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 grammatical 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 particular languages, typological tendencies, usage statistics from corpora, historical changes, and psycholinguistic findings The phenomena he deals with are similarly 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, linguistic typology, experimental psycholinguistics and historical linguistics 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 consequences 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 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 Contents Preface Abbreviations xi xiv Introduction 1.1 Performance–grammar correspondences: a hypothesis 1.2 Predictions of the PGCH 1.3 Efficiency and complexity 1.4 Issues of explanation 1.5 The challenge of multiple preferences 1 10 13 Linguistic Forms, Properties, and Efficient Signaling 2.1 Forms and properties 2.2 Property assignments in combinatorial and dependency relations 2.3 Efficiency and complexity in form–property signaling 15 15 18 25 Defining the Efficiency Principles and their Predictions 3.1 Minimize Domains (MiD) 3.2 Minimize Forms (MiF) 3.2.1 The logic of MiF 3.2.2 Form minimization predictions 3.2.3 Maximize the ease of processing enrichments 3.3 Maximize On-line Processing (MaOP) 3.3.1 Unassignments and misassignments 3.3.2 The quantitative metric 3.3.3 Predictions for performance and grammars 31 31 38 38 41 44 49 51 55 58 More on Form Minimization 4.1 Greenberg’s markedness hierarchies 4.2 Markedness hierarchies in diachrony 4.2.1 Morphological inventory predictions 4.2.2 Declining distinctions predictions 4.3 Grammaticalization and processing 4.4 The grammaticalization of definiteness marking 4.4.1 Semantic/pragmatic extensions 4.4.2 Syntactic extensions 63 64 68 69 73 79 82 84 86 www.IELTS4U.blogfa.com viii Contents 4.5 Processing enrichments through structural parallelism 4.6 The principle of conventionalized dependency 93 97 Adjacency Effects Within Phrases 5.1 EIC preferences for adjacency in performance 5.1.1 EIC in head-initial structures 5.1.2 EIC in head-final structures 5.2 Multiple preferences for adjacency in performance 5.2.1 Multiple preferences in English 5.2.2 Multiple preferences in Japanese 5.2.3 Total domain differentials 5.3 EIC preferences for adjacency in grammars 5.3.1 The Greenbergian correlations 5.3.2 Other ordering universals 5.4 Multiple preferences for adjacency in grammars 5.5 Competitions between domains and phrases 5.5.1 Relative clause extrapositions in German 103 104 104 108 111 111 118 119 123 123 127 131 136 142 Minimal Forms in Complements/Adjuncts and Proximity 6.1 Minimal formal marking in performance 6.1.1 Wh, that/zero relativizers 6.1.2 Other alternations 6.2 Minimal formal marking in grammars 6.3 Morphological typology and Sapir’s ‘drift’ 147 148 148 154 159 166 Relative Clause and Wh-movement Universals 7.1 The grammar and processing of filler–gap dependencies 7.2 The Keenan–Comrie Accessibility Hierarchy 7.2.1 Performance support for the FGD complexity ranking 7.2.2 Grammatical support for the FGD complexity ranking 7.3 Wh-fronting and basic word order 7.4 Other complexity hierarchies 7.4.1 A clause-embedding hierarchy 7.4.2 Reduce additional syntactic processing 7.4.3 Reduce additional semantic processing 7.5 MaOP effects 7.5.1 Fillers First 7.5.2 Relative clause ordering asymmetries 7.5.3 Grammatical conventions that facilitate filler–gap processing 7.6 That-trace in English and processing enrichments in Japanese 169 171 177 180 186 190 192 193 197 201 203 203 205 www.IELTS4U.blogfa.com 210 215 Contents Symmetries, Asymmetric Dependencies, and Earliness Effects 8.1 Some cross-linguistic generalizations 8.1.1 Symmetries 8.1.2 Asymmetries 8.2 Asymmetric dependencies 8.2.1 Antecedent precedes anaphor 8.2.2 Subjects precede direct objects 8.2.3 Topic to the left of a dependent predication 8.2.4 Restrictive before appositive relatives 8.3 Symmetrical dependencies 8.4 A hypothesis for symmetries and asymmetries 8.5 Morphosyntactic asymmetries 8.6 Processing in relation to antisymmetry in formal grammar Conclusions 9.1 Support for the PGCH 9.2 The performance basis of grammatical generalizations 9.3 The ultimate causality of the performance–grammar preferences 9.4 Some further issues 9.5 Acquisition and learnability ix 223 224 224 225 226 227 228 235 240 242 244 246 251 255 256 259 265 270 272 www.IELTS4U.blogfa.com References 277 Author Index 295 Language Index 299 Subject Index 301 This page intentionally left blank www.IELTS4U.blogfa.com Preface Has performance had any significant impact on the basic design features of grammars? Putting this another way: the variation patterns that we observe in the world’s 6,000 languages point to a causal role for principles of processing 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 processing 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 certain 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 correspond or 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 performance If, moreover, the properties of grammars that can be linked to patterns and preferences in performance include the very parameters and constraints 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 performance 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 efficiency 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 constraints 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 indifferent 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 www.IELTS4U.blogfa.com I realize that these conclusions will be unwelcome to many, especially those with philosophical commitments to the status quo But the current compartmentalization 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 critical assessment and testing of ideas that have become almost dogmatic There needs to be a reassessment of the grammar–performance relationship at this 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 this And my finding is that there is a deep correspondence between performance 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 www.IELTS4U.blogfa.com Leipzig 2003 JAH Abbreviations A Abs Acc Accus Adj AdjP Adjwk AdvP Agr AH Anaph Ant C CL Classif CNPC Comp CRD CV CVC Dat Def Dem Det DO DP EIC Erg F Fams Fem FGD F:P appositive absolutive accusative accusative adjective adjective phrase weak adjective adverb phrase agreement accessibility hierarchy anaphor antecedent complementizer classifier classifier complex noun phrase constraint complementizer constituent recognition domain consonant-vowel (syllable) consonant-vowel-consonant (syllable) dative definite determiner demonstrative determiner determiner direct object determiner phrase early immediate constituents ergative form families (of languages) feminine filler–gap domain form–property pairing www.IELTS4U.blogfa.com Abbreviations Fut G Gen HPD IC IO IP L LD Lgs L to R M m (mXP/XPm) Masc MaOP MHG Mid MiD Mod MiF N Neut NHG Nom Nomin NP NRel O OBL OCOMP OHG OP Op OP/UP OT OV OVS P future genitive genitive head–pronoun domain immediate constituent indirect object inflection phrase lenition lexical domain languages left to right mother node a mother-node-constructing category on the left/right periphery of XP masculine maximize on-line processing Middle High German middle minimize domains modern minimize forms noun neuter New High German nominative nominative noun phrase noun before relative clause object oblique object of comparison Old High German on-line property operator on-line property to ultimate property (ratios) optimality theory object before verb object before verb before subject preposition or postposition xv www.IELTS4U.blogfa.com xvi Abbreviations P Part PCD Pd Perf PGCH Pi Pl Plur Poss Possp PP Prep Pro P-set Qu R R-agr R-case Rel RelN RelPro Relv R-pronoun RQ R to L S S S′ Sg Sing SOV SVO SU TDD θ UG UP V property particle phrasal combination domain dependent prepositional phrase perfect(ive) performance–grammar correspondence hypothesis independent prepositional phrase plural plural possessive possessive phrase prepositional phrase preposition pronoun pragmatic set quantifier restrictive rich agreement rich case marking relative (clause) relative clause before noun relative pronoun relativizer reflexive pronoun requires right to left sentence subject sentence bar singular singular subject before object before verb subject before verb before object subject total domain differential theta (role) universal grammar ultimate property verb www.IELTS4U.blogfa.com Abbreviations VO VOS VP VSO XP xvii verb before object verb before object before subject verb phrase verb before subject before object an arbitrary phrase www.IELTS4U.blogfa.com This page intentionally left blank www.IELTS4U.blogfa.com Introduction An interesting general correlation appears to be emerging between performance and grammars, as more data become available from each There are patterns of preference in performance in languages possessing several structures 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 www.IELTS4U.blogfa.com 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 Morphological 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 grammatical hierarchy correlated with the processing ease of relativizing on these different positions and with corpus frequencies in a single language (English) ... between domains and phrases 5.5 .1 Relative clause extrapositions in German 10 3 10 4 10 4 10 8 11 1 11 1 11 8 11 9 12 3 12 3 12 7 13 1 13 6 14 2 Minimal Forms in Complements/Adjuncts and Proximity 6 .1 Minimal formal... enrichments in Japanese 16 9 17 1 17 7 18 0 18 6 19 0 19 2 19 3 19 7 2 01 203 203 205 www.IELTS4U.blogfa.com 210 215 Contents Symmetries, Asymmetric Dependencies, and Earliness Effects 8 .1 Some cross-linguistic... 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