Ebook efficiency and complexity in grammars part 1

John Hawkins has long been a trail blazer in the attempt to reconcile the results of formal and functional linguistics Efficiency and Complexity in Grammars charts new territory in this domain The boo[.]

John Hawkins has long been a trail-blazer in the attempt to reconcile the results

of formal and functional linguistics Efficiency and Complexity in Grammars

charts new territory in this domain The book argues persuasively that a small number of performance-based principles combine to account for many gram-matical constraints proposed by formal linguists and also explain the origins

of numerous typological generalizations discovered by functionalists

Frederick J Newmeyer, University of Washington.

The central claim in Hawkins’s new book is that grammar facilitates language processing This rather natural idea is by no means novel: attempts to explain aspects of linguistic structure on the basis of processing considerations go back

at least to the 1950s But such attempts have characteristically been little more than “just so stories” – that is, post hoc accounts of isolated observations What has been lacking until now is anything that could be called a theory of how constraints on the human processor shape grammatical structure

Hawkins has filled this lacuna Starting with three very general and intuitive principles about efficient processing of language, he derives a rich array of predictions about what kinds of grammatical structures should be preferred

He then adduces a wealth of evidence to demonstrate that his predictions hold His data are of a variety of types, including grammatical patterns in particu-lar languages, typological tendencies, usage statistics from corpora, historical changes, and psycholinguistic findings The phenomena he deals with are sim-ilarly varied, including word order, case making, filler-gap dependencies, island constraints, and anaphoric binding

Efficiency and Complexity in Grammars is a landmark work, setting a new

standard in the study of the relationship between linguistic competence and performance

Tom Wasow, Stanford University.

Hawkins argues that grammars are profoundly affected by the way humans process language He develops a simple but elegant theory of performance and grammar by drawing on concepts and data from generative grammar, lin-guistic typology, experimental psycholinlin-guistics and historical linlin-guistics In

so doing, he also makes a laudable attempt to bridge the schism between the

two research traditions in linguistics, the formal and the functional Efficiency

and Complexity in Grammars is a major contribution with far-reaching

con-sequences and implications for many of the fundamental issues in linguistic theory This is a tremendous piece of scholarship that no linguist can afford to neglect

Jae Jung Song, University of Otago.

This page intentionally left blank

Efficiency and Complexity in Grammars

JOHN A HAWKINS

1

Great Clarendon Street, Oxford ox2 6dp

Oxford University Press is a department of the University of Oxford.

It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in

Oxford New York

Auckland Bangkok Buenos Aires Cape Town Chennai

Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata

Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi São Paulo Shanghai Taipei Tokyo Toronto

Oxford is a registered trade mark of Oxford University Press

in the UK and in certain other countries

Published in the United States

by Oxford University Press Inc., New York

© John A Hawkins

The moral rights of the author have been asserted

Database right Oxford University Press (maker)

First published 2004

All rights reserved No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press,

or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above

You must not circulate this book in any other binding or cover

and you must impose this same condition on any acquirer

A catalogue record for this title is available from the British Library Library of Congress Cataloging in Publication Data

Data available

ISBN 0–19–925268–8 (hbk.)

ISBN 0–19–925269–6 (pbk.)

Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India

Printed in Great Britain on acid-free paper by

Biddles Ltd., King’s Lynn

To Kathryn and Kirsten,

who delayed this book beautifully

This page intentionally left blank

1.1 Performance–grammar correspondences: a hypothesis 1

1.5 The challenge of multiple preferences 13

2.2 Property assignments in combinatorial and

2.3 Efficiency and complexity in form–property signaling 25

3.2.3 Maximize the ease of processing enrichments 44

3.3.1 Unassignments and misassignments 51

3.3.3 Predictions for performance and grammars 58

4.2.1 Morphological inventory predictions 69 4.2.2 Declining distinctions predictions 73

4.4 The grammaticalization of definiteness marking 82

www.IELTS4U.blogfa.com

viii Contents

4.5 Processing enrichments through structural parallelism 93 4.6 The principle of conventionalized dependency 97

5.1 EIC preferences for adjacency in performance 104

5.2 Multiple preferences for adjacency in performance 111 5.2.1 Multiple preferences in English 111 5.2.2 Multiple preferences in Japanese 118

5.3 EIC preferences for adjacency in grammars 123

5.4 Multiple preferences for adjacency in grammars 131 5.5 Competitions between domains and phrases 136 5.5.1 Relative clause extrapositions in German 142

6.1 Minimal formal marking in performance 148

6.3 Morphological typology and Sapir’s ‘drift’ 166

7.1 The grammar and processing of filler–gap dependencies 171 7.2 The Keenan–Comrie Accessibility Hierarchy 177 7.2.1 Performance support for the FGD complexity ranking 180 7.2.2 Grammatical support for the FGD complexity ranking 186 7.3 Wh-fronting and basic word order 190

7.4.2 Reduce additional syntactic processing 197 7.4.3 Reduce additional semantic processing 201

7.5.2 Relative clause ordering asymmetries 205 7.5.3 Grammatical conventions that facilitate filler–gap

7.6 That-trace in English and processing enrichments in Japanese 215

www.IELTS4U.blogfa.com

Contents ix

8.1 Some cross-linguistic generalizations 224

8.2.2 Subjects precede direct objects 228 8.2.3 Topic to the left of a dependent predication 235 8.2.4 Restrictive before appositive relatives 240

8.4 A hypothesis for symmetries and asymmetries 244

8.6 Processing in relation to antisymmetry in formal grammar 251

9.2 The performance basis of grammatical generalizations 259 9.3 The ultimate causality of the performance–grammar

www.IELTS4U.blogfa.com

This page intentionally left blank

www.IELTS4U.blogfa.com

Has performance had any significant impact on the basic design features of grammars? Putting this another way: do the variation patterns that we observe

in the world’s 6,000 languages point to a causal role for principles of pro-cessing and use? One tradition of research, following Chomsky’s theory of

an autonomous and ultimately innate grammar, answers this question in the negative Other researchers, some following Greenberg’s early work on performance and cross-linguistic markedness hierarchies, answer it in the affirmative These two research traditions, the formal and the functional, use different methodologies and they formulate different kinds of linguistic generalizations, and this makes it hard for an uncommitted observer to assess their respective arguments Compounding this difficulty has been the absence for many years of systematic data from performance, on the basis

of which we could test whether principles of structural selection and pro-cessing have left any imprint on grammars and grammatical variation The field of experimental psycholinguistics has now reached the point, however, where we have a growing body of performance data from English and cer-tain other languages, and the advent of corpora has made available large quantities of usage data that can be accessed in the pursuit of theoretical questions

The time has come when we can return to the big question about the role of performance in explaining grammars and give some answers based not on philosophical speculation but on the growing body of empirical data from grammars and from performance Do these two sets of data cor-respond or do they not? Are distributional patterns and preferences that

we find in the one found in the other? If such correspondences can be found, this will provide evidence against the immunity of grammars to per-formance If, moreover, the properties of grammars that can be linked to patterns and preferences in performance include the very parameters and con-straints of Chomskyan Universal Grammar, then we will have evidence for a strong causal role for performance in explaining the basic design features of grammars

I argue in this book that there is a profound correspondence between per-formance and grammars, and I show this empirically for a large number of syntactic and morphosyntactic properties and constructions Specifically the

www.IELTS4U.blogfa.com

xii Preface

data of this book support the following hypotheses and conclusions:

• Grammars have conventionalized syntactic structures in proportion to their degrees of preference in performance, as evidenced by patterns

of selection in corpora and by ease of processing in psycholinguistic experiments (the ‘performance–grammar correspondence hypothesis’)

• These common preferences of performance and grammars are structured

by general principles of efficiency and complexity that are clearly visible in both usage data and grammatical conventions Three of these principles are defined and illustrated here: Minimize Domains, Minimize Forms, and Maximize On-line Processing

• Greater descriptive and explanatory adequacy can be achieved when effi-ciency and complexity principles are incorporated into the theory of grammar; stipulations are avoided, many exceptions can be explained, and improved formalisms incorporating significant generalizations from both performance and grammars can be proposed

• Psycholinguistic models need to broaden the explanatory basis for many performance preferences beyond working-memory load and capacity con-straints The data presented here point to multiple factors and to degrees

of preference that operate well within working memory limits, while some preferred structures actually increase working memory load as currently defined

• The innateness of human language resides primarily in mechanisms for processing and for learning The innateness of grammar is reduced to the extent that efficiency and complexity provide a more adequate description

of the facts, in conjunction with a theory of adaptation and change and the performance–grammar correspondence proposed here

• The language sciences are currently fragmented into often mutually indif-ferent subdisciplines: generative grammar, typology, psycholinguistics, and historical linguistics It is important, if we are to advance to the next stage

of descriptive adequacy and if we are to make progress in understanding why grammars are the way they are, that we try to integrate key findings and insights from each of these areas

I realize that these conclusions will be unwelcome to many, especially those with philosophical commitments to the status quo But the current com-partmentalization in our field and the absence of any real exchange of ideas and generalizations between many of the research groups is not satisfactory Peer-group conformist pressures also encourage acceptance rather than crit-ical assessment and testing of ideas that have become almost dogmatic There needs to be a reassessment of the grammar–performance relationship at this

www.IELTS4U.blogfa.com

Preface xiii point And in particular somebody needs to juxtapose the kinds of data and generalizations that these different fields have discovered and see whether there

is, or is not, some unity that underlies them all My primary goal is to attempt

to do this And my finding is that there is a deep correspondence between per-formance data and grammars and that grammatical theorizing needs to take account of this, both descriptively and at an explanatory level

There are so many people that I am indebted to for ideas and assistance in writing this book that I have decided not to list names at the outset but to make very clear in the text whose contributions I am using and how I have been fortunate over the years to have had colleagues and mentors in typology, formal grammar, psycholinguistics, and historical linguistics without whom I could not have undertaken the kind of synthesis I am attempting here At an institutional level I must mention the German Max Planck Society which has generously supported my work over a long period, first at the psycholinguistics institute in Nijmegen and more recently at the evolutionary anthropology institute in Leipzig Most of this book was written in Leipzig and I am grateful

to this institute, and to its co-director Bernard Comrie in particular, for the opportunity to complete it there The University of Southern California in Los Angeles has also supported me generously over many years

Leipzig 2003www.IELTS4U.blogfa.comJAH

AdjP adjective phrase

Adjwk weak adjective

AH accessibility hierarchy

Classif classifier

CNPC complex noun phrase constraint

CRD constituent recognition domain

CV consonant-vowel (syllable)

CVC consonant-vowel-consonant (syllable)

Dem demonstrative determiner

EIC early immediate constituents

Fams families (of languages)

F : P form–property pairing

www.IELTS4U.blogfa.com

Abbreviations xv

L to R left to right

m (mXP/XPm) a mother-node-constructing category on the left/right

periphery of XP

MaOP maximize on-line processing

NRel noun before relative clause

OCOMP object of comparison

OP/UP on-line property to ultimate property (ratios)

OVS object before verb before subject

P preposition or postposition

www.IELTS4U.blogfa.com

xvi Abbreviations

PCD phrasal combination domain

Pd dependent prepositional phrase

PGCH performance–grammar correspondence hypothesis

Pi independent prepositional phrase

Possp possessive phrase

PP prepositional phrase

P-set pragmatic set

R-agr rich agreement

R-case rich case marking

Rel relative (clause)

RelN relative clause before noun

RelPro relative pronoun

R-pronoun reflexive pronoun

R to L right to left

SOV subject before object before verb

SVO subject before verb before object

TDD total domain differential

www.IELTS4U.blogfa.com

Abbreviations xvii

VOS verb before object before subject

VSO verb before subject before object

www.IELTS4U.blogfa.com

This page intentionally left blank

www.IELTS4U.blogfa.com

Introduction

An interesting general correlation appears to be emerging between perform-ance and grammars, as more data become available from each There are patterns of preference in performance in languages possessing several struc-tures of a given type These same preferences can also be found in the fixed conventions of grammars, in languages with fewer structures of the same type The performance data come from corpus studies and processing experiments, the grammatical data from typological samples and from the growing number

of languages that have now been subjected to in-depth formal analysis The primary goal of this book is to explore this correlation in a broad range

of syntactic and morphosyntactic data I will argue that many of these common preferences of performance and grammars can be explained by efficiency and complexity, and some general and predictive principles will be defined that give substance to this claim In this introductory chapter I define my goals and show how they are relevant to current issues in linguistics and psycholinguistics

1.1 Performance–grammar correspondences: a hypothesis

An early example of the correlation between grammars and performance data can be found in Greenberg’s (1966) book on feature hierarchies such as Singular > Plural > Dual and Nominative > Accusative > Dative Morpho-logical inventories across languages, declining allomorphy and increased formal marking all provided evidence for the hierarchies, while declining frequencies of use for lower positions on each hierarchy, in languages like Sanskrit with productive morphemes of each type, showed a clear performance correlation with the patterns of grammars

Another early example, involving syntax, was proposed by Keenan & Comrie (1977) when motivating their Accessibility Hierarchy (SU>DO>IO/OBL> GEN) for cross-linguistic relativization patterns They argued that this gram-matical hierarchy correlated with the processing ease of relativizing on these different positions and with corpus frequencies in a single language (English)

www.IELTS4U.blogfa.com