <strong>of</strong> negation <strong>in</strong> this thesis<br />

either. Negation is a complex phenomenon which deserves an <strong>in</strong>dependent <strong>in</strong>vestigation. For<br />

the treatment <strong>of</strong> degree modifier scope, see 3.7.<br />

Besides the Magisterarbeit, we shall refer to Ste<strong>in</strong>berger (1992a and 1992b). These papers<br />

present an implementation <strong>of</strong> <strong>German</strong> and English modifier treatment <strong>with</strong><strong>in</strong> the CAT2<br />

Mach<strong>in</strong>e Translation (MT) system. CAT2, a sidel<strong>in</strong>e <strong>of</strong> the European MT system Eurotra, is<br />

a rule-based, unification and constra<strong>in</strong>t-based formalism (Sharp, 1989, 1993). In Ste<strong>in</strong>berger<br />

(1992a), we describe the implementation <strong>of</strong> <strong>German</strong> and English word order treatment, as<br />

suggested <strong>in</strong> chapter 6 <strong>of</strong> this thesis, <strong>with</strong> m<strong>in</strong>or differences. In 1992b, we present a way <strong>of</strong><br />

recognis<strong>in</strong>g and translat<strong>in</strong>g degree modifier scope <strong>in</strong> the same formalism. Degree modifier<br />

treatment is only partially handled <strong>in</strong> this thesis (3.7 and 6.7). The 1992 papers are thus an<br />

application <strong>of</strong> the f<strong>in</strong>d<strong>in</strong>gs presented <strong>in</strong> this thesis.<br />

1.2 WORD ORDER IN A WIDER CONTEXT<br />

On l'apprend difficilement et est encore plus fascheux à prononcer: de sorte que les<br />

enfants mêmes, qui sont naiz au pays, sont bien grandeletz avant qu'ils puissent<br />

bien former les mots et pr<strong>of</strong>érer les paroles. (Claude Duret on the <strong>German</strong><br />

language, 17 th century, quoted <strong>in</strong> Scaglione, 1981: 39) 7<br />

Some people seem to believe that word order differences between languages are not just<br />

differences <strong>in</strong> the order <strong>of</strong> words. It was said that the language <strong>of</strong> a nation shows how its<br />

people th<strong>in</strong>k. One claim is that some languages encourage clear and precise th<strong>in</strong>k<strong>in</strong>g,<br />

6 For an analysis <strong>of</strong> this special use <strong>of</strong> the adverb very, see Brugmann (1984).<br />

7 "It is difficult to learn and even worse to pronounce, so that even children who are born <strong>in</strong> the country are<br />

quite old before they are able to shape and utter the words."


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 7<br />

whereas others are confus<strong>in</strong>g, so that they provide ideal means to evoke emotions, and to<br />

deceive. In 1751, the French philosopher <strong>of</strong> the Enlightenment, Denis Diderot, affirmed his<br />

conviction that his language was superior to others:<br />

il faut parler français dans la société et dans les écoles de philosophie; et grec, lat<strong>in</strong>, anglais dans les<br />

chaires et sur les théâtres [...] Le français est fait pour <strong>in</strong>struire, éclairer et conva<strong>in</strong>cre; le grec, le lat<strong>in</strong>,<br />

l'italien, l'anglais, pour persuader, émouvoir et tromper." 8<br />

Denis Diderot, as well as his colleague Anto<strong>in</strong>e de Rivarol, claimed that a major reason for<br />

French be<strong>in</strong>g so correct, clear and precise is its direct word order, which is due to a lack <strong>of</strong><br />

<strong>in</strong>version. In 1784, the Prussian Academy even awarded Rivarol a prize for his essay, <strong>in</strong><br />

which he used this argument to expla<strong>in</strong> why French was, and deserved to be, the "universal<br />

language <strong>of</strong> <strong>in</strong>tellectual <strong>in</strong>tercourse" (Scaglione, 1981: 5).<br />

<strong>German</strong>, on the other hand, is traditionally regarded as a language which is "both unusually<br />

systematic and ultimately illogical, even irrational" <strong>with</strong> respect to word order (Scaglione,<br />

1981: 3). One peculiarity <strong>of</strong> <strong>German</strong> is exactly what, from the French po<strong>in</strong>t <strong>of</strong> view, looks<br />

like <strong>in</strong>version, namely the possibility to shift the subject beh<strong>in</strong>d the verb, by start<strong>in</strong>g the<br />

sentence <strong>with</strong> an object or an adverbial. Does the flexibility <strong>of</strong> word order prevent <strong>German</strong><br />

from be<strong>in</strong>g clear and precise?<br />

Another equally popular and doubtful myth is one <strong>of</strong> "<strong>German</strong> be<strong>in</strong>g the philosophical<br />

language par excellence, or at least the ideal language for philosophy" (Scaglione, 1981: 4).<br />

If one believes Mart<strong>in</strong> Heidegger, the lack <strong>of</strong> direct order, and the illogicality <strong>of</strong> <strong>German</strong> do<br />

not particularly prevent <strong>German</strong> philosophers from hav<strong>in</strong>g deep <strong>in</strong>sights 9 :<br />

8 "One should speak French <strong>in</strong> society and <strong>in</strong> schools <strong>of</strong> philosophy; and Greek, Lat<strong>in</strong> and English <strong>in</strong><br />

universities and theatres. French is suitable to teach, to enlighten and to conv<strong>in</strong>ce; Greek, Lat<strong>in</strong>, Italian<br />

and English are suitable to persuade, move and deceive." In: Lettre sur les sourds et muets (1751), quoted<br />

<strong>in</strong> Scaglione (1981: 5)<br />

9 Mart<strong>in</strong> Heidegger <strong>in</strong> an <strong>in</strong>terview recorded on September 23, 1966 by the staff <strong>of</strong> Der Spiegel. Due to the<br />

philosopher's wishes, the <strong>in</strong>terview was only published after his death (Der Spiegel, 30-23, May 31,<br />

1976). Found <strong>in</strong> Scaglione (1981: 4).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 8<br />

Ich denke an die besondere <strong>in</strong>nere Verwandtschaft der deutschen Sprache mit der Sprache der Griechen<br />

und deren Denken. Das bestätigen mit heute immer wieder die Franzosen. Wenn sie zu denken<br />

anfangen, sprechen sie Deutsch; sie versichern, sie kämen mit ihrer Sprache nicht durch [...] Weil sie<br />

sehen, daß sie mit ihrer ganzen großen Rationalität nicht mehr durchkommen <strong>in</strong> der heutigen Welt,<br />

wenn es sich darum handelt, diese <strong>in</strong> der Herkunft ihres Wesens zu verstehen.<br />

Another more scientific context <strong>in</strong> which <strong>German</strong> word order has been discussed is its<br />

comparison <strong>with</strong> other languages. Vennemann (1982), for <strong>in</strong>stance, contrasts <strong>German</strong> <strong>with</strong><br />

Korean and English. The po<strong>in</strong>t <strong>of</strong> departure for his discussion is that the unmarked order <strong>of</strong><br />

elements <strong>in</strong> the <strong>German</strong> sentence 13 is similar to the one <strong>in</strong> Korean (14), but it is nearly a<br />

complete <strong>in</strong>version <strong>of</strong> the English order <strong>in</strong> 15 (Vennemann, 1982: 10ff):<br />

<strong>German</strong>:<br />

13 Maria gab gestern <strong>in</strong> Seoul e<strong>in</strong>em Freund das Buch.<br />

5 0 4 3 2 1<br />

Korean:<br />

14 Maria yesterday Seoul_<strong>in</strong> friend_DAT book_ACC gave. 10<br />

5 4 3 2 1 0<br />

English:<br />

15 Mary gave the book to a friend <strong>in</strong> Seoul yesterday.<br />

5 0 1 2 3 4<br />

At the sentential level, the Korean word order thus corresponds to the order <strong>in</strong> <strong>German</strong><br />

subord<strong>in</strong>ate clauses, <strong>in</strong> which the verb is <strong>in</strong> the f<strong>in</strong>al position. Us<strong>in</strong>g Greenberg's (1966)<br />

classification, one would say that Korean is an SOV language, a language <strong>in</strong> which the verb<br />

is <strong>in</strong> the f<strong>in</strong>al position, and <strong>in</strong> which the subject precedes the object. English, however, is<br />

SVO: the subject precedes the verb, whereas the object follows it. <strong>German</strong> seems to be SOV<br />

<strong>in</strong> subord<strong>in</strong>ate clauses, and SVO <strong>in</strong> ma<strong>in</strong> declarative clauses. What does this <strong>in</strong>volve for the<br />

classification <strong>of</strong> <strong>German</strong> <strong>in</strong> Greenberg's typology? Is it SVO, because every sentence has a<br />

10 In Vennemann's paper, the Korean sentence is pr<strong>in</strong>ted us<strong>in</strong>g Korean letters. We are not able to show these<br />

here due to a lack <strong>of</strong> appropriate pr<strong>in</strong>t<strong>in</strong>g facilities.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 9<br />

ma<strong>in</strong> clause, whereas subord<strong>in</strong>ate clauses can be avoided? Or is it SOV, because the order <strong>of</strong><br />

the other elements is very similar to the Korean order?<br />

Whatever the answer is, the complementary distribution <strong>of</strong> verb positions imposes the view,<br />

that there is an underly<strong>in</strong>g word order. But can we exclude <strong>in</strong> that case, that <strong>German</strong> is a<br />

VSO language, as Beckmann (1980) dares to assume? To answer these questions, one has to<br />

consider further facts.<br />

In part <strong>of</strong> his implicational universals, Greenberg (1966) claims that there is a<br />

correspondence between the position <strong>of</strong> the verb <strong>in</strong> the sentence, and other sequences. These<br />

<strong>in</strong>clude the position <strong>of</strong> the preposition relative to the noun, the genitive attribute relative to<br />

its head, the position <strong>of</strong> adjectives and adverbs relative to the elements they modify, and<br />

many more.<br />

The idea <strong>of</strong> implicational universals and their correlation was developed further by<br />

Vennemann (1977), who formulated the pr<strong>in</strong>ciple <strong>of</strong> natural serialisation. This pr<strong>in</strong>ciple<br />

says that languages tend to be consistent <strong>in</strong> the direction <strong>in</strong> which elements are specified<br />

(modified or complemented). When modifiers and complements precede their head, the<br />

construction is called pre-specify<strong>in</strong>g (16), when they follow, it is post-specify<strong>in</strong>g (17).<br />

16 complements/modifiers head (e.g. sehr gut)<br />

17 head complements/modifiers (e.g. Hit der_Woche)<br />

If all specification relations <strong>with</strong><strong>in</strong> one language have the same direction, this language is<br />

called consistent. However, many languages, <strong>in</strong>clud<strong>in</strong>g <strong>German</strong>, are mixed <strong>with</strong> respect to<br />

specification direction. The <strong>German</strong> degree modification shown <strong>in</strong> 16, for <strong>in</strong>stance, is a prespecify<strong>in</strong>g<br />

construction, whereas the genitive attribute <strong>in</strong> 17 post-specifies its head (the noun<br />

Hit).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 10<br />

Accord<strong>in</strong>g to Vennemann (1974, 1975), the pr<strong>in</strong>ciple <strong>of</strong> natural serialisation is one <strong>of</strong> the<br />

forces which act on languages diachronically. As only consistent languages satisfy the<br />

pr<strong>in</strong>ciple, mixed languages tend to change towards consistency.<br />

If <strong>German</strong> was an SOV language, Greenberg's fourth universal would suggest that it should<br />

have postpositions, as opposed to prepositions (1966: 79).<br />

Universal 4: With overwhelm<strong>in</strong>gly greater than chance frequency, languages <strong>with</strong> normal SOV order<br />

are postpositional.<br />

Accord<strong>in</strong>g to Greenberg and Vennemann, adpositions 11 are the heads <strong>of</strong> PPs (pre-<br />

/postpositional phrases). Prepositional phrases are hence post-specify<strong>in</strong>g constructions, as<br />

the noun phrase complements the preposition. The similarity <strong>of</strong> <strong>German</strong> <strong>with</strong> Korean shown<br />

<strong>in</strong> 1 suggests that <strong>German</strong> is pre-specify<strong>in</strong>g at the sentential level. At the PP level, however,<br />

it is post-specify<strong>in</strong>g. Here aga<strong>in</strong>, <strong>German</strong> proves to be a mixed language <strong>with</strong> respect to<br />

specification direction. If Vennemann's claim is true that languages diachronically tend<br />

towards consistency, where is <strong>German</strong> go<strong>in</strong>g? Is it drift<strong>in</strong>g towards consistent pre- or postspecification?<br />

Scaglione (1981: 29ff) claims that <strong>German</strong> is develop<strong>in</strong>g away from a former basic order<br />

SOV towards the order SVO, and thus towards post-specification. Lehmann (1978b: 49ff) <strong>in</strong><br />

pr<strong>in</strong>ciple seems to agree <strong>with</strong> this view. He ma<strong>in</strong>ta<strong>in</strong>s that, <strong>in</strong> earlier times, <strong>German</strong>, as well<br />

as English, French and Spanish, were SOV languages (see also Fanselow, 1987: 128ff).<br />

While English and the Romance languages have nearly completed their drift towards SVO,<br />

<strong>German</strong> is still mixed. Lehmann expla<strong>in</strong>s the different position <strong>of</strong> <strong>German</strong> by the<br />

<strong>in</strong>terference <strong>of</strong> social forces, which prevented its development towards consistency. The<br />

crucial difference <strong>with</strong> the neighbour<strong>in</strong>g languages was that, at the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> the 16 th<br />

century, the sentence-f<strong>in</strong>al verb position was <strong>in</strong>troduced artificially for <strong>German</strong> subord<strong>in</strong>ate<br />

clauses (Lehmann, 1978c: 410ff; see also section 1.5.2).<br />

11 Accord<strong>in</strong>g to Vennemann's term<strong>in</strong>ology, we shall use the term adposition as a superord<strong>in</strong>ate for pre- and<br />

postpositions.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 11<br />

S<strong>in</strong>ce that time, two conflict<strong>in</strong>g patterns acted on the <strong>German</strong> language: the post-specify<strong>in</strong>g<br />

construction SVO <strong>in</strong> ma<strong>in</strong> clauses, and the pre-specify<strong>in</strong>g SOV sequence <strong>in</strong> subord<strong>in</strong>ate<br />

clauses. Accord<strong>in</strong>g to Lehmann (1978c: 409ff), the new SOV pattern is responsible for the<br />

emergence <strong>of</strong> several pre-modify<strong>in</strong>g structures, which did not exist before the 16 th century.<br />

These <strong>in</strong>clude the appearance <strong>of</strong> several postpositions, the use <strong>of</strong> preposed relative<br />

constructions, and others.<br />

Accord<strong>in</strong>g to Vennemann (1982: 27f), today's <strong>German</strong> is closer to be<strong>in</strong>g a post-specify<strong>in</strong>g<br />

than a pre-specify<strong>in</strong>g language. It is thus closer to English than it is to Korean, which is a<br />

consistent pre-specify<strong>in</strong>g system. The follow<strong>in</strong>g graph by Vennemann (1982: 34) visualises<br />

the scale <strong>of</strong> pre- and post-specification. Consistent post-specify<strong>in</strong>g languages such as Maori<br />

(Vennemann, 1982: 32) belong at the far right <strong>of</strong> the scale, consistent pre-specify<strong>in</strong>g ones at<br />

the far left.<br />

Korean<br />

<strong>German</strong> English<br />

+________________________________________________________+_______+______+<br />

consistent<br />

consistent<br />

pre-speci-<br />

post-specification<br />

fication<br />

A further context <strong>in</strong> which <strong>German</strong> word order is frequently mentioned is when discuss<strong>in</strong>g<br />

free word order languages. <strong>German</strong> is opposed to languages <strong>with</strong> relatively fixed word order<br />

such as English, <strong>in</strong> that <strong>German</strong> allows many permutations <strong>of</strong> constituents at sentence level.<br />

In a sentence <strong>with</strong> two verb arguments and one sentence modifier, for <strong>in</strong>stance, there can be<br />

six comb<strong>in</strong>atorial possibilities: ABC, ACB, BAC, BCA, CAB and CBA. In English,<br />

however, only modifiers have a certa<strong>in</strong> choice <strong>of</strong> position. When subject and object<br />

<strong>in</strong>terchange, the mean<strong>in</strong>g <strong>of</strong> the sentence is distorted (cf. 3.3.2):<br />

18a<br />

18b<br />

Der Mann sieht den Tisch.<br />

Den Tisch sieht der Mann.<br />

19a The man sees the table.<br />

19b * The table sees the man.<br />

The term free word order is rather confus<strong>in</strong>g, and even wrong, as it suggests that all<br />

elements <strong>of</strong> the sentence can be scrambled and put together <strong>in</strong> whatever new order.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 12<br />

Obviously, this is not possible <strong>in</strong> <strong>German</strong>, as determ<strong>in</strong>ers have to precede the noun,<br />

adjectives must be adjacent to the noun they qualify, and the verb position has to be<br />

sentence-<strong>in</strong>itial <strong>in</strong> yes/no questions, but verb-f<strong>in</strong>al <strong>in</strong> subord<strong>in</strong>ate clauses, just to give a few<br />

examples. For this reason, Schäufele (1991: 365f) suggests the more accurate term free<br />

phrase order for <strong>German</strong>, and pleads that the term free word order language be used for<br />

languages which also allow the permutation <strong>of</strong> elements <strong>with</strong><strong>in</strong> a phrase, such as Sanskrit,<br />

Lat<strong>in</strong>, Ancient Greek, and Warlpiri. However, as the expression free word order is wellestablished<br />

<strong>in</strong> l<strong>in</strong>guistic literature, we shall keep it.<br />

It is probably less well-known that even the constituents <strong>of</strong> a sentence cannot be permuted<br />

<strong>with</strong> each other that easily, and that some restrictions apply. Some constituents must precede<br />

others, and <strong>in</strong> other cases, one word order variation is more natural. The naturalness <strong>of</strong> a<br />

sentence is <strong>of</strong>ten determ<strong>in</strong>ed by the context, but sometimes no context makes a sentence<br />

sound natural. Even the replacement <strong>of</strong> one element by another one, which is closely related<br />

semantically, can make a natural sentence sound odd or even make it ungrammatical.<br />

These last questions, <strong>in</strong>clud<strong>in</strong>g both the reasons and the limits for word order freedom <strong>in</strong><br />

today's <strong>German</strong>, are the ones we 12 shall concentrate on. We mentioned language universals<br />

and diachronic change <strong>in</strong> order to put our research <strong>in</strong> a larger context. However, an ultimate<br />

12 It is worth po<strong>in</strong>t<strong>in</strong>g out our use <strong>of</strong> personal pronouns at the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> this thesis: One <strong>of</strong> the difficulties<br />

<strong>of</strong> scientific writ<strong>in</strong>g is to express oneself <strong>in</strong> an objective way <strong>with</strong>out produc<strong>in</strong>g heavy and artificial<br />

language. Objectivity requires the use <strong>of</strong> impersonal constructions such as passive voice, nom<strong>in</strong>alisations<br />

and expletives. Readability, on the other hand, calls for an agent who carries out the verbal action. As<br />

psychological tests have shown that active versions <strong>of</strong> sentences are processed and recalled better<br />

(Macdonald, 1983), I made the conscious decision to use personal pronouns. As there seems to be a strong<br />

convention not to use the first person s<strong>in</strong>gular pronoun, we shall stick to the plural.<br />

[...] many scientists and scientific editors now recognise and even promote the use <strong>of</strong> the active<br />

voice, <strong>in</strong>clud<strong>in</strong>g the use <strong>of</strong> first-person pronouns.<br />

Why should we avoid the passive voice? The rhetoric books describe it as "dron[<strong>in</strong>g] like noth<strong>in</strong>g<br />

under the sun," wordy and unclear, and less direct and less vigorous than the active voice. It may<br />

<strong>in</strong>deed be all those th<strong>in</strong>gs, but <strong>in</strong> addition, psychological research has shown that the active versions<br />

<strong>of</strong> a sentence are recalled better and verified faster. Scientific texts written <strong>in</strong> the third person<br />

passive, as "It was concluded that ..." are remembered less well and appreciated less than the same<br />

content written <strong>in</strong> the active voice. (Macdonald, 1983: 1893)


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 13<br />

answer to the questions regard<strong>in</strong>g language change and language universals is not crucial for<br />

a correct description <strong>of</strong> today's <strong>German</strong> language system.<br />

In the next section, we shall present the contents <strong>of</strong> this thesis.<br />

1.3 CONTENTS<br />

Besides clarify<strong>in</strong>g the scope <strong>of</strong> this thesis (1.1), and sett<strong>in</strong>g the problem <strong>of</strong> word order<br />

description <strong>in</strong> a larger frame (1.2), this first chapter <strong>in</strong>tends to present the problems related to<br />

our task (1.4), as well as the different practical benefits which can be reaped from our<br />

results. We shall po<strong>in</strong>t out the advantages for language teach<strong>in</strong>g (1.5.3), and for Natural<br />

Language Process<strong>in</strong>g, dist<strong>in</strong>guish<strong>in</strong>g analysis (1.5.1) and synthesis (1.5.2).<br />

Chapter 2 provides the def<strong>in</strong>itions and dist<strong>in</strong>ctions we need for a concise word order<br />

description. These are made up <strong>of</strong> the dichotomous concepts <strong>of</strong> modifier and complement<br />

(2.1), as well as the def<strong>in</strong>ition for the part-<strong>of</strong>-speech adverb, which is more heterogeneous<br />

than all other word classes. In 2.2, we give a def<strong>in</strong>ition which opposes it to related word<br />

classes. As a practical simplification, we shall roughly dist<strong>in</strong>guish three major modifier<br />

types, which will be <strong>in</strong>troduced <strong>in</strong> 2.3. Section 2.4 summarises the scarce and contradictory<br />

<strong>in</strong>formation on the position <strong>of</strong> modifiers <strong>in</strong> grammar books. The chapter ends <strong>with</strong> a few<br />

statistical facts about adverbs (2.5).<br />

Chapter 3 is dedicated to the explanation <strong>of</strong> the <strong>German</strong> word order<strong>in</strong>g mechanism. We list<br />

eleven different factors which expla<strong>in</strong> <strong>German</strong> native speakers' <strong>in</strong>tuitions on why one word<br />

order variation is better than another. Most <strong>of</strong> these pr<strong>in</strong>ciples act on verb arguments as well<br />

as modifiers. The strongest factors are l<strong>in</strong>ked to theme-rheme structure (3.1), functional<br />

sentence perspective (3.3), verb bond<strong>in</strong>g (3.4), animacy (3.5), and semantic roles (3.6).<br />

Another powerful determ<strong>in</strong>ant is the scope <strong>of</strong> scope-<strong>in</strong>clud<strong>in</strong>g elements (3.7), which we<br />

discuss aga<strong>in</strong> <strong>in</strong> 4.4.3. Further regularities which have an <strong>in</strong>fluence on our word order<br />

preferences are l<strong>in</strong>ked to heav<strong>in</strong>ess (3.2), rhythm (3.8), the natural gender <strong>of</strong> the persons that


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 14<br />

nouns refer to (3.9), hardened grammatical structures (3.10), and f<strong>in</strong>ally a regularity which<br />

Lenerz (1977) mentions as "Satzklammerbed<strong>in</strong>gung" (3.11).<br />

Most <strong>of</strong> these pr<strong>in</strong>ciples are extracted from diverse l<strong>in</strong>guistic literature and grammar books.<br />

The ones concern<strong>in</strong>g natural gender (3.9) and harden<strong>in</strong>g <strong>of</strong> syntactic structures (3.10) are the<br />

result <strong>of</strong> our own <strong>in</strong>vestigation.<br />

Chapter 4 deals <strong>with</strong> the question <strong>of</strong> how these different determ<strong>in</strong>ants <strong>in</strong>teract. The most<br />

likely explanation seems to be that all factors act <strong>in</strong> each sentence. As, depend<strong>in</strong>g on the<br />

specific order<strong>in</strong>g, the factors can require different sequences (4.1), the question arises <strong>of</strong> how<br />

strong the various pr<strong>in</strong>ciples are (4.2). A model <strong>of</strong> how to calculate the acceptability <strong>of</strong><br />

sentences is presented <strong>in</strong> 4.3. A problem related to the calculation is, that other factors<br />

restrict the <strong>in</strong>teraction <strong>of</strong> factors. These constra<strong>in</strong>ts are listed <strong>in</strong> 4.4.<br />

Due to the problems <strong>of</strong> word order calculation, we suggest further ways <strong>of</strong> deal<strong>in</strong>g <strong>with</strong><br />

word order <strong>in</strong> chapter 5. The most efficient method proves to be a sophisticated canonical<br />

form (5.2). After discuss<strong>in</strong>g two different canonical form versions suggested by Engel<br />

(1988) and Hoberg (1981), we present a new, prelim<strong>in</strong>ary canonical form (5.3), which will<br />

be ref<strong>in</strong>ed <strong>in</strong> chapter 6. In section 5.4, we compare some sentences <strong>with</strong> the canonical form<br />

<strong>in</strong> order to show its appropriateness. We also throw light on the reasons why some sentences<br />

differ from the fixed order which is represented <strong>in</strong> the canonical form. As the Vorfeld has not<br />

been considered yet, neither <strong>in</strong> the word order pr<strong>in</strong>ciples nor <strong>in</strong> the canonical forms, we<br />

discuss its special status <strong>in</strong> 5.5. Section 5.6 is devoted to the importance <strong>of</strong> theme, rheme and<br />

focus for the treatment <strong>of</strong> word order. These categories are essential for the further<br />

suggestions we shall make on word order variation <strong>in</strong> Natural Language Process<strong>in</strong>g.<br />

Chapter 6 concentrates more specifically on the computational treatment <strong>of</strong> word order<br />

variation. A major problem for computers is the structural and lexical ambiguity <strong>of</strong> language.<br />

Reliable grammar rules are crucial for their resolution. Concern<strong>in</strong>g word order, very few<br />

such rules can be formulated <strong>in</strong> the free word order language <strong>German</strong>, because some<br />

sequences may be less natural, but they can hardly be excluded. In section 6.1, we present a


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 15<br />

list <strong>of</strong> the few hard word order rules we have discovered. The discussion <strong>of</strong> theme, rheme<br />

and focus <strong>in</strong> 5.6 showed how important these categories are for the translation from one<br />

language <strong>in</strong>to another. In 6.2 and 6.3, we therefore suggest a mechanism to identify the three<br />

categories automatically dur<strong>in</strong>g the analysis <strong>of</strong> <strong>German</strong>.<br />

Section 6.4 conta<strong>in</strong>s a number <strong>of</strong> details which are not mandatory for a successful treatment<br />

<strong>of</strong> <strong>German</strong> <strong>in</strong> Natural Language Process<strong>in</strong>g, but which may help to understand the free word<br />

order phenomenon better. They concern the idiosyncrasies we identified when <strong>in</strong>vestigat<strong>in</strong>g<br />

the permutations <strong>of</strong> different modifier subclasses. In the follow<strong>in</strong>g section (6.5), we<br />

accomplish the prelim<strong>in</strong>ary version <strong>of</strong> the canonical form presented <strong>in</strong> 5.3.3. Through the<br />

<strong>in</strong>sertion <strong>of</strong> the flexible categories theme, rheme and focus, we achieve the result that<br />

generated word order differs, depend<strong>in</strong>g on the analysis <strong>of</strong> the source language. A<br />

constituent can have different positions, depend<strong>in</strong>g on whether it is thematic, rhematic, or<br />

neutral <strong>with</strong> respect to these categories.<br />

In 6.6, we present a preference rule for the resolution <strong>of</strong> some cases <strong>of</strong> PP attachment. It is<br />

based on our f<strong>in</strong>d<strong>in</strong>gs concern<strong>in</strong>g which constituent sequences are more natural than others.<br />

Section 6.7 gives an overview <strong>of</strong> the suggested modifier classes, and the 10 features that are<br />

necessary to treat them <strong>in</strong> NLP. These are the features used for the classification and cod<strong>in</strong>g<br />

<strong>of</strong> approximately four hundred adverbs listed <strong>in</strong> the appendix (8.2 and 8.3). In 6.8, we<br />

summarise the procedure presented <strong>in</strong> chapter 6.<br />

We end the thesis <strong>with</strong> conclud<strong>in</strong>g remarks (chapter 7). These comprise a brief summary <strong>of</strong><br />

our f<strong>in</strong>d<strong>in</strong>gs, an evaluation <strong>of</strong> the research carried out, as well as suggestions for future<br />

<strong>in</strong>vestigations, which could be carried out to complement the research presented here.<br />

The appendix consists <strong>of</strong> Hoberg's list <strong>of</strong> modifier classes (8.1), an extended alphabetical list<br />

<strong>of</strong> adverbs which we encoded us<strong>in</strong>g the features presented <strong>in</strong> 6.7 (8.2), and a third list, which<br />

presents the same modifiers ordered accord<strong>in</strong>g to position classes (8.3). Appendix 8.3 is<br />

particularly <strong>in</strong>terest<strong>in</strong>g, as it allows the comparison <strong>of</strong> the feature values <strong>of</strong> the adverbs<br />

belong<strong>in</strong>g to the same class, and to the same subgroups.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 16<br />

1.4. PROBLEMS OF GERMAN WORD ORDER DESCRIPTION<br />

Daß es <strong>in</strong> der deutschen Sprache ziemlich viele <strong>Variation</strong>smöglichkeiten im<br />

Bereich der Abfolge gibt, bedeutet nicht, daß ke<strong>in</strong>e Regeln vorhanden wären. Im<br />

Gegenteil, die Abfolge ist besonders kompliziert geregelt; denn für jede<br />

Folgevariante gibt es spezielle Bed<strong>in</strong>gungen, die sich <strong>in</strong> speziellen Regeln<br />

niederschlagen. (Engel, 1988: 303)<br />

In this section, we want to present the difficulties related to word order <strong>in</strong> general, and to<br />

modifier placement <strong>in</strong> particular, <strong>with</strong>out go<strong>in</strong>g <strong>in</strong>to too much detail at present.<br />

One difficulty regard<strong>in</strong>g word order description is that the data is generally fuzzy. There is<br />

no clear cut-<strong>of</strong>f po<strong>in</strong>t between grammatical and ungrammatical data, <strong>in</strong>stead there seems to<br />

be a graded borderl<strong>in</strong>e <strong>with</strong> different shades. Furthermore, spoken language differs from<br />

written language, and different concepts which are relevant for the description <strong>of</strong> word<br />

order<strong>in</strong>g can easily be mixed up, namely style, mean<strong>in</strong>g, scope, focus, and f<strong>in</strong>ally the<br />

semantic versus the syntactic classification <strong>of</strong> elements.<br />

When deal<strong>in</strong>g <strong>with</strong> the more specific task <strong>of</strong> modifier placement, we are confronted <strong>with</strong> a<br />

further problem, namely that the part-<strong>of</strong>-speech adverb is one <strong>of</strong> the most heterogeneous<br />

word classes <strong>in</strong> <strong>German</strong> (see also 2.2). This statement also seems to be true for other<br />

languages, as Schachter (1985: 20ff) and Givón (1984: 77ff) confirm <strong>in</strong> their universaloriented<br />

<strong>in</strong>vestigations. The differences between the large amount <strong>of</strong> different adverb classes<br />

make the description <strong>of</strong> their syntactic behaviour rather difficult.<br />

<strong>German</strong> word order variation usually changes the relative acceptability <strong>of</strong> sentences, <strong>with</strong>out<br />

creat<strong>in</strong>g either ungrammatical or completely natural utterances. The grammarian<br />

nevertheless has to decide which word order variation is grammatical and which one is not.<br />

As <strong>in</strong> other doma<strong>in</strong>s <strong>of</strong> l<strong>in</strong>guistic description native speakers differ <strong>in</strong> their judgements.<br />

Therefore, we have to rely on our own <strong>in</strong>tuition, sometimes backed up by other native<br />

speakers' judgements. To express the grades <strong>in</strong> the range <strong>of</strong> acceptability between<br />

grammatical and ungrammatical sentences we shall use the follow<strong>in</strong>g means:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 17<br />

? odd, or less acceptable than examples <strong>with</strong>out "?"<br />

?? less acceptable than "?" (to express grades <strong>of</strong> acceptability)<br />

# ungrammatical; can be considered acceptable if very strongly accentuated (contrast)<br />

! unacceptable for semantic reasons<br />

* def<strong>in</strong>itely ungrammatical<br />

<strong>German</strong> grammars only give a partial and rather unsatisfy<strong>in</strong>g explanation <strong>of</strong> where to put<br />

adverbials and what the differences between the positions are (see section 2.4). The general<br />

trend is to make a l<strong>in</strong>k between semantic classes and syntactic behaviour. Hoberg's<br />

positional classification <strong>of</strong> modifiers (1981: 106-131) shows, however, that the semantic and<br />

the syntactic classifications have to be partly dissociated to arrive at an appropriate<br />

description <strong>of</strong> adverb placement 13 . This makes the explanation <strong>of</strong> their syntactic behaviour<br />

even more difficult.<br />

A further problem is the question <strong>of</strong> what the differences between word order variations <strong>of</strong><br />

the same sentence, and subsets <strong>of</strong> these variations, are. Are they purely stylistical? In this<br />

case it would be <strong>in</strong>terest<strong>in</strong>g to f<strong>in</strong>d out what stylistical means. Do they <strong>in</strong>volve a different<br />

mean<strong>in</strong>g? Or does the difference consist <strong>of</strong> someth<strong>in</strong>g else, which would then have to be<br />

stated more precisely?<br />

20a<br />

20b<br />

20c<br />

Paul gab gestern der Frau das Buch.<br />

Paul gab der Frau das Buch gestern.<br />

Paul gab das Buch gestern der Frau.<br />

21a Paul gab nur der Frau das Buch.<br />

21b ! Paul gab der Frau das Buch nur. 14<br />

21c Paul gab das Buch nur der Frau.<br />

The sentences <strong>in</strong> 21 def<strong>in</strong>itely <strong>in</strong>volve different mean<strong>in</strong>gs, as nur is a degree adverb which<br />

changes its scope when moved. But what about 20? Accord<strong>in</strong>g to Engel (1988: 337) all<br />

modifiers have a scope, which means that the movement <strong>of</strong> gestern <strong>in</strong> 20 also <strong>in</strong>volves a<br />

change <strong>of</strong> mean<strong>in</strong>g. What then is this difference and how can we master it <strong>in</strong> order to be able<br />

13 Greenbaum (1969: 231) makes a similar statement about English.<br />

14 For reasons to be expla<strong>in</strong>ed <strong>in</strong> 4.4.4, the only possible place for the sentence focus is the verb gab which<br />

has to be stressed contrastively. 21b is ungrammatical, as it does not make sense to contrast nur geben<br />

s<strong>in</strong>ce there is noth<strong>in</strong>g to oppose it to. If it was Paul LIEH der Frau das Buch nur, 21b would be perfectly<br />

acceptable, as nur leihen can be imag<strong>in</strong>ed <strong>in</strong> opposition to schenken.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 18<br />

to translate the difference <strong>in</strong>to other languages? This is one <strong>of</strong> the ma<strong>in</strong> questions we shall<br />

focus on.<br />

What are the criteria to decide on the acceptability <strong>of</strong> sentences? Is 22a more acceptable than<br />

22b?:<br />

22a Hans traf sie gestern im K<strong>in</strong>o.<br />

22b # Hans traf gestern SIE im K<strong>in</strong>o. 15<br />

22b is a perfectly acceptable sentence but is contextually much more restricted than 22a<br />

because the personal pronoun sie has to be strongly stressed. It is very easy to mix up<br />

acceptability, focalization and scope, and this <strong>in</strong>creases the difficulty <strong>of</strong> word order<br />

description.<br />

Spoken and written <strong>German</strong> certa<strong>in</strong>ly differ as speakers <strong>of</strong>ten form their ideas while<br />

speak<strong>in</strong>g, whereas most written utterances are more thought-through 16 . Indeed, utterances<br />

like 23, <strong>with</strong> the temporal adverb beh<strong>in</strong>d the local verb complement, are likely to be heard<br />

but very unlikely to be read:<br />

23 ? Ich war im K<strong>in</strong>o gestern.<br />

We shall concentrate on our own <strong>in</strong>tuition <strong>of</strong> what contemporary written <strong>German</strong> is like,<br />

supported by some judgements <strong>of</strong> other native <strong>German</strong> speakers. For parts <strong>of</strong> the work, we<br />

shall also use a corpus <strong>of</strong> written <strong>German</strong> to verify our <strong>in</strong>tuition (cf. sections 6.7 and 6.8).<br />

We shall try to differentiate clearly between the different phenomena and to give answers to<br />

the questions asked <strong>in</strong> this <strong>in</strong>troductory chapter.<br />

15 Throughout this work, we capitalise words or syllables which have to be stressed strongly <strong>in</strong> order to<br />

make a sentence acceptable. 3b would be ungrammatical if the personal pronoun sie was not stressed<br />

contrastively.<br />

16 Further differences between spoken and written language will be discussed <strong>in</strong> sections 1.5.1.4 and 3.1.3.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 19<br />

1.5. WHY DESCRIBE WORD ORDER?<br />

There are three ma<strong>in</strong> doma<strong>in</strong>s to which the precise description <strong>of</strong> word order <strong>in</strong> general, and<br />

adverb placement <strong>in</strong> particular, could be a useful contribution: (a) the general knowledge <strong>of</strong><br />

how <strong>German</strong> and related languages work, (b) foreign language teach<strong>in</strong>g, for which rules-<strong>of</strong>thumb<br />

can be formulated, and (c) Natural Language Process<strong>in</strong>g. We are ma<strong>in</strong>ly <strong>in</strong>terested <strong>in</strong><br />

the last <strong>of</strong> these. Concern<strong>in</strong>g NLP, we shall dist<strong>in</strong>guish between problems aris<strong>in</strong>g whilst<br />

analys<strong>in</strong>g a text and whilst generat<strong>in</strong>g text. As Mach<strong>in</strong>e Translation (MT) is confronted <strong>with</strong><br />

both analysis and synthesis <strong>of</strong> sentences, it is an ideal application for which to discuss the<br />

use <strong>of</strong> our f<strong>in</strong>d<strong>in</strong>gs. However, we believe that our work is useful for other NLP applications,<br />

as well.<br />

When talk<strong>in</strong>g about Mach<strong>in</strong>e Translation, we refer to the rule-based approach as applied <strong>in</strong><br />

the MT systems Eurotra, CAT2, METAL, Rosetta and others 17 . Due to our previous work,<br />

we are most familiar <strong>with</strong> this approach. Furthermore, it is the only approach which relies on<br />

the precise formulation <strong>of</strong> l<strong>in</strong>guistic regularities. We do not want to exclude the fact, that our<br />

f<strong>in</strong>d<strong>in</strong>gs can be <strong>of</strong> use for other methods, such as the statistical approach. Whether, and how,<br />

our f<strong>in</strong>d<strong>in</strong>gs could be used for other approaches would have to be subject to a separate<br />

exam<strong>in</strong>ation.<br />

1.5.1 ANALYSIS OF GERMAN SENTENCES IN NLP<br />

Whilst automatically analys<strong>in</strong>g a given sentence we need to know which word order<br />

variations are ungrammatical, which ones are unlikely or marked, and which ones are likely<br />

to occur. We shall argue <strong>in</strong> chapter 4 that these three stages are grades <strong>of</strong> grammaticality.<br />

17 For an <strong>in</strong>troduction to Eurotra, METAL and Rosetta, see Hutch<strong>in</strong>s and Somers (1992: 239ff). For a more<br />

detailed presentation <strong>of</strong> Eurotra, see Mach<strong>in</strong>e Translation Volume 6, Numbers 2 and 3 (1991), and<br />

especially Bech/Maegaard/Nygaard (1991) and Durand et al. (1991). For an <strong>in</strong>troduction to CAT2, see<br />

Mesli (1991: 34ff) and Sharp (1989, 1993). Different approaches to MT, such as the rule-based,<br />

knowledge-based, example-based and statistics-based methods are presented <strong>in</strong> Ste<strong>in</strong>berger (1993) and<br />

Hutch<strong>in</strong>s/Somers (1992).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 20<br />

Through the precise description <strong>of</strong> possible and impossible word order variations we hope to<br />

contribute to the resolution <strong>of</strong> several problematic doma<strong>in</strong>s. These <strong>in</strong>clude the<br />

disambiguation <strong>of</strong> homonyms (1.5.1.1), resolution <strong>of</strong> prepositional phrase (PP) attachment<br />

(1.5.1.2), recognition <strong>of</strong> emphasis l<strong>in</strong>ked to word order (1.5.1.3), embedd<strong>in</strong>g <strong>of</strong> sentences <strong>in</strong><br />

the context (1.5.1.4), and the recognition <strong>of</strong> degree modifier scope (1.5.1.5). Concern<strong>in</strong>g PP<br />

attachment and degree modifier scope, we <strong>of</strong>fer only partial solutions <strong>in</strong> this work. However,<br />

as they are related to word order, we shall discuss them briefly for the sake <strong>of</strong> completeness.<br />

1.5.1.1 DISAMBIGUATION OF HOMONYMS<br />

Disambiguation <strong>of</strong> homonyms is necessary <strong>in</strong> both human and mach<strong>in</strong>e translation.<br />

In stratificational approaches to Mach<strong>in</strong>e Translation, such as Eurotra and CAT2, analysis is<br />

done <strong>in</strong> several steps. The first level is dedicated to morphological analysis, followed by the<br />

syntactic and the semantic analyses. When a word or a structure is ambiguous, the system<br />

generates two (or more) analysis objects accord<strong>in</strong>g to the different read<strong>in</strong>gs. The general<br />

procedure is to process all read<strong>in</strong>gs <strong>in</strong> parallel, until the ambiguity is resolved. Sometimes<br />

disambiguation is not feasible at all, and <strong>of</strong>ten it can be achieved at the semantic level only.<br />

As the parallel translation <strong>of</strong> several analyses typical for rule-based systems is timeconsum<strong>in</strong>g,<br />

it is important to provide tools to solve ambiguity as soon as possible.<br />

The two adverbs ehrlich <strong>in</strong> 24 and <strong>in</strong> 25 are word class homonyms, as they are semantically<br />

closely related adverbs which belong to two different sub-classes (for a further explanation<br />

<strong>of</strong> the term, see 2.3).<br />

Due to their different positional behaviour, they can already be recognised at the syntactic<br />

level, as ehrlich 1 <strong>in</strong> 24 has to precede the negational adverb nicht, whereas ehrlich 2 <strong>in</strong> 25<br />

must follow it:<br />

24 Peter kann ehrlich 2 nicht sprechen.<br />

(Honestly, Peter cannot speak)<br />

25 Peter kann nicht ehrlich 2 sprechen.<br />

(Peter cannot speak honestly)


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 21<br />

In MT systems <strong>of</strong> the afore-mentioned type, the first step <strong>of</strong> the analysis procedure would be<br />

to generate two analysis objects, one <strong>with</strong> ehrlich 1 and one <strong>with</strong> ehrlich 2 (26 and 27). In the<br />

next step, filters <strong>of</strong> the form: a certa<strong>in</strong> group <strong>of</strong> elements conta<strong>in</strong><strong>in</strong>g ehrlich 2 may not<br />

precede the negational modifier nicht would apply to the two objects. With their help, the<br />

system discovers which read<strong>in</strong>gs among 26a, 26b, 27a and 27b are the correct ones. These<br />

homonyms can thus be disambiguated, before enter<strong>in</strong>g the phase <strong>of</strong> semantic analysis, by<br />

check<strong>in</strong>g their position relative to the negational adverb nicht.<br />

26a Peter kann ehrlich 1 nicht sprechen.<br />

26b Peter kann ehrlich 2 nicht sprechen.<br />

27a Peter kann nicht ehrlich 2 sprechen.<br />

27b Peter kann nicht ehrlich 1 sprechen.<br />

Our goal is to identify a set <strong>of</strong> impossible sequences and to formulate them <strong>in</strong> as general a<br />

way as possible, <strong>in</strong> order to provide means to disambiguate homonyms, as well as cases <strong>of</strong><br />

structural ambiguity.<br />

The real ambiguity <strong>in</strong> 28, however, cannot be solved <strong>with</strong>out knowledge <strong>of</strong> the context:<br />

28 Peter kann ehrlich sprechen.<br />

(Honestly, Peter can speak; Peter can speak honestly)<br />

1.5.1.2 RESOLUTION OF PP-ATTACHMENT<br />

Another reason for ambiguity is the ambiguous attachment <strong>of</strong> prepositional phrases (PP).<br />

The problem <strong>with</strong> PPs is that very <strong>of</strong>ten it is difficult to decide whether they are verb<br />

arguments (prepositional objects), sentence modifiers (adverbials), or attributes or arguments<br />

to nouns, which thus belong to the preced<strong>in</strong>g noun phrase.<br />

We hope that the precise description <strong>of</strong> word order regularities is a useful contribution to the<br />

identification <strong>of</strong> the most likely attachment <strong>of</strong> PPs. The idea is to use our f<strong>in</strong>d<strong>in</strong>gs on which<br />

constituent sequences are either ungrammatical or unnatural. When one <strong>of</strong> the read<strong>in</strong>gs <strong>of</strong><br />

structurally ambiguous sentences is such an unlikely sequence, the other analysis should be<br />

preferred.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 22<br />

An example <strong>in</strong> which probability can help to choose one read<strong>in</strong>g over the other, is sentence<br />

29. The PP vor der Bank could either be analysed as an adjunct, namely a modifier <strong>of</strong> the<br />

nom<strong>in</strong>ative NP (29a), or as a sentence modifier express<strong>in</strong>g the location <strong>of</strong> the whole event<br />

expressed <strong>in</strong> 29 (29b). We shall see below (e.g. 6.2) that the read<strong>in</strong>g <strong>in</strong> 29b is unlikely and<br />

should thus be avoided if an alternative analysis, such as 29a, is available. In 30, the strength<br />

<strong>of</strong> the preference is easier to see, s<strong>in</strong>ce the attachment <strong>of</strong> the PP to the pronoun is not<br />

possible (30a). If we want vor der Bank to be a sentence modifier, we tend to use the order<br />

<strong>in</strong> 30c:<br />

29 Deshalb hat der Mann vor der Bank ihn e<strong>in</strong>fach ignoriert.<br />

29a Deshalb hat {der Mann vor der Bank} ihn e<strong>in</strong>fach ignoriert.<br />

29b ?? Deshalb hat {der Mann} {vor der Bank} ihn e<strong>in</strong>fach ignoriert.<br />

30a * Deshalb hat {er vor der Bank} ihn e<strong>in</strong>fach ignoriert.<br />

30b ?? Deshalb hat er {vor der Bank} ihn e<strong>in</strong>fach ignoriert.<br />

30c Deshalb hat er ihn {vor der Bank} e<strong>in</strong>fach ignoriert.<br />

Structural ambiguity as shown <strong>in</strong> 29 and 30 is not limited to PPs (cp 4.4.1) but it is much<br />

less <strong>of</strong> a problem for one-word modifiers. As the latter are our ma<strong>in</strong> concern, we shall treat<br />

PP attachment resolution only briefly (6.6). We believe, however, that the regularities and<br />

the means discussed <strong>in</strong> this work can be used to formulate more rules <strong>of</strong> the same k<strong>in</strong>d.<br />

1.5.1.3 RECOGNITION OF EMPHASIS<br />

The <strong>German</strong> variation <strong>in</strong> word order is sometimes used to stress a certa<strong>in</strong> element <strong>in</strong> the<br />

sentence. In 31a, the personal pronoun sie has to be stressed and is hence the focus <strong>of</strong> the<br />

sentence, whereas <strong>in</strong> 31b all five elements <strong>of</strong> the sentence can be stressed:<br />

31a Paul hat gestern SIE e<strong>in</strong>geladen.<br />

31b Paul hat sie gestern e<strong>in</strong>geladen.<br />

It is <strong>in</strong>deed a regularity that def<strong>in</strong>ite personal pronouns must not follow adverbs, unless they<br />

are strongly (contrastively) stressed. This, and other regularities, can be used to detect<br />

focus<strong>in</strong>g due to word order variation <strong>in</strong> <strong>German</strong> sentences.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 23<br />

Sentences <strong>with</strong> contrastive stress, such as 31a, are strongly restricted <strong>with</strong> respect to the<br />

possible contexts <strong>in</strong> which they can occur. For <strong>in</strong>stance, 31a is natural when follow<strong>in</strong>g<br />

context 32 but impossible after context 33:<br />

32 War Paul gestern bei Marie-Christ<strong>in</strong>e oder war sie bei ihm?<br />

31a' Paul hat gestern SIE e<strong>in</strong>geladen.<br />

33 Wann war Marie-Christ<strong>in</strong>e bei Paul?<br />

31a'' * Paul hat gestern SIE e<strong>in</strong>geladen.<br />

If possible, and depend<strong>in</strong>g on the language <strong>in</strong> which or from which it will be translated, the<br />

differences <strong>in</strong> mean<strong>in</strong>g and context compatibility should be equally rendered. 31a', for<br />

<strong>in</strong>stance, is equivalent to the Italian sentence 34a <strong>in</strong> which the stressed pronoun lei is used. A<br />

translation <strong>of</strong> 31a' (<strong>with</strong> context 32) by 34b <strong>in</strong>volv<strong>in</strong>g the unstressed pronoun la (also called<br />

clitic), and where the sentence focus is on <strong>in</strong>vitata would be wrong:<br />

34a Ieri, Paul ha <strong>in</strong>vitato LEI.<br />

34b * Ieri, Paul la ha <strong>in</strong>vitata. (as a translation <strong>of</strong> 31a' <strong>in</strong> the context 32)<br />

34 and 35 show that render<strong>in</strong>g the contrastive stress when translat<strong>in</strong>g <strong>in</strong>to Italian is essential<br />

to get a sentence which is correct <strong>in</strong> the context. It is impossible to use the clitic pronoun la<br />

when there is a not-clause (non-clause, <strong>in</strong> Italian) as this makes the contrast explicit (35a).<br />

Not render<strong>in</strong>g the Italian stress on lei <strong>in</strong> 34a <strong>in</strong>to <strong>German</strong> is less harmful. 36a shows that, if<br />

the context requires it, even the word order <strong>in</strong> 36a allows stress on sie, even if 36b would be<br />

more natural:<br />

35a * Ieri, Paul la ha <strong>in</strong>vitata, non Maria.<br />

35b Ieri, Paul ha <strong>in</strong>vitato lei, non Maria.<br />

36a<br />

36b<br />

Paul hat SIE gestern e<strong>in</strong>geladen, und nicht sie ihn.<br />

Paul hat gestern SIE e<strong>in</strong>geladen, und nicht sie ihn.<br />

One <strong>of</strong> the goals <strong>of</strong> our work is to describe stress due to word order variation appropriately,<br />

and to provide means to recognise, translate and render it <strong>in</strong> Mach<strong>in</strong>e Translation.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 24<br />

1.5.1.4 CONTEXTUAL EMBEDDING OF SENTENCES<br />

Other word order variations do not lead to obligatory contrastive stress, but rather to a<br />

change <strong>of</strong> the position <strong>of</strong> the sentence focus. The differences between 37a, 37b, 37c and 37d<br />

are more subtle than the ones <strong>in</strong> 31 and 34. Nevertheless, 37a to 37d have different uses and<br />

belong to different contexts:<br />

37a Paul gab gestern der Frau das Buch.<br />

37b Paul gab der Frau das Buch gestern.<br />

37c Paul gab das Buch gestern der Frau.<br />

37d Der Frau gab Paul das Buch gestern.<br />

...<br />

The ma<strong>in</strong> differences between the sentences <strong>in</strong> 37 concern the theme-rheme structure and<br />

functional sentence perspective (discussed <strong>in</strong> detail <strong>in</strong> 3.1 and 3.2). Languages differ <strong>in</strong> the<br />

degree to which they (have to) express these differences. In Russian, Czech and <strong>German</strong>, for<br />

<strong>in</strong>stance, word order largely depends on the context. <strong>Word</strong> order <strong>in</strong> more rigid languages,<br />

like French or English, depends less on context. Even these languages, however, render<br />

some <strong>of</strong> the differences, by choos<strong>in</strong>g different places for the adverb and other means (3.3.2):<br />

38a Er traf diesen Mann sehr <strong>of</strong>t.<br />

38b Il a rencontré cet homme très souvent.<br />

38c ? Il a très souvent rencontré cet homme. (as a translation <strong>of</strong> 38a)<br />

39a Er traf sehr <strong>of</strong>t diesen Mann.<br />

39b Il a très souvent rencontré cet homme.<br />

39c ? Il a rencontré cet homme très souvent. (as a translation <strong>of</strong> 39a)<br />

It is not useful to discuss one language <strong>in</strong> detail <strong>with</strong>out carry<strong>in</strong>g out research on a second<br />

language, as it does not help to recognise a focus, for <strong>in</strong>stance, <strong>with</strong>out render<strong>in</strong>g it <strong>in</strong> the<br />

translation. We therefore hope that such work will be carried out for further languages.<br />

Shorter descriptions already exist for English <strong>in</strong> Hajicová/Sgall/Skoumalová (1993) and<br />

Ste<strong>in</strong>berger (1992a).<br />

The goal <strong>of</strong> our work is (a) to identify the differences between the afore-mentioned sets <strong>of</strong><br />

sentences <strong>in</strong> <strong>German</strong>, and (b) to provide means for their automatic recognition and<br />

translation.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 25<br />

We want to address one problem related to discuss<strong>in</strong>g categories, such as focus and stress,<br />

when deal<strong>in</strong>g <strong>with</strong> written language. How can we recognise focus <strong>in</strong> written text? And what<br />

is its importance <strong>in</strong> written language? In order to recognise as well as to express focus <strong>in</strong><br />

written language, we have to rely upon our personal <strong>in</strong>tuition. Although there are no<br />

objective means to verify our judgement, we believe that there is little doubt, if any, that <strong>in</strong><br />

sentence 40 the personal pronoun has to be stressed contrastively for the sentence to be<br />

acceptable.<br />

40 Paul hat gestern SIE e<strong>in</strong>geladen.<br />

For sentences such as 41, we shall make statements similar to: It is likely that either gestern<br />

or e<strong>in</strong>geladen carry the sentence focus. This is also difficult to prove. The only means we<br />

have is to formulate possible contexts which require some phrases to be thematic and others<br />

to be rhematic. Then, we can judge (accord<strong>in</strong>g to our own <strong>in</strong>tuition) <strong>in</strong> which <strong>of</strong> these<br />

contexts our sentence sounds natural and <strong>in</strong> which ones it does not.<br />

41 Paul hat sie gestern e<strong>in</strong>geladen.<br />

However, our <strong>in</strong>tuition corresponds to what is generally said about the order <strong>of</strong> elements <strong>in</strong><br />

the <strong>German</strong> sentence s<strong>in</strong>ce the f<strong>in</strong>d<strong>in</strong>gs <strong>of</strong> the Prague School <strong>of</strong> l<strong>in</strong>guists. After hav<strong>in</strong>g<br />

carried out psychol<strong>in</strong>guistic experiments, Hajicová, Sgall and Skoumalová, for <strong>in</strong>stance, take<br />

it for granted that the sentence focus normally falls on the last words <strong>of</strong> the sentence (for<br />

further references, see sections 3.1 and 3.3):<br />

In technical texts (spoken or written), there is a strong tendency to arrange words so that the <strong>in</strong>tonation<br />

centre falls on the last word <strong>of</strong> the sentence (where it need not be phonetically manifested), <strong>with</strong> the<br />

exception, <strong>of</strong> course, <strong>of</strong> enclitic words (Hajicová/Sgall/Skoumalová, 1993: 179).<br />

L<strong>in</strong>guists and grammarians are becom<strong>in</strong>g more and more aware <strong>of</strong> the importance <strong>of</strong> large<br />

corpora to verify their <strong>in</strong>tuition. We acknowledge this necessity and therefore want to po<strong>in</strong>t<br />

out that written corpora are <strong>of</strong> no help to our research, as they do not carry <strong>in</strong>formation on<br />

stress etc. Spoken language, however, <strong>in</strong> which <strong>in</strong>formation on <strong>in</strong>tonation and stress is<br />

available, differs from written language <strong>in</strong> that the latter only has syntactical means to<br />

express the <strong>in</strong>tonation centre. For this reason, the syntax <strong>of</strong> written language has to be less


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 26<br />

ambiguous. Accord<strong>in</strong>g to Hajicová/Sgall/Skoumalová (1993: 179), the focalization <strong>of</strong> an<br />

element which is not at the end <strong>of</strong> the sentence is unlikely <strong>in</strong> written language. This proves<br />

our claim:<br />

In languages <strong>with</strong> a high degree <strong>of</strong> "free" word order (as <strong>in</strong> most Slavonic languages), a secondary<br />

position <strong>of</strong> the <strong>in</strong>tonation center is frequent only <strong>in</strong> spoken dialogues.<br />

We thus see no way <strong>of</strong> verify<strong>in</strong>g our <strong>in</strong>tuition <strong>in</strong> corpora. For the part <strong>of</strong> our work <strong>in</strong> which<br />

we could use corpora, namely the cod<strong>in</strong>g <strong>of</strong> modifiers, we made use <strong>of</strong> the Mannheim<br />

Korpus HK87 (see section 6.7.1). Concern<strong>in</strong>g the classification <strong>of</strong> modifiers <strong>in</strong> position<br />

classes (see section 5.3.2 and the appendix), our work is based on Hoberg (1981), who used<br />

corpora for her work as well. We hope and believe that, <strong>in</strong> spite <strong>of</strong> the amount <strong>of</strong> judgement<br />

based on <strong>in</strong>tuition, our research will be valid and useful for future applications and research.<br />

1.5.1.5 SCOPE OF DEGREE MODIFIERS<br />

Degree modifiers are a subgroup <strong>of</strong> the word class adverb (cf. section 2.2). Most degree<br />

modifiers can modify various word classes and they vary <strong>with</strong> respect to their position<br />

relative to the modified element. The movement <strong>of</strong> degree modifiers, from one position <strong>in</strong><br />

the sentence to another, changes the mean<strong>in</strong>g <strong>of</strong> the sentence considerably. Therefore, the<br />

description <strong>of</strong> the positional behaviour <strong>of</strong> this word group is particularly important.<br />

42a Paul hat nur der Frau e<strong>in</strong> Buch gegeben.<br />

42b (Paul only gave a book to the woman.)<br />

42c (It is only to the woman that Paul gave a book.)<br />

43a Paul hat der Frau nur e<strong>in</strong> Buch gegeben.<br />

43b (It is only a book that Paul gave to the woman.)<br />

Nur <strong>in</strong> 42a is likely to refer to the dative NP der Frau but it can also refer to the sentence as<br />

a whole. In 43a, nur is likely to refer to the accusative NP e<strong>in</strong> Buch, but can also refer to the<br />

whole sentence (Ste<strong>in</strong>berger, 1992b: 13ff). The English equivalent 42b is as ambiguous as<br />

42a. However, 42a and 43a could also be translated as 42c and 43b, which clearly show the<br />

change <strong>of</strong> mean<strong>in</strong>g l<strong>in</strong>ked to a change <strong>of</strong> position <strong>of</strong> nur.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 27<br />

Due to the considerable change <strong>of</strong> mean<strong>in</strong>g l<strong>in</strong>ked to the scope <strong>of</strong> elements, it is extremely<br />

important to recognise the possible scope <strong>of</strong> degree modifiers and hence the mean<strong>in</strong>gs <strong>of</strong> the<br />

sentence.<br />

The goal <strong>of</strong> this thesis is to suggest how to deal <strong>with</strong> word order variation under the<br />

exclusion <strong>of</strong> elements <strong>with</strong> scope 18 . Even though the dist<strong>in</strong>ction between scope and emphasis<br />

can be difficult, and many degree modifiers can also refer to the whole sentence, we shall<br />

nevertheless discuss degree modifier scope briefly (3.7, 6.7).<br />

1.5.2 SYNTHESIS OF GERMAN IN NLP<br />

<strong>German</strong> synthesis is probably easier than its analysis but it is by no means trivial. A good<br />

synthesis <strong>of</strong> <strong>German</strong> sentences does not only <strong>in</strong>clude the generation <strong>of</strong> correct word order,<br />

which is a much more complicated task than it may seem at a first glance, but it also<br />

<strong>in</strong>volves the render<strong>in</strong>g <strong>of</strong> more f<strong>in</strong>e-tuned <strong>in</strong>formation, which we obta<strong>in</strong> by analys<strong>in</strong>g a<br />

source language <strong>in</strong> the way described <strong>in</strong> 1.5.1.<br />

We dist<strong>in</strong>guish two types <strong>of</strong> problems which have to be solved when generat<strong>in</strong>g <strong>German</strong><br />

sentences. The first group comprises requirements which are obligatory for the production <strong>of</strong><br />

grammatical output. The second is related to less basic requirements, which can be neglected<br />

<strong>with</strong>out generat<strong>in</strong>g ungrammatical output, but are important to produce a natural flow <strong>of</strong><br />

sentences.<br />

One reason why a translation can result <strong>in</strong> ungrammaticality is because languages differ<br />

essentially <strong>in</strong> syntactic structure. A further common problem is that, <strong>in</strong> spite <strong>of</strong> <strong>German</strong> word<br />

order freedom, some sequences <strong>of</strong> phrases are impossible. This is particularly relevant for<br />

the cumulation <strong>of</strong> modifiers, as these have as <strong>of</strong> yet not been studied <strong>in</strong> so much detail. And<br />

18 Degree modifiers can generally refer to different phrases <strong>in</strong> the sentence. To avoid too many analyses a<br />

preference mechanism can be applied which choses among the several possible read<strong>in</strong>gs. For a suggestion<br />

on how to recognise, translate and render degree modifier scope see Ste<strong>in</strong>berger (1992b).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 28<br />

f<strong>in</strong>ally, translated sentences may not correspond to the source language sentence, because<br />

scope is rendered <strong>in</strong>correctly.<br />

A natural flow <strong>of</strong> sentences, and thus a coherent and eligible text, can be reached by<br />

embedd<strong>in</strong>g sentences appropriately, and by render<strong>in</strong>g contrastive stress. We shall discuss<br />

these basic requirements, as well as the desirable improvements, <strong>in</strong> the follow<strong>in</strong>g sections.<br />

Before that, we want to po<strong>in</strong>t out some basic facts concern<strong>in</strong>g the structure <strong>of</strong> <strong>German</strong>,<br />

which we will then take for granted, <strong>with</strong>out describ<strong>in</strong>g them explicitly <strong>in</strong> our discussion.<br />

This section will also serve to def<strong>in</strong>e some terms generally used to describe <strong>German</strong><br />

structure.<br />

The follow<strong>in</strong>g examples show that the order <strong>of</strong> words generally has to be changed when<br />

translat<strong>in</strong>g from one language <strong>in</strong>to another:<br />

44a I shall probably see him tomorrow.<br />

44b * Je peut-être verrai le dema<strong>in</strong>.<br />

44c * Ich werde vielleicht sehen ihn morgen.<br />

44b'<br />

44c'<br />

Je le verrai peut-être dema<strong>in</strong>.<br />

Ich werde ihn vielleicht morgen sehen.<br />

The structure <strong>of</strong> <strong>German</strong> declarative sentences is generally described as be<strong>in</strong>g composed <strong>of</strong><br />

Vorfeld, Mittelfeld, Nachfeld and the two verb positions which, together, form the<br />

Satzklammer (verbal bracket) (e.g. Engel, 1988: 303ff). In declarative ma<strong>in</strong> clauses (45a)<br />

<strong>with</strong> one simple verb, the verb follows one element and is thus <strong>in</strong> second position (verbsecond,<br />

or V2-clause). In subclauses (45b), it is <strong>in</strong> f<strong>in</strong>al position (verb-letzt/verb-last, or VLclauses).<br />

When the verbal part <strong>of</strong> the sentence <strong>in</strong> declarative ma<strong>in</strong> clauses consists <strong>of</strong> more<br />

than one part, one <strong>of</strong> the parts is <strong>in</strong> V2 position and the other one <strong>in</strong> VL position. The latter<br />

can be the case when auxiliaries or separable verb prefixes occur. Together, these two<br />

groups <strong>of</strong> verbal elements form the verbal bracket:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 29<br />

VORFELD V2 MITTELFELD VL NACHFELD<br />

45a T<strong>in</strong>a sprach gestern mit Wolf.<br />

[Ich weiß, daß ...]<br />

45b T<strong>in</strong>a gestern mit Wolf sprach.<br />

45c T<strong>in</strong>a hat gestern mit Wolf gesprochen.<br />

45d T<strong>in</strong>a hat mit Wolf gesprochen gestern.<br />

I__________________________________I<br />

verbal bracket<br />

The Vorfeld (pre-field) is the position preced<strong>in</strong>g the verb <strong>in</strong> second position, Mittelfeld<br />

(middle-field) is what is embraced by the verbal bracket, and Nachfeld (after-field) is what<br />

follows the verb <strong>in</strong> f<strong>in</strong>al position.<br />

As we have seen, the position <strong>of</strong> the verb is fixed rigidly. It has to be <strong>in</strong> second position <strong>in</strong><br />

declarative ma<strong>in</strong> clauses, and <strong>in</strong> end position <strong>in</strong> subclauses. This has not always been the<br />

case (Lehmann, 1978c: 410ff). Up until the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> the 16th century, the verb tended to<br />

be <strong>in</strong> second position, even <strong>in</strong> subord<strong>in</strong>ate clauses. The shift to the sentence-f<strong>in</strong>al position<br />

was an act <strong>of</strong> will by writers who tried to imitate the superior language Lat<strong>in</strong>. For a<br />

collection <strong>of</strong> orig<strong>in</strong>al quotations by philosophers and grammarians who shaped todays'<br />

<strong>German</strong> see Scaglione (1981: 29ff):<br />

The OV pattern <strong>of</strong> subord<strong>in</strong>ate clauses was adopted as the regular construction by learned writers<br />

around the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> the sixteenth century on the basis <strong>of</strong> Lat<strong>in</strong>, though it was not wholly absent <strong>in</strong><br />

earlier forms <strong>of</strong> <strong>German</strong>. Gradually the tw<strong>of</strong>old pattern<strong>in</strong>g <strong>of</strong> VO order <strong>in</strong> <strong>in</strong>dependent clauses, OV <strong>in</strong><br />

dependent, was <strong>in</strong>stalled, so that it is now regular <strong>in</strong> the standard written language. (Lehmann, 1978c:<br />

410)<br />

Another basic requirement for <strong>German</strong> is that every verb-second clause generally has to<br />

have one element <strong>in</strong> the Vorfeld 19 . The other elements, <strong>with</strong> the exception <strong>of</strong> the verb(s),<br />

belong <strong>in</strong> the Mittelfeld. In 5.5, we shall discuss how to choose which element goes <strong>in</strong>to the<br />

Vorfeld. The position beh<strong>in</strong>d VL is less important for written <strong>German</strong> 20 .<br />

19 Cases <strong>in</strong> which more than one element fills the Vorfeld are discussed <strong>in</strong> Altmann (1981: 285ff)<br />

20 For details on the role <strong>of</strong> the Nachfeld <strong>in</strong> <strong>German</strong> sentences, see Engel (1970: 83ff) and Altmann (1981).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 30<br />

It should be clear now why the English and French sentences <strong>in</strong> 46 and 47 cannot be<br />

translated <strong>in</strong>to <strong>German</strong> <strong>with</strong>out reorder<strong>in</strong>g. The English verb is followed by two constituents,<br />

and the French verb is preceded by two constituents. Both sequences are not possible <strong>in</strong><br />

<strong>German</strong>:<br />

46 I shall probably see him tomorrow.<br />

* Ich werde vielleicht sehen ihn morgen.<br />

47 Je le verrai peut-être dema<strong>in</strong>.<br />

* Ich ihn werde sehen vielleicht morgen.<br />

It is obvious that the order <strong>of</strong> elements has to be changed dur<strong>in</strong>g the translation process. The<br />

reorder<strong>in</strong>g <strong>with</strong><strong>in</strong> these basic limits generally does not pose a problem, as the rules are<br />

transparent and clear-cut. More important questions are which element should go <strong>in</strong> the<br />

Vorfeld, and <strong>in</strong> which order should the constituents <strong>in</strong> the Mittelfeld follow. These are the<br />

questions we focused on <strong>in</strong> our research.<br />

1.5.2.1 BASIC ORDERING OF CONSTITUENTS<br />

As phrases have to be reordered we have to ask, accord<strong>in</strong>g to which pr<strong>in</strong>ciples and rules<br />

should this be done. Both <strong>German</strong> language learners and Mach<strong>in</strong>e Translation systems are<br />

likely to produce ungrammatical sentences like 1a and 1b:<br />

48a # Ich habe e<strong>in</strong> Geschenk gestern zum Geburtstag bekommen.<br />

48b * Ich habe e<strong>in</strong> Geschenk zum Geburtstag gestern bekommen.<br />

48c Ich habe gestern e<strong>in</strong> Geschenk zum Geburtstag bekommen.<br />

Ungrammatical sequences such as <strong>in</strong> 48 can be avoided by obey<strong>in</strong>g some basic rules <strong>of</strong><br />

thumb. The rule neglected <strong>in</strong> 48a, for <strong>in</strong>stance, is that adverbs, <strong>with</strong> the exception <strong>of</strong> manner<br />

adverbs 21 , should not follow <strong>in</strong>def<strong>in</strong>ite noun phrases. In 48b, the additional rule is violated<br />

that the same group <strong>of</strong> modifiers should not follow prepositional objects.<br />

21 We shall def<strong>in</strong>e adverbs and their subclasses <strong>in</strong> 2.2.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 31<br />

1.5.2.2 CUMULATION OF MODIFIERS<br />

Although word order rules for arguments and for a subset <strong>of</strong> modifiers are well-known, little<br />

research has been carried out on the position <strong>of</strong> most modifier classes. The cumulation <strong>of</strong><br />

modifiers is a problem which has been neglected by most grammar writers (see 2.4). We<br />

<strong>in</strong>tend to provide means to avoid sentences such as 49a:<br />

49a * Er fuhr gleichfalls dennoch nach München.<br />

49b Er fuhr dennoch gleichfalls nach München.<br />

As we shall see later, modifier classes are so heterogeneous and diverse that it is very<br />

difficult to formulate rules <strong>of</strong> thumb like the one suggested <strong>in</strong> the previous section. For this<br />

reason, we shall suggest a canonical form which helps to generate a correct sequence <strong>of</strong> all<br />

constituents <strong>in</strong>clud<strong>in</strong>g a large number <strong>of</strong> modifier classes (sections 5.2 and 5.3). In chapters<br />

3 and 4, we shall give at least a partial explanation <strong>of</strong> why some sequences are<br />

ungrammatical. These are also the chapters which will probably result <strong>in</strong> be<strong>in</strong>g the most<br />

useful for language learners and teachers.<br />

1.5.2.3 CORRECT SCOPE<br />

Accord<strong>in</strong>g to an unambiguous and correct recognition <strong>of</strong> scope as described <strong>in</strong> 1.5.1.5, we<br />

also have to render scope appropriately <strong>in</strong> synthesis. The French ne que (only) <strong>in</strong> 50a refers<br />

to hier and not to Marie, and therefore translation 50b is correct whereas 50c is wrong .<br />

50a Je n'ai vu Marie qu'hier.<br />

50b Ich habe Marie erst gestern gesehen.<br />

50c * Ich habe nur Marie gestern gesehen.<br />

The fact that ne que is translated as erst when referr<strong>in</strong>g to gestern but as nur when referr<strong>in</strong>g<br />

to Marie is secondary here. Even though it is important that the correct translation is chosen,<br />

this is not directly related to word order, but rather to the class <strong>of</strong> the modified element. The<br />

translation <strong>of</strong> ne que as either erst or nur is not an isolated case. In Ste<strong>in</strong>berger (1992b: 31),<br />

for <strong>in</strong>stance, we show that the modifier gleich has to be rendered by either <strong>of</strong> the five<br />

translations: equivalent, same, identical, immediate or equal. Which <strong>of</strong> the translations is


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 32<br />

correct depends on whether gleich is an adjective (the first two options) or an adverb (last<br />

three options), and on the word class they modify.<br />

1.5.2.4 SENTENCE EMBEDDING<br />

Similar to what we said about embedd<strong>in</strong>g a sentence translated from <strong>German</strong> <strong>in</strong>to another<br />

language, we should also take care that a text translated <strong>in</strong>to <strong>German</strong> fits <strong>in</strong>to the context<br />

naturally. If the sentences 51a and 52a are translated by 51b, 51c and 52b, the sentence focus<br />

<strong>in</strong> the Spanish and the <strong>German</strong> sentences are on the constituents correspond<strong>in</strong>g to each other.<br />

In the translations 51d and 52c, they are not:<br />

51a A Maria le voy a dar el libro.<br />

51b Maria werde ich das Buch geben.<br />

51c Ich werde Maria das Buch geben.<br />

51d ? Ich werde das Buch Maria geben.<br />

52a Le voy a dar el libro a Maria.<br />

52b Ich werde das Buch Maria geben.<br />

52c ? Ich werde Maria das Buch geben.<br />

PP attachment and the correct choice <strong>of</strong> homonyms are no problems for synthesis. Once we<br />

know what a PP refers to - a difficulty which must be solved dur<strong>in</strong>g analysis - it is easy to<br />

f<strong>in</strong>d its appropriate place <strong>in</strong> the target sentence. The same applies for words which are<br />

homonyms <strong>in</strong> the source language. Once we know which homonym read<strong>in</strong>g is correct, the<br />

word can be translated exactly like every other word, as source and target language words<br />

are l<strong>in</strong>ked through a bil<strong>in</strong>gual dictionary.<br />

1.5.3 FOREIGN LANGUAGE TEACHING<br />

The f<strong>in</strong>e-tuned differences between word order variations <strong>in</strong> general and modifier placement<br />

<strong>in</strong> particular, are a major problem for foreigners speak<strong>in</strong>g the <strong>German</strong> language. Our own<br />

teach<strong>in</strong>g experience has shown that even students <strong>with</strong> a very broad vocabulary and a sound<br />

knowledge <strong>of</strong> the grammar make strik<strong>in</strong>g errors l<strong>in</strong>ked to <strong>in</strong>correct adverb position<strong>in</strong>g.<br />

Grammars have little to say about this (see 2.4) so that even most zealous study is no<br />

remedy.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 33<br />

We hope that our f<strong>in</strong>d<strong>in</strong>gs are a contribution to improve foreign language teach<strong>in</strong>g on two<br />

levels. Firstly, we want to expla<strong>in</strong> more clearly than <strong>in</strong> exist<strong>in</strong>g grammar books how free<br />

<strong>German</strong> word order is, and what its limits are. Secondly, we <strong>in</strong>tend to provide concrete<br />

means <strong>of</strong> avoid<strong>in</strong>g the noticeable errors which, <strong>in</strong> spite <strong>of</strong> some foreigners perfect<br />

pronunciation, make it so <strong>of</strong>ten obvious that a foreigner is speak<strong>in</strong>g.<br />

In the chapters 6 and 7, we concentrate on how to deal <strong>with</strong> word order variation <strong>in</strong> Natural<br />

Language Process<strong>in</strong>g. These chapters may not be very important for language learners and<br />

teachers, but they <strong>in</strong>clude however data which is <strong>in</strong>terest<strong>in</strong>g for language acquisition, <strong>in</strong> that<br />

they describe which sentences are natural and which ones are contextually more restricted<br />

(or marked).<br />

The canonical form which we develop <strong>in</strong> these two chapters comprises over eighty position<br />

classes and therefore, it may be tiresome to use for humans. The section about compulsory<br />

orders, however, can be seen as a list <strong>of</strong> useful rules <strong>of</strong> thumb to avoid severe mistakes<br />

l<strong>in</strong>ked to word order.<br />

It is certa<strong>in</strong>ly an advantage for language learners if they are made aware <strong>of</strong> the <strong>in</strong>teraction <strong>of</strong><br />

preference rules described <strong>in</strong> chapter 3. The understand<strong>in</strong>g <strong>of</strong> some <strong>of</strong> these preferences<br />

probably requires sound l<strong>in</strong>guistic knowledge which the average language learner does not<br />

have. We believe, however, that the mere statement <strong>of</strong> the fact how they <strong>in</strong>teract (chapter 4)<br />

can be <strong>of</strong> help for foreigners, whose language has more fixed word order.<br />

Last but not least, it may be helpful for language teachers to be made aware <strong>of</strong> the fuzzy<br />

borderl<strong>in</strong>e which exists between the word classes described <strong>in</strong> chapter 2. As we show <strong>in</strong><br />

2.2.1 and 2.2.4, grammar books provide contradictory def<strong>in</strong>itions <strong>of</strong> the word class adverb. It<br />

may help language students to get over the problems they have <strong>with</strong> <strong>German</strong> when they see<br />

that the terms they confuse have their roots <strong>in</strong> vague def<strong>in</strong>itions.<br />

We believe that our most important contribution to language teach<strong>in</strong>g and learn<strong>in</strong>g is the<br />

general explanation and the formulation <strong>of</strong> rules <strong>of</strong> thumb.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 34<br />

2. COMPLEMENT, MODIFIER, ADVERB AND ADVERB<br />

SUBTYPES<br />

In this chapter, we want to discuss the terms we shall use. One dist<strong>in</strong>ction we make is<br />

between modifiers (Angaben) and complements (Ergänzungen) (2.1), another one between<br />

different adverb subtypes (2.2, 2.3). We shall also give an overview <strong>of</strong> what grammar books<br />

say about adverb placement (2.4), and some statistical <strong>in</strong>formation on adverbs <strong>in</strong> <strong>German</strong><br />

(2.5).<br />

We want to po<strong>in</strong>t out that the term adverb designates a word class, whereas modifier and<br />

complement name different functions or relations <strong>with</strong><strong>in</strong> the sentence. Adverbs, NPs, PPs<br />

and clauses can thus function as adverbials (1 to 4) or as verb complements (5 to 8).<br />

1 Harold spielt dort S<strong>of</strong>tball.<br />

2 Harold spielt den ganzen Tag S<strong>of</strong>tball.<br />

3 Harold spielt mit Vergnügen S<strong>of</strong>tball.<br />

4 Harold spielt S<strong>of</strong>tball, weil es ihm Vergnügen bereitet.<br />

5 Paul wohnt dort.<br />

6 Paul liest das ganze Buch.<br />

7 Paul spricht mit Christoph.<br />

8 Paul weiß, daß S<strong>of</strong>tball Harold Vergnügen bereitet.<br />

We do not agree <strong>with</strong> Givón (1984: 77), who uses adverb <strong>in</strong> the sense <strong>of</strong> adverbial:<br />

Unlike the categories noun, verb and adjective, adverbs are a rather mixed lexical class, semantically,<br />

morphologically and syntactically. Many adverbs are full sentential constructions rather than one-word<br />

lexical items [...]. But even s<strong>in</strong>gle adverbial words ...<br />

Another area where some grammarians lack consistency is the dist<strong>in</strong>ction between some<br />

parts-<strong>of</strong>-speech. The reason for the confusion is the badly drawn borderl<strong>in</strong>e between the<br />

word class adverb and other particles. Schulz and Griesbach (1980: 211), for <strong>in</strong>stance, use<br />

the expression "Adverb als Konjunktion", and Liebsch/Doer<strong>in</strong>g (1976: 135) mix up several<br />

word classes:<br />

Das Adverb wird zur Präposition, wenn es sich mit Nomen verb<strong>in</strong>det. [...] Adverbien können als<br />

Konjunktionen auftreten, aber auch als Adverbialbestimmungen mit Satzgliedwert.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 35<br />

Although we understand how this confusion can arise, we shall only use the word classes for<br />

the doma<strong>in</strong> <strong>in</strong> which they are def<strong>in</strong>ed. If an adverb takes over the function <strong>of</strong> a conjunction,<br />

we shall say it is a conjunction, which is homonymous to an adverb.<br />

The dist<strong>in</strong>ction between modifiers and complements is not trivial. We shall discuss this<br />

subject briefly before say<strong>in</strong>g more about adverbs and their subclasses <strong>in</strong> 2.2.<br />

2.1. MODIFIERS (ANGABEN) VS. COMPLEMENTS<br />

(ERGÄNZUNGEN))<br />

In Deutsche Grammatik (1988: 862ff), Engel def<strong>in</strong>es the two complementary notions as<br />

follows:<br />

Ergänzung: "syntaktisches Glied, das se<strong>in</strong>em Regens subklassenspezifisch zugeordnet ist; kann<br />

obligatorisch oder fakultativ se<strong>in</strong>; kommt vor allem bei Verb, Nomen, Adjektiv, Pronomen vor<br />

(Gegensatz: Angabe)".<br />

Angabe: "syntaktisches Glied, das beliebigen Elementen e<strong>in</strong>er regierenden Wortklasse (vor allem<br />

Verben, Nom<strong>in</strong>a, Adjektiven) aspezifisch zugeordnet werden kann, meist fakultativ (Gegensatz:<br />

Ergänzung)".<br />

However, <strong>in</strong> the same grammar book (1988: 219) he cuts down the use <strong>of</strong> sentence modifiers<br />

say<strong>in</strong>g that they are never compulsory ("es gibt ke<strong>in</strong>e obligatorischen Satzangaben"). At<br />

least at sentence level, modifiers are thus always optional.<br />

Bußmann's (1983: 125) def<strong>in</strong>ition slightly differs. Accord<strong>in</strong>g to her, complements are:<br />

"Von der Valenz sprachlicher Ausdrücke (Verben, Adjektive, Nomen) geforderte syntaktische Elemente<br />

(im Unterschied zu den beliebig h<strong>in</strong>zufügbaren oder weglaßbaren Angaben)"<br />

Bußmann thus def<strong>in</strong>es complements as obligatory elements. Engel expla<strong>in</strong>s his differ<strong>in</strong>g<br />

view <strong>with</strong> a considerable number <strong>of</strong> complements, which may or may not be omitted,<br />

<strong>with</strong>out result<strong>in</strong>g <strong>in</strong> a dist<strong>in</strong>ct verb mean<strong>in</strong>g. Most <strong>of</strong> these optional complements are<br />

accusative objects:<br />

9 Peter ißt (e<strong>in</strong> Steak).<br />

10 Helga liest (die Schmutzigen Hände von Jean Paul Sartre).<br />

11 Holga schwimmt (auf der Isar nach München).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 36<br />

Accord<strong>in</strong>g to Engel (1988: 183), the optional phrases <strong>in</strong> brackets must be considered as<br />

complements, because they cannot be comb<strong>in</strong>ed <strong>with</strong> other verbs:<br />

12 * Der Kaktus blüht e<strong>in</strong> Steak.<br />

13 * Ihm graut die schmutzigen Hände von Jean Paul Sartre.<br />

14 * Holga liest e<strong>in</strong> Buch nach München.<br />

Engel dist<strong>in</strong>guishes eleven subtypes <strong>of</strong> verb complements, which can be realised by<br />

pronouns, def<strong>in</strong>ite NPs, <strong>in</strong>def<strong>in</strong>ite NPs, NPs accompanied by a modifier or a complement,<br />

proper names etc (Engel, 1988: 185). The subtypes are (1988: 185-198):<br />

- Nom<strong>in</strong>ativergänzung (subject): Der Mann liest.<br />

- Akkusativergänzung: Er sieht die Frau.<br />

- Genitivergänzung: Er bedarf des Buches.<br />

- Dativergänzung: Er folgt ihr.<br />

- Präpositivergänzung: Wir verlassen uns auf dich.<br />

- Situativergänzung: Sie wohnt <strong>in</strong> München.<br />

- Direktivergänzung: Sie fährt nach Paris.<br />

- Expansivergänzung: Er war e<strong>in</strong>en Kilometer gelaufen.<br />

- Nom<strong>in</strong>alergänzung: Me<strong>in</strong> Bruder ist Beamter.<br />

Er hat sich als Betrüger erwiesen.<br />

- Adjektivalergänzung: Sie wird krank.<br />

Sie benimmt sich anständig.<br />

Sie war wie wild.<br />

Sie war es auch.<br />

Sie war eben so.<br />

- Verbativergänzung: Peter läßt die Puppen tanzen.<br />

Es heißt, er wolle zurücktreten.<br />

Engel's argumentation is problematic, as modifiers can be selectionally restricted as well.<br />

Modifiers <strong>in</strong>dicat<strong>in</strong>g an <strong>in</strong>strument (Instrumentalangabe), a goal (F<strong>in</strong>alangabe) and an<br />

iteration (Iterativangabe), for <strong>in</strong>stance, cannot be freely comb<strong>in</strong>ed <strong>with</strong> any verb:<br />

15 * Etwas fällt mir auf + F<strong>in</strong>alangabe (mit diesem Ziel/zu diesem Zweck).<br />

16 * Vor dem Buch graut ihm + Instrumentalangabe (mit der Lupe/mit dem Hammer).<br />

17 * Er wurde <strong>in</strong> München geboren + Iterativangabe (wiederum, erneut).<br />

18 * Er las sehr.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 37<br />

It seems that these selectional restrictions are <strong>of</strong> a semantic nature and that they can be<br />

overcome by choos<strong>in</strong>g a semantically different word <strong>of</strong> the same word class (Er las schnell).<br />

On the other hand, syntactic valency certa<strong>in</strong>ly is <strong>of</strong> semantic or pragmatic orig<strong>in</strong> as well. The<br />

verb geben, for <strong>in</strong>stance, has three complements, as the action <strong>of</strong> giv<strong>in</strong>g concerns humans<br />

<strong>with</strong> respect to three aspects: AGENT, (given) OBJECT and RECEIVER 22 . These are<br />

realised by the nom<strong>in</strong>ative, accusative and dative complements. Although time and place can<br />

be relevant <strong>in</strong> certa<strong>in</strong> situations, they are usually less dom<strong>in</strong>ant than the roles represented by<br />

the complements. The verb wohnen, however, calls for local <strong>in</strong>formation, and this is the<br />

reason we would say that the phrase referr<strong>in</strong>g to the PLACE is a complement, rather than an<br />

adverbial:<br />

The two criteria (a) obligator<strong>in</strong>ess and (b) the ability to occur <strong>with</strong> every verb, do not<br />

provide a clear-cut borderl<strong>in</strong>e between the concepts <strong>of</strong> complement and modifier. As for<br />

obligator<strong>in</strong>ess, it seems to us that the dist<strong>in</strong>ction is rather one <strong>of</strong> degree, <strong>in</strong> the sense that<br />

some complements can be omitted more easily than others. The phrases express<strong>in</strong>g location<br />

<strong>with</strong> the verb wohnen, for <strong>in</strong>stance, can hardly be dropped (19). The only exceptions are rare<br />

cases such as 19c. The accusative object <strong>of</strong> the verb essen, however, is omitted quite<br />

frequently, <strong>with</strong>out caus<strong>in</strong>g odd or ungrammatical sentences (20). One could thus argue that<br />

optional complements have less the quality <strong>of</strong> complements, and more the quality <strong>of</strong><br />

modifiers, than obligatory complements.<br />

19a * Ralf wohnt.<br />

19b Ralf wohnt <strong>in</strong> München.<br />

19c Ralf wohnt modern.<br />

20a<br />

20b<br />

Christoph ißt.<br />

Christoph ißt e<strong>in</strong>en Apfel.<br />

We shall not follow this question any further, because it is not <strong>of</strong> immediate relevance to us.<br />

We shall use the dichotomous dist<strong>in</strong>ction because <strong>in</strong>tuitively it is quite clear, and we did not<br />

encounter any problems us<strong>in</strong>g it. Furthermore, Engel and Hoberg, to whose work we refer a<br />

22 Throughout this work, we write semantic roles <strong>in</strong> capitals.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 38<br />

lot, use it. However, it is not essential for the relevance <strong>of</strong> our work what the exact<br />

borderl<strong>in</strong>e between the two concepts is. If we dropped the dist<strong>in</strong>ction completely, we would<br />

have to rename the categories which we use but even this would not put <strong>in</strong>to question our<br />

core f<strong>in</strong>d<strong>in</strong>gs.<br />

We thus use the notion modifier (Angabe) as equivalent to the term optional adverbials<br />

which can be realised by the follow<strong>in</strong>g categories (Bußmann, 1983: 9):<br />

- pure adverbs (heute, dort)<br />

- de-adjectival adverbs (schön, neu) 23<br />

- pronom<strong>in</strong>al adverbs (Pronom<strong>in</strong>aladverbien) (dar<strong>in</strong>, deshalb)<br />

- PPs (auf dem Tisch)<br />

- Genitive NPs (e<strong>in</strong>es Morgens)<br />

- Accusative NPs (den ganzen Tag)<br />

- adverbial clauses (Adverbialsätze) (Er folgte ihr, woh<strong>in</strong> sie wollte.) 24<br />

Some l<strong>in</strong>guists do not <strong>in</strong>clude toners (Abtönungs- or Modalpartikel: ja, doch, halt,<br />

eigentlich) and degree modifiers (Gradadverbien: sehr, nur) <strong>in</strong> the word class adverb (cf.<br />

2.2). We <strong>in</strong>clude toners <strong>in</strong> our research and <strong>in</strong>clude degree modifiers if they refer to the<br />

sentence, as opposed to a phrase <strong>of</strong> the sentence. Most degree modifiers can refer to the<br />

whole sentence as well as to parts <strong>of</strong> it, so that nearly the whole group is <strong>in</strong>cluded.<br />

The only realization <strong>of</strong> modifiers we exclude <strong>in</strong> our work are adverbial clauses. Among the<br />

other realizations, we focus on one-word modifiers, as they are a limited set (cf. 1.2).<br />

23 An example is schnell <strong>in</strong> Carol<strong>in</strong>e liest schnell. For a more detailed discussion <strong>of</strong> these elements which<br />

are sometimes called Adjektivadverbien (Bußmann, 1983: 7) or adverbiale Adjektive (Eisenberg, 1989:<br />

220), see section 2.2.3.3.<br />

24 Bußmann does not mention the realization <strong>of</strong> adverbials by <strong>in</strong>f<strong>in</strong>itival clauses (Er setzte sich, um sich<br />

auszuruhen) which we shall not treat here either.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 39<br />

2.2. DEFINITION AND CLASSIFICATION OF ADVERBS<br />

Die Adverbien gehören zum Widerspenstigsten und Unübersichtlichsten, was die<br />

deutsche Grammatik zu bieten hat. Kaum e<strong>in</strong>e andere Kategorie wird ähnlich<br />

une<strong>in</strong>heitlich abgegrenzt und <strong>in</strong>tern nach so verschiedenen Gesichtspunkten<br />

gegliedert, e<strong>in</strong>e den Gesamtbereich des Adverbs überzeugend ordnende<br />

Darstellung fehlt bisher. Auch im Term<strong>in</strong>ologischen besteht e<strong>in</strong> Wirrwarr, der<br />

se<strong>in</strong>esgleichen sucht. (Eisenberg, 1989: 204)<br />

For our purposes, we shall dist<strong>in</strong>guish three ma<strong>in</strong> modifier classes (2.3):<br />

a) existimatorische Angaben<br />

b) situative Angaben<br />

c) modale Angaben<br />

which will be subdivided later. Further terms for word classes used <strong>in</strong> our research are toners<br />

(Abtönungspartikeln) and degree adverbs (Gradpartikeln). In addition to these terms, other<br />

ways <strong>of</strong> classify<strong>in</strong>g adverbs have been suggested <strong>in</strong> l<strong>in</strong>guistic literature and grammar books.<br />

We shall also mention these, so as to give an overview.<br />

Adverbs are particularly difficult to def<strong>in</strong>e, as they share a lot <strong>of</strong> properties <strong>with</strong> other word<br />

classes. There is no set <strong>of</strong> dist<strong>in</strong>ctive features which would be sufficient to dist<strong>in</strong>guish this<br />

heterogeneous word class from all other word classes. This fact expla<strong>in</strong>s why the term<br />

adverb is <strong>of</strong>ten used as a waste paper basket for other l<strong>in</strong>guistic elements, which cannot be<br />

categorised appropriately. Sommerfeldt and Starke (1988: 159) therefore state that the term<br />

adverb is reduced to be<strong>in</strong>g a "Sammelwortart". In The Structure <strong>of</strong> <strong>German</strong>, Fox (1990: 160)<br />

comes to the same conclusion:<br />

In practice, however, the class <strong>of</strong> adverb covers a very wide area, and any word which cannot otherwise<br />

be accomodated tends to be called an adverb by default.<br />

This confusion is not restricted to <strong>German</strong>. French and English grammar writers seem to<br />

have the same problem:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 40<br />

La classe des adverbes est un des plus remarquables "fourre-tout" de la grammaire traditionelle. A2 sa<br />

décharge, on relèvera que l'analyse de ces unités est extrèmement délicate. (Jolivet, 1982: 387) 25<br />

Grammarians are not <strong>in</strong> general agreement on what to <strong>in</strong>clude among sentence modifiers or sentence<br />

adverbs. (Greenbaum, 1969: 2)<br />

In his description <strong>of</strong> universal part-<strong>of</strong>-speech systems, Schachter (1985: 20ff) questions<br />

[...] whether there is sufficient similarity among the various types <strong>of</strong> adverbs that may be recognised <strong>in</strong><br />

a language to justify their be<strong>in</strong>g assigned to a s<strong>in</strong>gle parts-<strong>of</strong>-speech class.<br />

The difficulties experienced by grammar writers are reflected <strong>in</strong> the handl<strong>in</strong>g <strong>of</strong> modifiers <strong>in</strong><br />

l<strong>in</strong>guistic theory, such as Head-Driven Phrase Structure Grammar (HPSG) and X-Bar<br />

Theory:<br />

In HPSG theory, as <strong>in</strong> l<strong>in</strong>guistic theory at large, the analysis <strong>of</strong> adjuncts is at a very primitive stage.<br />

(Pollard/Sag, 1987: 158)<br />

Remarkably absent is the lexical category preposition-postposition [-N, -V] (as well as adverb). This<br />

fact emphasises the deficient treatment <strong>of</strong> this lexical category <strong>in</strong> current X-bar theory. (Zonnefeld,<br />

1991: 147)<br />

For an overview <strong>of</strong> the different classifications <strong>of</strong> the word class systems see<br />

Sommerfeldt/Starke (1988: 58ff).<br />

To specify what we understand by adverb and modifier, we shall have to look at how it has<br />

been def<strong>in</strong>ed <strong>in</strong> l<strong>in</strong>guistic literature (2.2.1 and 2.2.2), as well as what the borderl<strong>in</strong>es <strong>with</strong> the<br />

related word classes particle, conjunction, preposition and adjective are (2.2.3). We shall<br />

evaluate the results <strong>of</strong> the comparison <strong>in</strong> a conclusion, <strong>in</strong> which we def<strong>in</strong>e the term adverb<br />

(2.2.4).<br />

2.2.1. DIFFERENT DEFINITIONS OF THE ADVERB<br />

The word adverb suggests that these words accompany verbs and thus modify their mean<strong>in</strong>g.<br />

Admoni (1970: 198, 142), Liebsch/Doer<strong>in</strong>g (1976: 134f), Erben (1972: 166) and<br />

Dreyer/Schmitt (1985: 218) thus see the primary use <strong>of</strong> adverbs as verb modifiers, and all<br />

25 The class adverb is one <strong>of</strong> the biggest catchalls <strong>in</strong> traditional grammar. It must be said <strong>in</strong> its defence that<br />

its analysis is extremely delicate.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 41<br />

other uses as secondary uses. Lyons (1971: 331), however, emphasises that the Lat<strong>in</strong> word<br />

"verbum" is more general than the <strong>German</strong> or English word verb, and that it means "word".<br />

Accord<strong>in</strong>gly, Erben (1972: 166) translates "ad-verbium" as "Bei-wort" or Nebenwort and<br />

thus concludes that adverbs are dependent words <strong>with</strong> a support<strong>in</strong>g ("dienend") function.<br />

This criterion cannot be taken <strong>in</strong> a very strict sense, as adverbs can realise the same functions<br />

as, for <strong>in</strong>stance, prepositional objects (21; PO) and a few <strong>of</strong> them even have a valency <strong>with</strong><br />

dependent complements (22; Duden, 1984: 346 and Dreyer/Schmitt, 1985: 220ff):<br />

21 Er freute sich darüber (über den Besuch; PO).<br />

22 Er ist mir <strong>in</strong> der Musik voraus.<br />

A dist<strong>in</strong>ctive feature <strong>of</strong> adverbs <strong>in</strong> which most 26 sources agree is that adverbs are elements<br />

that cannot be <strong>in</strong>flected. They share this characteristic <strong>with</strong> prepositions, conjunctions and a<br />

few smaller word classes (Engel, 1988: 18). We shall list further characteristics that<br />

grammars use to def<strong>in</strong>e and describe adverbs.<br />

Eisenberg (1989: 204) mentions that <strong>in</strong> most cases adverbs situate other entities <strong>in</strong> a local,<br />

temporal or modal way. Adverbs are an open class because there are morphological means to<br />

multiply the otherwise restricted number. Generally, they have a lexical mean<strong>in</strong>g and they<br />

are no-function words.<br />

Engel (1988: 749) states that adverbs can fill the preverbal position (Vorfeld) <strong>in</strong> Verb-<br />

Second (V2) sentences and that they can either be answers to w-questions (wer, wen, wann,<br />

warum etc), or they are w-elements themselves. They can be modifiers, be equivalent to a<br />

sentence, be verb complements, or attributes <strong>of</strong> nouns. Engel gives an allegedly exhaustive<br />

list <strong>of</strong> 150 simple adverbs (i.e. derivation or compound<strong>in</strong>g has not been applied) 27 .<br />

26 Some sources po<strong>in</strong>t out that adverbs used to have <strong>in</strong>flectional end<strong>in</strong>gs <strong>in</strong> Middle-High-<strong>German</strong>, and Erben<br />

(1972: 57f) comb<strong>in</strong>es adverbs and adjectives <strong>in</strong> a common class <strong>of</strong> elements which is <strong>in</strong>flectable <strong>with</strong> a<br />

subgroup hav<strong>in</strong>g the feature that they are not <strong>in</strong>flectable (see also below <strong>in</strong> this section and 2.2.3.3).<br />

27 The small number can be expla<strong>in</strong>ed by the strict application <strong>of</strong> tests, for <strong>in</strong>stance (a) adverbs must be able<br />

to fill the Vorfeld position, which excludes adverbs like m<strong>in</strong>destens and (b) they must not be <strong>in</strong>flectable,<br />

which excludes schnell, bereitwillig and others (Engel, 1988: 754).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 42<br />

Bußmann (1984: 8f) def<strong>in</strong>es adverbs as modifiers <strong>of</strong> verbs, adjectives, adverbials and entire<br />

sentences. They occur freely (frei vorkommende Adverbien; abends, bergauf, gern) or are<br />

pro-forms <strong>of</strong> prepositional objects or adverbials (Pronom<strong>in</strong>aladverbien: darauf, dorth<strong>in</strong>,<br />

deswegen). The subgroup <strong>of</strong> sentence adverbs expresses the speakers' judgement, which<br />

refers to the whole sentence.<br />

Liebsch & Doer<strong>in</strong>g (1976: 135ff) state that adverbs do not have comparative forms, although<br />

they admit that there are exceptions to this.<br />

Helbig/Buscha (1988: 337ff) call attention to the fact that, although adverbs cannot be<br />

<strong>in</strong>flected, some <strong>of</strong> them do have comparative forms (23). Furthermore, adverbs have three<br />

possible uses: adverbial (24, 25), predicative (26) and attributive (27):<br />

23 <strong>of</strong>t - öfter - am öftesten<br />

gut - besser - am besten<br />

24 Der Mann arbeitet dort. (adverbial use)<br />

25 Der dort arbeitende Mann ist Kameruner. (adverbial use)<br />

26 Der Mann ist dort. (predicative use)<br />

27 Der Mann dort arbeitet den ganzen Tag. (attributive use)<br />

In the Duden (1984: 90f), adverbs are def<strong>in</strong>ed as non-<strong>in</strong>flectables specify<strong>in</strong>g circumstances<br />

(nähere Umstände). They can also be used as attributes to nouns.<br />

Erben (1972: 166ff) subsumes adverbs as well as adjectives <strong>in</strong> one class called<br />

charakterisierendes Beiwort. He gives it this name, because adverbs and adjectives<br />

characterise other words, which they accompany. The elements <strong>of</strong> this class have<br />

comparative forms and can be comb<strong>in</strong>ed <strong>with</strong> degree modifiers. As they have comparative<br />

end<strong>in</strong>gs, Erben (1972: 57ff) lists these by-words together <strong>with</strong> nouns and verbs under the<br />

header <strong>in</strong>flectables.<br />

Accord<strong>in</strong>g to Hammond (1981: 185ff) adverbs cannot only be used to extend or modify<br />

verbs, adjectives, other adverbs, prepositions, adverbial phrases and conjunctions, but also to<br />

form compound verbs (herumfahren) and to ask questions (warum).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 43<br />

Jung (1971: 318) reveals that some adverbs diachronically are not only hardened genitive<br />

forms <strong>of</strong> NPs (flugs, abends), but also dative (zuweilen, vergebens) or accusative forms<br />

(weg, allzeit), which can no longer be recognised as such. Some adverbs were formed <strong>in</strong><br />

analogy to genitive forms, <strong>with</strong>out actually be<strong>in</strong>g genitives (nachts, neuerd<strong>in</strong>gs).<br />

Most sources, but not all <strong>of</strong> them, accept the term adverb as the name for a word category,<br />

and see it <strong>in</strong> opposition to the relational term adverbial (cf. Eisenberg, 1989: 204).<br />

2.2.2. SEMANTIC CLASSIFICATION<br />

All grammars classify adverbs semantically. The Duden (1984: 346ff) for <strong>in</strong>stance<br />

dist<strong>in</strong>guishes the follow<strong>in</strong>g semantic sub-groups to express:<br />

- local <strong>in</strong>formation (dort, l<strong>in</strong>ks, wor<strong>in</strong>, h<strong>in</strong>, herum, fort)<br />

- temporal <strong>in</strong>formation (heute, wann, seither, vorerst, manchmal, erstens)<br />

- modal <strong>in</strong>formation<br />

- manner ("Art und Weise") (quality) (gerne, geradeaus, nebenher, umsonst, so)<br />

- degree, amount (quantity)<br />

- high degree (zu, wieviel, besonders, sehr, gar)<br />

- limited degree, approximation, uncerta<strong>in</strong>ty (be<strong>in</strong>ahe, etwa, halbwegs, rund, etwa,<br />

höchstens, über)<br />

- extension (zudem, außerdem, weiterh<strong>in</strong>, ebenfalls)<br />

- restriction, adversative (<strong>in</strong>s<strong>of</strong>ern, allerd<strong>in</strong>gs, immerh<strong>in</strong>, weder - noch, vielmehr)<br />

- emphasis (ausgerechnet, genau, nur, sogar, besonders)<br />

- judgement (vielleicht, gewiß, leider, h<strong>of</strong>fentlich, nicht)<br />

- toners (personal feel<strong>in</strong>gs like astonishment, anger, doubt ...) (aber, ja, doch, denn,<br />

mal, nur, schon, vielleicht)<br />

- <strong>in</strong>formation about the cause<br />

- cause and consequence (weshalb, darum, folglich, me<strong>in</strong>etwegen)<br />

- <strong>in</strong>strument (dadurch, hiermit, womit)<br />

- condition (notfalls, sonst, strenggenommen)<br />

- concession (dennoch, nichtsdestoweniger, doch)<br />

- f<strong>in</strong>ality, <strong>in</strong>tention (dazu, hierfür, warum)<br />

Due to the vastness <strong>of</strong> the semantic field covered by adverbs, the classifications by different<br />

authors vary considerably. Engel for <strong>in</strong>stance (1988: 750ff) dist<strong>in</strong>guishes:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 44<br />

- local (l<strong>in</strong>ks, daneben, wo)<br />

- directive (dah<strong>in</strong>, davor, woh<strong>in</strong>)<br />

- temporal (anfangs, neulich, wann)<br />

- "komitativ" (describ<strong>in</strong>g the surround<strong>in</strong>g circumstances) (damit, womit)<br />

- causal (daher, me<strong>in</strong>etwegen, warum)<br />

- conditional (dann)<br />

- <strong>in</strong>strumental (damit, dadurch, womit)<br />

- f<strong>in</strong>al (dafür, dazu, w<strong>of</strong>ür)<br />

- modificational (gerne, heimlicherweise, wie)<br />

- grad<strong>in</strong>g (sehr, teilweise, <strong>in</strong>wiefern)<br />

Two word classes which by some grammarians are <strong>in</strong>cluded <strong>in</strong> the group <strong>of</strong> adverbs (Duden,<br />

1984) and by others are not (Heidolph et. al., 1981; Altmann, 1976), deserve to be discussed<br />

more specifically: degree modifiers and toners.<br />

Degree modifiers like nur, e<strong>in</strong>zig, auch, sogar, schon, erst etc can modify the whole<br />

sentence, as well as parts <strong>of</strong> it. The modified constituent is called the scope <strong>of</strong> the modifier<br />

(cf. 3.7). Accord<strong>in</strong>g to Engel (1988: 764) they cannot be <strong>in</strong>flected or compared, cannot fill<br />

the Vorfeld position on their own and can be positioned between conjunctions and Vorfeld<br />

elements. Altmann's (1978, 1976: 1f) def<strong>in</strong>ition is more semantically oriented and does not<br />

<strong>in</strong>clude the condition, that the degree modifiers must not fill the Vorfeld position on their<br />

own. It therefore covers wenigstens, be<strong>in</strong>ahe and others, which Engel classifies as<br />

Rangierpartikel and Modalpartikel.<br />

Semantically, they rank the modified constituent (<strong>in</strong> the follow<strong>in</strong>g example: wir) on a scale<br />

(Altmann, 1976: 122ff). Sogar wir has the mean<strong>in</strong>g, that besides us there were others who<br />

saw Sartre, and that these others were more likely to see him. Sogar can also be seen as an<br />

expression <strong>of</strong> the speaker's personal judgement: we put the likelihood <strong>of</strong> see<strong>in</strong>g Sartre low<br />

on the scale, so that one would not have expected that we would see him.<br />

28 Sogar wir haben J.-P. Sartre gesehen.<br />

Degree modifiers have word class homonyms among conjunctions, toners and other adverbs<br />

(Eisenberg, 1989: 207). We use the term word class homonym to express the fact, that some


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 45<br />

lexemes have two or more related mean<strong>in</strong>gs, whilst be<strong>in</strong>g members <strong>of</strong> different word classes.<br />

The term is a relative one, because degree modifiers, conjunctions and toners, for <strong>in</strong>stance,<br />

are <strong>in</strong> the same word class, the non-<strong>in</strong>flectables, but they are <strong>in</strong> different sub-classes. We<br />

shall use the term strictly for all sets <strong>of</strong> homonymous elements, which belong to at least two<br />

different modifier subclasses.<br />

Toners (Abtönungspartikeln) are a very limited group <strong>of</strong> particles like aber, doch, wohl, eben<br />

etc which cannot fill the Vorfeld position on their own (Thurmair, 1989: 25ff). Engel (1988:<br />

19) lists the further characteristics: (a) they cannot be asked for and (b) they cannot be<br />

negated 28 . Toners have also been called Modalpartikeln (Thurmair, 1988), Würzwörter or<br />

Füllwörter (Duden, 1984: 351), Färbewörter (Sommerfeldt/Starke, 1988: 163),<br />

Fülladverbien and Gefühlsträger (Pelz, 1963: 104).<br />

Toners cannot refer to a part <strong>of</strong> the sentence, but they are always at the same level as the<br />

sentence itself (Eisenberg (1989, 206): Sie "treten als Konstituente neben den ganzen Satz";<br />

see also section 2.3). They cannot fill the Vorfeld and they are rarely stressed<br />

(Sommerfeldt/Starke, 1988: 164) 29 . Semantically, they relate the contents <strong>of</strong> the sentence to<br />

the speaker's attitude or judgement. The mean<strong>in</strong>g <strong>of</strong> these typically <strong>German</strong> particles is <strong>of</strong>ten<br />

difficult to grasp. They are <strong>in</strong>deed <strong>of</strong>ten omitted when translat<strong>in</strong>g <strong>in</strong>to other languages (Pelz,<br />

1963: 102ff; Barth, 1961: 102-104). Weydt (1969: 73) estimates that one out <strong>of</strong> ten, or even<br />

only one out <strong>of</strong> a hundred toners, are translated <strong>in</strong>to French.<br />

29 Männer s<strong>in</strong>d eben so.<br />

30 T<strong>in</strong>a ist aber klug.<br />

The toner eben <strong>in</strong> 29 expresses someth<strong>in</strong>g like "I always knew that" and aber <strong>in</strong> 30 expresses<br />

a feel<strong>in</strong>g <strong>of</strong> astonishment comb<strong>in</strong>ed <strong>with</strong> admiration. For a semantic description <strong>of</strong> these<br />

elements see Engel (1988: 231-238). A peculiar characteristic <strong>of</strong> toners is that every s<strong>in</strong>gle<br />

28 We do not want to add Engel's claim here that they never cumulate, as he himself gives several examples<br />

which show more than one toner together (1989: 327).<br />

29 Accord<strong>in</strong>g to our own judgement, toners can never be stressed (see the cod<strong>in</strong>g <strong>in</strong> appendix 8.3, as well as<br />

the discussion <strong>in</strong> section 6.7)


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 46<br />

one <strong>of</strong> them has at least one word class homonym. Aber, for <strong>in</strong>stance, can also be a<br />

conjunction, and eben can be a (temporal) adverb as well as an adjective.<br />

2.2.3. ADVERBS AND RELATED WORD CLASSES<br />

2.2.3.1. ADVERBS AND PARTICLES<br />

Grosse Unterschiede [<strong>in</strong> der Klassifikation] gibt es auch im H<strong>in</strong>blick auf<br />

Gliederung und Zuordnung der Pronomen, Adverbien und Partikeln, auch des<br />

Artikels. (Sommerfeldt/Starke, 1988: 58)<br />

Duden (1984: 91) sees adverbs as one out <strong>of</strong> three subgroups <strong>of</strong> particles (adverbs,<br />

prepositions and conjunctions) and opposes this supergroup to the other ma<strong>in</strong> categories<br />

verb, noun, adjective, article/pronoun and <strong>in</strong>terjection.<br />

verb<br />

noun<br />

adjective<br />

article / pronoun<br />

<strong>in</strong>terjection<br />

particles<br />

- adverbs<br />

- prepositions<br />

- conjunctions<br />

Heidolph's et. al. classification (1981: 683), however, subsumes adverbs, prepositions and<br />

conjunctions, <strong>in</strong> addition to the particles, under the header non-<strong>in</strong>flectables<br />

(Nichtflektierbare). They see particles as the subgroup <strong>of</strong> toners (Abtönungspartikel), degree<br />

modifiers and others.<br />

non-<strong>in</strong>flectables<br />

- adverbs<br />

- prepositions<br />

- conjunctions<br />

- particles - toners<br />

- degree modifiers<br />

- ...<br />

Altmann (1976: 3) <strong>in</strong>cludes all word classes, which cannot be <strong>in</strong>flected, as direct daughters<br />

<strong>in</strong> the super-class particle:<br />

particles<br />

- adverb particles<br />

- conjunction particles<br />

- preposition particles<br />

- toners (Abtönungspartikeln)<br />

- degree modifiers (Gradpartikeln)<br />

- ...


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 47<br />

Sommerfeldt and Starke (1988: 58ff and 159ff) see adverbs as the superclass to pure<br />

adverbs, modal words and particles, <strong>with</strong> particles be<strong>in</strong>g elements which cannot fill the<br />

Vorfeld on their own.<br />

For a critical discussion <strong>of</strong> the different ways <strong>of</strong> classify<strong>in</strong>g adverbs and particles see<br />

Eisenberg (1989: 204ff).<br />

Another related word class which should be mentioned briefly here are <strong>in</strong>terjections like ah,<br />

oh, äh, hm, miaou, hoppla etc. Engel (1988: 772ff) classifies them as a subgroup <strong>of</strong> the<br />

elements which can be equivalent to a sentence (Satzäquivalente). The specific feature <strong>of</strong><br />

this word class is that they are syntactically isolated elements (Duden, 1984: 90f). This<br />

characteristic is sufficient to dist<strong>in</strong>guish them from adverbs, some <strong>of</strong> which can also<br />

constitute one-word sentences (31), but which always can be part <strong>of</strong> a longer sentence.<br />

Isolated elements, such as äh, do not pose a problem <strong>with</strong> respect to word order. We shall<br />

leave them aside <strong>in</strong> our research.<br />

31 A: Liebt sie He<strong>in</strong>z?<br />

B: Vermutlich.<br />

In the next section, we try to shed some light on the borderl<strong>in</strong>es between adverbs on one<br />

hand side, and prepositions, conjunctions and adjectives on the other.<br />

2.2.3.2. ADVERBS VS. CONJUNCTIONS AND PREPOSITIONS<br />

A few words can be prepositions as well as adverbs (e.g. abseits, l<strong>in</strong>ks, oberhalb), and<br />

elements <strong>of</strong> both word classes can denote similar th<strong>in</strong>gs, such as a position. For Liebsch &<br />

Doer<strong>in</strong>g (1976: 135) the difference between the two is that adverbs refer to the verbal<br />

statement, whereas prepositions refer to nouns. As we have seen that adverbs can also refer<br />

to nouns, Eisenberg's (1989: 206) dist<strong>in</strong>ctive feature seems more appropriate: prepositions<br />

call for one or even several cases (32a, 33a), whereas adverbs do not (32b, 33b):<br />

32a Sie lebt abseits des Dorfes. (abseits + genitive)<br />

32b Sie lebt abseits (vom Dorf). (abseits + optional von-PP)<br />

33a Er kleidete sich entsprechend der Vorschrift. (entsprechend + genitive)<br />

33b Er kleidete sich entsprechend. (entsprechend + ∅)


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 48<br />

Eisenberg's dist<strong>in</strong>ction thus allows for words such as abseits to be both prepositions and<br />

adverbs, depend<strong>in</strong>g on whether the complement is realised or not. As we want to avoid<br />

claim<strong>in</strong>g that elements belong to two classes, we suggest a different borderl<strong>in</strong>e, by allow<strong>in</strong>g<br />

adverbs to have valency: The two word classes differ <strong>in</strong> that the arguments <strong>of</strong> prepositions<br />

are compulsory, whereas adverbs can occur on their own. Accord<strong>in</strong>g to this criterion, abseits<br />

(32) and entsprechend (33) are adverbs <strong>with</strong> valency, but für is a preposition, as it cannot<br />

occur on its own (34):<br />

34a Paul gab Harold das Buch für Jock.<br />

34b * Paul gab Harold das Buch für.<br />

Examples <strong>of</strong> elements that can be adverbs or conjunctions are da and seitdem (Duden, 1984:<br />

380). The difference between these two classes are that conjunctions l<strong>in</strong>k more than one<br />

word or word group <strong>of</strong> a certa<strong>in</strong> category to each other (35a, 36a), whereas adverbs appear<br />

<strong>with</strong> one word or word group (35b, 36b; Eisenberg, 1989: 206):<br />

35a Da (=weil) sie ke<strong>in</strong>e Zeit hatte, konnte sie nicht kommen. (conj)<br />

35b Da (=dort) liegt e<strong>in</strong> Buch. (adv)<br />

36a Seitdem Peter nach Köln gezogen ist, haben wir nichts mehr von ihm gehört. (conj)<br />

36b Peter ist nach Köln gezogen. Seitdem haben wir nichts mehr von ihm gehört. (adv)<br />

A subgroup <strong>of</strong> adverbs, namely the <strong>in</strong>terrogative adverbs (Fragewörter,<br />

Interrogativadverbien, also called Relativadverbien (Duden, 1984: 355)) like weshalb,<br />

superficially resembles conjunctions: both can <strong>in</strong>troduce subord<strong>in</strong>ate clauses. The difference<br />

between them is that the adverb is a Stellungsglied <strong>of</strong> the subord<strong>in</strong>ate clause (that is to say a<br />

constituent belong<strong>in</strong>g to it), whereas the conjunction is not (Eisenberg, 1989: 338ff). ob <strong>in</strong><br />

37a cannot be replaced <strong>in</strong> 37b by any element hav<strong>in</strong>g the same or a similar mean<strong>in</strong>g.<br />

Weshalb <strong>in</strong> 38a, however, corresponds to deshalb, or to aus diesem Grund, <strong>in</strong> 38b. Weshalb<br />

asks for a constituent, which is part <strong>of</strong> the subord<strong>in</strong>ate clause, whereas ob is not part <strong>of</strong> the<br />

subord<strong>in</strong>ate clause:<br />

37a Er fragt, ob du das machst. (conj)<br />

7b * Du machst das wie/so/aus diesem Grund/... .<br />

38a<br />

38b<br />

Er fragt, weshalb du das machst. (adv)<br />

Du machst das deshalb / aus diesem Grund.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 49<br />

Interrogative adverbs always have a non-<strong>in</strong>terrogative equivalent (weshalb - deshalb etc),<br />

which Helbig/Buscha (1988: 341) call conjunctional adverbs. Although <strong>in</strong>terrogative<br />

adverbs normally <strong>in</strong>troduce the subord<strong>in</strong>ate clause, they can stand <strong>in</strong> the same position as<br />

their non-<strong>in</strong>terrogative counterpart. The result<strong>in</strong>g sentences, which conta<strong>in</strong> <strong>in</strong>terrogative<br />

adverbs, are generally called echo-questions (Stechow/Sternefeld, 1988: 354; examples from<br />

Eisenberg, 1989: 341):<br />

39a Wann/Morgen trifft Luise den Herrn Direktor (?).<br />

39b Luise trifft wann/morgen den Herrn Direktor (?).<br />

39c Luise trifft den Herrn Direktor wann/morgen (?).<br />

2.2.3.3. ADVERBS VS. ADJECTIVES<br />

[...] adjectives refer to properties <strong>of</strong> <strong>in</strong>dividuals, and adverbs refer to properties <strong>of</strong><br />

actions. (Kempson, 1977: 13)<br />

A further dist<strong>in</strong>ction to make is the one between adverbs and adjectives. Adverbs and<br />

adjectives can modify nouns (40), and both categories can be used predicatively (41):<br />

40a Der Mann dort spricht sehr gut Deutsch.<br />

40b Der große Mann spricht sehr gut Deutsch.<br />

41a Der Mann ist dort.<br />

41b Der Mann ist <strong>in</strong>telligent.<br />

The difference between der Mann dort and der große Mann is that the adverb is not <strong>in</strong>flected<br />

and it follows the noun, whereas the adjective precedes the noun and is <strong>in</strong>flected. A case<br />

where it is more difficult to dist<strong>in</strong>guish between adverbs and adjectives is the non-<strong>in</strong>flected<br />

adjective follow<strong>in</strong>g the noun <strong>in</strong> antiquated <strong>German</strong> (42), and <strong>in</strong> advertis<strong>in</strong>g language (43).<br />

Engel (1988: 612f) states that these constructions have already left the limited doma<strong>in</strong> <strong>of</strong> this<br />

sub-language and hazards the guess that they are grow<strong>in</strong>g more and more popular:<br />

42 E<strong>in</strong> Rösle<strong>in</strong> rot steht auf der Weide.<br />

43a Campari bitter<br />

43b Benz<strong>in</strong> bleifrei<br />

43c Putenschenkel bratfertig<br />

43d Fahrspass total<br />

Engel's exclusive def<strong>in</strong>ition <strong>of</strong> adverbs as words which must never be <strong>in</strong>flected, clearly<br />

identifies the non-<strong>in</strong>flected elements <strong>in</strong> 42 and 43 as adjectives. Our def<strong>in</strong>ition will be


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 50<br />

broader (see the conclusion below) and therefore we have to give another criterion for<br />

classify<strong>in</strong>g them as adjectives. The number <strong>of</strong> adverbs which can post-specify nouns is very<br />

limited, and a feature <strong>in</strong>dicat<strong>in</strong>g this syntactic behaviour is necessary. We shall therefore say<br />

that all post-nom<strong>in</strong>al, non-<strong>in</strong>flected elements, which are not adverbs hav<strong>in</strong>g the explicit<br />

feature that they can follow nouns, are adjectives.<br />

A further difficulty for the dist<strong>in</strong>ction between these two word classes is that most adjectives<br />

can be used adverbially when not <strong>in</strong>flected.<br />

44 Hans schreit laut.<br />

45 Hans verwischt sorgfältig die Spuren.<br />

As a result <strong>of</strong> this similarity between the two word classes, they have been classified <strong>in</strong><br />

different ways. Russon/Russon (1978: 31f), Admoni (1970: 146f), Helbig/Buscha (1988:<br />

338f) and Dreyer/Schmitt (1985: 218ff) see them as adverbs. Hammond (1981: 176ff)<br />

ma<strong>in</strong>ta<strong>in</strong>s they are adjectives used as adverbs. Eisenberg (1989: 219ff), Engel (1988: 558ff),<br />

Bußmann (1983: 7), Gl<strong>in</strong>z (1971b: 231) and the Duden (1984: 581ff) classify them as<br />

adjectives, giv<strong>in</strong>g them the additional name <strong>of</strong> Adjektivadverb and Satzadjektiv.<br />

Sommerfeldt and Starke (1988: 58ff), who give an overview <strong>of</strong> different word class systems<br />

<strong>in</strong> l<strong>in</strong>guistic literature, also conclude that nowadays these elements are widely accepted as<br />

adjectives, which have the function <strong>of</strong> adverbials. Erben (1972: 166ff) tries to generally<br />

avoid the dist<strong>in</strong>ction between adverbs and adjectives by assign<strong>in</strong>g them the common class<br />

name charakterisierendes Beiwort, but nevertheless uses the term adverb on a few occasions<br />

when speak<strong>in</strong>g about sentence adjectives, such as schnell (1972: 177f).<br />

Even the grammars which claim that these elements are adjectives, are sometimes not<br />

consistent. The authors <strong>of</strong> the Duden, for <strong>in</strong>stance, who clearly state that these elements are<br />

adjectives, are <strong>in</strong>consistent, as they classify adjectives like früh and schön as adverbs when<br />

expla<strong>in</strong><strong>in</strong>g how to generate the comparative forms (1984: 606).<br />

Eisenberg (1989: 219ff) summarises the arguments for and aga<strong>in</strong>st the classification <strong>of</strong> these<br />

elements as adverbs.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 51<br />

Reasons why laut and sorgfältig could be considered as adverbs are that (a) they can stand <strong>in</strong><br />

the same positions as adverbs like hier and niemals. (b) They refer to the verb and (c) they,<br />

as well as adverbs, are not <strong>in</strong>flected.<br />

However, Eisenberg prefers their classification as adverbial adjectives because (a) there are<br />

some positions that they do not share <strong>with</strong> adverbs, like for <strong>in</strong>stance:<br />

46 * Hans ist sorgfältig Lehrer.<br />

47 Hans ist hier Lehrer.<br />

(b) Adverbs very <strong>of</strong>ten do not refer to the verb which <strong>in</strong>validates the second argument, and<br />

(c) laut and sorgfältig are not <strong>in</strong>flected, whereas adverbs generally are not <strong>in</strong>flectable. A<br />

consequence <strong>of</strong> classify<strong>in</strong>g non-<strong>in</strong>flected adjectives as adverbs would be that predicative<br />

adjectives, such as <strong>in</strong> 48, would have to be categorised as adverbs as well:<br />

48 Hans ist sorgfältig.<br />

We want to po<strong>in</strong>t out two arguments aga<strong>in</strong>st Eisenberg's classification which have not been<br />

brought up <strong>in</strong> any grammar. The first one is that, diachronically, these elements were<br />

adverbs, which had special adverb end<strong>in</strong>gs <strong>in</strong> Old Greek, Lat<strong>in</strong> and <strong>in</strong> Old and Middle High-<br />

<strong>German</strong> (-o and -e). The end<strong>in</strong>gs disappeared on the way to New High-<strong>German</strong> so that<br />

nowadays these adverbs are form-identical <strong>with</strong> adjectives (Wahrig, 1986: 32). Accord<strong>in</strong>g to<br />

Jung (1971: 317), a few adverbs <strong>with</strong> an -e end<strong>in</strong>g still exist <strong>in</strong> parallel <strong>in</strong> today's <strong>German</strong><br />

(49a, 49b), although some <strong>of</strong> them are restricted to poetic language (49c):<br />

49a gern - gerne<br />

49b lang - lange<br />

49c Guter Mond, du gehst so stille. (poetic)<br />

A second po<strong>in</strong>t worth mention<strong>in</strong>g is that, apparently, <strong>in</strong> most languages the equivalent<br />

expressions <strong>of</strong> <strong>German</strong> deutlich (50a) are adverbs, which are clearly identifiable as such.<br />

Mostly this is because <strong>of</strong> an adverb-specific end<strong>in</strong>g (-ly, -ment, -mente etc). The only<br />

languages <strong>in</strong> our list which behave like <strong>German</strong> <strong>in</strong> that the adverb and the un<strong>in</strong>flected<br />

adjective have the same form are the Indo-European languages Dutch (50c), Greek (50l) and<br />

Persian (50p).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 52<br />

The fact that the equivalent expressions <strong>of</strong> deutlich <strong>in</strong> other languages are adverbs, is not a<br />

proper argument for <strong>German</strong> deutlich to be an adverb, as languages sometimes do express<br />

semantic contents by different parts <strong>of</strong> speech (cf. <strong>German</strong> adverb gern vs. English: to like).<br />

In H<strong>in</strong>di, for example (50o), clearly is translated as a postpositional phrase (clarity <strong>with</strong>),<br />

and we would not like to argue that deutlich is a PP. In teach<strong>in</strong>g practice and NLP, however,<br />

it would probably make th<strong>in</strong>gs easier, if we assume that deutlich, <strong>in</strong> analogy to most other<br />

languages, is an adverb:<br />

50a <strong>German</strong>:<br />

50b English:<br />

50c Dutch:<br />

50d French:<br />

50e Spanish:<br />

50h Italian:<br />

50i Portuguese:<br />

Er spricht deutlich.<br />

He speaks clearly.<br />

Hij spreekt duidelijk.<br />

Il parle nettement.<br />

Habla claramente.<br />

Habla claro. 30<br />

Parla chiaramente.<br />

Ele fala claramente<br />

50k Swedish: Han talar tydligt 31<br />

he speaks clearly<br />

50l Greek: Milaei kaqara. 32<br />

speaks clearly<br />

50m Welsh: Maén siarad yn glir. 33<br />

he speaks clearly<br />

30 In most cases, Spanish allows for an alternative to the adverb which is the mascul<strong>in</strong>e s<strong>in</strong>gular adjective<br />

form (claramente - claro). However, a few adverbs do not allow the use <strong>of</strong> the mascul<strong>in</strong>e s<strong>in</strong>gular form,<br />

like for <strong>in</strong>stance (Thanks to Elena Bárcena):<br />

50f trabaja concienzudamente<br />

* trabaja concienzudo<br />

50g corrige descuidadamente<br />

* corrige descuidado<br />

31 The Swedish adjective tydlig has the adverb suffix -t. The adverb form is identical to the form <strong>of</strong><br />

adjectives accompany<strong>in</strong>g -ett nouns. However, as no such noun occurs <strong>in</strong> the sentence to agree <strong>with</strong>, the<br />

suffix cannot be the adjective end<strong>in</strong>g (Thanks to Kristi<strong>in</strong>a Jok<strong>in</strong>en). It seems that most Danish adverbs are<br />

also formed by add<strong>in</strong>g the adverb suffix -t to adjectives (Thanks to Neil Tipper).<br />

32 The Greek equivalent <strong>of</strong> <strong>German</strong> deutlich is kaqara. The form is identical to the nom<strong>in</strong>ative or accusative<br />

form <strong>of</strong> the neuter adjective (Thanks to Mel<strong>in</strong>a Alexa).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 53<br />

50n Irish (Ghaeilge): Labhraíon sé go soileir. 34<br />

Speaks he adv clear<br />

50o H<strong>in</strong>di: Woh sufaayi say boolta hai. 35<br />

he clarity <strong>with</strong> speaks<br />

50p Persian: Vaazeh sohbat miikonad. 36<br />

He clear to speak<br />

50q Russian: On govorit jasno. 37<br />

He speaks clearly<br />

50r F<strong>in</strong>nish: Hän puhuu selvästi. 38<br />

S/he speaks clearly<br />

50s Basque: Argi hitz egiten du. 39<br />

clearly speaks he<br />

50t Arabic: Yatakalam waadihan. 40<br />

He speaks clearly<br />

50u Japanese: Kare-wa hakkiri-to hanasu. 41<br />

he -theme clarity-adv speak<br />

50v Mandar<strong>in</strong> Ch<strong>in</strong>ese: Ta q<strong>in</strong>gchu-de jianghua. 42<br />

he clear -ly speak<br />

33 The Welsh equivalent <strong>of</strong> deutlich, yn glir, is composed by the adverb marker yn and the follow<strong>in</strong>g<br />

adjective glir (Thanks to Debbie Sapsed and her Welsh helpers).<br />

34 Irish behaves the same as Welsh, <strong>in</strong> that the obligatory adverb marker precedes the adjective (Thanks to<br />

Liam Murray).<br />

35 H<strong>in</strong>di expresses clearly as the postpositional phrase sufaayi say (clarity <strong>with</strong>) (Thanks to Archana<br />

H<strong>in</strong>duja).<br />

36 The Persian equivalent <strong>of</strong> <strong>German</strong> deutlich is sohbat. The form is identical to the un<strong>in</strong>flected adjective<br />

(Thanks to Farid-Ali Khazaee).<br />

37 The Russian adverb jasno can be clearly identified as such, because <strong>of</strong> the adverb end<strong>in</strong>g -o and because<br />

<strong>of</strong> its position <strong>in</strong> the sentence (Thanks to Anne Reck).<br />

38 The F<strong>in</strong>nish adverb end<strong>in</strong>g is -sti (Thanks to Kristi<strong>in</strong>a Jok<strong>in</strong>en).<br />

39 In Basque, the adjective has to agree <strong>in</strong> number <strong>with</strong> the element it modifies, even <strong>in</strong> predicative structures<br />

(sg: argia, pl: argiak). Thus, argi is an adverb because it does not have an end<strong>in</strong>g (Thanks to Maria<br />

Victoria Arranz).<br />

40 In Arabic, the adjective clear is waadih and the adverb is formed by affix<strong>in</strong>g the accusative end<strong>in</strong>g (-an)<br />

(Thanks to Cather<strong>in</strong>e Pease).<br />

41 In Japanese, clearly is derived from a noun (hakkiri - clarity) by us<strong>in</strong>g the adverbial affix end<strong>in</strong>g (Thanks<br />

to Masaki Kiyono).<br />

42 Modern Mandar<strong>in</strong> Ch<strong>in</strong>ese behaves like most <strong>of</strong> the other languages, <strong>in</strong> that the adverb marker -de is<br />

added to the adjective q<strong>in</strong>gchu. It is <strong>in</strong>terest<strong>in</strong>g that, <strong>in</strong> parallel to this modern way <strong>of</strong> speak<strong>in</strong>g, an older<br />

one exists, which is hard to analyse <strong>in</strong> terms <strong>of</strong> Indo-European traditions: Ta jianghua q<strong>in</strong>gchu. Either he<br />

(ta) and speak (jianghua) are analysed as double subject, <strong>with</strong> the first one be<strong>in</strong>g marked as topic (cf.<br />

Japanese), or speak and clear (q<strong>in</strong>gchu) are analysed as double predicate. However, none <strong>of</strong> the analyses


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 54<br />

50w Esperanto: Li parol-as klar-e. 43<br />

he (nom) speak (pres) clear-ly<br />

50x Malay: Percakapan-nya jelas. 44<br />

speak<strong>in</strong>g his clear<br />

Dia ber- jacap dengan jelas.<br />

He do<strong>in</strong>g speak <strong>with</strong> clarity<br />

However, follow<strong>in</strong>g the general trend <strong>of</strong> some grammars, which we consider as be<strong>in</strong>g among<br />

the most powerful ones, we shall call these elements adjectives. The names adverbial<br />

adjectives and sentence adjectives will help us to dist<strong>in</strong>guish their use from the predicative<br />

and the attributive uses.<br />

51a e<strong>in</strong> zuverlässiger Partner<br />

51b T<strong>in</strong>a ist zuverlässig.<br />

51c Wir erledigen die Sache zuverlässig.<br />

(attributive)<br />

(predicative)<br />

(adverbial adjective)<br />

Another problem, which is l<strong>in</strong>ked to this type <strong>of</strong> adjective, should be mentioned briefly:<br />

Adverbial adjectives can refer to several elements <strong>in</strong> the sentence, probably depend<strong>in</strong>g on<br />

the semantics <strong>of</strong> all <strong>in</strong>volved words. The adjectives below refer to the subject (52a/53a), to<br />

the object (52b/53b) or to the verbal action (52c/53c; Duden, 1984: 582f):<br />

52a Der Beamte verlangt den Ausweis zerstreut.<br />

52b Der Beamte verlangt den Ausweis aufgeschlagen.<br />

52c Der Beamte verlangt den Ausweis laut.<br />

53a Der zerstreute Beamte<br />

53b Der aufgeschlagene Ausweis<br />

53c Das laute Verlangen / (Der laute Beamte)<br />

The complexity <strong>in</strong>volved <strong>in</strong> this problem is particularly clear <strong>in</strong> the last example, where one<br />

cannot exclude the modification <strong>of</strong> der Beamte by laut. We shall not follow up this question<br />

any further, as it does not have any impact on word order. For further discussion and<br />

references see Eisenberg (1989: 222ff), Duden (581ff) and Helbig/Buscha (1988: 554ff).<br />

would see q<strong>in</strong>gchu as an adverb, as old Mandar<strong>in</strong> Ch<strong>in</strong>ese does not formally dist<strong>in</strong>guish between parts-<strong>of</strong>speech<br />

(Thanks to Jip<strong>in</strong>g Sun).<br />

43 In Esperanto, the adverb end<strong>in</strong>g -e must be added to adjectives, such as klar, when they modify the verb<br />

(Thanks to Wei Li).<br />

44 In Malay, the dist<strong>in</strong>ction between adverb and adjective does not exist. Example v seems to be a copula<br />

construction <strong>with</strong> the copula verb be<strong>in</strong>g left out. Furthermore, the adjective/adverb jelas is the same word<br />

as the noun clarity <strong>in</strong> He speaks <strong>with</strong> clarity (Thanks to Seet Wykeen):


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 55<br />

2.2.4 CONCLUSION, FINAL DEFINITION<br />

Due to the heterogeneity <strong>of</strong> the word class, adverbs have been classified and def<strong>in</strong>ed <strong>in</strong> very<br />

different ways. Criteria concerned the follow<strong>in</strong>g features:<br />

- etymology <strong>of</strong> the word ad-verbum (accompany<strong>in</strong>g a verb/word)<br />

- <strong>in</strong>flection (not <strong>in</strong>flectable, not <strong>in</strong>flected)<br />

- semantics (situate other entities locally, temporally etc)<br />

- lexical mean<strong>in</strong>g<br />

- can or cannot fill the Vorfeld position<br />

- can be an answer to w-questions or are w-elements themselves<br />

- function as modifiers or attributes, or accompany sentences<br />

- open class<br />

- modify verbs, adjectives, adverbials, and sentences<br />

- occur freely or as pro-forms<br />

- sentence adverbs can express the speakers' <strong>in</strong>tentions, judgement etc<br />

- adverbial, predicative and attributive use<br />

- some have comparative forms<br />

- adverb is a word category or/and a relational term<br />

- no case mark<strong>in</strong>g (as opposed to prepositions)<br />

- constitute a member <strong>of</strong> the subord<strong>in</strong>ate clause when l<strong>in</strong>k<strong>in</strong>g clauses (as opposed<br />

to conjunctions)<br />

- historically, adverbially-used adjectives are adverbs (end<strong>in</strong>gs -e and -o <strong>in</strong> Middle<br />

High-<strong>German</strong>)<br />

- <strong>in</strong> many languages, adverbially used adjectives have adverb end<strong>in</strong>gs<br />

Furthermore, we saw that the semantic classification, as well as the categorization <strong>with</strong><br />

respect to the term particle, vary considerably from one grammar to the other.<br />

We want to def<strong>in</strong>e the term adverb <strong>in</strong> the follow<strong>in</strong>g way:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 56<br />

DEFINITION: The term adverb denotes a non-<strong>in</strong>flectable word class. The<br />

characteristic concern<strong>in</strong>g <strong>in</strong>flectability dist<strong>in</strong>guishes the adverb from the <strong>in</strong>flectable<br />

word classes verb, noun, adjective, article and pronoun. Adverbs share this feature<br />

<strong>with</strong> prepositions, conjunctions and <strong>in</strong>terjections. In opposition to prepositions,<br />

adverbs call<strong>in</strong>g for a specific case can also occur on their own. Conjunctions are not a<br />

constituent (Satzglied) <strong>of</strong> the subord<strong>in</strong>ate clause when l<strong>in</strong>k<strong>in</strong>g clauses, whereas<br />

adverbs are. Interjections always occur syntactically isolated, whereas adverbs can be<br />

part <strong>of</strong> a sentence.<br />

This def<strong>in</strong>ition is purely exclusive and does not consider the large amount <strong>of</strong> semantic and<br />

syntactic similarities adverbs share <strong>with</strong> adjectives, PPs, conjunctions etc. We also do not<br />

want to restrict the use <strong>of</strong> the term adverb to elements which can fill the Vorfeld position 45 .<br />

Without hav<strong>in</strong>g said so explicitly <strong>in</strong> the def<strong>in</strong>ition, toners (54a), degree modifiers (54b),<br />

modals (Modalwörter) (54c) and pronom<strong>in</strong>al adverbs (54d) are part <strong>of</strong> what we consider as<br />

adverbs:<br />

54a Er arbeitet ja noch.<br />

54b Er arbeitet sehr konzentriert.<br />

54c Er arbeitet vermutlich noch.<br />

54d Er war dabei sehr konzentriert. (bei der Arbeit)<br />

This def<strong>in</strong>ition <strong>of</strong> adverb does not <strong>in</strong>clude all one-word modifiers at sentence level which we<br />

would like to cover <strong>in</strong> our work, as sentence adjectives (which by some are categorised as<br />

adverbs) are excluded. This constitutes a problem, as we do not have a name for the whole<br />

group <strong>of</strong> one-word modifiers at sentence level. Therefore, we shall use the terms modifier<br />

and adverbial to name this group. These two terms normally <strong>in</strong>clude NPs and PPs, which<br />

have the same function as adverbs, and therefore our use <strong>of</strong> them is not entirely correct.<br />

We believe, however, that the confusion <strong>of</strong> the terms can be pardoned, as we state here<br />

explicitly that, <strong>in</strong> the context <strong>of</strong> our research, we understand by modifiers and adverbials the<br />

group <strong>of</strong> one-word modifiers at sentence level, unless otherwise stated. Furthermore, NPs<br />

45 Our reason for this is that this criterion can easily be stated <strong>in</strong>dependently. Although we honour Engel's<br />

attempt to <strong>of</strong>fer clear syntactic criteria for the classification <strong>of</strong> the different word classes, this results <strong>in</strong> too<br />

great a number <strong>of</strong> word classes.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 57<br />

and PPs used adverbially behave very similarly to one-word modifiers, <strong>with</strong> respect to their<br />

position<strong>in</strong>g <strong>in</strong> the sentence. As we stated <strong>in</strong> 1.2, the only reason we do not consider NPs and<br />

PPs, is that they are an open class and therefore difficult to recognise automatically.<br />

2.3. MODIFIER TYPES (ANGABEKLASSEN)<br />

Hoberg (1981: 132ff) and Engel (1988: 219ff) dist<strong>in</strong>guish four ma<strong>in</strong> modifier classes, which<br />

we want to discuss briefly, as we shall <strong>of</strong>ten refer to their classification: (a) existimatorial<br />

modifiers, (b) situatives, (c) modifiers to express negation and (d) modal adverbials. Hoberg<br />

calls the last group modale Angaben, whereas Engel (1988: 219f) calls it modifizierende or<br />

modifikative Angaben. The borderl<strong>in</strong>e between the four groups is not clear-cut. Therefore,<br />

the follow<strong>in</strong>g description <strong>of</strong> these superclasses should not be taken too strictly. There is also<br />

a large number <strong>of</strong> relevant subclasses.<br />

The group <strong>of</strong> existimatorial modifiers <strong>in</strong>cludes elements, such as the follow<strong>in</strong>g and many<br />

more: denn, ja, eben (not temporal), jedoch, freilich, besonders, primär, immerh<strong>in</strong>,<br />

theoretisch, vielleicht, sicher(lich), h<strong>of</strong>fentlich, leider, normalerweise, wirklich, selbst, ruhig,<br />

... (see adverb lists <strong>in</strong> appendix 8.1 and 8.3).<br />

Existimatorial modifiers can be realised as adverbs (<strong>in</strong> particular degree adverbs, toners and<br />

modal adverbs), non-<strong>in</strong>flected adjectives, PPs and subclauses. The Lat<strong>in</strong> word existimare<br />

means to assess or to evaluate (Engel, 1981: 226). Hoberg (1981: 132f) also calls this<br />

modifier class sentence adverbials or pragmatical modifiers. By us<strong>in</strong>g these modifiers the<br />

speaker gives an assessment, evaluation or judgement, <strong>with</strong> respect to the proposition <strong>of</strong> the<br />

sentence. In future, we shall use the terms evaluative, existimatorial and pragmatical<br />

modifiers as synonyms. Existimatorial modifiers <strong>in</strong>dicate the role <strong>of</strong> the sentence as a<br />

pragmatic entity <strong>with</strong><strong>in</strong> the discourse. The term sentence adverbial relates to the fact that<br />

these elements represent a superord<strong>in</strong>ate sentence (or hyper-sentence), which can generally<br />

be paraphrased by: "Es ist [modifier] der Fall, daß ..." (55b) or an equivalent (55c):


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 58<br />

55a Peter spielt leider Klavier.<br />

55b Es ist leider der Fall, daß Peter Klavier spielt.<br />

55c Es tut mir leid, daß Peter Klavier spielt.<br />

One could also say that evaluative modifiers modify the verb <strong>of</strong> the hyper-sentence and that,<br />

<strong>in</strong> an underly<strong>in</strong>g structure, they are not part <strong>of</strong> the sentence <strong>in</strong> which they occur, but rather<br />

part <strong>of</strong> a superord<strong>in</strong>ate sentence. They cannot be replaced by a pr<strong>of</strong>orm (5d) and cannot be<br />

asked for (55e):<br />

55d * Peter spielt so Klavier. (as an equivalent <strong>of</strong> a)<br />

55e * Wie / warum / weshalb... spielt Peter Klavier? (as a question to 55a)<br />

The group <strong>of</strong> situative modifiers <strong>in</strong>cludes the follow<strong>in</strong>g elements and many more: trotzdem,<br />

rechtlich, deshalb, gestern, hier, plötzlich, endlich, ebenfalls, nochmals, selten, bloß, erst, ...<br />

These elements give the temporal, causal, local etc circumstances for the verb, for its<br />

complements and sometimes also for further modifiers. They are the biggest and most<br />

frequent subclass <strong>of</strong> modifiers, hav<strong>in</strong>g the most subgroups, and they can be realised as<br />

adverbs, PPs, accusative NPs, adjectives (adjective phrases) and subclauses (Engel, 1988:<br />

220ff). Situatives are less restricted, <strong>with</strong> respect to their placement, than the other modifier<br />

classes.<br />

The group <strong>of</strong> modal modifiers <strong>in</strong>cludes schnell, laut, so, sehr, ganz, gern, mite<strong>in</strong>ander, damit<br />

and many more. These elements modify, degree-modify or quantify the verb (exclud<strong>in</strong>g its<br />

complements). They can be realised by adverbs, un<strong>in</strong>flected adjectives, PPs, wie-phrases<br />

(Sie spricht wie ihre Tante) and others (Engel, 1988: 219f). They can never be placed beh<strong>in</strong>d<br />

the ma<strong>in</strong> verb, if the verb is <strong>in</strong> clause-f<strong>in</strong>al position. Hoberg (1981: 133f) claims that modal<br />

modifiers can never be comb<strong>in</strong>ed <strong>with</strong> the copula se<strong>in</strong>, as se<strong>in</strong> is semantically empty (56b),<br />

but at least the adverb gern can (56c). It is however correct that de-adjectival adverbs cannot<br />

modify the verb se<strong>in</strong>:<br />

56a Er rechnet die Beträge sorgfältig zusammen.<br />

56b * Er ist sorgfältig Lehrer.<br />

56c Er ist gern Lehrer.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 59<br />

It is <strong>in</strong>terest<strong>in</strong>g to see that, among the few dozen manner adverbs listed <strong>in</strong> appendix 8.3<br />

(class a 43 ), only three can modify se<strong>in</strong>: gern, umsonst and vergebens:<br />

57a Er ist vergebens Lehrer.<br />

57b * Er ist laut/schnell/gut/sorgfältig/... Lehrer.<br />

Umsonst and vergebens are adverbs which, for semantic reasons, we were unhappy to<br />

classify as manner modifiers (a 43 elements). We categorised them as members <strong>of</strong> a 43 for<br />

their positional behaviour only (see section 6.7.1, C: position class).<br />

The third exception, the adverb gern, is slightly peculiar <strong>in</strong> several respects, <strong>in</strong>clud<strong>in</strong>g the<br />

fact that it can occur <strong>in</strong> positions as <strong>in</strong> c. We want to use this occasion to highlight briefly<br />

some further aspects <strong>of</strong> this special modifier.<br />

Accord<strong>in</strong>g to Engel's strict def<strong>in</strong>ition (cf. 2.2.3.3), gern is the only adverb among the<br />

elements which can play the role <strong>of</strong> verb modifier (Engel, 1988: 219). All other one-wordelements<br />

<strong>of</strong> this group are either degree modifiers or they are <strong>in</strong>flectable, and thus adjectives.<br />

Another peculiarity about this adverb is that a lot <strong>of</strong> languages render its semantic contents<br />

(its mean<strong>in</strong>g) by a verb, as opposed to by an adverb. These <strong>in</strong>clude Indo-European languages<br />

such as English, French, Spanish, Italian, Irish Gaelic, Welsh and Persian, as well as non-<br />

Indo-European ones such as Basque, Nkwen, Arabic, Malay and Mandar<strong>in</strong> Ch<strong>in</strong>ese 46 :<br />

58 Ich lese gern.<br />

59 I like read<strong>in</strong>g.<br />

60 Persian: Khaandan-raa duust-daaram.<br />

read<strong>in</strong>g I to like<br />

61 Mandar<strong>in</strong>: Wo xihuan yuedu.<br />

I like read<strong>in</strong>g<br />

46 For their help <strong>with</strong> discuss<strong>in</strong>g the equivalent <strong>of</strong> gern <strong>in</strong> other languages, we want to thank Masaki Kiyono,<br />

Maria-Victoria Arranz, Farid-Ali Khazaee, Cather<strong>in</strong>e Pease, Jip<strong>in</strong>g Sun, Kristi<strong>in</strong>a Jok<strong>in</strong>en, Sylvana<br />

S<strong>of</strong>kova, Wim Peters, Siety Meyer, Emma Mumford, Liam Murray, Neil Tipper, Seet Wykeen and Blaise<br />

Nkwenti Azeh.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 60<br />

To our knowledge, the only language, which behaves like <strong>German</strong> <strong>in</strong> that it normally<br />

expresses the mean<strong>in</strong>g <strong>of</strong> gern by an adverb is Dutch (62). In F<strong>in</strong>nish, gern can be expressed<br />

by a postpositional phrase (63) or by a verb (64). Both ways are equally possible and likely<br />

to occur:<br />

62 Ik lees graag.<br />

63 Hän lukee miele+llä+än<br />

S/he reads pleasure+<strong>with</strong>+3pers.poss.suff<br />

64 Hän pitää luke+mise+sta<br />

S/he likes read+<strong>in</strong>g +elative case (nom<strong>in</strong>al form <strong>of</strong> the verb, not <strong>in</strong>f<strong>in</strong>itive)<br />

The artificially created language Esperanto is very flexible and therefore allows for<br />

constructions us<strong>in</strong>g an adverb (65), as well as three different verb constructions (66, 67, 68):<br />

65 Mi shat-e leg-as.<br />

I lik<strong>in</strong>gly (-e is Adverb morpheme) read (pres.)<br />

I read lik<strong>in</strong>gly (I read enjoy<strong>in</strong>gly)<br />

66 Mi shat-as leg-i.<br />

I(nom) like (pres.) read (<strong>in</strong>f)<br />

Lit.: I like to read.<br />

67 Mi shat-as leg-on (or: leg-ad-on).<br />

read (noun, acc.)<br />

Lit.: I like read<strong>in</strong>g.<br />

68 Leg-i/leg-o/leg-ad-o plach-as al mi.<br />

Lit.: To read/Read<strong>in</strong>g pleases me.<br />

Danish and Swedish do have adverbs express<strong>in</strong>g <strong>German</strong> gern, namely gerne and gärna, but<br />

it seems that both languages would rather express <strong>German</strong> gern by a verb. The part-<strong>of</strong>speech<br />

<strong>of</strong> the Japanese equivalent <strong>of</strong> gern is not clear to us, as it seems that our Indo-<br />

European word classes cannot be applied directly. Accord<strong>in</strong>g to different l<strong>in</strong>guistic analyses,<br />

suki da (69) is either an adjectival verb or it is an adjectival noun (suki), followed by the<br />

verb to be (da/desu):<br />

69 Kare wa (hon wo) yomu no ga suki da.<br />

(he) (read) (like)


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 61<br />

In Irish Gaelic (Ghaeilge), gern is expressed as maith which, apparently, could be a noun, an<br />

adjective or an adverb (70). When the mean<strong>in</strong>g He likes to read is <strong>in</strong>tended, Irish Gaelic can<br />

form an adverb on the basis <strong>of</strong> the noun pleasure (71) or the adjective pleasant (72):<br />

70 Is maith liom leamh.<br />

Is good/well/like "<strong>with</strong>"-1st-sg read<br />

read<strong>in</strong>g is good <strong>with</strong> me<br />

71 Léann sé go pléisiúrtha.<br />

read he adv pleasure<br />

72 Léann sé go laoibhneach.<br />

read he adv pleasant<br />

We do not have anyth<strong>in</strong>g more to say about the special status <strong>of</strong> gern but want to po<strong>in</strong>t out<br />

that this adverb may also behave slightly differently from the other elements <strong>of</strong> its group,<br />

<strong>with</strong> regard to position.<br />

To come back to the three ma<strong>in</strong> groups, one can say that evaluative modifiers modify the<br />

utterance, situative adverbials the sentence and modal modifiers the verb.<br />

Negational modifiers <strong>in</strong>clude a limited number <strong>of</strong> one-word elements (nicht, ke<strong>in</strong>eswegs,<br />

ke<strong>in</strong>esfalls, kaum, nirgends and niemals) and comb<strong>in</strong>ations <strong>of</strong> these (gar nicht, nicht mehr<br />

etc). Semantically, they negate clauses (73) or parts <strong>of</strong> them (74) (Engel, 1988: 226), and<br />

they cannot be asked for (75). The negational particle nicht, and its comb<strong>in</strong>ations <strong>with</strong><br />

adverbs, cannot fill the Vorfeld, whereas the others can.<br />

73 Peter hat Paul das Buch nicht gegeben.<br />

74 Peter hat Paul das Buch nicht <strong>in</strong> der Kirche gegeben.<br />

75 * Wie/warum/w<strong>of</strong>ür/... hat Peter Paul das Buch gegeben? (as a question to 73)<br />

As the scope <strong>of</strong> negation is a very complex subject, which deserves more thorough<br />

treatment, we shall not <strong>in</strong>clude negational modifiers <strong>in</strong> our word order discussion. For some<br />

relations between word order and negation see section 5.6, as well as Ste<strong>in</strong>berger (1990:<br />

117ff) and (1992b).<br />

We shall subdivide these four ma<strong>in</strong> groups later and discuss the subgroups <strong>in</strong> the chapters 3,<br />

5, and 6.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 62<br />

2.4. SOME INFORMATION ON THE POSITION OF MODIFIERS<br />

Adverb Placement: Slippery Johnnies, adverbs. [...] However, it is not clear what<br />

factors <strong>in</strong>fluence the choice <strong>of</strong> rest<strong>in</strong>g place. (Balkan et. al., 1991: 171)<br />

In this section, we want to summarise the <strong>in</strong>formation given by different grammars, on the<br />

placement <strong>of</strong> adverbs <strong>in</strong> relation to each other. We shall also po<strong>in</strong>t out contradictions<br />

between them. The ma<strong>in</strong> goal is to give an overview <strong>of</strong> what <strong>in</strong>formation is currently<br />

available to both language learners and writers <strong>of</strong> computer grammars. We shall limit<br />

ourselves to list<strong>in</strong>g the sequences <strong>of</strong> adverb classes, <strong>with</strong>out mention<strong>in</strong>g how the authors<br />

expla<strong>in</strong> them. Explanations will be given <strong>in</strong> chapter 3.<br />

The knowledge most grammars provide regard<strong>in</strong>g modifier sequences is very scarce. Only<br />

some more recent ones, namely Engel (1988), Sommerfeldt/Starke (1988) and<br />

Helbig/Buscha (1988), have a more sophisticated approach, show<strong>in</strong>g the <strong>in</strong>teraction <strong>of</strong><br />

multiple factors. All grammars but one (Liebsch/Doer<strong>in</strong>g, 1976) give at least some<br />

<strong>in</strong>formation on the position <strong>of</strong> adverbs <strong>in</strong> the sentence and/or their relative position to each<br />

other.<br />

The development <strong>in</strong> the treatment <strong>of</strong> word order <strong>in</strong> grammar books shows, that the l<strong>in</strong>guistic<br />

research <strong>of</strong> the last decades has been <strong>of</strong> immediate use for the language learner. We hope<br />

that our work will be a further contribution to the explanation and treatment <strong>of</strong> this<br />

particularly confus<strong>in</strong>g subject. In this section, we shall not go <strong>in</strong>to much detail, because our<br />

discussion <strong>of</strong> the problem depends on the facts given <strong>in</strong> chapter 3, those be<strong>in</strong>g the factors<br />

which determ<strong>in</strong>e <strong>German</strong> word order.<br />

Russon and Russon (1978: 7-9) place adverbial expressions after pronouns, between two<br />

noun objects and before prepositional objects, "s<strong>in</strong>gle noun objects" and predicative<br />

adjectives. They dist<strong>in</strong>guish six adverbial position classes, the order <strong>of</strong> which tends to be:<br />

cause < time < manner < place < purpose or result < degree


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 63<br />

The flash ("A < B") <strong>in</strong>dicates that A stands before B. They add that when there is a<br />

cumulation <strong>of</strong> several modifiers <strong>of</strong> one subtype, the more general precedes the more specific.<br />

The negational modifier nicht precedes the past participle, the <strong>in</strong>f<strong>in</strong>itive, the separable<br />

prefix, the predicative adjective or noun and adverbial phrases <strong>of</strong> place, manner or degree.<br />

Jung (1971: 147) claims the same order, but does not mention purpose or result modifiers. If<br />

any <strong>of</strong> these elements is focused on, the order can change:<br />

cause < time < manner < place < goal<br />

Helbig and Buscha (1988: 564-583) give a very good explanation <strong>of</strong> some pr<strong>in</strong>ciples, and<br />

stress their <strong>in</strong>teraction, but dist<strong>in</strong>guish only four types <strong>of</strong> modifiers which appear <strong>in</strong> the weak<br />

order:<br />

temporal and causal < local and modal<br />

Conjunctional adverbials, like deswegen, precede the others. Furthermore, they mention that<br />

adverbials tend to fill the Vorfeld position and that facultative elements (<strong>in</strong>clud<strong>in</strong>g modifiers)<br />

generally precede compulsory elements (Ergänzungen), but that pronouns precede<br />

adverbials:<br />

pronom<strong>in</strong>al arguments < free adverbials < <strong>in</strong>def<strong>in</strong>ite/article-less arguments<br />

Accord<strong>in</strong>g to Schulz and Griesbach (1980: 389ff) time and place modifiers <strong>of</strong>ten fill the<br />

Vorfeld position and thus tend to precede other modifiers and arguments. The sequence <strong>of</strong><br />

the other modifiers is not fixed but is likely to be:<br />

temporal < causal < modal < local<br />

When the causal modifier is realised as a pronoun (e.g. deshalb) it precedes the temporal<br />

adverbial.<br />

Schulz and Griesbach furthermore mention a list <strong>of</strong> four adverbs which have the order:<br />

auch < mehr < noch < schon


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 64<br />

but which vary <strong>in</strong> their order <strong>with</strong> respect to the negational element nicht: nicht precedes<br />

mehr <strong>in</strong> nicht mehr, but follows noch <strong>in</strong> noch nicht. This apparently contradicts the above<br />

order:<br />

76 nicht mehr noch nicht<br />

auch nicht mehr<br />

auch noch nicht<br />

auch noch<br />

auch nicht<br />

auch schon<br />

Accord<strong>in</strong>g to the model <strong>of</strong> degree modifier treatment presented <strong>in</strong> Ste<strong>in</strong>berger (1992b), there<br />

is no <strong>in</strong>consistency here, as nicht mehr is analysed as one positional element: nicht is<br />

modified by mehr (similar to gar nicht etc; cf. 4.4.1). If we assume that nicht, <strong>in</strong>clud<strong>in</strong>g its<br />

modifiers, follows the four adverbs, the order is not contradictory.<br />

The authors <strong>of</strong> the Duden (1984: 721-724) also stress that several grammatical and nongrammatical<br />

(pragmatical) factors <strong>in</strong>teract, and that one can only give rules-<strong>of</strong>-thumb.<br />

Tak<strong>in</strong>g over Engel's (1970: 48ff) classification, they state the follow<strong>in</strong>g order:<br />

1a: temporal, local, causal<br />

< 2: existimatorial (evaluative)<br />

< 1b: (like 1a but which also express judgement/assessment or <strong>in</strong>def<strong>in</strong>iteness<br />

(e.g. manchmal))<br />

< 3: negation particles<br />

< 4: adverbial sentence adjectives<br />

They do not commit themselves to a certa<strong>in</strong> sequence <strong>with</strong> respect to verb arguments, but<br />

state that some elements have a very strong tendency to the right (some adverbials, POs,<br />

genitive objects; less when they are pronom<strong>in</strong>al), whereas others are relatively free<br />

(<strong>in</strong>def<strong>in</strong>ite NPs).<br />

Heidolph, Flämig and Motsch (1981: 702ff) give an order for all elements <strong>in</strong> the sentence,<br />

which is roughly the follow<strong>in</strong>g:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 65<br />

Adverbials III (temporal)<br />

< Adverbials III (causal)<br />

< Adverbials II (local)<br />

< Adverbials II (modal, <strong>in</strong>strumental)<br />

< accusative and dative and prepositional objects<br />

< absolute directional adverbial I (<strong>in</strong>itial po<strong>in</strong>t)<br />

< absolute directional adverbial I (medium)<br />

< absolute directional adverbial I (goal)<br />

< relative directional adverbial<br />

< predicative element<br />

< verb<br />

The sequence <strong>of</strong> the second group <strong>of</strong> elements is fixed, whereas the order <strong>of</strong> the elements <strong>of</strong><br />

the first group can change.<br />

Sommerfeldt and Starke (1988: 277ff) give a whole range <strong>of</strong> factors, which determ<strong>in</strong>e the<br />

order <strong>of</strong> elements <strong>in</strong> the sentence. The overall order <strong>of</strong> complements is:<br />

Subj < Dat < Acc < Gen < PO/"Objektprädikativum"<br />

This however varies <strong>with</strong> the realization <strong>of</strong> the arguments as pronouns, and def<strong>in</strong>ite and<br />

<strong>in</strong>def<strong>in</strong>ite NPs.<br />

Just as Schulz and Griesbach, they mention that local and temporal modifiers <strong>of</strong>ten fill the<br />

Vorfeld, but add that local modifiers can vary quite a lot. Modal modifiers also tend to vary<br />

<strong>in</strong> their position relatively to each other. Directional elements have a fixed place at the end<br />

<strong>of</strong> the Mittelfeld. The only order Sommerfeldt and Starke commit themselves to is:<br />

<strong>in</strong>strumental modifier < local modifier < objects < local complement<br />

Accord<strong>in</strong>g to them, even this order is not obligatory, but any change results <strong>in</strong> marked word<br />

order and the focus<strong>in</strong>g <strong>of</strong> some element.<br />

The most f<strong>in</strong>e-gra<strong>in</strong>ed description is given by Engel (1988: 325-344), who dist<strong>in</strong>guishes<br />

compulsory and non-compulsory sequences. He also focuses on the existence <strong>of</strong> different


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 66<br />

factors, which have an impact on the order <strong>of</strong> non-compulsory elements. Engel gives a basic<br />

word order and only then lists its possible variations. The general sequence <strong>of</strong> modifiers is:<br />

existimatorial < situative < negational < modal<br />

Modals follow <strong>in</strong>def<strong>in</strong>ite complements and precede prepositional, directive and predicative<br />

complements, as well as the nom<strong>in</strong>al parts <strong>of</strong> support verb constructions. Situative and<br />

evaluative modifiers normally follow non-stressed pronouns and precede the <strong>in</strong>def<strong>in</strong>ite<br />

complements. The negational element must follow evaluative modifiers, and it generally<br />

follows situatives.<br />

With<strong>in</strong> groups <strong>of</strong> existimatorial modifiers the order is the follow<strong>in</strong>g:<br />

toners < judicative < ord<strong>in</strong>ative < verificative < cautive < selective<br />

The sequence <strong>of</strong> toners relative to each other is strictly ordered <strong>with</strong><strong>in</strong> seven subgroups,<br />

which are listed by Engel. The subgroup <strong>of</strong> ord<strong>in</strong>atives is strictly ordered <strong>in</strong> part, the reason<br />

for which seems not to be known.<br />

Most grammars stress that it is impossible to give a concise description <strong>of</strong> modifier classes,<br />

and their relative order as the sequence is determ<strong>in</strong>ed by several factors. However, most <strong>of</strong><br />

them dist<strong>in</strong>guish classes, which generally follow a certa<strong>in</strong> sequence, but which allow for<br />

variations. One should assume that the basic orders given by different grammars are<br />

compatible. There are however several contradictions. Russon and Russon assume that the<br />

basic order <strong>of</strong> causal and temporal modifiers is cause < time, whereas Helbig/Buscha and<br />

the Duden do not favour any sequence, and Schulz/Griesbach, as well as<br />

Heidolph/Flämig/Motsch, favour the opposite order, namely time < cause.<br />

Sommerfeldt and Starke are the only authors who situate <strong>in</strong>strumental modifiers ahead <strong>of</strong> all<br />

other modifiers, whereas the other grammarians, who specify the position <strong>of</strong> <strong>in</strong>strumentals<br />

(e.g. Heidolph/Flämig/Motsch), place them after the temporal, causal and local modifiers.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 67<br />

The different places for modal adverbials might have to do <strong>with</strong> the large differences <strong>in</strong><br />

def<strong>in</strong><strong>in</strong>g this subgroup. Heidolph/Flämig/Motsch and Engel assume their position after all<br />

the other modifiers, whereas Schulz/Griesbach position them <strong>in</strong> front <strong>of</strong> the locals.<br />

All <strong>in</strong> all, it seems obvious that the modifier sequences given by the grammars differ<br />

considerably, as no two use the same classification for their description. Furthermore, there<br />

are different op<strong>in</strong>ions regard<strong>in</strong>g the relative order <strong>of</strong> time and cause modifiers, as well as<br />

others.<br />

2.5. SOME STATISTICAL FACTS ABOUT ADVERBS<br />

The authors <strong>of</strong> the Duden (1984: 386) estimate the amount <strong>of</strong> <strong>German</strong> words, exclud<strong>in</strong>g<br />

specialised language, at between 300,000 and 500,000. This <strong>in</strong>cludes the most common<br />

compounds and derivations, but certa<strong>in</strong>ly not all possible words that can be comb<strong>in</strong>ed by<br />

us<strong>in</strong>g the over 200 morphological means to create new words. This number must be handled<br />

<strong>with</strong> care, as the borderl<strong>in</strong>e between the general and the specialised lexicon is very vague.<br />

The specialised lexicon <strong>of</strong> the medical language alone has been estimated at about 170,000<br />

words (Duden, 1984: 386).<br />

The three to five hundred thousand words are derived from a few thousand simple lemmas<br />

("Simplizia"; Duden, 1984: 386f). Ortmann (1975: vol. 3, XXXI) counted 5849 lemmas. The<br />

list below shows the number <strong>of</strong> verbs, nouns, adjectives and adverbs and their percentage<br />

compared to the total number <strong>of</strong> lemmas:<br />

Verbs 1129 19,3 %<br />

Nouns 2692 46,0 %<br />

Adjectives 1323 22,6 %<br />

Adverbs 389 + 6,7 %<br />

Total <strong>of</strong> these word classes: 94,6 %<br />

There are 69 prepositions, 75 conjunctions, 24 question words and 47 pronouns. In<br />

agreement <strong>with</strong> our classification, Ortmann classified the un<strong>in</strong>flected adjectives as<br />

adjectives, and stated the non-<strong>in</strong>flectability as a criterion for the classification <strong>of</strong> adverbs.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 68<br />

The difference between Ortmann's classification and ours is that Ortmann also <strong>in</strong>cludes non<strong>in</strong>flectable<br />

adjectives <strong>in</strong> the group <strong>of</strong> adverbs (1975, vol. 2: 5f). This group is very small and<br />

<strong>in</strong>cludes adjectives like extra, prima, rosa, amongst others (Helbig/Buscha, 1988: 312).<br />

Ortmann's 389 simple adverbs by far exceed Engel's count <strong>of</strong> 247 elements (1988: 749ff). As<br />

Engel's def<strong>in</strong>ition <strong>of</strong> adverbs is narrower than ours, we <strong>in</strong>clude the follow<strong>in</strong>g <strong>of</strong> Engel's<br />

classes:<br />

adverbs 150<br />

modal particles 22<br />

order<strong>in</strong>g particles 17 ("Rangierpartikeln")<br />

degree modifiers 26<br />

copula particles 8<br />

toners + 24<br />

Total: 247<br />

Although we do not know from where this discrepancy arises, it is quite probable that Engel<br />

forgot to consider a few adverbs. Indeed, we found some additional adverbs <strong>in</strong> the Siemens<br />

METAL lexicon, which do not appear <strong>in</strong> either <strong>of</strong> Engel's lists.<br />

Ortmann's and Engel's numbers have to be extended by the elements that can be generated<br />

by us<strong>in</strong>g morphological means, as they only list simple lemmas.<br />

The most frequent adverb <strong>in</strong> Ortmann's corpus <strong>of</strong> nearly 11 million words (Ortmann, 1975,<br />

vol. 1: 3) is the particle nicht <strong>with</strong> 114,518 occurrences (1975, vol. 2: L142ff). The most<br />

frequent word form altogether is die <strong>with</strong> 349,553 occurrences (1975, vol. 2: S51). Die thus<br />

occurs only three times as frequently as the most common modifier nicht. This is even more<br />

surpris<strong>in</strong>g, as the occurrences <strong>of</strong> die <strong>in</strong>clude all <strong>in</strong>stances <strong>of</strong> die, whatever its category<br />

(article, relative pronoun, etc). Accord<strong>in</strong>g to Engel's (1988: 876) classification, nicht belongs<br />

to two categories, as well: it can be a degree modifier or a toner.<br />

Nicht is the eighth most frequent word form <strong>in</strong> Ortmann's whole corpus. The follow<strong>in</strong>g are<br />

the 20 most frequent adverbs extracted from Ortmann's list (1975, vol. 2: L142ff):


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 69<br />

Place<br />

Adverb Occurrences out <strong>of</strong> 10,910,777 <strong>Word</strong>s<br />

8 nicht 114518<br />

22 so 74273<br />

24 auch 60750<br />

32 aber 44201<br />

36 nur 39507<br />

38 noch 39179<br />

60 da 23497<br />

64 doch 21994<br />

72 mehr 18549<br />

74 denn 18488<br />

75 nun 17891<br />

77 sehr 17293<br />

78 selbst 16911<br />

79 schon 16727<br />

80 hier 16667<br />

84 dann 15545<br />

88 wieder 14693<br />

101 ja 12527<br />

103 jetzt 11859<br />

104 immer 11664<br />

Accord<strong>in</strong>g to Meier (1978, vol. II: 111f), 7994 word forms 47 , correspond<strong>in</strong>g to 3.1% <strong>of</strong> all<br />

counted word forms, represent 87.44% <strong>of</strong> <strong>German</strong> texts. The word forms die, der and und<br />

alone represent 9.27% <strong>of</strong> text. The fact that 17 adverbs are among the 100 most frequent<br />

<strong>German</strong> word forms makes it obvious that a satisfy<strong>in</strong>g treatment <strong>of</strong> modifiers <strong>in</strong> grammars<br />

and NLP is mandatory.<br />

47 Every different <strong>in</strong>flection form <strong>of</strong> a lexeme is one word form. Gehe, gehst, geht, g<strong>in</strong>g etc are thus different<br />

word forms <strong>of</strong> the verb gehen.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 70<br />

3. FACTORS WHICH DETERMINE GERMAN WORD ORDER<br />

Many <strong>of</strong> the universals [...] suggest strongly that performance is the driv<strong>in</strong>g force<br />

beh<strong>in</strong>d l<strong>in</strong>earization pr<strong>in</strong>ciples <strong>of</strong> the competence grammar. (Hawk<strong>in</strong>s, 1990: 225)<br />

In this chapter, we describe the factors which can have an <strong>in</strong>fluence on the order <strong>of</strong> elements<br />

<strong>in</strong> the <strong>German</strong> sentence. Part <strong>of</strong> our goal was to describe and expla<strong>in</strong> <strong>in</strong> which order several<br />

modifiers follow each other, but the task is not limited to this. Another question is where to<br />

position modifiers <strong>in</strong> relation to verb arguments. As we cannot describe modifier position<strong>in</strong>g<br />

<strong>in</strong> absolute terms, but only <strong>with</strong> respect to other elements, namely verb arguments and the<br />

verb itself, we did not limit ourselves to the description <strong>of</strong> adverb placement, but tried to<br />

gather <strong>in</strong>formation on <strong>German</strong> word order <strong>in</strong> general.<br />

The examples 1, 2 and 3 show that the position <strong>of</strong> a s<strong>in</strong>gle modifier differs, depend<strong>in</strong>g on the<br />

realization <strong>of</strong> the verb arguments as def<strong>in</strong>ite NPs (1), <strong>in</strong>def<strong>in</strong>ite NPs (2) and pronouns (3). In<br />

the first case, the adverb morgen can either precede or follow both complements or stand <strong>in</strong><br />

between them (1). When the verb arguments are both <strong>in</strong>def<strong>in</strong>ite, the modifier can follow one<br />

<strong>of</strong> them but not both (2). When the objects are realised as pronouns, the adverb cannot<br />

precede both <strong>of</strong> them, and the middle-position is acceptable only when the dative pronoun is<br />

stressed contrastively (3):<br />

1a<br />

1b<br />

1c<br />

2a<br />

2b<br />

2c<br />

3a<br />

3b<br />

3c<br />

Er gibt morgen dem Mann das Buch.<br />

Er gibt dem Mann morgen das Buch.<br />

Er gibt dem Mann das Buch morgen.<br />

Er gibt morgen e<strong>in</strong>em Mann e<strong>in</strong> Buch.<br />

Er gibt e<strong>in</strong>em Mann morgen e<strong>in</strong> Buch.<br />

* Er gibt e<strong>in</strong>em Mann e<strong>in</strong> Buch morgen.<br />

* Er gibt morgen es ihm.<br />

# Er gibt es morgen IHM.<br />

Er gibt es ihm morgen.<br />

It is obvious, <strong>with</strong>out hav<strong>in</strong>g to discuss the other comb<strong>in</strong>ations <strong>of</strong> pronom<strong>in</strong>al, def<strong>in</strong>ite and<br />

<strong>in</strong>def<strong>in</strong>ite complements, that the potential places for the modifier differ depend<strong>in</strong>g on the


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 71<br />

realization <strong>of</strong> the verb arguments. Therefore, we shall discuss <strong>German</strong> word order <strong>in</strong> general<br />

and shall focus on modifiers when necessary.<br />

This chapter tries to answer the question, "what are the reasons for word order variation <strong>in</strong><br />

<strong>German</strong>?". As one could expect <strong>in</strong> a free word order language, the result <strong>of</strong> our <strong>in</strong>vestigation<br />

is not easy and straight-forward. Instead, we found eleven pr<strong>in</strong>ciples which can all apply to<br />

determ<strong>in</strong>e <strong>German</strong> word order <strong>in</strong> one sentence, namely:<br />

1) Theme-rheme structure<br />

2) Behaghel's "Gesetz der wachsenden Glieder"<br />

3) Functional sentence perspective<br />

4) Semantic-syntactic closeness to the verb ("Verbnähe")<br />

5) The Animacy-first pr<strong>in</strong>ciple<br />

6) Semantic roles<br />

7) Scope<br />

8) Rhythm<br />

9) Natural gender<br />

10) Grammaticalisation (habit)<br />

11) Lenerz' "Satzklammerbed<strong>in</strong>gung"<br />

These pr<strong>in</strong>ciples are only tendencies which <strong>in</strong>teract, their violation does not necessarily lead<br />

to ungrammaticality. We compiled this list by comb<strong>in</strong><strong>in</strong>g what we found <strong>in</strong> l<strong>in</strong>guistic<br />

literature <strong>with</strong> the results <strong>of</strong> our own <strong>in</strong>vestigation. None <strong>of</strong> the sources known to us have<br />

mentioned all <strong>of</strong> these factors, or even the majority <strong>of</strong> them. Furthermore, the factors they<br />

mention differ considerably from one source to another. Some <strong>of</strong> the factors are very closely<br />

related to each other.<br />

In chapter 4, we shall show how these pr<strong>in</strong>ciples <strong>in</strong>teract, and <strong>in</strong> chapter 6 we shall discuss<br />

what it is possible to do <strong>with</strong> this knowledge <strong>in</strong> Natural Language Process<strong>in</strong>g. This <strong>in</strong>cludes<br />

the questions <strong>of</strong> how to featurise the differences between word order variation and <strong>of</strong> how to<br />

recognise the different values for these features.<br />

We shall now discuss these pr<strong>in</strong>ciples <strong>in</strong> turn.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 72<br />

3.1 THEME-RHEME STRUCTURE<br />

The theme-rheme structure (TRS) is widely accepted as a ma<strong>in</strong> factor determ<strong>in</strong><strong>in</strong>g the order<br />

<strong>of</strong> elements <strong>in</strong> <strong>German</strong> sentences (Engel, 1970: 12). It is very closely related to another<br />

factor, called functional sentence perspective (3.3), as well as to Behaghel's Gesetz der<br />

wachsenden Glieder (3.2). The pr<strong>in</strong>ciple claims that the theme, which could roughly be<br />

called the known <strong>in</strong>formation precedes the rheme, the new <strong>in</strong>formation <strong>in</strong> the sentence. In 4a,<br />

the pronoun he is thematic:<br />

4 Context: Yesterday, I met Peter.<br />

4a He was drunk as always.<br />

3.1.1 SOME DEFINITIONS OF THE TERMS THEME AND RHEME<br />

The two terms have been discussed a lot and have been used <strong>with</strong> different mean<strong>in</strong>gs.<br />

Furthermore, a lot <strong>of</strong> different terms are <strong>in</strong> use more or less synonymously:<br />

functor/argument, given/new, topic/comment, presupposition/focus and topic/focus (cf.<br />

Ste<strong>in</strong>er/W<strong>in</strong>ter, 1987: 3, and Eisenberg, 1989: 149ff) 48 . Normally, the theme is not only the<br />

known or given <strong>in</strong>formation but at the same time it is the topic <strong>of</strong> the sentence which we<br />

presuppose. The rheme, on the other hand, is the new <strong>in</strong>formation which constitutes a<br />

comment or a predication over the theme. It is generally the focus <strong>of</strong> our attention. In<br />

predicate logic, the functor would correspond to the theme and the argument to the rheme.<br />

Although these concepts are obviously <strong>in</strong>terrelated, they are not identical.<br />

We shall concentrate on the dichotomy given versus new <strong>in</strong>formation, as these concepts have<br />

a correspondence <strong>in</strong> the realization <strong>of</strong> NPs (see section 3.1.2), and therefore are most likely<br />

to be recognised <strong>in</strong> NLP. Given <strong>in</strong>formation is thus everyth<strong>in</strong>g that has been mentioned <strong>in</strong><br />

the previous context, explicitly or not. In spoken language, the given would <strong>in</strong>clude the<br />

immediate surround<strong>in</strong>gs <strong>of</strong> both speaker and listener, as they are accessible to both persons.<br />

48 For a further discussion <strong>of</strong> the terms theme and rheme, as well as some related concepts<br />

(argument/predication, presupposition/assertion, context restriction ("Kontextgebundenheit") and context<br />

<strong>in</strong>dependence ("Kontextfreiheit")), see Eroms (1986, and <strong>in</strong> particular pp. 9-29).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 73<br />

In written language, however, writer and reader are normally separated by time and space, so<br />

that the given <strong>in</strong>formation should be restricted to what is found <strong>in</strong> the context. Given<br />

<strong>in</strong>formation is the complementary term, namely the <strong>in</strong>formation that has not been referred to<br />

previously.<br />

One problem related to the theme-rheme dichotomy is that the borderl<strong>in</strong>e between old and<br />

new <strong>in</strong>formation is <strong>of</strong>ten not clear cut. Consequently, some l<strong>in</strong>guists claim that the difference<br />

between the two should be seen as a question <strong>of</strong> degree (Bußmann, 1983: 541f). The theme<br />

would be the element <strong>with</strong> the least degree <strong>of</strong> "communicative dynamism", whereas the<br />

rheme is the element <strong>with</strong> the highest degree 49 .<br />

In Eroms' (1986: 54) classification, there are several grades <strong>of</strong> thematic elements 50 :<br />

A) the thematic basis (TB (=T0))<br />

B) situative, thematic adverbials (TSit)<br />

C) further thematic elements T1, T2, T3 etc, <strong>with</strong> T1 hav<strong>in</strong>g less communicative<br />

importance ("ger<strong>in</strong>gerer Mitteilungswert") than T2, and so on<br />

5 Ich habe heute Otto das Buch, von dem ich dir gestern erzählt habe, geschenkt.<br />

TB R TSit T1 T2 R<br />

Accord<strong>in</strong>g to the TRS, elements l<strong>in</strong>k<strong>in</strong>g the sentence to the preced<strong>in</strong>g text tend to stand at<br />

the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> the sentence. Among them are coord<strong>in</strong>at<strong>in</strong>g and subord<strong>in</strong>at<strong>in</strong>g<br />

conjunctions, as well as adverbials express<strong>in</strong>g cause, goal or contrast. Local and temporal<br />

adverbials also can have this function, ma<strong>in</strong>ly when they are realised as pro-adverbs (davor,<br />

seither etc):<br />

6 Sie g<strong>in</strong>g nach Hause. Dort wartete ihr Mann mit dem Nudelholz auf sie.<br />

49 For the dist<strong>in</strong>ction between the terms "communicative dynamism" and "communicative importance" <strong>in</strong><br />

relation to the notions theme and rheme, see Lenerz (1977: 13f).<br />

50 Eroms (1986: 30f) po<strong>in</strong>ts out that <strong>in</strong> addition to the serialisation <strong>of</strong> elements, the <strong>in</strong>tonation and the voice<br />

<strong>of</strong> the verb can express the TRS. For the discussion <strong>of</strong> voice, cf. pp. 73-80. These three components are "i<br />

n t e g r a t i v e Signale der Funktion des Satzes im Diskurs" (emphasis <strong>in</strong> the orig<strong>in</strong>al).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 74<br />

Eroms (1986: 31 and 81ff) po<strong>in</strong>ts out that the TRS even plays a role below sentence level,<br />

namely at phrase level ("Satzgliedebene"). Therefore, a precise treatment would make it<br />

necessary to dist<strong>in</strong>guish the phrase and the sentence level. In contracted structures<br />

("verdichtete Strukturen"), secondary accents can help to identify the thematic and the<br />

rhematic parts. Attributes replac<strong>in</strong>g whole clauses, for <strong>in</strong>stance, can conta<strong>in</strong> themes and<br />

rhemes.<br />

7a Nach e<strong>in</strong>er Entscheidung des Gerichts müssen ...<br />

T1<br />

7b Nach e<strong>in</strong>er Entscheidung, die das Gericht getr<strong>of</strong>fen hat, müssen ...<br />

TSit TB R1 R0<br />

In his discussion <strong>of</strong> the multi-level approach, Eroms (1986: 82f) concludes however that<br />

the sophisticated analysis <strong>of</strong> (7a) as (7b) is not necessary. All <strong>in</strong>formation relevant to the<br />

dialogue or monologue structure is available <strong>in</strong> the contracted sentence. However, if a<br />

sentence is realised <strong>in</strong> its extended (not contracted) form, it must be analysed at all its levels,<br />

i.e. at the matrix and at the subclause level.<br />

3.1.2 THE REALIZATION OF THEMATIC AND RHEMATIC ELEMENTS<br />

With respect to the realization <strong>of</strong> themes and rhemes <strong>in</strong> the sentence, Engel (1988: 322)<br />

dist<strong>in</strong>guishes three groups, which will appear <strong>in</strong> the sequence A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 75<br />

C) Indef<strong>in</strong>ite verb arguments: They are the most rhematic elements, realised either as<br />

<strong>in</strong>def<strong>in</strong>ite pronouns (jemand, nichts etc) or as <strong>in</strong>def<strong>in</strong>ite NPs. Indef<strong>in</strong>ite NPs either do<br />

not have any article (exclud<strong>in</strong>g proper names), have <strong>in</strong>def<strong>in</strong>ite articles (e<strong>in</strong>) or<br />

<strong>in</strong>def<strong>in</strong>ite determ<strong>in</strong>ers (e<strong>in</strong>ige, ke<strong>in</strong> etc). While the <strong>in</strong>def<strong>in</strong>ite subject still tends to stand<br />

at the beg<strong>in</strong>n<strong>in</strong>g, all other complements have a clear tendency towards the right <strong>of</strong> the<br />

sentence. The nom<strong>in</strong>al genitive complement always tends to the right, <strong>in</strong>dependently<br />

from whether it is def<strong>in</strong>ite or <strong>in</strong>def<strong>in</strong>ite.<br />

Reis (1987: 158) po<strong>in</strong>ts out that, although def<strong>in</strong>ite NPs can be as thematic as pronom<strong>in</strong>al<br />

elements, the latter have a much stronger tendency towards the left than def<strong>in</strong>ite NPs. As a<br />

result, the reason for this has to be <strong>in</strong>dependent from the TRS.<br />

Eroms (1986: 47) dist<strong>in</strong>guishes three groups as well, but these differ from Engel's:<br />

A) Nom<strong>in</strong>al thematic elements<br />

B) Pronom<strong>in</strong>al elements, reflexive pronouns and proper names<br />

C) Demonstrative elements<br />

Eroms classifies the proper names <strong>with</strong> the pronouns, as opposed to <strong>with</strong> the def<strong>in</strong>ite<br />

arguments, as Engel does. As personal pronouns can never refer to unknown persons,<br />

whereas def<strong>in</strong>ite NPs and proper names can, Engel's classification seems more appropriate to<br />

us.<br />

Hawk<strong>in</strong>s (1986: 46f) po<strong>in</strong>ts out that an ill-def<strong>in</strong>ed dist<strong>in</strong>ction between theme and rheme is<br />

not enough to expla<strong>in</strong> the order <strong>of</strong> elements <strong>in</strong> the sentence. He pleads that a more f<strong>in</strong>egra<strong>in</strong>ed<br />

classification, consider<strong>in</strong>g concepts such as shared <strong>in</strong>formation and mutual<br />

knowledge, be applied.<br />

Lenerz (1977: 46ff) underp<strong>in</strong>s Hawk<strong>in</strong>s' idea, by show<strong>in</strong>g that there is no one-to-one<br />

relationship between def<strong>in</strong>iteness and thematicity. In 8a for <strong>in</strong>stance, the def<strong>in</strong>ite NP is not<br />

thematic and the <strong>in</strong>def<strong>in</strong>ite NP is not rhematic (example from Lenerz, 1977: 46):<br />

8 Context: Wem hast du e<strong>in</strong> Buch geschenkt?<br />

8a Ich habe dem NACHbark<strong>in</strong>d e<strong>in</strong> Buch geschenkt.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 76<br />

Because <strong>of</strong> these cases, Lenerz pleads for the dist<strong>in</strong>ction [+/- known] <strong>in</strong>stead <strong>of</strong> [+/-<br />

def<strong>in</strong>ite]. Furthermore, he dist<strong>in</strong>guishes three subgroups <strong>with</strong><strong>in</strong> the <strong>in</strong>def<strong>in</strong>ite NPs, which are<br />

three possible realizations <strong>of</strong> [-def] (1977: 46ff):<br />

SPEC (spezifisch bzw. spezifiziert = [+bekannt])<br />

9 Maria kennt e<strong>in</strong>en Schornste<strong>in</strong>feger.<br />

INDEF (<strong>in</strong>def<strong>in</strong>ite = [-bekannt])<br />

10 Maria möchte e<strong>in</strong>en Schornste<strong>in</strong>feger kennenlernen, hat aber noch ke<strong>in</strong>en bestimmten im Auge.<br />

GEN ([+generic])<br />

11 Schornste<strong>in</strong>feger werden als Glücksbr<strong>in</strong>ger geschätzt.<br />

The def<strong>in</strong>ite article can have a specific as well as a generic use. The <strong>in</strong>def<strong>in</strong>ite article has<br />

even more uses.<br />

Although we agree <strong>with</strong> Lenerz' classification, we shall adopt the one suggested by Engel.<br />

The reason is, that it can be very difficult to automatically calculate whether NPs and PPs<br />

are specific, <strong>in</strong>def<strong>in</strong>ite or generic. This means that we take the three-fold classification <strong>of</strong><br />

unstressed pronouns, def<strong>in</strong>ite and <strong>in</strong>def<strong>in</strong>ite NPs as an approximation for thematicity. We<br />

hope that this will give us the right result <strong>in</strong> a lot <strong>of</strong> sentences, although we know that some<br />

cases will not be analysed correctly. The major problem concerns the classification <strong>of</strong><br />

def<strong>in</strong>ite NPs as thematic elements, because they are <strong>of</strong>ten rhematic. Indef<strong>in</strong>ite NPs and<br />

pronouns, on the other hand, are almost <strong>in</strong>variably rhematic and thematic, respectively.<br />

3.1.3 THE ORDER OF THEMATIC AND RHEMATIC COMPLEMENTS<br />

In most <strong>German</strong> sentences, the theme precedes the rheme. This has to do <strong>with</strong> two factors.<br />

The first is that, when speak<strong>in</strong>g or writ<strong>in</strong>g, we tend to start <strong>of</strong>f by referr<strong>in</strong>g to someth<strong>in</strong>g<br />

which has been mentioned earlier (the topic). By this procedure, we make a l<strong>in</strong>k between our<br />

statement and its context. It is only at this stage, that we add the new <strong>in</strong>formation (the<br />

comment).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 77<br />

Bl<strong>in</strong>kenberg (1928: 27ff), and many others s<strong>in</strong>ce, have called this, <strong>in</strong> various ways,<br />

psychological order (as opposed to grammatical order). Accord<strong>in</strong>g to the psychological<br />

order, the psychological subject ("la notion <strong>in</strong>itiale") precedes the psychological predicate,<br />

which is the goal <strong>of</strong> the statement ("le but de l'énoncé"). Bl<strong>in</strong>kenberg claims that people tend<br />

to match the grammatical and the psychological order, so as to make sentences as easily<br />

understandable as possible. Free word order languages provide better means to achieve this<br />

than fixed word order languages, but even the latter have means to shift the psychological<br />

predicate towards the end <strong>of</strong> the sentence.<br />

Accord<strong>in</strong>g to Eisenberg (1989: 401), the reward for an achieved match between surface and<br />

psychological order is improved "perceptual and cognitive processability" 51 . It seems that we<br />

take longer to understand a sentence if we do not <strong>in</strong>itially know what it refers to.<br />

Because <strong>of</strong> grammatical requirements, psychological and grammatical order do not <strong>of</strong>ten<br />

co<strong>in</strong>cide. This causes a tension, which we tend to avoid. Oliva (1991: 10) claims that our<br />

need to mention thematic elements before rhematic ones is so strong, that it can even lead to<br />

cross<strong>in</strong>g dependency, which is a strong violation <strong>of</strong> grammatical order. In long-distance<br />

scrambl<strong>in</strong>g, for <strong>in</strong>stance, as described by Becker/Joshi/Rambow (1991), a verb argument<br />

belong<strong>in</strong>g to the deeper-embedded verb precedes arguments <strong>of</strong> the higher-embedded one,<br />

because it is thematic (12b, example from Becker/Joshi/Rambow, 1991: 22):<br />

12a ... daß ich i dem Kunden [PRO i den Kühlschrank zu reparieren] versprochen habe.<br />

12b ... daß ich i [den Kühlschrank] j dem Kunden [PRO i t j zu reparieren] versprochen habe.<br />

The second fact which expla<strong>in</strong>s why thematic elements should precede rhematic ones is that<br />

the sentence focus, which is the <strong>in</strong>tonation centre <strong>of</strong> the sentence, tends to be at the end <strong>of</strong><br />

the sentence. Most l<strong>in</strong>guists seem to agree on this, not only for <strong>German</strong>, but for other<br />

languages as well 52 . As we are generally more <strong>in</strong>terested <strong>in</strong> new <strong>in</strong>formation than <strong>in</strong> what we<br />

51 "perzeptuelle und kognitive Verarbeitbarkeit"<br />

52 For <strong>German</strong>, for <strong>in</strong>stance, Rochemont (1989: 3), Fox (1990: 253), Oliva (1991: 9f), Reis (1987: 169) and<br />

Erben (1970: 275) take it for granted, that the <strong>in</strong>tonation centre is at the end <strong>of</strong> the sentence. Quirk et al.<br />

(1979: 938-955), Taglicht (1984: 172), and Ste<strong>in</strong>er/W<strong>in</strong>ter (1987: 4ff) have the same assumption for


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 78<br />

know already, and as we tend to stress what we are <strong>in</strong>terested <strong>in</strong>, it seems quite natural that<br />

new <strong>in</strong>formation and sentence focus co<strong>in</strong>cide.<br />

However, it is not clear whether the sentence focus tends to be at the end, because the new<br />

<strong>in</strong>formation tends to be located there, or whether there are <strong>in</strong>dependent (e.g. phonetic)<br />

reasons for it. An argument for the sentence focus hav<strong>in</strong>g its own right is that even fixed<br />

word order languages try to match the psychological and the grammatical orders (cf. last<br />

footnote and section 3.3.2).<br />

With this background knowledge, it is easier to understand why verb arguments <strong>in</strong> the<br />

<strong>German</strong> Mittelfeld 53 tend to have the order (Engel, 1988: 322):<br />

pronouns < def<strong>in</strong>ite NPs < <strong>in</strong>def<strong>in</strong>ite NPs<br />

Pronouns refer to known <strong>in</strong>formation and are thus thematic, <strong>in</strong>def<strong>in</strong>ite NPs <strong>of</strong>ten refer to<br />

rhematic <strong>in</strong>formation, and def<strong>in</strong>ite NPs can be both, <strong>with</strong> a tendency towards be<strong>in</strong>g thematic.<br />

Although this order applies <strong>in</strong> general, we have to add some details. Accord<strong>in</strong>g to Engel<br />

(1988: 323), the genitive pronoun tends less to the left than other pronouns, and nom<strong>in</strong>atives<br />

have a stronger tendency to the left. Pronouns tend to follow <strong>in</strong> the order N


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 79<br />

N pron /N +d < (Nom - A - D) pron < N -d < (D - A) +d < G pron < (D - A) -d < G +/-d<br />

55<br />

This means that the unmarked order <strong>of</strong> elements <strong>in</strong> the <strong>German</strong> sentence is: nom<strong>in</strong>ative<br />

complement realised as a pronoun or as a def<strong>in</strong>ite NP, pronom<strong>in</strong>al nom<strong>in</strong>al argument 56 ,<br />

pronom<strong>in</strong>al accusative and dative arguments, <strong>in</strong>def<strong>in</strong>ite nom<strong>in</strong>ative complement, def<strong>in</strong>ite<br />

dative and accusative complements, pronom<strong>in</strong>al genitive NP, <strong>in</strong>def<strong>in</strong>ite dative and<br />

accusative NP, genitive arguments <strong>in</strong>dependently from their def<strong>in</strong>iteness.<br />

The follow<strong>in</strong>g sentences correspond to this basic order (Engel, 1988: 323):<br />

14 Darum habe ich es me<strong>in</strong>er Mutter gegeben. (N pron < A pron < D +d )<br />

15 Darum hat es ihr niemand gebracht. (A pron < D pron < N -d )<br />

16 Deshalb ist unsere Freund<strong>in</strong> es bis an ihr Lebensende geblieben. (N +d < NOM pron )<br />

Heidolph/Flämig/Motsch (1981: 737) po<strong>in</strong>t out that prepositional objects (PO) always tend<br />

to the right, so that they will follow the NP arguments. We shall see below that, <strong>with</strong><strong>in</strong> the<br />

relative order <strong>of</strong> several POs, the dist<strong>in</strong>ction <strong>of</strong> pronom<strong>in</strong>al, def<strong>in</strong>ite, and <strong>in</strong>def<strong>in</strong>ite PPs<br />

nevertheless holds.<br />

We mentioned before that the preference rule concern<strong>in</strong>g TRS can be violated <strong>with</strong>out<br />

result<strong>in</strong>g <strong>in</strong> an ungrammatical sentence. Indeed, sentences <strong>with</strong> the rheme preced<strong>in</strong>g the<br />

theme exist. However, this phenomenon seems to be rare <strong>in</strong> written sentences <strong>of</strong> free word<br />

order languages (Oliva, 1991: 9f; Hajicová/Sgall/Skoumalová, 1993: 179). The reason for<br />

this discrepancy between spoken and written language is that <strong>in</strong> written language <strong>in</strong>tonation<br />

and stress are not available as means <strong>of</strong> expression, so that word order which does not<br />

55 From now on, we shall use the abbreviations +/-d for the feature values def<strong>in</strong>ite/<strong>in</strong>def<strong>in</strong>ite and +/-a for<br />

animate/<strong>in</strong>animate.<br />

56 Hoberg (1981) and Engel (1970) name this group "Nom<strong>in</strong>alergänzung". Unlike Engel, Hoberg (1981: 85)<br />

not only <strong>in</strong>cludes nom<strong>in</strong>ative (13a) and accusative predicative nouns (13b), but also prepositional<br />

predicative elements (13c):<br />

13a Das ist e<strong>in</strong> übler Scherz.<br />

13b Das nenne ich e<strong>in</strong>en üblen Scherz.<br />

13c Das halte ich für e<strong>in</strong>en üblen Scherz.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 80<br />

correspond to the TRS would be confus<strong>in</strong>g and unnatural. Siewierska (1988: 85) mentions a<br />

further reason why spoken language can differ from written utterances:<br />

The readily accessible <strong>in</strong>formation is not <strong>of</strong> prime <strong>in</strong>terest to the speaker, his aim be<strong>in</strong>g to convey some<br />

salient <strong>in</strong>formation which has not yet been activated - the focus <strong>of</strong> the utterance. [...] Consequently, the<br />

salient <strong>in</strong>formation could be expected to be placed prior to the more accessible material. [...] The shared<br />

situational context and <strong>of</strong>ten actual knowledge <strong>of</strong> each other's experiences, wants and expectations, plus<br />

the possibility <strong>of</strong> immediate feedback, reduce the need for stat<strong>in</strong>g a frame for the <strong>in</strong>terpretation <strong>of</strong> the<br />

utterance.<br />

Accord<strong>in</strong>g to Engel (1970: 15), the follow<strong>in</strong>g sentences contradict the TRS, as <strong>in</strong> 17 there is<br />

no theme, <strong>in</strong> 18 drei Männer and <strong>in</strong> 19 kamen are the themes:<br />

17 Es regnet.<br />

18 Es ritten drei Männer zum Tore h<strong>in</strong>aus.<br />

19 Da kamen drei Direktoren.<br />

With respect to 17, we agree <strong>with</strong> Engel, although we do not th<strong>in</strong>k the fact that the whole<br />

sentence is rhematic constitutes any contradiction. The TRS just does not apply to 17. We<br />

disagree, however, concern<strong>in</strong>g the explanation <strong>of</strong> the second and the third sentences: The<br />

existence <strong>of</strong> the expletive es <strong>in</strong> the Vorfeld <strong>in</strong> 18 shows that there is no theme <strong>in</strong> the<br />

sentence. It is for the same reason that the prototypical <strong>German</strong> fairy tale starts <strong>of</strong>f <strong>with</strong> the<br />

expression "Es war e<strong>in</strong>mal ...", namely because there is no theme at the very beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong><br />

the text. In 19, the temporal or causal adverb da is thematic. It expresses that the whole<br />

proposition is consequent upon what has been said before. None <strong>of</strong> these sentences<br />

contradict the TRS.<br />

3.1.4 THE SEPARATION OF THEME AND RHEME BY MODIFIERS<br />

Several authors mention that modifiers separate theme and rheme but they differ <strong>with</strong> respect<br />

to the subclasses, to which they attribute this task. Accord<strong>in</strong>g to Zemb (1968: 111) and<br />

Eroms (1986: 19ff), it is the negation (20b). Accord<strong>in</strong>g to Engel (1988: 340) and Hoberg<br />

(1981: 153), the situative modifiers have this function. Thurmair (1989: 29ff) suspects the<br />

toners and Waltz<strong>in</strong>g (1986: 128ff) attributes this quality to the whole group <strong>of</strong> evaluative<br />

modifiers. When the modifiers are adjacent, these statements are not contradictory (20c), but<br />

they are sometimes separated (21):


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 81<br />

20a Er ist per Anhalter nach Katmandu gefahren.<br />

20b Er ist nicht per Anhalter nach Katmandu gefahren.<br />

20c Er ist damals sit wohl toner vermutlich exist nicht per Anhalter nach Katmandu gefahren.<br />

21 Er hat gestern sit dem Mann wohl exist wieder sit nicht neg geholfen.<br />

Waltz<strong>in</strong>g (1986: 128ff) argues that the negation nicht, when a sentence modifier, cannot<br />

move <strong>in</strong> a flexible way, and therefore it is not likely to move when dependent on the context.<br />

22a confirms his doubt, as nicht def<strong>in</strong>itely follows the rhematic elements se<strong>in</strong>e Eltern,<br />

gestern and e<strong>in</strong>en Wunsch:<br />

22 Context: Warum ist Peter heute so traurig?<br />

22a Er ist traurig, weil ihm se<strong>in</strong>e Eltern gestern e<strong>in</strong>en Wunsch nicht erfüllt haben.<br />

Situative modifiers do have more flexibility but they are an important part <strong>of</strong> the proposition,<br />

and can therefore be thematic or rhematic themselves. This makes the borderl<strong>in</strong>e between<br />

old and new <strong>in</strong>formation unclear. Waltz<strong>in</strong>g's (1986: 129) argument for the evaluative<br />

modifiers as separators is that they are not really elements <strong>of</strong> the sentence itself, but<br />

represent performative (hyper-) verbs. They represent the speaker's op<strong>in</strong>ion and cannot be<br />

stressed. Therefore, they cannot be part <strong>of</strong> the rheme and allow for a much more precise<br />

borderl<strong>in</strong>e between old and new:<br />

Nichts ist schließlich naheliegender, als daß der Sprecher gerade die Elemente, mit denen er se<strong>in</strong>e<br />

Stellungnahme zum verbalisierten Sachverhalt ausdrückt, gleichzeitig zur Markierung der von ihm<br />

<strong>in</strong>tendierten kommunikativen Gewichtung, der Verteilung von Thema und Rhema <strong>in</strong>nerhalb e<strong>in</strong>er<br />

Äußerung benutzt. (1986: 129f)<br />

The follow<strong>in</strong>g examples show that situative as well as existimatorial modifiers can have this<br />

separat<strong>in</strong>g function. Neither 4a nor 4b are good answers to question 23, which asks for the<br />

dative NP se<strong>in</strong>em Onkel. Instead, 23c is the only satisfy<strong>in</strong>g answer, as the dative<br />

complement follows both adverbs:<br />

23 Context: Wem hat er gestern zu viel We<strong>in</strong> gegeben?<br />

23a ?? Er hat gestern sit se<strong>in</strong>em ONKEL wohl exist zu viel We<strong>in</strong> gegeben.<br />

23b ? Er hat wohl exist se<strong>in</strong>em ONKEL gestern sit zu viel We<strong>in</strong> gegeben.<br />

23c Er hat wohl gestern/gestern wohl se<strong>in</strong>em ONKEL zu viel We<strong>in</strong> gegeben.<br />

23b seems to be slightly more acceptable than 23a, but 23c def<strong>in</strong>itely is the only appropriate<br />

answer. Consequently, we can say that elements <strong>of</strong> both modifier groups can separate theme<br />

and rheme. When there is more than one modifier, the last one is relevant. If it is a situative


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 82<br />

modifier, we cannot decide whether the modifier itself is thematic or rhematic <strong>with</strong>out<br />

recurr<strong>in</strong>g to the context. We shall come back to the separat<strong>in</strong>g function <strong>of</strong> modifiers <strong>in</strong><br />

section 3.3, which handles functional sentence perspective.<br />

3.2 BEHAGHEL'S "GESETZ DER WACHSENDEN GLIEDER"<br />

The law <strong>of</strong> the <strong>in</strong>creas<strong>in</strong>g constituents goes back to Behaghel (1932: 5f) and is known <strong>in</strong><br />

English literature as the heav<strong>in</strong>ess pr<strong>in</strong>ciple. It states that the longer the constituents are, the<br />

later they serialise <strong>in</strong> the sentence. This does not seem to be a very rigid rule, as it applies for<br />

some sentences (24), but not for the majority <strong>of</strong> others (25, 26):<br />

24 Gestern kam endlich die ersehnte Nachricht.<br />

25 Ich esse me<strong>in</strong>e Suppe nicht.<br />

26 Er kehrte mit unermeßlichen Schätzen heim.<br />

The negational modifier nicht <strong>in</strong> 25 is less heavy than the accusative NP, but cannot be<br />

moved <strong>in</strong> front <strong>of</strong> the complement <strong>with</strong>out a considerable change <strong>of</strong> mean<strong>in</strong>g. The separable<br />

verb prefix heim cannot be moved <strong>in</strong> front <strong>of</strong> the sentence-modify<strong>in</strong>g PP at all, although it is<br />

much shorter than the PP. Behaghel's law only applies to the order <strong>of</strong> some constituents. It<br />

should be reduced to the statement that pronouns tend to precede full NPs and that, <strong>in</strong> some<br />

cases, longer NPs, which are extended by an attribute, tend to follow shorter NPs (Engel,<br />

1970: 14):<br />

27a Paul hat der Frau den langen Brief über moderne Kleidung gegeben.<br />

27b ? Paul hat den langen Brief über moderne Kleidung der Frau gegeben.<br />

28a Paul hat den Brief der langen Frau <strong>in</strong> moderner Kleidung gegeben.<br />

28b ? Paul hat der langen Frau <strong>in</strong> moderner Kleidung den Brief gegeben.<br />

Eisenberg (1989: 401), aga<strong>in</strong>, expla<strong>in</strong>s Behaghel's law by the easier perceptual and cognitive<br />

process<strong>in</strong>g <strong>of</strong> sentences which conform to the law. For the same reason, relative clauses<br />

<strong>of</strong>ten are extraposed to the end <strong>of</strong> the sentence (Engel, 1988: 333f).<br />

Behaghel's Gesetz der wachsenden Glieder is closely related to the TRS as heavy NPs are<br />

mostly rhematic. Attributes to nouns generally carry new <strong>in</strong>formation, and thematic elements<br />

do not normally need to be specified. We nevertheless list the pr<strong>in</strong>ciple <strong>in</strong>dependently,


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 83<br />

because heav<strong>in</strong>ess might be measurable and this could be <strong>of</strong> use for the automatic treatment<br />

<strong>of</strong> language. A def<strong>in</strong>ite NP, for example, is typically thematic, but there is a greater chance<br />

that it is rhematic <strong>with</strong> an attribute rather than <strong>with</strong>out.<br />

3.3 FUNCTIONAL SENTENCE PERSPECTIVE<br />

Für die Bestimmung e<strong>in</strong>er Normalfolge von Ergänzungen im Mittelfeld spielen<br />

statt morphosyntaktischer viel eher pragmatische, das heißt sachverhalts- und<br />

sprecherbezogene Faktoren e<strong>in</strong>e Rolle. (Lötscher, 1981: 44)<br />

The term functional sentence perspective (FSP) goes back to the Prague l<strong>in</strong>guist Vilém<br />

Mathesius (1929). It is an extremely important factor for <strong>German</strong>, and is very closely related<br />

to the theme-rheme relationship and the heav<strong>in</strong>ess <strong>of</strong> phrases. The pr<strong>in</strong>ciple requires that not<br />

only should new <strong>in</strong>formation follow known <strong>in</strong>formation, but also that elements should<br />

serialise accord<strong>in</strong>g to their <strong>in</strong>formational value. Speakers will generally first mention the<br />

facts they consider less important, so that the facts they want to stress appear at the end <strong>of</strong><br />

the sentence. We mentioned before that the sentence focus normally falls on the end <strong>of</strong> the<br />

sentence. Accord<strong>in</strong>g to Hajicová/Sgall/Skoumalová (1993: 179), a sentence focus not be<strong>in</strong>g<br />

at the end <strong>of</strong> the sentence is rare <strong>in</strong> written language. Thus, the important parts <strong>of</strong> the<br />

statement would be stressed automatically.<br />

The FSP <strong>in</strong>cludes the TRS, because new <strong>in</strong>formation is normally more important than old<br />

<strong>in</strong>formation. The two pr<strong>in</strong>ciples are however not identical, as we see <strong>in</strong> 29a:<br />

29a Er wird es morgen IHM geben.<br />

29b * Er wird es morgen ihm GEben.<br />

29c Er wird es ihm morgen geben.<br />

29d Er wird es morgen IHM geben, und nicht IHR.<br />

The pronoun ihm refers necessarily to known <strong>in</strong>formation, as otherwise the use <strong>of</strong> the<br />

pronoun would not be possible. However, 29b shows that ihm <strong>in</strong> a is the only element, which<br />

can carry stress. The stress is so strong, that 29a can only be understood by contrast<strong>in</strong>g the<br />

pronoun ihm to another pronoun such as ihr (29d). In the more natural order <strong>of</strong> 29c, all<br />

elements but es can carry the ma<strong>in</strong> accent, but either morgen or geben are the most likely to


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 84<br />

be stressed. The dichotomy known versus new <strong>in</strong>formation fails to expla<strong>in</strong> 29. Another term,<br />

referr<strong>in</strong>g to communicative importance, is required.<br />

In addition to the old/new dist<strong>in</strong>ction, the FSP represents the <strong>in</strong>tention <strong>of</strong> the speaker or<br />

writer. This pr<strong>in</strong>ciple can override the theme-rheme structure as a thematic pronoun can have<br />

the largest <strong>in</strong>formational content, and therefore goes to the end <strong>of</strong> the sentence. Although<br />

this pr<strong>in</strong>ciple seems more important than the TRS, and <strong>in</strong> any case <strong>in</strong>cludes the latter, we<br />

have to mention them separately. The reason for this is that theme and rheme can be<br />

recognised approximately by the form <strong>of</strong> the arguments (cf. 3.1.2), whereas the<br />

<strong>in</strong>formational value, l<strong>in</strong>ked to the <strong>in</strong>tention <strong>of</strong> the speaker or writer, cannot.<br />

3.3.1 THEMATISATION AND RHEMATISATION<br />

FSP expla<strong>in</strong>s what has been called thematization and rhematization (Engel, 1988: 340f).<br />

Thematization <strong>of</strong> elements means assign<strong>in</strong>g little <strong>in</strong>formational value to them, and<br />

rhematization means focus<strong>in</strong>g on them.<br />

These terms are slightly mislead<strong>in</strong>g, as the thematicity <strong>of</strong> elements does not change <strong>with</strong> the<br />

writer's attitude towards them. What the writer can do, however, is to assign relative<br />

importance to some elements. In spite <strong>of</strong> the slight imprecision <strong>of</strong> the terms, we shall keep<br />

us<strong>in</strong>g them, as they seem to be well-accepted <strong>in</strong> the literature. Foley and Van Val<strong>in</strong> (1985)<br />

use the more fortunate term Information Packag<strong>in</strong>g. However, it <strong>in</strong>cludes many more<br />

devices than just thematization and rhematization (cf. 3.3.2).<br />

As elements <strong>with</strong> little <strong>in</strong>formational value tend to precede elements <strong>with</strong> high content, one<br />

can actively thematise elements by plac<strong>in</strong>g them at the left. Rhematization is the reverse<br />

procedure: the speaker places subjectively important <strong>in</strong>formation at the right. This is shown<br />

<strong>in</strong> 30. 30a and 30b are similar sentences, which only differ <strong>in</strong> that different phrases are<br />

thematised and rhematised:<br />

30a Die K<strong>in</strong>der sollten während des Gewitters ihre Zimmer aufräumen.<br />


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 85<br />

30b Ihre Zimmer sollten die K<strong>in</strong>der während des Gewitters aufräumen.<br />

<br />

Thematization and rhematization are frequently used <strong>in</strong> newspaper articles:<br />

31 Gestern Nachmittag um 15.30 Uhr <strong>in</strong> | hat e<strong>in</strong>e Unbekannte<br />

der Balanstraße | e<strong>in</strong>en Mann geohrfeigt.<br />

<br />

Even if time and place <strong>in</strong> 31 are not thematic (known), they can be represented as<br />

background <strong>in</strong>formation to the ma<strong>in</strong> action. Phrases can be thematised artificially,<br />

<strong>in</strong>dependently from whether they are realised as pr<strong>of</strong>orms (dort, bei ihr) or full forms (letzte<br />

Woche, bei Tante Emma) (Heidolph/Flämig/Motsch, 1981: 736).<br />

As some modifiers separate theme and rheme (3.1.4), their movement to the left or to the<br />

right changes the theme-rheme structure. In 32, the normal position <strong>of</strong> the adverb gestern<br />

(not <strong>in</strong>volv<strong>in</strong>g rhematization) would be as <strong>in</strong> 32b (Hoberg, 1981: 149). When gestern<br />

precedes the dative complement (32a), the dative NP dem Verleger Axel Spr<strong>in</strong>ger is<br />

rhematised. The difference between 32a and 32b is very little, as the number <strong>of</strong> potential<br />

rhematic elements is nevertheless quite large <strong>in</strong> (32b). The movement <strong>of</strong> gestern to the right<br />

(32c) has a more dist<strong>in</strong>ctive effect. Only the modifier itself and the ma<strong>in</strong> verb can be<br />

rhematic (example from Hoberg, 1981: 154):<br />

32a Bundespräsident He<strong>in</strong>rich Lübke hat gestern ... dem Verleger Axel Spr<strong>in</strong>ger das große<br />

Verdienstkreuz ... überreicht.<br />

32b Bundespräsident He<strong>in</strong>rich Lübke hat ... dem Verleger Axel Spr<strong>in</strong>ger gestern das große<br />

Verdienstkreuz ... überreicht.<br />

32c Bundespräsident He<strong>in</strong>rich Lübke hat ... dem Verleger Axel Spr<strong>in</strong>ger das große<br />

Verdienstkreuz ... gestern überreicht.<br />

3.3.2 FURTHER MEANS TO EXPRESS FUNCTIONAL SENTENCE<br />

PERSPECTIVE<br />

<strong>Word</strong> order variation, as discussed <strong>in</strong> 3.3.1, is a means available to free word order<br />

languages only. The question arises, what can languages <strong>with</strong> more or less fixed word order<br />

do to express theme, rheme and the relative <strong>in</strong>formational value <strong>of</strong> elements. Do languages


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 86<br />

such as English and French actually express it, or do they use the same order <strong>of</strong> words for<br />

different contexts?<br />

At least for the position<strong>in</strong>g <strong>of</strong> modifiers, English and French have some choice.<br />

Complements, however, are only recognised by their position so that a change <strong>of</strong> position<br />

would consequently lead to a change <strong>of</strong> roles 57 :<br />

34a The man gave the dog bones.<br />

34b * The dog gave the man bones. (as an equivalent <strong>of</strong> a)<br />

34c * Bones gave the dog the man. (as an equivalent <strong>of</strong> a)<br />

There are a few exceptions to this, which are generally called subject-verb <strong>in</strong>version (35, 36)<br />

(Quirk et al., 1979: 948ff). They <strong>of</strong>ten belong to literary style. Declarative sentences<br />

beg<strong>in</strong>n<strong>in</strong>g <strong>with</strong> a verb such as 37 seem to be hardly possible <strong>in</strong> English (37 is chosen from<br />

the expla<strong>in</strong><strong>in</strong>g text <strong>in</strong> Wagner/P<strong>in</strong>chon, 1962: 393):<br />

35 Here comes the bus.<br />

36 Under no circumstances must the switch be left on.<br />

37 Appartiennent à cette catégorie les adverbes circonstanciels (de temps ou de lieu) et les<br />

adverbes par lesquels un locuteur exprime son op<strong>in</strong>ion sur D.<br />

(Belong to this category circumstantial adverbs (<strong>of</strong> time and place) and the adverbs by<br />

which a speaker expresses his/her op<strong>in</strong>ion on D.)<br />

(Circumstantial adverbs (<strong>of</strong> time and place) and adverbs by which speakers express their<br />

op<strong>in</strong>ion on D belong to this category)<br />

In fixed, as well as free word order languages, functional sentence perspective can <strong>of</strong>ten be<br />

expressed by other means than word order variation. Accord<strong>in</strong>g to Lenerz (1977: 15f), these<br />

<strong>in</strong>clude passive, cleft<strong>in</strong>g, change <strong>of</strong> the <strong>in</strong>tonation peak <strong>in</strong> spoken language and sometimes<br />

morphological means, as <strong>in</strong> Japanese. To these, we would like to add the syntactical means<br />

<strong>of</strong> topicalisation, dislocation, pseudo-cleft<strong>in</strong>g, as well as lexical means. For an overview <strong>of</strong><br />

universally available packag<strong>in</strong>g devices see Foley/Van Val<strong>in</strong> (1985).<br />

57 A similar th<strong>in</strong>g happens <strong>in</strong> <strong>German</strong> when case mark<strong>in</strong>g fails and when the context does not help <strong>with</strong> the<br />

disambiguation <strong>of</strong> the read<strong>in</strong>gs. In this case, there is a strong tendency for the reader to assume a normal<br />

default word order (Höhle, 1982: 130). In contextless sentences <strong>with</strong>out case-mark<strong>in</strong>g such as 33a and<br />

33b below, one is thus tempted to read both sentences <strong>with</strong> the subject verb object order:<br />

33a Stefan liebt Marie.<br />

33b Marie liebt Stefan.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 87<br />

In addition to the permutation <strong>of</strong> elements <strong>in</strong> the sentence, <strong>German</strong> has access to the means<br />

listed <strong>in</strong> the preced<strong>in</strong>g paragraph as well, although some <strong>of</strong> them are less natural than <strong>in</strong><br />

English or French. Explicit morphological theme mark<strong>in</strong>g, as <strong>with</strong> the Japanese marker wa,<br />

is not at all available. Instead, some verbs allow the use <strong>of</strong> prefixes to change the theme<br />

rheme perspective <strong>in</strong> <strong>German</strong>.<br />

We shall briefly discuss the relation between the listed constructions and FSP, and shall<br />

compare them <strong>with</strong> the use <strong>in</strong> English, French and Spanish. The word order <strong>of</strong> the former<br />

two is very restricted. Spanish (as well as Italian) does not have case-mark<strong>in</strong>g, but<br />

nevertheless has relatively free word order.<br />

An essential prerequisite for this discussion is the fact that, <strong>in</strong> these languages, the sentence<br />

focus also tends to be at the end <strong>of</strong> the sentence, and the topic has a propensity to stand<br />

sentence-<strong>in</strong>itial (cf. footnote 51 <strong>in</strong> 3.1.3).<br />

In passive sentences, the rhematic deep subject stands at the end. In this position, it cooccurs<br />

<strong>with</strong> the sentence focus (38).<br />

38 Context: Wer hat dieses Haus gebaut.<br />

38a Dieses Haus wurde von Gaudí gebaut.<br />

38b This house was built by Gaudí.<br />

38c Cette maison a été construite par Gaudí.<br />

38d Esta casa fue construida por Gaudí. 58<br />

Cleft constructions are used to position focused elements at the very beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> the<br />

sentence. They are particularly stressed <strong>in</strong> this position (Quirk et al., 1979: 951ff) but this is<br />

l<strong>in</strong>ked much more to the mark<strong>in</strong>g by the cleft<strong>in</strong>g construction "es ist X, der" than by the<br />

position at the head <strong>of</strong> the sentence. Although cleft<strong>in</strong>g is possible <strong>in</strong> <strong>German</strong>, it seems to be<br />

less frequently used than <strong>in</strong> Romance languages and English:<br />

58 Although this sentence is perfectly well-formed, Spanish speakers would rather express it by us<strong>in</strong>g<br />

dislocation:<br />

38e Esta casa, la construió Gaudí.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 88<br />

39 Context: Wer hat das Haus gebaut.<br />

39a Gaudí war es, der das Haus gebaut hat.<br />

39b Es war Gaudí, der das Haus gebaut hat.<br />

39c Das Haus hat Gaudí gebaut.<br />

39d It was Gaudí who built the house.<br />

39e C'était Gaudí qui a construit cette maison.<br />

39f Era Gaudí que construió la casa.<br />

Topicalisation is a means to place the topic, and thus a thematic element, at the head <strong>of</strong> the<br />

sentence. This results <strong>in</strong> a relative right movement <strong>of</strong> the rhematic elements. In <strong>German</strong>,<br />

topicalisation <strong>of</strong> the English type is not possible (40c). In English, both the topic and the<br />

subject precede the verb, whereas <strong>in</strong> <strong>German</strong> only one <strong>of</strong> them can precede the verb, when it<br />

is <strong>in</strong> verb second position. This object or modifier-front<strong>in</strong>g construction is very common <strong>in</strong><br />

<strong>German</strong>. Accord<strong>in</strong>g to Gadler's (1982: 163f) small Austrian newspaper corpus, four out <strong>of</strong><br />

five topicalised elements are adverbials (example a from Quirk et al., 1979: 946):<br />

40a Really good cocktails they made at that hotel.<br />

40b Wirklich gute Cocktails mixten sie <strong>in</strong> diesem Hotel.<br />

40c * Wirklich gute Cocktails sie mixten <strong>in</strong> diesem Hotel.<br />

40d Vraiment, de bons cocktails ils faisaient à cet hotel.<br />

40e Realmente buenos cocteles hicieron en ese hotel.<br />

Dislocation has the same effect as topicalisation, namely the relative right movement <strong>of</strong><br />

rhematic elements towards the end <strong>of</strong> the sentence. In contrast to topicalisation, the<br />

dislocated element is picked up by a pronoun. Spanish and French <strong>of</strong>ten use dislocation as an<br />

equivalent <strong>of</strong> the English topicalisation construction.<br />

41a Die Agnès, die werde ich morgen treffen.<br />

41b Agnès, I shall meet her tomorrow.<br />

41c Agnès, je la reverrai dema<strong>in</strong>.<br />

41d Agnès, la vedré otra vez mañana.<br />

By us<strong>in</strong>g the pseudo-cleft structure, a rhematic subject can be moved right <strong>of</strong> the theme so<br />

that it receives the sentence focus. Although pseudo-cleft<strong>in</strong>g is possible <strong>in</strong> <strong>German</strong>, there is<br />

no need for this structure and it is therefore not very common:<br />

42 Context: Was ärgert dich?<br />

42a Was mich ärgert, ist der Lärm.<br />

42b Mich ärgert der Lärm.<br />

42c What is bother<strong>in</strong>g me is the noise.<br />

42d Ce qui m'ennuie c'est le bruit.<br />

42e Lo que me molesta es el rumor.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 89<br />

In spoken language, the <strong>in</strong>tonation peak is an explicit means <strong>of</strong> focus<strong>in</strong>g a particular element<br />

<strong>in</strong> the sentence. In <strong>German</strong> and English, as well as <strong>in</strong> the Romance languages, even phrases<br />

at the beg<strong>in</strong>n<strong>in</strong>g or <strong>in</strong> the middle <strong>of</strong> the sentence can be stressed. When a non-f<strong>in</strong>al element<br />

is stressed, the elements follow<strong>in</strong>g the focus are always contextually bound (thematic)<br />

(Hajicová/Sgall/Skoumalová, 1993:179). When non-f<strong>in</strong>al focus occurs, this is normally<br />

required by the context:<br />

43 Context: Who is it who will cook tomorrow?<br />

43a ALAN will cook tomorrow.<br />

43b ! Alan will COOK tomorrow.<br />

43c ! Alan will cook toMORrow.<br />

In Japanese, a specific particle (wa) exists to mark the topic:<br />

44 Kiyono-wa watashi-tachi-no tomodachi desu.<br />

Kiyono-topic 1stpers-plur -Gen/Poss friend be<br />

(Kiyono is our friend)<br />

This means is not available <strong>in</strong> any <strong>of</strong> the mentioned European languages. However, by us<strong>in</strong>g<br />

the <strong>German</strong> verb prefix ver- <strong>with</strong> a limited set <strong>of</strong> verbs, the semantic roles <strong>of</strong> verb arguments<br />

can be changed for repackag<strong>in</strong>g purposes (Foley/Van Val<strong>in</strong>, 1985: 294ff). This perspective<br />

change is not a productive morphological means:<br />

45a Ralf kaufte e<strong>in</strong> Buch von Harold.<br />

45b Harold verkaufte Ralf e<strong>in</strong> Buch.<br />

The same conversive relation exists <strong>in</strong> other languages on a lexical level (46a/46b, 47, 48). It<br />

has identical consequences for the sentence perspective (Foley/Van Val<strong>in</strong>, 1985: 291ff).<br />

Note that English has the additional means <strong>of</strong> dative shift (Foley/Van Val<strong>in</strong>, 1985: 347ff).<br />

When dative shift is used, the sentence focus is on the direct object (46c), as opposed to on<br />

the <strong>in</strong>direct object (46b):<br />

46a Ralf bought a book from Harold.<br />

46b Harold sold a book to Ralf.<br />

46c Harold sold Ralf a book.<br />

47a Ralf a acheté un livre de Harold.<br />

47b Harold a vendu un livre a Ralf.<br />

48a Ralf compró un libro de Harold.<br />

48b Harold vendí un libro a Ralf


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 90<br />

Oliva (1991: 10f) claims that even the phenomenon <strong>of</strong> rais<strong>in</strong>g can be expla<strong>in</strong>ed by our<br />

tendency to avoid the:<br />

[...] tension between syntactic requirements such as (multiple) subcategorization or (especially <strong>in</strong> the<br />

case <strong>of</strong> <strong>German</strong>) word order constra<strong>in</strong>ts imposed onto the govern<strong>in</strong>g elements by the grammar on one<br />

hand and communicative dynamism <strong>of</strong> the subcategorised elements on the other hand.<br />

This actually br<strong>in</strong>gs us very near to answer<strong>in</strong>g the question <strong>of</strong> raison d'être <strong>of</strong> the phenomenon <strong>of</strong><br />

rais<strong>in</strong>g: primarily, it is but a means <strong>of</strong> remov<strong>in</strong>g this tension from the structure <strong>of</strong> the sentence.<br />

Rais<strong>in</strong>g structures only change the topic position, but not the position <strong>of</strong> the sentence focus.<br />

If we assume that Paul is the topic <strong>in</strong> 49, the rais<strong>in</strong>g construction <strong>in</strong> 49b decreases the<br />

tension between the tendency <strong>of</strong> the topic to move <strong>in</strong>to sentence-<strong>in</strong>itial position and the<br />

syntactic requirement, which is that the nom<strong>in</strong>ative complement <strong>of</strong> kaufen has to be <strong>in</strong> the<br />

middle <strong>of</strong> the sentence:<br />

49a Es sche<strong>in</strong>t, daß Paul auch e<strong>in</strong> Buch von Harold kauft.<br />

49b Paul sche<strong>in</strong>t auch e<strong>in</strong> Buch von Harold zu kaufen.<br />

We have seen that a lot <strong>of</strong> l<strong>in</strong>guistic constructions have the same goal, namely allow<strong>in</strong>g us to<br />

order <strong>in</strong>formation as required by the context, or accord<strong>in</strong>g to our <strong>in</strong>tention. We want to po<strong>in</strong>t<br />

out that some <strong>of</strong> these means can entail other consequences, such as the change <strong>of</strong> quantifier<br />

scope. In 50, for <strong>in</strong>stance, the active and the passive sentences have different mean<strong>in</strong>gs<br />

(example from Primus, 1987: 61):<br />

50a Jeder Mann küsst e<strong>in</strong>e Frau.<br />

50b E<strong>in</strong>e Frau wird von jedem Mann geküßt.<br />

3.4. VERBNÄHE<br />

Another parameter, which has a great <strong>in</strong>fluence on the order <strong>of</strong> elements <strong>in</strong> the <strong>German</strong><br />

clause is what has been called semantic-syntactic closeness <strong>of</strong> elements to the verb, or<br />

Verbnähe. In the English literature, the expressions verb bond<strong>in</strong>g (e.g. Toml<strong>in</strong>, 1986: 73ff)<br />

and obliqueness (e.g. Pollard/Sag, 1987: 174ff) are used but these terms are <strong>of</strong>ten limited to<br />

the verb object relation. We shall use the term verb bond<strong>in</strong>g synonymously <strong>with</strong> Verbnähe<br />

and semantic-syntactic closeness.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 91<br />

Behaghel (1932: 4) formulates <strong>in</strong> his first law concern<strong>in</strong>g the order <strong>of</strong> elements <strong>in</strong> <strong>German</strong>:<br />

Das oberste Gesetz ist dieses, daß das geistig eng zusammengehörige auch eng zusammengestellt wird.<br />

Behaghel's law is <strong>of</strong> a very general nature and is not restricted to the order <strong>of</strong> constituents <strong>in</strong><br />

the sentence. Ursula Hoberg (1981: 63) specifies the closeness pr<strong>in</strong>ciple for the sentence<br />

level:<br />

Je enger die strukturelle Relation, desto enger auch die positionelle Relation zum Verb (<strong>in</strong> Endstellung).<br />

This pr<strong>in</strong>ciple thus says that there is a tendency <strong>in</strong> <strong>German</strong> to express the semantic-syntactic<br />

closeness between the verb and its arguments and modifiers topologically. Verb arguments,<br />

for <strong>in</strong>stance, which have a strong semantic relationship <strong>with</strong> the verb tend to be situated at<br />

the end <strong>of</strong> the middle field. In this position, they are adjacent to the ma<strong>in</strong> verb <strong>in</strong> verb-f<strong>in</strong>al<br />

sentences. Conversely, elements which are not semantically close to the verb tend to the left,<br />

far away from the verb <strong>in</strong> clause-f<strong>in</strong>al position.<br />

It is important to po<strong>in</strong>t out that the sequence <strong>of</strong> verb arguments and modifiers tends to be the<br />

same <strong>in</strong> verb-second and verb-<strong>in</strong>itial sentences. If we accept the verb bond<strong>in</strong>g pr<strong>in</strong>ciple as<br />

stated here - and there are good reasons to do so - it is an argument for the assumption that<br />

the underly<strong>in</strong>g word order <strong>of</strong> <strong>German</strong> is verb-f<strong>in</strong>al (cf. 1.1).<br />

Accord<strong>in</strong>g to Hoberg (1981: 63), the only criterion to f<strong>in</strong>d out which elements are<br />

semantically close to the verb is the substitution test:<br />

Je ger<strong>in</strong>ger die syntaktischen und semantischen Substitutionsmöglichkeiten für e<strong>in</strong> Element s<strong>in</strong>d, desto<br />

enger ist se<strong>in</strong>e B<strong>in</strong>dung ans Verb, und umgekehrt: Je größer se<strong>in</strong>e Substituierbarkeit ist, desto ferner<br />

steht es dem Verb.<br />

Am deutlichsten ist dieser Zusammenhang bei Funktionsnomen: Sie s<strong>in</strong>d nicht pronom<strong>in</strong>alisierbar<br />

(Letzte Woche rettete er e<strong>in</strong>em alten Mann das Leben. * Gestern hat er es e<strong>in</strong>em Mädchen gerettet.),<br />

nicht durch e<strong>in</strong>en Gliedsatz ersetzbar, <strong>in</strong> der Regel auf e<strong>in</strong>e bestimmte Artikelart (def<strong>in</strong>ite, <strong>in</strong>def<strong>in</strong>ite<br />

oder ohne Artikel) festgelegt, nur beschränkt attribuierbar und jeweils auf e<strong>in</strong>ige wenige Lexeme (e<strong>in</strong><br />

und desselben Wortfeldes - z.B. Gedanke, Idee, ...) beschränkt, s<strong>of</strong>ern überhaupt e<strong>in</strong>e Wahl <strong>in</strong> den<br />

Lexemen besteht. Oft ist nur das gesamte Funktionsverbgefüge, also FN [Funktionsnomen, RS] + Verb,<br />

substituierbar, und zwar gegen e<strong>in</strong> e<strong>in</strong>faches Verb. (italics <strong>in</strong> the orig<strong>in</strong>al)<br />

Toml<strong>in</strong> (1986), who looks at order<strong>in</strong>g pr<strong>in</strong>ciples from a universal po<strong>in</strong>t <strong>of</strong> view, specifies<br />

that verb object bond<strong>in</strong>g "is not to be taken as an absolute language universal, but rather as


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 92<br />

another fundamental pr<strong>in</strong>ciple that may shape natural languages" (p. 74). He describes the<br />

effect <strong>of</strong> the pr<strong>in</strong>ciple <strong>in</strong> the follow<strong>in</strong>g way (1986: 73):<br />

<strong>in</strong> general <strong>in</strong> transitive clauses it is more difficult to <strong>in</strong>terfere <strong>with</strong> the syntactic juxtaposition and<br />

semantic unity <strong>of</strong> the verb and object than it is to <strong>in</strong>terfere <strong>with</strong> that <strong>of</strong> the verb and subject. Various<br />

<strong>in</strong>dependent syntactic, semantic, and even phonological processes appear to conspire to prevent the<br />

separation <strong>of</strong> the object from the verb; and these same processes <strong>of</strong>ten permit separation <strong>of</strong> the subject<br />

from the verb <strong>in</strong> order to ma<strong>in</strong>ta<strong>in</strong> the bond between the verb and object.<br />

We want to extend the term bond<strong>in</strong>g to elements which are not verb arguments (similarly<br />

Pollard/Sag, 1987: 181).<br />

Pollard and Sag (1987: 175) differ from Hoberg <strong>in</strong> that they mention four different k<strong>in</strong>ds <strong>of</strong><br />

evidence for obliqueness: control, b<strong>in</strong>d<strong>in</strong>g, passivisation and agreement. Regard<strong>in</strong>g the first<br />

phenomenon, they state:<br />

The theory <strong>of</strong> control [...] requires that the unexpressed subject <strong>of</strong> a VP complement must be controlled<br />

by a less oblique complement [...]. [...], the theory <strong>of</strong> control implies that these elements [the controllers,<br />

RS] are less oblique than their VP[INF] sisters.<br />

Their po<strong>in</strong>t is that the <strong>in</strong>f<strong>in</strong>itival verb arguments as a whole must be more oblique than the<br />

controller <strong>of</strong> the non-realised subject PRO <strong>of</strong> the VP, namely the direct and the <strong>in</strong>direct<br />

objects <strong>in</strong> 51 and 52 (examples by Pollard and Sag):<br />

51 He persuaded Kim i PRO i to cooperate.<br />

52 He appealed to Lou i PRO i to be polite.<br />

Pollard and Sag are not very explicit on the relation between b<strong>in</strong>d<strong>in</strong>g and obliqueness. They<br />

refer to the second, still forthcom<strong>in</strong>g volume <strong>in</strong>stead.<br />

Similarly, the b<strong>in</strong>d<strong>in</strong>g theory [...] expla<strong>in</strong>s familiar constra<strong>in</strong>ts on the distribution <strong>of</strong> reflexive and<br />

nonreflexive pronouns <strong>in</strong> terms <strong>of</strong> the same hierarchical obliqueness relations required for the theory <strong>of</strong><br />

control.<br />

Accord<strong>in</strong>g to Pollard and Sag, passivisation can only apply to sentences <strong>with</strong> a direct object,<br />

which <strong>in</strong> their op<strong>in</strong>ion is a sign <strong>of</strong> the direct objects be<strong>in</strong>g more oblique than <strong>in</strong>direct objects.<br />

Their assumption is: the more oblique a verb argument, the more likely it is that its verbal<br />

governor can be passivised. It is worth po<strong>in</strong>t<strong>in</strong>g out that some <strong>German</strong> verbs do allow for<br />

dative (53) and even PO passivisation (54) but that the datives and POs are not realised as<br />

nom<strong>in</strong>atives (or subjects). It seems that it would be more accurate to limit Pollard and Sag's


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 93<br />

statement to passivisation <strong>in</strong>volv<strong>in</strong>g the realization <strong>of</strong> oblique case NPs or PPs as a<br />

nom<strong>in</strong>ative NP:<br />

53a Er hilft dem Mann.<br />

53b Dem Mann kann geholfen werden.<br />

54a Er spricht mit dem Direktor.<br />

54b Mit dem Direktor wird gerade gesprochen.<br />

And f<strong>in</strong>ally, Pollard and Sag ma<strong>in</strong>ta<strong>in</strong> that the least oblique verb argument, which typically<br />

is the subject, is the one which agrees <strong>with</strong> the verb.<br />

We have seen that both the names <strong>of</strong> the verb bond<strong>in</strong>g phenomenon and the criteria to<br />

describe its effect differ. What we understand by Verbnähe will become clearer <strong>in</strong> the next<br />

section, <strong>in</strong> which we describe which arguments and modifiers are more bound than others.<br />

3.4.1. WHICH ELEMENTS ARE SEMANTICALLY CLOSE TO THE VERB?<br />

In l<strong>in</strong>guistic literature, the verb object relationship has been discussed <strong>in</strong> much more detail<br />

than the verb modifier relationship. As we extend the term verb bond<strong>in</strong>g to modifiers, we<br />

shall discuss both, start<strong>in</strong>g <strong>with</strong> verb arguments, and then review Hoberg's statements on the<br />

verb bond<strong>in</strong>g <strong>of</strong> modifiers.<br />

3.4.1.1. ARGUMENTS<br />

Accord<strong>in</strong>g to Hoberg (1981: 63f), nouns <strong>of</strong> support verb constructions (SVC), which we<br />

shall from now on call SVC-nouns, are semantically very close to the verb.<br />

SVC are noun verb comb<strong>in</strong>ations such as etwas <strong>in</strong> Anwendung br<strong>in</strong>gen <strong>in</strong> which the noun<br />

carries the ma<strong>in</strong> <strong>in</strong>formation. Accord<strong>in</strong>g to Mesli (1991: 4ff), the nouns are abstract nouns<br />

express<strong>in</strong>g a process, a state, an event or a quality. They generally have an argument<br />

structure (55a, 55b), and they can represent a sentence (55c). One could say that the noun is<br />

"conjugated" (Mesli, 1991: 4) by the support verb, <strong>in</strong> that the verb carries the <strong>in</strong>formation on<br />

time, mode and number for the action expressed by the noun.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 94<br />

55a Die Anwendung des Gesetzes durch das Parlament.<br />

55b Anwendung (Parlament, Gesetz)<br />

55c Das Parlament brachte das Gesetz <strong>in</strong> Anwendung.<br />

Due to the special status <strong>of</strong> SVC-nouns, and their close relationship <strong>with</strong> the support verbs,<br />

they cannot be substituted by a pronoun (56b), and <strong>of</strong>ten one s<strong>in</strong>gle verb corresponds to the<br />

whole support verb construction (56c). And <strong>in</strong>deed these NPs cannot be separated from the<br />

verb <strong>in</strong> end position (56d, 56e):<br />

56a Stefan will die emanzipatorischen Gedanken endlich <strong>in</strong> Anwendung br<strong>in</strong>gen.<br />

56b * Stefan will die emanzipatorischen Gedanken endlich <strong>in</strong> sie br<strong>in</strong>gen.<br />

56c Stefan will die emanzipatorischen Gedanken endlich anwenden.<br />

56d * Stefan will die emanzipatorischen Gedanken <strong>in</strong> Anwendung endlich br<strong>in</strong>gen.<br />

56e * Stefan will <strong>in</strong> Anwendung die emanzipatorischen Gedanken endlich br<strong>in</strong>gen.<br />

Further elements which are semantically close to the verb are prepositional objects (PO).<br />

They are verb complements realised as a PP <strong>with</strong> a fixed preposition, such as glauben an and<br />

sich verlassen auf (Bußmann, 1983: 402). They differ from obligatory adverbials <strong>in</strong> that the<br />

adverbial PP can vary and <strong>in</strong> that its preposition is mean<strong>in</strong>gful (57), whereas the fixed PO<br />

prepositions are semantically empty (58):<br />

57a Man kann sich auf Archana verlassen.<br />

57b * Man kann sich <strong>in</strong> Archana verlassen.<br />

57c * Man kann sich vor Archana verlassen.<br />

57d * Man kann sich bei Archana verlassen.<br />

58a<br />

58b<br />

Archana wohnt <strong>in</strong> Wilmslow.<br />

Archana wohnt nahe Macclesfield.<br />

POs tend to follow case objects (accusative, dative, nom<strong>in</strong>ative or genitive NPs),<br />

<strong>in</strong>dependently from their realization as a full NP or as a pronoun (Heidolph/Flämig/Motsch,<br />

1981: 737).<br />

Lenerz (1977: 65ff) claims that accusative complements are more closely l<strong>in</strong>ked<br />

semantically to the verb than POs. He bases his view on the fact that <strong>in</strong> sentences like 3a the<br />

ellipsis e<strong>in</strong>en Brief schreiben is more <strong>of</strong> a unit ("E<strong>in</strong>heit") (Lenerz, 1977: 78) than an me<strong>in</strong>en


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 95<br />

Vater schreiben 59 . Lenerz agrees, however, that POs tend to follow case arguments. He<br />

expla<strong>in</strong>s this tendency towards the right by the <strong>German</strong> development from an SOV to an<br />

SVO language (cf. Vennemann's theory <strong>of</strong> Natural Serialisation and word order change,<br />

1974 and others). Accord<strong>in</strong>g to Lenerz, the order NP argument before PP argument is a sign<br />

for the new order SVO.<br />

59a Ich schreibe e<strong>in</strong>en Brief an me<strong>in</strong>en Vater.<br />

59b e<strong>in</strong>en Brief schreiben<br />

59c ? an me<strong>in</strong>en Vater schreiben<br />

There are several reasons why we do not agree <strong>with</strong> Lenerz' assumption: Firstly, <strong>in</strong> sentences<br />

like (60), the object is not more closely l<strong>in</strong>ked to the verb than the PO. Secondly, Lenerz<br />

does not account for the <strong>in</strong>fluence <strong>of</strong> the other factors <strong>in</strong> his examples, so that they have to<br />

be reevaluated. Thirdly, we shall see that, at sentence level, <strong>German</strong> is a very consistent<br />

SOV language, apart from the verb, which can occur <strong>in</strong> verb second position. And fourthly,<br />

the preposition <strong>of</strong> POs is semantically empty and subcategorised for by the verb, <strong>in</strong>stead <strong>of</strong><br />

match<strong>in</strong>g the semantics <strong>of</strong> the PP (see example 57, above):<br />

60a<br />

60b<br />

60c<br />

Er berichtet dem Freund über die Ause<strong>in</strong>andersetzung.<br />

dem Freund berichten<br />

über die Ause<strong>in</strong>andersetzung berichten<br />

Universally, objects are more closely l<strong>in</strong>ked to the verb than subjects (Toml<strong>in</strong>, 1986: 73).<br />

For this reason, generative grammar (e.g. Stechow/Sternefeld, 1988) assigns the subject a<br />

special status: it gets its case assignment from Infl (the head <strong>of</strong> the Inflectional Phrase)<br />

whereas objects get it from the verb. Inf<strong>in</strong>itive verb forms cannot be accompanied by a<br />

subject 60 :<br />

62a e<strong>in</strong>em Mann e<strong>in</strong> Buch geben.<br />

62b * er e<strong>in</strong>em Mann e<strong>in</strong> Buch geben.<br />

59 Lenerz' decision <strong>of</strong> what a unit is is more or less based on <strong>in</strong>tuition (ellipsis test). Constituent tests do not<br />

seem to prove the existence <strong>of</strong> these units clearly, as their outcome is contradictory (Lenerz, 1977: 75ff).<br />

For a discussion <strong>of</strong> the validity <strong>of</strong> Lenerz' data see Gadler (1982: 156ff).<br />

60 Note that <strong>in</strong> exceptional structures the subject can receive its (accusative) case form the matrix verb:<br />

61 Ich sah ihn e<strong>in</strong>em Mann e<strong>in</strong> Buch geben.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 96<br />

For most <strong>German</strong> speakers the accusative object <strong>in</strong>tuitively is more closely l<strong>in</strong>ked to the verb<br />

geben than the dative object. Therefore, they feel that e<strong>in</strong> Buch geben is more likely to be a<br />

unit than e<strong>in</strong>em Mann geben. However, a questionnaire on this subject described and<br />

analysed <strong>in</strong> Ste<strong>in</strong>berger (1990) shows that the bond<strong>in</strong>g <strong>of</strong> the object types is verb-specific<br />

(cf. 4.2).<br />

3.4.1.2. MODIFIERS<br />

We suggested <strong>in</strong> 2.3 that we should dist<strong>in</strong>guish three modifier types which refer to the<br />

utterance (existimatorial, pragmatic), to the sentence (situative) and to the verb (modal). It is<br />

obvious that the verb-modify<strong>in</strong>g modal modifiers are the ones which are closest to the verb,<br />

whereas existimatorial adverbials are least close (cf. Hoberg, 1981: 132ff). Accord<strong>in</strong>g to<br />

their l<strong>in</strong>ks <strong>with</strong> the verb the elements <strong>of</strong> the modifier classes can permute <strong>with</strong> different<br />

degrees <strong>of</strong> freedom:<br />

63a<br />

63b<br />

Peter hat Frauenbücher gern gelesen.<br />

Peter hat gern Frauenbücher gelesen.<br />

63c Peter hat leider Frauenbücher gelesen.<br />

63d * Peter hat Frauenbücher leider gelesen.<br />

The modal modifier gern tends to stand at the right (the end) <strong>of</strong> the sentence and can<br />

permute <strong>with</strong> the <strong>in</strong>def<strong>in</strong>ite plural noun Frauenbücher, whereas the evaluative adverb leider<br />

tends to the left. Its position beh<strong>in</strong>d the <strong>in</strong>def<strong>in</strong>ite accusative noun is not possible.<br />

However, the dist<strong>in</strong>ction <strong>of</strong> four ma<strong>in</strong> modifier classes is not enough to describe <strong>German</strong><br />

word order. In Regeln zur Wortstellung (1970), Engel dist<strong>in</strong>guishes 22 modifier classes.<br />

Both his classification, and the order <strong>in</strong> which he claims they follow, are based on his<br />

<strong>in</strong>tuition. Hoberg (1981: 98-149) verified the order <strong>of</strong> these classes empirically, us<strong>in</strong>g the<br />

Mannheimer Duden-Korpus 61 . She found that Engel's <strong>in</strong>tuition did not always correspond to<br />

61 The Mannheimer Duden-Korpus consists <strong>of</strong> a collection <strong>of</strong> written <strong>German</strong> contemporary texts (s<strong>in</strong>ce<br />

1945). It is composed <strong>of</strong> five text types: fiction and poetry, light fiction, popular scientific texts, memoirs<br />

and newspapers/magaz<strong>in</strong>es. Out <strong>of</strong> 20 texts, she chose blocks from the beg<strong>in</strong>n<strong>in</strong>g, the middle and the end,


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 97<br />

the real-life occurrence <strong>in</strong> the corpus. Furthermore, some <strong>of</strong> the co-occurr<strong>in</strong>g elements were<br />

<strong>of</strong> the same class, so that Engel's list did not make any prediction on their order. On the basis<br />

<strong>of</strong> her f<strong>in</strong>d<strong>in</strong>gs, Hoberg split Engel's modifier groups and reordered some <strong>of</strong> them. The result<br />

are 44 modifier position classes which are ordered accord<strong>in</strong>g to their <strong>in</strong>dexes from 1 to 44.<br />

They are described <strong>in</strong> detail <strong>in</strong> Hoberg (1981: 106-131). A shorter version is listed <strong>in</strong><br />

appendix 8.1.<br />

What we identified as the group <strong>of</strong> existimatorial modifiers corresponds to the <strong>in</strong>dexes a1 to<br />

a18 (a for Angaben). The situatives are represented by a19 to a40, and the modals by a42 to<br />

a44 (Hoberg, 1981: 132f). The class a41 denotes the negation. Her classification is not<br />

restricted to adverbs, but also <strong>in</strong>cludes adverbial phrases such as PPs and NPs.<br />

In the follow<strong>in</strong>g examples, the modifier sequence is accord<strong>in</strong>g to her classification. The<br />

<strong>in</strong>dexes <strong>in</strong>dicate to which group the modifiers belong (Examples from Hoberg, 1981: 131):<br />

64 Es hat beispielsweise 8 zweifellos 12 erst 39 e<strong>in</strong>er sehr langen Erfahrung bedurft.<br />

65 ... und wenn auch..., so ist der Sprung doch 4 immer wieder 37 deutlich 43 zu spüren.<br />

66 Wie solche von den Nationalsozialisten verstanden wurde, hatten wir ja 1 <strong>in</strong>zwischen 26<br />

e<strong>in</strong> paarmal 37 erfahren können.<br />

The <strong>in</strong>dexes roughly reflect the semantic closeness to the verb. The higher the <strong>in</strong>dex <strong>of</strong> a<br />

modifier, the closer it is to the verb. Hoberg (1981: 134) admits, however, that it is not<br />

possible to see the difference <strong>in</strong> verb bond<strong>in</strong>g from one class to the next. It would be difficult<br />

to argue, for <strong>in</strong>stance, that adverbials express<strong>in</strong>g condition are less bound than those<br />

express<strong>in</strong>g a concession, etc (see Hoberg's classes, 1981: 106-132):<br />

... condition 19 < concession 20 < cause 22 < consecutive 23 < f<strong>in</strong>al 24 < medial 25 <<br />

time_period 26 < local 27 < accompany<strong>in</strong>g_circumstances 28 ...<br />

In l<strong>in</strong>guistic literature, a small number <strong>of</strong> modifier classes have been discussed more<br />

frequently, <strong>with</strong> respect to their serialisation. These groups always tend to be the same,<br />

<strong>of</strong> approximately 1,400 words each. Hoberg's test corpus thus comprises about 85,000 words<br />

correspond<strong>in</strong>g to 11,000 simple sentences (sentences <strong>with</strong> one ma<strong>in</strong> verb each) (Hoberg, 1981: 27).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 98<br />

namely time, place, manner, and sometimes cause modifiers (cf. 1.2). All the others are<br />

generally neglected.<br />

Lenerz (1977: 78ff), for <strong>in</strong>stance, discusses the order <strong>of</strong> local (LOC) and temporal (TEMP) 62<br />

adverbials, us<strong>in</strong>g Vennemann's universal pr<strong>in</strong>ciple <strong>of</strong> natural serialisation (cf. 1.1). He states<br />

that TEMP, which is more abstract, is less closely related to the verb than the more concrete<br />

LOC and that the SOV language <strong>German</strong> therefore prefers the order TEMP < LOC. English,<br />

which is known as an SVO language, prefers LOC < TEMP. However, Lenerz restricts his<br />

claim by say<strong>in</strong>g that the modifiers' degree <strong>of</strong> verb bond<strong>in</strong>g differs depend<strong>in</strong>g on the verb<br />

concerned. Although it should be the more closely l<strong>in</strong>ked element which is able to be<br />

topicalised <strong>with</strong> the verb, <strong>in</strong> 67 only the temporal modifier can move to the Vorfeld <strong>with</strong> the<br />

verb. Lenerz encounters even more contradictions when compar<strong>in</strong>g the relative verb bond<strong>in</strong>g<br />

<strong>of</strong> LOC and case arguments (1977: 89ff):<br />

67 Vor Mitternacht e<strong>in</strong>schlafen kann man <strong>in</strong> diesem Hotel nie.<br />

* In diesem Hotel e<strong>in</strong>schlafen kann man vor Mitternacht nie.<br />

It seems that verb bond<strong>in</strong>g can roughly expla<strong>in</strong> the order <strong>of</strong> modifier classes, namely for the<br />

superclasses existimatorial, situative and modal modifiers. However, it does not provide<br />

enough motivation for the more f<strong>in</strong>e-gra<strong>in</strong>ed dist<strong>in</strong>ction <strong>of</strong> position classes as <strong>of</strong>fered by<br />

Hoberg.<br />

3.4.2. LIMITS OF THE VERBNÄHE PRINCIPLE<br />

A sequence which cannot be expla<strong>in</strong>ed by the Verbnähe pr<strong>in</strong>ciple is the relative order <strong>of</strong><br />

toners (examples from Engel, 1988: 327):<br />

68 Er ist aber 3 eben 8 viel zu hartnäckig.<br />

69 Wie hieß er denn 1 noch 34 ?<br />

62 Local modifiers correspond to Hoberg's class a 27 . Temporal modifiers are spread over the classes a 26 ,<br />

a 33 , a 36 , a 37 , a 39 and a 40 (cf. 6.7.1).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 99<br />

Verb bond<strong>in</strong>g is a very strong pr<strong>in</strong>ciple but, as far as we are aware, it has limited explanatory<br />

power for the more f<strong>in</strong>e-gra<strong>in</strong>ed order <strong>of</strong> modifiers. Another problem l<strong>in</strong>ked to it is that it is<br />

difficult to grasp, other than by <strong>in</strong>tuition. The arguments given by Pollard and Sag are <strong>of</strong> no<br />

great assistance: The control criterion only applies for <strong>in</strong>f<strong>in</strong>itival clauses, and the fact that<br />

subjects agree <strong>with</strong> the verb only identifies subjects as be<strong>in</strong>g least oblique. The latter<br />

criterion even causes a contradiction, as the most oblique verb arguments are direct objects<br />

(passivisation), and direct objects are the ones which become subjects <strong>in</strong> passivised<br />

sentences. None <strong>of</strong> Pollard and Sag's criteria is applicable to modifiers.<br />

It is <strong>in</strong>trigu<strong>in</strong>g to see that a pr<strong>in</strong>ciple <strong>with</strong> the strength <strong>of</strong> verb bond<strong>in</strong>g is so much based on<br />

<strong>in</strong>tuition. One wonders why the verb geben immediately calls for the question "Was?", and<br />

only then "Wem?"? Why does the statement: "Er starb" call for the questions: "Woran?" and<br />

"Wann?" and only much later: "Wo?" whereas "Er wohnt" needs the answer to the question<br />

"Wo?". It looks as if these questions cannot be answered <strong>with</strong>out revert<strong>in</strong>g to pragmatics.<br />

Research on valency could also probably contribute to f<strong>in</strong>d<strong>in</strong>g an appropriate answer (cf.<br />

also section 3.6 on semantic roles).<br />

3.5. THE ANIMACY-FIRST PRINCIPLE<br />

Another pr<strong>in</strong>ciple is the tendency <strong>of</strong> elements <strong>with</strong> animate reference to precede other<br />

elements which do not (Eisenberg, 1988: 424f). This factor applies only to nouns and is<br />

furthermore restricted to subdoma<strong>in</strong>s: Pronom<strong>in</strong>alised NPs, animate or not, have a strong<br />

tendency to precede full NPs, and def<strong>in</strong>ite NPs precede <strong>in</strong>def<strong>in</strong>ite ones. To name the<br />

features, we shall use the follow<strong>in</strong>g convention:<br />

N, A, D, G for the cases nom<strong>in</strong>ative, accusative, dative and genitive<br />

[+/-a] for +/- animate NPs<br />

[pron] for pronouns<br />

FN for nouns (Funktionsnomen) <strong>in</strong> support verb constructions (SVC)<br />

As pronom<strong>in</strong>ality and animacy have a strong <strong>in</strong>fluence on <strong>German</strong> word order, Hoberg<br />

(1981: 79) chose these features to describe the order <strong>of</strong> verb arguments <strong>in</strong> her canonical


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 100<br />

form. Accord<strong>in</strong>g to her, the follow<strong>in</strong>g is the unmarked sequence <strong>in</strong> the <strong>German</strong> Mittelfeld: 1)<br />

pronouns [pron], 2) animate nouns [+a], 3) <strong>in</strong>animate nouns [-a] and 4) nouns <strong>of</strong> support<br />

verb constructions [FN]. Hoberg claims that the sequence <strong>of</strong> the cases <strong>with</strong><strong>in</strong> these four<br />

groups is N, A, D and G (the slash "/" between two or more groups means that elements <strong>of</strong><br />

these groups do not occur together):<br />

(N - A - D/G) pron - (N - A - D/G) +a - (N - A - D/G) -a - (N/A/D/G) FN<br />

The animacy-first pr<strong>in</strong>ciple expla<strong>in</strong>s that datives normally precede accusatives when the<br />

verb is a verb <strong>of</strong> giv<strong>in</strong>g, tak<strong>in</strong>g, communication or concealment. It is worth po<strong>in</strong>t<strong>in</strong>g out that<br />

Hoberg does not take <strong>in</strong>to consideration the def<strong>in</strong>iteness feature, but only refers to<br />

pronom<strong>in</strong>ality and animacy. We shall discuss this <strong>in</strong> section 4.2.<br />

Eisenberg (1989: 424f) calls attention to the <strong>in</strong>terest<strong>in</strong>g fact that non-prototypical verbs like<br />

the psychological verbs (ärgern, beruhigen, begeistern, <strong>in</strong>teressieren, ...) <strong>of</strong>ten differ from the<br />

order N


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 101<br />

It may be assumed, and it has <strong>of</strong>ten been asserted, that among the <strong>in</strong>f<strong>in</strong>ity <strong>of</strong> potential referents that may<br />

engage our attention some are <strong>in</strong>tr<strong>in</strong>sically more salient than others, just as certa<strong>in</strong> potential dist<strong>in</strong>ctions<br />

upon which the classification <strong>of</strong> phenomena might be based are, by virtue <strong>of</strong> our biological endowment,<br />

<strong>in</strong>tr<strong>in</strong>sically more salient than others are [...]. What is known <strong>of</strong> course, almost by def<strong>in</strong>ition, more<br />

salient than what is unknown; and, other th<strong>in</strong>gs be<strong>in</strong>g equal, the more recently that someth<strong>in</strong>g has been<br />

mentioned and put <strong>in</strong>to the universe-<strong>of</strong>-discourse, or the more familiar that someth<strong>in</strong>g is to the<br />

participants <strong>in</strong> a conversation, the greater will be its psychological salience.<br />

Lyons assumes that man is more <strong>in</strong>terested <strong>in</strong> humans than <strong>in</strong> animals and more <strong>in</strong>terested <strong>in</strong><br />

the latter than <strong>in</strong> <strong>in</strong>animate entities (similar: Toml<strong>in</strong>, 1986: 102ff). This is the reason why the<br />

follow<strong>in</strong>g passive clause seems more normal than the active counterpart (example from<br />

Lyons, 1977: 511):<br />

73a A man was stung by a bee <strong>in</strong> the High Street to-day.<br />

73b ? A bee stung a man <strong>in</strong> the High Street to-day.<br />

Foley and Van Val<strong>in</strong> (1985: 287ff) claim that the universal animacy hierarchy is even more<br />

f<strong>in</strong>e-tuned:<br />

speaker < addressee < human proper < human common < other animate < <strong>in</strong>animate<br />

To underp<strong>in</strong> the relevance <strong>of</strong> this hierarchy, they mention the Mexican language Mixe, <strong>in</strong><br />

which it determ<strong>in</strong>es the order <strong>of</strong> phrases <strong>in</strong> the sentence. It seems, though, that Foley and<br />

Van Val<strong>in</strong> confuse animacy <strong>with</strong> a more general hierarchy <strong>of</strong> mental presence or <strong>of</strong> general<br />

<strong>in</strong>terest. Speaker, addressee and other human be<strong>in</strong>gs certa<strong>in</strong>ly do not differ <strong>in</strong> how animate<br />

they are, but rather <strong>in</strong> how <strong>in</strong>terested we are <strong>in</strong> the different persons and objects. We, the<br />

speakers, are always aware <strong>of</strong> what we feel and <strong>of</strong> what we want. The addressee also<br />

deserves a considerable part <strong>of</strong> our momentary <strong>in</strong>terest. Other people, as well as other th<strong>in</strong>gs,<br />

are probably <strong>of</strong> least <strong>in</strong>terest to us, when we are speak<strong>in</strong>g to someone else. Foley and Van<br />

Val<strong>in</strong>'s hierarchy is thus more likely to represent a hierarchy mix<strong>in</strong>g animacy, semantic roles<br />

(3.6) and probably someth<strong>in</strong>g else.<br />

Animacy is not only relevant for <strong>German</strong>, where it def<strong>in</strong>itely plays an important role for<br />

word order (cf. our weigh<strong>in</strong>g up aga<strong>in</strong>st def<strong>in</strong>iteness <strong>in</strong> 4.2). Toml<strong>in</strong> (1986) and Foley/Van<br />

Val<strong>in</strong>, (1985) show that the effect <strong>of</strong> this pr<strong>in</strong>ciple is relevant universally, or at least for a<br />

considerable part <strong>of</strong> the other languages <strong>of</strong> the world.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 102<br />

3.6. SEMANTIC ROLES<br />

Hoberg (1981: 58) <strong>in</strong>dicates that there could be a relationship between the semantic roles <strong>of</strong><br />

NPs <strong>in</strong> a sentence and their animacy. The roles AGENT and EXPERIENCER, for <strong>in</strong>stance,<br />

are necessarily animate and the role OBJECT rarely is. It seems reasonable to assume that<br />

the animacy-first pr<strong>in</strong>ciple could be replaced by a precedence pr<strong>in</strong>ciple, based on semantic<br />

roles. In his book on universal word order<strong>in</strong>g pr<strong>in</strong>ciples, Toml<strong>in</strong> (1986: 104) l<strong>in</strong>ks animacy<br />

directly to Fillmore's hierarchy <strong>of</strong> case roles:<br />

One NP will be more animated than another if it is higher on a hierarchy <strong>of</strong> semantic roles derived from<br />

Fillmore (1968: 24-25, 33): [...]<br />

Agent > Instrumental > Benefactive/Dative > Patient<br />

He furthermore claims that semantic roles "take precedence over animacy" (1986: 106).<br />

Siewierska (1988: 56ff) gives an example, <strong>in</strong> which the exact opposite holds, namely the<br />

language Sesotho, <strong>in</strong> which animacy takes precedence over semantic roles. She claims<br />

generally, that there "are relatively few <strong>in</strong>stances <strong>of</strong> order<strong>in</strong>g phenomena directly<br />

attributable to the semantic role hierarchy" (Siewierska, 1988: 51).<br />

For <strong>German</strong>, Jacobs (1988) seems to share Toml<strong>in</strong>'s assumption as he does not even mention<br />

animacy <strong>in</strong> Probleme der freien Wortstellung, where he expla<strong>in</strong>s <strong>German</strong> word order<br />

variation by eight position<strong>in</strong>g pr<strong>in</strong>ciples, four <strong>of</strong> which are based on semantic roles (1988:<br />

19ff):<br />

P1: AGENT < X<br />

P2: DATIVE < PATIENT<br />

P7: GOAL < THEME<br />

P8: OBJECT < DIRECTIONAL<br />

X refers to any semantic role other than AGENT, and DATIVE subsumes the semantic roles<br />

which are <strong>of</strong>ten realised by dative NPs, namely RECIPIENT, ANIMATE GOAL,<br />

BENEFACTIVE etc (1988: 19). Jacobs understands by THEME the "moved or localised<br />

entity" ("die bewegte oder lokalisierte Entität", 1988: 23).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 103<br />

Pr<strong>in</strong>ciples based on semantic roles can be very strong. Some verb groups differ <strong>in</strong> their<br />

syntactical behaviour from the prototypical verbs (Eisenberg, 1989: 424f). The psychic verbs<br />

call<strong>in</strong>g for an accusative, for example (ärgern, beruhigen, begeistern, erfreuen, entsetzen,<br />

<strong>in</strong>teressieren), have A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 104<br />

3.7. SCOPE<br />

Another factor which can determ<strong>in</strong>e the order <strong>of</strong> elements is what Engel (1970: 14f) calls<br />

"Zemb's thesis" mak<strong>in</strong>g reference to Les structures logiques de la phrase allemande (Zemb,<br />

1968). Simplified, this thesis states that elements at the left determ<strong>in</strong>e elements at the right.<br />

Behaghel (1932: 5) refers to the same regularity when he says "daß das unterscheidende<br />

Glied dem unterschiedenen vorausgeht". At sentence level, this pr<strong>in</strong>ciple is confirmed by<br />

negational elements and degree modifiers. In 76, they refer to the NPs (76a-76c) or the verb<br />

(76d) which immediately follow them:<br />

76a {Nicht/Nur} ICH habe den Leuten die Bücher geliehen.<br />

76b Ich habe {nicht/nur} den LEUten die Bücher geliehen.<br />

76c Ich habe den Leuten {nicht/nur} die BÜcher geliehen.<br />

76d Ich habe den Leuten die Bücher {nicht/nur} geLIEhen.<br />

77 and 78 show that there are exceptions to this rule, which claims that elements <strong>with</strong> scope<br />

precede those they refer to:<br />

77 Peter gerade sollte aber ke<strong>in</strong> Glück haben.<br />

78 Männer zum<strong>in</strong>dest neigen dazu, <strong>in</strong> ihrem Verhalten unpolitisch zu se<strong>in</strong>.<br />

If we assume that determ<strong>in</strong>ation means specification or modification, this pr<strong>in</strong>ciple does not<br />

apply to the order <strong>of</strong> verb arguments (79), temporal and local adverbials (80) and toners<br />

(81). In 79 to 81, this pr<strong>in</strong>ciple does not apply, as there is no scope:<br />

79a<br />

79b<br />

80a<br />

80b<br />

Ich habe der Frau e<strong>in</strong> Buch geschenkt.<br />

Ich habe das Buch e<strong>in</strong>er Frau geschenkt.<br />

Ich habe den Mann hier letzte Woche gesehen.<br />

Ich habe den Mann letzte Woche hier gesehen.<br />

81a Männer s<strong>in</strong>d eben durchaus noch nicht emanzipiert.<br />

81b * Männer s<strong>in</strong>d durchaus eben noch nicht emanzipiert.<br />

The formulation left determ<strong>in</strong>es right seems to be too general, as only some elements have<br />

the capacity <strong>of</strong> scope <strong>in</strong>clusion. Jacobs (1988: 20) therefore limits the pr<strong>in</strong>ciple to scope<strong>in</strong>clud<strong>in</strong>g<br />

elements and thus to the relevant part <strong>of</strong> the lexicon. NPs can have a scope when<br />

they are accompanied by a quantifier. Evidence <strong>of</strong> this is the follow<strong>in</strong>g example, <strong>in</strong> which<br />

the active and the passive sentences have different mean<strong>in</strong>gs (Primus, 1987: 61). In 82,


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 105<br />

every man kisses one woman, and this woman can differ from one man to the other. In 83,<br />

however, it is the same woman who is kissed by the men:<br />

82 Jeder Mann küsst e<strong>in</strong>e Frau.<br />

83 E<strong>in</strong>e Frau wird von jedem Mann geküßt.<br />

Concern<strong>in</strong>g modifiers, we shall have to specify which ones have scope. The scope <strong>of</strong><br />

operators <strong>in</strong> natural language is a complex area and therefore cannot be discussed<br />

exhaustively here. In the follow<strong>in</strong>g paragraphs, we shall sketch some problems l<strong>in</strong>ked to this<br />

phenomenon. For a more <strong>in</strong>-depth description <strong>of</strong> degree modifiers and their treatment <strong>in</strong><br />

Mach<strong>in</strong>e Translation, see Ste<strong>in</strong>berger (1992b) and the literature quoted there.<br />

3.7.1. DEFINITIONS OF SCOPE<br />

In Lexikon der Sprachwissenschaft (1983: 465), Bußmann def<strong>in</strong>es scope as follows:<br />

Skopus [griech. scopos >Zielpunkt Logische Partikeln und Quantoren) bezeichnet, wird<br />

Skopus <strong>in</strong> der Sprachwissenschaft für den semantischen Bezugsbereich von Negation, sprachlichen<br />

Quantoren und Partikeln verwendet. Dem Skopus e<strong>in</strong>es Operators <strong>in</strong> der Logik entspricht <strong>in</strong> der<br />

Sprachwissenschaft die Konstituente, die durch Quantoren oder Partikeln modifiziert wird; vgl. auch <strong>in</strong><br />

Philip hatte auch Hunger (nicht nur Durst) vs. Philip hatte auch Hunger (nicht nur die anderen).<br />

Fixierung und Interpretation des Skopus hängt häufig von der Akzentsetzung ab, vgl. Intonation.<br />

Scope-<strong>in</strong>clud<strong>in</strong>g elements are thus the negation (84), quantifiers (85) (jeder, alle, ...) and<br />

degree modifiers (86) (sogar, nur, ...):<br />

84 Sie gibt nicht den Männern die Schuld.<br />

85 Jede Frau sollte drei Männer haben.<br />

86 Sie mag nur Männer.<br />

Engel (1988: 764) specifies the effect <strong>of</strong> degree modifiers:<br />

[Gradpartikeln] präzisieren [...] unter Umständen den Grad e<strong>in</strong>er Eigenschaft des folgenden Elements<br />

(z.B. nahezu), zugleich und <strong>of</strong>t ausschließlich spezifizieren sie aber die Erwartbarkeit (sogar), oder sie<br />

setzen das Folgeelement <strong>in</strong> Beziehung zu gleichartigen Elementen (besonders). So kann der Sprecher<br />

auch mit ihrer Hilfe vor allem se<strong>in</strong>e E<strong>in</strong>stellung zu e<strong>in</strong>em Sachverhalt ausdrücken.<br />

Thus, the functions <strong>of</strong> scope-<strong>in</strong>clud<strong>in</strong>g elements are:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 106<br />

A) modification <strong>of</strong> a constituent<br />

B) graduation <strong>of</strong> a quality<br />

C) specification <strong>of</strong> how expected an event is<br />

D) to make a relation between similar elements<br />

E) expression <strong>of</strong> the speaker's attitude<br />

Due to the close relationship between the scope-<strong>in</strong>clud<strong>in</strong>g element and the modified lexemes<br />

and phrases, Engel (1988) and Hoberg (1981) call this relationship adjunction.<br />

3.7.2. PROBLEMS WITH THE TERM SCOPE<br />

Bußmann's and Engel's def<strong>in</strong>itions are not as clear as they seem at a first glance. Accord<strong>in</strong>g<br />

to Hoberg, particles such as aber, nämlich and allerd<strong>in</strong>gs, <strong>in</strong> (87), (88) and (89) are adjuncts<br />

to the other elements <strong>in</strong> the Vorfeld. However, none <strong>of</strong> Engel's (1988: 764) descriptions is<br />

appropriate for their relation (examples 87, 88 and 89 accord<strong>in</strong>g to Hoberg, 1981: 178):<br />

87 Dann aber bekam sie etwas Unerwartetes zu hören.<br />

88 In der Antike nämlich war jede poetische Gattung erst <strong>in</strong> e<strong>in</strong>er beschränkten Zahl von<br />

Mustern vertreten.<br />

89 E<strong>in</strong>es allerd<strong>in</strong>gs steht fest: ... (wobei e<strong>in</strong>es dafür steht, daß "se<strong>in</strong> Ansehen <strong>in</strong> der<br />

arabischen Welt gelitten hat")<br />

Waltz<strong>in</strong>g (1986: 141f) also has doubts concern<strong>in</strong>g the adjunction relation <strong>of</strong> these particles<br />

and the Vorfeld elements:<br />

Es verhält sich doch wohl eher so, daß das durch allerd<strong>in</strong>gs zum Ausdruck gebrachte e<strong>in</strong>schränkende<br />

Zugeständnis die g a n z e Proposition (daß etwas feststeht) betrifft und daß diese Äußerung - bis auf<br />

noch zu klärende Nuancen - gar nichts anderes besagt als:<br />

90a Es steht allerd<strong>in</strong>gs e<strong>in</strong>es fest: ...<br />

90b Allerd<strong>in</strong>gs steht e<strong>in</strong>es fest: ...<br />

90c Feststehen tut allerd<strong>in</strong>gs e<strong>in</strong>es: ...<br />

90d E<strong>in</strong>es steht allerd<strong>in</strong>gs fest: ...<br />

[...] Alles deutet darauf h<strong>in</strong>, daß die I ex [existimatorischen Angaben, RS], die ja im Mittelfeld die<br />

Grenze zu den (rechts davon stehenden) rhematischen Mittelfeldelementen markieren, auch im Vorfeld<br />

e<strong>in</strong>e ähnliche Funktion erfüllen, nämlich die, das l i n k s neben ihnen stehende Element als rhematisch<br />

- und zwar mit hohem Fokussierungsgrad/hohem Mitteilungswert - zu kennzeichnen. (Waltz<strong>in</strong>g, 1986:<br />

142; emphasis <strong>in</strong> the orig<strong>in</strong>al; number<strong>in</strong>g changed)<br />

In all four variations, e<strong>in</strong>es is rhematic, but the degree <strong>of</strong> its focalization varies from one<br />

sentence to the other. 90a is the most neutral (i.e. ambiguous) <strong>with</strong> respect to the<br />

specification <strong>of</strong> the relative importance <strong>of</strong> the elements, but <strong>in</strong> all four variations e<strong>in</strong>es is


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 107<br />

focused. Further evaluative modifiers which can be used to stress the Vorfeld element are<br />

freilich, immerh<strong>in</strong>, jedenfalls, schließlich, übrigens, zum Beispiel and beispielsweise<br />

(Waltz<strong>in</strong>g, 1986: 143).<br />

A further problem is the fact that sometimes degree modifiers can also refer to the sentence<br />

as a whole (Altmann, 1976: 248ff). Therefore the borderl<strong>in</strong>e between degree modifiers, on<br />

one side, and existimatorial and situative modifiers, on the other, is <strong>of</strong>ten unclear. In<br />

particular the dist<strong>in</strong>ction between degree modifiers and toners is gradual. There are <strong>in</strong>deed<br />

no elements <strong>in</strong> the group <strong>of</strong> toners which exclusively belong to this group (Eisenberg, 1989:<br />

208).<br />

91 Geh Du mir nur noch e<strong>in</strong>mal alle<strong>in</strong>e <strong>in</strong> die Diskothek! (toner)<br />

92 Er wiegt nur noch vierzig Kilogramm. (degree modification)<br />

As further particles, such as be<strong>in</strong>ahe, can refer to a phrase as well as to a sentence, Hoberg<br />

(1981: 145) formulates the behaviour <strong>of</strong> these elements as hav<strong>in</strong>g a tendency to be adjuncts.<br />

In 93, be<strong>in</strong>ahe refers to the NP, <strong>in</strong> 94 it refers to the whole sentence:<br />

93 Er hat be<strong>in</strong>ahe jeden Tag mit ihr telefoniert.<br />

94 Er hätte be<strong>in</strong>ahe mit ihr telefoniert.<br />

The difference between 93 and 94 seems to depend on the semantics <strong>of</strong> the elements which<br />

follow be<strong>in</strong>ahe: When be<strong>in</strong>ahe is a degree modifier, it needs someth<strong>in</strong>g which can be<br />

considered <strong>in</strong> terms <strong>of</strong> degree, such as frequency adverbials (jeden Tag), measurements (10<br />

meters), an evaluation (häßlich, gut) etc.<br />

3.7.3. THE DIFFERENCE BETWEEN SCOPE AND FOCALISATION<br />

Accord<strong>in</strong>g to Engel (1988: 337), all modifiers have a scope. In the sentence<br />

Solange haben sie im Vorraum unruhig auf die Polizei gewartet.<br />

[ist] die modifikative Angabe unruhig auf das Gefüge auf die Polizei warten zu beziehen, die<br />

Lokalangabe im Vorraum aber auf das größere Gefüge unruhig auf die Polizei warten. Das der Polizei<br />

geltende Warten wird also als unruhig charakterisiert und dieses unruhige Warten wird im Vorraum<br />

lokalisiert. Vertauscht man die beiden Angaben, so ändern sich die Bezugsbereiche und damit die<br />

Bedeutung des Satzes entsprechend. (italics <strong>in</strong> the orig<strong>in</strong>al)


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 108<br />

We do not agree <strong>with</strong> Engel's <strong>in</strong>terpretation <strong>of</strong> Bezug (scope): Although it is true that the<br />

restless wait<strong>in</strong>g takes place <strong>in</strong> the hall, the persons referred to as sie are also <strong>in</strong> the hall while<br />

wait<strong>in</strong>g. Furthermore, it is very difficult to imag<strong>in</strong>e what is meant when im Vorraum and<br />

unruhig are permuted relative to each other, as <strong>in</strong> that case unruhig should refer to Vorraum.<br />

Contrary to his <strong>in</strong>tention, Engel's example makes it clear that one has to dist<strong>in</strong>guish between<br />

scope <strong>in</strong>clusion and focalization. When elements <strong>with</strong> scope move, their scope, and thus the<br />

mean<strong>in</strong>g <strong>of</strong> the sentence, change. <strong>Word</strong> order variation <strong>in</strong>volv<strong>in</strong>g all other elements changes<br />

the theme-rheme structure, but no change <strong>of</strong> scope takes place.<br />

Although the scope <strong>of</strong> elements has an impact on the order <strong>of</strong> words, it is an impact which<br />

also obstructs word order variation. When we want to express that the scope-<strong>in</strong>clud<strong>in</strong>g<br />

element nur refers to the subject (95a), rather than to the <strong>in</strong>direct object (95b), the positions<br />

<strong>of</strong> nur are very limited. For this reason, we shall list scope <strong>in</strong> section 4.4.3 aga<strong>in</strong>, when<br />

discuss<strong>in</strong>g restrictions to the <strong>in</strong>teraction <strong>of</strong> the word order pr<strong>in</strong>ciples.<br />

95a {Nicht/Nur} ICH habe den Leuten die Bücher geliehen.<br />

95b Ich habe {nicht/nur} den LEUten die Bücher geliehen.<br />

When we refer to elements such as nur from now on, we refer to their sentence-modify<strong>in</strong>g<br />

read<strong>in</strong>g. In the analysis suggested <strong>in</strong> Ste<strong>in</strong>berger (1992b), degree modifiers are part <strong>of</strong> the<br />

phrase they modify, and move <strong>with</strong> it <strong>in</strong> the sentence. Nur ich <strong>in</strong> 95a would thus be analysed<br />

as one NP.<br />

With this short discussion, we <strong>in</strong>tended to show what is generally referred to by the term<br />

scope. Furthermore, we mentioned some problems which are l<strong>in</strong>ked to it, namely the<br />

different use <strong>of</strong> the term by different authors, as well as the difficult borderl<strong>in</strong>e between<br />

scope-<strong>in</strong>clud<strong>in</strong>g elements and others.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 109<br />

3.8. RHYTHM<br />

Eroms (1986: 50) and Lötscher (1983: 185) suggest that rhythm plays a role when we decide<br />

on the order <strong>of</strong> elements <strong>in</strong> the <strong>German</strong> sentence. However, they do not specify this<br />

pr<strong>in</strong>ciple. Behaghel (1932: 6f) states more precisely:<br />

Das Deutsche hat das Bestreben, stärker und schwächer betonte Glieder abwechseln zu lassen.<br />

A similar formulation can be found <strong>in</strong> Sommerfeldt/Starke (1988: 287f) who also see a<br />

relationship between rhythm and Behaghel's Gesetz der wachsenden Glieder, which they<br />

comprehend as a variation on the factor rhythm. Furthermore, some <strong>of</strong> the people who<br />

answered the questionnaire discussed <strong>in</strong> Ste<strong>in</strong>berger (1990) mentioned that rhythm has<br />

<strong>in</strong>fluenced their acceptability decision. We do not have more <strong>in</strong>formation or <strong>in</strong>tuition on this<br />

subject, and therefore we cannot discuss it any further.<br />

In French there seems to be no doubt that rhythm has an <strong>in</strong>fluence on the position <strong>of</strong><br />

adverbials. This factor is mentioned <strong>in</strong> several grammars (see e.g. Grevisse, 1986: 1421).<br />

3.9. NATURAL GENDER<br />

We have not come across literature mention<strong>in</strong>g natural gender as a factor determ<strong>in</strong><strong>in</strong>g word<br />

order, but the questionnaire discussed <strong>in</strong> Ste<strong>in</strong>berger (1990) gave some evidence that the<br />

natural gender <strong>of</strong> the referents <strong>of</strong> nouns can <strong>in</strong>fluence people's decisions on the order <strong>of</strong> NPs.<br />

38 <strong>German</strong> native speakers were asked to decide on which sentence, out <strong>of</strong> a sentence pair,<br />

they thought was more natural. All sentences were given <strong>with</strong>out context, and the answer<br />

that both sentences are equally natural was allowed. The numbers show how many<br />

<strong>in</strong>formants chose the respective sentences to be the more natural ones (test 2, page III:<br />

96a Sie hatten dem Mann die Frau gleichgestellt. 6<br />

96b Sie hatten die Frau dem Mann gleichgestellt. 32<br />

97a Sie hatten der Frau den Mann gleichgestellt. 13<br />

97b Sie hatten den Mann der Frau gleichgestellt. 25


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 110<br />

When decid<strong>in</strong>g whether 96a or 96b was more natural, 32 out <strong>of</strong> the 38 tested persons<br />

preferred the order accusative before dative (96b). In the sentence pair 97, the same order<br />

was preferred but the amount <strong>of</strong> votes for the dative-accusative order was larger (13<br />

compared to 6 answers). As the factors animacy, def<strong>in</strong>iteness and semantic roles are<br />

identical <strong>in</strong> both sentence pairs, the answers show that another pr<strong>in</strong>ciple must have<br />

<strong>in</strong>fluenced the tested persons. Indeed several <strong>of</strong> the <strong>in</strong>formants said <strong>in</strong>dependently that they<br />

wanted to give precedence to the woman ("ich wollte der Frau den Vortritt lassen"), or that<br />

they were used to lett<strong>in</strong>g the woman go first.<br />

In another test (test 5, page VI) we wanted to f<strong>in</strong>d out how the tested persons <strong>in</strong>terpreted the<br />

semantic roles <strong>of</strong> NPs if they are not expressed morphologically. The test is based on Höhle's<br />

(1982: 130) claim that hearers prefer unmarked order for contextless sentences, as this<br />

allows them most liberal <strong>in</strong>terpretation:<br />

Außerhalb von disambiguierenden Kontexten bevorzugt der Hörer e<strong>in</strong>e Interpretation, die der normalen<br />

Wortstellung entspricht [..., denn ...] unter normaler Wortstellung [ist] die Anzahl der pr<strong>in</strong>zipiell<br />

möglichen Kontexttypen am größten; diese Interpretation ist daher e<strong>in</strong>e relativ sichere Interpretation.<br />

We asked our <strong>in</strong>formants to put articles <strong>in</strong> front <strong>of</strong> the proper names <strong>in</strong> sentences, such as 3,<br />

<strong>in</strong> order to identify their syntactic and semantic roles. From this procedure, we hoped to<br />

discover which word order they thought was more natural, or unmarked.<br />

In five test sentence pairs we used male names. The result was that <strong>in</strong> 87% <strong>of</strong> the cases<br />

people preferred the order dative before accusative. In only one sentence was a female name<br />

<strong>in</strong>volved (Helga <strong>in</strong> 98), which changed the result drastically: Only 45% gave the answer<br />

98b, which shows that they preferred the order dative before accusative. 50% <strong>of</strong> the<br />

<strong>in</strong>formants preferred the opposite order (98b). A possible explanation for these diverg<strong>in</strong>g<br />

results is that the tested persons preferred the idea that a man (Fritz) <strong>in</strong>troduces a man<br />

(Wolfgang) to a woman (Helga) to the one that a man <strong>in</strong>troduces a woman to a man:<br />

98 Dann stellte Fritz Wolfgang Helga vor. (morphologically and syntactically ambiguous)<br />

98a Der Fritz dem Wolfgang die Helga D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 111<br />

We want to stress that it is not the 5% difference between the D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 112<br />

pr<strong>in</strong>ciple could be formulated: Dative verb complements tend to precede accusative<br />

complements.<br />

3.11. LENERZ' "SATZKLAMMERBEDINGUNG"<br />

Lenerz (1977: 63) formulates a further rule:<br />

Es besteht die stilistische Tendenz, Sätze ohne h<strong>in</strong>ten geschlossene Satzklammer möglichst nicht auf e<strong>in</strong><br />

gewichtsloses Satzglied enden zu lassen.<br />

This tendency expla<strong>in</strong>s the perception-related difference <strong>in</strong> acceptability between 102a and<br />

102b (example, stress mark<strong>in</strong>g and acceptability judgement accord<strong>in</strong>g to Lenerz, 1977: 61):<br />

102a ? Ich widme den vielen überaus hilfreichen KolLEgen, die durch ihre Kritik erst se<strong>in</strong>e<br />

Entstehung ermöglicht haben, dieses Buch.<br />

102b<br />

Ich habe den vielen überaus hilfreichen KolLEgen, die durch ihre Kritik erst se<strong>in</strong>e<br />

Entstehung ermöglicht haben, dieses Buch gewidmet.<br />

Both sentences contradict Behaghel's Gesetz der wachsenden Glieder, which Lenerz also<br />

lists under the header stylistic tendencies. Lenerz claims that 102b is more acceptable than<br />

102a, because <strong>in</strong> 102b the light element dieses Buch is supported by gewidmet. Sommerfeldt<br />

and Starke (1988: 283ff) mention the same argument as "zu schwacher Klammerrand",<br />

lead<strong>in</strong>g to the extraposition <strong>of</strong> middle field elements <strong>in</strong>to the Nachfeld. Furthermore, they<br />

expla<strong>in</strong> the related phenomenon that, if the middle field is too crowded, some or all <strong>of</strong> its<br />

elements are extraposed.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 113<br />

4. THE INTERACTION OF PREFERENCE RULES, AND SOME<br />

RESTRICTIONS<br />

In the preced<strong>in</strong>g chapter we discussed eleven factors which can have an impact on the order<br />

<strong>of</strong> elements <strong>in</strong> the <strong>German</strong> sentence, namely:<br />

1) Theme-rheme structure<br />

2) Behaghel's "Gesetz der wachsenden Glieder"<br />

3) Functional sentence perspective<br />

4) Semantic-syntactic closeness to the verb ("Verbnähe")<br />

5) The Animacy-first pr<strong>in</strong>ciple<br />

6) Semantic roles<br />

7) Scope<br />

8) Rhythm<br />

9) Natural gender<br />

10) Grammaticalisation (habit)<br />

11) Lenerz' "Satzklammerbed<strong>in</strong>gung"<br />

For demonstration purposes, we discussed the different pr<strong>in</strong>ciples <strong>in</strong>dependently, but all act<br />

and <strong>in</strong>teract <strong>in</strong> every s<strong>in</strong>gle sentence. In 4.1. we shall discuss this <strong>in</strong>teraction and <strong>in</strong> 4.2 the<br />

relative weight <strong>of</strong> certa<strong>in</strong> factors. 4.3. will be dedicated to the calculation <strong>of</strong> acceptability,<br />

and some limitations to the free <strong>in</strong>teraction <strong>of</strong> the pr<strong>in</strong>ciples will be the subject <strong>of</strong> 4.4.<br />

4.1. INTERACTION OF THE PRINCIPLES<br />

All pr<strong>in</strong>ciples discussed <strong>in</strong> 3 apply at the sentence level. However, not all can be dealt <strong>with</strong><br />

purely at sentential level. Instead, some rely on contextual <strong>in</strong>formation (i.e. theme-rheme<br />

structure and functional sentence perspective). The factors can be categorised as pragmatic<br />

(TRS, FSP), semantic (semantic roles), syntactic (grammaticalization, scope), socio-cultural<br />

(natural gender) and prosodical (rhythm) but all have an impact on the syntax, namely the<br />

word order <strong>of</strong> the sentence. The pr<strong>in</strong>ciples are not rigid but are tendencies towards a certa<strong>in</strong>


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 114<br />

order <strong>of</strong> elements. Therefore, contradiction among them does not necessarily make a<br />

sentence ungrammatical.<br />

Some <strong>of</strong> the pr<strong>in</strong>ciples are <strong>in</strong>terrelated. Long and heavy elements tend to be rhematic,<br />

rhemes tend to have a high <strong>in</strong>formational value (sentence perspective), and the scope <strong>of</strong><br />

degree modifiers also tends to be the focus <strong>of</strong> the sentence (sentence perspective). Rhythm,<br />

as well as Behaghel's Gesetz der wachsenden Glieder have been subsumed under the<br />

common header stylistics. The latter restricts the isolated occurrence <strong>of</strong> short verbs, or verb<br />

prefixes, at the end <strong>of</strong> the sentence, and is thus also l<strong>in</strong>ked to Behaghel's Gesetz. Moreover,<br />

there is a relationship between semantic roles and animacy, as some roles are necessarily<br />

animate (e.g. AGENT, EXPERIENCER) whereas others tend to be <strong>in</strong>animate (e.g.<br />

OBJECT). One can assume that Verbnähe and the importance <strong>of</strong> semantic roles are<br />

<strong>in</strong>terrelated, too: AGENTs tend to be realised as subjects, and are thus m<strong>in</strong>imally bound by<br />

the verb, whereas OBJECTs tend to be realised as closely verb-bound direct objects. And<br />

f<strong>in</strong>ally, Verbnähe can be a result <strong>of</strong> grammaticalisation, as can be seen by the rigidity <strong>of</strong><br />

support verb constructions.<br />

The politeness <strong>of</strong> the persons who tend to give precedence to the woman, and the restriction<br />

that some people would rather expect a man to <strong>in</strong>troduce a man to a woman than a woman to<br />

a man (see 3.9) are not <strong>in</strong> an obvious relationship <strong>with</strong> any <strong>of</strong> the other pr<strong>in</strong>ciples. However,<br />

although this factor is worth mention<strong>in</strong>g and might be found amus<strong>in</strong>g, it probably does not<br />

play a very important role compared to some other preference rules.<br />

It is obvious that normally not all preference rules can be satisfied, and that therefore, they<br />

prefer constituent sequences, which are not compatible <strong>with</strong> each other. We want to show<br />

this us<strong>in</strong>g example sentence 1:<br />

1a<br />

Ich habe das sehr dicke Buch e<strong>in</strong>em Mann gegeben.<br />

In 1a, the rules concern<strong>in</strong>g theme-rheme structure (1), scope (7) and rhythm (8) are satisfied,<br />

whereas the rules 2 (wachsende Glieder), 4 (verb bond<strong>in</strong>g), 5 (animacy) and 6 (semantic<br />

roles) are violated. We cannot say anyth<strong>in</strong>g about the subjective factor <strong>of</strong> functional sentence


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 115<br />

perspective (3) as we do not know the speaker's <strong>in</strong>tention but can only deduce that the<br />

speaker apparently assigns more <strong>in</strong>formational value to e<strong>in</strong>em Mann than to das sehr dicke<br />

Buch. Pr<strong>in</strong>ciple 9 (natural gender) does not apply because only a man is referred to <strong>in</strong> the<br />

third person. Preference rule 10 concern<strong>in</strong>g grammaticalisation does not apply here either, as<br />

there occurs no rigid structure <strong>in</strong> example 1, except probably the tendency to the order D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 116<br />

It seems reasonable to assume that the grammaticality <strong>of</strong> sentences is a function <strong>of</strong> the<br />

pr<strong>in</strong>ciples which apply (cf. Jacobs, 1988). Example 1 shows that the violation <strong>of</strong> several<br />

rules does not lead to ungrammaticality and that apparently one rule can have the same<br />

weight as several others.<br />

The <strong>in</strong>teraction <strong>of</strong> the preference rules suggests that grammaticality and ungrammaticality<br />

are merg<strong>in</strong>g concepts. The best sentence is one which satisfies all preference rules and the<br />

more pr<strong>in</strong>ciples are satisfied the better a sentence is. To simplify, one could thus say that our<br />

<strong>in</strong>tuition on the grammaticality <strong>of</strong> sentences can approximately be measured by count<strong>in</strong>g the<br />

pr<strong>in</strong>ciples which are satisfied, and those which are violated. In this assumption we are <strong>in</strong><br />

agreement <strong>with</strong> Sampson (1987):<br />

The majority <strong>of</strong> computational l<strong>in</strong>guists develop systems which analyse NLs us<strong>in</strong>g some form <strong>of</strong><br />

generative grammar which def<strong>in</strong>es a clearcut class <strong>of</strong> well-formed sentences. But computational<br />

l<strong>in</strong>guists who work <strong>with</strong> corpora <strong>of</strong> authentic NL material <strong>of</strong>ten doubt the validity <strong>of</strong> any clearcut<br />

dist<strong>in</strong>ction between grammatical and ungrammatical sequences. Statistics on the different types <strong>of</strong> noun<br />

phrase <strong>in</strong> a 40,000-word sample <strong>of</strong> written English are used to show (i) that there is a cont<strong>in</strong>uous<br />

gradient from very common to very rare constructions, and (ii) that alternative constructions grow more<br />

numerous at lower frequency-levels <strong>in</strong> a regular fashion which implies that a significant proportion <strong>of</strong><br />

grammatical constituents <strong>in</strong> a text will belong to extremely rare types. (Sampson, 1987: 219)<br />

The limitations we see to this approach will be discussed <strong>in</strong> 4.3. and 4.4.<br />

The question <strong>of</strong> whether the pr<strong>in</strong>ciples are universal or <strong>German</strong>-specific has to be exam<strong>in</strong>ed<br />

separately.<br />

4.2. RELATIVE WEIGHT OF SOME PRINCIPLES<br />

Most likely there are also absolute weight differences among order<strong>in</strong>g pr<strong>in</strong>ciples;<br />

for <strong>in</strong>stance, the pr<strong>in</strong>ciple that orders pronouns before nonpronom<strong>in</strong>al NPs seems<br />

stronger throughout than the heav<strong>in</strong>ess pr<strong>in</strong>ciple. (Uszkoreit, 1987: 123)<br />

E<strong>in</strong>e verallgeme<strong>in</strong>erte Beschreibung auch von nichtnormalen Satzgliedfolgen im<br />

Mittelfeld muß e<strong>in</strong>e Darstellung des Zusammenspiels verschiedener,<br />

unterschiedlich gewichteter E<strong>in</strong>flüsse be<strong>in</strong>halten. (Lötscher, 1981: 44)<br />

We saw <strong>in</strong> 4.1 that the pr<strong>in</strong>ciples have different weights, so that sometimes the contradiction<br />

<strong>of</strong> one s<strong>in</strong>gle rule can make a sentence less acceptable than the contradiction <strong>of</strong> several<br />

weaker ones. A rule which turns out to be more important than all other s<strong>in</strong>gle rules is the


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 117<br />

theme-rheme structure (1) (Engel, 1970: 12). Weaker rules are the ones concern<strong>in</strong>g natural<br />

gender (9), Behaghel's Gesetz der wachsenden Glieder (2), rhythm (8) and Lenerz'<br />

Satzklammerbed<strong>in</strong>gung (11).<br />

It is very difficult to f<strong>in</strong>d out exactly how powerful the s<strong>in</strong>gle factors are, as some cannot be<br />

separated. To weigh them properly we would have to be able to isolate them. However, this<br />

is not possible, as some rules, such as semantic roles for <strong>in</strong>stance, apply <strong>in</strong> every s<strong>in</strong>gle<br />

sentence. What we can do, though, is to compare n-tuples <strong>of</strong> sentences which are identical<br />

and to vary one feature only. This is what we have done <strong>in</strong> a test to f<strong>in</strong>d out the relative<br />

importance <strong>of</strong> animacy and def<strong>in</strong>iteness, which are two features which compete strongly, as<br />

well as the unmarked order <strong>of</strong> dative and accusative objects. The tests and the results are<br />

described <strong>in</strong> detail <strong>in</strong> Ste<strong>in</strong>berger (1990).<br />

The reason for do<strong>in</strong>g this test was that Ursula Hoberg and Ulrich Engel base their canonical<br />

forms on different features. Hoberg's is based on the features pronom<strong>in</strong>ality and animacy<br />

(1981: 94, 194), whereas Engel (1988: 326) uses def<strong>in</strong>iteness <strong>in</strong>stead. Both agree <strong>in</strong> that<br />

pronouns have the order N


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 118<br />

Engel's forecast<br />

correct : false<br />

D +d < A +d They 26:17 cases<br />

D +d < A -d do not 48: 0 cases<br />

A +d < D -d agree 1: 5 cases<br />

D -d < A -d<br />

8: 0 cases<br />

Total result: (out <strong>of</strong> 105 cases) 83:22 cases<br />

In the second group <strong>of</strong> NP sequences, Engel and Hoberg do not really disagree, but use<br />

different features for their forecasts, which makes them unable to be compared. If one NP is<br />

animate and def<strong>in</strong>ite, and the other one is <strong>in</strong>animate and <strong>in</strong>def<strong>in</strong>ite, both approaches foresee<br />

the same order. Only if the features are distributed differently do the theses contradict each<br />

other.<br />

The table shows that <strong>in</strong> the cases <strong>in</strong> which Engel and Hoberg do not agree, namely the order<br />

<strong>of</strong> full NPs, Engel's forecast is wrong <strong>in</strong> 21% <strong>of</strong> the cases (22 out <strong>of</strong> 105 occurrences). In one<br />

po<strong>in</strong>t, namely the order <strong>of</strong> an animate accusative and a <strong>in</strong>animate dative, Engel is even<br />

wrong <strong>in</strong> the relation 1:5. Unfortunately, Hoberg does not <strong>in</strong>vestigate the statistical<br />

relevance <strong>of</strong> her assumption so that we do not know whether the use <strong>of</strong> the animacy feature<br />

is more appropriate than the use <strong>of</strong> def<strong>in</strong>iteness. However, Hoberg (1981: 59f) lists several<br />

examples which should underp<strong>in</strong> the relevance <strong>of</strong> her canonical form:<br />

3 Mit e<strong>in</strong>em Fünfzigmarksche<strong>in</strong> rettete der Eßl<strong>in</strong>ger Holzfabrikant Hermann Bisch<strong>of</strong>f e<strong>in</strong>er<br />

Frau das Leben. (D -d+a - A +d-a )<br />

4 ..., was im E<strong>in</strong>zelfall auch bedeuten kann, das persönliche dem allgeme<strong>in</strong>en Wohl<br />

e<strong>in</strong>zuordnen. (A +d-a - D +d-a )<br />

5 ...; er hatte den Tod der Vertreibung von Heimat und Besitz vorgezogen. (A +d-a - D +d-a )<br />

6 Wenn wir den schwarzen Bankbeamten <strong>in</strong> Aruscha dem Stallknecht <strong>in</strong> e<strong>in</strong>em entlegenen<br />

bayrischen Dorf gegenüberstellen ... (A +d+a - D +d+a )<br />

3 is one <strong>of</strong> the cases where an <strong>in</strong>def<strong>in</strong>ite dative precedes a def<strong>in</strong>ite accusative argument. The<br />

strong tendency <strong>of</strong> das Leben <strong>in</strong> das Leben retten suggests to us that the whole term should<br />

probably be categorised as an idiomatic phrase. In this case, the example would not<br />

constitute a counter-example.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 119<br />

In 4, Hoberg is def<strong>in</strong>itely right, as the opposite order would be unnatural. The NPs das<br />

persönliche (Wohl) and das allgeme<strong>in</strong>e Wohl could be <strong>in</strong>terpreted respectively as start<strong>in</strong>g<br />

po<strong>in</strong>t and goal <strong>of</strong> an action. The A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 120<br />

large amount <strong>of</strong> verbs which call for animate dative and <strong>in</strong>animate accusative arguments<br />

expla<strong>in</strong>s why most grammarians assume basic D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 121<br />

The only application <strong>of</strong> such a weigh<strong>in</strong>g mechanism we know <strong>of</strong> was carried out by Jacobs<br />

(1988: 27ff) who uses addition and subtraction as the weigh<strong>in</strong>g function, and an open set <strong>of</strong><br />

values <strong>in</strong>dicat<strong>in</strong>g the importance <strong>of</strong> the pr<strong>in</strong>ciples. The higher the number, the more<br />

important the preference is. Jacobs' pr<strong>in</strong>ciples P1, P2, P7 and P8 are based on semantic roles.<br />

P3, P4 and P5 are more or less related to theme-rheme structure and functional sentence<br />

perspective:<br />

Jacobs' preference rule<br />

assigned score<br />

P1: AGENT < X 3<br />

P2: DATIVE < PATIENT 2<br />

P3: PPro < full-NP 3<br />

P4: def<strong>in</strong>ite < <strong>in</strong>def<strong>in</strong>ite 2<br />

P5: background < focus 1<br />

P6: scope-<strong>in</strong>clud<strong>in</strong>g < scope-<strong>in</strong>cluded<br />

P7: GOAL < THEME<br />

P8: OBJECT < DIRECTIONAL<br />

Accord<strong>in</strong>g to Jacobs (1987: 30), P6 is a super-pr<strong>in</strong>ciple which cannot be assigned a score as<br />

it rules out all the others (cf. our section 4.4.3 below). P7, on the other hand, is very weak<br />

(Jacobs, 1988: 23), which is probably why he did not assign it a score at all. P8 is supposed<br />

to be a strong pr<strong>in</strong>ciple (Jacobs, 1987: 30). Nevertheless, Jacobs mentions it only briefly and<br />

fails to assign it a value <strong>with</strong>out giv<strong>in</strong>g a reason for this.<br />

When a pr<strong>in</strong>ciple is satisfied, its value is added to the preference value <strong>of</strong> the sentence, when<br />

it is violated, its value is subtracted. The <strong>in</strong>tuitive <strong>in</strong>terpretation <strong>of</strong> the score is that the<br />

violation <strong>of</strong> P5, for example, is three times less harmful than the violation <strong>of</strong> P1 or <strong>of</strong> P3.<br />

Jacobs' approach is more f<strong>in</strong>e-gra<strong>in</strong>ed than another one suggested by Erbach (1993). Erbach<br />

recommends to <strong>in</strong>troduce preference <strong>in</strong>to typed feature formalisms such as head-driven<br />

phrase structure grammar (HPSG) by associat<strong>in</strong>g every feature structure <strong>with</strong> a preference<br />

rule (Erbach, 1993: 177):<br />

The preference value is <strong>in</strong>tended to model the degree <strong>of</strong> confidence that the feature structure is an<br />

appropriate representation <strong>of</strong> a l<strong>in</strong>guistic utterance. In the case <strong>of</strong> ambiguity, several feature structures<br />

can be found each <strong>of</strong> which has a preference value. The feature structure <strong>with</strong> the highest preference is<br />

the one which is given most confidence. The order<strong>in</strong>g imposed on the feature structures is just the<br />

numerical order <strong>of</strong> the associated preferences.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 122<br />

For the treatment <strong>of</strong> word order, Erbach suggests us<strong>in</strong>g the subcat list to express unmarked<br />

word order, as is usual <strong>in</strong> HPSG (Pollard/Sag, 1987: 172ff). If the arguments are taken from<br />

the subcat list <strong>in</strong> the unmarked order, the default preference value 1.0 (100%) is assigned. If<br />

the element which has to be bound first is not first on the subcat list, the preference value<br />

decreases (e.g. by the factor 0.8). The further away the element is from the head <strong>of</strong> the list,<br />

the worse the preference. Erbach (1992: 179) shows his idea us<strong>in</strong>g the follow<strong>in</strong>g examples:<br />

7 (weil) der Mann dem Mädchen das Buch gibt.<br />

gibt 1.000<br />

das Buch gibt 1.000<br />

dem Mädchen das Buch gibt 1.000<br />

der Mann dem Mädchen das Buch gibt 1.000<br />

8 (weil) dem Mädchen der Mann das Buch gibt.<br />

gibt 1.000<br />

das Buch gibt 1.000<br />

der Mann das Buch gibt 0.800<br />

dem Mädchen der Mann das Buch gibt 0.800<br />

9 (weil) dem Mädchen das Buch der Mann gibt.<br />

gibt 1.000<br />

der Mann gibt 0.640<br />

das Buch der Mann gibt 0.512<br />

dem Mädchen das Buch der Mann gibt 0.512<br />

In 7, all elements are bound <strong>in</strong> the unmarked order, which corresponds to the order <strong>of</strong> the<br />

subcat list <strong>of</strong> the verb geben:<br />

lexicon (gibt) ==><br />

synsem: local:<br />

(head: cat: v &<br />

subcat: [np(acc), np(dat), np(nom)] ).<br />

The preference for the whole sentence thus is 1.00. In 8, the nom<strong>in</strong>ative NP is bound before<br />

the dative NP, which reduces the preference value to 0.8 because the nom<strong>in</strong>ative NP is the<br />

second element <strong>of</strong> the rema<strong>in</strong><strong>in</strong>g subcat list, and not the first one. Then only is the dative NP<br />

taken. The dative NP is the only rema<strong>in</strong><strong>in</strong>g element on the subcat list and therefore does not<br />

reduce the preference value any further.<br />

In 9, der Mann, which is the third element on the list, is bound first, which reduces the value<br />

to 0.64 (0.8 multiplied by 0.8). Then, das Buch is bound. As the accusative NP is the second<br />

element on the rema<strong>in</strong><strong>in</strong>g subcat list, the preference value is reduced aga<strong>in</strong> by the factor 0.8


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 123<br />

(result<strong>in</strong>g <strong>in</strong> 0,512). The dative NP is the only element left on the list so that the overall<br />

value does not decrease any further. The preference values reflect the fact that 7 is the most<br />

natural and 9 the least natural <strong>of</strong> the three sentences.<br />

After the broad discussion <strong>of</strong> the complexity <strong>of</strong> our word order rules it is obvious that<br />

Erbach's approach is not suitable for the description <strong>of</strong> <strong>German</strong> word order. The problems<br />

l<strong>in</strong>ked to his suggestion are (a) that it uses the subcat list, which automatically limits the<br />

effect to arguments, and (b) that it is far too crude. It does not account for differences<br />

between pronouns and full NPs, for def<strong>in</strong>iteness differences, animacy, rhythm and gender.<br />

The merit <strong>of</strong> Erbach's work, however, is that he <strong>in</strong>troduces preference <strong>in</strong> a formalism which<br />

does not normally allow preference at all (cf. Uszkoreit, 1991: 237). It would be <strong>in</strong>terest<strong>in</strong>g<br />

to see whether it could be applied to the more sophisticated l<strong>in</strong>ear precedence rules<br />

suggested by Uszkoreit (1987), or to comb<strong>in</strong>e them <strong>with</strong> our f<strong>in</strong>d<strong>in</strong>gs on word order<br />

regularities.<br />

As for Jacobs' score calculation model, we do sympathise <strong>with</strong> it. If we do not want to assign<br />

our pr<strong>in</strong>ciples a score here it is because it is difficult to isolate the rules <strong>in</strong> order to weigh<br />

them properly. Furthermore, there are too many <strong>in</strong>consistencies, or hardened structures,<br />

which <strong>in</strong>terfere <strong>with</strong> the free <strong>in</strong>teraction <strong>of</strong> the pr<strong>in</strong>ciples. These are the restrictions<br />

discussed <strong>in</strong> 4.4, but also, and ma<strong>in</strong>ly, hardened structures such as the strict order <strong>of</strong> toners,<br />

the impossibility <strong>of</strong> stress<strong>in</strong>g the accusative neuter pronoun es, and many others which are<br />

listed <strong>in</strong> section 6.1. Instead, we shall use a detailed canonical form <strong>in</strong>clud<strong>in</strong>g flexible<br />

categories such as theme, rheme and focus to describe word order. This alternative method,<br />

as well as the arguments for it, will be discussed <strong>in</strong> section 5.2. For this reason, we shall<br />

leave the subject <strong>of</strong> relative pr<strong>in</strong>ciple weigh<strong>in</strong>g now and discuss some restrictions which<br />

<strong>in</strong>terfere <strong>with</strong> the pr<strong>in</strong>ciples.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 124<br />

4.4. RESTRICTION ON THE INTERACTION OF PREFERENCE<br />

RULES<br />

Several general restrictions apply to the free <strong>in</strong>teraction <strong>of</strong> word-order<strong>in</strong>g pr<strong>in</strong>ciples. These<br />

are the limitation to the sentence or VP level, the occurrence <strong>of</strong> possessive relations or<br />

quantificational elements, and the pragmatic need <strong>of</strong> a sentence focus<br />

("Mitteilungszentrum"). More idiosyncratic limitations will be listed <strong>in</strong> 6.1.<br />

4.4.1. SYNTACTIC SUBORDINATION<br />

We claim <strong>in</strong> this section that modifiers and arguments which are part <strong>of</strong> a PP, NP, AP or an<br />

AdvP, are not subject to the effect <strong>of</strong> the word-order<strong>in</strong>g pr<strong>in</strong>ciples. The <strong>in</strong>fluence <strong>of</strong> the<br />

preference rules is restricted to the verb arguments (<strong>in</strong>clud<strong>in</strong>g the nom<strong>in</strong>ative or subject<br />

argument), and the modifiers which refer to anyth<strong>in</strong>g else, namely to the verb and the<br />

sentence.<br />

Preference rules thus apply to phrases as a whole, <strong>in</strong>clud<strong>in</strong>g their complements and<br />

modifiers. Noun attributes or adjective complements are not determ<strong>in</strong>ed by the word order<br />

preferences and will thus not be described by the canonical form which we shall develop<br />

below, either.<br />

The follow<strong>in</strong>g examples show that subord<strong>in</strong>ate elements do not behave the same way that<br />

phrases at sentence and VP level do:<br />

10 ..., weil ihr Heimweh nach Berl<strong>in</strong> ihm auf die Nerven g<strong>in</strong>g.<br />

11 ..., weil die e<strong>in</strong>es Ordens würdige Nonne ihm Heilung brachte.<br />

12 Gestern brachte der Vater des K<strong>in</strong>des der Hebamme Blumen.<br />

In 10, the PP nach Berl<strong>in</strong> is a PO <strong>of</strong> the noun Heimweh. Therefore, it must be adjacent to the<br />

noun which precedes the pronoun ihm. If the PP was a PO <strong>of</strong> the verb and thus a PO at<br />

sentence or VP level, it would have to follow the pronoun. The same applies for the genitive<br />

complement <strong>of</strong> würdige (e<strong>in</strong>es Ordens), and for the genitive attribute <strong>of</strong> Vater (des K<strong>in</strong>des).<br />

If they were verb complements they would have to follow the dative objects.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 125<br />

In a few cases, this observation may help to disambiguate sentences <strong>with</strong> respect to their<br />

structures, as one possible analysis is either ungrammatical or unlikely. In sentences such as<br />

10 to 12, the verb dependency read<strong>in</strong>g <strong>of</strong> PPs or NPs can be excluded. In 6.6, we shall<br />

briefly discuss how such a disambiguation could be realised <strong>in</strong> a Mach<strong>in</strong>e Translation<br />

system.<br />

More cases <strong>of</strong> ambiguity arise <strong>with</strong> respect to modifiers. These can <strong>of</strong>ten refer either to the<br />

sentence or to a noun (or NP, PP etc). Accord<strong>in</strong>g to Eisenberg (1989: 210), all local and a few<br />

temporal modifiers can be noun attributes. Modal adverbs never occur <strong>with</strong> this function. To<br />

guarantee correct analysis, adverbs need a feature <strong>in</strong>dicat<strong>in</strong>g whether they can modify other<br />

parts <strong>of</strong> speech. In 6.7, we shall propose features which help to f<strong>in</strong>d out whether their position<br />

has to be dealt <strong>with</strong> at sentence level or not. 13 and 14 show an adverb and a PP as attributes to<br />

nouns:<br />

13 Der Mann dort schielt.<br />

14 Die Vorlesung um 12 Uhr wird sicherlich langweilig werden.<br />

Generally, the only situation <strong>in</strong> which one can be sure that a phrase modifies the preced<strong>in</strong>g<br />

noun is <strong>in</strong> the Vorfeld. Dort <strong>in</strong> 13 and um 12 Uhr <strong>in</strong> 14 must be noun modifiers because there<br />

is only space for one constituent <strong>in</strong> the Vorfeld. In the Mittelfeld, nearly all occurrences <strong>of</strong><br />

temporal and local modifiers are ambiguous (15a). 15b is one <strong>of</strong> the few cases where, for<br />

semantic reasons, dort can unambiguously be analysed as a noun attribute, as the woman<br />

cannot be a teacher here and there simultaneously:<br />

15a Ich weiß das, weil die Frau dort früher Lehrer<strong>in</strong> war.<br />

15b Ich weiß das, weil die Frau dort früher hier Lehrer<strong>in</strong> war.<br />

We have seen that it is necessary to dist<strong>in</strong>guish subord<strong>in</strong>ated elements from those which are<br />

not. The former and the latter follow different position<strong>in</strong>g regularities. Another fact worth<br />

po<strong>in</strong>t<strong>in</strong>g out is that some subord<strong>in</strong>ated elements have to be adjacent to their regent (Heimweh<br />

nach Berl<strong>in</strong> <strong>in</strong> 16), whereas others do not (dessen verdächtig <strong>in</strong> 17):<br />

16 ..., weil ihr Heimweh nach Berl<strong>in</strong> dem Mann Angst machte.<br />

* ..., weil ihr Heimweh dem Mann nach Berl<strong>in</strong> Angst machte.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 126<br />

17 Er hat sich dessen gewiß verdächtig gemacht.<br />

Er hat sich gewiß dessen verdächtig gemacht.<br />

4.4.2. POSSESSIVE RELATIONS<br />

Possessive relations also restrict the power <strong>of</strong> preference rules. When there is a possessive<br />

relation <strong>with</strong><strong>in</strong> a clause, the possessor must antecede the possessed (Lenerz, 1977: 99f;<br />

examples from Lenerz):<br />

18a Ich habe das Buch se<strong>in</strong>em Besitzer zurückgegeben.<br />

18b * Ich habe se<strong>in</strong>em Besitzer das Buch zurückgegeben.<br />

(if Besitzer is the possessor <strong>of</strong> das Buch)<br />

Behaghel (1929: 203f) claims that this even applies for implicit possessive relations:<br />

19a Wenn e<strong>in</strong> K<strong>in</strong>d den Eltern gleicht, ist niemand überrascht.<br />

19b * Wenn den Eltern e<strong>in</strong> K<strong>in</strong>d gleicht, ist niemand überrascht.<br />

4.4.3. QUANTIFICATIONAL ELEMENTS<br />

The scope <strong>of</strong> quantifiers is a fundamental part <strong>of</strong> the mean<strong>in</strong>g <strong>of</strong> a sentence, which changes<br />

<strong>with</strong> their different positions <strong>in</strong> the sentence. As the above-mentioned preference rules must<br />

not change the semantics <strong>of</strong> the sentence, the presence <strong>of</strong> quantifiers can counteract the play<br />

<strong>of</strong> the preference rules, or even cancel them out. The mean<strong>in</strong>g <strong>of</strong> 20a, for <strong>in</strong>stance, differs<br />

from that <strong>of</strong> 20b (examples and explanation from Lenerz, 1977: 58):<br />

20a Ich habe vielen Leuten drei Bücher gezeigt.<br />

20b Ich habe drei Bücher vielen Leuten gezeigt.<br />

[Der Unterschied besteht dar<strong>in</strong>, daß (20a)] auch dann wahr [ist], wenn ich nicht jedem der Leute die<br />

gleichen Bücher gezeigt habe. (20b) jedoch ist nur dann wahr, wenn jeder der Leute die gleichen drei<br />

Bücher gezeigt bekommen hat. Unter diesen Bed<strong>in</strong>gungen s<strong>in</strong>d aber alle Verteilungen von def<strong>in</strong>iten und<br />

nicht-def<strong>in</strong>iten Determ<strong>in</strong>atoren sowohl bei IO DO wie bei DO IO akzeptabel. [Lenerz uses IO and DO<br />

for <strong>in</strong>direct and direct objects; number<strong>in</strong>g changed]<br />

The same reason leads Jacobs (1988: 30) to not <strong>in</strong>clude the super pr<strong>in</strong>ciple scope <strong>in</strong> his<br />

acceptability calculation. Even a high score would not be enough to account for the qualitative<br />

difference between the effect <strong>of</strong> scope and the other pr<strong>in</strong>ciples. The fact that scope-<strong>in</strong>clud<strong>in</strong>g<br />

elements have to precede the scope-<strong>in</strong>cluded ones can rule out all other pr<strong>in</strong>ciples. Jacobs<br />

(1988: 30f) assumes that the violation <strong>of</strong> restrictions <strong>in</strong>volv<strong>in</strong>g c-command (<strong>in</strong> terms <strong>of</strong><br />

Government and B<strong>in</strong>d<strong>in</strong>g Theory, Chomsky, 1981) will always make sentences ungrammatical.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 127<br />

4.4.4. THE PRAGMATIC NEED OF A SENTENCE FOCUS<br />

Every sentence needs a focus, or Mitteilungszentrum (central <strong>in</strong>formation unit, or centre <strong>of</strong><br />

<strong>in</strong>terest), <strong>in</strong> order to be acceptable. By focus we understand a potential rheme, or an element<br />

which ranks relatively high on the scale <strong>of</strong> communicative importance. This focus can<br />

consist <strong>of</strong> a rhematic (i.e. unknown) element, or <strong>of</strong> a thematic element which is stressed<br />

contrastively. This constra<strong>in</strong>t is a pragmatic truth which is self-evident, as we normally<br />

speak <strong>in</strong> order to say someth<strong>in</strong>g. This speech act constra<strong>in</strong>t can have an impact on the syntax<br />

<strong>of</strong> sentences.<br />

We have mentioned <strong>in</strong> 3.1.4. and 3.3.1. that pragmatic and situative modifiers normally<br />

separate theme and rheme. Although they can precede the theme, their position beh<strong>in</strong>d the<br />

rheme leads to ungrammatical sentences. For this reason the right-movement <strong>of</strong> elements <strong>of</strong><br />

these modifier groups limits the potentially rhematic doma<strong>in</strong>. The follow<strong>in</strong>g examples show<br />

that the further right the position <strong>of</strong> the existimatorial modifier wohl is, the smaller the<br />

potential rhemes <strong>of</strong> the sentences are (the potential rheme is marked by italics):<br />

21a Der Mann hat wohl der Frau das Buch geschenkt.<br />

21b Der Mann hat der Frau wohl das Buch geschenkt.<br />

21c Der Mann hat der Frau das Buch wohl geSCHENKT.<br />

In 21a, the dative and the accusative objects, as well as the past participle, can carry the<br />

sentence accent, whereas <strong>in</strong> 21b the dative NP cannot. In 21c the focus is limited to the past<br />

participle (cf. 21d and 21e). If the focus is on either <strong>of</strong> the objects, wohl has to precede them<br />

(21f, 21g):<br />

21d * Der Mann hat der FRAU das Buch wohl geschenkt.<br />

21e * Der Mann hat der Frau das BUCH wohl geschenkt.<br />

21f Der Mann hat wohl der FRAU das Buch geschenkt.<br />

21g Der Mann hat der Frau wohl das BUCH geschenkt.<br />

Some <strong>German</strong> native speakers may disagree <strong>with</strong> the harsh acceptability judgement<br />

concern<strong>in</strong>g 21d and want to replace it by ??, but they will certa<strong>in</strong>ly agree that the other<br />

permutations (21a, 21b, 21c, 21f and 21g) are much better.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 128<br />

If <strong>in</strong> a correct sentence there is no typically rhematic element such as an <strong>in</strong>f<strong>in</strong>ite NP, a noun<br />

<strong>of</strong> a support verb construction, a predicative element, a modal modifier or a ma<strong>in</strong> verb<br />

(Hoberg, 1981: 175), a typically thematic element (e.g. a def<strong>in</strong>ite NP) has to function as a<br />

rheme. Consequently, the sentence "Der Mann ist wohl der Präsident" is acceptable <strong>in</strong> 22<br />

but not <strong>in</strong> 23 as the NP der Präsident cannot be rhematic <strong>in</strong> the latter:<br />

22 Context: August, schau mal den stattlichen Herrn da drüben an!<br />

22a Der Mann ist wohl der PräsiDENT.<br />

23 Context: Der Präsident muß irgendwo hier im Raum se<strong>in</strong>!<br />

23b * Der Mann ist wohl der Präsident.<br />

23b could probably be regarded as acceptable if the determ<strong>in</strong>er der is <strong>in</strong>terpreted as a<br />

demonstrative which carries a contrastive accent (DER Mann ist wohl der Präsident). This<br />

has to do <strong>with</strong> the exceptional status <strong>of</strong> the Vorfeld which we shall discuss <strong>in</strong> 5.5.<br />

If the focus is reduced to an element which cannot be stressed, such as the copula verb se<strong>in</strong><br />

for example, the sentence is ungrammatical:<br />

24a ..., weil der Mann wohl der PräsiDENT ist.<br />

24b * ..., weil der Mann der Präsident wohl ist.<br />

The toner wohl cannot itself be focused, as toners are generally not focusable (Bußmann,<br />

1983: 325) 66 and therefore cannot carry the sentence accent. However, it would be wrong to<br />

conclude that these particles cannot follow def<strong>in</strong>ite NPs, or that they must not f<strong>in</strong>ish the<br />

sentence. If there is another possible rheme or focus <strong>in</strong> the sentence (26), or if the f<strong>in</strong>ite verb<br />

<strong>in</strong> verb-second position can be focused (27), this order is perfectly possible:<br />

26 Der Mann ist dem Präsidenten wohl zuWIder.<br />

27 Der Mann SCHENKte das Buch dem Präsidenten wohl.<br />

A topicalised verb complement can carry the sentence focus, too (28a). However, if it is<br />

moved back to the Mittelfeld, it has to follow the particle which limits the potential focus<br />

66 Wohl <strong>in</strong> 4 is no exception to this regularity. Instead, it is a homonym to wohl <strong>in</strong> 3 (see modifier list <strong>in</strong><br />

appendix 8.2):<br />

25 Der M<strong>in</strong>ister lügt sehr WOHL.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 129<br />

(28b versus 28c). Alternatively, the verb has to function as the sentence focus (28d). The<br />

phenomenon that, even if the potential rheme is limited to the right, the Vorfeld element and<br />

the verb <strong>in</strong> verb-second position can function as focus will be discussed <strong>in</strong> 5.5:<br />

28a Dem MANN gab er das Buch wohl.<br />

28b Er gab das Buch wohl dem MANN.<br />

28c * Er gab dem MANN das Buch wohl.<br />

28d Er GAB dem Mann das Buch wohl.<br />

The compulsory existence <strong>of</strong> a sentence focus must be taken <strong>in</strong>to consideration when<br />

describ<strong>in</strong>g word order variations, as we do not want to exclude sentences which would be<br />

correct if this constra<strong>in</strong>t was satisfied.<br />

In this context we wish to mention another regularity: A simple sentence not only has to<br />

have at least one potential focus, but it should not have more than one either (Altmann,<br />

1976: 30f; Rochemont, 1989: 6). In 6.2, we shall list the constructions which can <strong>in</strong>flict<br />

focus on phrases. In 29, jetzt is focused by the degree modifier erst, and the pronoun er must<br />

be stressed because <strong>of</strong> its position beh<strong>in</strong>d the adverb deshalb:<br />

29 * Erst JETZT hat deshalb ER ihn besucht.<br />

Note that <strong>in</strong> some cases context can also call for the focus<strong>in</strong>g <strong>of</strong> elements. 30 is<br />

ungrammatical because the degree modifier nur focuses er, and Sibylle carries a contrastive<br />

accent because <strong>of</strong> the context:<br />

30 Context: Peter hat nicht MARIA geküßt.<br />

30a * Nur ER hat SiBYLle geküßt.<br />

4.5. SUMMARY<br />

The goal <strong>of</strong> this chapter was to show how the preference rules described <strong>in</strong> chapter 3<br />

<strong>in</strong>teract. The factors can have different weights, and they can favour compet<strong>in</strong>g word orders.<br />

Our <strong>in</strong>tuition on whether a phrase sequence is natural depends on how many pr<strong>in</strong>ciples<br />

favour it, and on how strong these pr<strong>in</strong>ciples are (cf. 4.1 and 4.2). In 4.3, we presented two<br />

models to imitate people's <strong>in</strong>tuitions on which word order is most natural. Jacobs (1987)<br />

suggested to give each preference rule a weight. If the rule is satisfied <strong>in</strong> a given sentence,


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 130<br />

the value is added to the acceptability score <strong>of</strong> the sentence, if it is violated, its score is<br />

subtracted. The sentence <strong>with</strong> the the highest score is the most natural.<br />

Another proposition is made by Erbach (1993), who claims that the most natural sentences<br />

are the ones whose phrases are ordered accord<strong>in</strong>g to the obliqueness hierarchy. Each<br />

diversion from this ideal order decreases the acceptability score <strong>of</strong> the sentence by a factor<br />

0.8. Although this latter approach seems far too crude, as it does not account for the vary<strong>in</strong>g<br />

effect <strong>of</strong> different variations, and as it does not <strong>in</strong>clude the position <strong>of</strong> modifiers, it could<br />

probably be modified. The <strong>in</strong>terest<strong>in</strong>g part <strong>of</strong> Erbach's proposition is that he <strong>in</strong>troduces the<br />

notion <strong>of</strong> preference <strong>in</strong> feature-based formalisms, so that it could be applied to more<br />

sophisticated preference rules.<br />

In 4.4, we show that the phrase order<strong>in</strong>g pr<strong>in</strong>ciples only apply to verb arguments, and to<br />

sentence and verb modifiers. Subord<strong>in</strong>ate elements, such as noun arguments and modifiers<br />

behave differently (4.4.1). Further factors which can <strong>in</strong>terfere <strong>with</strong> the free <strong>in</strong>teraction <strong>of</strong><br />

position<strong>in</strong>g pr<strong>in</strong>ciples are possessive relations (4.4.2) and the scope <strong>of</strong> quantificational<br />

elements (4.4.3). 4.4.4 f<strong>in</strong>ally shows that the condition that every sentence needs a possible<br />

rheme can also have an impact on order<strong>in</strong>g, as rheme-less sentences must be avoided. The<br />

latter condition ma<strong>in</strong>ly concerns the position <strong>of</strong> the subgroup <strong>of</strong> modifiers which restrict the<br />

possible rheme <strong>of</strong> a sentence, and which cannot be rhematic themselves.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 131<br />

5. HOW TO DESCRIBE GERMAN FREE WORD ORDER<br />

FORMALLY<br />

In this chapter we mention three formal possibilities <strong>of</strong> how to describe a free word order<br />

language (5.1). One <strong>of</strong> these methods is the calculation <strong>of</strong> the score <strong>of</strong> different permutation<br />

variations, as mentioned <strong>in</strong> section 4.3. Another is the use <strong>of</strong> complex l<strong>in</strong>ear precedence (LP)<br />

rules, suggested by Uszkoreit (1987). The third possibility is the use <strong>of</strong> a canonical form<br />

(5.2). We claim that the latter method is more efficient for applied l<strong>in</strong>guistics. We then<br />

discuss what the term canonical form (CF) implies for a free word order language, and<br />

discuss those suggested by Hoberg (1981), Engel (1988) and Ste<strong>in</strong>berger (1990) (5.3). The<br />

latter will be developed further <strong>in</strong> chapter 6. In the follow<strong>in</strong>g sections, we discuss why some<br />

sentences do not comply <strong>with</strong> our CF (5.4), exam<strong>in</strong>e the role <strong>of</strong> the Vorfeld (5.5) and show<br />

how important it can be for translation to recognise theme, rheme and focus <strong>of</strong> a sentence<br />

(5.6).<br />

5.1. ACCEPTABILITY CALCULATION AND LP RULE DISJUNCTION<br />

In 4.3, we discussed the possibility <strong>of</strong> describ<strong>in</strong>g <strong>German</strong> word order by calculat<strong>in</strong>g the<br />

score <strong>of</strong> different permutation variations. This method would not only be helpful for the<br />

acceptability comparison <strong>of</strong> several permutation variations <strong>in</strong> analysis, but it could also be<br />

used <strong>in</strong> synthesis, by generat<strong>in</strong>g all possible variations, and then choos<strong>in</strong>g the one <strong>with</strong> the<br />

highest score, <strong>in</strong> order to get the most acceptable or natural one. However, we mentioned <strong>in</strong><br />

the same section that this procedure is problematic, because the calculation does not take<br />

<strong>in</strong>to account the restrictions formulated <strong>in</strong> 4.4. Rigid sequences such as support verb<br />

constructions, the fixed order <strong>of</strong> toners, and other idiosyncrasies could not be accounted for<br />

either.<br />

One could possibly avoid the problem by giv<strong>in</strong>g the rules <strong>in</strong>volv<strong>in</strong>g these regularities<br />

extraord<strong>in</strong>arily high scores. Alternatively, one could th<strong>in</strong>k <strong>of</strong> hav<strong>in</strong>g an <strong>in</strong>teraction <strong>of</strong>


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 132<br />

preference and hard rules, the latter <strong>of</strong> which must not be violated. Although the named<br />

possibilities should not be excluded, this method seems relatively complicated. Furthermore,<br />

it is computationally expensive, as all possible variations would have to be created and<br />

compared <strong>in</strong> sentence generation, before the word order <strong>with</strong> the highest score could be<br />

identified.<br />

An alternative method was suggested for the treatment <strong>of</strong> free word order <strong>in</strong> Generalized<br />

Phrase Structure Grammar (GPSG) by Uszkoreit (1987: 115ff). In GPSG, L<strong>in</strong>ear<br />

Precedence (LP) Rules are traditionally used to order phrases <strong>in</strong> the sentence. Some <strong>of</strong> the<br />

rules that Uszkoreit (1987: 114 and 167) suggests concern compulsory word order<strong>in</strong>g<br />

regularities, such as the position <strong>of</strong> the f<strong>in</strong>ite verb <strong>in</strong> ma<strong>in</strong> and subord<strong>in</strong>ate clauses, and the<br />

position <strong>of</strong> the conjunction daß <strong>in</strong> the subord<strong>in</strong>ate clause. A further subset <strong>of</strong> Uszkoreit's LP<br />

rules is more relevant for our purpose, as it is related to the preference rules listed <strong>in</strong> chapter 3:<br />

A) +NOM < +DAT<br />

B) +NOM < +ACC<br />

C) +DAT < +ACC<br />

D) -FOCUS < +FOCUS<br />

E) +PRONOUN < -PRONOUN<br />

F) +TOP < X (The topic precedes everyth<strong>in</strong>g else)<br />

Uszkoreit (1987: 118) concedes that these rules are only a subset <strong>of</strong> the rules apply<strong>in</strong>g <strong>in</strong><br />

<strong>German</strong>. They are all related to the preference rules mentioned earlier. LP rules A and B are<br />

vaguely l<strong>in</strong>ked to the preference rules discussed <strong>in</strong> 3.4 (verb bond<strong>in</strong>g), 3.5 (animacy-first)<br />

and 3.6 (semantic roles). Rule D corresponds to functional sentence perspective (3.3), and E<br />

and F are part <strong>of</strong> what we discussed <strong>in</strong> 3.1 (theme-rheme structure). Concern<strong>in</strong>g the order<strong>in</strong>g<br />

<strong>of</strong> dative and accusative arguments (C), we expressed our reservations <strong>in</strong> 4.2.<br />

As far as we are concerned however, the ma<strong>in</strong> <strong>in</strong>terest <strong>of</strong> Uszkoreit's word order treatment is<br />

that he solves a major problem l<strong>in</strong>ked to LP rules <strong>in</strong> standard GPSG, namely that the<br />

violation <strong>of</strong> one <strong>of</strong> these rules <strong>in</strong>volves the rejection <strong>of</strong> the whole structure. We have seen <strong>in</strong><br />

4.3 that even several preference rules can be violated <strong>with</strong>out the sentence be<strong>in</strong>g


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 133<br />

ungrammatical. Uszkoreit (1987: 115ff) solves this problem by comb<strong>in</strong><strong>in</strong>g the atomic LP<br />

rules <strong>in</strong>to one complex LP rule, consist<strong>in</strong>g <strong>of</strong> a major LP disjunction. The idea is that<br />

any <strong>of</strong> the atomic LP rules <strong>with</strong><strong>in</strong> the complex LP rule can be violated as long as the violations are<br />

sanctioned by at least one <strong>of</strong> the atomic LP rules. (Uszkoreit, 1987: 118)<br />

Only if none <strong>of</strong> the LP rules legitimises the violation <strong>of</strong> the other(s) is a sentence<br />

ungrammatical. Uszkoreit exemplifies his method us<strong>in</strong>g the sentences 1 to 3:<br />

1 Dann will der Doktor dem Mann die Pille geben.<br />

-FOCUS +FOCUS -FOCUS<br />

+NOM +DAT +ACC<br />

2 Dann will der Doktor die Pille dem Mann geben.<br />

-FOCUS -FOCUS +FOCUS<br />

+NOM +ACC +DAT<br />

3 ?? Dann will der Doktor die Pille dem Mann geben.<br />

-FOCUS +FOCUS -FOCUS<br />

+NOM +ACC +DAT<br />

Accord<strong>in</strong>g to his model, 1 and 2 are grammatical, whereas 3 is not, as <strong>in</strong> 3 the atomic LP<br />

rules +DAT < +ACC and -FOCUS < +FOCUS are violated, and no other atomic LP rule<br />

legitimises these violations.<br />

Uszkoreit's model is <strong>in</strong>terest<strong>in</strong>g <strong>in</strong> that it <strong>in</strong>troduces a way to deal <strong>with</strong> the comb<strong>in</strong>ation <strong>of</strong><br />

several weak word order<strong>in</strong>g rules <strong>in</strong> sentence analysis. In synthesis, it would generate all<br />

permutations sanctioned by any atomic LP rule. Although this is a good result for a<br />

formalism which <strong>in</strong>tends to imitate the speaker's competence, it generates too many<br />

sentences for a Mach<strong>in</strong>e Translation system, where one s<strong>in</strong>gle word sequence is required.<br />

In the follow<strong>in</strong>g sections, we suggest the use <strong>of</strong> a flexible canonical form<br />

(Grundwortstellung) for Mach<strong>in</strong>e Translation.<br />

5.2. THE RELEVANCE OF A CANONICAL FORM FOR GERMAN<br />

The canonical form <strong>of</strong> a language <strong>in</strong>dicates the sequence <strong>of</strong> different elements <strong>in</strong> the<br />

sentence. If a language has fixed word order <strong>with</strong> an obligatory place for subject, objects, PO<br />

and modifiers, the canonical form is relatively simple. In <strong>German</strong>, however, we encounter at


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 134<br />

least two complications. First, we have to dist<strong>in</strong>guish between the different realizations <strong>of</strong><br />

the verb arguments as pronoun, def<strong>in</strong>ite NP/PP or <strong>in</strong>def<strong>in</strong>ite NP/PP. This is because<br />

pronouns tend to precede full NPs, and <strong>in</strong>def<strong>in</strong>ite NPs normally follow both pronouns and<br />

def<strong>in</strong>ite NPs. Second, we have to be able to cope <strong>with</strong> the fact that even one s<strong>in</strong>gle subgroup,<br />

such as that <strong>of</strong> def<strong>in</strong>ite object NPs, does not always take the same place. Instead, its position<br />

varies depend<strong>in</strong>g on the theme-rheme structure, on the speaker's <strong>in</strong>tention as to what the<br />

sentence focus should be, the semantic roles, the animacy <strong>of</strong> the referents, and others.<br />

<strong>German</strong> word order is the result <strong>of</strong> the <strong>in</strong>teraction <strong>of</strong> various order<strong>in</strong>g pr<strong>in</strong>ciples. The natural<br />

or unmarked order varies depend<strong>in</strong>g on the parametrisation <strong>of</strong> the s<strong>in</strong>gle factors. Jacobs<br />

(1988: 29) concludes that the pr<strong>in</strong>ciples apply <strong>in</strong> every s<strong>in</strong>gle sentence, even <strong>in</strong> the ones <strong>with</strong><br />

most natural or normal word order. As long as all parameters rema<strong>in</strong> constant there is one<br />

s<strong>in</strong>gle unmarked word order. Any <strong>German</strong> canonical form can thus represent only one<br />

possible order <strong>of</strong> elements. We want to show this aga<strong>in</strong> <strong>in</strong> the follow<strong>in</strong>g three examples:<br />

4 Without context: Er hat es ihm gestern geliehen.<br />

5 Context: Wem hat er es gestern geliehen?<br />

5a ?? Er hat es IHM gestern geliehen.<br />

5b Er hat es gestern IHM geliehen.<br />

6 Context: Wann hat er es ihm geliehen?<br />

6a * Er hat es GEStern ihm geliehen.<br />

6b Er hat es ihm GEStern geliehen.<br />

The sentences differ <strong>with</strong> respect to their theme-rheme structure. The elements which are<br />

asked for are the rhemes or focuses <strong>of</strong> the sentences. <strong>Word</strong> order <strong>in</strong> 4, 5a and 6b complies<br />

<strong>with</strong> Engel's canonical form (5.3.1), and <strong>in</strong>deed it seems to be the most natural one <strong>in</strong><br />

isolated sentences (4). However, 5a is considerably less good an answer to 5 than 5b, as the<br />

rhematic element ihm <strong>in</strong> 5a precedes the thematic element gestern. Context 5 requires the<br />

answer 5b. As an answer to 6, the same order is unacceptable. Context 6 requires the order<br />

<strong>of</strong> Engel's canonical form (5.3.1).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 135<br />

It is obvious that no canonical form satisfies all context requirements. Because <strong>of</strong> vary<strong>in</strong>g<br />

parametrisation, every context requires its own canonical form <strong>in</strong>stead. A fixed canonical<br />

form which would produce correct sentences <strong>in</strong> all possible contexts does not exist.<br />

Due to this conflict, we would like to put forward the view that canonical form describes the<br />

order <strong>of</strong> elements which generates natural sentences <strong>in</strong> the highest possible number <strong>of</strong><br />

contexts. This def<strong>in</strong>ition co<strong>in</strong>cides <strong>with</strong> what Höhle (1982) calls "normale Wortstellung"<br />

(summarised by Eisenberg, 1989: 420):<br />

Je mehr mögliche Rhemata e<strong>in</strong> Satz hat, desto normaler ist er. Denn je mehr mögliche Rhemata er hat,<br />

<strong>in</strong> desto mehr Kontexten kann er stehen. Der Satz mit der maximalen Zahl von Rhemata ist kontextuell<br />

am wenigsten restr<strong>in</strong>giert und <strong>in</strong>s<strong>of</strong>ern normal. Zwei Sätze mit derselben Anzahl von Rhemata s<strong>in</strong>d<br />

gleich normal.<br />

Eisenberg's example sentences:<br />

7a<br />

7b<br />

Emma hat dem Studenten das Auto gegeben.<br />

Emma hat das Auto dem Studenten gegeben.<br />

have the same amount <strong>of</strong> rhemes. Therefore, each variation is as normal or basic as the<br />

other. This does not necessarily mean that the serialisations <strong>in</strong> 7a and 7b statistically are<br />

equally frequent, as the number <strong>of</strong> occurrences <strong>of</strong> different contexts is not the same.<br />

Eisenberg (1989: 420ff) suggests several tests which help to f<strong>in</strong>d the unmarked order <strong>in</strong> the<br />

sense expla<strong>in</strong>ed earlier. The question test, for <strong>in</strong>stance, shows how many elements <strong>of</strong> a given<br />

sentence are possible rhemes. The pr<strong>in</strong>ciple <strong>of</strong> the question test is to ask for every s<strong>in</strong>gle<br />

phrase <strong>in</strong> the sentence, and to exam<strong>in</strong>e whether the sentence is 7a possible answer to this<br />

question. One problem we see <strong>with</strong> this method is that it does not account for more f<strong>in</strong>etuned<br />

acceptability differences. It is our <strong>in</strong>tuition, for example, that 7a is a better answer to<br />

the question 8c than it is to the question 9e (compare 8a and 9a). In our <strong>in</strong>tuition, 7a is<br />

<strong>in</strong>deed more natural than 7b:<br />

8 Wem hat Emma das Auto gegeben?<br />

8a Emma hat dem Studenten das Auto gegeben.<br />

9 Was hat Emma dem Studenten gegeben?<br />

9a ? Emma hat das Auto dem Studenten gegeben.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 136<br />

A similar test is the accentuation test (Engel, 1989: 422f). The goal is to stress every s<strong>in</strong>gle<br />

phrase <strong>of</strong> a sentence and to count the amount <strong>of</strong> stressable constituents. The more phrases<br />

can be stressed, the more normal the sentence. This test does not really differ from the<br />

question test, as the exact purpose <strong>of</strong> the questions is to focus on different constituents.<br />

A better alternative is the def<strong>in</strong>ite/<strong>in</strong>def<strong>in</strong>ite test. Eisenberg (1989: 423) suggests to test all<br />

word order variations and all def<strong>in</strong>ite/<strong>in</strong>def<strong>in</strong>ite comb<strong>in</strong>ations to see which word order<br />

allows for more comb<strong>in</strong>ations:<br />

10 Emma hat dem Studenten e<strong>in</strong> Auto geliehen.<br />

11 Emma hat e<strong>in</strong>em Studenten das Auto geliehen.<br />

12 Emma hat das Auto e<strong>in</strong>em Studenten geliehen.<br />

13 * Emma hat e<strong>in</strong> Auto dem Studenten geliehen.<br />

Eisenberg po<strong>in</strong>ts out that the focus should lie on a constituent which is not <strong>in</strong>volved <strong>in</strong> the<br />

variations <strong>of</strong> order, such as the subject or the verb. The result <strong>of</strong> the test is that the order<br />

D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 137<br />

The order <strong>in</strong> 14c for example, <strong>in</strong>volv<strong>in</strong>g a def<strong>in</strong>ite dative NP, seems to us as acceptable as<br />

the one <strong>in</strong> 14a. However, the order <strong>in</strong> 14a seems natural even <strong>with</strong> a def<strong>in</strong>ite dative<br />

complement (14d):<br />

14c<br />

14d<br />

dem Schüler mit e<strong>in</strong>em Füller e<strong>in</strong>e Note <strong>in</strong> e<strong>in</strong> Heft schreiben.<br />

mit e<strong>in</strong>em Füller dem Schüler e<strong>in</strong>e Note <strong>in</strong> e<strong>in</strong> Heft schreiben.<br />

Although all the tests are not <strong>with</strong>out contradictions, they are helpful as they can underp<strong>in</strong><br />

our <strong>in</strong>tuition. Note, however, that <strong>in</strong>tuition plays an important role <strong>in</strong> all tests. The only non<strong>in</strong>tuitive<br />

evidence for word order distribution we can th<strong>in</strong>k <strong>of</strong> is statistical data. It would be<br />

<strong>in</strong>terest<strong>in</strong>g to see how frequent different sequences are <strong>in</strong> multi-million word corpora. Note,<br />

however, that even statistical data is not a reliable source for unmarkedness because real-life<br />

texts are contextually bound, and not all contexts occur <strong>with</strong> equal frequency.<br />

Although the canonical form statistically does not necessarily represent the most frequent<br />

order (cf. Engel, 1988: 304 and Eisenberg, 1989: 421), there is a strong reason for assum<strong>in</strong>g<br />

a canonical form. It seems reasonable to assume one order as the basic word order so as to<br />

take it as a po<strong>in</strong>t <strong>of</strong> departure for the description <strong>of</strong> variations. In order to discuss<br />

serialisation variations, we shall use the movement metaphor such as: "The result <strong>of</strong> mov<strong>in</strong>g<br />

a dative pronoun beh<strong>in</strong>d a def<strong>in</strong>ite accusative object NP is that the pronoun has to be<br />

stressed contrastively". We want to po<strong>in</strong>t out, however, that we do not <strong>in</strong>tend to imply by<br />

us<strong>in</strong>g this formulation that we follow a transformational theory such as Government and<br />

B<strong>in</strong>d<strong>in</strong>g (e.g. Chomsky, 1981) which is based on the transformation move-alpha and deep<br />

and surface structures.<br />

5.3. CANONICAL FORMS FOR GERMAN IN LITERATURE<br />

In this chapter, we want to present, compare and evaluate the canonical forms suggested by<br />

Engel (1988) and Hoberg (1981). We shall exam<strong>in</strong>e the few sequences where they contradict<br />

each other and suggest a new canonical form <strong>in</strong> which we try to comb<strong>in</strong>e the advantages <strong>of</strong><br />

both models.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 138<br />

5.3.1. ENGEL’S CANONICAL FORM<br />

Engel's canonical form (Grundwortstellung) <strong>in</strong> Deutsche Grammatik (1988: 326) is the<br />

follow<strong>in</strong>g (we replace Hoberg's and Engel's conventions by ours <strong>in</strong> order to keep the<br />

description as clear and consistent as possible):<br />

N pron / N +d < (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 139<br />

groups any further as it has been already mentioned <strong>in</strong> 2.4. Instead, we shall concentrate on<br />

Hoberg's classification which is more f<strong>in</strong>e-tuned.<br />

There are several comments to make <strong>with</strong> respect to Engel's order. Our tests (cf. section 4.2<br />

and Ste<strong>in</strong>berger, 1990: annex, Iff) have shown that the unmarked order <strong>of</strong> verb objects is not<br />

D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 140<br />

(N < A < D / G / Nom / Adj) pron < (N < A < D / G / PO) +a <<br />

^<br />

^<br />

a pragm<br />

a sit<br />

(a 1 -a 18 ) (a 19 -a 40 )<br />

< (N < A) -a < (D / G / PO / Sit / Dir) -a <<br />

^<br />

a neg


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 141<br />

As we have po<strong>in</strong>ted out <strong>in</strong> section 4.2 the def<strong>in</strong>iteness feature is more powerful than the one<br />

concern<strong>in</strong>g animacy. Consequently, Hoberg's canonical form must be complemented by the<br />

def<strong>in</strong>iteness feature. In the follow<strong>in</strong>g section, we shall develop a new canonical form based<br />

on Engel's and Hoberg's.<br />

5.3.3. NEW PRELIMINARY CANONICAL FORM<br />

In a new canonical form both features, animacy and def<strong>in</strong>iteness, should be present, <strong>with</strong><br />

def<strong>in</strong>iteness be<strong>in</strong>g super-ord<strong>in</strong>ated to animacy (cf. section 4.2). These features should also<br />

apply to POs. We want to present here a prelim<strong>in</strong>ary version <strong>of</strong> this new canonical form. It is<br />

based on earlier work carried out <strong>in</strong> Ste<strong>in</strong>berger (1990), <strong>with</strong> m<strong>in</strong>or changes. We compiled it<br />

by comb<strong>in</strong><strong>in</strong>g groups <strong>of</strong> two constituents or more <strong>in</strong> hundreds <strong>of</strong> sentence pairs <strong>with</strong> vary<strong>in</strong>g<br />

word order. We decided <strong>in</strong>tuitively for each sentence pair which order is more natural, and<br />

comb<strong>in</strong>ed these groups <strong>of</strong> two <strong>in</strong>to a complex canonical form. A basic assumption <strong>of</strong> our work<br />

was that word sequences are transitive. This means that from A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 142<br />

Interpretation: Accord<strong>in</strong>g to this canonical form, the verb complements serialise <strong>in</strong> the<br />

follow<strong>in</strong>g order:<br />

A) pronom<strong>in</strong>al or def<strong>in</strong>ite and animate subject NPs (=nom<strong>in</strong>ative NPs)<br />

B) pronom<strong>in</strong>al complements <strong>in</strong> the accusative and dative cases, pronom<strong>in</strong>al nom<strong>in</strong>al or<br />

adjectival complements<br />

C) nom<strong>in</strong>ative complements <strong>with</strong> the features [+d,-a]<br />

D) nom<strong>in</strong>ative complements <strong>with</strong> the features [-d,+a]<br />

E) animate and def<strong>in</strong>ite accusative and dative complements<br />

F) pronom<strong>in</strong>al genitive complements<br />

G) <strong>in</strong>def<strong>in</strong>ite and <strong>in</strong>animate nom<strong>in</strong>ative complements<br />

H) def<strong>in</strong>ite and <strong>in</strong>animate accusative and dative complements<br />

I) accusative and dative complements <strong>with</strong> the features [-d,+a]<br />

K) pronom<strong>in</strong>al prepositional objects<br />

L) accusative and dative complements <strong>with</strong> the features [-d,-a]<br />

M) prepositional complements <strong>with</strong> the features <strong>in</strong> the order [+d,+a], [+d,-a], [d,+a],<br />

[-d,-a]<br />

N) nom<strong>in</strong>al genitive complements, <strong>in</strong>dependently from the features<br />

O) situative, directive or expansive complements<br />

P) nom<strong>in</strong>al or adjectival complements <strong>in</strong> full form<br />

Q) NPs or PPs <strong>of</strong> support verb constructions<br />

We take over Hoberg's f<strong>in</strong>e-tuned modifier classification. Although we restrict ourselves to<br />

the treatment <strong>of</strong> one-word modifiers here, we shall not omit <strong>in</strong> the canonical form the<br />

modifier classes which can only be realised by PPs. The reason is that it is easier for the user<br />

to take out some classes, which are not relevant for some purposes and applications, than it<br />

is to <strong>in</strong>sert new classes when a complete canonical form is needed. In our Mach<strong>in</strong>e<br />

Translation implementation us<strong>in</strong>g the CAT2 formalism for <strong>in</strong>stance, we reduced the 44<br />

modifier groups to 17, as PPs and toners were reduced to one s<strong>in</strong>gle class each (Ste<strong>in</strong>berger,<br />

1992a: 28f and 38ff). Some modifier classes are very text-specific, so that several classes<br />

can be comb<strong>in</strong>ed <strong>with</strong><strong>in</strong> sublanguages such as the language used <strong>in</strong> technical manuals.<br />

The groups <strong>of</strong> pragmatic (a 1 -a 18 ) and situative modifiers (a 19 -a 40 ) follow pronouns and<br />

def<strong>in</strong>ite nom<strong>in</strong>ative complements, and they precede <strong>in</strong>def<strong>in</strong>ite accusative and dative<br />

complements. The negational modifier a 41 and the modal modifier group a 42


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 143<br />

("Konkomitanzangabe") 69 follow the situative and pragmatic modifier groups and the<br />

def<strong>in</strong>ite accusative and dative arguments. The ma<strong>in</strong> modal modifier group a 43 either<br />

precedes <strong>in</strong>def<strong>in</strong>ite and animate accusative and dative complements, or is placed <strong>in</strong> between<br />

them, or follows them. The <strong>in</strong>strumental modifier group a 44 , which Hoberg classifies as a<br />

modal modifier, follows the other modifiers and thus precedes pronom<strong>in</strong>al and <strong>in</strong>def<strong>in</strong>ite<br />

verb complements <strong>in</strong> accusative and dative cases.<br />

It is difficult to be more precise <strong>with</strong> respect to the pragmatic and situative modifiers, as<br />

their position can vary greatly. In section 6.5 we shall narrow down their position. In<br />

Ste<strong>in</strong>berger (1990: 73ff), we positioned pragmatic modifiers after pronom<strong>in</strong>al genitive<br />

complements, and situatives <strong>in</strong> the area between pragmatic modifiers and the negational<br />

modifier. Although a more restricted position would be desirable, we believe that it would be<br />

wrong to limit the places where modifiers can occur <strong>in</strong> that way. Pragmatic and situative<br />

modifiers can thus appear at both extremes <strong>of</strong> the range. In 19, the adverbs should precede<br />

the <strong>in</strong>def<strong>in</strong>ite nom<strong>in</strong>ative. 20 shows that their occurrence beh<strong>in</strong>d a <strong>in</strong>animate def<strong>in</strong>ite<br />

accusative is also perfectly natural:<br />

19a Deshalb erschien irrtümlicherweise 14 / gestern 26 e<strong>in</strong> Mann.<br />

19b ?? Deshalb erschien e<strong>in</strong> Mann irrtümlicherweise 14 / gestern 26 .<br />

20a<br />

20b<br />

Sie schenkte irrtümlicherweise 14 / gestern 26 das Buch e<strong>in</strong>em Fremden.<br />

Sie schenkte das Buch irrtümlicherweise 14 / gestern 26 e<strong>in</strong>em Fremden.<br />

This canonical form differs from Hoberg's and Engel's ma<strong>in</strong>ly <strong>in</strong> that ours is more detailed:<br />

NP and PP objects are split <strong>in</strong>to four subgroups differ<strong>in</strong>g <strong>in</strong> the features [+/-animate] and<br />

[+/-def<strong>in</strong>ite]. Furthermore, we placed the modifier groups a 42 to a 44 <strong>in</strong> separate places as,<br />

accord<strong>in</strong>g to our <strong>in</strong>tuition, they behave differently. We also changed a few sequences, such<br />

as, for <strong>in</strong>stance, <strong>in</strong> the doma<strong>in</strong> <strong>of</strong> the nom<strong>in</strong>ative complements. These changes are aga<strong>in</strong> due<br />

69 The term Konkomitanz is based on the Lat<strong>in</strong> word concomitatus which means accompanied (Bußmann,<br />

1983: 260). Konkomitanzangaben are adverbials which refer to accompany<strong>in</strong>g persons (Hoberg, 1981:<br />

129), such as mit den K<strong>in</strong>dern or mite<strong>in</strong>ander. Throughout this work, we shall use the <strong>German</strong> term.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 144<br />

to our <strong>in</strong>tuition, based on our own tests, <strong>in</strong> which we evaluated the markedness <strong>of</strong> sets <strong>of</strong><br />

sentences <strong>in</strong>volv<strong>in</strong>g representatives <strong>of</strong> different position classes.<br />

The follow<strong>in</strong>g examples verify the order <strong>of</strong> the new canonical form (examples accord<strong>in</strong>g to<br />

Hoberg, 1981: 152ff):<br />

21 ...; den S<strong>in</strong>n von dramatisch prägt mir vielleicht 12 e<strong>in</strong> Wortwechsel e<strong>in</strong>.<br />

(D pron -a 12(=pragm) -A -d-a )<br />

22 Ich überlegte auch, ob ich wirklich 15 die Lauretanische Litanei s<strong>in</strong>gen sollte...<br />

(N pron -a 15(=pragm) -A +d-a )<br />

23 ... <strong>in</strong> die Lage kommen, für sich oder se<strong>in</strong>e Angehörigen 30 die Hilfe gerade dieses Arztes <strong>in</strong><br />

Anspruch nehmen zu müssen. (a 30(=sit) -A +d-a )<br />

24 Deshalb müßte dem Landarzt <strong>in</strong> solchen Fällen 19 e<strong>in</strong> Zeugnisverweigerungsrecht<br />

zugebilligt werden. (D +d+a -a 19(=sit) -N -d-a )<br />

25 ..., "ich habe nur gehört, daß es so etwas [e<strong>in</strong> Tempelchen] gegeben hat, und es lag auch 7<br />

bei me<strong>in</strong>en Eltern 27 e<strong>in</strong>e Zeichnung herum." (a 7(=pragm) -a 27(=sit) -N -d-a )<br />

25 Um den alten Kornmarkt zu verschönern, ließ der französische Reichstagsabgesandte<br />

im Jahr 1788 26 auf dem Platz 27 e<strong>in</strong>e Baumallee anlegen.<br />

(N +d+a -a 26(=sit-temp) -a 27(=sit-lok) -A -d-a )<br />

26 ...; kann man den Defekt e<strong>in</strong>es elektrischen Gerätes durch bloßes Berühren 25 feststellen?<br />

(A +d-a -a 25 )<br />

The next examples were quoted by Engel (1988: 326f):<br />

27 ..., daß Hans mich <strong>of</strong>fensichtlich 12 letzte Woche 26 auf diesen Umstand hat h<strong>in</strong>weisen<br />

wollen. (N +d+a -A pron -a 12 -a 26 -PO +d-a )<br />

28 Sie konnte sich eben 5 nicht mehr 41 dieses Vorfalls ents<strong>in</strong>nen. (A pron -a 5 -a 41 -G nom )<br />

29 Sie wollen dich immer wieder 36 <strong>in</strong> die Geschichte h<strong>in</strong>e<strong>in</strong>ziehen. (A pron -a 36 -PO +d-a )<br />

30 Er ist e<strong>in</strong>fach 14 allmählich 26 alt geworden. (a 14 -a 26 -Adj nom )<br />

31 Man könnte natürlich 12 diese D<strong>in</strong>ge viel schneller 43 erledigen. (a 12 -A +d-a -a 43 )<br />

32 Das würde gut 43 zu ihrem sonstigen Verhalten passen. (a 43 -PO +d-a )<br />

33 Hat er nicht obendre<strong>in</strong> noch Grundstücke gekauft? (N pron -a 5 -a 7 -a 34 -A -d-a )<br />

34 Sie hatte allerd<strong>in</strong>gs 6 sowieso 9 gekündigt. (a 6 -a 9 )<br />

35 Du kannst doch 4 wohl 12 nicht 41 e<strong>in</strong>fach 43 absagen. (a 4 -a 12 -a 41 -a 43 )<br />

36 ..., weil der Kollege mir das gestern Abend 26 erzählt hat. (N +d+a -D pron -A +d-a -a 26 )<br />

37 Hierzu bedarf er nicht 41 der Zustimmung se<strong>in</strong>es gesetzlichen Vertreters.<br />

(N pron -a neg -G nom )


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 145<br />

38 ... <strong>in</strong> die Lage kommen, für sich oder se<strong>in</strong>e Angehörigen 30 die Hilfe gerade dieses Arztes<br />

<strong>in</strong> Anspruch nehmen zu müssen. (a 30 -A +d-a -PO SVC )<br />

39 Sie konnte sich eben 5 nicht mehr 41 dieses Vorfalls ents<strong>in</strong>nen. (A pron -a 5 -a 41 -G +d-a )<br />

Most <strong>of</strong> the position classes behave quite straightforwardly <strong>in</strong> that there is no doubt <strong>with</strong><br />

respect to the unmarked order <strong>of</strong> elements. These <strong>in</strong>clude the pronom<strong>in</strong>al verb complements,<br />

dative and accusative objects, and all complements which have a strong tendency to be<br />

positioned to the right <strong>of</strong> the sentence, namely the situative and directive complements, the<br />

nom<strong>in</strong>al and adjectival complements, as well as the NPs and PPs <strong>of</strong> support verb<br />

constructions.<br />

However, there are some classes which have less clear syntactic behaviour, namely<br />

prepositional objects, <strong>in</strong>def<strong>in</strong>ite nom<strong>in</strong>ative complements, genitive objects and the<br />

negational modifier a 41 . We discussed the position <strong>of</strong> these problematic constituents <strong>in</strong><br />

Ste<strong>in</strong>berger (1990: 76-88).<br />

5.4. WHY DO SOME SENTENCES DIFFER FROM THE BASIC WORD<br />

ORDER<br />

In this section, we discuss a few sentences which do not comply <strong>with</strong> the canonical form.<br />

First, we shall have a look at the possible reasons for the order <strong>of</strong> some sentences to differ<br />

from the canonical form. Then, we want to list the effects <strong>of</strong> word order variation which are<br />

relevant for translation.<br />

40 ... alle Päckchen und Pakete aus dem Westen zu beschlagnahmen, die nicht haargenau den von<br />

Pankow erlassenen Bestimmungen entsprachen. (a 41 -a 43 -D +d-a )<br />

Sentence 40 is from Hoberg's corpus (1981: 153). It could easily be reversed <strong>in</strong>to the CF<br />

order "..., die den von Pankow erlassenen Bestimmungen nicht haargenau entsprachen." A<br />

reason why the order <strong>in</strong> 40 differs from that foreseen <strong>in</strong> the CF is the application <strong>of</strong> the<br />

Gesetz der wachsenden Glieder. The heav<strong>in</strong>ess pr<strong>in</strong>ciple expla<strong>in</strong>s, why the long dative NP<br />

follows the negational particle nicht and the manner adverb haargenau.<br />

41 Er hat immerh<strong>in</strong> 9 obendre<strong>in</strong> 7 bezahlt. (a 9 -a 7 )


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 146<br />

41 was quoted by Engel (1988: 326) <strong>in</strong> order to show that the position <strong>of</strong> some modifiers can<br />

change <strong>with</strong>out a change <strong>in</strong> the mean<strong>in</strong>g: Er hat obendre<strong>in</strong> 7 immerh<strong>in</strong> 9 bezahlt.<br />

42a Er hat e<strong>in</strong>em E<strong>in</strong>brecher das Messer entgegengesetzt. (D -d+a - A +d-a )<br />

42b ? Er hat das Messer e<strong>in</strong>em E<strong>in</strong>brecher entgegengesetzt.<br />

43a Er hat e<strong>in</strong>em Mann das Buch aufgenötigt. (D -d+a - A +d-a )<br />

43b ? Er hat das Buch e<strong>in</strong>em Mann aufgenötigt.<br />

42 and 43 are sentences which were part <strong>of</strong> the questionaire. 23 <strong>of</strong> the 38 participants <strong>in</strong> the<br />

case <strong>of</strong> 42, and 21 participants <strong>in</strong> the case <strong>of</strong> 43 decided that they preferred the order D


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 147<br />

We shall first list the mean<strong>in</strong>gful variations, and then discuss some mean<strong>in</strong>gless ones.<br />

Deviation from the canonical form can fulfil three ma<strong>in</strong> functions: focus<strong>in</strong>g (rhematization)<br />

<strong>of</strong> some elements, change <strong>of</strong> theme-rheme structure, and the change <strong>of</strong> scope <strong>of</strong> scope<strong>in</strong>clud<strong>in</strong>g<br />

elements. It is obvious that focus<strong>in</strong>g on one phrase, such as <strong>in</strong> 44, should be<br />

rendered <strong>in</strong> translation. It is not important by which means the target language expresses<br />

focus<strong>in</strong>g (cf. 3.3.2).<br />

The theme-rheme structure seems to be slightly less important than focus<strong>in</strong>g. Accord<strong>in</strong>g to<br />

the theme-rheme structure, elements referr<strong>in</strong>g to known <strong>in</strong>formation tend to precede the<br />

others. <strong>German</strong> sentences are concatenated through this procedure. Languages differ <strong>in</strong> their<br />

requirement for concatenation:<br />

45a Als kle<strong>in</strong>er Junge fiel Obelix <strong>in</strong> den Topf mit dem Zaubertrank. Deshalb darf er ihn jetzt nicht<br />

mehr tr<strong>in</strong>ken.<br />

45b Étant enfant, Obélix est tombé dans la marmite de potion magique. C'est pourquoi il ne peut<br />

plus en boire ma<strong>in</strong>tenant.<br />

45c When Obelix was a little boy, he fell <strong>in</strong>to the pot <strong>with</strong> the magic potion. That's why he must<br />

not dr<strong>in</strong>k it any more.<br />

46a Als K<strong>in</strong>d fuhr ich <strong>of</strong>t zu me<strong>in</strong>er Großmutter. In ihrem Garten pflanzte ich Tulpen.<br />

46b Lorsque j'étais enfant, je me rendais souvent chez ma grand-mère. Je plantais des tulipes<br />

dans son jard<strong>in</strong>.<br />

46c When I was a child, I <strong>of</strong>ten went to our grandmother's. I planted tulips <strong>in</strong> her garden.<br />

The adverb deshalb <strong>in</strong> 45a is a thematic element which makes the l<strong>in</strong>k between both<br />

sentences. Therefore, its place is <strong>in</strong> the Vorfeld where it precedes the rest <strong>of</strong> the sentence.<br />

This concatenation is rendered <strong>in</strong> the French and English translations. The render<strong>in</strong>g <strong>of</strong> the<br />

<strong>German</strong> structure <strong>in</strong> 46, however, is possible but less natural. In ihrem Garten l<strong>in</strong>ks the<br />

<strong>German</strong> sentences together but its equivalents are likely to stand at the end <strong>of</strong> the French and<br />

English sentences. In 46, <strong>German</strong> differs from French and English <strong>with</strong> respect to the<br />

render<strong>in</strong>g <strong>of</strong> theme-rheme structure. Theme and rheme should be expressed <strong>in</strong> the other<br />

language if it has means to express them. Therefore, these categories have to be recognised<br />

and featurised dur<strong>in</strong>g analysis.<br />

A further change <strong>of</strong> mean<strong>in</strong>g is achieved by mov<strong>in</strong>g scope-<strong>in</strong>clud<strong>in</strong>g elements. In <strong>German</strong>, it<br />

is possible to move degree-modifiers and negational particles to the left (and more rarely to


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 148<br />

the right) <strong>of</strong> an element which one wants to be scope-<strong>in</strong>cluded. Other languages do not allow<br />

this (47c, 48a), or are more restrictive than <strong>German</strong> (48a, 48b). In these languages the<br />

element which should be <strong>in</strong>cluded should probably be clefted so that only this and no other<br />

elements are <strong>in</strong>cluded. In French, seulement and ne que cannot be moved freely <strong>with</strong><strong>in</strong> the<br />

sentence so that dislocation (47d) or cleft<strong>in</strong>g (47e) are necessary. <strong>German</strong> and English<br />

behave the same for nur/only (47b) but differ for example <strong>with</strong> respect to negation.<br />

Sondernegation (partial negation) is possible <strong>in</strong> <strong>German</strong> but very restricted <strong>in</strong> English:<br />

47a Nur ich habe diesem Mann e<strong>in</strong> Buch gegeben.<br />

47b Only I gave a book to this man.<br />

47c * Seulement je/moi ai donné un livre à ce Monsieur.<br />

47d Je suis le seul à avoir donné un livre à ce Monsieur.<br />

47e Il n'y a que moi qui ai donné un livre à ce Monsieur.<br />

48a * Not I gave a book to this man.<br />

48b Not I but my brother gave a book to this man.<br />

In addition to the volitional divergences already mentioned, there are further reasons for<br />

deviations which do not imply a change <strong>of</strong> mean<strong>in</strong>g, and therefore do not have to be<br />

rendered <strong>in</strong> translation. An example <strong>of</strong> the latter is the clarification <strong>of</strong> complex syntactic<br />

structures, which we mentioned <strong>in</strong> 3.2 and 3.11. Extraposition <strong>of</strong> heavy constituents has its<br />

reason <strong>in</strong> better processability. Besides this, extraposition <strong>of</strong> constituents does not have any<br />

impact:<br />

49a ? Ich habe auch DEM Mann, dem wir letztes Jahr <strong>in</strong> Spanien <strong>in</strong> den Gärten der Alhambra<br />

<strong>in</strong> Grenada unseren Reiseführer geschenkt haben, die E<strong>in</strong>ladung zu unserer Party<br />

geschickt.<br />

49b<br />

Ich habe unsere E<strong>in</strong>ladung auch DEM Mann geschickt, dem wir letztes Jahr <strong>in</strong> Spanien<br />

<strong>in</strong> den Gärten der Alhambra <strong>in</strong> Granada unseren Reiseführer geschenkt haben.<br />

49a is marked compared to 49b, as the ma<strong>in</strong> verb as well as the accusative object follow very<br />

late. In 49b, the relative clause is extraposed and, <strong>in</strong> contradiction to the canonical form, the<br />

animate dative NP follows the <strong>in</strong>animate accusative NP <strong>in</strong> order to keep the distance from<br />

the relative clause smaller. As rhematic NPs and PPs tend to be longer than thematic ones<br />

and tend to the right anyway, it is quite unlikely that the clarification <strong>of</strong> complex syntactic<br />

structures leads to a divergence from the canonical form. In 49, both NPs are def<strong>in</strong>ite but it<br />

seems very likely that most parts <strong>of</strong> the complex dative NP are rhematic. The clarification <strong>of</strong>


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 149<br />

complex syntactic structures as a reason for the divergence from the canonical form does not<br />

<strong>in</strong>volve any change <strong>of</strong> mean<strong>in</strong>g, and therefore must not necessarily be rendered <strong>in</strong> a target<br />

language when translat<strong>in</strong>g.<br />

A further mean<strong>in</strong>gless process is the addition <strong>of</strong> <strong>in</strong>formation <strong>in</strong> the Nachfeld (Nachtrag). The<br />

reason for Nachtrag is that speakers forgot to mention someth<strong>in</strong>g before, or that they<br />

suddenly decide that the hearer/reader needs some more <strong>in</strong>formation (Engel, 1988: 333):<br />

Häufig kommt es vor, daß die Struktur von Sätzen sich erst während des Sprechens bildet und daß der<br />

Sprecher bei den ersten Worten das Gesamtgerüst e<strong>in</strong>er Äußerung noch nicht völlig durchschaut: Er<br />

redet, und während des Redens wird ihm bewußt, daß er e<strong>in</strong> Element vergessen hat, das nach den<br />

geltenden Regeln schon zuvor hätte ersche<strong>in</strong>en müssen. In solchen Fällen wird das Element <strong>in</strong>s<br />

Nachfeld gesetzt; es ist dann meistens unbetont und hat <strong>of</strong>fensichtlich die Funktion e<strong>in</strong>es Nachtrags.<br />

Allerd<strong>in</strong>gs ist die Nachfeldstellung von Ergänzungen nur sehr begrenzt möglich. Vor allem kann<br />

praktisch nur die Präpositivergänzung im Nachfeld auftauchen.<br />

Nachtrag is uncommon <strong>in</strong> written language because the slower speed <strong>of</strong> the writ<strong>in</strong>g process<br />

allows authors to plan their utterances better (Engel, 1988: 316ff). In Early New High-<br />

<strong>German</strong> (1500-1650), the position <strong>of</strong> all verb arguments <strong>in</strong> the Nachfeld was possible. When<br />

such word order<strong>in</strong>g is used nowadays, it has a strongly archais<strong>in</strong>g function (Engel, 1988:<br />

316). Nachtrag <strong>of</strong> modifiers <strong>in</strong> spoken language does not have an archais<strong>in</strong>g effect.<br />

We want to po<strong>in</strong>t out that other languages may behave the same as <strong>German</strong> <strong>with</strong> respect to<br />

the clarification <strong>of</strong> complex syntactic structures and to Nachtrag. One may thus want to<br />

render these constructions when translat<strong>in</strong>g <strong>in</strong>to these languages. However, we do not see<br />

any need for a featurisation <strong>of</strong> these as they do not contribute to the mean<strong>in</strong>g <strong>of</strong> the sentence.<br />

The same applies to all the other preferences listed <strong>in</strong> chapter 3, and which we have not<br />

mentioned <strong>in</strong> this section. Theme, rheme, focus and scope, on the other hand, convey more<br />

<strong>of</strong> a mean<strong>in</strong>g, <strong>in</strong> the sense that they are important for the use <strong>of</strong> the sentence <strong>with</strong><strong>in</strong> a certa<strong>in</strong><br />

context.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 150<br />

5.5. THE VORFELD POSITION<br />

So far, we have restricted our statements on word order<strong>in</strong>g pr<strong>in</strong>ciples, as well as on<br />

canonical form, to the <strong>German</strong> Mittelfeld. The Mittelfeld is the part <strong>of</strong> the sentence which is<br />

limited by the verb <strong>in</strong> verb-second position to the left, and by the participle or verb prefix to<br />

the right. For this reason, we want to discuss the Vorfeld now.<br />

The Vorfeld is the part <strong>of</strong> the sentence which precedes the verb <strong>in</strong> verb-second position. In<br />

sentences other than verb-second (i.e. verb-<strong>in</strong>itial and verb-f<strong>in</strong>al) it does not make sense to<br />

speak <strong>of</strong> a Vorfeld. One s<strong>in</strong>gle phrase <strong>of</strong> the constituents which occur <strong>in</strong> the Mittelfeld <strong>of</strong><br />

verb-f<strong>in</strong>al clauses, can be chosen and put <strong>in</strong>to the Vorfeld. The most likely phrase to move to<br />

the Vorfeld is the nom<strong>in</strong>ative complement, but most other elements can be <strong>in</strong> this position as<br />

well. The order <strong>of</strong> the rema<strong>in</strong><strong>in</strong>g constituents <strong>in</strong> the Mittelfeld does not change.<br />

The X-bar syntax model comb<strong>in</strong>ed <strong>with</strong> the transformational approach, as described <strong>in</strong><br />

Stechow/Sternefeld (1988) and Fanselow/Felix (1987), for example, describes this<br />

mechanism very conveniently. We can assume that the order <strong>of</strong> elements <strong>in</strong> the deep<br />

structure is similar to the one found <strong>in</strong> <strong>German</strong> subord<strong>in</strong>ate clauses. In declarative sentences,<br />

the verb is moved by move-alpha from its f<strong>in</strong>al position to verb-second. Any other XP<br />

(maximal projection, such as NP, PP, AP etc) is shifted by the same transformation to the<br />

Vorfeld position. In terms <strong>of</strong> X-bar syntax, the latter is the structural position Spec-Comp<br />

(complement phrase specifier position). The other elements are not affected by the<br />

movement and their order does not change.<br />

The order <strong>of</strong> the elements <strong>in</strong> the Mittelfeld is decided by the word-order<strong>in</strong>g pr<strong>in</strong>ciples. Once<br />

these have applied, one <strong>of</strong> the elements is moved to the Vorfeld. It seems reasonable to<br />

assume nevertheless a place (a trace) for the moved elements <strong>in</strong> the Mittelfeld. Gestern <strong>in</strong><br />

50a, for example, belongs between the pronom<strong>in</strong>al NP ihr and the <strong>in</strong>def<strong>in</strong>ite NP e<strong>in</strong>en<br />

Bären, and not <strong>in</strong> a different position such as <strong>in</strong> 50c:<br />

50a Gestern hat er ihr __ e<strong>in</strong>en Bären aufgebunden.<br />

50b Ich sagte, daß er ihr gestern e<strong>in</strong>en Bären aufgebunden hat.<br />

50c * Ich sagte, daß er ihr e<strong>in</strong>en Bären gestern aufgebunden hat.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 151<br />

If we use this model, which assumes a certa<strong>in</strong> order <strong>with</strong> one element be<strong>in</strong>g moved <strong>in</strong>to the<br />

Vorfeld, it is sufficient to describe the order <strong>in</strong> the Mittelfeld and then to provide means to<br />

choose the element which moves to the sentence-<strong>in</strong>itial position. We have seen that a ma<strong>in</strong><br />

criterion <strong>in</strong> choos<strong>in</strong>g the moved element is the theme-rheme structure, which requires that a<br />

thematic element makes the l<strong>in</strong>k to one <strong>of</strong> the last sentences (concatenation). Waltz<strong>in</strong>g<br />

(1986: 144) furthermore mentions that <strong>in</strong> sentences <strong>with</strong> too many existimatorial modifiers<br />

(51a), one <strong>of</strong> them <strong>of</strong>ten moves to the Vorfeld <strong>in</strong> order to avoid an overloaded Mittelfeld<br />

(51b). This overflow<strong>in</strong>g probably does not happen very <strong>of</strong>ten, as it seems that the Mittelfeld<br />

is quite robust <strong>in</strong> this respect (examples accord<strong>in</strong>g to Waltz<strong>in</strong>g):<br />

51a ? Er hat damals ja natürlich außerdem dankenswerterweise se<strong>in</strong>en Vertreter geschickt.<br />

51b Außerdem hat er damals ja natürlich dankenswerterweise se<strong>in</strong>en Vertreter geschickt.<br />

We assume that whatever element is marked as be<strong>in</strong>g thematic moves to the Vorfeld<br />

position. This is a very simple mechanism and does not need further discussion (for an<br />

implementation <strong>in</strong> a synthesis grammar see Ste<strong>in</strong>berger, 1992a: 38ff). However, there are<br />

some more details to mention concern<strong>in</strong>g how we can make use <strong>of</strong> the particular status <strong>of</strong> the<br />

Vorfeld.<br />

In a few cases, the position <strong>of</strong> an element can help us to disambiguate homonyms as not all<br />

elements can fill the Vorfeld position. For <strong>in</strong>stance, if an element A is <strong>in</strong> the Vorfeld, and its<br />

homonym A' generally cannot occur there, we can deduce that the word <strong>in</strong> question is A, and<br />

not A'. The lexeme eben for <strong>in</strong>stance can be either a toner, a temporal adverb or an adjective.<br />

In 52, it is nearly certa<strong>in</strong> that eben is the temporal adverb as toners cannot stand <strong>in</strong> the<br />

Vorfeld, and we can almost exclude the possibility that it is an adjective, as the verb se<strong>in</strong><br />

only calls for either two NPs, or for one NP and an AP. However, we have to consider the<br />

possibility that eben is a sentence adjective. This does not make any sense <strong>in</strong> 52, but the<br />

occurrence <strong>of</strong> adjectives as sentence modifiers <strong>in</strong> the Vorfeld cannot generally be excluded,<br />

as shown <strong>in</strong> 53:<br />

52 Eben war Ralf noch ke<strong>in</strong> Herr Doktor.<br />

53 Behutsam schrieb Ralf Kapitel für Kapitel.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 152<br />

The Vorfeld can also help <strong>with</strong> structural ambiguity as one can assume that it is filled by one<br />

phrase only. This means that the occurrence <strong>of</strong> two phrases <strong>in</strong> the Vorfeld means that one <strong>of</strong><br />

them is likely to be an adjunct to the other (54). This is not an absolute rule, though, as there<br />

are a few exceptions. In 55 this analysis is wrong 70 :<br />

54 Der Prüfer von Ralfs Doktorarbeit erkennt sich wieder.<br />

55 Im Stadion auf der Tribüne steht me<strong>in</strong> Doktorvater.<br />

A further regularity l<strong>in</strong>ked to the Vorfeld, which can be used to improve analysis, is that<br />

some elements have to be stressed when they are <strong>in</strong> the Vorfeld (56). This fact is not only<br />

important for the recognition <strong>of</strong> focus (cf. 6.2), but also for the scope identification <strong>of</strong> degree<br />

modifiers and negation <strong>in</strong> the Mittelfeld. We shall see later (cf. 5.6) that nicht <strong>in</strong> sentences<br />

such as 56 has to refer to the focused Vorfeld element:<br />

56 Dem MANN gab er das Buch nicht.<br />

For the analysis <strong>of</strong> <strong>German</strong> sentences we can simply assume that if the Vorfeld element is<br />

not focused, it is thematic. This makes it easy to identify the theme. The fill<strong>in</strong>g <strong>of</strong> the<br />

Vorfeld by a rhematic element is quite rare <strong>in</strong> written language but exists. The most frequent<br />

rhematic elements to occur there are modal modifiers (example from Hoberg, 1981: 173):<br />

57 Ohne Pilot startete <strong>in</strong> Frederikshaven (Dänemark) e<strong>in</strong> Flugzeug.<br />

However, this k<strong>in</strong>d <strong>of</strong> sentence is likely to occur at the beg<strong>in</strong>n<strong>in</strong>g <strong>of</strong> texts as it does not leave<br />

any space for contextual embedd<strong>in</strong>g. Accord<strong>in</strong>g to Hoberg (1981: 172), this rare structure<br />

ma<strong>in</strong>ly occurs <strong>in</strong> emotional language such as headl<strong>in</strong>es <strong>of</strong> the popular press and spoken<br />

language (Hoberg, 1981: 173):<br />

Die Besetzung des Vorfelds mit dem Rhema, d.h. also das Aufrollen der Äußerung vom<br />

Mitteilungskern aus, läßt <strong>in</strong> der Regel ke<strong>in</strong>en Raum für die kontextualle E<strong>in</strong>bettung der Äußerungsbasis,<br />

obwohl sich Rhema- und Satzverknüpfungsfunktion grundsätzlich nicht ausschließen müssen.<br />

For the generation <strong>of</strong> <strong>German</strong> sentences it is <strong>in</strong>terest<strong>in</strong>g to look at statistical data. When a<br />

theme can be identified dur<strong>in</strong>g analysis <strong>of</strong> the source language sentence, there is no need to<br />

70 See Altmann (1981: 285ff) on the double fill<strong>in</strong>g <strong>of</strong> the Vorfeld position.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 153<br />

use statistics, but if not, we can choose the element which statistically is most likely to fill<br />

the Vorfeld. In both columns <strong>of</strong> the follow<strong>in</strong>g table, only sentences hav<strong>in</strong>g a Vorfeld are<br />

considered (V2-sentences). The first column shows how <strong>of</strong>ten different elements occur <strong>in</strong> the<br />

Vorfeld <strong>of</strong> all V2-sentences. We see that the nom<strong>in</strong>ative complement is clearly ahead <strong>of</strong> all<br />

other constituents. The second column displays how <strong>of</strong>ten these elements occur as a<br />

percentage <strong>of</strong> all V2-sentences which <strong>in</strong>volve these elements (Hoberg, 1981: 162):<br />

Vorfeld Element <strong>in</strong> Percentage <strong>of</strong> all Percentage <strong>in</strong><br />

verb-second sentences elements occurr<strong>in</strong>g <strong>in</strong> sentences only<br />

the Vorfeld<br />

<strong>in</strong> which they occur<br />

Nom<strong>in</strong>ative complement 63.15 % 55.60 %<br />

Other complements 10.15 % 9.27 %<br />

Modifiers 26.06 % 28.32 %<br />

among these:<br />

a pragm 5.38 % 38.37 %<br />

a sit 17.74 % 28.13 %<br />

a mod 1.93 % 16.98 %<br />

For our purpose, the second column is more <strong>in</strong>terest<strong>in</strong>g. The data shows that <strong>in</strong> over 55% <strong>of</strong><br />

the cases, the nom<strong>in</strong>ative complement fills the Vorfeld. If pragmatic modifiers are present<br />

these are the second most likely to occur there (over 38%). As toners are pragmatic<br />

modifiers which cannot precede the verb <strong>in</strong> verb-second position, the percentage <strong>of</strong> the<br />

rema<strong>in</strong><strong>in</strong>g pragmatic modifiers should even be higher. As one may have expected, the<br />

default candidate for the Vorfeld position is the nom<strong>in</strong>ative complement. It is closely<br />

followed by the subgroup <strong>of</strong> existimatorial modifiers which can fill the Vorfeld. If no theme<br />

can be identified dur<strong>in</strong>g analysis, the nom<strong>in</strong>ative complement should precede the verb <strong>in</strong><br />

second position.<br />

Hoberg deduces that the further left elements normally occur <strong>in</strong> the Mittelfeld, the more<br />

likely they are to go <strong>in</strong>to the Vorfeld if there is one (Hoberg, 1981: 164):<br />

[...] wenn man sich <strong>in</strong> e<strong>in</strong>em Satzerzeugungsprozeß die Anordnung der Elemente im Mittelfeld als<br />

Basisoperation und die Besetzung des Vorfelds als Permutation e<strong>in</strong>es dieser Mittelfeldelemente an den<br />

Satzanfang vorstellt: es wird aus der Menge der Mittelfeldelemente am häufigsten dasjenige Element als<br />

Vorfeldelement ausgewählt, das am weitesten l<strong>in</strong>ks steht und entsprechend am wenigsten häufig<br />

dasjenige, das am weitesten rechts steht.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 154<br />

Accord<strong>in</strong>g to Hoberg (1981: 156ff), the particle groups a 2 (da, dann, nun, jetzt (not<br />

temporal)) and the speech act-structur<strong>in</strong>g adverbials a 6 , a 7 and a 8 (jedoch, <strong>in</strong>dessen,<br />

allerd<strong>in</strong>gs, ..., erstens, ferner, ..., z.B., <strong>in</strong> erster L<strong>in</strong>ie, ...) are the modifiers which occur <strong>in</strong> the<br />

Vorfeld most frequently. She assumes that thematicity and the concatenat<strong>in</strong>g function are not<br />

the only pr<strong>in</strong>ciples responsible for the choice <strong>of</strong> the element which moves <strong>in</strong>to the Vorfeld.<br />

Further <strong>in</strong>vestigation should be carried out on this subject, probably <strong>in</strong>volv<strong>in</strong>g large corpora<br />

for reliable statistical <strong>in</strong>formation. We consider that we have discussed the Vorfeld enough to<br />

ensure reasonable translational results and will therefore go back to the treatment <strong>of</strong> word<br />

order <strong>in</strong> the Mittelfeld.<br />

5.6. THE IMPORTANCE OF THEME, RHEME AND FOCUS<br />

The <strong>in</strong>tention <strong>of</strong> this section is to show the importance <strong>of</strong> the categories theme, rheme and<br />

focus. Their ma<strong>in</strong> relevance to this dissertation concerns the concatenation <strong>of</strong> sentences,<br />

which is necessary to guarantee a natural flow <strong>with</strong><strong>in</strong> the text. Furthermore, the categories<br />

play an important role when treat<strong>in</strong>g negation and quantifier scope. We shall discuss the<br />

latter only briefly.<br />

Scope as an essential part <strong>of</strong> the propositional content <strong>of</strong> a sentence should always be<br />

rendered correctly. It is mostly <strong>in</strong>dependent from theme-rheme structure and functional<br />

sentence perspective, and therefore overrules any order suggested by these categories (cf.<br />

4.4.3). However, scope and focus are not completely <strong>in</strong>dependent, at least as far as negation<br />

is concerned (Payne, 1985: 232):<br />

One well-known feature <strong>of</strong> standard negation is its tendency to associate <strong>with</strong> the focused elements <strong>in</strong> a<br />

sentence.<br />

Accord<strong>in</strong>g to Altmann (1976: 30f) and Rochemont (1989: 6), simple sentences (<strong>in</strong>volv<strong>in</strong>g<br />

one ma<strong>in</strong> verb) should have only one focus. It is possible that a simple sentence has two or<br />

more focuses, but this is rare and stylistically marked. In 58 (found <strong>in</strong> Engel, 1988: 764), for<br />

<strong>in</strong>stance, the author focuses two different phrases by us<strong>in</strong>g the degree modifiers auch and


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 155<br />

vor allem. The focused phrases are mit ihrer Hilfe and se<strong>in</strong>e E<strong>in</strong>stellung zu e<strong>in</strong>em<br />

Sachverhalt:<br />

58 So kann der Sprecher auch mit ihrer Hilfe [der Gradpartikel, RS] vor allem se<strong>in</strong>e<br />

E<strong>in</strong>stellung zu e<strong>in</strong>em Sachverhalt ausdrücken.<br />

It is thus very likely that the focus <strong>of</strong> degree adverbs, as well as the negational particle nicht,<br />

co<strong>in</strong>cides <strong>with</strong> the sentence focus. In the case <strong>of</strong> sentence negation, nicht can generally focus<br />

on all s<strong>in</strong>gle elements <strong>of</strong> the sentence (59a-59e) 71 :<br />

59a Die Schwester von Wolf hat dem Lehrer das Buch nicht geliehen (sondern Wolf<br />

SELBST/die Schwester von JOHANNES).<br />

59b Die Schwester von Wolf hat dem Lehrer das Buch nicht geliehen (sondern WIRD es<br />

ihm leihen).<br />

59c Die Schwester von Wolf hat dem Lehrer das Buch nicht geliehen (sondern e<strong>in</strong>em<br />

ANDEREN Lehrer/dem SCHÜLER).<br />

59d ? Die Schwester von Wolf hat dem Lehrer das Buch nicht geliehen (sondern e<strong>in</strong><br />

ANDERES Buch/die ZEITSCHRIFT).<br />

59e Die Schwester von Wolf hat dem Lehrer das Buch nicht geliehen (sondern<br />

GESCHENKT).<br />

When any <strong>of</strong> these elements is focused by other means, the negation nicht is likely to focus<br />

on the same element (60a). Although this is not obligatory, it is very likely, as the unnatural<br />

sentence (60b) suggests. Therefore the recognition <strong>of</strong> focused elements may help to identify<br />

the scope <strong>of</strong> the negational element nicht. For a discussion <strong>of</strong> the relationship between<br />

quantifier scope and focus, see Ste<strong>in</strong>er/W<strong>in</strong>ter (1987: 3ff).<br />

60a Nur WOLF hat dem Lehrer das Buch nicht geliehen.<br />

60b ?? Nur Wolf hat dem Lehrer das Buch nicht geLIEhen.<br />

What we shall rather focus on <strong>in</strong> this thesis is the correct placement <strong>of</strong> thematic, rhematic<br />

and focused elements <strong>in</strong> order to guarantee natural sentence concatenation. We shall<br />

<strong>in</strong>tuitively dist<strong>in</strong>guish rheme and focus. A focused constituent will be much more strongly<br />

stressed than other rhematic elements. In 61a, ihn carries a heavy (contrastive) stress,<br />

whereas its full form equivalent den Mann <strong>in</strong> 61b is likely to be rhematic. It is not as strongly<br />

stressed as ihn <strong>in</strong> 61a. The accusative pronoun <strong>in</strong> 1a is thus a focus, whereas we would say that<br />

71 A m<strong>in</strong>or restriction to this statement is brought forward by Höhle (1982: 426). He claims that, supposedly<br />

for pragmatic reasons, sentence negation never <strong>in</strong>cludes all constituents <strong>of</strong> the sentence.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 156<br />

the full form <strong>in</strong> 61b is likely to be a rheme. The difference between 61b and 61c is that, <strong>in</strong> the<br />

former, the accusative NP is likely to be rhematic, whereas <strong>in</strong> the latter it is not marked as either<br />

thematic, rhematic or focused. In 61c, gestern is likely to be rhematic. In the model we suggest,<br />

gestern <strong>in</strong> 61a and 61b, and den Mann <strong>in</strong> 61c are neutral <strong>with</strong> respect to the three categories:<br />

61a Ich habe gestern IHN gesehen.<br />

61b Ich habe gestern den MANN gesehen.<br />

61c Ich habe den Mann GEStern gesehen.<br />

It is a well-known fact that context can require focus<strong>in</strong>g on a particular constituent <strong>in</strong> a<br />

sentence. M<strong>in</strong>imal contexts requir<strong>in</strong>g focus<strong>in</strong>g are wh-questions, because they make it clear<br />

which elements are thematic and which ones are not. Question 62 asks for the PP mit dem<br />

Nudelholz, whereas 63 asks for the accusative complement den Mann. In our <strong>in</strong>tuition, the<br />

best order <strong>of</strong> constituents <strong>in</strong> 62 is the order accusative NP before PP (62a), whereas the best<br />

answer to question 63 has the order PP before accusative NP (63a). Answer<strong>in</strong>g question 62<br />

<strong>with</strong> the phrase order <strong>of</strong> 63a, however, is very unnatural (62c):<br />

62 Mit was erwartete die Frau ihren Mann gestern?<br />

62a Die Frau erwartete ihren Mann mit dem NUdelholz.<br />

62b ? Mit dem NUdelholz erwartete die Frau ihren Man.<br />

62c ?? Die Frau erwartete mit dem NUdelholz ihren Mann.<br />

62d * Die Frau erwartete mit dem Nudelholz ihren MANN.<br />

63 Wen erwartete die Frau gestern mit dem Nudelholz?<br />

63a Die Frau erwartete mit dem Nudelholz ihren MANN.<br />

63b ? Die Frau erwartete ihren MANN mit dem Nudelholz.<br />

63c ?? Mit dem Nudelholz erwartete die Frau ihren MANN.<br />

In the contexts 62 and 63, the phrase orders <strong>of</strong> the answers should not be <strong>in</strong>terchanged.<br />

Many other contexts probably do not allow the recognition <strong>of</strong> theme and rheme so clearly, so<br />

that the differences <strong>in</strong> acceptability judgement should be less pronounced. We believe,<br />

however, that even <strong>in</strong> these cases, proper sentence concatenation produces a more natural<br />

flow <strong>of</strong> the text. In order to achieve sentences <strong>with</strong> vary<strong>in</strong>g, and appropriate, word order, we<br />

suggest to identify theme, rheme and focus, and to assign them places <strong>in</strong> the canonical form.<br />

In chapter 6, we shall show how to recognise the categories dur<strong>in</strong>g analysis, and how to<br />

<strong>in</strong>corporate them <strong>in</strong> the canonical form to generate natural sentences.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 157<br />

As we deal <strong>with</strong> written language only, a focus will be recognised as a focus only when it is<br />

clearly marked by syntactical means, as is the case <strong>in</strong> 61a. We assume that the occurrence <strong>of</strong><br />

focused phrases is very restricted <strong>in</strong> written language. The accusative NP den Mann <strong>in</strong> 61b<br />

could be as strongly stressed as ihn <strong>in</strong> 61a, but <strong>with</strong>out a context it is impossible to f<strong>in</strong>d out<br />

whether it is or not. Therefore, we shall treat it as not focused. Rhemes are less easy to<br />

recognise and also play a less important role when decid<strong>in</strong>g on word order. This will become<br />

clearer <strong>in</strong> the sections to come.<br />

We have to dist<strong>in</strong>guish the generation <strong>of</strong> <strong>German</strong> sentences from their analysis. We start<br />

<strong>with</strong> the former. We suggest that a neutral situative modifier, for example, will be positioned<br />

at a different place than the same modifier when it is thematic or focused. When thematic, it<br />

should be placed earlier (more to the left), when rhematic or focused, it should be placed<br />

later. Some focused elements could be stressed by us<strong>in</strong>g a cleft<strong>in</strong>g construction.<br />

We shall show a few simplified examples for the distribution <strong>of</strong> phrases which are thematic,<br />

rhematic and focused. The follow<strong>in</strong>g sentences would be generated on the basis <strong>of</strong> the<br />

canonical form and the three categories:<br />

A) No theme, rheme or focus <strong>in</strong>formation is available; the word order is accord<strong>in</strong>g to<br />

the canonical form. The same sentence would be produced <strong>with</strong> ich carry<strong>in</strong>g the theme<br />

feature, and/or den Mann carry<strong>in</strong>g the rheme feature:<br />

64a Ich habe gestern den Mann gesehen.<br />

B) If gestern is marked as rhematic, and the other elements are not marked at all or ich<br />

is marked as thematic, 64b would be generated:<br />

64b Ich habe den Mann gestern gesehen.<br />

C) c is the sentence generated if the theme is gestern, and the other elements are either<br />

not marked at all or Mann is rhematic:<br />

64c Gestern habe ich den Mann gesehen.<br />

D) Ich is focused and gestern is thematic:<br />

64d Gestern habe den Mann ICH gesehen.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 158<br />

It is easy to imag<strong>in</strong>e more variations <strong>of</strong> feature comb<strong>in</strong>ations.<br />

The ma<strong>in</strong> idea is that <strong>with</strong>out any <strong>in</strong>formation available on the theme-rheme structure and<br />

the functional sentence perspective, sentences are generated accord<strong>in</strong>g to the canonical form.<br />

If an element has been recognised as theme, and it is a potential Vorfeld element, it starts <strong>of</strong>f<br />

the sentence. If an element is rhematic, its place is towards the end <strong>of</strong> the sentence (<strong>in</strong> the<br />

examples a to d it precedes the verb <strong>in</strong> f<strong>in</strong>al position). If a phrase has been recognised as<br />

be<strong>in</strong>g focused, it takes special positions which are to be described <strong>in</strong> 6.5.3. The different<br />

parametrisations can also be comb<strong>in</strong>ed, e.g. the theme precedes the verb <strong>in</strong> second position<br />

and the rheme position is late <strong>in</strong> the sentence etc.<br />

Before def<strong>in</strong><strong>in</strong>g the places <strong>of</strong> the three categories <strong>in</strong> the new canonical form (cf. 6.5.3), we<br />

shall describe how to recognise theme, rheme and focus automatically <strong>in</strong> a Mach<strong>in</strong>e<br />

Translation system. Such a procedure is presented <strong>in</strong> Ste<strong>in</strong>berger (1992a: 20ff) for <strong>German</strong><br />

and English. Its application to the Mach<strong>in</strong>e Translation formalism CAT2 <strong>in</strong> a slightly<br />

simplified version is described <strong>in</strong> Ste<strong>in</strong>berger (1992a: 33ff). In addition to list<strong>in</strong>g means to<br />

recognise these three categories, we shall also list elements <strong>with</strong> compulsory order, as these<br />

facts are particularly important for automatic treatment.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 159<br />

6. AIDS FOR COMPUTATIONAL ANALYSIS<br />

[...] no implemented NL system has been demonstrated so far that handles partially<br />

free word order <strong>of</strong> <strong>German</strong> and many other languages <strong>in</strong> a satisfactory way<br />

(Engelkamp/Erbach/Uszkoreit, 1992: 201)<br />

Chapter 6 is dedicated to the computational treatment <strong>of</strong> <strong>German</strong> free word order <strong>in</strong> Mach<strong>in</strong>e<br />

Translation. In 6.1, we list compulsory orders, which are particularly important <strong>in</strong> Natural<br />

Language Process<strong>in</strong>g. Then, we concentrate on how to recognise focused, thematic and<br />

rhematic elements <strong>in</strong> a sentence (6.2 and 6.3). Section 5.6 made it clear that the recognition<br />

<strong>of</strong> theme, rheme and focus is needed for <strong>German</strong> synthesis. We therefore want to provide the<br />

same <strong>in</strong>formation for the generation <strong>of</strong> sentences <strong>in</strong> other languages, when <strong>German</strong> is the<br />

source language.<br />

In 6.4, we mention some details which are not absolutely mandatory for the computational<br />

treatment <strong>of</strong> word order variation, but which certa<strong>in</strong>ly give us a better understand<strong>in</strong>g <strong>of</strong> the<br />

problem. They concern the consequences <strong>of</strong> modifier sequence variations. In 6.5, we develop<br />

the f<strong>in</strong>al version <strong>of</strong> the canonical form which <strong>in</strong>cludes positions for theme, rheme and focus.<br />

For cases <strong>in</strong> which the grammar allows several analyses <strong>of</strong> one sentence, the formulation <strong>of</strong><br />

preference rules could help to choose the most likely one. We formulate such a preference <strong>in</strong><br />

6.6. Before summaris<strong>in</strong>g the results <strong>of</strong> chapter 6 <strong>in</strong> 6.8, we give a list <strong>of</strong> features which are<br />

both necessary and sufficient for the satisfy<strong>in</strong>g syntactic treatment <strong>of</strong> adverbs (6.7). These<br />

are the features used to encode the modifiers listed <strong>in</strong> the appendix (8.2 and 8.3).<br />

6.1. COMPULSORY ORDERS<br />

A major problem <strong>of</strong> automatic analysis is ambiguity. Too many analyses are created for most<br />

sentences, and only few means are available to choose among them. The parallel treatment<br />

<strong>of</strong> all <strong>of</strong> them is computationally expensive and should be avoided. A commercial Mach<strong>in</strong>e<br />

Translation system not only has to be fast but should also <strong>of</strong>fer a s<strong>in</strong>gle output sentence, as<br />

opposed to list<strong>in</strong>g a translation correspond<strong>in</strong>g to each possible ambiguity.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 160<br />

In some cases, our list <strong>of</strong> obligatory order <strong>of</strong> elements may help to rule out wrong analyses<br />

and to recognise lexically or structurally ambiguous elements. How these facts can be<br />

formulated <strong>with</strong><strong>in</strong> the grammar depends entirely on the formalism. In Eurotra (Bech, 1991,<br />

17f) as well as <strong>in</strong> CAT2 (Sharp, 1993: 16ff), obligatory sequences can be expressed as strict<br />

rules. Impossible order can be ruled out by killer rules.<br />

What we can do to recognise whether a sentence is grammatical is to list all strong<br />

restrictions found dur<strong>in</strong>g our research. One <strong>of</strong> these restrictions is for example, that<br />

existimatorial modifiers must not be <strong>in</strong> the scope <strong>of</strong> the negational particle nicht. This<br />

regularity can help to identify sicher <strong>in</strong> 1a as a pragmatic modifier, and <strong>in</strong> 1b as a manner<br />

adverb:<br />

1a<br />

1b<br />

Er spricht sicher nicht. (It is certa<strong>in</strong> that he does not speak)<br />

Er spricht nicht sicher.<br />

(He does not speak confidently / He is not confident when he speaks)<br />

We want to po<strong>in</strong>t out that the number <strong>of</strong> constituent sequences which are absolutely<br />

impossible is restricted. Our experience corresponds to Reis' (1987: 167) claim that, <strong>in</strong><br />

<strong>German</strong>, strong stress (and thus focalization) can overrule most word order variations which<br />

would be impossible <strong>with</strong> normal <strong>in</strong>tonation:<br />

überhaupt vermag Betonung grundsätzlich auch sonst relativ strenge Abfolgeregeln (z.B. DOPO to DO


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 161<br />

The same regularity expla<strong>in</strong>s why we can work <strong>with</strong> isolated (contextless) sentences: Some<br />

word order variations call for a certa<strong>in</strong> theme-rheme structure, or even for contrastive stress,<br />

so that the context <strong>in</strong> which these sentences can occur is restricted.<br />

As we have described the obligatory-order rules (Obligatorische Folge-, or OF-Regeln) <strong>in</strong><br />

detail <strong>in</strong> previous work, we shall only mention some <strong>of</strong> them briefly to specify the k<strong>in</strong>d <strong>of</strong><br />

rules we are th<strong>in</strong>k<strong>in</strong>g <strong>of</strong>. For a more explicit discussion and examples, see Ste<strong>in</strong>berger (1990:<br />

102-116), to which the OF-rules <strong>in</strong> this section refer.<br />

A) Rule OF1 says that NPs and PPs <strong>of</strong> support verb constructions (as described <strong>in</strong><br />

3.4.1.1) must follow all other complements and modifiers <strong>in</strong> the sentence.<br />

B) OF2 and OF5 state the fact that obligatory reflexive pronouns, as well as neutral<br />

accusative pronouns, can never be contrastively focused. Therefore, they must precede<br />

the negation nicht and all situative modifiers. In this they differ from other pronouns.<br />

C) In OF3, we mention that full form nom<strong>in</strong>ative complements can follow either<br />

accusative or dative full form complements, but not both.<br />

D) Another regularity is that an element A referr<strong>in</strong>g to an element B by a possessive<br />

pronoun must follow the referred-to element B (OF7). The order B


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 162<br />

I) Modal modifiers, <strong>with</strong> the exception <strong>of</strong> gern (2), cannot precede personal pronouns<br />

(3). Gern has a special status <strong>with</strong><strong>in</strong> the modal adverbs (cf. section 2.3). This is<br />

confirmed by Hoberg (1981: 136):<br />

Gern "sche<strong>in</strong>t nicht so fest an die Endstellung gebunden zu se<strong>in</strong> wie die übrigen modalen Elemente."<br />

2 Ich küsse gern SIE.<br />

3a<br />

3b<br />

* Ich küsse laut SIE.<br />

Ich küsse sie laut.<br />

K) Indef<strong>in</strong>ite genitive complements cannot be followed by a modifier:<br />

4a<br />

4b<br />

* Ich habe gehört, daß Peter e<strong>in</strong>er Frau früher bedurfte.<br />

Ich habe gehört, daß Peter früher e<strong>in</strong>er Frau bedurfte.<br />

L) Manner adverbs (a 43 ) must not precede def<strong>in</strong>ite, animate nom<strong>in</strong>ative complements.<br />

We cannot th<strong>in</strong>k <strong>of</strong> any reasonable stress which would make sentences like 5<br />

acceptable:<br />

5 * Deshalb hat laut der Papst gepredigt.<br />

M) No modifiers can follow <strong>in</strong>def<strong>in</strong>ite, <strong>in</strong>animate verb complements other than the<br />

nom<strong>in</strong>ative:<br />

6a<br />

6b<br />

* ..., weshalb ich e<strong>in</strong>en Tisch GEStern geschre<strong>in</strong>ert habe.<br />

* ..., weshalb ich mich e<strong>in</strong>es Tisches GEStern entledigt habe.<br />

N) No modifiers can follow <strong>in</strong>def<strong>in</strong>ite, animate verb complements other than the<br />

nom<strong>in</strong>ative when the arguments have an article (7a). (7b) shows that <strong>with</strong>out an article<br />

this sentence is grammatical:<br />

7a<br />

7b<br />

* Deshalb grüße ich e<strong>in</strong>en Unbekannten <strong>of</strong>t.<br />

Deshalb grüße ich Unbekannte OFT.<br />

O) No elements must follow complements express<strong>in</strong>g an expansion<br />

(Expansivergänzungen):<br />

8 * Der Sportler warf den Speer 100 Meter weit vor 10 Jahren.<br />

P) In our <strong>in</strong>tuition, the comb<strong>in</strong>ation <strong>of</strong> two manner adverbs <strong>in</strong> a simple sentence is also<br />

ungrammatical:<br />

9a<br />

9b<br />

* Er spielte laut eifrig Klavier.<br />

* Er spielte eifrig laut Klavier.<br />

The first part <strong>of</strong> this section is very short given the importance <strong>of</strong> obligatory order rules <strong>in</strong><br />

Natural Language Process<strong>in</strong>g. The reason for this is that these rules have been sufficiently


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 163<br />

discussed <strong>in</strong> Ste<strong>in</strong>berger (1990). We shall now concentrate on the recognition <strong>of</strong> focus,<br />

theme and rheme.<br />

6.2. RECOGNITION OF FOCUS<br />

In this section, we shall look at focalization from the po<strong>in</strong>t <strong>of</strong> view <strong>of</strong> analysis. There are<br />

several means to express focus<strong>in</strong>g <strong>of</strong> a phrase <strong>in</strong> a sentence. We shall discuss them one by<br />

one <strong>in</strong> order to isolate the cases <strong>in</strong> which a phrase is focused. A general po<strong>in</strong>t is that<br />

positional change <strong>of</strong> elements which have a strong tendency to the left or to the right causes<br />

stronger focalization than the permutation <strong>of</strong> other elements.<br />

The position <strong>of</strong> some elements <strong>in</strong> the Vorfeld leads to their strong accentuation. Reis (1987:<br />

169) confirms that stressed verb arguments tend to be positioned <strong>in</strong> the Vorfeld:<br />

[Es] sei daran er<strong>in</strong>nert, daß betonte Verbargumente sehr häufig, ja normalerweise, ihre nichtnormale<br />

(mit m<strong>in</strong>imalem Fokus verbundene) Stellung im Vorfeld haben, das den Satzanfang bildet.<br />

Elements which can automatically be recognised as be<strong>in</strong>g stressed when <strong>in</strong> the Vorfeld are<br />

manner adverbs (10), Konkomitanzangaben (11), <strong>in</strong>strumental modifiers (12), situative (13)<br />

and directional complements (14), <strong>in</strong>def<strong>in</strong>ite verb complements <strong>with</strong> the exception <strong>of</strong> the<br />

nom<strong>in</strong>ative complement (15) as well as <strong>in</strong>f<strong>in</strong>itives (16):<br />

10 SCHNELL bist Du gelaufen.<br />

11 Mit BERNhard will sie <strong>in</strong> den Urlaub fahren.<br />

12 Mit dem HAMmer hat er Mücken gejagt.<br />

13 In MÜNchen wohnt Günter.<br />

14 Nach MÜNchen fährt er.<br />

15 E<strong>in</strong>e INder<strong>in</strong> hat Anne im Bus kennengelernt.<br />

16 SCHENken will er mir das Buch nicht.<br />

The focalization <strong>of</strong> these seven types <strong>of</strong> elements is probably due to the fact that the Vorfeld is<br />

typically filled by elements <strong>with</strong> a l<strong>in</strong>k<strong>in</strong>g function but that these seven classes are typically<br />

rhematic. They are also quite closely l<strong>in</strong>ked to the verb and have a strong tendency towards the<br />

end <strong>of</strong> the sentence. The elements <strong>in</strong> sentences such as 10 to 16 sometimes do not carry the<br />

ma<strong>in</strong> focus <strong>of</strong> the sentence but they are always contrastive elements. It is for example very<br />

difficult to f<strong>in</strong>d a possible context for 12 <strong>in</strong> which the <strong>in</strong>strumental modifier is not focused:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 164<br />

17a ? Mit dem HAMmer hat er MÜCken gejagt, mit dem SCHRAUBENzieher AMeisen.<br />

17b ?? Mit dem Hammer hat er Ameisen gejagt und nicht FLIEgen.<br />

17c ?? Mit dem Hammer hat er Ameisen geJAGT und nicht geSTREIchelt.<br />

Nom<strong>in</strong>ative and other def<strong>in</strong>ite verb complements, as well as existimatorial and situative<br />

modifiers, on the other hand <strong>of</strong>ten have l<strong>in</strong>k<strong>in</strong>g function and are likely to be the topic <strong>of</strong> a<br />

sentence. These elements can be focused (21a) when they are positioned <strong>in</strong> the Vorfeld but<br />

the neutral read<strong>in</strong>g is as likely (21b, 18, 19, 20). Therefore, the Vorfeld position <strong>of</strong> these<br />

phrases should not be analysed as focalization:<br />

18 Ralf hätte gerne im Bus e<strong>in</strong>e Inder<strong>in</strong> kennengelernt.<br />

19 Wahrsche<strong>in</strong>lich hat Axel Fieber.<br />

20 Gestern habe ich Wim Wenders' Reise ans Ende der Welt gesehen.<br />

21a DIEses Buch will Paul lesen (... und ke<strong>in</strong> ANderes).<br />

21b Dieses Buch habe ich LETZtes Jahr schon gelesen.<br />

Schwartz and Tomaselli (1991: 252) po<strong>in</strong>t out the <strong>in</strong>terest<strong>in</strong>g fact that the position <strong>of</strong> oblique<br />

pronouns <strong>in</strong> the Vorfeld is marked (22). This is unexpected because pronom<strong>in</strong>al elements are<br />

thematic and should thus be unmarked <strong>in</strong> sentence-<strong>in</strong>itial position. This can probably also be<br />

expla<strong>in</strong>ed by the larger verb bond<strong>in</strong>g <strong>of</strong> non-nom<strong>in</strong>ative complements. We did not however<br />

list this case among the focalis<strong>in</strong>g sequences shown <strong>in</strong> 10 to 16, as a read<strong>in</strong>g <strong>with</strong> another<br />

focused phrase <strong>in</strong> the sentence is possible. As <strong>in</strong> the sentences 18 to 21, ihm <strong>in</strong> 22 can be<br />

stressed contrastively, but this is not compulsory.<br />

22 Ihm hat Piklu e<strong>in</strong>e runtergehauen.<br />

As def<strong>in</strong>ite pronouns tend to the left, their right-movement leads to their focalization (23, 124).<br />

We would expect def<strong>in</strong>ite and animate nom<strong>in</strong>ative complements to behave <strong>in</strong> the same way but<br />

they are not necessarily focused when follow<strong>in</strong>g a modifier (25). We mentioned <strong>in</strong> 6.1. (po<strong>in</strong>t<br />

L) that modal modifiers cannot precede this group <strong>of</strong> nom<strong>in</strong>ative complements at all.<br />

23 Axel hat Cornelia IHM vorgezogen (und nicht etwa umgekehrt).<br />

24 Philippe wollte dennoch IHN küssen.<br />

25 Deshalb hat gestern der Papst se<strong>in</strong> Konto überzogen.<br />

For elements which tend strongly to the right the opposite holds: When other elements<br />

follow them, the ones <strong>with</strong> right-tendency are thematised and the ones moved to the right are


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 165<br />

focused. The elements whose left-movement has this effect are: POs <strong>in</strong> full form (26),<br />

directional (27) and situative complements (28), genitive full NPs (29) and <strong>in</strong>def<strong>in</strong>ite and<br />

animate NPs <strong>with</strong>out a determ<strong>in</strong>er (30):<br />

26 Kutluk unterrichtete uns über den Sieg e<strong>in</strong>e Woche zu SPÄT.<br />

27 Ralf schrieb Upamanyu nach Indien immer GERN.<br />

28 Ralf wohnt <strong>in</strong> diesem Haus schon LANge.<br />

29 Er wollte sich se<strong>in</strong>er Frau am MONtag entledigen.<br />

30 Deshalb grüße ich Unbekannte OFT.<br />

We mentioned <strong>in</strong> 6.1 (A, H, M, N and O) that NPs and PPs <strong>of</strong> support verb constructions,<br />

predicative elements, <strong>in</strong>def<strong>in</strong>ite and <strong>in</strong>animate NPs, <strong>in</strong>def<strong>in</strong>ite and animate NPs <strong>with</strong> a<br />

determ<strong>in</strong>er, as well as expansion complements cannot be moved to the left at all. Pronom<strong>in</strong>al<br />

POs and modal modifiers do not have enough right-tendency to actually focus other<br />

elements. We shall therefore mention them when discuss<strong>in</strong>g the recognition <strong>of</strong> rhemes.<br />

When whatever element is modified by a degree modifier (31), or when it is partially<br />

negated (32) (Sondernegation), it is focused, as well:<br />

31 Till küßt sogar MÄNner.<br />

32 Bärbel schickte den Brief nicht der FRAU.<br />

As degree modification and partial negation cause the focus<strong>in</strong>g <strong>of</strong> the modified phrase, it<br />

should not be necessary to use a focus<strong>in</strong>g construction, such as cleft<strong>in</strong>g, to express the<br />

focus<strong>in</strong>g <strong>in</strong> a fixed-word order language (cf. 3.3.2).<br />

Some languages, such as English, tend to express focus<strong>in</strong>g less than <strong>German</strong> (Sgall, 1982:<br />

68). When translat<strong>in</strong>g <strong>in</strong>to <strong>German</strong>, we thus cannot rely on the fact that focus has been<br />

recognised dur<strong>in</strong>g analysis. However, as partial negation and degree modification always<br />

lead to focalization, we can automatically add the focus feature to the degree-modified<br />

phrase before generat<strong>in</strong>g the <strong>German</strong> sentence. In this way, we can recover at least some <strong>of</strong><br />

the focuses we failed to recognise dur<strong>in</strong>g analysis.<br />

We have seen several examples <strong>in</strong> which the verb had to be focused. The reason for this is<br />

not that the verb itself was moved, as its position depends on the sentence type rather than on<br />

functional sentence perspective. Its focalization can rather be expla<strong>in</strong>ed by the restriction


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 166<br />

concern<strong>in</strong>g the Mitteilungszentrum, discussed <strong>in</strong> 4.4.4: When a non-focusable pragmatic<br />

modifier is the last non-verbal element <strong>in</strong> the sentence, the verb has to be focused. Pragmatic<br />

modifiers exclude preced<strong>in</strong>g elements from be<strong>in</strong>g thematic, and the condition says that at<br />

least one element <strong>in</strong> the sentence must be a possible rheme. The verb is the only available<br />

element left (33). In sentences <strong>with</strong> auxiliary or modal verbs, we do not have any means to<br />

dist<strong>in</strong>guish whether the auxiliary or the ma<strong>in</strong> verb are focused (34):<br />

33 Randy ASS den Regenwurm wohl.<br />

34 Peter will heute h<strong>of</strong>fentlich KOMmen.<br />

In a lot <strong>of</strong> other sentences we have no means <strong>of</strong> f<strong>in</strong>d<strong>in</strong>g out whether the verb is focused or<br />

not as this simply depends on the context, to which we do not have access.<br />

Independently <strong>of</strong> whether other languages have means to express focalization <strong>of</strong> the verb,<br />

the identification <strong>of</strong> such a verb focus can be <strong>of</strong> use. As we assume that every sentence has<br />

only one focus it helps to avoid double-assignment <strong>of</strong> focus.<br />

To summarise this section, we can say that focus<strong>in</strong>g constructions <strong>in</strong> <strong>German</strong> are:<br />

A) Marked Vorfeld position<br />

B) Right-movement <strong>of</strong> def<strong>in</strong>ite personal pronouns<br />

C) Right-movement <strong>of</strong> elements beh<strong>in</strong>d some elements which strongly tend to the right<br />

D) Degree modification<br />

E) Partial negation (negation <strong>of</strong> a s<strong>in</strong>gle element <strong>in</strong> the sentence)<br />

F) Restriction <strong>of</strong> the focus on the verb due to the Mitteilungszentrum-condition (cf.<br />

4.4.4)<br />

In Ste<strong>in</strong>berger (1992a: 33ff), we describe how we realised the implementation <strong>of</strong> focus<br />

recognition <strong>in</strong> the CAT2 formalism. English focus<strong>in</strong>g constructions are discussed at the same<br />

place, pages 23f.<br />

Two po<strong>in</strong>ts are worth remark<strong>in</strong>g upon: One is the fact that focus<strong>in</strong>g constructions may not be<br />

frequent <strong>in</strong> written language at all. When analys<strong>in</strong>g spoken language, we may not need the<br />

word order analysis as phonetic <strong>in</strong>formation on the sentence stress is available. The other is<br />

the problem that we have no means but our <strong>in</strong>tuition to dist<strong>in</strong>guish between what we call


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 167<br />

focalization and rhematization. Other native speakers may disagree <strong>in</strong> their judgement <strong>of</strong><br />

how strongly the phrases <strong>in</strong> the sentences <strong>of</strong> this chapter are focused.<br />

The analysis may also depend on the comparison <strong>with</strong> the language <strong>in</strong>to which we want to<br />

translate. If the equivalent focus<strong>in</strong>g construction (such as cleft<strong>in</strong>g) is much stronger than the<br />

<strong>German</strong> word-order<strong>in</strong>g, translation by this means would sound exaggerated or wrong. In that<br />

case, it would be better to analyse the <strong>German</strong> construction as rheme-identify<strong>in</strong>g <strong>in</strong>stead. To<br />

decide on how to analyse the different phenomena, we should compare the relevant language<br />

pair. When implement<strong>in</strong>g these phenomena for the CAT2 Mach<strong>in</strong>e Translation formalism<br />

we did not dist<strong>in</strong>guish between focus and rheme. The disadvantage <strong>of</strong> that approach is that<br />

strong focalization such as <strong>in</strong> (35) is analysed the same as the weak one <strong>in</strong> (36):<br />

35 Er hat gestern IHN gesehen.<br />

36 Er hat Petra gestern e<strong>in</strong>en Bären aufgebunden.<br />

We shall discuss these problems aga<strong>in</strong> <strong>in</strong> 6.8.<br />

6.3. RECOGNITION OR THEME AND RHEME<br />

Recognis<strong>in</strong>g the thematic element <strong>in</strong> a sentence is probably easiest. We assign the feature<br />

{theme=yes} (expressed <strong>in</strong> attribute value notation {attribute=value}) to the first element <strong>in</strong><br />

the sentence if this is not already marked as be<strong>in</strong>g focused. By this sequence <strong>of</strong> feature<br />

assignment, we avoid phrases such as the accusative NP <strong>in</strong> (37) be<strong>in</strong>g analysed as be<strong>in</strong>g both<br />

focus and theme. We do not claim that focus and theme generally exclude each other, but<br />

there is no practical use for us <strong>in</strong> comb<strong>in</strong><strong>in</strong>g these two features:<br />

37 E<strong>in</strong>en REgenwurm will er essen.<br />

All other elements automatically get the complementary feature {theme=no}. By this<br />

procedure, nom<strong>in</strong>atives are analysed as be<strong>in</strong>g thematic when they are sentence-<strong>in</strong>itial. If<br />

other elements such as modifiers or object NPs etc start the sentence, these are marked as<br />

be<strong>in</strong>g the theme. By sentence-<strong>in</strong>itial position we either mean the Vorfeld or the first position<br />

<strong>in</strong> verb-f<strong>in</strong>al sentences.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 168<br />

Rhematic elements can be recognised <strong>in</strong> a similar way: If no element <strong>in</strong> the sentence has<br />

been marked as be<strong>in</strong>g the focus, the last constituent <strong>in</strong> the sentence but the verb 73 will be<br />

recognised as rheme (1, 2, 3):<br />

38 Nadia sprach mit Randy über die Unterschiedlichkeit der Geschmäcker.<br />

39 Daraufh<strong>in</strong> hat Randy Hans kürzlich e<strong>in</strong>en Regenwurm geschenkt.<br />

40 Dieser verschlang ihn eilig.<br />

Accord<strong>in</strong>g to what we stated <strong>in</strong> the preced<strong>in</strong>g paragraph, the toner wohl <strong>in</strong> (41) would be<br />

analysed as be<strong>in</strong>g rhematic. This is wrong, however, as wohl cannot carry the sentence<br />

focus 74 . Instead, the verb must be focused. To avoid wrong analysis, non-stressable<br />

modifiers should be marked <strong>with</strong> a feature <strong>in</strong>dicat<strong>in</strong>g that they cannot be rhematic. We<br />

suggest such a feature <strong>in</strong> section 6.7, and apply it to the adverb lists <strong>in</strong> the appendix (8.2 and<br />

8.3). One can roughly say that pragmatic modifiers cannot be rhematic or stressed, whereas<br />

manner adverbs (a 43 ) can. However, this generalization allows some exceptions, such as the<br />

existimatorial adverb wirklich 15 <strong>in</strong> 42:<br />

41 Er GLAUBT es deshalb 22 wohl 12 .<br />

42 Peter wird heute h<strong>of</strong>fentlich KOMmen.<br />

43 Er glaubt es deshalb 22 WIRKlich 14 .<br />

In addition to most modal modifiers, full-PP and full-NP complements can be rhematic.<br />

Many situative modifiers can also be rhematic and focused (45), but some <strong>of</strong> them cannot<br />

(46a). This quality should also be mentioned <strong>in</strong> the dictionary.<br />

45 Er sprach deshalb 22 GEStern 26 .<br />

46a * Er sprach deshalb 22 SCHON 39 .<br />

46b Er SPRACH deshalb 22 schon 39 .<br />

73 For the reason why the verb is excluded, see below <strong>in</strong> this section. Note also that, accord<strong>in</strong>g to what we<br />

said <strong>in</strong> 4.4.1, the whole constituent über die Unterschiedlichkeit der Geschmäcker <strong>in</strong> 38 is marked as<br />

rhematic, and not the modify<strong>in</strong>g NP der Geschmäcker.<br />

74 Note, however, that every word can be focused when used <strong>in</strong> metalanguage:<br />

44 Ich sagte "WOHL", und nicht "HOHL".


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 169<br />

Intuitively, we assumed that the elements which cannot be rhematic are the same elements<br />

which cannot be negated. The comparison <strong>of</strong> the features <strong>in</strong> appendix 8.3 shows, however,<br />

that this is not always the case.<br />

Although elements <strong>with</strong> right-tendency such as modal adverbs, <strong>in</strong>def<strong>in</strong>ite NPs, etc have a<br />

stronger tendency to be rhematic than other constituents, no further rules will be necessary to<br />

identify the rheme. If a modal modifier, for <strong>in</strong>stance, does not appear at the very end <strong>of</strong> the<br />

sentence, it will not have the ma<strong>in</strong> sentence accent. In (47), either e<strong>in</strong>en Regenwurm or zum<br />

Geburtstag can have the ma<strong>in</strong> accent. In (48), eilig cannot and vor dem Abendessen is most<br />

likely to be the rheme. In both cases, our algorithm assigns the rheme feature to the last<br />

constituent. Other elements, such as directional complements etc, will normally be <strong>in</strong><br />

sentence-f<strong>in</strong>al position and they will thus be analysed correctly as be<strong>in</strong>g rhematic.<br />

47 Randy hat Hans kürzlich e<strong>in</strong>en Regenwurm zum Geburtstag geschenkt.<br />

48 Er verschlang ihn eilig 43 vor dem Abendessen.<br />

Pronom<strong>in</strong>al prepositional objects cannot be rhematic, as (49) and (50) show, and should<br />

therefore get a similar feature <strong>in</strong> the lexicon:<br />

49a Kate dachte daran zu SPÄT.<br />

49b * Kate dachte zu spät daRAN.<br />

49c ? Kate dachte zu SPÄT daran.<br />

50a Kate gab gestern 10 PFUND dafür aus.<br />

50b * Kate gab gestern 10 Pfund DAFÜR aus.<br />

Although verbs are frequently rhematic, this algorithm excludes the possibility that verbs are<br />

identified as rhemes. The reason for this is that it is difficult to be sure about the status <strong>of</strong> the<br />

verb. Reis (1987: 169) mentions some verbs which attract the sentence focus, <strong>in</strong>clud<strong>in</strong>g<br />

tanzen, beten and gefallen, but their number is very small. The suggestion by Hajicová, Sgall<br />

and Skoumalová (1993: 180), who <strong>of</strong>fer an algorithm for English to identify rhematic verbs,<br />

is <strong>of</strong> more help:<br />

if the ma<strong>in</strong> verb <strong>of</strong> sentence n has the same mean<strong>in</strong>g as (or a mean<strong>in</strong>g <strong>in</strong>cluded <strong>in</strong>) that <strong>of</strong> sentence n-1,<br />

then it belongs to the topic; also verbs <strong>with</strong> very general lexical mean<strong>in</strong>gs (such as be, have, happen,<br />

carry out, become may be handled as belong<strong>in</strong>g to the topic. Otherwise (i.e. <strong>in</strong> the unmarked case), the<br />

verb generally belongs to the focus.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 170<br />

The ma<strong>in</strong> problems l<strong>in</strong>ked to this proposition are that a complex semantic system is needed<br />

to recognise whether the mean<strong>in</strong>g <strong>of</strong> a verb is <strong>in</strong>cluded <strong>in</strong> that <strong>of</strong> the preced<strong>in</strong>g sentence.<br />

Furthermore, depend<strong>in</strong>g on the complexity <strong>of</strong> the sentence n-1, there can be several verbs,<br />

and the algorithm <strong>in</strong>volves operation over sentence borders, which most Mach<strong>in</strong>e<br />

Translation systems cannot handle. Alternatively, one could leave aside the first condition<br />

(<strong>in</strong>clusion <strong>of</strong> mean<strong>in</strong>g), and decide for every verb whether it can be rhematic or not. This<br />

<strong>in</strong>formation can be encoded <strong>in</strong> the dictionary, similarly to our suggestion for adverbs.<br />

However, we decided to leave verbs aside <strong>in</strong> our implementation <strong>of</strong> theme and rheme<br />

recognition, because the verb position is generally fixed by other factors, so that it does not<br />

matter whether the verb is rhematic. In 10 to 13, for <strong>in</strong>stance, it would be irrelevant for<br />

<strong>German</strong> whether the verbs are rhematic or not.<br />

We suggest that only one rheme feature be assigned <strong>in</strong> a simple sentence. This means that <strong>in</strong><br />

47, for <strong>in</strong>stance, only zum Geburtstag should be marked as rhematic, although e<strong>in</strong>en<br />

Regenwurm could <strong>in</strong> pr<strong>in</strong>ciple be rhematic as well. The algorithm is not perfectly reliable, as<br />

there is the possibility that either e<strong>in</strong>en Regenwurm or zum Geburtstag can be rhematic. As<br />

<strong>with</strong> the theme and the focus assignment, it happens to be the case that whatever choice we<br />

make, we can go wrong. The only possibility we have is to go for the most likely analysis.<br />

In this and the preced<strong>in</strong>g section, we gathered rules and regularities to identify focus, theme<br />

and rheme <strong>of</strong> sentences <strong>with</strong>out referr<strong>in</strong>g to the context. These categories may help us to<br />

concatenate sentences by assign<strong>in</strong>g them a place <strong>in</strong> the canonical form presented <strong>in</strong> 5.3.3.<br />

The f<strong>in</strong>al canonical form, <strong>in</strong>clud<strong>in</strong>g these categories, will be developed <strong>in</strong> 6.5.<br />

6.4. SOME MORE DETAILS<br />

In this section, we want to po<strong>in</strong>t out some more details concern<strong>in</strong>g the idiosyncratic<br />

behaviour <strong>of</strong> some adverbs, or rather adverb sequences. These aspects are not necessary for<br />

a successful treatment <strong>of</strong> word order, but they underp<strong>in</strong> the relevance <strong>of</strong> focus recognition.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 171<br />

Furthermore, they can help language learners to better understand the mechanics and the<br />

effects <strong>of</strong> word order variation.<br />

The goal is to f<strong>in</strong>d out to which extent the effect <strong>of</strong> modifier comb<strong>in</strong>ations and permutations<br />

is predictable by discuss<strong>in</strong>g the comb<strong>in</strong>ation <strong>of</strong> the different subtypes <strong>with</strong> each other. We<br />

shall see that, after all that has been said so far, we do not as yet know everyth<strong>in</strong>g about the<br />

effects <strong>of</strong> modifier placement and their permutation. In that sense, this section encourages<br />

further l<strong>in</strong>guistic <strong>in</strong>vestigation.<br />

6.4.1. PERMUTATION OF PRAGMATIC MODIFIERS<br />

We mentioned <strong>in</strong> 6.1 that toners are strictly ordered. Other pragmatic modifiers can permute<br />

(51a, 51b, 51c). Although some <strong>of</strong> the sentences <strong>in</strong> (51) seem more natural than others, and<br />

(51d) is even ungrammatical, none <strong>of</strong> the existimatorial modifiers can be rhematic. For this<br />

reason, their permutation does not have any effect on the theme-rheme structure. For our<br />

purpose, it follows that we can always generate the order <strong>of</strong> pragmatic modifiers accord<strong>in</strong>g<br />

to the canonical form, <strong>with</strong>out variation. A further consequence is that, dur<strong>in</strong>g analysis, we<br />

do not have to care about the order <strong>of</strong> pragmatic modifiers, as it has no significance.<br />

51a Er war h<strong>of</strong>fentlich 13 klugerweise 14 tatsächlich 15 VOLLkaskoversichert.<br />

51b Er war klugerweise 14 h<strong>of</strong>fentlich 13 tatsächlich 15 VOLLkaskoversichert.<br />

51c Er war h<strong>of</strong>fentlich 13 tatsächlich 15 klugerweise 14 VOLLkaskoversichert.<br />

51d * Er war tatsächlich 15 klugerweise 14 h<strong>of</strong>fentlich 13 VOLLkaskoversichert.<br />

There is however a subgroup <strong>of</strong> pragmatic modifiers which must be exempt from this<br />

statement. The order <strong>of</strong> elements <strong>of</strong> the modifier group a 7 , which Hoberg calls<br />

sprechaktgliedernd (speech act-order<strong>in</strong>g), is relevant, as one can see <strong>in</strong> (52a) and (53a):<br />

52a Er ist bekanntlich 12 erstens 7 e<strong>in</strong> Mörder, zweitens 7 säuft er, ...<br />

53a Er ist erstens 7 bekanntlich 12 e<strong>in</strong> Mörder, zweitens 7 säuft er, ...<br />

In (52a) bekanntlich refers to what is said under erstens and zweitens. In (53a), we only<br />

know that the person <strong>in</strong> question is a murderer, whereas the fact that this person is a dr<strong>in</strong>ker<br />

is not known to everybody. These speech act-order<strong>in</strong>g modifiers create a strict hierarchy,<br />

which can be shown graphically by us<strong>in</strong>g a colon:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 172<br />

52b Er ist bekanntlich 12 :<br />

erstens 7 e<strong>in</strong> Mörder<br />

zweitens 7 säuft er, ...<br />

53b Er ist:<br />

erstens 7 bekanntlich 12 e<strong>in</strong> Mörder<br />

zweitens 7 säuft er, ...<br />

6.4.2. PERMUTATION OF MODAL MODIFIERS<br />

Modal modifiers <strong>in</strong>clude Hoberg's position classes a 42 , a 43 and a 44 . Sentences <strong>with</strong> two<br />

modal adverbs, such as (54a), <strong>of</strong>ten sound heavy, or are even ungrammatical. The reason for<br />

this could be that both attract the sentence focus, which contradicts Altmann's and<br />

Rochemont's claim that simple sentences should only have one focus. The sentence sounds<br />

better when one <strong>of</strong> the modifiers moves to the Vorfeld (54b):<br />

54a * Die Radfahrer fuhren geme<strong>in</strong>sam 42 SCHNELL 43<br />

54b GeMEINsam 42 fuhren die Motorradfahrer SCHNELL 43 .<br />

When Konkomitanzangaben (a 42 ) follow manner adverbs (a 43 ) there must be another<br />

element com<strong>in</strong>g after both <strong>of</strong> them (54c and 54d). It seems that Konkomitanzangaben must<br />

not carry the sentence focus when a manner adverb is present (54c), whereas they can when<br />

no manner adverb is <strong>in</strong>volved (55). Probably this data can be expla<strong>in</strong>ed by say<strong>in</strong>g that<br />

manner adverbs attract the sentence stress more strongly than Konkomitanzangaben.<br />

54c * Die Radfahrer fuhren schnell43 geme<strong>in</strong>sam42.<br />

54d Die Radfahrer fuhren schnell43 geme<strong>in</strong>sam42 <strong>in</strong> die Stadt.<br />

55 Die Radfahrer fuhren geme<strong>in</strong>sam42.<br />

When manner (a 43 ) and <strong>in</strong>strumental modifiers (a 44 ) are comb<strong>in</strong>ed, the opposite holds: the<br />

manner adverb must precede the <strong>in</strong>strumental modifier (56a, 56b), except if a follow<strong>in</strong>g<br />

element attracts the sentence stress (56c):<br />

56a ? Er predigte die Worte deutlich 43 über den Lautsprecher 44 .<br />

56b * Er predigte die Worte über den Lautsprecher 44 deutlich 43 .<br />

56c Er predigte über den Lautsprecher44 deutlich43 biblische Worte.<br />

Konkomitanzangaben (a 42 ) and <strong>in</strong>strumental modifiers (a 44 ) are more compatible. Both<br />

sequences are grammatical. The latter element is more strongly stressed (57a, 57b). When


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 173<br />

the latter is pronom<strong>in</strong>al, it cannot carry the stress (57c) and the preced<strong>in</strong>g modifier takes this<br />

role (57d):<br />

57a Er sprach mit Brigitte 42 über das Telefon 44 .<br />

57b Er sprach über das Telefon 44 mit Brigitte 42 .<br />

57c * Er sprach mit Brigitte 42 daRÜber 44 .<br />

57d Er sprach mit BriGITte 42 darüber 44 .<br />

We mentioned <strong>in</strong> 6.1 that, <strong>in</strong> our <strong>in</strong>tuition, the comb<strong>in</strong>ation <strong>of</strong> two manner adverbs is not<br />

grammatical at all.<br />

6.4.3. PERMUTATION OF PRAGMATIC AND SITUATIVE/MODAL<br />

MODIFIERS<br />

Pragmatic modifiers tend to precede situative modifiers, as they refer to the super-ord<strong>in</strong>ated<br />

verb (cf. 2.3 and 3.4.1.2). When elements <strong>of</strong> these two groups are permuted we can assume<br />

that the situative adverb is thematic. The pragmatic adverb, however, is not rhematic. This is<br />

why the verb <strong>in</strong> (58b), and either the verb or the adverb spät <strong>in</strong> (58c) and (58d), must be<br />

rhematic:<br />

58a ? Er kam vielleicht 12 DEShalb 22 .<br />

58b ? Er KAM deshalb 22 vielleicht 12 .<br />

58c Er kam deshalb 22 vielleicht 12 zu SPÄT 40 .<br />

58d ? Er KAM deshalb 22 vielleicht 12 zu spät 40 .<br />

Although the modal modifier group a 43 cannot precede pragmatic and situative modifiers,<br />

the two other modal modifier classes (a 42 : Konkomitanzangaben and a 44 : <strong>in</strong>strumental<br />

modifiers) can. In that case, the modal modifiers are thematised:<br />

59 Sie fuhren mite<strong>in</strong>ander 42 eben 5 <strong>in</strong> URlaub.<br />

60 Sie sprach damit 44 (mit dem Megaphon) <strong>in</strong>dessen 6 zu den DemonsTRANten.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 174<br />

6.4.4. PERMUTATION OF SITUATIVE MODIFIERS<br />

The group <strong>of</strong> situative modifiers comprises the largest amount <strong>of</strong> subclasses 75 . It is also the<br />

most complicated and <strong>in</strong>trigu<strong>in</strong>g group.<br />

Some sequences <strong>of</strong> situative modifiers are ungrammatical, or at least less natural than others.<br />

We have not found out yet what the reasons for this are. Before go<strong>in</strong>g <strong>in</strong>to details, we want<br />

to po<strong>in</strong>t out that the facts mentioned <strong>in</strong> this section pose no problems for the computational<br />

treatment <strong>of</strong> modifiers. When analys<strong>in</strong>g sentences, we can assume that the order <strong>of</strong> elements<br />

is grammatical. The means provided so far are enough to f<strong>in</strong>d out theme, rheme and focus <strong>of</strong><br />

the sentence, and this is all we need. When generat<strong>in</strong>g <strong>German</strong> sentences, we can use the<br />

order foreseen by the canonical form. In this manner, we avoid all possible problems.<br />

It is however unsatisfactory to accept the existence <strong>of</strong> this <strong>in</strong>sufficiently expla<strong>in</strong>ed section on<br />

the mechanism <strong>of</strong> modifier position<strong>in</strong>g. We do not have suggestions on how to solve this<br />

problem, but we would like to mention it <strong>in</strong> order to encourage further research on this<br />

subject.<br />

Situative modifiers have been discussed from different angles, such as natural serialisation<br />

(Lenerz, 1977; Vennemann, 1982), word order treatment <strong>in</strong> grammar theory (Oliva, 1992b),<br />

and restrictions on their comb<strong>in</strong>ation (Ste<strong>in</strong>itz, 1969; Bartsch, 1972). Depend<strong>in</strong>g on their<br />

purposes, most l<strong>in</strong>guists choose and discuss only a small subset <strong>of</strong> classes, such as local,<br />

causal and temporal modifiers (<strong>with</strong> the subclasses temporal, durative and iterative). To our<br />

knowledge, no research has been carried out on the whole range <strong>of</strong> comb<strong>in</strong>ations <strong>of</strong> Hoberg's<br />

22 situative modifier classes.<br />

Accord<strong>in</strong>g to what has been said so far, one should assume that situative modifiers follow <strong>in</strong><br />

the order <strong>of</strong> the canonical form, and that the right-movement <strong>of</strong> each modifier causes its<br />

75 We suspect that the biggest modifier subgroup <strong>in</strong> terms <strong>of</strong> amount <strong>of</strong> elements belong<strong>in</strong>g to this subclass<br />

is the group a 43 . It <strong>in</strong>cludes all manner adverbs, as well as the adjectives that we classified as sentence<br />

adjectives.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 175<br />

focalisation. However, this is not always the case. The permutation <strong>of</strong> modifiers has more<br />

effect when there is no other possible rheme beh<strong>in</strong>d them (61). This is due to the fact that <strong>in</strong><br />

this case the last modifier carries the stress. When there is another phrase follow<strong>in</strong>g the<br />

modifiers, it carries the sentence stress <strong>in</strong>stead <strong>of</strong> the modifiers (62). There seems to be very<br />

little difference between (62a) and (62b) whereas there is a noticeable difference between<br />

(61a) and (61b):<br />

61a Er hält den Vortrag bei gutem Wetter 28 auf der TAgung 29 .<br />

61b Er hält den Vortrag auf der Tagung 29 bei gutem WETter 28 .<br />

62a Er will bei gutem Wetter 28 auf der Tagung 29 e<strong>in</strong>en VORtrag halten.<br />

62b Er will auf der Tagung 29 bei gutem Wetter 28 e<strong>in</strong>en VORtrag halten.<br />

(61a) and (61b) belong to different contexts but they are equally grammatical. However,<br />

when local and temporal modifiers co<strong>in</strong>cide, one <strong>of</strong> the two possible orders is less natural<br />

than the other:<br />

63a Der Meister sprach letzte Woche 26 am M.I.T. 27 .<br />

63b ? Der Meister sprach am M.I.T. 27 letzte WOche 26 .<br />

In 63a, both modifiers can have the sentence accent. In 63b, the accent is restricted to the<br />

second modifier (compare Lenerz, 1977: 80ff). 63a is thus less contextually restricted. It<br />

<strong>of</strong>fers more possible rhemes.<br />

In addition to the two different effects <strong>of</strong> modifier permutation shown <strong>in</strong> 61, 62 and <strong>in</strong> 63,<br />

there is a third case. The permutation <strong>of</strong> the situative modifiers <strong>in</strong> 64 and 65 is not<br />

grammatical at all. 64 and 65 are particularly astonish<strong>in</strong>g, as the directional complement<br />

follow<strong>in</strong>g the modifiers is likely to be rhematic, so that the permutation <strong>of</strong> the modifiers<br />

should have less importance. Furthermore, we see <strong>in</strong> 66 that the concessive adverb dennoch<br />

can be stressed:<br />

64a Er fuhr dennoch 19 gleichfalls 35 nach München.<br />

64b * Er fuhr gleichfalls 35 dennoch 19 nach München.<br />

65a Er fuhr daher 22 endlich 33 nach München.<br />

65b * Er fuhr endlich 33 daher 22 nach München.<br />

66 Es war ihm egal. Er fuhr DENnoch 19 nach München.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 176<br />

What is it, that makes (64b) and (65b) ungrammatical? We shall have to look at these<br />

examples separately, start<strong>in</strong>g <strong>with</strong> the first one. In (67a), our <strong>in</strong>tuition tells us that dennoch<br />

can refer to the whole sentence: There is an argument for Paul not to go to Munich but he<br />

goes nevertheless. Gleichfalls <strong>in</strong> (67b), however, must refer to Paul: Paul, too, goes to<br />

Munich. If we wanted to express that Paul not only goes to Berl<strong>in</strong> but also to Munich, we<br />

would have to express this by us<strong>in</strong>g the adverb auch (67c):<br />

67a<br />

67b<br />

67c<br />

Paul fuhr dennoch 19 nach München.<br />

Paul fuhr gleichfalls 35 nach München.<br />

Paul fuhr auch 35 nach MÜNchen.<br />

Gleichfalls is thus a scope-<strong>in</strong>clud<strong>in</strong>g element. In parallel to other degree modifiers and the<br />

negation nicht, gleichfalls has to be stressed strongly when separated from the scope<strong>in</strong>cluded<br />

element. Gleichfalls differs from the other degree modifier auch, <strong>in</strong> that it cannot<br />

refer to the directional complement nach München. Furthermore, it cannot precede the<br />

scope-<strong>in</strong>cluded subject <strong>in</strong> the Vorfeld (67d):<br />

67d * Gleichfalls 35 Paul fuhr nach München.<br />

67e Auch 35 PAUL fuhr nach München.<br />

Apparently, the position <strong>of</strong> dennoch after gleichfalls <strong>in</strong> (64b) requires that dennoch has to be<br />

stressed contrastively. The reason for this sentence to be ungrammatical thus seems to be<br />

that both gleichfalls and dennoch need to be stressed heavily.<br />

We mentioned the second example, 65, <strong>in</strong> order to show that, even <strong>with</strong>out scope <strong>in</strong>clusion<br />

be<strong>in</strong>g <strong>in</strong>volved, some situative modifiers cannot be permuted. Both endlich and daher<br />

def<strong>in</strong>itely refer to the sentence. Nevertheless, the modifier daher express<strong>in</strong>g a cause must not<br />

follow the evaluative-temporal adverb endlich. We cannot see any <strong>in</strong>dependent reason for<br />

this impossible word order, and thus have to accept it for the moment as be<strong>in</strong>g an<br />

idiosyncratic behaviour <strong>of</strong> the two adverbs, or <strong>of</strong> the adverb sequence.<br />

To summarise, we can say that the description <strong>of</strong> the permutation <strong>of</strong> situative modifiers<br />

relative to each other is more complex than the permutation <strong>of</strong> the other modifier groups. We<br />

could dist<strong>in</strong>guish three cases: (a) The permutation results <strong>in</strong> the second modifier be<strong>in</strong>g more


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 177<br />

stressed. (b) Permutation is possible, but one <strong>of</strong> the two sequences is more natural <strong>in</strong> that<br />

both modifiers can be stressed, whereas <strong>in</strong> the other only the latter can. (c) One <strong>of</strong> the two<br />

sequences is ungrammatical. The reason why modifier sequences differ <strong>with</strong> respect to<br />

grammaticality is not clear to us. We have to accept it as idiosyncratic behaviour <strong>of</strong> modifier<br />

sequences.<br />

This result is unsatisfy<strong>in</strong>g and leaves an unexpla<strong>in</strong>ed gap for language teachers and learners.<br />

However, it does not pose a problem for the computational treatment <strong>of</strong> modifier placement.<br />

In generation, we can simply use the order <strong>of</strong> modifiers suggested by the canonical form. In<br />

analysis, we can assume that the <strong>in</strong>put sentences are grammatical. The means to identify<br />

theme, rheme and focus <strong>of</strong> the sentence are sufficient to guarantee correct translation.<br />

6.5. FINAL VERSION OF THE CANONICAL FORM<br />

In 5.6, we discussed the importance <strong>of</strong> the categories theme, rheme and focus for the<br />

generation <strong>of</strong> <strong>German</strong> sentences. In 6.2 and 6.3, we gave means to identify them dur<strong>in</strong>g<br />

analysis. Now, we have to decide on the place <strong>of</strong> these categories <strong>in</strong> the canonical form. We<br />

shall start by determ<strong>in</strong><strong>in</strong>g the position <strong>of</strong> thematic elements, and then do the same for rheme<br />

and focus.<br />

6.5.1. PLACEMENT OF THE THEME<br />

The category theme is primarily, but not exclusively, <strong>of</strong> importance for our treatment <strong>of</strong><br />

verb-second sentences, as the theme is the category which is likely to fill the Vorfeld, <strong>in</strong><br />

order to guarantee sentence cohesion (cf. 5.5). In verb-f<strong>in</strong>al sentences, thematicity is<br />

implicitly realised <strong>in</strong> the canonical form through the features pronom<strong>in</strong>ality and def<strong>in</strong>iteness.<br />

Thematic complements are likely to be realised either as a pronoun or as a def<strong>in</strong>ite NP, and<br />

therefore they precede most other elements, even <strong>with</strong>out be<strong>in</strong>g marked as be<strong>in</strong>g the theme.<br />

Pronouns are always thematic, whereas def<strong>in</strong>ite NPs can be rhematic. Above all, we thus<br />

have to decide where to place thematic modifiers <strong>in</strong> the Mittelfeld, and whether thematic<br />

def<strong>in</strong>ite verb complements are placed differently from rhematic ones.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 178<br />

As themes tend to the left <strong>of</strong> the sentence, one should assume that the category theme has its<br />

place <strong>in</strong> the canonical form somewhere among the elements <strong>with</strong> left-tendency. We shall<br />

<strong>in</strong>deed argue that thematic modifiers immediately follow pronom<strong>in</strong>al elements, so that the<br />

first l<strong>in</strong>e <strong>of</strong> the canonical form is the follow<strong>in</strong>g:<br />

N pron /N +d+a < (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 179<br />

71 Context: Was passierte dann? Warum ist Patrick so blau im Gesicht?<br />

71a # ..., weil der Mann dann 26 ihn geohrfeigt hat. 76<br />

71b ..., weil der Mann ihn dann 26 geohrfeigt hat.<br />

We mentioned before that manner adverbs tend to be rhematic. If they are not rhematic, they<br />

are normally not thematic either, but unmarked <strong>with</strong> respect to both features. It is <strong>in</strong>deed<br />

quite difficult to imag<strong>in</strong>e sentences <strong>with</strong> thematic manner adverbs, as they would probably<br />

not be repeated, except <strong>in</strong> comb<strong>in</strong>ation <strong>with</strong> the verb group. However, as (72a) <strong>in</strong> which<br />

schwer precedes the <strong>in</strong>def<strong>in</strong>ite animate nom<strong>in</strong>ative is grammatical, we shall assume the same<br />

place for thematic manner adverbs, bear<strong>in</strong>g <strong>in</strong> m<strong>in</strong>d that the theme will normally not be<br />

realised by a manner adverb:<br />

72a ..., weshalb unbeholfen 43 e<strong>in</strong> Betrunkener here<strong>in</strong>wankte.<br />

72b ..., weshalb e<strong>in</strong> Betrunkener unbeholfen 43 here<strong>in</strong>wankte.<br />

The preced<strong>in</strong>g examples <strong>in</strong>volved thematic modifiers. The other group we have to consider<br />

are thematic def<strong>in</strong>ite verb complements. The follow<strong>in</strong>g examples show that they behave the<br />

same as modifiers:<br />

73 Context: Wo ist das Paket?<br />

73a Ich vermute, daß der Postbote das Paket der Frau gab.<br />

73b # Ich vermute, daß der Postbote der Frau das Paket gab.<br />

73c * Ich vermute, daß das Paket der Postbote der Frau gab.<br />

As a conclusion, we can confirm that, <strong>in</strong> verb-f<strong>in</strong>al sentences, thematic elements tend to<br />

follow the groups <strong>of</strong> def<strong>in</strong>ite and animate nom<strong>in</strong>ative complements, and <strong>of</strong> pronom<strong>in</strong>al verb<br />

complements. In verb-second sentences, they are likely to fill the Vorfeld.<br />

N pron /N +d+a < (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 180<br />

6.5.2. PLACEMENT OF THE RHEME<br />

We mentioned <strong>in</strong> 6.5.1 that thematicity <strong>of</strong> arguments is implicitly realised <strong>in</strong> the canonical<br />

form, so that we do not have to care about pronouns and <strong>in</strong>def<strong>in</strong>ite NPs. The goal <strong>of</strong> this<br />

section will thus be to f<strong>in</strong>d out the position <strong>of</strong> rhematic modifiers, and <strong>of</strong> def<strong>in</strong>ite NP<br />

complements which are marked as be<strong>in</strong>g rhematic.<br />

In 5.3.3, we could not be precise about the position <strong>of</strong> most modifiers but had to give a range<br />

<strong>with</strong><strong>in</strong> which they can appear. As now we dist<strong>in</strong>guish thematic, rhematic and neutral 77<br />

position classes, we shall be able to be more specific.<br />

Rhematic elements tend to follow thematic ones. Therefore the position RHEME will have<br />

to appear quite late <strong>in</strong> the sequence <strong>of</strong> the canonical form. We suggest to place the rheme<br />

beh<strong>in</strong>d the modifier groups a 41 to a 43 , and before the <strong>in</strong>def<strong>in</strong>ite and animate accusative and<br />

dative complements. We shall discuss this order below and give a few examples which<br />

underp<strong>in</strong> its accuracy:<br />

THEME < N +d-a /N -d+a < (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 181<br />

75a ?? Anthony hat e<strong>in</strong>en Indianer DAmals 26 getr<strong>of</strong>fen.<br />

75b ?? Anthony hat e<strong>in</strong>en Indianer damals 26 geTROFfen.<br />

75c Anthony hat damals 26 e<strong>in</strong>en InDIAner getr<strong>of</strong>fen.<br />

Rhematic modifiers tend to follow the group <strong>of</strong> def<strong>in</strong>ite verb complements (76) whereas they<br />

are likely to precede it when thematic (77):<br />

76 Context: Wann war Mel<strong>in</strong>a betrunken?<br />

Sie holte sich den Rausch GEStern 26 .<br />

? Sie holte sich GEStern 26 den Rausch.<br />

77 Context: Was tat Mel<strong>in</strong>a gestern?<br />

Sie traf gestern 26 den Mann ihrer Träume.<br />

* Sie traf den Mann ihrer Träume gestern 26 .<br />

The example sentences 74 to 77 underp<strong>in</strong> our claim that the position <strong>of</strong> the category rheme<br />

<strong>in</strong> the canonical form should be <strong>in</strong> between def<strong>in</strong>ite and <strong>in</strong>def<strong>in</strong>ite accusative and dative<br />

arguments.<br />

After hav<strong>in</strong>g <strong>in</strong>troduced the categories theme and rheme, we can now specify the canonical<br />

form position <strong>of</strong> neutral pragmatic and situative modifiers: They follow the def<strong>in</strong>ite<br />

accusative and dative verb arguments (78, 79). The a-sentences are clearly better than the b-<br />

sentences, <strong>in</strong> spite <strong>of</strong> the fact that neither the modifier nor the accusative NP have to carry<br />

stress, as the manner adverb gern attracts the sentence focus:<br />

Context: Was sagtest Du gerade über Christoph?<br />

78a Ich sagte, er schenkte Paul<strong>in</strong>e das Buch damals 22 gern 43 .<br />

78b ? Ich sagte, er schenkte Paul<strong>in</strong>e damals 22 das Buch gern 43 .<br />

Context: Was sagtest Du gerade über Christoph?<br />

79a Ich sagte, er schenkte Paul<strong>in</strong>e das Buch sicherlich 12 gern 43 .<br />

79b * Ich sagte, er schenkte Paul<strong>in</strong>e sicherlich 12 das Buch gern 43 .<br />

Note that, <strong>in</strong> spite <strong>of</strong> neutral and rhematic modifiers hav<strong>in</strong>g nearly the same position <strong>in</strong> the<br />

canonical form, it is crucial that they are <strong>in</strong> separate classes. The dist<strong>in</strong>ction is necessary,<br />

because several modifiers can occur <strong>in</strong> one sentence, so that we need means to regulate the<br />

order <strong>of</strong> rhematic and neutral modifiers (80). Furthermore, the category rheme also <strong>in</strong>cludes<br />

complements. 81b can only be generated if neutral modifiers precede the category rheme:<br />

80a Er küßte Maria deshalb 26 meistens 37 .<br />

80b Er küßte Maria meistens 37 deshalb 26+rheme .


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 182<br />

81a Er küßte Maria gestern 26 .<br />

81b Er küßte gestern 26 Maria +rheme .<br />

Note that modifiers can only be marked as rhematic if the relevant feature <strong>in</strong> the dictionary<br />

entry allows it (cf. 6.7 and appendix 8.2/8.3). Therefore, sequences such as 82b will not<br />

occur:<br />

82a Er küßte Maria allenfalls 16 ehrenhalber 22 .<br />

82b * Er küßte Maria ehrenhalber 22 allenfalls 16+rheme .<br />

The second group <strong>of</strong> elements for which we could not def<strong>in</strong>e a precise position <strong>in</strong> the earlier<br />

version <strong>of</strong> our canonical form are manner adverbs relative to the complement group (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 183<br />

rheme. In 3.1.4, we mentioned the contradictory claims put forward by Thurmair (1989:<br />

29ff) and Engel (1988: 340), who say, respectively, that toners and situative modifiers<br />

separate theme and rheme. We can specify now that all modifier classes have this separat<strong>in</strong>g<br />

function, if they are not marked as be<strong>in</strong>g thematic or rhematic themselves.<br />

... < N -d-a < (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 184<br />

could only be shifted to the left, which would have a thematis<strong>in</strong>g, rather than a rhematis<strong>in</strong>g,<br />

function (87). The place for the rhematic PO <strong>in</strong> 87 should thus be the one <strong>in</strong> 87b, and not the<br />

one <strong>in</strong> 87a:<br />

87a ?? Er hat dem Autor für die WIDmung +rheme herzlich 43 gedankt.<br />

87b Er hat dem Autor herzlich 43 für die WIDmung +rheme gedankt.<br />

For the categories <strong>with</strong> absolute right-tendency, such as the nom<strong>in</strong>al part <strong>of</strong> support verb<br />

constructions (6.1., condition A), a shift to the left would even lead to ungrammaticality. The<br />

category rheme should thus only apply to elements which are to the left <strong>of</strong> the category<br />

rheme <strong>in</strong> the canonical form. It should never cause a left-movement <strong>of</strong> whatever category.<br />

The rheme category is therefore limited to rhematic elements <strong>of</strong> the classes: N +d-a , N -d ,<br />

(A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 185<br />

A) Marked Vorfeld position<br />

B) Right-movement <strong>of</strong> elements which strongly tend to the left<br />

C) Right-movement <strong>of</strong> elements beh<strong>in</strong>d elements which strongly tend to the right<br />

D) Degree modification<br />

E) Sondernegation (negation <strong>of</strong> a s<strong>in</strong>gle element <strong>in</strong> the sentence)<br />

F) Restriction <strong>of</strong> the focus on the verb due to the Mitteilungszentrum-condition<br />

As the focus<strong>in</strong>g constructions differ <strong>with</strong> the category <strong>of</strong> the focused elements, we need<br />

different focus positions <strong>in</strong> the canonical form, depend<strong>in</strong>g on what k<strong>in</strong>d <strong>of</strong> element should be<br />

focused. Therefore, we shall discuss the possibilities (A) to (F) one by one.<br />

Focus<strong>in</strong>g <strong>of</strong> degree-modified (D) and partially negated (sondernegiert) phrases (E) does not<br />

depend on their position <strong>in</strong> the sentence, as they are automatically focused by the degree<br />

modification and the negation. Nevertheless, their position should ideally co<strong>in</strong>cide <strong>with</strong> the<br />

sentence focus. Stress<strong>in</strong>g the verb through word order permutation (F) is not possible <strong>in</strong><br />

<strong>German</strong>, so that we cannot consider the last case. The means (A), (B) and (C), however, are<br />

realised by word order variation.<br />

A general problem l<strong>in</strong>ked to the treatment <strong>of</strong> focus is that specific sequences focus <strong>in</strong> some<br />

cases, but not <strong>in</strong> others. For <strong>in</strong>stance, the position <strong>of</strong> the temporal adverb heute beh<strong>in</strong>d an<br />

<strong>in</strong>def<strong>in</strong>ite accusative NP leads to focalization <strong>in</strong> (88), but is ungrammatical <strong>in</strong> (89). This does<br />

not pose a problem for analysis, as there is no danger <strong>in</strong> say<strong>in</strong>g that if a modifier follows an<br />

<strong>in</strong>def<strong>in</strong>ite and <strong>in</strong>animate verb complement, the modifier is focused. In synthesis, however,<br />

we should try to avoid the generation <strong>of</strong> wrong order, such as <strong>in</strong> 89.<br />

88 Peter fürchtet Autos HEUte 26 noch 39 .<br />

89 * Peter sah Autos gestern 26 noch 39 .<br />

We have seen <strong>in</strong> 6.2 that the follow<strong>in</strong>g elements can be focused by their placement <strong>in</strong> the<br />

Vorfeld (A): manner adverbs, Konkomitanzangaben, <strong>in</strong>strumental modifiers, situative and<br />

directional complements, <strong>in</strong>def<strong>in</strong>ite verb complements (90a) <strong>with</strong> the exception <strong>of</strong> the<br />

nom<strong>in</strong>ative, as well as <strong>in</strong>f<strong>in</strong>itives. When generat<strong>in</strong>g verb-second clauses <strong>in</strong>volv<strong>in</strong>g any <strong>of</strong><br />

these focused categories, it is thus appropriate to replace the theme category <strong>in</strong> the Vorfeld<br />

by the focused phrase. In verb-f<strong>in</strong>al clauses, however, these elements cannot be focused, as


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 186<br />

the place at the end <strong>of</strong> the sentence is the neutral position (90b), and left-movement would<br />

have the opposite effect, namely topicalisation. The only means <strong>of</strong> focalis<strong>in</strong>g them <strong>in</strong> written<br />

language would be through a cleft<strong>in</strong>g structure, such as <strong>in</strong> (90c):<br />

90a E<strong>in</strong>e ROSE schenkte er der Frau.<br />

90b ..., weshalb er der Frau e<strong>in</strong>e Rose schenkte.<br />

90c ..., weshalb es e<strong>in</strong>e ROSE war, die er der Frau schenkte.<br />

There is thus no focus position for the categories listed <strong>in</strong> the preced<strong>in</strong>g paragraph <strong>in</strong> verbf<strong>in</strong>al<br />

sentences. To avoid the use <strong>of</strong> cleft<strong>in</strong>g structures, because <strong>of</strong> their heav<strong>in</strong>ess, and<br />

because <strong>of</strong> the computational complexity <strong>in</strong>volved, one can either choose the normal<br />

canonical form position <strong>of</strong> these elements, or the rheme position. This position (90b) does<br />

not require stress, but it allows focalization. One can thus hope for a free ride, as the context<br />

may make it obvious that Rose is focused.<br />

Elements that tend strongly to the left (B) are personal pronouns and def<strong>in</strong>ite, animate<br />

nom<strong>in</strong>ative complements. They can be stressed by be<strong>in</strong>g moved to the right. The question is<br />

what exactly is the position <strong>of</strong> these elements. In 90, the position <strong>of</strong> the focused nom<strong>in</strong>ative<br />

complement beh<strong>in</strong>d the A +d-a constituent is grammatical, whereas <strong>in</strong> 91, it is not. The<br />

difference between 90 and 91 seems to be that the pronoun <strong>in</strong> 91 follows several<br />

constituents, whereas <strong>in</strong> 90 it follows only one. 91 makes it clear that a position too far<br />

towards the end <strong>of</strong> the sentence can cause ungrammaticality. A better place to focus<br />

elements <strong>with</strong> left-tendency seems to be the position follow<strong>in</strong>g the nom<strong>in</strong>ative class N -d+a ,<br />

and preced<strong>in</strong>g the classes (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 187<br />

In 92 and 93, this focus position <strong>with</strong><strong>in</strong> the canonical form gives good results, but <strong>in</strong><br />

sentences <strong>in</strong>volv<strong>in</strong>g other categories, it does not. The nom<strong>in</strong>ative complements <strong>in</strong> 94 and 95<br />

are marked as focus, but word order accord<strong>in</strong>g to the canonical form does not require them<br />

to be stressed. The nom<strong>in</strong>atives can be stressed, but this is not <strong>in</strong>dicated by the word order.<br />

In 94 and 95, the canonical form does not generate an unambiguous sentence, as the<br />

nom<strong>in</strong>atives could be either focus or neutral constituents. In such cases, we have to rely on<br />

context to make it clear that the focus constituents have to be stressed:<br />

94 ..., weil er +focus die Schönheitskönig<strong>in</strong> gestern 26 geküßt hat.<br />

95 ..., weil gestern +theme der Besucher +focus die Schönheitskönig<strong>in</strong> geküßt hat.<br />

As we are not able to f<strong>in</strong>d an unambiguous place for the category focus, it is worth<br />

consider<strong>in</strong>g to always use the cleft<strong>in</strong>g construction <strong>in</strong>stead <strong>of</strong> word order. However, cleft<strong>in</strong>g<br />

structures are stylistically heavy, and therefore we suggest to use this focus position, to the<br />

detriment <strong>of</strong> cases such as 94 and 95.<br />

The elements which neither tend strongly to the left, nor to the right can be focused by be<strong>in</strong>g<br />

positioned beh<strong>in</strong>d elements <strong>with</strong> strong right-tendency (C). We consider the latter to be most<br />

elements from the modal element a 42 to the right <strong>in</strong> the canonical form, namely modifier<br />

groups a 42 to a 44 (96), <strong>in</strong>def<strong>in</strong>ite, <strong>in</strong>animate (cf. 6.2) accusative and dative complements,<br />

non-pronom<strong>in</strong>al genitive complements (97) and prepositional objects (98, 99). The other<br />

complement groups <strong>with</strong> right-tendency, such as the directional complement, for <strong>in</strong>stance,<br />

cannot be moved to the left (100):<br />

96 Wolf schrieb mit dieser Feder 44 e<strong>in</strong>en ganzen RoMAN +focus .<br />

97 T<strong>in</strong>a er<strong>in</strong>nerte sich des Geschehnisses GEStern 26+focus .<br />

98 T<strong>in</strong>a bewarb sich für die Stelle GEStern 26+focus .<br />

99 Deshalb bewarb sich damals 26 für die Stelle e<strong>in</strong> MANN +focus .<br />

100 * T<strong>in</strong>a fuhr <strong>in</strong> die Arbeit GEStern 26+focus .<br />

The problem l<strong>in</strong>ked to the focus position <strong>of</strong> elements <strong>with</strong> left-tendency, namely that not all<br />

sentences generated by this canonical form require strong stress, applies here, as well. In 97<br />

to 99, strong focus on the last constituents is required, whereas the accusative NP <strong>in</strong> 96 is not<br />

more stressed than a normal rhematic constituent.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 188<br />

This focus category comprises all elements which are positioned to its left, exclud<strong>in</strong>g the<br />

elements <strong>in</strong>cluded <strong>in</strong> the first focus category, and thus all accusative, dative and genitive<br />

complements and all modifiers, as well as the nom<strong>in</strong>ative complements <strong>with</strong> the features<br />

N +d-a and N -d+a . Note that the focus elements <strong>with</strong><strong>in</strong> the complex class can be unordered,<br />

as there should not be more than one focus <strong>in</strong> the sentence (<strong>in</strong>dicated by the slash "/").<br />

(A/D/G/PO/N +d-a /N -d /a pragm /a sit /a mod ) +focus<br />

The f<strong>in</strong>al canonical form, <strong>in</strong>clud<strong>in</strong>g theme, rheme and the focus categories is thus the<br />

follow<strong>in</strong>g:<br />

N pron /N +d+b < (A


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 189<br />

ungrammatical. Instead, a default word order is generated, which is correct <strong>in</strong> a large number<br />

<strong>of</strong> sentences.<br />

6.6. PREFERENTIAL PP ATTACHMENT<br />

One <strong>of</strong> the major problems <strong>in</strong> natural language analysis is the highly ambiguous attachment<br />

<strong>of</strong> prepositional phrases. Whittemore, Ferrara and Brunner (1990) discuss several PP<br />

attachment schemes. The attachment predictors <strong>in</strong>clude (1990: 23f):<br />

A) Right association, namely "the tendency for constituents to associate <strong>with</strong> adjacent<br />

items to their right [...], also known as low attachment."<br />

B) M<strong>in</strong>imal attachment, which is the "tendency to attach <strong>in</strong> a manner <strong>in</strong> which the least<br />

number <strong>of</strong> syntactic rules are employed".<br />

C) The lexical preference <strong>of</strong> PPs to attach to certa<strong>in</strong> verbs or nouns, and the preference<br />

<strong>of</strong> certa<strong>in</strong> prepositions for specific k<strong>in</strong>ds <strong>of</strong> constructions.<br />

D) Referential success, which is the prediction that PPs are likely to attach to<br />

<strong>in</strong>def<strong>in</strong>ite NPs and PPs, and to verbs, rather than to def<strong>in</strong>ite NPs, because the former<br />

"require less search over discourse space". This attachment predictor is l<strong>in</strong>ked to the<br />

statement made <strong>in</strong> section 3.1.2 that rhematic elements (<strong>in</strong>def<strong>in</strong>ite NPs) are more likely<br />

to be modified than thematic ones.<br />

To these four methods, we want to add the prediction that PPs <strong>in</strong> sentences <strong>in</strong>volv<strong>in</strong>g strong<br />

stress, as described <strong>in</strong> 6.2, are likely to be attached to the preced<strong>in</strong>g constituent, if by this<br />

attachment the existence <strong>of</strong> contrastive stress can be avoided. The reason for this suggestion<br />

is that word order <strong>in</strong>volv<strong>in</strong>g strong focus<strong>in</strong>g is stylistically marked. It cannot be excluded <strong>in</strong><br />

analysis because it may occur, and if it occurs the focus<strong>in</strong>g should be recognised, so that it<br />

can be expressed <strong>in</strong> the target language <strong>of</strong> the translation. However, if there is a more natural<br />

analysis available, it should be preferred.<br />

We shall thus assume that focus<strong>in</strong>g constructions are relatively unlikely to occur <strong>in</strong> written<br />

text, and suggest that the analysis <strong>in</strong>volv<strong>in</strong>g focus where another analysis is possible should


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 190<br />

be avoided. This is the case when the analysis <strong>of</strong> the PP as an adjunct results <strong>in</strong> a sentence<br />

<strong>with</strong>out contrastive stress.<br />

An example <strong>of</strong> this is sentence 101, <strong>in</strong> which vor der Bank could either be an adjunct,<br />

namely a modifier <strong>of</strong> the nom<strong>in</strong>ative NP der Mann (101a), or a sentence modifier (a 27 )<br />

express<strong>in</strong>g the location <strong>of</strong> the whole event expressed <strong>in</strong> 101 (101b). In the latter case, the<br />

accusative pronoun ihn would have to be stressed contrastively, as shows 102, <strong>in</strong> which the<br />

PP cannot modify the nom<strong>in</strong>ative NP. 101a is def<strong>in</strong>itely a more natural read<strong>in</strong>g than 101b.<br />

Note that, <strong>in</strong> spite <strong>of</strong> the markedness <strong>of</strong> the focalis<strong>in</strong>g construction <strong>in</strong> 101b, we cannot<br />

generally exclude the position <strong>of</strong> a modifier before pronouns. If we did, our grammar would<br />

wrongly exclude sentences such as (102b), where the adjunction shown <strong>in</strong> (102a) is not<br />

possible:<br />

101 Deshalb hat der Mann vor der Bank ihn ignoriert.<br />

101a Deshalb hat {der Mann vor der Bank} ihn ignoriert.<br />

101b ? Deshalb hat der Mann {vor der Bank} 27 IHN ignoriert.<br />

102a * Deshalb hat {er vor der Bank} ihn ignoriert.<br />

102b ?? Deshalb hat er {vor der Bank} 27 IHN ignoriert.<br />

Instead <strong>of</strong> a preference, an absolute word order restriction can be formulated regard<strong>in</strong>g the<br />

last few complement classes <strong>of</strong> the canonical form, namely NPs and PPs <strong>of</strong> support verb<br />

constructions, predicative nouns and adjectives, expansion complements, and complements<br />

other than the nom<strong>in</strong>ative which have the features X -d-a (cf. 6.1, rules H, K, M and O). The<br />

PP <strong>in</strong> der Sprachenschule is an adjunct to Lehrer<strong>in</strong> <strong>in</strong> 103a, as modifiers cannot follow<br />

predicative elements. The word order restriction is clearer <strong>in</strong> 103b, where the adjunct<br />

read<strong>in</strong>g <strong>of</strong> the PP is not possible:<br />

103a Assunta sagt, daß Louisa vor e<strong>in</strong>em Jahr Lehrer<strong>in</strong> <strong>in</strong> der Sprachenschule war.<br />

103b * Assunta sagt, daß Louisa vor e<strong>in</strong>em Jahr gut <strong>in</strong> der Sprachenschule war.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 191<br />

6.7. DICTIONARY ENTRIES FOR ADVERBS<br />

The adverb is the most complicated, and perhaps also the most <strong>in</strong>terest<strong>in</strong>g part <strong>of</strong><br />

speech. Past research <strong>in</strong> natural language process<strong>in</strong>g, however, has not dealt<br />

seriously <strong>with</strong> adverbs, [...]. (Conlon/Evens, 1992: 1192)<br />

Hav<strong>in</strong>g presented our suggestions <strong>of</strong> how to deal <strong>with</strong> word order variation <strong>in</strong> <strong>German</strong>, we<br />

now want to list the adverb-related features we need to realise this method <strong>in</strong> Natural<br />

Language Process<strong>in</strong>g. Section 6.7.1 conta<strong>in</strong>s a list <strong>of</strong> features and their values, as well as a<br />

description <strong>of</strong> how we proceded to do the encod<strong>in</strong>g <strong>of</strong> the adverbs listed <strong>in</strong> 8.2 and 8.3.<br />

Section 6.7.2 discusses some generalizations, namely the fact that some position classes tend<br />

to have specific feature values. Based on these facts, we also suggest default values which<br />

can be used <strong>in</strong> future cod<strong>in</strong>g.<br />

Some <strong>of</strong> the features, such as the classes <strong>of</strong> elements degree modifiers can modify, do not<br />

relate directly to word order. We want to list them nevertheless, for the sake <strong>of</strong><br />

completeness, and because most degree modifiers can also modify the whole sentence. The<br />

specific degree modifier <strong>in</strong>formation is needed <strong>in</strong> order to f<strong>in</strong>d out whether it refers to the<br />

sentence or to part <strong>of</strong> it. As far as we are aware, the follow<strong>in</strong>g <strong>in</strong>formation, which can be<br />

formulated as features, is enough to deal syntactically <strong>with</strong> adverbs <strong>in</strong> Natural Language<br />

Process<strong>in</strong>g. The appendices 8.2 and 8.3 conta<strong>in</strong> four hundred one-word modifiers which are<br />

encoded us<strong>in</strong>g these features.<br />

6.7.1. CODING OF ADVERBS IN THE DICTIONARY<br />

The adverbs listed are the ones Hoberg (1981) found <strong>in</strong> her corpus, and which she classified<br />

accord<strong>in</strong>g to her list <strong>of</strong> adverbial position classes. We added Engel's (1988) lists <strong>of</strong> adverbs<br />

(749ff), modal particles (762f), order<strong>in</strong>g particles (Rangierpartikeln, 763f), degree modifiers<br />

(764ff) and toners (774f) 78 . Engel's lists are supposed to be exhaustive, but they only <strong>in</strong>clude<br />

simple, namely non-derived, modifiers. We added to this list all modifiers we came across<br />

78 Note that Engel (1988) uses his own, strictly syntactic classification which can differ from the ones used<br />

by other l<strong>in</strong>guists (cf. section 2.2).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 192<br />

while cod<strong>in</strong>g the elements <strong>of</strong> this list. However, the list<strong>in</strong>g is by far not exhaustive, as nearly<br />

all adjectives can be used adverbially. For the classification <strong>of</strong> these see section C, POS<br />

(position), below.<br />

We coded the adverbs alphabetically, as opposed to by position groups, <strong>in</strong> order to avoid<br />

be<strong>in</strong>g biased or complacent. Furthermore, we coded them feature by feature, as opposed to<br />

modifier by modifier. This means that we first checked, whether all adverbs were able to fill<br />

the Vorfeld on their own, then whether they could be compared, and so on.<br />

There were an astonish<strong>in</strong>g number <strong>of</strong> cases <strong>in</strong> which our <strong>in</strong>tuition failed to be 100% <strong>in</strong><br />

favour <strong>of</strong> one value. For <strong>in</strong>stance, it was not clear to us, whether the adverb<br />

durchweg/durchwegs 37 could be negated (+) or not (-). A feature value which turned out to be<br />

particularly difficult to judge was rheme. The question <strong>of</strong> whether a modifier such as<br />

dere<strong>in</strong>st 26 could carry the sentence focus cannot be answered objectively, at least when limited<br />

to written language. For spoken language, there might be means to measure relative loudness,<br />

or the <strong>in</strong>tensity one puts on a word. In written texts, one is reduced to one's own <strong>in</strong>tuition.<br />

For this reason, we checked the occurrences <strong>of</strong> the modifiers for which we lacked a clear<br />

<strong>in</strong>tuition <strong>in</strong> a corpus. The <strong>German</strong> corpus we used is called HK 87 (Handbuchkorpus). It has<br />

been compiled by the Institut für deutsche Sprache (IdS) <strong>in</strong> Mannheim. It consists <strong>of</strong> articles<br />

written for the newspaper Mannheimer Morgen <strong>in</strong> the year 1987, and it comprises 3 million<br />

words. This corpus turned out to be too small, as a limited number <strong>of</strong> modifiers only<br />

occurred a few times, which makes our encod<strong>in</strong>g less reliable. However, no other corpus was<br />

accessible at the time we carried out our work.<br />

We did not check all values for all modifiers <strong>in</strong> the corpus. In order to make the <strong>in</strong>formation<br />

given <strong>in</strong> our list as objective a possible, we added a star (*) to the values we have checked <strong>in</strong><br />

the corpus. A star <strong>with</strong> a plus feature is more significant than a star <strong>with</strong> a m<strong>in</strong>us feature. The<br />

reason is that the value "+*" <strong>in</strong>dicates that we have found at least one occurrence which<br />

proves that the modifier really deserves this value (e.g. <strong>in</strong>sgeheim 28 did occur at least once<br />

<strong>in</strong> the Vorfeld). If a value is accompanied by a m<strong>in</strong>us sign <strong>with</strong> a star (-*), this only means


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 193<br />

that <strong>in</strong> our corpus there was no positive evidence. However, <strong>in</strong> a bigger corpus, there might<br />

be positive evidence. Diesseits 27 , for <strong>in</strong>stance, did not occur <strong>in</strong> the Vorfeld <strong>in</strong> our corpus,<br />

and thus got the value "-*", but this may be l<strong>in</strong>ked to the fact that HK87 is too small.<br />

In a very small number <strong>of</strong> cases, our <strong>in</strong>tuition differed strongly from what we found <strong>in</strong> the<br />

corpus. An example is ausnahmsweise 14 , which occurred only 12 times <strong>in</strong> the corpus, and<br />

which did not occur once <strong>in</strong> the Vorfeld. As we firmly believe that ausnahmsweise can fill<br />

the Vorfeld, we added a "&" sign to the m<strong>in</strong>us sign (-) <strong>of</strong> VF (-&) <strong>in</strong>dicat<strong>in</strong>g that we believe<br />

the value should be plus (+).<br />

We shall now discuss the s<strong>in</strong>gle features and their values, as well as the problems we<br />

encountered while encod<strong>in</strong>g.<br />

A) Modifier:<br />

This feature designates the one-word modifier we encoded. Accord<strong>in</strong>g to the def<strong>in</strong>ition <strong>in</strong><br />

2.2.4, we <strong>in</strong>cluded elements such as beiderseits 27 , as beiderseits can occur <strong>with</strong>out its<br />

genitive complement, <strong>in</strong> spite <strong>of</strong> its classification as a preposition <strong>in</strong> Wahrig (1986: 245).<br />

About 40 modifiers <strong>in</strong> the list occurred less than 50 times. As the cod<strong>in</strong>g <strong>in</strong>formation is not<br />

really reliable for such a small amount <strong>of</strong> occurrences, we added <strong>in</strong> brackets how <strong>of</strong>ten the<br />

modifier occurred <strong>in</strong> the corpus, when the numbers were small.<br />

A lot <strong>of</strong> modifiers have homographs either among the one-word sentence modifiers, or<br />

among other word classes. In order to dist<strong>in</strong>guish them, we added a short comment<br />

<strong>in</strong>dicat<strong>in</strong>g the mean<strong>in</strong>g, such as:<br />

bloß 5 (Wunsch, Aufforderung)<br />

bloß 38 (nur, alle<strong>in</strong>)<br />

to differentiate the two mean<strong>in</strong>gs <strong>of</strong> bloß <strong>in</strong> 104a and 104b:<br />

104a Ach, käme Peter doch bloß 5 heute.<br />

104b Bloß 38 Peter kommt heute.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 194<br />

B) Who: Source <strong>of</strong> the position class<br />

The position class is the most difficult feature for which to identify the values. In order to<br />

make the value-f<strong>in</strong>d<strong>in</strong>g process more transparent, we mention <strong>in</strong> the column Who from<br />

which source we got the <strong>in</strong>formation on the position class. The adverbs and their position<br />

classification which we took over from Hoberg (1981) are <strong>in</strong>dicated by the value HO. Where<br />

possible, we directly assigned Engel's (1988: 749ff) semantic classification to Hoberg's<br />

position classes. All local adverbs (1988: 751) could for example be assigned position class<br />

27. When there was no doubt about the assignment, we gave the Who column the value EN<br />

(for Engel). However, the assignment <strong>of</strong> other semantic classes was much less obvious.<br />

Among others, this is the case for the group <strong>of</strong> temporal adverbs (Engel, 1988: 752).<br />

Temporals split <strong>in</strong>to the position classes 26, 33, 36, 37, 39 and 40. As the assignment <strong>of</strong> the<br />

modifiers to these six groups is not based on Engel's classification, the value <strong>of</strong> the Who<br />

column <strong>in</strong> such cases is RS (Ralf Ste<strong>in</strong>berger). All further modifiers added to Engel's and<br />

Hoberg's lists carry the value RS, as well.<br />

C) POS: Position class<br />

Our goal was to assign all one-word modifiers to any <strong>of</strong> the 44 position classes. We<br />

mentioned <strong>in</strong> the previous paragraph that we took over the classification <strong>of</strong> all modifiers<br />

made by Hoberg. Furthermore, we assigned the adverb groups classified semantically by<br />

Engel (1988:, 751ff) to the follow<strong>in</strong>g position classes:<br />

local adverbs a 27<br />

causal adverbs a 22<br />

conditional adverb a 19<br />

<strong>in</strong>strumental adverbs a 44<br />

f<strong>in</strong>al adverbs a 24<br />

Another group mentioned by Engel (1988:751), the adverbs express<strong>in</strong>g direction, are not<br />

classified as modifiers but as complements.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 195<br />

The cod<strong>in</strong>g was not always easy, as several <strong>of</strong> the classes are very similar. Do<strong>in</strong>g this<br />

classification is quite time-consum<strong>in</strong>g, but the effort is limited because adverbs, toners and<br />

degree modifiers are closed classes. There is only a limited amount <strong>of</strong> means to generate<br />

new adverbs (cf. Engel, 1988: 755ff). As far as we can see, adverbs which can be derived<br />

from adjectives belong to a small subset <strong>of</strong> position classes only, namely a 21<br />

(wirtschaftlich), a 33 (plötzlich), a 36 (erneut), a 37 (selten) and the manner adverbs a 43 (laut,<br />

schön, ...). a 43 is by far the biggest group 79 (Ste<strong>in</strong>berger, 1992a: 29ff). Accord<strong>in</strong>g to<br />

Schachter (1985:21), <strong>German</strong> shares the fact that manner adverbs are derivable from<br />

adjectives <strong>with</strong> a lot <strong>of</strong> other languages.<br />

The three position classes 10, 23 and 32 are not represented by one-word modifiers but by<br />

PPs or NPs only. We nevertheless did not change the number<strong>in</strong>g, as we wanted to do our<br />

classification <strong>in</strong> conformity <strong>with</strong> Hoberg and other authors who use her classification (e.g.<br />

Waltz<strong>in</strong>g, 1986).<br />

In order to f<strong>in</strong>d the position class a modifier belongs to, we did the follow<strong>in</strong>g: We narrowed<br />

down the possible position classes <strong>of</strong> an adverb by compar<strong>in</strong>g it <strong>with</strong> the description<br />

provided by Hoberg (1981: 106-131). A good share <strong>of</strong> the adverbs could be assigned a<br />

position class because <strong>of</strong> their semantics (e.g. somit expresses a cause and should thus<br />

belong to the group a 22 ). With<strong>in</strong> the diverse group <strong>of</strong> temporal modifiers, we used the<br />

restrictions mentioned by Hoberg, namely:<br />

a 26 (denotes a period <strong>of</strong> time; Zeitraum, -<strong>in</strong>tervall, -erstreckung)<br />

a 33 (evaluative temporal modifiers, denot<strong>in</strong>g a specific moment (Zeitpunkt)) precede<br />

a 36 , whereas a 40 follow a 36 .<br />

a 36 (repetition; "Wiederholung e<strong>in</strong>es Vorgangs") precede a 37<br />

a 37 (frequency (how <strong>of</strong>ten?); Häufigkeit) follow a 36<br />

a 39 (temporal-pragmatic modifiers) are <strong>of</strong>ten adjo<strong>in</strong>ed to other elements<br />

a 40 (either denot<strong>in</strong>g a period <strong>of</strong> time or a specific moment) follow a 37<br />

79 In our list it looks as if the temporal modifier group a 26 is the one <strong>with</strong> most entries. However, the reason<br />

for this is that we deliberately limited the amount <strong>of</strong> manner adverbs because they are an open class.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 196<br />

Several adverbs belong to both a 26 and a 40 and both are frequently comb<strong>in</strong>ed <strong>with</strong> other<br />

modifiers <strong>of</strong> the same type (gestern um drei Uhr).<br />

We then searched for all occurrences <strong>of</strong> the adverbs for which we could not yet identify a<br />

s<strong>in</strong>gle position class <strong>in</strong> the corpus, and made a list <strong>of</strong> all other modifiers which occurred <strong>in</strong><br />

the corpus, to the left and to the right <strong>of</strong> the unclassified adverb. We only considered adverbs<br />

<strong>in</strong> the ma<strong>in</strong> part <strong>of</strong> the sentence (Mittelfeld) which were not adjuncts. With the help <strong>of</strong> these<br />

left and right-hand side occurrences, we could f<strong>in</strong>ally choose the position class which fitted<br />

best <strong>in</strong> the overall order <strong>of</strong> the 44 position classes. The result <strong>of</strong> this procedure was not<br />

always obvious, as it happens regularly that modifiers vary from the order <strong>of</strong> the position<br />

classes. We nevertheless believe that the result is satisfy<strong>in</strong>g.<br />

Due to this procedure, we had to assign the position class a 43 (manner adverbs) to the two<br />

adverbs umsonst and vergebens, <strong>in</strong> spite <strong>of</strong> the fact that we felt that these modifiers would fit<br />

perfectly well <strong>in</strong>to the modifier group a 14 . The reason is that they follow all other modifiers<br />

<strong>of</strong> the corpus they occur <strong>with</strong>. Most modifiers, however, could be assigned more<br />

straightforwardly.<br />

D) CLASS: Modifier class<br />

As mentioned <strong>in</strong> 2.3, we dist<strong>in</strong>guish three ma<strong>in</strong> modifier classes, existimatorial (pragmatic)<br />

modifiers (a 1 -a 18 ), situatives (a 19 -a 40 ), and modal modifiers (a 42 -a 44 ). The feature<br />

modifier class is redundant, as the <strong>in</strong>formation to which class modifiers belong is expressed<br />

<strong>in</strong>directly through their position class. However, we want to mention it because three<br />

superclasses are <strong>in</strong>tuitively more accessible than Hoberg's 44 position classes, and because<br />

we want to formulate some generalizations based on them. These ma<strong>in</strong>ly concern scope,<br />

discussed under (G) below.<br />

E) VF: Ability to occur <strong>in</strong> the Vorfeld<br />

Some modifiers can occur <strong>in</strong> the Vorfeld on their own (105a), others cannot (105b, 105c).<br />

This <strong>in</strong>formation is expressed by the feature VF <strong>with</strong> its boolean values +/-. It is necessary <strong>in</strong>


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 197<br />

order to avoid the generation <strong>of</strong> <strong>in</strong>correct sentences and can be used to improve the analysis<br />

by disambiguat<strong>in</strong>g some homonyms (such as e<strong>in</strong>fach 18/43 , erst 34/39 , eigentlich 9/11 ,<br />

also 3/22 and others) if one <strong>of</strong> them can appear <strong>in</strong> the Vorfeld and the other cannot<br />

(Ste<strong>in</strong>berger, 1992a: 12f and 37).<br />

105a E<strong>in</strong>fach 43 geht das nicht! (manner)<br />

105b Er g<strong>in</strong>g e<strong>in</strong>fach 18 nicht. (pragmatic)<br />

105c * E<strong>in</strong>fach 18 g<strong>in</strong>g er nicht.<br />

F) NEG: Negability<br />

This is also a boolean feature <strong>in</strong>dicat<strong>in</strong>g whether a modifier follow<strong>in</strong>g the negator nicht can<br />

be partially negated or not. It may help to disambiguate homonyms if one <strong>of</strong> them is<br />

negatable (107), and the other is not (106) (Ste<strong>in</strong>berger, 1992b: 20f):<br />

106a Ich kann gerade 33 nicht 41 gehen. (temporal: now)<br />

106b * Ich kann nicht 41 gerade 33 gehen.<br />

107a * Ich kann gerade 43 nicht 41 gehen. (manner: straight)<br />

107b Ich kann nicht 41 gerade 43 gehen.<br />

G) SCOPE: 80<br />

Scope is to be understood here as what the (degree) modifier refers to 81 . In Ste<strong>in</strong>berger<br />

(1992b), we dist<strong>in</strong>guished seven possible classes that can be referred to: (a) modal adverbs<br />

(a 42 -a 44 ), (b) situative (a 19 -a 40 ) and (c) pragmatic adverbs (a 1 -a 18 ), (d) adjective phrases<br />

(AP), (e) NPs and PPs, (f) card<strong>in</strong>al phrases and (g) the sentence. We have added to this (h)<br />

the negation particle nicht (gar/überhaupt nicht) and conjunctions (nur wenn).<br />

For degree modifiers, this feature dist<strong>in</strong>guishes what they can modify. Rund 16 for <strong>in</strong>stance<br />

can only modify card<strong>in</strong>al numbers (rund 10 Leute, *rund heute), whereas nur 38 can modify<br />

everyth<strong>in</strong>g but pragmatic modifiers and the negation. Adverbials which are not degree<br />

modifiers (<strong>in</strong> the sense used here) only refer to the sentence. We are aware <strong>of</strong> the fact that<br />

80 The features scope and grad play only a marg<strong>in</strong>al role <strong>in</strong> the treatment <strong>of</strong> word order. For this reason, they<br />

are not expla<strong>in</strong>ed <strong>in</strong> detail. For a more detailed description, see Ste<strong>in</strong>berger (1992b).<br />

81 For the def<strong>in</strong>ition <strong>of</strong> the word class degree modifier, see 3.7.1.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 198<br />

modal adverbs, for example, rather modify the verb or the verb phrase than the sentence, but<br />

this simplification does not matter <strong>in</strong> our case 82 . Modal adverbs are coded as such (D), and<br />

so are existimatorial and situative adverbs.<br />

H) PRE/POST: Position <strong>of</strong> modifiers<br />

In addition to what a modifier can modify we have to specify whether it has to precede (pre)<br />

or follow (post) the modified phrase. Some modifiers can stand <strong>in</strong> both positions and thus<br />

get the value both.<br />

J) DIST: Distance <strong>of</strong> modifiers<br />

Furthermore, we have to specify whether the modifier has to be adjacent (-) to the modified<br />

phrase, or whether it can stand <strong>in</strong> a position distant from it (+).<br />

K) GRAD: Gradability 83<br />

Not all elements can be modified or graded. Therefore, we need a boolean feature which<br />

<strong>in</strong>dicates this. The feature scope (G) expresses, for <strong>in</strong>stance, that nur can modify situatives.<br />

With the gradability feature we have to specify that heute and dort can be modified (nur<br />

heute) whereas sogar and schon cannot (*nur sogar) (Ste<strong>in</strong>berger, 1992b: 18f).<br />

We believe that the two features, scope and gradability, are enough to f<strong>in</strong>d out whether a<br />

modifier should be seen as <strong>in</strong>dependent, or rather as modify<strong>in</strong>g, and thus dependent on<br />

another element. One could certa<strong>in</strong>ly specify <strong>in</strong> more detail that ganz, for <strong>in</strong>stance, can<br />

modify wahrsche<strong>in</strong>lich, but not leider. However, we hope that a sequence such as ganz<br />

leider will simply not occur <strong>in</strong> a text so that the <strong>in</strong>formation we give is sufficient.<br />

82 Also, this corresponds more to our non-configurational approach which is strongly <strong>in</strong>fluenced by our<br />

f<strong>in</strong>d<strong>in</strong>gs on word order, namely that word order is the result <strong>of</strong> the <strong>in</strong>teraction <strong>of</strong> preference rules rather<br />

than be<strong>in</strong>g the consequence <strong>of</strong> a strongly hierarchical constituent structure.<br />

83 For a more detailed discussion <strong>of</strong> this feature, see Ste<strong>in</strong>berger (1992b).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 199<br />

L) RHEME: Potential rhematicity<br />

This boolean feature is important to guarantee the correct recognition <strong>of</strong> the rheme <strong>of</strong> the<br />

sentence. We assume <strong>in</strong> 6.3 that the rheme is the last element <strong>in</strong> the sentence which can be<br />

rhematic. In 108, it is thus gestern 26 , whereas <strong>in</strong> 109 it is sang, as bereits 39 is not a<br />

potential rheme:<br />

108 Er sang GEStern 26 .<br />

109a * Er sang BeREITS 39 .<br />

109b Er SANG bereits 39 .<br />

Whether an element can be rhematic or not can be found out by construct<strong>in</strong>g short sentences<br />

<strong>with</strong> a structure such as <strong>in</strong> 108, <strong>in</strong> which the last element should be stressed. If the verb <strong>in</strong><br />

verb-second position has to be stressed, the modifier is not a potential rheme.<br />

However, sometimes the decision is not as obvious as <strong>in</strong> 108 and 109. For <strong>in</strong>stance, it was<br />

less clear to us whether the adverb unverzüglich 33 is a potential rheme. In all cases <strong>in</strong> which<br />

we did not have a clear-cut <strong>in</strong>tuition, we searched the corpus for occurrences <strong>of</strong> the modifier<br />

<strong>in</strong> question, and checked where we would put the sentence accent when read<strong>in</strong>g the<br />

sentences <strong>in</strong> the given contexts. If <strong>in</strong> any <strong>of</strong> them, the modifier carried the sentence accent,<br />

we gave it the value "+*", and otherwise "-*".<br />

M) Valency:<br />

Not a lot <strong>of</strong> modifiers take arguments but some <strong>of</strong> them do. They are never obligatory.<br />

Entsprechend 31 , for <strong>in</strong>stance, can be accompanied by a genitive NP (110). We have shown<br />

<strong>in</strong> 4.4.1 that it is important to recognise syntactic subord<strong>in</strong>ation such as <strong>in</strong> 110:<br />

110 Er verhielt sich entsprechend 31 der Abmachung.<br />

N) PRED: Predicative use<br />

Some adverbs can be used predicatively (111a), others cannot (111b). This feature helps to<br />

dist<strong>in</strong>guish adverbial homonyms such as so 22/43 , eher 6/26/40 , gleich 18/33 and others:


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 200<br />

111a Harold war so 43 . (pr<strong>of</strong>orm for manner adverbs)<br />

111b * Harold war so 22 . (cause: somit)<br />

O) COMP: Comparability<br />

Another feature which does not directly have to do <strong>with</strong> word order is the one <strong>in</strong>dicat<strong>in</strong>g<br />

comparability. This boolean feature <strong>in</strong>dicates whether a modifier can be compared or not. If<br />

it can, a set <strong>of</strong> further features is necessary to generate the correct comparison and<br />

superlative forms morphologically (cf. Zähner/Gupta/Ste<strong>in</strong>berger, forthcom<strong>in</strong>g).<br />

N<strong>in</strong>e <strong>of</strong> these features are necessary to cope <strong>with</strong> word order variation, and the related<br />

degree modification, <strong>in</strong> Natural Language Process<strong>in</strong>g (C, D, E, F, G, H, I, K and L). In<br />

addition to these, one may want to express the compatibility <strong>of</strong> modifiers <strong>with</strong> each other.<br />

Work on this subject has been carried out by Ste<strong>in</strong>itz (1963) and by Bartsch (1972). Such<br />

<strong>in</strong>formation may be necessary to solve some cases <strong>of</strong> ambiguity caused by homonyms, if the<br />

syntactic features suggested <strong>in</strong> this section fail to provide help. Another use <strong>of</strong> cooccurrence<br />

restrictions could be made <strong>in</strong> <strong>German</strong> sentence generation, <strong>in</strong> order to facilitate the choice<br />

among several lexemes.<br />

Semanticists may want to add further features, such as the <strong>in</strong>fluence <strong>of</strong> adverbs on time and<br />

aspect. The <strong>German</strong> adverb gerade, for <strong>in</strong>stance, bears aspect <strong>in</strong>formation which <strong>in</strong> English<br />

has to be rendered morphologically (112). And temporal expressions can be <strong>of</strong> importance to<br />

decide on the tense <strong>of</strong> the translated sentence (113):<br />

112a Juan las gerade e<strong>in</strong> Buch.<br />

112b Juan was read<strong>in</strong>g a book.<br />

112c * Juan read a book.<br />

113a Archana steht morgen um acht Uhr auf.<br />

113b * Tomorrow, Archana gets up at eight o'clock.<br />

113c Tomorrow, Archana will get up at eight o'clock.<br />

Although the number <strong>of</strong> modifier-related features probably has to be extended for these<br />

tasks, we believe that the means <strong>of</strong>fered <strong>in</strong> this section are sufficient to treat word order<br />

problems satisfactorily.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 201<br />

6.7.2. SOME GENERALIZATIONS<br />

Once the cod<strong>in</strong>g work <strong>in</strong> alphabetical order was f<strong>in</strong>ished, we ordered the entries accord<strong>in</strong>g to<br />

their position classes to see whether the position classes shared the same feature values.<br />

Furthermore, we wanted to verify some <strong>in</strong>tuitive assumptions we had, such as the one that<br />

manner adverbs are all potential rhemes, and that most pragmatic modifiers are not, among<br />

others.<br />

We found out that position classes do not share all feature values but that, <strong>in</strong> the position<br />

classes a 1 to a 25 , there is a strong tendency to share values, <strong>with</strong> only a few exceptions. It<br />

seems that the higher the position classes, the less homogenous their values.<br />

It is particularly surpris<strong>in</strong>g that the 45 adverbs <strong>of</strong> the position classes a 1 to a 7 have the same<br />

values for the features VF, KOMP, NEG, MOD, RHEME and PRED, <strong>with</strong> the exception <strong>of</strong><br />

one s<strong>in</strong>gle feature each for two adverbs: eher 6 (rather) can be modified (viel eher), and<br />

weiterh<strong>in</strong> 7 is a potential rheme. Some more exceptions can be seen <strong>in</strong> the table below.<br />

In the follow<strong>in</strong>g, we shall list the position classes which are consistent for some features.<br />

This <strong>in</strong>formation could be <strong>in</strong>terest<strong>in</strong>g for further research, for <strong>in</strong>stance <strong>in</strong>to why different<br />

groups are so consistent, and what they have <strong>in</strong> common <strong>with</strong> other classes which share the<br />

same feature values. Another <strong>in</strong>terest<strong>in</strong>g question is whether modifiers share the same<br />

feature values <strong>with</strong> their equivalents <strong>in</strong> other languages. Generalizations can furthermore be<br />

<strong>of</strong> importance for future cod<strong>in</strong>g work as both, consistencies and strong tendencies, can be<br />

used to formulate default values.<br />

The list below shows all position classes, and the values they share for different features. It<br />

also conta<strong>in</strong>s the classes which have a strong tendency towards a certa<strong>in</strong> value. For the latter,<br />

we add the <strong>in</strong>formation on how many exceptions there are. "-1/12" thus means, for <strong>in</strong>stance,<br />

that the class comprises 12 elements which generally have a negative value, and that there is<br />

only one exception <strong>with</strong> a positive value.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 202<br />

Class members VF COMP NEG GRAD RHEME PRED<br />

1 2 - - - - - -<br />

2 4 - - - - - -<br />

3 2 - - - - - -<br />

4 1 - - - - - -<br />

5 7 - - - - - -<br />

6 12 + - - - 1/12 - -<br />

7 18 + - - - - 1/18 -<br />

8 10 - - 1/10 - 1/10<br />

9 6 - - - -<br />

10 -<br />

11 11 - - 1/11 - 1/11 - 1/11<br />

12 22 + 1/22 - -<br />

13 7 + - - - - -<br />

14 13 -<br />

15 2 - - + +<br />

16 7 - - - - -<br />

17 4 - - - - - -<br />

18 9 - - 1/9 -<br />

19 1 + - - + - -<br />

20 2 + - - - + -<br />

21 3 + - + - -<br />

22 21 + -<br />

23 -<br />

24 3 + - + +<br />

25 2 + - + + -<br />

26 46 + 3/46 - 2/46<br />

26/40 18 + - 1/18 + 1/18 + 1/18<br />

27 43 + 3/43 -<br />

28 1 + - - - - -<br />

29 2 + - + +<br />

30 2 + - + +<br />

31 1 + - + - + +<br />

32 -<br />

33 18 - 2/18 -<br />

34 7 - - 1/7 -<br />

35 5 -<br />

36 5 + - +<br />

37 29 +<br />

38 5 - - -<br />

39 5 -<br />

40 6 +<br />

41 2 + - + + + -<br />

42 1 + - + + + -<br />

43 33+<br />

44 1 + - + + - +


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 203<br />

6.8. SUMMARY OF CHAPTER 6<br />

We dedicated chapter 6 to a discussion <strong>of</strong> the consequences that our research could have for<br />

Natural Language Process<strong>in</strong>g. We started by giv<strong>in</strong>g a list <strong>of</strong> compulsory orders which apply<br />

<strong>in</strong> the free word order language <strong>German</strong> (6.1). Their value cannot be underestimated as they<br />

are part <strong>of</strong> the small number <strong>of</strong> hard and reliable word order rules we can work <strong>with</strong> <strong>in</strong><br />

sentence analysis.<br />

In the next two sections (6.2 and 6.3) we suggested means <strong>of</strong> automatically recognis<strong>in</strong>g<br />

theme, rheme and focus <strong>in</strong> a <strong>German</strong> sentence. We understand focus as be<strong>in</strong>g an element<br />

which must be very strongly, or contrastively, stressed. Although the procedure <strong>of</strong><br />

recognis<strong>in</strong>g thematic, rhematic and focused phrases is not completely safe, we believe that it<br />

provides us <strong>with</strong> correct results <strong>in</strong> most cases.<br />

Section 6.4 conta<strong>in</strong>s some f<strong>in</strong>d<strong>in</strong>gs which are not immediately relevant for the treatment <strong>of</strong><br />

word order <strong>in</strong> Natural Language Process<strong>in</strong>g, but which help to understand <strong>German</strong> word<br />

order better. They concern the positional behaviour <strong>of</strong> several modifiers relative to each<br />

other. The regularities concern<strong>in</strong>g situative modifiers (6.4.4) proved to be more complex<br />

than the ones concern<strong>in</strong>g other groups. When situative modifiers permute, this can either not<br />

have any effect at all, or it can result <strong>in</strong> focalization <strong>of</strong> the latter, or it can cause contrastive<br />

stress, or it can even lead to ungrammaticality, depend<strong>in</strong>g on the modifiers <strong>in</strong>volved.<br />

Although we could not expla<strong>in</strong> these different results satisfactorily, from a l<strong>in</strong>guistic po<strong>in</strong>t <strong>of</strong><br />

view, we believe that they do not represent a problem for Mach<strong>in</strong>e Translation. The reason is<br />

that, <strong>in</strong> sentence generation, we can simply order the modifiers accord<strong>in</strong>g to the canonical<br />

form. When analys<strong>in</strong>g a <strong>German</strong> sentence, we can assume that the <strong>in</strong>put is correct, and use<br />

all the means described <strong>in</strong> this chapter to extract theme, rheme and focus, <strong>in</strong> order to<br />

guarantee a correct translation.<br />

Section 6.5 is dedicated to the elaboration <strong>of</strong> the f<strong>in</strong>al version <strong>of</strong> the canonical form. The<br />

changes <strong>with</strong> respect to the one suggested <strong>in</strong> 5.3.3 ma<strong>in</strong>ly concern the <strong>in</strong>corporation <strong>of</strong> the


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 204<br />

flexible categories theme, rheme and focus. Complements and modifiers which have been<br />

identified as be<strong>in</strong>g thematic can take the position <strong>of</strong> theme, and the same applies for rheme<br />

and focus constituents. When <strong>German</strong> sentences are generated <strong>with</strong> the help <strong>of</strong> this canonical<br />

form, their word order differs depend<strong>in</strong>g on the analysis <strong>of</strong> the source sentences.<br />

One problem encountered <strong>in</strong> the earlier canonical form presented <strong>in</strong> 5.3.3 was that we could<br />

not identify a specific place for modifiers, because <strong>of</strong> their vary<strong>in</strong>g positional behaviour.<br />

Section 6.5 showed that this problem was due to the fact that modifier position varies<br />

considerably depend<strong>in</strong>g on the theme-rheme structure. Thanks to the <strong>in</strong>sertion <strong>of</strong> the three<br />

new categories <strong>in</strong> the f<strong>in</strong>al version <strong>of</strong> the canonical form, we could now identify a specific<br />

position for each neutral modifier subtype.<br />

In section 6.6, we suggested a preference for PP attachment resolution, which is based on the<br />

fact that focus<strong>in</strong>g word order is marked. Whenever constructions lead<strong>in</strong>g to contrastive<br />

stress can be avoided by analys<strong>in</strong>g the PP as an adjunct, this should be done.<br />

In 6.7, f<strong>in</strong>ally, we gave a list <strong>of</strong> the modifier-related features we need <strong>in</strong> order to deal <strong>with</strong><br />

word order variation. We hope that the <strong>in</strong>formation they conta<strong>in</strong> is sufficient for the<br />

syntactic treatment <strong>of</strong> adverbs <strong>in</strong> Natural Language Process<strong>in</strong>g.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 205<br />

7. CONCLUDING REMARKS<br />

To conclude, we want to summarise the contents briefly 84 , evaluate the results <strong>of</strong> our work,<br />

and suggest areas <strong>of</strong> future research, which could make use <strong>of</strong> the outcome <strong>of</strong> this thesis.<br />

We have discussed the complexity <strong>of</strong> the mechanism <strong>of</strong> <strong>German</strong> word order from the<br />

l<strong>in</strong>guistic po<strong>in</strong>t <strong>of</strong> view, and we made suggestions <strong>of</strong> how to cope <strong>with</strong> this complexity <strong>in</strong><br />

Mach<strong>in</strong>e Translation. We paid special attention to the treatment <strong>of</strong> modifiers, as most word<br />

order <strong>in</strong>formation <strong>in</strong> l<strong>in</strong>guistic literature concerns verb complements.<br />

We started the work by giv<strong>in</strong>g the motivation for word order description (1.5), and by<br />

def<strong>in</strong><strong>in</strong>g the relevant term<strong>in</strong>ology (2). The def<strong>in</strong>itions <strong>of</strong> the word class adverb (2.2) found <strong>in</strong><br />

grammar books turned out to be contradictory. Chapter 3 is dedicated to the factors which<br />

determ<strong>in</strong>e <strong>German</strong> word order. We could identify eleven factors, <strong>in</strong>clud<strong>in</strong>g the theme-rheme<br />

structure (3.1), functional sentence perspective (3.3), and the scope <strong>of</strong> scope-<strong>in</strong>clud<strong>in</strong>g<br />

elements (3.7). All <strong>of</strong> these factors can be seen as preferences, as opposed to clear-cut<br />

order<strong>in</strong>g rules (4). It cannot be said, for <strong>in</strong>stance, that thematic phrases always precede<br />

rhematic ones, but it can only be formulated that themes tend to precede rhemes. Depend<strong>in</strong>g<br />

on the parameters set for each factor <strong>in</strong> a specific sentence, the different factors <strong>of</strong>ten prefer<br />

different phrase sequences. It can therefore occur that, <strong>in</strong> a given sentence, a phrase referr<strong>in</strong>g<br />

to the AGENT <strong>of</strong> an action tends to precede the EXPERIENCER <strong>of</strong> the action (3.6), but that<br />

the theme-rheme structure prefers the opposite order. In chapter 4, we described this<br />

<strong>in</strong>teraction <strong>of</strong> order<strong>in</strong>g factors, and gave evidence for the fact that some preferences are<br />

stronger than others. It was also po<strong>in</strong>ted out that this mechanism <strong>of</strong> weigh<strong>in</strong>g several factors<br />

is restricted by possessive relations, the <strong>in</strong>volvement <strong>of</strong> quantificational elements, amongst<br />

others (4.4).<br />

84 For a more detailed summary, see section 1.3.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 206<br />

On the basis <strong>of</strong> this data, we discussed <strong>in</strong> chapter 5 recent suggestions <strong>of</strong> how to deal <strong>with</strong><br />

free word order <strong>in</strong> Natural Language Process<strong>in</strong>g (5.1). As no proposition seems to be<br />

satisfy<strong>in</strong>g, we suggest another method, namely the use <strong>of</strong> a flexible canonical form. As it is<br />

not evident how appropriate a canonical form is for a free word order language, we discuss<br />

this question <strong>in</strong> 5.2. The idea is that a canonical form expresses some <strong>of</strong> the preferences<br />

implicitly. The requirement <strong>of</strong> the theme-rheme structure, for <strong>in</strong>stance, that def<strong>in</strong>ite NPs and<br />

PPs should precede <strong>in</strong>def<strong>in</strong>ite ones, can be satisfied by formulat<strong>in</strong>g the canonical form<br />

accord<strong>in</strong>gly. A further essential component is that a canonical form can be made flexible, by<br />

<strong>in</strong>clud<strong>in</strong>g categories such as theme, rheme and (contrastive) focus. These position categories<br />

can <strong>in</strong>clude unlike elements such as complements and most modifier types (5.6 and 6.5).<br />

Depend<strong>in</strong>g on which elements have been recognised as theme, rheme and focus, this flexible<br />

canonical form produces sentences which are naturally embedded <strong>in</strong> their context. In order<br />

to make the best use <strong>of</strong> such a canonical form, the relevant categories should be identified<br />

dur<strong>in</strong>g analysis. In sections 6.2, 6.3, and 6.7 we suggest the means for recognis<strong>in</strong>g theme,<br />

rheme and focus <strong>in</strong> <strong>German</strong>, this be<strong>in</strong>g the source language.<br />

Free word order poses major problems for both Natural Language Process<strong>in</strong>g and language<br />

learn<strong>in</strong>g, especially <strong>with</strong> respect to the neglected field <strong>of</strong> modifier position<strong>in</strong>g. We believe<br />

that the suggested procedure us<strong>in</strong>g the canonical form is a good means <strong>of</strong> cop<strong>in</strong>g <strong>with</strong> word<br />

order variation <strong>in</strong> <strong>German</strong>.<br />

However, several problems are l<strong>in</strong>ked to our suggestion, and we are still far away from<br />

hav<strong>in</strong>g solved all difficulties. An unsatisfy<strong>in</strong>g aspect is that the data we used to develop the<br />

canonical form is largely based on our own <strong>in</strong>tuition. However, we could take over Hoberg's<br />

(1981) f<strong>in</strong>d<strong>in</strong>gs on modifier classes, which are based on the Mannheimer Duden corpus.<br />

Furthermore, we did a survey on people's preferences regard<strong>in</strong>g word order permutation<br />

(Ste<strong>in</strong>berger, 1990), and used a <strong>German</strong> three million word corpus (6.7) to decide on the<br />

feature values for the adverbs listed <strong>in</strong> the appendix (8.2 and 8.3).


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 207<br />

If a lot <strong>of</strong> our work had to rely on l<strong>in</strong>guistic <strong>in</strong>tuition, this is because, as far as we are aware,<br />

there is no tagged corpus which would comprise <strong>of</strong> tags for thematic, rhematic and focused<br />

phrases, or for the sophisticated modifier classes we are us<strong>in</strong>g. An automatic assignment <strong>of</strong><br />

these categories by us<strong>in</strong>g our own f<strong>in</strong>d<strong>in</strong>gs, if possible, would lead to circular argumentation.<br />

If one wanted to test the newly developed canonical form empirically, one would thus have<br />

to put a lot <strong>of</strong> hard work <strong>in</strong>to hand-tagg<strong>in</strong>g a large corpus. An alternative possibility would<br />

be to apply the canonical form to a Mach<strong>in</strong>e Translation system and to evaluate the results it<br />

produces <strong>in</strong> real-life applications.<br />

Another weak po<strong>in</strong>t <strong>of</strong> this work is the dist<strong>in</strong>ction between rheme and focus, which cannot<br />

be checked automatically either. Everybody will agree that there is a difference between the<br />

stresses <strong>in</strong> examples <strong>in</strong>volv<strong>in</strong>g contrastive focus and others <strong>in</strong>volv<strong>in</strong>g a simple rhematic<br />

element, but <strong>of</strong>ten the borderl<strong>in</strong>e between the two is unclear. We cannot th<strong>in</strong>k <strong>of</strong> any means<br />

to verify this dist<strong>in</strong>ction empirically for written texts.<br />

The recognition <strong>of</strong> the categories theme, rheme and focus seems to work well for <strong>German</strong>. In<br />

our research, we took <strong>German</strong> as both a source and target language. In a Mach<strong>in</strong>e<br />

Translation system, however, we will be confronted <strong>with</strong> differences between languages<br />

regard<strong>in</strong>g their means to express thematicity and functional sentence perspective. English<br />

and French, for <strong>in</strong>stance, express theme and rheme far less than <strong>German</strong>, so that the<br />

<strong>in</strong>formation gathered dur<strong>in</strong>g the analysis <strong>of</strong> any <strong>of</strong> these languages will not be enough to<br />

make full use <strong>of</strong> the sophisticated canonical form for <strong>German</strong>. On the other hand, when<br />

<strong>German</strong> is the source language, a lot <strong>of</strong> <strong>in</strong>formation on theme and rheme will be found, but<br />

English and French cannot make use <strong>of</strong> all <strong>of</strong> it. This is not a real problem, as the canonical<br />

form can be used perfectly well <strong>with</strong>out the categories theme, rheme and focus. Any further<br />

<strong>in</strong>formation on thematicity will be an advantage. The best use <strong>of</strong> it, however, could probably<br />

be made when translat<strong>in</strong>g from another free word order language, such as Czech, <strong>in</strong>to <strong>German</strong>,<br />

or vice versa.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 208<br />

<strong>German</strong> analysis should depend on the language it is compared <strong>with</strong>. Focus<strong>in</strong>g, for <strong>in</strong>stance,<br />

may or may not be expressed by special means <strong>in</strong> the other language. In 1, we have to<br />

choose between render<strong>in</strong>g the very strong focus<strong>in</strong>g <strong>of</strong> the <strong>German</strong> sentence and the stylistic<br />

heav<strong>in</strong>ess <strong>of</strong> the equivalent English cleft<strong>in</strong>g construction. However, as the word order <strong>in</strong> 1a<br />

is very strongly marked, 1c is a more appropriate translation than 1b. Accord<strong>in</strong>g to the<br />

Coll<strong>in</strong>s <strong>German</strong> dictionary (1981: 582), the degree modifier schon <strong>in</strong> sentences such as 2a<br />

should disappear when translat<strong>in</strong>g <strong>in</strong>to English. In the <strong>German</strong> sentence 2a, schon is<br />

necessary to underl<strong>in</strong>e the focus on gestern. For a <strong>German</strong> speaker's <strong>in</strong>tuition, 2b seems to be<br />

an <strong>in</strong>approprate translation <strong>of</strong> 2a, as both an equivalent <strong>of</strong> schon and the forced stress on<br />

yesterday are miss<strong>in</strong>g. 2 shows to what extent <strong>German</strong> and English differ <strong>with</strong> respect to<br />

express<strong>in</strong>g differences <strong>in</strong> functional sentence perspective.<br />

1a<br />

1b<br />

1c<br />

2a<br />

2b<br />

2c<br />

Er hat gestern IHN gesehen.<br />

?? He saw him yesterday.<br />

It was HIM that he saw yesterday.<br />

Er sprach mit der Frau GEStern schon.<br />

He talked to the woman yesterday.<br />

?? It was yesterday that he talked to the woman.<br />

1 and 2 show that our dist<strong>in</strong>ctions and suggestions should be reviewed accord<strong>in</strong>g to the<br />

language <strong>in</strong>to or from which it is translated. It did nevertheless make sense to analyse<br />

<strong>German</strong> <strong>with</strong>out referr<strong>in</strong>g to another language, as our <strong>in</strong>tention was to show the pr<strong>in</strong>ciple.<br />

We provided all the means and, depend<strong>in</strong>g on the other languages, either all <strong>of</strong> them or an<br />

appropriate subset can now be chosen.<br />

Another problem is that our canonical form is only an approximation. We found example<br />

pairs which showed that no canonical form generates the correct order <strong>of</strong> elements for all<br />

sentences. On whatever order <strong>of</strong> categories <strong>in</strong> the canonical form we decide, some sentences<br />

will be generated <strong>in</strong>correctly. This <strong>in</strong>dicates either that these examples are highly<br />

idiosyncratic, or that we still have not found all dist<strong>in</strong>ctive features necessary for an<br />

automatic treatment <strong>of</strong> word order. However, we hope and believe that most sentences can<br />

be generated correctly.


Ralf Ste<strong>in</strong>berger – <strong>Word</strong> <strong>Order</strong> <strong>Variation</strong> <strong>in</strong> <strong>German</strong>, Modifiers (Ph.D. Thesis, UMIST, 1994) 209<br />

In an evaluation <strong>of</strong> the suggested method, it should also be mentioned that it can only treat<br />

adverbs and that we had to leave aside the bigger group <strong>of</strong> prepositional phrases. PPs still<br />

cannot be disambiguated and classified automatically. However, the existence <strong>of</strong> a sound<br />

classification system based on the closed word class <strong>of</strong> adverbs, comb<strong>in</strong>ed <strong>with</strong> our<br />

explanatory f<strong>in</strong>d<strong>in</strong>gs on word order variation, might be a first step <strong>in</strong> the direction <strong>of</strong> solv<strong>in</strong>g<br />

this problem. Interest<strong>in</strong>g research that could be carried out <strong>in</strong> the future would be, for<br />

<strong>in</strong>stance, to categorise the PPs occurr<strong>in</strong>g <strong>in</strong> a corpus accord<strong>in</strong>g to the classification suggested<br />

<strong>in</strong> this thesis. Are there any regularities which can be used for an automatic assignment <strong>of</strong><br />

PPs to the modifier classes? Are certa<strong>in</strong> prepositions and nouns l<strong>in</strong>ked to specific modifier<br />

classes or to the feature values shown <strong>in</strong> the appendix (8.2 and 8.3)? And is there a relation<br />

between the modifier position classes and the def<strong>in</strong>iteness <strong>of</strong> NPs or PPs? The answers to<br />

these and other questions would certa<strong>in</strong>ly help <strong>in</strong> progress concern<strong>in</strong>g the automatic<br />

treatment <strong>of</strong> prepositional phrases. Another future application based on the f<strong>in</strong>d<strong>in</strong>gs <strong>of</strong> this<br />

thesis would be the formulation <strong>of</strong> further preferences such as the one presented <strong>in</strong> 6.6.<br />

We implemented a slightly simpler version <strong>of</strong> our model <strong>with</strong><strong>in</strong> the CAT2 Mach<strong>in</strong>e<br />

Translation system, and for the several hundred test sentences, it seemed satisfactory. The<br />

next step would now be to apply this method to a commercial system on a larger scale, <strong>in</strong><br />

order to f<strong>in</strong>d out its strength as well as its limits.


210<br />

8. APPENDIX<br />

8.1. ANGABESTELLUNGSKLASSEN (ACCORDING TO HOBERG,<br />

1981: 106-131)<br />

a 1<br />

denn, ja (meist <strong>in</strong> direkten oder <strong>in</strong>direkten Fragesätzen)<br />

a 2<br />

da, dann, nun, jetzt (nicht temporal)<br />

a 3<br />

aber (nicht nur adversativ),<br />

also (nicht kausal)<br />

a 4<br />

doch (nicht adversativ oder konzessiv, sondern bekräftigend)<br />

a 5<br />

eben/halt (nicht temporal)<br />

nämlich<br />

nicht, etwa (<strong>in</strong> Fragesätzen)<br />

nur, bloß (Wunsch oder Aufforderung)<br />

(pragmatischer Charakter des Satzes)<br />

a 6<br />

jedoch, <strong>in</strong>dessen, allerd<strong>in</strong>gs, nur (=allerd<strong>in</strong>gs), freilich, zwar, wohl (=zwar),<br />

vielmehr, dagegen, übrigens, jedenfalls ...<br />

(<strong>of</strong>t Gegensatz, E<strong>in</strong>schränkung, Korrektur e<strong>in</strong>es anderen Satzes)<br />

a 7<br />

erstens, zweitens ..., ferner, weiter(h<strong>in</strong>), außerdem, auch (=außerdem), sonst<br />

(=außerdem), zudem, überdies, obendre<strong>in</strong>, daneben, zugleich, endlich/schließlich<br />

(nicht temporal), e<strong>in</strong>erseits, andererseits<br />

(zur Gliederung des Sprechaktes)<br />

a 8<br />

beispielsweise, z.B., etwa (=z.B.), u.a.<br />

<strong>in</strong> erster L<strong>in</strong>ie, primär, sekundär, vor allem, besonders, <strong>in</strong>sbesondere, nicht zuletzt,<br />

vorwiegend, zumal, eben, gerade<br />

(Ordnungsfunktion)<br />

a 9<br />

sowieso, ohneh<strong>in</strong>, ohnedies, überhaupt, eigentlich (=überhaupt), immerh<strong>in</strong><br />

(= wie dem auch sei, Unabänderlichkeit)


211<br />

a 10<br />

laut Protokoll, nach Chomsky<br />

(referentiell, mit existimatorisch-pragmatischem Charakter)<br />

a 11<br />

an sich, im Grunde genommen, eigentlich (=im Grunde), genau/streng genommen,<br />

im Pr<strong>in</strong>zip, pr<strong>in</strong>zipiell, letztlich, letzten Endes, theoretisch, praktisch,<br />

gewissermaßen, sozusagen, gleichsam, im allgeme<strong>in</strong>en, im großen und ganzen,<br />

im wesentlichen, z,T., teilweise, mehr oder weniger/m<strong>in</strong>der, fast, be<strong>in</strong>ahe<br />

(betreffen die Gültigkeit der Aussage, relativierend)<br />

a 12<br />

eventuell, vielleicht, womöglich, möglicherweise, wohl, vermutlich, voraussichtlich,<br />

wahrsche<strong>in</strong>lich, sicher(lich), gewiß, bestimmt, zweifellos, ohne Zweifel, ...<br />

natürlich, selbstverständlich, bekanntlich, angeblich, ansche<strong>in</strong>end, sche<strong>in</strong>bar,<br />

<strong>of</strong>fenbar, <strong>of</strong>fensichtlich, augensche<strong>in</strong>lich ...<br />

(modifizieren den assertorischen Modus / Evidenz wird bekräftigt oder<br />

abgeschwächt)<br />

a 13<br />

h<strong>of</strong>fentlich, erfreulicherweise, Gottseidank, zum Glück<br />

unglücklicherweise, bedauerlicherweise, leider, ...<br />

(erwünscht oder unerwünscht)<br />

a 14<br />

dummerweise, klugerweise, merkwürdigerweise, eigentuümlicherweise,<br />

billigerweise, zu (Un-)Recht, unmöglich, besser, ...<br />

normalerweise, irrtümlicherweise, ausnahmsweise, notwendigerweise,<br />

zwangsläufig, ...<br />

(kommentieren stark subjektiv wertend)<br />

a 15<br />

<strong>in</strong> Wirklichkeit, <strong>in</strong> Wahrheit, wirklich, tatsächlich, <strong>in</strong> der Tat<br />

(nicht performatorisch, sondern beziehen sich auf das Tatsache-se<strong>in</strong>)<br />

a 16<br />

wenigstens, zum<strong>in</strong>dest, höchstens, allenfalls, bestenfalls<br />

(Maß-Adverbiale, Vergleich / <strong>of</strong>t adjungiert)<br />

a 17 (so)gar, selbst (=sogar) (ähnlich a 18 )<br />

a 18 ruhig, getrost, e<strong>in</strong>fach, geradezu, leicht, regelrecht, ...<br />

(existimatorischer Charakter)


212<br />

a 19<br />

dann, <strong>in</strong> diesem Fall, unter der Bed<strong>in</strong>gung/Voraussetzung, bei diesen Verhältnissen)<br />

(konditional)<br />

a 20<br />

trotzdem, dennoch, ... (konzessiv)<br />

a 21 <strong>in</strong> dieser H<strong>in</strong>sicht, rechtlich, wirtschaftlich, -mäßig, ...<br />

(Pr<strong>of</strong>orm: <strong>in</strong> dieser H<strong>in</strong>sicht, limitieren die Prädikation auf e<strong>in</strong>en bestimmten<br />

Aspekt)<br />

a 22<br />

deshalb, daher, (al)so, aus diesem Grunde, <strong>in</strong>folge von, (kausal)<br />

a 23<br />

mit der Folge, daß (konsekutiv)<br />

a 24<br />

zu diesem Zweck, mit diesem Ziel, ... (f<strong>in</strong>al)<br />

a 25<br />

mit diesen Mitteln, damit, auf diesem Wege, durch solche Maßnahmen,<br />

mithilfe dieses Verfahrens (medial)<br />

a 26 gestern, nach dem Essen, 1980, jetzt, damals, später, mittlerweile, ...<br />

(Zeitraum, <strong>in</strong> dem sich der Sachverhalt abspielt)<br />

a 27<br />

hier, auf der Straße, <strong>in</strong> Bonn<br />

(lokal, Pr<strong>of</strong>orm: dort) "Quasi-Lokale", (echte Lokale s<strong>in</strong>d Ergänzungen)<br />

a 28<br />

bei gutem Wetter, mit zweistündiger Verspätung, unter großem Beifall<br />

(Begleitumstände, Pr<strong>of</strong>orm: wobei-Satz)<br />

a 29<br />

bei der Feier, auf der Tagung, anläßlich ... (Anlaß, Gelegenheit, Pr<strong>of</strong>orm: dabei)<br />

a 30<br />

hier, <strong>in</strong> der Kunst, auf diesem Gebiet, bei Goethe, für mich, bezülich, betreffs<br />

(limitierende Funktion, lokal oder abstrakt-geistig)<br />

a 31<br />

entgegen allen Voraussagen, im Verhältnis/Vergleich zu früher,<br />

entsprechend/gemäß der Vere<strong>in</strong>barung<br />

(Gegensatz oder Entsprechung e<strong>in</strong>es Sachverhaltes im Vergleich)<br />

a 32<br />

von der Polizei, durch die Amerikaner<br />

((Personaler) Urheber e<strong>in</strong>es Geschehens, Agensangabe im Passivsatz)


213<br />

a 33<br />

plötzlich, auf e<strong>in</strong>mal, endlich, schließlich, bald, s<strong>of</strong>ort, (so)gleich, unverzüglich,<br />

gerade, (so)eben, ... (valuativ-temporal, geben Zeitpunkt an)<br />

a 34<br />

doch, durchaus, (sehr) wohl, schon/noch/erst (nicht temporal).<br />

(ca. pragmatisch, sprecherbezogen)<br />

a 35<br />

auch, ebenfalls, gleichfalls.<br />

a 36 wieder(um), erneut, noch e<strong>in</strong>mal, nochmals, abermals, immer wieder, ...<br />

(Wiederholung, iterativ)<br />

a 37<br />

e<strong>in</strong> paarmal, jedesmal, selten, zeitweise, vielfach, wiederholt, häufig, <strong>of</strong>t,<br />

regelmäßig, stets, immer(zu)... (Frequenz)<br />

a 38<br />

nur, bloß, lediglich, alle<strong>in</strong>, e<strong>in</strong>zig. (limitierend)<br />

a 39<br />

schon, bereits, noch, immer noch, erst.<br />

(temporal-pragmatisch, schon: (Erwartung des Sprechers)<br />

a 40 3 Wochen, den ganzen Tag, von 8-10 Uhr, lange, am Montag, um 9 Uhr, ...<br />

(zeitliche Dauer, Zeitpunkt)<br />

a 41<br />

nicht, gar nicht, überhaupt nicht, nicht mehr, nicht etwa, nicht e<strong>in</strong>mal, ke<strong>in</strong>eswegs,<br />

<strong>in</strong> ke<strong>in</strong>er Weise, kaum, nie(mals), ...<br />

(Negationsangaben, auch graduierend (ganz und gar nicht, kaum))<br />

a 42 mit den K<strong>in</strong>dern, <strong>in</strong> Begleitung e<strong>in</strong>es Erwachsenen, mite<strong>in</strong>ander, ...<br />

(konkomitant, begleitend-kooperational)<br />

a 43<br />

so, schnell, laut, mit großem Eifer, gut, gern, auf diese Weise, sehr, wenig, völlig,<br />

ganz, e<strong>in</strong> bißchen, ...<br />

modifizieren, graduieren, quantifizieren<br />

(Modificativa (unter Ausschluß der attributiv verwendeten Modalangaben)),<br />

Pr<strong>of</strong>ormen: so/<strong>in</strong> diesem Maße<br />

a 44 mit dem Messer, mit Hilfe e<strong>in</strong>es Bohrers, ...<br />

(Instrumentalangaben) (konkreter als Mediale)


8.2. ALPHABETICAL LISTING OF MODIFIERS<br />

Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

1980 26 sit HO + - + + + + s - - -<br />

abends 26,40 sit EN +* - + + + + s - - -<br />

aber (nicht nur adversativ) 03 pragm HO - - - - - - s,man,sit,npp,ap post - -<br />

abermals 36 sit HO + - - - + - s - - -<br />

abseits (17; 5 ohne GEN) 27 sit RS +*# -* -*& +* +* +* s - - GEN<br />

alle<strong>in</strong> (e<strong>in</strong>zig, nur) 38 sit HO -* -* +* +* -* - s,man,sit,npp,card both - -<br />

allenfalls 16 pragm HO - - - - - - s,man,sit,npp,card,ap pre - -<br />

allenthalben 27 sit EN + - - - -* -* s - - -<br />

allerd<strong>in</strong>gs 06 pragm HO +* -* -* -* -* -* s,sit post - -<br />

allezeit (-) 37 sit RS -? - +? - -? -? s - - -<br />

also 22 sit HO + - - - - - s,sit,npp post - -<br />

also (nicht kausal) 03 pragm HO - - - - - - s - - -<br />

andererseits 07 pragm HO + - - - - - s - - -<br />

anders 43 mod RS + - + + + + s - - -<br />

anderswo 27 sit EN + - + + + + s - - -<br />

anfangs 26 sit EN + - - + + + s - - -<br />

angeblich 12 pragm HO +* - -* -* -* -* s - - -<br />

anläßlich ... 29 sit HO + - + + + + s - - -<br />

ansche<strong>in</strong>end 12 pragm HO + - - - - - s - - -<br />

auch (außerdem) 07 pragm HO +* - -* -* -* -* s - - -<br />

auch (gleichfalls) 35 sit HO -* - +* -* +* -* s,man,sit,npp,ap,koord pre - -<br />

augensche<strong>in</strong>lich (4) 12 pragm HO +* - - + - -* s - - -<br />

außen 27 sit EN + - + + + + s - - -<br />

außerdem 07 pragm HO + - -* - - - s - - -<br />

ausnahmsweise (12) 14 pragm HO -*& -* -* +* -* -* s - - -<br />

214


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

auswärts 27 sit EN + - + + + + s - - -<br />

bald 33 sit HO + - + + + + s - - -<br />

bedauerlicherweise 13 pragm HO +* -* -* -* -* -* s - - -<br />

beiderseits 27 sit EN + - + + + + s - - GEN<br />

be<strong>in</strong>ahe 11 pragm HO + - - - -* - s,man,sit,ap,card pre - -<br />

beispielsweise 08 pragm HO + - - - - - s,sit,npp,ap post - -<br />

beizeiten (2) 39 sit RS + - + + + + s - - -<br />

bekanntlich 12 pragm HO + - - - - - s - - -<br />

bereits 39 sit HO - - - - - - s,man,sit,npp,card,ap both yes -<br />

besonders (Ordnungsfunktion) 08 pragm HO +* -* - +* +* -* s,man,sit,pragm,npp,ap pre - -<br />

besser 43 mod RS + - + + + + s - - -<br />

besser (subjektiv wertend) 14 pragm HO -* -* -* - - - s - - -<br />

bestenfalls 16 pragm HO + - - - - - s,man,sit,card,ap pre - -<br />

bestimmt (96) 12 pragm HO +* - - + + -* s,npp pre - -<br />

billigerweise (-) 14 pragm HO - - - - - - s - - -<br />

bisher 26 sit EN + - - + +* + s - - -<br />

bislang (328) 26 sit EN +* - - -* -* + s - - -<br />

bisweilen 26 sit EN + - - - -* -* s - - -<br />

bitte (81) 18 pragm RS + - - - -* - s - - -<br />

bloß (Wunsch, Aufforderung) 05 pragm HO - - - - - - s - - -<br />

bloß (nur, alle<strong>in</strong>) 38 sit HO +*& - +* -* - -* s,man,sit,npp,card,ap pre yes -<br />

da 02 pragm HO - - - - - - s - - -<br />

da 26 sit EN + - + + -* + s - - -<br />

dabei 27 sit EN + - + +* + +* s - - -<br />

dabei (Anlaß) 29 sit HO + - + + - -* s - - -<br />

215


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

dadurch (medial) 25 sit RS + - + + - - s - - -<br />

dafür (f<strong>in</strong>al: zu diesem Zweck) 24 sit EN + - + + - - s - - -<br />

dagegen (<strong>in</strong>dessen) 06 pragm HO + - - - - - s,npp post - -<br />

daher (kausal) 22 sit HO + - - - - - s - - -<br />

dah<strong>in</strong>ter 27 sit EN + - + + + + s - - -<br />

damals 26 sit HO + - + + + + s - - -<br />

damit (<strong>in</strong>strumental) 44 mod EN + - + + -* + s - - -<br />

damit (medial, weniger konkret als a 44 ) 25 sit HO + - + + -* + s - - -<br />

danach 26 sit EN + - + + + + s - - -<br />

danach 27 sit EN + - + + + + s - - -<br />

daneben 27 sit EN + - + + + + s - - -<br />

daneben (obendre<strong>in</strong>) 07 pragm HO + - - - - - s - - -<br />

dann 26 sit EN + - - + -* -* s - - -<br />

dann (da, nun, jetzt; ähnlich a 19/26 ) 02 pragm HO - - - - -* - s - - -<br />

dann (<strong>in</strong> diesem Fall) 19 sit HO +* - - +* - -* s - - -<br />

daran 27 sit EN + - + + + + s - - -<br />

darauf 27 sit EN + - + + + + s - - -<br />

dar<strong>in</strong> 27 sit EN + - + + + + s - - -<br />

darüber 27 sit EN +* - + + + + s - - -<br />

darum 22 sit EN + - + + - +* s - - -<br />

dauernd 37 sit HO + - + + + + s - - -<br />

davon 27 sit EN + - + + - - s - - -<br />

dazu (f<strong>in</strong>al: zu diesem Zweck) 24 sit EN + - + + - -* s - - -<br />

dazwischen 27 sit EN + - + + + + s - - -<br />

de<strong>in</strong>ethalben 22 sit EN + - + + + + s - - -<br />

216


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

de<strong>in</strong>etwegen 22 sit EN + - + + + + s - - -<br />

demnächst 26 sit EN + - + - + + s - - -<br />

denn 01 pragm HO - - - - - - s - - -<br />

dennoch 20 sit HO + - - - +* -* s - - -<br />

dere<strong>in</strong>st (5) 26 sit EN -* -* -* -* -* -* s - - -<br />

derzeit 26 sit EN +* - -* - - -* s - - -<br />

deshalb 22 sit HO + - + + -* -* s - - -<br />

deswegen 22 sit EN + - + + -* -* s - - -<br />

dienstags 26,40 sit EN + - + + + + s - - -<br />

diesbezüglich 30 sit HO +* - + + -* -* s - - -<br />

diesseits (5) 27 sit EN -* -* -* -* -* + s - - GEN<br />

doch (nicht adversativ/konzessiv; bekräftigend) 04 pragm HO - - - - - - s - - -<br />

doch (pragmatisch: durchaus) 34 sit HO + - - - +* - s - - -<br />

donnerstags 26,40 sit EN + - + + + + s - - -<br />

dort 27 sit HO + - + + + + s - - -<br />

draußen 27 sit EN + - + + + + s - - -<br />

dr<strong>in</strong>nen 27 sit EN + - + + + + s - - -<br />

droben 27 sit EN + - + + + + s - - -<br />

drüben 27 sit EN + - + + + + s - - -<br />

drunten 27 sit EN + - + + + + s - - -<br />

dummerweise 14 pragm HO + - - - - - s - - -<br />

durchaus 34 sit HO - - - - + - s,sit pre - -<br />

durchweg (ausnahmslos) (52) 37 sit RS +* -* +* +* -* -* s,ap pre - -<br />

durchwegs (ausnahmslos) (1) 37 sit RS +* -* +* +* -* -* s,ap pre - -<br />

eben 08 pragm HO - - - - - - s,sit,npp pre - -<br />

217


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

eben (halt; nicht temporal) 05 pragm HO - - - - - - s - - -<br />

eben (temporal) 33 sit HO + - - - + + s - - -<br />

ebenfalls 35 sit HO +* - - -* + - s,npp - yes -<br />

ebenso 35 sit RS + - - - + - s,man,sit,ap pre - -<br />

ehedem (14) 26 sit EN +* -* -* -* +* - s - - -<br />

ehemals 26 sit EN -* - - - - - s - - -<br />

eher 26,40 sit EN + + + + + + s - - -<br />

eher (rather) 06 pragm RS + - - +* - - s,man,sit,pragm,npp,card,ap pre - -<br />

ehrenhalber 22 sit EN + - + + + - s - - -<br />

eigentlich (im Grunde) 11 pragm HO + - +* - - - s - - -<br />

eigentlich (überhaupt) (nur Fragen) 09 pragm HO - - - - - - s - - -<br />

eigentümlicherweise (2) 14 pragm HO +* -* -* -* -* -* s - - -<br />

e<strong>in</strong>erseits 07 pragm HO + - - - - - s - - -<br />

e<strong>in</strong>fach 18 pragm HO -* -* +* +* -* - s - - -<br />

e<strong>in</strong>fach 43 mod RS +* + + +* +* + s - - -<br />

e<strong>in</strong>mal (1 Mal) 37 sit RS + - + +* + -* s - - -<br />

e<strong>in</strong>mal (toner) 14 pragm RS - - - - - - s - - -<br />

e<strong>in</strong>st 26 sit EN + - - - - -* s - - -<br />

e<strong>in</strong>stmals 26 sit EN + - - - - -* s - - -<br />

e<strong>in</strong>stweilen 26 sit EN + - -* - - - s - - -<br />

e<strong>in</strong>zig (nur) 38 sit HO - - -* - - - s,sit,npp pre - -<br />

endlich 33 sit HO + - - - - - s - - -<br />

endlich (nicht temporal) 07 pragm HO + - - - - - s - - -<br />

entsprechend 31 sit HO + - + - + + s,man,sit,ap pre - GEN<br />

erfreulicherweise 13 pragm HO + - - - - - s - - -<br />

218


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

erneut (210) 36 sit HO +* - -*& -* + - s - - -<br />

erst 39 sit HO + - - - - - s,sit,npp,card pre - -<br />

erst (nicht temporal) 34 sit HO - - - - - - s - - -<br />

erstens 07 pragm HO + - - - - - s - - -<br />

etwa (<strong>in</strong> Fragesätzen) 05 pragm HO - - -* - - - s - - -<br />

etwa (ungefähr) 16 pragm RS - - -* - - - s,sit,npp,card pre - -<br />

etwa (z.B.) 08 pragm HO - - - - - - s,sit,npp post - -<br />

eurethalben 22 sit EN + - + + + + s - - -<br />

euretwegen 22 sit EN + - + + + + s - - -<br />

eventuell (40) 12 pragm HO + - - -* - - s - - -<br />

fast 11 pragm HO + - - - - - s,sit,npp,card,ap pre yes -<br />

ferner (148) 07 pragm HO +* -* -* -* - - s - - -<br />

fortan 26 sit EN +* -* -* -* -* -* s - - -<br />

freilich 06 pragm HO + - - - - - s,sit,npp post - -<br />

freitags 26,40 sit EN + - + + + + s - - -<br />

früh 40 sit RS + + + + + + s - - -<br />

früher (damals) 26 sit RS + - - + + + s - - -<br />

ganz 43 mod HO + - +* - +* +* s,man,sit,pragm,npp,ap pre - -<br />

gar 17 pragm HO -* - - - - - s,neg pre - -<br />

gelegentlich 37 sit HO + - + - -* -* s - - -<br />

genauso 43 mod RS + - - + + + s - - -<br />

genauso (ebenso) 35 sit RS + - - + + + s,man,sit,pragm pre - -<br />

gerade 43 mod RS -* + + + + + s - - -<br />

gerade (eben; erst recht) 08 pragm HO -* - +* - - -* s,sit,npp,coord pre - -<br />

gerade (jetzt) 33 sit HO +* - - +* - + s - - -<br />

219


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

geradeheraus (1) 43 mod RS -* -* -* -* -* -* s - - -<br />

geradenwegs (-) 43 mod RS + - -? -? + - s - - -<br />

geradeswegs (-) 43 mod RS + - -? -? + - s - - -<br />

geradewegs (7) 43 mod RS -* -* -* -* -* -* s - - -<br />

geradezu 18 pragm HO - - - - - - s,man,ap pre - -<br />

gern 43 mod HO + - + + + - s - - -<br />

gestern 26 sit HO + - + + + + s - - -<br />

getrost 18 pragm HO - - - + -* - s - - -<br />

gewiß 12 pragm HO + - - + + + s - - -<br />

gewissermaßen 11 pragm HO -* - - - - - s - - -<br />

glatt (ohne weiteres) 43 mod RS +* + +* +* +* +* s - - -<br />

glattweg (ohne weiteres) (2) 43 mod RS -* -* -* -* -* -* s - - -<br />

gleich 18 pragm RS - - -* - - - s - - -<br />

gleich 33 sit HO + - +* - + + s - - -<br />

gleichfalls 35 sit HO - - - - + - s - - -<br />

gleichsam 11 pragm HO + - - - -* - s - - -<br />

gleichzeitig 26,40 sit RS + - + + + + s - - -<br />

glücklicherweise 13 pragm HO + - - - - - s - - -<br />

Gottseidank 13 pragm HO + - - - -* - s - - -<br />

gut 43 mod HO - + + + + + s,card pre - -<br />

häufig 37 sit HO + + + + + + s - - -<br />

halt (eben) 05 pragm HO - - - - - - s - - -<br />

hernach (8) 26 sit EN - -* -* -* -* -* s - - -<br />

heute 26 sit EN + - -* + -* -* s - - -<br />

heutzutage (43) 26 sit EN +* - - + -* -* s - - -<br />

220


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

hier 27 sit HO + - + + + + s - - -<br />

hier 30 sit HO + - + + + + s - - -<br />

hierfür 24 sit EN + - + + + + s - - -<br />

h<strong>in</strong>ten 27 sit EN + - + + + + s - - -<br />

h<strong>in</strong>terher 26 sit EN + - +* + + + s - - -<br />

h<strong>in</strong>terrücks 43 mod RS -*& -* -* -* -* - s - - -<br />

höchstens 16 pragm HO +* - - - - - s,npp,card,coord pre yes -<br />

h<strong>of</strong>fentlich 13 pragm HO + - - - - - s - - -<br />

ihrethalben 22 sit EN + - + -*& + + s - - -<br />

ihretwegen 22 sit EN + - + + + + s - - -<br />

immer 37 sit HO + - + + + - s,man,sit,npp pre - -<br />

immerh<strong>in</strong> 09 pragm HO + - - - - - s,ap,card pre - -<br />

immerzu 37 sit HO + - - - + - s - - -<br />

<strong>in</strong>dessen 06 pragm HO + - - - - - s,npp post - -<br />

<strong>in</strong>nen 27 sit EN + - + + + + s - - -<br />

<strong>in</strong>sbesondere 08 pragm HO +* - - - - - s,sit,npp pre - -<br />

<strong>in</strong>sgeheim (14) 28 sit RS +* -* -* -* -* -* s - - -<br />

<strong>in</strong>zwischen 26 sit HO + - - - - -* s - - -<br />

irgendwann (97) 26,40 sit RS + - - - - -* s - - -<br />

irgendwo 27 sit EN + - - - + + s - - -<br />

irrtümlicherweise 14 pragm HO + - - - -* - s - - -<br />

ja 01 pragm HO - - - - - - s - - -<br />

jäh (15) 33 sit RS - - -* -* +* -* s - - -<br />

jähl<strong>in</strong>gs (-) 33 sit RS + - + - + - s - - -<br />

jahrelang 40 sit HO + - + + + - s - - -<br />

221


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

jedenfalls 06 pragm HO +* - - - - - s,sit,npp both - -<br />

jederzeit 37 sit RS + - + - + - s - - -<br />

jedesmal 37 sit HO + - + + + - s - - -<br />

jedoch 06 pragm HO +* - - - - - s,sit,npp post - -<br />

jenseits 27 sit EN + - + + + + s - - GEN<br />

jetzt 26 sit HO + - + + + + s - - -<br />

jetzt (nicht temporal) 02 pragm HO - - - - - - s - - -<br />

kaum 41 neg HO + - - - + - s - - -<br />

ke<strong>in</strong>eswegs (224) 41 neg HO +* - +* +* +* - s,man,sit,npp,ap pre - -<br />

klugerweise 14 pragm HO + - - - - - s - - -<br />

kopfüber 43 mod RS -* -* -* -* -* -* s - - -<br />

künftig 26 sit HO + - - - + - s - - -<br />

kürzlich (126) 26 sit RS +* - -* +* -* -* s - - -<br />

längsseits (-) 27 sit EN + - + + -*& -* s - - GEN<br />

längst (400) 33 sit RS +* - - +* - - s,neg pre - -<br />

lange 40 sit HO + + + + + + s - - -<br />

laufend 37 sit HO +* - -* -* +* + s - - -<br />

laut 43 mod HO + + + + + + s - - -<br />

lediglich 38 sit HO - - - - - - s,man,sit,npp,card,ap,coord pre - -<br />

leicht 18 pragm HO + - +* + +* - s - - -<br />

leicht 43 mod RS + + + + + + s,ap pre - -<br />

leider 13 pragm HO + - - - - - s - - -<br />

letztens 07 pragm RS + - - - - - s - - -<br />

letztlich 11 pragm HO + - - - - - s - - -<br />

l<strong>in</strong>ks 27 sit EN + - + + + + s - - -<br />

222


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

mal 17 pragm RS - - - - - - s - - -<br />

manchmal 37 sit HO + - - - + + s - - -<br />

mehrmals 37 sit RS +* - - +* + + s - - -<br />

me<strong>in</strong>ethalben 22 sit EN + - + + + + s - - -<br />

me<strong>in</strong>etwegen 22 sit EN + - + + + + s - - -<br />

meist 37 sit RS + - - - + + s - - -<br />

meistens 37 sit HO + - - - + + s - - -<br />

merkwürdigerweise 14 pragm HO + - - - - - s - - -<br />

mite<strong>in</strong>ander 42 mod HO + - + + + - s - - -<br />

mittendr<strong>in</strong> 27 sit EN + - - + + + s - - -<br />

mittlerweile 26 sit HO + - - - - - s - - -<br />

mittwochs 26,40 sit EN + - + + + + s - - -<br />

mitunter 37 sit HO + - - - - - s - - -<br />

möglicherweise 12 pragm HO +* - - - - - s - - -<br />

momentan 26 sit RS + - - - -* - s - - -<br />

montags 26,40 sit EN + - + + + + s - - -<br />

morgen 26 sit EN + - + + + + s - - -<br />

morgens 26,40 sit EN + - + + + + s - - -<br />

nächstens 26 sit EN + - - + - + s - - -<br />

nämlich 05 pragm HO - - - - - - s - - -<br />

nachher 26 sit EN + - + + - + s - - -<br />

nachmittags 26,40 sit EN + - + + + + s - - -<br />

natürlich 12 pragm HO + - - - - -* s - - -<br />

natürlich 43 mod RS + + + + + + s - - -<br />

nebenan 27 sit EN + - + + + + s - - -<br />

223


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

neulich 26 sit EN + - - + + + s - - -<br />

nicht (nur Fragen) 05 pragm HO - - - - - - s - - -<br />

noch 39 sit HO + - - +* -* + s,sit,npp both yes -<br />

noch (nicht temporal) 34 sit HO - - - - - - s - - -<br />

nochmals (71) 36 sit HO + - +* +* + - s - - -<br />

normalerweise 14 pragm HO + - - - - - s - - -<br />

notwendigerweise (10) 14 pragm HO -* - +* + -* -* s - - -<br />

nun 02 pragm HO - - - - - - s - - -<br />

nun 26 sit EN + - - - - - s - - -<br />

nur (Wunsch/Aufforderungssatz) 05 pragm HO -* - - - - - s - - -<br />

nur (allerd<strong>in</strong>gs) 06 pragm HO +* - - - - - s - - -<br />

nur (bloß; limitierend) 38 sit HO -* - + - - - s,man,sit,pragm,npp,card,coord pre yes -<br />

öfters 37 sit HO + - + + + + s - - -<br />

oben 27 sit EN + - + + + + s - - -<br />

obenauf (1) 27 sit EN +* -* -* -* -* -* s - - -<br />

obendre<strong>in</strong> 07 pragm HO + - - - - - s - - -<br />

obenh<strong>in</strong> (-) 43 mod RS -? - +? + + - s - - -<br />

<strong>of</strong>fenbar (311) 12 pragm HO +* - - +* - +* s,npp pre - -<br />

<strong>of</strong>fensichtlich 12 pragm HO + - -* + - + s - - -<br />

<strong>of</strong>t 37 sit HO + + + + + + s - - -<br />

<strong>of</strong>tmals 37 sit RS + - - - + -* s - - -<br />

ohnedies 09 pragm HO - - - - + - s - - -<br />

ohneh<strong>in</strong> 09 pragm HO - - - - + - s - - -<br />

plötzlich 33 sit HO + - + + + + s - - -<br />

praktisch (140) 11 pragm HO +* - -* - - + s,neg pre - -<br />

224


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

primär (21) 08 pragm HO + - - - - - s - - -<br />

pr<strong>in</strong>zipiell (18) 11 pragm HO +* - -*& + + - s - - -<br />

rechtlich 21 sit HO + - - + + - s - - -<br />

rechts 27 sit EN + - + + + + s - - -<br />

regelmäßig 37 sit HO + + + + + + s - - -<br />

regelrecht 18 pragm HO - - - - - - s - - -<br />

r<strong>in</strong>gs (5) 27 sit RS +* -* -* -* -* -* s - - PP (um)<br />

r<strong>in</strong>gsherum (1) 27 sit EN -* -* -* -* +* -* s - - -<br />

r<strong>in</strong>gsum (10) 27 sit EN +* -* -* -* +* -* s - - -<br />

ruhig 18 pragm HO - - - - - - s - - -<br />

ruhig 43 mod RS + + + + + + s - - -<br />

rund 16 pragm RS - - - - - - card pre - -<br />

rundheraus (<strong>of</strong>fen) (1) 43 mod RS + - - + + - s - - -<br />

rundweg (unumwunden) (7) 43 mod RS + - - -* + - s - - -<br />

samstags 26,40 sit EN + - + + + + s - - -<br />

sche<strong>in</strong>bar 12 pragm HO + - - - - - s,man,sit,npp pre - -<br />

schließlich 33 sit HO + - - - - - s - - -<br />

schließlich (nicht temporal) 07 pragm HO + - - - - - s - - -<br />

schnell 18 pragm RS - - - + - - s - - -<br />

schnell 43 mod HO + + + + + + s - - -<br />

schnurstracks (5) 43 mod RS + - + -* + - s - - -<br />

schon 39 sit HO + - + - - - s,sit,npp both yes -<br />

schon (nicht temporal) 34 sit HO - - - - - - s - - -<br />

sehr 43 mod HO -* - + - + - s,man,sit,pragm,ap pre - -<br />

se<strong>in</strong>erzeit (58) 26 sit EN +* - - +* - - s - - -<br />

225


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

se<strong>in</strong>ethalben 22 sit EN + - + + + + s - - -<br />

se<strong>in</strong>etwegen 22 sit EN + - + + + + s - - -<br />

seitdem 26 sit EN + - + + + -* s - - -<br />

seither 26 sit EN + - + - - -* s - - -<br />

sekundär 08 pragm HO + - - - - - s - - -<br />

selbst ("sogar", nicht "selber") 17 pragm HO - - - - - - s,man,sit,npp,ap,coord pre - -<br />

selbstverständlich 12 pragm HO + - - + + + s - - -<br />

selten 37 sit HO + + + + + + s - - -<br />

sicher 43 mod RS -* + + + + + s - - -<br />

sicher (sicherlich) 12 pragm HO + - - + + - s - - -<br />

sicherlich 12 pragm HO + - - - + - s - - -<br />

so 43 mod HO + - + + + + s,man,sit,pragm,ap pre - -<br />

so (kausal) 22 sit HO + - - + - - s - - -<br />

soeben (26) 33 sit HO +* - -* +* + -* s - - -<br />

s<strong>of</strong>ort (312) 33 sit HO +* - +* -* +* -* s - - -<br />

sogar 17 pragm HO - - - - - - s,man,sit,pragm,npp,card, pre yes -<br />

sogleich 33 sit HO + - - - -* -* s - - -<br />

somit (kausal) 22 sit RS + - - - - - s - - -<br />

sonntags 26,40 sit EN + - + + + + s - - -<br />

sonst (außerdem) 07 pragm HO + - - - - - s - - -<br />

sowieso 09 pragm HO -* - - - + - s - - -<br />

sozusagen 11 pragm HO - - - - - - s - - -<br />

später 26 sit HO + + + + + + s - - -<br />

ständig 37 sit HO + + + - + + s - - -<br />

stets 37 sit HO + - - - + + s - - -<br />

226


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

tagelang 40 sit RS + - + + + - s - - -<br />

tagsüber (28) 26 sit EN +* - -* +* + + s - - -<br />

tatsächlich (63) 15 pragm HO +* - -* - + + s - - -<br />

teilweise 11 pragm HO + - - + - - s - - -<br />

theoretisch 11 pragm HO + - - - - - s - - -<br />

trotzdem 20 sit HO + - - - + - s - - -<br />

überall 27 sit EN + - + - + + s - - -<br />

überdies 07 pragm HO + - - - - - s - - -<br />

überhaupt 09 pragm HO + - - - + - s,neg pre - -<br />

übermorgen 26 sit EN + - + + + -* s - - -<br />

übrigens 06 pragm HO + - - - - - s - - -<br />

umher 27 sit EN - - - + + - s - - -<br />

umsonst (31) (Bedeutung=a 14 ?) 43 mod RS +* - +* + + + s - - -<br />

unbed<strong>in</strong>gt (27) 34? sit RS + - +* + + - s - - -<br />

unglücklicherweise 13 pragm HO + - - - - - s - - -<br />

unlängst 26 sit EN +* - - + -* -* s - - -<br />

unmöglich 14 pragm HO + - - + + + s - - -<br />

unserthalben 22 sit EN + - + + + + s - - -<br />

unsertwegen 22 sit EN + - + + + + s - - -<br />

unsretwegen 22 sit EN + - + + + + s - - -<br />

unten 27 sit EN + - + + + + s - - -<br />

unterdessen 26 sit HO + - - - - - s - - -<br />

unumwunden 43 mod RS + - - + + - s - - -<br />

ununterbrochen (7) 37 sit EN + - + - -*& -* s - - -<br />

unversehens 33 sit RS + - - + + - s - - -<br />

227


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

unverzüglich 33 sit HO -* - - -*& -* -* s - - -<br />

vergebens (21) (Bed=a 14 ?) 43 mod RS +* - + + + + s - - -<br />

vermutlich 12 pragm HO + - - - - - s,sit,npp,ap pre - -<br />

vielfach 37 sit HO + - - - + - s - - -<br />

vielleicht 12 pragm HO +* - -* - - -* s - - -<br />

vielmehr 06 pragm HO + - - - - - s - - -<br />

völlig 43 mod HO -* - + - + - s,man,npp,ap pre - -<br />

voraussichtlich (55) 12 pragm HO +* - - -* - -* s - - -<br />

vorerst (89) 33 sit EN +* - -* - - - s - - -<br />

vorgestern 26 sit EN + - -* + + + s - - -<br />

vorher 33 sit EN +* - + + + + s - - -<br />

vorh<strong>in</strong> 26 sit EN + - - + -* -* s - - -<br />

vorne 27 sit EN + - + + + + s - - -<br />

vorwiegend 08 pragm HO +* - -* -* -* -* s,npp,ap pre - -<br />

währenddessen 26 sit EN + - -* + -*& -* s - - -<br />

wahrsche<strong>in</strong>lich 12 pragm HO + - - + - + s,sit,card,ap pre - -<br />

weiter (142) 07 pragm HO + - - - - - s - - -<br />

weiterh<strong>in</strong> 07 pragm HO + - - - + - s - - -<br />

wenig 43 mod HO -* + + +* + + s,sit,npp,ap pre - -<br />

wenigstens 16 pragm HO + - - - - - s,man,sit,npp,card pre - -<br />

wieder 36 sit HO + - - + + + s - - -<br />

wiederholt 37 sit HO + - - + -*& -* s - - -<br />

wiederum 36 sit HO + - - - + + s - - -<br />

wirklich 15 pragm HO - - + - + + s,man,sit,ap pre - -<br />

wirtschaftlich (60) 21 sit HO + - + + + - s - - -<br />

228


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

wirtschaftsmäßig (-) 21 sit HO + - - +? + - s - - -<br />

wochenlang 40 sit RS + - + + + - s - - -<br />

wohl (adversativ/entgegen d.allgeme<strong>in</strong>en Auffassung) 34 sit HO -*& - - +* +* -* s - - -<br />

wohl (toner: E<strong>in</strong>schränkung im nächsten Satz) 06 pragm HO +* - - - - -* s - - -<br />

wohl (vermutlich; bestimmt) 12 pragm HO -* - - +* - -* s,sit,npp,card,ap pre yes -<br />

womöglich 12 pragm HO + - - - - - s - - -<br />

zeitlebens (17) 40 sit RS + - - - -*& -* s - - -<br />

zeitweise 37 sit HO + - - + -*& -* s - - -<br />

zudem 07 pragm HO +* - - - - - s - - -<br />

zuerst 26,40 sit EN + - + + + + s - - -<br />

zugleich 26,40 sit EN + - + + -*& -* s - - -<br />

zugleich (obendre<strong>in</strong>, ferner) 07 pragm HO +* - - - -* -* s - - -<br />

zuletzt 26,40 sit EN + - + + -*& -* s - - -<br />

zumal 08 pragm HO - - - - - - s,sit,npp pre - -<br />

zum<strong>in</strong>dest 16 pragm HO + - - - - - s,man,sit,pragm,npp,card pre yes -<br />

zunächst (642) 33 sit RS +* - - -* - - s - - -<br />

zuvor 26,40 sit EN + - + + + -* s - - -<br />

zuweilen 37 sit RS + - - - - - s - - -<br />

zwangsläufig 14 pragm HO +* - +* +* + +* s - - -<br />

zwar 06 pragm HO + - - - - - s - - -<br />

zweifellos 12 pragm HO + - - + + - s - - -<br />

zweitens 07 pragm HO + - - - - - s - - -<br />

zwischendurch 26,40 sit EN + - + + + - s - - -<br />

229


8.3. LISTING OF MODIFIERS ACCORDING TO POSITION CLASSES<br />

Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

denn 01 pragm HO - - - - - - s - - -<br />

ja 01 pragm HO - - - - - - s - - -<br />

da 02 pragm HO - - - - - - s - - -<br />

dann (da, nun, jetzt; ähnlich a 19/26 ) 02 pragm HO - - - - -* - s - - -<br />

jetzt (nicht temporal) 02 pragm HO - - - - - - s - - -<br />

nun 02 pragm HO - - - - - - s - - -<br />

aber (nicht nur advers.) 03 pragm HO - - - - - - s,man,sit,npp,ap post - -<br />

also (nicht kausal) 03 pragm HO - - - - - - s - - -<br />

doch (nicht adversativ/konz.; bekräftgend) 04 pragm HO - - - - - - s - - -<br />

bloß (Wunsch, Aufforderung) 05 pragm HO - - - - - - s - - -<br />

eben (halt; nicht temporal) 05 pragm HO - - - - - - s - - -<br />

etwa (<strong>in</strong> Fragesätzen) 05 pragm HO - - -* - - - s - - -<br />

halt (eben) 05 pragm HO - - - - - - s - - -<br />

nämlich 05 pragm HO - - - - - - s - - -<br />

nicht (nur Fragen) 05 pragm HO - - - - - - s - - -<br />

nur (Wunsch/Aufforderungssatz) 05 pragm HO -* - - - - - s - - -<br />

230


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

allerd<strong>in</strong>gs 06 pragm HO +* -* -* -* -* -* s,sit post - -<br />

dagegen (<strong>in</strong>dessen) 06 pragm HO + - - - - - s,npp post - -<br />

eher (rather) 06 pragm RS + - - +* - - s,man,sit,pragm,npp,card,ap pre - -<br />

freilich 06 pragm HO + - - - - - s,sit,npp post - -<br />

<strong>in</strong>dessen 06 pragm HO + - - - - - s,npp post - -<br />

jedenfalls 06 pragm HO +* - - - - - s,sit,npp both - -<br />

jedoch 06 pragm HO +* - - - - - s,sit,npp post - -<br />

nur (allerd<strong>in</strong>gs) 06 pragm HO +* - - - - - s - - -<br />

übrigens 06 pragm HO + - - - - - s - - -<br />

vielmehr 06 pragm HO + - - - - - s - - -<br />

wohl (toner: E<strong>in</strong>schränkung im nächst. Satz) 06 pragm HO +* - - - - -* s - - -<br />

zwar 06 pragm HO + - - - - - s - - -<br />

231


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

andererseits 07 pragm HO + - - - - - s - - -<br />

auch (außerdem) 07 pragm HO +* - -* -* -* -* s - - -<br />

außerdem 07 pragm HO + - -* - - - s - - -<br />

daneben (obendre<strong>in</strong>) 07 pragm HO + - - - - - s - - -<br />

e<strong>in</strong>erseits 07 pragm HO + - - - - - s - - -<br />

endlich (nicht temporal) 07 pragm HO + - - - - - s - - -<br />

erstens 07 pragm HO + - - - - - s - - -<br />

ferner (148) 07 pragm HO +* -* -* -* - - s - - -<br />

letztens 07 pragm RS + - - - - - s - - -<br />

obendre<strong>in</strong> 07 pragm HO + - - - - - s - - -<br />

schließlich (nicht temporal) 07 pragm HO + - - - - - s - - -<br />

sonst (außerdem) 07 pragm HO + - - - - - s - - -<br />

überdies 07 pragm HO + - - - - - s - - -<br />

weiter (142) 07 pragm HO + - - - - - s - - -<br />

weiterh<strong>in</strong> 07 pragm HO + - - - + - s - - -<br />

zudem 07 pragm HO +* - - - - - s - - -<br />

zugleich (obendre<strong>in</strong>, ferner) 07 pragm HO +* - - - -* -* s - - -<br />

zweitens 07 pragm HO + - - - - - s - - -<br />

232


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

beispielsweise 08 pragm HO + - - - - - s,sit,npp,ap post - -<br />

besonders (Ordnungsfunktion) 08 pragm HO +* -* - +* +* -* s,man,sit,pragm,npp,ap pre - -<br />

eben 08 pragm HO - - - - - - s,sit,npp pre - -<br />

etwa (z.B.) 08 pragm HO - - - - - - s,sit,npp post - -<br />

gerade (eben; erst recht) 08 pragm HO -* - +* - - -* s,sit,npp,coord pre - -<br />

<strong>in</strong>sbesondere 08 pragm HO +* - - - - - s,sit,npp pre - -<br />

primär (21) 08 pragm HO + - - - - - s - - -<br />

sekundär 08 pragm HO + - - - - - s - - -<br />

vorwiegend 08 pragm HO +* - -* -* -* -* s,npp,ap pre - -<br />

zumal 08 pragm HO - - - - - - s,sit,npp pre - -<br />

eigentlich (überhaupt (nur Fragen) 09 pragm HO - - - - - - s - - -<br />

immerh<strong>in</strong> 09 pragm HO + - - - - - s,card,ap pre - -<br />

ohnedies 09 pragm HO - - - - + - s - - -<br />

ohneh<strong>in</strong> 09 pragm HO - - - - + - s - - -<br />

sowieso 09 pragm HO -* - - - + - s - - -<br />

überhaupt 09 pragm HO + - - - + - s,neg pre - -<br />

be<strong>in</strong>ahe 11 pragm HO + - - - -* - s,man,sit,ap,card pre - -<br />

eigentlich (im Grunde) 11 pragm HO + - +* - - - s - - -<br />

fast 11 pragm HO + - - - - - s,sit,npp,card,ap pre yes -<br />

gewissermaßen 11 pragm HO -* - - - - - s - - -<br />

gleichsam 11 pragm HO + - - - -* - s - - -<br />

letztlich 11 pragm HO + - - - - - s - - -<br />

praktisch (140) 11 pragm HO +* - -* - - + s,neg pre - -<br />

pr<strong>in</strong>zipiell (18) 11 pragm HO +* - -*& + + - s - - -<br />

sozusagen 11 pragm HO - - - - - - s - - -<br />

teilweise 11 pragm HO + - - + - - s - - -<br />

theoretisch 11 pragm HO + - - - - - s - - -<br />

233


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

angeblich 12 pragm HO +* - -* -* -* -* s - - -<br />

ansche<strong>in</strong>end 12 pragm HO + - - - - - s - - -<br />

augensche<strong>in</strong>lich (4) 12 pragm HO +* - - + - -* s - - -<br />

bekanntlich 12 pragm HO + - - - - - s - - -<br />

bestimmt (96) 12 pragm HO +* - - + + -* s,npp pre - -<br />

eventuell (40) 12 pragm HO + - - -* - - s - - -<br />

gewiß 12 pragm HO + - - + + + s - - -<br />

möglicherweise 12 pragm HO +* - - - - - s - - -<br />

natürlich 12 pragm HO + - - - - -* s - - -<br />

<strong>of</strong>fenbar (311) 12 pragm HO +* - - +* - +* s,npp pre - -<br />

<strong>of</strong>fensichtlich 12 pragm HO + - -* + - + s - - -<br />

sche<strong>in</strong>bar 12 pragm HO + - - - - - s,man,sit,npp pre - -<br />

selbstverständlich 12 pragm HO + - - + + + s - - -<br />

sicher (sicherlich) 12 pragm HO + - - + + - s - - -<br />

sicherlich 12 pragm HO + - - - + - s - - -<br />

vermutlich 12 pragm HO + - - - - - s,sit,npp,ap pre - -<br />

vielleicht 12 pragm HO +* - -* - - -* s - - -<br />

voraussichtlich (55) 12 pragm HO +* - - -* - -* s - - -<br />

wahrsche<strong>in</strong>lich 12 pragm HO + - - + - + s,sit,card,ap pre - -<br />

wohl (vermutlich; bestimmt) 12 pragm HO -* - - +* - -* s,sit,npp,card,ap pre yes -<br />

womöglich 12 pragm HO + - - - - - s - - -<br />

zweifellos 12 pragm HO + - - + + - s - - -<br />

234


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

bedauerlicherweise 13 pragm HO +* -* -* -* -* -* s - - -<br />

erfreulicherweise 13 pragm HO + - - - - - s - - -<br />

glücklicherweise 13 pragm HO + - - - - - s - - -<br />

Gottseidank 13 pragm HO + - - - -* - s - - -<br />

h<strong>of</strong>fentlich 13 pragm HO + - - - - - s - - -<br />

leider 13 pragm HO + - - - - - s - - -<br />

unglücklicherweise 13 pragm HO + - - - - - s - - -<br />

ausnahmsweise (12) 14 pragm HO -*& -* -* +* -* -* s - - -<br />

besser (subjektiv wertend) 14 pragm HO -* -* -* - - - s - - -<br />

billigerweise (-) 14 pragm HO - - - - - - s - - -<br />

dummerweise 14 pragm HO + - - - - - s - - -<br />

eigentümlicherweise (2) 14 pragm HO +* -* -* -* -* -* s - - -<br />

e<strong>in</strong>mal (toner) 14 pragm RS - - - - - - s - - -<br />

irrtümlicherweise 14 pragm HO + - - - -* - s - - -<br />

klugerweise 14 pragm HO + - - - - - s - - -<br />

merkwürdigerweise 14 pragm HO + - - - - - s - - -<br />

normalerweise 14 pragm HO + - - - - - s - - -<br />

notwendigerweise (10) 14 pragm HO -* - +* + -* -* s - - -<br />

unmöglich 14 pragm HO + - - + + + s - - -<br />

zwangsläufig 14 pragm HO +* - +* +* + +* s - - -<br />

tatsächlich (63) 15 pragm HO +* - -* - + + s - - -<br />

wirklich 15 pragm HO - - + - + + s,man,sit,ap pre - -<br />

235


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

allenfalls 16 pragm HO - - - - - - s,man,sit,npp,card,ap pre - -<br />

bestenfalls 16 pragm HO + - - - - - s,man,sit,card,ap pre - -<br />

etwa (ungefähr) 16 pragm RS - - -* - - - s,sit,npp,card pre - -<br />

höchstens 16 pragm HO +* - - - - - s,npp,card,coord pre yes -<br />

rund 16 pragm RS - - - - - - card pre - -<br />

wenigstens 16 pragm HO + - - - - - s,man,sit,npp,card pre - -<br />

zum<strong>in</strong>dest 16 pragm HO + - - - - - s,man,sit,pragm,npp,card pre yes -<br />

gar 17 pragm HO -* - - - - - s,neg pre - -<br />

mal 17 pragm RS - - - - - - s - - -<br />

selbst ("sogar", nicht "selber") 17 pragm HO - - - - - - s,man,sit,npp,ap,coord pre - -<br />

sogar 17 pragm HO - - - - - - s,man,sit,pragm,npp,card, pre yes -<br />

bitte (81) 18 pragm RS + - - - -* - s - - -<br />

e<strong>in</strong>fach 18 pragm HO -* -* +* +* -* - s - - -<br />

geradezu 18 pragm HO - - - - - - s,man,ap pre - -<br />

getrost 18 pragm HO - - - + -* - s - - -<br />

gleich 18 pragm RS - - -* - - - s - - -<br />

leicht 18 pragm HO + - +* + +* - s - - -<br />

regelrecht 18 pragm HO - - - - - - s - - -<br />

ruhig 18 pragm HO - - - - - - s - - -<br />

schnell 18 pragm RS - - - + - - s - - -<br />

dann (<strong>in</strong> diesem Fall) 19 sit HO +* - - +* - -* s - - -<br />

dennoch 20 sit HO + - - - +* -* s - - -<br />

trotzdem 20 sit HO + - - - + - s - - -<br />

236


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

rechtlich 21 sit HO + - - + + - s - - -<br />

wirtschaftlich (60) 21 sit HO + - + + + - s - - -<br />

wirtschaftsmäßig (-) 21 sit HO + - - +? + - s - - -<br />

also 22 sit HO + - - - - - s,sit,npp post - -<br />

daher (kausal) 22 sit HO + - - - - - s - - -<br />

darum 22 sit EN + - + + - +* s - - -<br />

de<strong>in</strong>ethalben 22 sit EN + - + + + + s - - -<br />

de<strong>in</strong>etwegen 22 sit EN + - + + + + s - - -<br />

deshalb 22 sit HO + - + + -* -* s - - -<br />

deswegen 22 sit EN + - + + -* -* s - - -<br />

ehrenhalber 22 sit EN + - + + + - s - - -<br />

eurethalben 22 sit EN + - + + + + s - - -<br />

euretwegen 22 sit EN + - + + + + s - - -<br />

ihrethalben 22 sit EN + - + -*& + + s - - -<br />

ihretwegen 22 sit EN + - + + + + s - - -<br />

me<strong>in</strong>ethalben 22 sit EN + - + + + + s - - -<br />

me<strong>in</strong>etwegen 22 sit EN + - + + + + s - - -<br />

se<strong>in</strong>ethalben 22 sit EN + - + + + + s - - -<br />

se<strong>in</strong>etwegen 22 sit EN + - + + + + s - - -<br />

so (kausal) 22 sit HO + - - + - - s - - -<br />

somit (kausal) 22 sit RS + - - - - - s - - -<br />

unserthalben 22 sit EN + - + + + + s - - -<br />

unsertwegen 22 sit EN + - + + + + s - - -<br />

unsretwegen 22 sit EN + - + + + + s - - -<br />

237


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

dafür (f<strong>in</strong>al: zu diesem Zweck) 24 sit EN + - + + - - s - - -<br />

dazu (f<strong>in</strong>al: zu diesem Zweck) 24 sit EN + - + + - -* s - - -<br />

hierfür 24 sit EN + - + + + + s - - -<br />

dadurch (medial) 25 sit RS + - + + - - s - - -<br />

damit (medial, weniger konkret als a 44 ) 25 sit HO + - + + -* + s - - -<br />

1980 26 sit HO + - + + + + s - - -<br />

anfangs 26 sit EN + - - + + + s - - -<br />

bisher 26 sit EN + - - + +* + s - - -<br />

bislang (328) 26 sit EN +* - - -* -* + s - - -<br />

bisweilen 26 sit EN + - - - -* -* s - - -<br />

da 26 sit EN + - + + -* + s - - -<br />

damals 26 sit HO + - + + + + s - - -<br />

danach 26 sit EN + - + + + + s - - -<br />

dann 26 sit EN + - - + -* -* s - - -<br />

demnächst 26 sit EN + - + - + + s - - -<br />

dere<strong>in</strong>st (5) 26 sit EN -* -* -* -* -* -* s - - -<br />

derzeit 26 sit EN +* - -* - - -* s - - -<br />

ehedem (14) 26 sit EN +* -* -* -* +* - s - - -<br />

ehemals 26 sit EN -* - - - - - s - - -<br />

e<strong>in</strong>st 26 sit EN + - - - - -* s - - -<br />

e<strong>in</strong>stmals 26 sit EN + - - - - -* s - - -<br />

e<strong>in</strong>stweilen 26 sit EN + - -* - - - s - - -<br />

fortan 26 sit EN +* -* -* -* -* -* s - - -<br />

früher (damals) 26 sit RS + - - + + + s - - -<br />

gestern 26 sit HO + - + + + + s - - -<br />

238


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

hernach (8) 26 sit EN - -* -* -* -* -* s - - -<br />

heute 26 sit EN + - -* + -* -* s - - -<br />

heutzutage (43) 26 sit EN +* - - + -* -* s - - -<br />

h<strong>in</strong>terher 26 sit EN + - +* + + + s - - -<br />

<strong>in</strong>zwischen 26 sit HO + - - - - -* s - - -<br />

jetzt 26 sit HO + - + + + + s - - -<br />

künftig 26 sit HO + - - - + - s - - -<br />

kürzlich (126) 26 sit RS +* - -* +* -* -* s - - -<br />

mittlerweile 26 sit HO + - - - - - s - - -<br />

momentan 26 sit RS + - - - -* - s - - -<br />

morgen 26 sit EN + - + + + + s - - -<br />

nächstens 26 sit EN + - - + - + s - - -<br />

nachher 26 sit EN + - + + - + s - - -<br />

neulich 26 sit EN + - - + + + s - - -<br />

nun 26 sit EN + - - - - - s - - -<br />

se<strong>in</strong>erzeit (58) 26 sit EN +* - - +* - - s - - -<br />

seitdem 26 sit EN + - + + + -* s - - -<br />

seither 26 sit EN + - + - - -* s - - -<br />

später 26 sit HO + + + + + + s - - -<br />

tagsüber (28) 26 sit EN +* - -* +* + + s - - -<br />

übermorgen 26 sit EN + - + + + -* s - - -<br />

unlängst 26 sit EN +* - - + -* -* s - - -<br />

unterdessen 26 sit HO + - - - - - s - - -<br />

vorgestern 26 sit EN + - -* + + + s - - -<br />

vorh<strong>in</strong> 26 sit EN + - - + -* -* s - - -<br />

währenddessen 26 sit EN + - -* + -*& -* s - - -<br />

239


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

abends 26,40 sit EN +* - + + + + s - - -<br />

dienstags 26,40 sit EN + - + + + + s - - -<br />

donnerstags 26,40 sit EN + - + + + + s - - -<br />

eher 26,40 sit EN + + + + + + s - - -<br />

freitags 26,40 sit EN + - + + + + s - - -<br />

gleichzeitig 26,40 sit RS + - + + + + s - - -<br />

irgendwann (97) 26,40 sit RS + - - - - -* s - - -<br />

mittwochs 26,40 sit EN + - + + + + s - - -<br />

montags 26,40 sit EN + - + + + + s - - -<br />

morgens 26,40 sit EN + - + + + + s - - -<br />

nachmittags 26,40 sit EN + - + + + + s - - -<br />

samstags 26,40 sit EN + - + + + + s - - -<br />

sonntags 26,40 sit EN + - + + + + s - - -<br />

zuerst 26,40 sit EN + - + + + + s - - -<br />

zugleich 26,40 sit EN + - + + -*& -* s - - -<br />

zuletzt 26,40 sit EN + - + + -*& -* s - - -<br />

zuvor 26,40 sit EN + - + + + -* s - - -<br />

zwischendurch 26,40 sit EN + - + + + - s - - -<br />

240


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

abseits (17; 5 ohne GEN) 27 sit RS +*# -* -*& +* +* +* s - - GEN<br />

allenthalben 27 sit EN + - - - -* -* s - - -<br />

anderswo 27 sit EN + - + + + + s - - -<br />

außen 27 sit EN + - + + + + s - - -<br />

auswärts 27 sit EN + - + + + + s - - -<br />

beiderseits 27 sit EN + - + + + + s - - GEN<br />

dabei 27 sit EN + - + +* + +* s - - -<br />

dah<strong>in</strong>ter 27 sit EN + - + + + + s - - -<br />

danach 27 sit EN + - + + + + s - - -<br />

daneben 27 sit EN + - + + + + s - - -<br />

daran 27 sit EN + - + + + + s - - -<br />

darauf 27 sit EN + - + + + + s - - -<br />

dar<strong>in</strong> 27 sit EN + - + + + + s - - -<br />

darüber 27 sit EN +* - + + + + s - - -<br />

davon 27 sit EN + - + + - - s - - -<br />

dazwischen 27 sit EN + - + + + + s - - -<br />

diesseits (5) 27 sit EN -* -* -* -* -* + s - - GEN<br />

dort 27 sit HO + - + + + + s - - -<br />

draußen 27 sit EN + - + + + + s - - -<br />

dr<strong>in</strong>nen 27 sit EN + - + + + + s - - -<br />

droben 27 sit EN + - + + + + s - - -<br />

drüben 27 sit EN + - + + + + s - - -<br />

drunten 27 sit EN + - + + + + s - - -<br />

hier 27 sit HO + - + + + + s - - -<br />

h<strong>in</strong>ten 27 sit EN + - + + + + s - - -<br />

241


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

<strong>in</strong>nen 27 sit EN + - + + + + s - - -<br />

irgendwo 27 sit EN + - - - + + s - - -<br />

jenseits 27 sit EN + - + + + + s - - GEN<br />

längsseits (-) 27 sit EN + - + + -*& -* s - - GEN<br />

l<strong>in</strong>ks 27 sit EN + - + + + + s - - -<br />

mittendr<strong>in</strong> 27 sit EN + - - + + + s - - -<br />

nebenan 27 sit EN + - + + + + s - - -<br />

oben 27 sit EN + - + + + + s - - -<br />

obenauf (1) 27 sit EN +* -* -* -* -* -* s - - -<br />

rechts 27 sit EN + - + + + + s - - -<br />

r<strong>in</strong>gs (5) 27 sit RS +* -* -* -* -* -* s - - PO (um)<br />

r<strong>in</strong>gsherum (1) 27 sit EN -* -* -* -* +* -* s - - -<br />

r<strong>in</strong>gsum (10) 27 sit EN +* -* -* -* +* -* s - - -<br />

überall 27 sit EN + - + - + + s - - -<br />

umher 27 sit EN - - - + + - s - - -<br />

unten 27 sit EN + - + + + + s - - -<br />

vorne 27 sit EN + - + + + + s - - -<br />

<strong>in</strong>sgeheim (14) 28 sit RS +* -* -* -* -* -* s - - -<br />

anläßlich ... 29 sit HO + - + + + + s - - -<br />

dabei (Anlaß) 29 sit HO + - + + - -* s - - -<br />

diesbezüglich 30 sit HO +* - + + -* -* s - - -<br />

hier 30 sit HO + - + + + + s - - -<br />

entsprechend 31 sit HO + - + - + + s,man,sit,ap pre - GEN<br />

242


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

bald 33 sit HO + - + + + + s - - -<br />

eben (temporal) 33 sit HO + - - - + + s - - -<br />

endlich 33 sit HO + - - - - - s - - -<br />

gerade (jetzt) 33 sit HO +* - - +* - + s - - -<br />

gleich 33 sit HO + - +* - + + s - - -<br />

jäh (15) 33 sit RS - - -* -* +* -* s - - -<br />

jähl<strong>in</strong>gs (-) 33 sit RS + - + - + - s - - -<br />

längst (400) 33 sit RS +* - - +* - - s,neg pre - -<br />

plötzlich 33 sit HO + - + + + + s - - -<br />

schließlich 33 sit HO + - - - - - s - - -<br />

soeben (26) 33 sit HO +* - -* +* + -* s - - -<br />

s<strong>of</strong>ort (312) 33 sit HO +* - +* -* +* -* s - - -<br />

sogleich 33 sit HO + - - - -* -* s - - -<br />

unversehens 33 sit RS + - - + + - s - - -<br />

unverzüglich 33 sit HO -* - - -*& -* -* s - - -<br />

vorerst (89) 33 sit EN +* - -* - - - s - - -<br />

vorher 33 sit EN +* - + + + + s - - -<br />

zunächst (642) 33 sit RS +* - - -* - - s - - -<br />

doch (pragmatisch: durchaus) 34 sit HO + - - - +* - s - - -<br />

durchaus 34 sit HO - - - - + - s,ap pre - -<br />

erst (nicht temporal) 34 sit HO - - - - - - s - - -<br />

noch (nicht temporal) 34 sit HO - - - - - - s - - -<br />

schon (nicht temporal) 34 sit HO - - - - - - s - - -<br />

unbed<strong>in</strong>gt (27) 34? sit RS + - +* + + - s - - -<br />

wohl (adversativ/entgegen d.allg.Auffassung) 34 sit HO -*& - - +* +* -* s - - -<br />

243


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

auch (gleichfalls) 35 sit HO -* - +* -* +* -* s,man,sit,npp,ap,koord pre - -<br />

ebenfalls 35 sit HO +* - - -* + - s,npp - yes -<br />

ebenso 35 sit RS + - - - + - s,man,sit,ap pre - -<br />

genauso (ebenso) 35 sit RS + - - + + + s,man,sit,pragm pre - -<br />

gleichfalls 35 sit HO - - - - + - s - - -<br />

abermals 36 sit HO + - - - + - s - - -<br />

erneut (210) 36 sit HO +* - -*& -* + - s - - -<br />

nochmals (71) 36 sit HO + - +* +* + - s - - -<br />

wieder 36 sit HO + - - + + + s - - -<br />

wiederum 36 sit HO + - - - + + s - - -<br />

allezeit (-) 37 sit RS -? - +? - -? -? s - - -<br />

dauernd 37 sit HO + - + + + + s - - -<br />

durchweg (ausnahmslos) (52) 37 sit RS +* -* +* +* -* -* s,ap pre - -<br />

durchwegs (ausnahmslos) (1) 37 sit RS +* -* +* +* -* -* s,ap pre - -<br />

e<strong>in</strong>mal (1 Mal) 37 sit RS + - + +* + -* s - - -<br />

gelegentlich 37 sit HO + - + - -* -* s - - -<br />

häufig 37 sit HO + + + + + + s - - -<br />

immer 37 sit HO + - + + + - s,man,sit,npp pre - -<br />

immerzu 37 sit HO + - - - + - s - - -<br />

jederzeit 37 sit RS + - + - + - s - - -<br />

jedesmal 37 sit HO + - + + + - s - - -<br />

laufend 37 sit HO +* - -* -* +* + s - - -<br />

manchmal 37 sit HO + - - - + + s - - -<br />

mehrmals 37 sit RS +* - - +* + + s - - -<br />

meist 37 sit RS + - - - + + s - - -<br />

244


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

meistens 37 sit HO + - - - + + s - - -<br />

mitunter 37 sit HO + - - - - - s - - -<br />

öfters 37 sit HO + - + + + + s - - -<br />

<strong>of</strong>t 37 sit HO + + + + + + s - - -<br />

<strong>of</strong>tmals 37 sit RS + - - - + -* s - - -<br />

regelmäßig 37 sit HO + + + + + + s - - -<br />

selten 37 sit HO + + + + + + s - - -<br />

ständig 37 sit HO + + + - + + s - - -<br />

stets 37 sit HO + - - - + + s - - -<br />

ununterbrochen (7) 37 sit EN + - + - -*& -* s - - -<br />

vielfach 37 sit HO + - - - + - s - - -<br />

wiederholt 37 sit HO + - - + -*& -* s - - -<br />

zeitweise 37 sit HO + - - + -*& -* s - - -<br />

zuweilen 37 sit RS + - - - - - s - - -<br />

alle<strong>in</strong> (e<strong>in</strong>zig, nur) 38 sit HO -* -* +* +* -* - s,man,sit,npp,card both - -<br />

bloß (nur, alle<strong>in</strong>) 38 sit HO +*& - +* -* - -* s,man,sit,npp,card,ap pre yes -<br />

e<strong>in</strong>zig (nur) 38 sit HO - - -* - - - s,sit,npp pre - -<br />

lediglich 38 sit HO - - - - - - s,man,sit,npp,card,ap,coord pre - -<br />

nur (bloß; limitierend) 38 sit HO -* - + - - - s,man,sit,pragm,npp,card,coord pre yes -<br />

beizeiten (2) 39 sit RS + - + + + + s - - -<br />

bereits 39 sit HO - - - - - - s,man,sit,npp,card,ap both yes -<br />

erst 39 sit HO + - - - - - s,sit,npp,card pre - -<br />

noch 39 sit HO + - - +* -* + s,sit,npp both yes -<br />

schon 39 sit HO + - + - - - s,sit,npp both yes -<br />

245


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

früh 40 sit RS + + + + + + s - - -<br />

jahrelang 40 sit HO + - + + + - s - - -<br />

lange 40 sit HO + + + + + + s - - -<br />

tagelang 40 sit RS + - + + + - s - - -<br />

wochenlang 40 sit RS + - + + + - s - - -<br />

zeitlebens (17) 40 sit RS + - - - -*& -* s - - -<br />

kaum 41 neg HO + - - - + - s - - -<br />

ke<strong>in</strong>eswegs (224) 41 neg HO +* - +* +* +* - s,man,sit,npp,ap pre - -<br />

mite<strong>in</strong>ander 42 mod HO + - + + + - s - - -<br />

anders 43 mod RS + - + + + + s - - -<br />

besser 43 mod RS + - + + + + s - - -<br />

e<strong>in</strong>fach 43 mod RS +* + + +* +* + s - - -<br />

ganz 43 mod HO + - +* - +* +* s,man,sit,pragm,npp,ap pre - -<br />

genauso 43 mod RS + - - + + + s - - -<br />

gerade 43 mod RS -* + + + + + s - - -<br />

geradeheraus (1) 43 mod RS -* -* -* -* -* -* s - - -<br />

geradenwegs (-) 43 mod RS + - -? -? + - s - - -<br />

geradeswegs (-) 43 mod RS + - -? -? + - s - - -<br />

geradewegs (7) 43 mod RS -* -* -* -* -* -* s - - -<br />

gern 43 mod HO + - + + + - s - - -<br />

glatt (ohne weiteres) 43 mod RS +* + +* +* +* +* s - - -<br />

glattweg (ohne weiteres) (2) 43 mod RS -* -* -* -* -* -* s - - -<br />

gut 43 mod HO - + + + + + s - - -<br />

h<strong>in</strong>terrücks 43 mod RS -*& -* -* -* -* - s - - -<br />

246


Adverb Pos Class Who VF Comp Neg Grad Rheme Pred Scope Pre/Post Dist Valency<br />

kopfüber 43 mod RS -* -* -* -* -* -* s - - -<br />

laut 43 mod HO + + + + + + s - - -<br />

leicht 43 mod RS + + + + + + s,ap pre - -<br />

natürlich 43 mod RS + + + + + + s - - -<br />

obenh<strong>in</strong> (-) 43 mod RS -? - +? + + - s - - -<br />

ruhig 43 mod RS + + + + + + s - - -<br />

rundheraus (<strong>of</strong>fen) (1) 43 mod RS + - - + + - s - - -<br />

rundweg (unumwunden) (7) 43 mod RS + - - -* + - s - - -<br />

schnell 43 mod HO + + + + + + s - - -<br />

schnurstracks (5) 43 mod RS + - + -* + - s - - -<br />

sehr 43 mod HO -* - + - + - s,man,sit,pragm,ap pre - -<br />

sicher 43 mod RS -* + + + + + s - - -<br />

so 43 mod HO + - + + + + s,man,sit,pragm,ap pre - -<br />

umsonst (31) (Bedeutung=a 14 ?) 43 mod RS +* - +* + + + s - - -<br />

unumwunden 43 mod RS + - - + + - s - - -<br />

vergebens (21) (Bed=a 14 ?) 43 mod RS +* - + + + + s - - -<br />

völlig 43 mod HO -* - + - + - s,man,npp,ap pre - -<br />

wenig 43 mod HO -* + + +* + + s,sit,npp,ap pre - -<br />

damit (<strong>in</strong>strumental) 44 mod EN + - + + -* + s - - -<br />

247


8.4. CANONICAL FORM (FINAL VERSION, CF. 6.5.3)<br />

Npron/N+d+b < (A


249<br />

9. BIBLIOGRAPHY<br />

Abraham, Werner (ed.) (1982): Satzglieder im Deutschen. Vorschläge zur syntaktischen,<br />

semantischen und pragmatischen Fundierung, Tüb<strong>in</strong>gen<br />

Abraham, Werner/ Kosmeijer, Wim/Reuland, Eric (eds.) (1991): Issues <strong>in</strong> <strong>German</strong>ic Syntax,<br />

Berl<strong>in</strong>/New York<br />

Admoni, Wladimir (1970): Der deutsche Sprachbau (3. erweiterte Auflage), München<br />

Altmann, Hans (1976): Die Gradpartikeln im Deutschen - Untersuchungen zu ihrer Syntax,<br />

Semantik und Pragmatik, Tüb<strong>in</strong>gen, 1976<br />

Altmann, Hans (1978): Gradpartikel-Probleme - Zur Beschreibung von gerade, genau, eben,<br />

ausgerechnet, vor allem, <strong>in</strong>sbesondere, zum<strong>in</strong>dest, wenigstens, Tüb<strong>in</strong>gen<br />

Altmann, Hans (1981): Formen der `Herausstellung' im Deutschen: Rechtsversetzung,<br />

L<strong>in</strong>ksversetzung, freies Thema und verwandte Konstruktionen, Tüb<strong>in</strong>gen<br />

Arrivé, Michel/Gadet, François/Galmiche, Michel (1986): La grammaire d'aujourd'hui:<br />

Guide alphabétique de l<strong>in</strong>guistique française, Paris<br />

Balkan, Lorna/Chambers, Chris/L<strong>in</strong>dop, Jeremy/Maxwell, Kerry/Meijer, Siety/Underwood,<br />

Nancy (1991): Eurotra-GB Implementation Report, 1 January - 30 November 1991,<br />

CEC-Luxembourg<br />

Barth (1961): Recherche sur la fréquence et la valeur des parties du discours en français, en<br />

anglais et en espagnol, Paris<br />

Bartsch, Renate (1972): Adverbialsemantik. Die Konstitution logisch-semantischer<br />

Repräsentationen von Adverbialkonstruktionen, Frankfurt/Ma<strong>in</strong><br />

Bech, Annelise (1991): Description <strong>of</strong> the Eurotra framework. In: C. Copeland/J. Durand/S.<br />

Krauwer/B. Maegaard (ed.): Studies <strong>in</strong> Mach<strong>in</strong>e Translation and Natural Language<br />

Process<strong>in</strong>g, Volume 2: The Eurotra formal specifications, pp. 7-40, Luxembourg<br />

Bech, Annelise/Maegaard, Bente/Nygaard, Anders (1991): The Eurotra MT Formalism. In:<br />

Mach<strong>in</strong>e Translation, Vol. 6, No. 2, 83-102<br />

Becker, Tilman/Joshi, Arav<strong>in</strong>d K./Rambow, Owen (1991): Long Distance Scrambl<strong>in</strong>g and<br />

Tree Adjo<strong>in</strong><strong>in</strong>g Grammars. In: EACL Proceed<strong>in</strong>gs 1991, 21-32, Berl<strong>in</strong>


250<br />

Beckmann, Barbara Joe (1980): Underly<strong>in</strong>g <strong>Word</strong> <strong>Order</strong> - <strong>German</strong> as a VSO language,<br />

Frankfurt<br />

Behaghel, Otto (1929): Zur Stellung des Subjekts im Nebensatz des Deutschen. In:<br />

Zeitschrift für deutsches Altertum und Literatur 66 , Wiesbaden<br />

Behaghel, Otto (1932): Die deutsche Syntax. E<strong>in</strong>e geschichtliche Darstellung, 4 Bände<br />

(1923-1932), Band IV: Wortstellung, Heidelberg<br />

Bl<strong>in</strong>kenberg, Andreas (1928): L'ordre des mots en français moderne, volume I, Kopenhagen<br />

Bl<strong>in</strong>kenberg, Andreas (1933): L'ordre des mots en français moderne, volume II, Kopenhagen<br />

Brugmann, Claudia (1984): The Very Idea: A Case <strong>Study</strong> <strong>in</strong> Polysemy and Cross-Lexical<br />

Generalizations. In: David Texten/Veena Mishra/Joseph Drogo (eds.): Papers from the<br />

Parasession on Lexical Semantics, 27-28 April 1984, Chicago L<strong>in</strong>guistic Society, 21-<br />

38, Chicago<br />

Bußmann, Hadumod (1983): Lexikon der Sprachwissenschaft, Stuttgart<br />

Chomsky, Noam (1981): Lectures on Government and B<strong>in</strong>d<strong>in</strong>g. Studies <strong>in</strong> Gnerative<br />

Grammar 9, Dordrecht<br />

Coll<strong>in</strong>s (1981) <strong>German</strong>-English English-<strong>German</strong> Dictionary, London/Glasgow/Stuttgart<br />

Conlon, Sumali P<strong>in</strong>-Ngern/Martha Evens (1992), Can Computers Handle Adverbs?. In:<br />

Proceed<strong>in</strong>gs <strong>of</strong> the 15th International Conference on Computational L<strong>in</strong>guistics<br />

(COLING-92), 1192-1196, Nantes<br />

Contreras, Heles (1976): A Theory <strong>of</strong> <strong>Word</strong> <strong>Order</strong> <strong>with</strong> <strong>Special</strong> Reference to Spanish,<br />

Amsterdam<br />

D<strong>in</strong>ser, Gudula (1974) (ed.): Zur Theorie der Sprachveränderung, Kronberg<br />

Dorr, Bonnie (1990): Solv<strong>in</strong>g Thematic Divergences <strong>in</strong> Mach<strong>in</strong>e Translation. In: 28 th<br />

Annual Meet<strong>in</strong>g <strong>of</strong> the Association for Computational L<strong>in</strong>guistics (ACL). Proceed<strong>in</strong>gs<br />

<strong>of</strong> the Conference, 127-134, Pittsburgh<br />

Dreyer, Hilke/Schmitt, Richard (1985): Lehr- und Übungsbuch der deutschen Grammatik,<br />

München<br />

Duden (1984): Grammatik der deutschen Gegenwartssprache (Band 4); völlig neu<br />

bearbeitete und erweiterte Auflage; Editor: Günter Drosdowski, Mannheim


251<br />

Durand, Jacques/Bennett, Paul/Allegranza, Valerio/Eynde, Frank Van/Humphreys,<br />

Lee/Schmidt, Paul/Ste<strong>in</strong>er, Erich (1991): The Eurotra L<strong>in</strong>guistic specifications: An<br />

Overview. In: Mach<strong>in</strong>e Translation, Volume 6, No. 2, 103-147<br />

Eisenberg, Peter (1989): Grundriß der deutschen Grammatik; zweite überarbeitete und<br />

erweiterte Auflage, Stuttgart<br />

Engel, Ulrich (1970): Regeln zur Wortstellung. In: Forschungsberichte des Instituts für<br />

deutsche Sprache 5, 7-148, Mannheim<br />

Engel, Ulrich (1973): Zur Abfolge der Adverbialia im deutschen Verbalsatz. E<strong>in</strong> Beitrag zur<br />

Wortstellung. In: Gerhard Nickel (ed.): Angewandte Sprachwissenschaft und<br />

Deutschunterricht, 168-191, München<br />

Engel, Ulrich (1988): Deutsche Grammatik, Heidelberg<br />

Engelkamp, Judith/Erbach, Gregor/Uszkoreit, Hans (1992): Handl<strong>in</strong>g L<strong>in</strong>ear Precedence<br />

Constra<strong>in</strong>ts by Unification. In: ACL Proceed<strong>in</strong>gs 1992, 201-208, Newark<br />

Erbach, Gregor (1993): Us<strong>in</strong>g Preference Values <strong>in</strong> Typed Feature Structures to Exploit<br />

Non-Absolute Constra<strong>in</strong>ts for Disambiguation. In: Harald Trost (ed.), 173-186,<br />

Chichester<br />

Erben, Johannes (1972): Deutsche Grammatik - E<strong>in</strong> Abriß, München<br />

Eroms, Hans Werner (1986): Funktionale Satzperspektive, Tüb<strong>in</strong>gen<br />

Fanselow, Gisbert (1987): Über Wortstellungstypologie. Anläßlich e<strong>in</strong>es Buches von John<br />

Hawk<strong>in</strong>s. In: Zeitschrift für Sprachwissenschaft 6,1, 114-133<br />

Fanselow, Gisbert/Felix, Sascha W. (1987): Sprachtheorie. E<strong>in</strong>e E<strong>in</strong>führung <strong>in</strong> die<br />

Generative Grammatik. Band 2: Die Rektions- und B<strong>in</strong>dungstheorie, Tüb<strong>in</strong>gen<br />

Foley/Van Val<strong>in</strong> (1985): Information Packag<strong>in</strong>g <strong>in</strong> the Clause. In: Timothy Shopen (ed.):<br />

Language Typology and Syntactic Description. Volume I: Clause Structure, 282-364,<br />

Cambridge<br />

Fox, Anthony (1990): The Structure <strong>of</strong> <strong>German</strong>, Oxford<br />

Gadler, Hanspeter (1982): Zur Serialisierung nom<strong>in</strong>aler Satzglieder im Mittelfeld und zur<br />

Topikalisierung. In: Abraham, Werner (ed.): Satzglieder im Deutschen. Vorschläge zur<br />

syntaktischen, semantischen und pragmatischen Fundierung (pp. 155-169).


252<br />

Givón, Talmy (1984): Syntax. A Functional-Typological Introduction, Volume I,<br />

Amsterdam/Philadelphia<br />

Gl<strong>in</strong>z, Hans (1971b): Deutsche Grammatik. Volume II: Kasussyntax - Nom<strong>in</strong>alstrukturen -<br />

Wortarten - Kasusfremdes, Frankfurt<br />

Greenbaum, Sydney (1969): Studies <strong>in</strong> English Adverbial Usage, London and Harlow<br />

Greenberg, Joseph H. (1966): Some Universals <strong>of</strong> Grammar <strong>with</strong> Particular Reference to the<br />

<strong>Order</strong> <strong>of</strong> Mean<strong>in</strong>gful Elements. In: Greenberg (ed.): Universals <strong>of</strong> Language, 73-113,<br />

Cambridge, Massachusetts<br />

Greenberg, Joseph H. (1966) (ed.): Universals <strong>of</strong> Language, Report <strong>of</strong> a Conference held at<br />

Dobbs Ferry, New York, April 13-15 1961, second edition (first edition 1963),<br />

Cambridge, Massachusetts<br />

Grevisse, Maurice (1986): Le bon usage. Grammaire française. 12 e édition refondue par<br />

André Goosse, Paris<br />

Hajicová, Eva/Sgall, Petr/Skoumalová, Hana (1993): Identify<strong>in</strong>g Topic and Focus by an<br />

Automatic Procedure. EACL Proceed<strong>in</strong>gs 1993, 178-182, Utrecht<br />

Hammond, Rob<strong>in</strong> (1981): A <strong>German</strong> Reference Grammar, Oxford<br />

Hawk<strong>in</strong>s, John A. (1986): A comparative typology <strong>of</strong> English and <strong>German</strong> - Unify<strong>in</strong>g the<br />

contrasts, London/Sydney<br />

Hawk<strong>in</strong>s, John A. (1990): A Pars<strong>in</strong>g Theory <strong>of</strong> <strong>Word</strong> <strong>Order</strong> Universals. In: L<strong>in</strong>guistic<br />

Inquiry, Volume 22/II, 223-261<br />

Heidolph, Karl-Erich/Flämig, W./Motsch, Wolfgang (1981): Grundzüge e<strong>in</strong>er deutschen<br />

Grammatik (Akademiegrammatik), Berl<strong>in</strong><br />

Helbig, Gerhard/Buscha, Joachim (1988): Deutsche Grammatik. E<strong>in</strong> Handbuch für den<br />

Ausländerunterricht, Leipzig<br />

Her<strong>in</strong>ger, Hans-Jürgen/Strecker, B./Wimmer, R. (1980): Syntax, Stuttgart<br />

Hoberg, Ursula (1981): Die Wortstellung <strong>in</strong> der geschriebenen deutschen<br />

Gegenwartssprache. In: L<strong>in</strong>guistische Grundlagen - Forschungen des Instituts für<br />

deutsche Sprache 10; Editors: Ulrich Engel, Horst Sitta, Hugo Steger; München


253<br />

Höhle, Tilman, N. (1982): Explikation für "normale Betonung" und "normale Wortstellung".<br />

In: Werner Abraham (ed.): Satzglieder im Deutschen. Vorschläge zur syntaktischen,<br />

semantischen und pragmatischen Fundierung, 75-153, Tüb<strong>in</strong>gen<br />

Hutch<strong>in</strong>s, W. John/Somers, Harold L. (1992): An Introduction to Mach<strong>in</strong>e Translation,<br />

London<br />

Jacobs, Joachim (1982): Syntax und Semantik der Negation im Deutschen, München<br />

Jacobs, Joachim (1988): Probleme der freien Wortstellung im Deutschen. In: Sprache und<br />

Pragmatik - Arbeitsberichte, 8-37, Lund<br />

Jung, Walter (1971): Grammatik der deutschen Sprache, Leipzig<br />

Kempson, Ruth M. (1977): Semantic Theory, Cambridge<br />

Kerpedjiev, Stephan M. (1992): Automatic Generation <strong>of</strong> Multimodal Weather Reports from<br />

Datasets, In: Proceed<strong>in</strong>gs <strong>of</strong> the Third Conference on Applied Language Process<strong>in</strong>g<br />

(ACL), 48-55, Trento<br />

Lehmann, W<strong>in</strong>fred P. (ed.) (1978a): Syntactic typology. Studies <strong>in</strong> the phenomenology <strong>of</strong><br />

language, Sussex<br />

Lehmann, W<strong>in</strong>fred P. (1978b): The Great Underly<strong>in</strong>g Ground-Plans. In: W<strong>in</strong>fred P.<br />

Lehmann (ed.) (1978a), 3-56, Sussex<br />

Lehmann, W<strong>in</strong>fred P. (1978c): Conclusion: Towards an Understand<strong>in</strong>g <strong>of</strong> the Pr<strong>of</strong>ound<br />

Unity Underly<strong>in</strong>g Languages. In: W<strong>in</strong>fred P. Lehmann (ed.) (1978a), 395-432, Sussex<br />

Lenerz, Jürgen (1977): Zur Abfolge nom<strong>in</strong>aler Satzglieder im Deutschen, Tüb<strong>in</strong>gen<br />

Li, Charles N. (ed.) (1975): <strong>Word</strong> <strong>Order</strong> and <strong>Word</strong> <strong>Order</strong> Change, Aust<strong>in</strong><br />

Liebsch, Helmut/Doer<strong>in</strong>g, Hellmut (eds.) (1976): Deutsche Sprache, von e<strong>in</strong>em<br />

Autorenkollektiv unter der Leitung von Liebsch/Doer<strong>in</strong>g, Leipzig<br />

L<strong>in</strong>dop, Jeremy/Tsujii, Jun-ichi (1991): Complex Transfer <strong>in</strong> MT: A Survey <strong>of</strong> Examples,<br />

CCL/UMIST Report No. 91/5, UMIST, Manchester<br />

Lötscher, Andreas (1981): Abfolgeregeln für Ergänzungen im Mittelfeld. In: Deutsche<br />

Sprache. Zeitschrift für Theorie, Praxis, Dokumentation, 44-60, Berl<strong>in</strong>, Tüb<strong>in</strong>gen


254<br />

Lötscher, Andreas (1983): Satzakzent und Funktionale Satzperspektive im Deutschen,<br />

Tüb<strong>in</strong>gen<br />

Lyons, John (1971): E<strong>in</strong>führung <strong>in</strong> die moderne L<strong>in</strong>guistik. München<br />

Lyons, John (1977): Semantics. Volume II, Cambridge<br />

Macdonald, N.H. (1983): The UNIX Writer's Workbench S<strong>of</strong>tware: Rationale and Design.<br />

In: The Bell System Technical Journal, Vol. 62, No. 6, July-August 1983, USA<br />

Mathesius, Vilém (1929): Zur Satzperspektive im modernen Englisch. In: Archiv für das<br />

Studium der neueren Sprachen und Literaturen, volume 155 (volume 55 <strong>of</strong> the new<br />

series), 202-210, Braunschweig/Berl<strong>in</strong>/Bad Homburg<br />

Meier, Helmut (1978): Deutsche Sprachstatistik, Hildesheim<br />

Mesli, Nadia (1991): Funktionsverbgefüge <strong>in</strong> der masch<strong>in</strong>ellen Analyse und Übersetzung:<br />

l<strong>in</strong>guistische Beschreibung und Implementierung im CAT2-Formalismus, Eurotra<br />

Work<strong>in</strong>g Papers No. 19, IAI, Saarbrücken<br />

Oliva, Karel (1991): On Cases <strong>of</strong> "Fixed" <strong>Word</strong> <strong>Order</strong> <strong>in</strong> a "Free" <strong>Word</strong> <strong>Order</strong> Language.<br />

CLAUS-Report Nr. 14, CL, Saarbrücken<br />

Oliva, Karel (1992a): The proper Treatment <strong>of</strong> <strong>Word</strong> <strong>Order</strong> <strong>in</strong> HPSG. In: Proceed<strong>in</strong>gs <strong>of</strong> the<br />

15th International Conference on Computational L<strong>in</strong>guistics (COLING-92), Vol. I,<br />

184-190, Nantes<br />

Oliva, Karel (1992b): <strong>Word</strong> <strong>Order</strong> Constra<strong>in</strong>ts <strong>in</strong> B<strong>in</strong>ary Branch<strong>in</strong>g Syntactic Structures.<br />

CLAUS Report 20, CL, Saarbrücken<br />

Ortmann, Wolf Dieter (1975): Hochfrequente deutsche Wortformen (3 volumes), München<br />

Payne, John R. (1985): Negation. In: Timothy Shopen (ed.) (1985), 197-242<br />

Pelz, Heidrun (1963): Das französische qualifizierende Adverb und se<strong>in</strong>e Übersetzung im<br />

Englischen und im Deutschen, Tüb<strong>in</strong>gen<br />

Pollard, Carl/Sag, Ivan A. (1987): Information-based Syntax and Semantics - Volume I:<br />

Fundamentals, Stanford<br />

Primus, Beatrice (1987): Grammatische Hierarchien. E<strong>in</strong>e Beschreibung und Erklärung von<br />

Regularitäten des Deutschen ohne grammatische Relationen, München


255<br />

Reis, Marga (1987): Die Stellung der Verbargumente im Deutschen. Stilübungen zum<br />

Grammatik:Pragmatik-Verhältnis. In: Inger Rosengren (ed.): Sprache und Pragmatik,<br />

Lunder Symposium 1986, 139-178, Lund<br />

Reiter, Ehud / Mellish, Chris / Lev<strong>in</strong>e, John (1992): Automatic Generation <strong>of</strong> On-L<strong>in</strong>e<br />

Documentation <strong>in</strong> the IDAS Project. In: Proceed<strong>in</strong>gs <strong>of</strong> the Third Conference on<br />

Applied Language Process<strong>in</strong>g (ACL), 64-71, Trento<br />

Rochemont, Michael (1989): Implement<strong>in</strong>g Focus <strong>in</strong> Mach<strong>in</strong>e Translation. Eurotra-D<br />

Work<strong>in</strong>g Papers No. 9, IAI, Saarbrücken<br />

Russon, A. & Russon, L.J. (1978): Advanced <strong>German</strong> Course, Revised Edition, Harlow<br />

(Essex)<br />

Sampson, Ge<strong>of</strong>frey (1987): Evidence aga<strong>in</strong>st the "grammatical"/"ungrammatical"<br />

dist<strong>in</strong>ction. In: Meijs, Willem (ed.): Corpus L<strong>in</strong>guistics and beyond. Proceed<strong>in</strong>gs <strong>of</strong> the<br />

Seventh International Conference on English Language Research on Computerized<br />

Corpora, 219-226, Amsterdam<br />

Scaglione, Aldo D. (1981): The Theory <strong>of</strong> <strong>German</strong> <strong>Word</strong> <strong>Order</strong> from the Renaissance to the<br />

Present, M<strong>in</strong>nesota<br />

Schachter, Paul (1985): Parts-<strong>of</strong>-Speech Systems. In: Shopen, Timothy (ed.): Language<br />

Typology and Syntactic Description, Volume I, pp. 3-61, Cambridge<br />

Schäufele, Steven (1991): A Note on the Term Scrambl<strong>in</strong>g. In: Natural Language &<br />

L<strong>in</strong>guistic Theory, volume 9-2, 365-368, Dordrecht/Boston/London<br />

Schulz, Dora/Griesbach, He<strong>in</strong>z (1980): Grammatik der deutschen Gegenwartssprache,<br />

München<br />

Schwartz, Bonnie D./Tomaselli, Alessandra (1991): Some Implications from an Analysis <strong>of</strong><br />

<strong>German</strong> <strong>Word</strong> <strong>Order</strong>. In: Werner Abraham/Wim Kosmeijer/Eric Reuland (eds.)<br />

(1991), 251-274, Berl<strong>in</strong>/New York<br />

Sgall, Petr (1982): Wortfolge und Fokus im Deutschen. In: Werner Abraham (1982), 59-74,<br />

Tüb<strong>in</strong>gen<br />

Sharp, Randall (1989): CAT2 - A Formalism for Multil<strong>in</strong>gual Mach<strong>in</strong>e Translation,<br />

Proceed<strong>in</strong>gs <strong>of</strong> the International Sem<strong>in</strong>ar on Mach<strong>in</strong>e Translation, Tblisi, Georgia<br />

(USSR)


256<br />

Sharp, Randall (1993): CAT2 Reference Manual, Version 3.1, Unf<strong>in</strong>ished Draft (March 19,<br />

1993), IAI, Saarbrücken<br />

Shopen, Timothy (ed.) (1985): Language Typology and Syntactic Description, Volume I:<br />

Clause Structure, Cambridge<br />

Siewierska, Anna (1988): <strong>Word</strong> <strong>Order</strong> Rules, Beckenham (Kent)<br />

Somers, H.L. / Tsujii, J. / Jones, D. (1990): Mach<strong>in</strong>e Translation <strong>with</strong>out a source text. In:<br />

Proceed<strong>in</strong>gs <strong>of</strong> the 13 th International Conference on Computational L<strong>in</strong>guistics<br />

(COLING-90), Vol.3, 271-276, Hels<strong>in</strong>ki<br />

Somers, H.L. (1992): Interactive multil<strong>in</strong>gual text generation for a monol<strong>in</strong>gual user. In:<br />

Actes du Quatrième Colloque International sur les Aspects Théoriques et<br />

Méthodologiques de la Traduction Automatique TMI-92, 35-43, Montréal<br />

Sommerfeldt, Karl-Ernst/Starke, Günther (1988): E<strong>in</strong>führung <strong>in</strong> die Grammatik der<br />

deutschen Gegenwartssprache (von e<strong>in</strong>em Autorenkollektiv unter der Leitung von<br />

Sommerfeldt/Starke, Leipzig (taken over by Max Niemeyer Verlag, Tüb<strong>in</strong>gen)<br />

Stechow, Arnim von/Sternefeld, Wolfgang (1988): Bauste<strong>in</strong>e syntaktischen Wissens. E<strong>in</strong><br />

Lehrbuch der Generative Grammatik, Opladen<br />

Ste<strong>in</strong>berger, Ralf (1990): Wortstellung und Masch<strong>in</strong>elle Übersetzung: Die Stellung von<br />

Angaben im deutschen Satz. Unpublished Magisterarbeit submitted to the Ludwig-<br />

Maximilians-Universität München<br />

Ste<strong>in</strong>berger, Ralf (1992a): Beschreibung der Adverbstellung im deutschen und englischen<br />

Satz im H<strong>in</strong>blick auf Masch<strong>in</strong>elle Übersetzung. Eurotra-D Work<strong>in</strong>g Papers No. 23,<br />

IAI, Saarbrücken<br />

Ste<strong>in</strong>berger, Ralf (1992b): Der Skopus von Gradpartikeln: Se<strong>in</strong>e Übersetzung und se<strong>in</strong>e<br />

Implementierung im Masch<strong>in</strong>ellen Übersetzungssystem CAT2. Eurotra-D Work<strong>in</strong>g<br />

Papers No. 24, IAI, Saarbrücken<br />

Ste<strong>in</strong>berger, Ralf (1993): Grenzen und Möglichkeiten der Masch<strong>in</strong>ellen Übersetzung. In:<br />

Informatik Forum - Fachzeitschrift für Informatik, Band 7, Doppelheft 1/2, 69-74,<br />

Vienna<br />

Ste<strong>in</strong>er, Erich/W<strong>in</strong>ter, Jutta (1987): The semantics <strong>of</strong> focus phenomena. In: Erich<br />

Ste<strong>in</strong>er/Jutta W<strong>in</strong>ter/Cornelia Zel<strong>in</strong>sky-Wibbelt (1987), 1-24, Saarbrücken


257<br />

Ste<strong>in</strong>er, Erich/W<strong>in</strong>ter, Jutta/Zel<strong>in</strong>sky-Wibbelt, Cornelia (1987): Aspects <strong>of</strong> Determ<strong>in</strong>ation<br />

and Focus <strong>in</strong> a Multil<strong>in</strong>gual MT System. Eurotra-D Work<strong>in</strong>g Papers No. 5, IAI,<br />

Saarbrücken<br />

Ste<strong>in</strong>itz, Renate (1969): Adverbialsyntax. Studia Grammatica X, Berl<strong>in</strong><br />

Taglicht, Josef (1984): Message and Emphasis - On Focus and Scope <strong>in</strong> English, London<br />

and New York<br />

Thurmair, Gregor (1990): Complex Lexical Transfer <strong>in</strong> METAL. In: Third International<br />

Conference on Theoretical and Methodological Issues <strong>in</strong> Mach<strong>in</strong>e Translation, Aust<strong>in</strong>,<br />

Texas<br />

Thurmair, Maria (1989): Modalpartikeln und ihre Komb<strong>in</strong>ationen, Tüb<strong>in</strong>gen<br />

Toml<strong>in</strong>, Russell S. (1986): Basic <strong>Word</strong> <strong>Order</strong>: Functional Pr<strong>in</strong>ciples, Beckenham (Kent)<br />

Trost, Harald (ed.) (1993): Feature Formalisms and L<strong>in</strong>guistic Ambiguity, Chichester<br />

Uszkoreit, Hans (1987): <strong>Word</strong> order and constituent structure <strong>in</strong> <strong>German</strong>. CSLI Lecture<br />

Notes No. 8, Stanford<br />

Uszkoreit, Hans (1991): Strategies for add<strong>in</strong>g control <strong>in</strong>formation to declarative grammars.<br />

In: ACL Proceed<strong>in</strong>gs 1991, 237-245, Berkeley<br />

Vennemann, Theo (1974): Zur Theorie der Wortstellungsveränderung: von SXV zu SVX<br />

über TVX. In: Gudula D<strong>in</strong>ser (ed.): Zur Theorie der Sprachveränderung, Kronberg<br />

Vennemann, Theo (1975): An Explanation <strong>of</strong> Drift. In: Charles N. Li (ed.): <strong>Word</strong> <strong>Order</strong> and<br />

<strong>Word</strong> <strong>Order</strong> Change, 269-305, Aust<strong>in</strong><br />

Vennemann, Theo (1977): Konstituenz und Dependenz <strong>in</strong> e<strong>in</strong>igen neueren<br />

Grammatiktheorien. In: Sprachwissenschaft 2, 259-301<br />

Vennemann, Theo (1982): Deutsche, englische und koreanische Wortstellungssyntax aus<br />

typologischer Sicht. Sonderdruck aus: Zeitschrift für deutsche Sprache und Literatur,<br />

Nr. 17, März 1982, Seoul, Korea<br />

Wagner, R.L./P<strong>in</strong>chon, J. (1962): Grammaire du français classique et moderne (Ouvrage<br />

couronné par l'Académie Française), Paris<br />

Wahrig, Gerhard et al. (1986): Deutsches Wörterbuch, München


258<br />

Waltz<strong>in</strong>g, Raymond (1986): Existimatorische Angaben - E<strong>in</strong>e semanto-syntaktische<br />

Untersuchung bestimmter Elemente des deutschen Satzes und ihrer französischen<br />

Entsprechungen, Frankfurt/Ma<strong>in</strong><br />

Weydt, Harald (1969): Abtönungspartikel - Die deutschen Modalwörter und ihre<br />

französischen Entsprechungen, Bad Homburg<br />

Weydt, Harald (ed.) (1977): Aspekte der Modalpartikeln. Studien zur deutschen Abtönung,<br />

Tüb<strong>in</strong>gen<br />

Whitley, Melv<strong>in</strong> Stanley (1986): Spanish/English Contrasts - A Course <strong>in</strong> Spanish<br />

L<strong>in</strong>guistics, Wash<strong>in</strong>gton<br />

Whittemore, Greg/Ferrara, Kathleen/Brunner, Hans (1990): Empirical <strong>Study</strong> <strong>of</strong> Predictive<br />

Powers <strong>of</strong> Simple Attachment Schemes for Post-modifier Prepositional Phrases. In:<br />

28 th Annual Meet<strong>in</strong>g <strong>of</strong> the Association for Computational L<strong>in</strong>guistics (ACL).<br />

Proceed<strong>in</strong>gs <strong>of</strong> the Conference, 23-30, Pittsburgh<br />

Zähner, Christoph/Gupta, Gautam/Ste<strong>in</strong>berger, Ralf (forthcom<strong>in</strong>g): An Onl<strong>in</strong>e Lexicon for<br />

<strong>German</strong> Second-Language Acquisition, CCL/UMIST Report No. 94/1, UMIST,<br />

Manchester<br />

Zemb, Jean-Marie (1968): Les structures logiques de la phrase allemande. Contribution à<br />

l'étude des rapports entre la langue et la pensée, Paris: O.C.D.L.<br />

Zonnefeld, Ron van (1991): Syntactic Nom<strong>in</strong>alization. In: Abraham, Werner/Kosmeijer,<br />

Wim/Reuland, Eric (eds.): Issues <strong>in</strong> <strong>German</strong>ic Syntax, 135-160. Trends <strong>in</strong> L<strong>in</strong>guistics -<br />

Studies and Monographs 44 (ed.: Werner W<strong>in</strong>ter), Berl<strong>in</strong>

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!