Noam Chomsky and Language Descriptions [1 ed.] 9789027285119, 9789027210692

Citation preview

Noam Chomsky and Language Descriptions

The Development of the Anglo-Saxon Language and Linguistic Universals (DASLU)

Volume 2 Noam Chomsky and Language Descriptions Edited by John Ole Askedal, Ian Roberts and Tomonori Matsushita

Noam Chomsky and Language Descriptions Edited by

John Ole Askedal University of Oslo

Ian Roberts University of Cambridge

Tomonori Matsushita Senshu University

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Noam Chomsky and language descriptions / edited by John Ole Askedal, Ian Roberts, Tomonori Matsushita. p. cm. (The Development of the Anglo-Saxon Language and Linguistic Universals, issn 1877-3451 ; v. 2) Includes bibliographical references and index. 1. Chomsky, Noam. 2. Grammar, Comparative and general. 3. Language acquisition. I. Askedal, John Ole, 1942- II. Roberts, Ian G. III. Matsushita, Tomonori. P85.C47N633   2010 415’.0182--dc22 2010007272 isbn 978 90 272 1069 2 (Hb ; alk. paper)

© 2010 – Senshu University No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

To Noam Chomsky

Table of Contents PREFACE Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA .............................. 1

1. Genesis of Generative Grammar Systems of Syntactic Analysis Noam CHOMSKY........................................................................................................ 13 Some Methodological Remarks on Generative Grammar Noam CHOMSKY........................................................................................................ 31 Knowledge of Language: Its Elements and Origins Noam CHOMSKY........................................................................................................ 55

2. Current Issues in Language Descriptions Germanic Passive Constructions John Ole ASKEDAL .................................................................................................... 75 Prosodic Constraints on Old English Alliteration Yasuaki FUJIWARA....................................................................................................111 The Historical Role of Genitives in the Emergence of DP Kazuyo MIYAMAE..................................................................................................... 125 The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose Akinobu TANI ............................................................................................................ 149 A Short Note on Movement and Control in the English Noun Phrase Junji HAMAMATSU .................................................................................................. 169 Coordinating and Subordinating Conjunctions in Spoken American English Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA ............................................ 179 Complement Capacities in German: Three Types of Complements Yasuhito HOSAKA ..................................................................................................... 197 Index of Names................................................................................................................... 213 Index of Subjects ................................................................................................................ 215 Editors & Contributors........................................................................................................ 225

PREFACE Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA GENESIS OF GENERATIVE GRAMMAR Noam Chomsky was born in Philadelphia, Pa., on Dec. 7, 1928. He attended the University of Pennsylvania, where he met the eminent linguist Zellig Harris and where he took his B. A. (1949), M. A. (1951), and Ph.D. (1955) degrees. He was Junior Fellow at Harvard in the early 1950s, and began working on generative grammar at this period. He began teaching at the Massachusetts Institute of Technology in 1955, and is now Institute Professor there. Chomsky is best known for the theory of generative grammar. Originally put forward in the 1950s (in Chomsky’s PhD dissertation The Logical Structure of Linguistic Theory, 1955, published in 1975, and Syntactic Structures, Mouton, 1957), the central idea in generative grammar is that the nature of natural-language syntax can be captured by a finite set of rules which are able to produce an infinite set of well-formed structures. This idea, which was given full formal expression in a rigorous algebraic notation from Chomsky’s earliest work, is now assumed by the majority of linguists, and has been highly influential in related fields such as philosophy, psychology and computer science. Many linguists feel that generative grammar has provided a true insight into the nature of human language, and thus into the workings of the human mind. Like many advances in human knowledge, generative grammar arose from a synthesis of earlier lines of thought. Arguably, Chomsky’s ideas can be best understood as a combination of American structuralist techniques in linguistics, the formal mechanisms of recursive-function theory, and Cartesian rationalism. Chomsky was trained as a linguist in 1940s America, in the heyday of post-Bloomfieldian structuralism. Indeed, his principal teacher, Zellig Harris, was a major practitioner of this approach. The key idea behind structuralism, in particular American structuralism, was the autonomy of linguistic structures: linguistic structures were looked at independently of psychology, of logic, even of meaning, and each “linguistic level” (phonemics, morphemics, syntax, etc.) was seen as independent of all others. This approach arguably led to greater rigour in analysis than had previously achieved and certainly freed linguistic analysis from certain aprioristic shackles. However, despite the best efforts of some, notably

2

Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA

Harris, the nature of syntax in particular remained difficult to understand. Chomsky saw that what was needed to understand syntax was a recursive rule system, or system of rule systems, which specified precisely how symbols could be combined and recombined in a limited number of ways to form the infinity of well-formed expressions of any natural language. The techniques of recursive-function theory, as it had been developed in mathematics largely in the interwar years, were adopted to this effect. In this way the key property of natural-language syntax, discrete infinity, could be formally captured for the first time (arguably this feature of language had been glimpsed by earlier thinkers, notably the 17th-century French Port Royal grammarians, but they had been unable to provide formal expression of their insights). Combining recursive rule systems with structuralist analytical rigour was already a huge step forward in our understanding of language, but Chomsky has furthermore argued that such rule systems must, at some level of abstraction from physical mechanisms, be instantiated in the human brain and determined by the human genome: thus the insight of discrete infinity becomes a biological fact about our species. We are only beginning to see the full consequences of this last idea now. Very largely thanks to Chomsky’s innovations in the 1950s, over the past fifty years linguists have arrived at a deeper understanding of syntax than had ever before been achieved. Chomsky also greatly enhanced the formal rigour of linguistics by providing an algebra for rule systems (this has had a major impact in computer science – see below). Moreover, Chomsky, by his advocacy of a rationalist view of mind and language, has firmly situated linguistic theory in relation to the cognitive sciences and philosophical questions more generally. It should also not be forgotten that Chomsky has contributed to the other areas of linguistics, most notably phonology in his monumental work, co-authored with Morris Halle, The Sound Pattern of English (Holt, 1968): this book redefined the subdisciplines of phonology and morphology entirely, and remains hugely influential. Chomsky has also written numerous works on the philosophy of language, and one monograph on the history of linguistics (Cartesian Linguistics, 1966; 3rd edition reissued by Cambridge University Press, 2009). Outside linguistics, Chomsky’s ideas have influenced other fields. Let us look first at psychology. It is sometimes stated that the “cognitive revolution” in psychology, the idea that behaviourist approaches were unrevealing and that human cognition could be studied on the basis of the computer metaphor of the mind, really began in 1957 with the publication of Syntactic Structures. Certainly the Chomskyan approach to language, in particular the idea that a central aspect of language is the manipulation of symbolic representations, is highly congenial to the cognitive-science view

PREFACE

3

of the mind in general. Chomsky’s thinking has thus been highly influential in cognitive psychology. Another major contribution to psychology was Chomsky’s 1959 review of Skinner’s Verbal Behavior. In that work, Skinner attempts to provide an account of first-language acquisition based on behaviourist stimulus-response theory. Chomsky demonstrates that this simply cannot work, and in so doing played a role in the downfall of behaviourism. A third contribution to psychology is the emphasis in Chomsky’s theorising on the importance of children’s acquisition of their mother tongue. Chomsky, in his celebrated “poverty-of-the-stimulus” argument, has emphasised the difficulty of the language-acquisition task, and the fact that children seem to have the capacity to acquire the grammar of their native language on the basis of highly limited and deficient evidence. He points out that these observations are consistent with the postulation of an innate predisposition to language, or more precisely to the acquisition of grammars of a particular form, while they remain very difficult to understand if we assume that the sole basis of knowledge is sensory data. These ideas virtually created the discipline of developmental psycholinguistics, the study of the linguistic competence of small children. Over the past forty years, our understanding of the linguistic abilities of babies and toddlers has grown enormously. Although Chomsky himself has never directly worked in this field, his theoretical positions are what made the field possible in the first place. This brings us naturally to philosophy. As mentioned above, Chomsky has advocated a form of Cartesian rationalism: the view that knowledge is determined largely by the nature of the mind itself, with experience playing only a triggering role. His principal argument for this position is the poverty-of-the-stimulus argument from first-language acquisition, described in the previous paragraph. Chomsky, however, eschews dualism, instead taking the view that the language faculty must be ultimately instantiated in neurological mechanisms (although we are a long way from understanding how this works) and that innate ideas must be part of the genome. Chomsky’s ideas about language and language acquisition did much to reignite discussion of rationalist theories of knowledge, especially in the English-speaking world. Another area where Chomsky’s contributions have been important is computer science. In the late 1950s and early 1960s Chomsky worked, with various colleagues, on the abstract question of how different kinds of rule systems can generate different kinds of infinite sets of expressions. This work led to what is still known as the “Chomsky hierarchy” of formal languages, and still appears in computer-science textbooks.

4

Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA

The above may give the impression that Chomsky’s work was mostly carried out decades ago. Although much ground-breaking work was done in the 1950s, and of course the benefit of hindsight makes it easy to see the importance of those ideas, Chomsky has never ceased to develop and revise the details of his theory, particularly in syntax. Over the past fifteen years or so, his ideas have taken a radically new turn with the postulation of the Minimalist Program for linguistic theory. The minimalist program asks a question that was never explicitly asked before: how well-adapted is the language faculty for its function in relation to other aspects of cognition? To put it another way: how little do we have to say in our linguistic theory given, on the one hand, the basic fact that the linguistic system relates sounds and meaning through a generative rule system, and, on the other hand, the inherent constraints imposed by the physical world and by logic? The minimalist program thus consists in developing analyses of natural-language phenomena which attempt to render the rule system as simple and general as possible. Chomsky has even suggested that natural language may, in some non-trivial way, be a perfect system, i.e. one which is optimally structured for the purpose of relating sound and meaning. This is now known as the Strong Minimalist Thesis. Although highly speculative, this idea is both provocative and profound. The minimalist program has led to a very drastic revision of many of the formal, technical aspects of the theory of syntax. Leaving those aside, however, it has also altered the conceptual basis of the field. Chomsky (2004) argues that the notion of explanatory adequacy, the idea that a linguistic phenomenon is explained if we can explain how children can acquire it given the poverty of the stimulus, which since 1964 he had argued to be the goal of linguistic theory, may not be the ultimate goal. There may be a still deeper goal: that of explaining how the highly imperfect-seeming properties of natural languages are in fact derived from an ultimately perfect system. A further recent development emerges from Chomsky’s collaboration with the evolutionary biologists Hauser and Fitch in a 2002 paper in Science on the evolution of language (Hauser, Chomsky, and Fitch 2002): here Chomsky and his co-authors suggest that rather little in language may be domain-specific, i.e. specifically evolved for the purposes of the language faculty. We will return briefly to the question of language evolution below. Professor Noam Chomsky has warmly accepted our request that the Senshu Project issue a volume including his three earlier papers as follows: “Systems of Syntactic Analysis” (1953), “Some Methodological Remarks on Generative Grammar” (1961) and “Knowledge of Language: its Elements and Origins” (1981).

PREFACE

5

PROSPECTS FOR THE FUTURE Language research in the 21st century covers a range of very diverse fields including philology, descriptive grammars of particular languages, historical linguistics, language typology, generative grammar, bio-linguistics, neuroscience, and bioscience. In a recent volume, Chomsky (2004: 187) states his concern with human biology as Language can and should be studied from many different points of view, hence in many different disciplinary contexts. … The parts of the study of language that we’ve been talking about seem to me to fall, in principle, into human biology. McGilvray (2009: 4) points out Chomsky’s change of focus concerning mind and language as follows: In recent years Chomsky’s label for his approach to mind and languages has changed from “rationalistic” to “biolinguistic.” … The label change highlights a characteristic work; the aim has always been to try to accommodate the science of language to some natural science, thus biology – for biology alone can explain how language is innate, why it is unique to humans, and how it grows. In this connection, one might ask what the connection may be between the biolinguistic programme as Chomsky has recently outlined it and historical linguistics, including in particular diachronic syntax. If we take the view that the language faculty is a system of principles and parameters, with the parametric variation determining and constraining how languages may vary and change, then we can see a very clear and intriguing connection. Because grammars are recreated by each cohort of acquirers within the tight constraints imposed by the language faculty, the variation and change that is so prevalent in language arises. All that seems to be required for this is for certain parts of the system to be open to variation and the poverty of the stimulus; given these conditions, different adult systems will emerge (this point is demonstrated very clearly in Niyogi (2006)). These constantly innovated, minutely varying systems stabilise after the critical period for language acquisition, and typically become associated with social and cultural value (in an ultimately quite arbitrary way, as far as the system itself is concerned). The ongoing, inevitable propensity for variation, determined by the parametrised parts of the formal system, leads to the creation of new systems which fit into the social value system in different

6

Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA

ways. Thus a parametrised language faculty, allowing random variation in a few small areas, gives rise to the phenomena of variation – both sociolinguistic and cross-linguistic – and change. And children have the ability to acquire these systems along with the variation and, under conditions whose precise nature remains to be determined, they have the capacity to subtly modify the system. The study of historical syntax can, in these terms, find its natural place in the cognitive sciences. What will hopefully develop is a greater understanding of each of the three elements which, according to Niyogi (2006), contribute to the dynamical system that is a language being spoken by a population. We need to better understand the nature of language learning and acquisition through empirical work on language acquisition and theoretical work on learnability; we need to better understand the relation between language variation and populations through empirical work in sociolinguistics and theoretical work on the computational modelling of population dynamics. Finally, we need to better understand the properties of the set of grammars through empirical work in language typology and theoretical work in grammatical theory. Principles and parameters theory is obviously central to this last enterprise, itself crucial to a full understanding of language change. The obvious question that these remarks give rise to is what the nature of the variation in the language faculty is, and why should it vary at all. Berwick & Chomsky (2008) suggest an interesting view on this question, which ultimately traces the answer to the way in which the language faculty may have evolved. They sketch an account of the evolution of language which involves four separate components: (i) the development of the syntactic system (which, in the context of the Minimalist Program, largely reduces to a single combinatorial operation, known as Merge); (ii) the development of the connection to the conceptual-intentional system of thought, which may be closely linked to (i); (iii) the development of the means of “externalization” of thought through speech, entailing the existence of phonology and morphology, and (iv) the development of the lexical atoms. Berwick & Chomsky argue that language may be fundamentally a “tool for thought”, and that, hence, (i) and (ii) are intrinsically linked. Leaving aside (iv), which to some degree remains mysterious, they suggest that “externalization” of thought was a secondary process, possibly one which combined pre-existing cognitive abilities rather than involving a special mutation. In this connection, they (ibid.: 15) say: Externalization is not a simple task. It has to relate two quite distinct systems: one is a sensorimotor system that appears to have been

PREFACE

7

basically intact for hundreds of thousands of years; the second is a newly emerged computational system for thought, which is perfect, insofar as the strong minimalist thesis is correct. We would expect, then, that morphology and phonology – the linguistic processes that convert internal syntactic objects to the entities accessible to the sensorimotor system – might turn out to be quite intricate, varied, and subject to accidental historical events. Parameterization and diversity, then, would be mostly – possibly entirely – restricted to externalization. That is pretty much what we seem to find: a computational system efficiently generating expressions interpretable at the semantic/pragmatic interface, with diversity resulting from complex and highly varied modes of externalization, which, furthermore, are readily susceptible to historical change. In this context, studying historical change may be a question of understanding the variation in the range of possible externalization strategies made available by the language faculty. This is clearly true for morphological and phonological change, on this view. However, syntactic change may raise further questions. One central question that this view raises is: how much of the syntactic variation that we see in the world’s languages is due to externalization processes and how much is due to the core computational system of syntax? Given the logic of the above quotation, this is one way of investigating the overall correctness of the Strong Minimalist Thesis. From another perspective, the core computational system may be perfect but still variable: its perfection may indeed entail a certain variability. Once again, the study of change and variation as observed in languages across time and space may shed light on these questions. In this way, we can understand how some the traditional concerns of historical linguistics, seen in a new light, may be directly relevant for the biolinguistic programme. CURRENT ISSUES IN LANGUAGE DESCRIPTIONS In his contribution to this volume, Askedal describes Germanic passive constructions and proposes a comparative classification of passive constructions in modern standard Germanic languages and general typological characteristics of the languages: the Insular Scandinavian languages, Icelandic and Faroese; the modern Mainland Scandinavian languages, Danish, Swedish and Norwegian; the Continental Germanic languages, Modern West Frisian, Dutch and German; and English. Fujiwara is in his paper “Prosodic Constraints on Old English Alliteration” concerned with Old English poetry and considers the distinction between compound words and complex words and argues that secondary

8

Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA

stresses on the second elements of compound words are qualified for alliteration in the same way as primary stresses on the first elements, while tertiary or zero stresses are irrelevant to alliteration. Miyamae offers in her article “The Historic Role of Genitives in the Emergence of DP” an investigation of the historical development of the English nominal construction in terms of emergence and establishment of a new functional category D. She proposes that many factors and motives in the preceding period contributed to this drastic syntactic change in the latter half of the 15th century. Tani discusses the word pairs in Chaucer’s verse in comparison with those in his prose and argues that the function of the Word Pairs in the verse is interlinked with rime in stark contrast to that in the prose and that the verse texts in Fragment I of the Canterbury Tales can be classified into three groups. In his paper “A Short Note on Movement and Control in the English Noun Phrase”, Hamamatsu claims that the Agent DP, realized as subject in ‘John’s attempt to gain attention (failed)’, appears to be ‘demoted’ in ‘The attempt of John to gain attention (failed)’ within the noun phrase and that the movement approach to control offers a neat explanation for the alternation. Hosaka argues in his article “Complement Capacities in German: Three Types of Complements” that the complement capacities have correlations with the syntactic properties shown in passivization in German. The passivizability of verbs with infinitival complements is dependent on some conditions. In their paper “Coordination and Subordinating Conjunctions in Spoken American English”, Iyeiri, Yaguchi and Baba discuss the employment of coordinating and subordinating conjunctions in spoken American English, paying special attention to their uses in different styles of professional English in conferences and meetings and investigating The Corpus of Spoken Professional American English (CSPAE). The papers from these various branches deal with fundamental issues in the fields of Generative Grammar and language descriptions. They share the common goal of contributing to our understanding of these areas. The Senshu Open Research Project ‘The Development of the Anglo-Saxon Language and Linguistic Universals’ was selected for funding as one of several promising and unique projects in Japan by the Ministry of Education, Sports, Culture, Science and Technology in 2005 and has been supported by Senshu University in conjunction with the Ministry.

PREFACE

9

The Senshu Project organized in the academic year 2008 and 2009 three International Conferences with lectures devoted to ‘Phonetics as Brain Training’, ‘On Parametric Syntax with special reference to English and East Asian Languages,’ and ‘Language Typology’. The following scholars were invited to the conferences: Michael Ashby (Experimental Phonetics, University College London), John Ole Askedal (Germanic Linguistics, University of Oslo), Berndt Heine (Language Typology, University of Cologne, emer.), C.–T. James Huang (Generative Grammar, Harvard University), Heiko Narrog (Japanese, Tohoku University), and Ian Roberts (Diachronic Syntax, Cambridge University). The publication of this book was supported by the “Open Research Center” Project for Private Universities: matching the funds provided by MEXT (Ministry of Education, Culture, Sports, Science and Technology), 2005-2009. REFERENCES Berwick, R. C. & N. Chomsky. 2008. “The Biolinguistic Program: The Current State of its Evolution and Development”. Ms. MIT. In Biolinguistic Investigations. Forthcoming. A. DiSciullo & C. Aguero eds. Cambridge: MIT Press. Chomsky, N. 1953. “Systems of Syntactic Analysis”. The Journal of Symbolic Logic, Vol. 18, No. 3. 242-256. 1955. The Logical Structure and Linguistic Theory. Ph.D. dissertaion. University of Pennsylvania. (Published in 1975. New York and London: Plenum Press.) 1957. Syntactic Structures. The Hague: Mouton. 1959. “Review of Verbal Behavior by B.F. Skinner”. Language 35, no. 1 (January-March 1959): 26-57. (Reprinted as no. A-34 in the Social Sciences by Bobbs-Merrill, Inc.; reprinted in The Structure of Language, edited by Fodor and Katz. New York: Prentice-Hall. 1964; reprint (in French). Language 16 (December 1969): 16-49; reprinted in Readings in Philosophy of Psychology, edited by N. Block, 48-63. Cambridge: Harvard University Press. 1980; reprinted in Ajattelu, Kieli, Merkitys: Analyyttisen Filosofian Avainkirjoituksia, 279-310. Helsinki: Gaudeamus. 1997.) 1961. “Some Methodological Remarks on Generative Grammar”. Word. Vol. 17. No. 2. 219-239. 1981. “Knowledge of Language: its Elements and Origins”. Philosophical Transactions of the Royal Society of London. B 295, 223-234.

10

Ian ROBERTS, John Ole ASKEDAL and Tomonori MATSUSHITA

2004. The Generative Enterprise, Revisited. Berlin: Mouton de Gruyter. 2009. Cartesian Linguistics, 3rd Edition. Cambridge: Cambridge University Press. & M. Halle. 1968. The Sound Pattern of English. New York: Holt, Rinehart and Winston. Hauser, M. D., N. Chomsky, and W. T. Fitch. 2002. “The Faculty of Language: What is it, Who has it, and How did it Evolve?” Science, Vol. 298, 1569-1579. McGilvray, J. 2009. “Introduction to the Third Edition”. In N. Chomsky 2009, 1-52. Niyogi, P. 2006. The Computational Nature of Language Learning and Evolution. Cambridge: MIT Press.

1. Genesis of Generative Grammar

Systems of Syntactic Analysis* Noam CHOMSKY 1. INTRODUCTION 1 During the past several decades, linguists have developed and applied widely techniques which enable them, to a considerable extent, to determine and state the structure of natural languages without semantic reference. It is of interest to inquire seriously into the formality of linguistic method and the adequacy of whatever part of it can be made purely formal, and to examine the possibilities of applying it, as has occasionally been suggested,2 to a wider range of problems. In order to pursue these aims it is first necessary to reconstruct carefully the set of procedures by which the linguist derives the statements of a linguistic grammar from the behaviour of language users, distinguishing clearly between formal and experimental in such a way that grammatical notions, appearing as definienda in a constructional system, will be formally derivable for any language from a fixed sample of linguistic material upon which the primitives of the system are experimentally defined. The present paper will be an attempt to formalize a certain part3 of the ©Association for Symbolic Logic * Reprinted by permission from Journal of Symbolic Logic, 18:3. 242-256 (September, 1953). The Center for Research on Language and Culture, Senshu University would like to express our thanks to Professor Noam Chomsky and the Association for Symbolic Logic. Received October 18, 1952. 1 Within linguistics, the source for these investigations is in the methods of structural analysis developed by Z. S. Harris; within philosophy and logic, it is in the work of N. Goodman on constructional systems and in the development of nominalistic syntax by Goodman and Quine. As general references, then, for this paper, see HARRIS, Methods in structural linguistics, Chicago, 1951, GOODMAN, The structure of appearance, Cambridge, 1951, and GOODMAN and QUINE, Steps towards a constructive nominalism, this JOURNAL, vol. 12 (1947), pp. 105-122. I am much indebted to Professors Harris, Goodman, and Quine, as well as to Y. Bar-Hillel, H. Hiż, and others, for many suggestions and criticisms. 2 E.g., W. V. QUINE, Notes on existence and necessity, Journal of philosophy, vol. 40 (1943), pp. 120. Also, see Z. S. HARRIS, Discourse analysis, Language, vol. 28(1952), pp. 1-30, for an investigation of the possibility of using methods of linguistics to determine the structure of a connected short text, thus, in a sense, setting up partial synonymity classes for it. 3 The constructions of this paper are roughly coextensive with the procedures of chapters 15, 16, Methods.

14

Noam CHOMSKY

linguist’s generalized syntax language. From another point of view, this paper is an attempt to develop an adequate notion of syntactic category within an inscriptional nominalistic framework. The inscriptional approach seems natural for linguistics, particularly in view of the fact that an adequate extension of the results of this paper will have to deal with the problem of homonymity, i.e., with a statement of the conditions under which tokens of the same type must be assigned to different syntactic classes. It will appear below that the calculus of individuals can often supply quite simple solutions to constructional problems that seem on the surface to require a set-theoretic solution, thus removing the necessity for an involved hierarchy of types and increasing the overall workability of the system. There are several ways in which we might approach the concept ‘belong to the same syntactic category.’ We might consider assigning elements to the same category only if they have all contexts in common (i.e., are mutually substitutable in all sentences), if they share some context, or if the ancestral of this latter relation holds between them.4 All three suggestions are too restrictive for the general case which we wish to consider.5 In particular, consider the following six-sentence text: (1)

‘ab’, ‘cb’, ‘de’, ‘fe’, ‘axd’, ‘cyf’.

To attain the purposes of this constructional attempt, we must be able to assign ‘x’ and ‘y’ to the same category. The general procedure which we wish to reconstruct is roughly as follows. If, in a given body of material, two elements occur in sentences which differ only in these elements (e.g., ‘a’ and ‘c’ occur in the context ‘..b’, ‘d’ and ‘f’ in ‘..e’, in (1)), then the two elements are assigned to the same class. But now two expressions differing term by term only in elements previously assigned to the same class (e.g., ‘a..d’ and ‘c..f’) are identified, thus shrinking the totality of contexts and allowing new elements (e.g., ‘x’ and ‘y’) to be put into the same class on the basis of occurrence in the same sentential context. When this process can be carried no further, considering expressions of any length and degree of discontinuity as elements, the resulting classes are the broadest syntactic categories for this text. Before proceeding with the actual constructions, it should be made clear that the present system as given here is not adequate for the analysis of 4 5

See Y. BAR-HILLEL, On syntactic categories, this JOURNAL, vol. 15 (1950), pp.1-16, for a development of these notions. The third suggestion is actually equivalent to the system adopted here for the special case of languages in which each sentence contains exactly two elements (morphemes).

Systems of Syntactic Analysis

15

natural languages. Several crucial problems have been explicitly avoided in this treatment. One is the problem of homonymity touched on above. The second is the problem posed by those sentential contexts in which members of various syntactic classes can occur, e.g., in English, ‘it was ....’ The third concerns the predictive character of the grammar. A syntactic analysis will result in a system of rules stating the permitted sequences of the syntactic categories of the analyzed sample of the language, and thus generating the possible or grammatical sentences of the language. We can state this problem as one of determining the limitations on distribution which characterize various subsets of the broad syntactic categories, and of determining for any given language which subsets should be established. These problems lie beyond the scope of this paper, 6 which is concerned solely with the derivation, from reasonably limited samples,7 of the broadest categories and the most general statement of the grammatical rules. These problems are related in various ways, and their solution seems to require auxiliary systems using richer means than those adopted here. Thus in actual linguistic procedure, problems of the second type, for example, are avoided by considerations involving the relative size of the logical product of the sets of contexts (distributions) of the elements involved.8 The final decision as to the general adequacy of the systems proposed here will of course depend on the success or failure of such auxiliary systems. 2. ELEMENTARY NOTIONS OF SYNTAX We therefore take as the general apparatus for these constructions quantification theory with variables ranging over inscriptions, that is, 6

7

8

The first two in particular are problems of how to apply the primitives of these systems. Thus ‘CON’ must not be predicated of homonyms, and ‘ENV’ must not be predicated of contexts such as ‘it was ...’ (see below, § 2). Thus we do not wish to require in principle that the ‘whole language’ be available as data. It is, however, of interest to consider this situation as well. Thus, if there are large significant classes which are subdivided into classes whose distributions cluster separately (see footnote 8), but such that the subclasses have similar distributions in terms of other classes, then the methods to be adopted here permit the construction of the large class as a ‘second-level’ class. Thus we might require, for an expression to be admissible into the class of contexts, that the distributions of the elements occurring in its ‘blank space’ form a single cluster of sets. It is therefore necessary on the one hand to clarify the sense in which a set of sets can be said to be most efficiently divided into a set of clusters of sets, on the other, to investigate the actual statistics of distribution in natural languages. Precisely the same researches are necessary to resolve at least part of the homonym problem, considering homonyms as the elements whose distributions overlap two clusters of distributions. Cf. HARRIS, Methods, pp. 257ff.

16

Noam CHOMSKY

morphemes 9 and sums (scattered or continuous) 10 of morphemes. As extra-logical primitives we take the following. ‘O’, read ‘overlaps.’ This is the primitive notion of the calculus of individuals.10 Two inscriptions overlap if there is some one inscription that is a part of both. ‘EQL’, read ‘equally long.’ Two inscriptions are equally long if they contain exactly the same number of morphemes (atoms). ‘CON’, read ‘conform.’ This is an equivalence relation holding between inscriptions of any length and shape (as long as they contain discontinuities at exactly the same places) which are, atom by atom, tokens of the same type, with the special considerations cited above in the case of homonyms and morpheme alternants.9 ‘PRE’, read ‘precedes.’ This is a total ordering among atoms, a partial ordering among longer inscriptions (see A6, A7, below). ‘ENV’, read ‘environment.’ The long inscription (the text) to be analyzed is divided into environments such that if the sum of a and b (non-over-lapping) is an environment, then a is a context of b and vice versa. Thus it might be useful to take sentences as the environment system for natural languages. We begin by constructing certain elementary notions.11 D1. ‘SEGab’ for ‘(x)(Oxa ⊃ Oxb)’. (‘a is a segment of b.’) D2. ‘=ab’ for ‘SEGab.SEGba’. (‘a is identical with b.’) D3. ‘∩bc’ for ‘(ιa)(x)(SEGxa ≡ .SEGxb.SEGxc)’. (‘the product of b and c.’) D4. ‘∪bc’ for ‘(ιa)(x)(Oxa ≡ .Oxb ∨ Oxc)’. (‘the sum of b and c.’) D5. ‘ATMa’ for ‘(x)(SEGxa ⊃ =xa)’. (‘a is an atom.’) D6. ‘TIab’ for ‘SEGab. (x)(y)(z)(SEGxa.SEGya.SEGzb.PRExz. PREzy. ⊃ SEGza)’. (‘a is a through inscription of b.’) 9

10

11

Actually, over morpheme occurrences. The linguist’s morphemes are classes of conforming minimal meaning-bearing units, e.g., ‘boy,’ ‘think,’ ‘of,’ ‘ing,’ the plural ‘s’, etc. Forms such as ‘wife’ and ‘wive,’ with selection predictable given the context (thus ‘wive’ occurs only before ‘s’ plural, ‘wife’ only elsewhere), are called morpheme alternants and are considered to belong to the same morpheme. They are here considered to conform. See Methods, chap. 12, 13. For a discussion of the Calculus of Individuals (and the notions of ‘sum’, ‘scattered individual’, etc.) see H. S. LEONARD and N. GOODMAN, The calculus of individuals and its uses, this JOURNAL, vol. 5 (1940), pp. 45-55, and Structure, pp. 42-55. D1-4 are, respectively, D2.042, D2.044, D2.045, and D2.047 of Structure, pp. 44-46.

Systems of Syntactic Analysis

17

Thus if we have the inscription ‘pqrstuv’ and if b is ‘pq..s..v’, then ‘pq’, ‘pq..s’, ‘q..s..v’, etc., are through inscriptions of b, but not ‘p..s’. Thus if TIab, a need not be continuous unless b is. But now we can readily define ‘a is continuous.’ D7. ‘CIa’ for ‘(x)(SEGax ⊃ TIax)’. (‘a is a continuous inscription.’) D8. ‘BEGab’ for ‘TIab.~(Ez)(SEGzb.PREza)’. (‘a is a beginning of b.’) D9. ‘ENDab’ for ‘TIab.~(Ez)(SEGzb.PREaz)’. (‘a is an ending of b.’) D10. ‘XLab’ for ‘BEGab.ATMa’. (‘a is to the extreme left of b.’) D11. ‘XRab’ for ‘ENDab.ATMa’. (‘a is to the extreme right of b.’) D12. ‘MCSab’ for ‘CIa.SEGab.(x)(SEGxb.CI∪ax. ⊃ SEGxa)’. (‘a is a maximal continuous segment of b.’) D13. ‘DCab’ for ‘~Oab.(Ex)(Ey)(SEGxb.PRExa.SEGyb.PREay. CI∪a∪xy)’. (‘a is a discontinuity of b.’)

Thus in the example for D6, ‘r’ and ‘tu’ are the discontinuities of b. We assign to every inscription a discontinuity index consisting of one atom from each discontinuity, and one additional atom to ensure that even continuous inscriptions have a discontinuity index, there being no null inscription. Two inscriptions will then be equally discontinuous if their discontinuity indices are of the same length. D14. ‘Idcab’ for ‘(x)(SEGxa.ATMx. ≡ .(Ey)(DCyb.XLxy) ∨ XLxb)’. (‘a is the discontinuity index of b.’) D15. ‘Edcab’ for ‘(x)(y)(Idcxa.Idcyb. ⊃ EQLxy)’. (‘a and b are equally discontinuous.’) D16. ‘Kab’ for ‘ENV∪ab.~Oab’. (‘a is a context of b.’) With the aid of these concepts we can state an axiom system.12 12

A1 and A10 are, respectively, 2.41 and 2.45 of Structure, pp. 44-46. The essential idea of A13 is discussed in Structure on pp. 47-48. This axiom system is adequate only if we assume that no inscription contains infinitely many atoms, and then carry out proofs in the metalanguage, using induction on the number of atoms in an inscription. Alternatively, we could adjoin several axioms involving ‘EQL’ which would permit the derivation of all theorems in which no schematically defined terms appear within the system.

18

Noam CHOMSKY A1. A2. A3. A4. A5. A6. A7. A8. A9. A10. A11. A12. A13.

Oab ≡ (Ex)(y)(Oyx ⊃ .Oya.Oyb). EQLab ≡ :ATMa.ATMb. ∨ (Ec)(Ed)(Ee)(Ef)(EQLce.EQLdf. ～Ocd. ～Oef. = a∪cd. = b∪ef). ATMa ⊃. CONaa.(CONab ⊃ CONba).(CONab.CONbc. ⊃ CONac). CONab ≡ .EQLab.(x)(y)(r)(s)(BEGxa.XRrx.BEGyb.XRsy. EQLxy. ⊃. CONrs.Edcxy). PREab.PREbc. ⊃ PREac. PREab.SEGxa.SEGyb. ⊃ PRExy. ATMa.ATMb. ⊃ .PREab ∨ PREba ∨ =ab. =ab.ENVa. ⊃ ENVb. =ab.CONac. ⊃ CONbc. (Ez) (=z∪xy). (Ex)(Ey)(XLxz.XRyz). ATMx ⊃ (Ey)(Ez)(CONxy.Kzy). (Ex)(..x..) ⊃ (Ea)(y)(ATMy ⊃ .SEGya ≡ (Ez)(SEGyx. .. z..)).

We can now derive such expected theorems as the substitution rule for identity, ‘Oab ≡ (Ex)(=x∩ab)’, ‘～PREaa’, ‘ ‘EQL’, ‘CON’, and ‘Edc’ are equivalences,’ etc. We may now proceed to the analysis of the concepts ‘comparable’ as applied to sequences, and ‘same position’, as applied to terms of comparable sequences. It will be necessary here to consider sequences as being bracketed into terms in certain specific ways.13 We will say that two sequences are comparable (with respect to given bracketings) if they contain the same number of bracketed terms and the same number of discontinuities, and if these discontinuities occur in the same places with reference to the bracketed terms. E.g., the following are similar with respect to the bracketing indicated by space, where ‘...’ is a discontinuity.14

13

14

The non-atomic terms will in the interesting cases be what are called ‘immediate constituents’ in linguistic terminology. Thus such a linguistic form as ‘that poor fellow on the corner missed his bus’ might be analyzed into two immediate constituents, a noun phrase (‘that .... corner’) and a verb phrase (‘missed his bus’), in which case it might be shown to be equivalent in the sense of the procedure to be adopted to a sentence consisting simply of a noun and a verb, e.g., ‘he fell.’ These phrases in turn can be analyzed into immediate constituents (e.g., ‘that poor fellow’ and ‘on the corner’), etc., until the ultimate constituents (morphemes) are reached. For a detailed discussion of constituent analysis and its problems see R. S. WELLS, Immediate constituents, Language, vol. 23 (1947), pp. 81-117, and Methods. For the time being, we restrict ourselves to terms which do not cross over discontinuities. See however systems III, IV, V, pp. 15-18.

Systems of Syntactic Analysis (2)

19

ab cdef . . . h . . . ijk l m nopq r . . . stu . . . v wx yz

Two bracketed terms are in the same position in comparable sequences if each is the nth term. Thus ‘comparable’ will be a four-place and ‘same position’ a six-place predicate. D17. ‘DVab’ for ‘(Er)(BEGra.SEGrb.(x)(y)(MCSyb.XRxy. ⊃ SEGxr))’. (‘a is a divisor (or bracketing) of b.’)

Thus any divisor of the first sequence in (2) will include at least the atoms ‘f’, ‘h’, and ‘m’, plus any number (or 0) of the other atoms of the sequence, plus any number (or 0) of atoms which may follow ‘m’, but none which precede ‘a’. The divisor is at least as long as the dicontinuity index. In D22 it will be convenient to consider a given divisor as being the divisor for a sequence and for all of its beginnings ending in an atom of the divisor. This explains why in D17 we consider not a itself, but only some beginning of it (r) to be included in b. A term of a sequence is a maximal part of the sequence occurring between two atoms of the divisor and containing the second of these on its extreme right. Thus in (2), with the given bracketings, each sequence has six terms. ‘Term’ is in many ways the analogue of ‘є’ in this system. D18. ‘TRMabc’ for ‘DVcb.(Ex)(XRxa. =x∩ac).(BEGab ∨ (Ey)(SEGyc. PREya. TI∪yab))’. (‘a is a term of b with respect to the divisor c.’)

It follows that terms are continuous inscriptions.14 We assign to an atom in a sequence (with respect to a given divisor) a term index in such a way that the term index contains n atoms just in case the given atom is in the nth term of the sequence. D19. ‘Itrmabcd’ for ‘DVdc.ATMb.SEGbc.(x)(SEGxa ≡ :PRExb. SEGxd. ∨ =xb)’. (‘a is the term index of the atom b in c bracketed by d.’)

Thus in the first sequence of (2), the term index of ‘j’ with the given bracketing is the discontinuous inscription ‘b...f...h...j’, and the index of ‘a’ is ‘a’ itself. Two atoms are preceded by the same number of terms if their term indices are of the same length.

20

Noam CHOMSKY D20. ‘Etrmabcdef’ for ‘(Ex)(Ey)(Itrmxabc.Itrmydef.EQLxy)’. (‘the atoms a (in the sequence b bracketed by c) and d (in the sequence e bracketed by f) each occur in the nth term, for some n, of these sequences bracketed in this way.’)

We can define a predicate analogous to ‘Edc’ as follows. D21. ‘Edvabcd’ for ‘(x)(y)(XRxa.XRyc. ⊃ Etrmxabycd)’. (‘the sequence a is divided into the same number of terms by the divisor b as is the sequence c by the divisor d.’)

We can now define ‘comparable’ and ‘same position’ in the sense described above. D22. ‘CMPabcd’ for ‘Edvabcd.(x)(y)(BEGxa.BEGyc.Edvxbyd. ⊃ Edcxy)’. (‘the sequence a with divisor b is comparable to c with divisor d.’) D23. ‘SPabcdef’ for ‘TRMabc.TRMdef.CMPbcef.(x)(y)(XRxa.XRyd. ⊃ Etrmxbcyef)’. (‘a and d are in the same position in b (bracketed by c) and e (bracketed by f), respectively.’)

We see that ‘Etrm’, ‘Edv’, ‘CMP’, and ‘SP’ are symmetrical and transitive, and are reflexive if their places are significantly filled. Before proceeding with the actual formulation of the procedure of syntactic analysis, it will be useful to provide the following auxiliary notions. D24. ‘EItrmab’ for ‘DVba.(x)(TRMxab ⊃ (Ey)(Ew)(CONxw.Kyw))’. (‘a is bracketed by b into terms conforming to environment-included15 sequences.’) D25. ‘EIa’ for ‘(x)(XRxa ⊃ EItrmax)’. (‘a conforms to an environment-included sequence.’) D26. ┌EQ4(R)abcd┐ for ┌CMPabcd.(x)(y)(SPxabycd ⊃Rxy)┐, where ‘R’ is a syntactic variable ranging over two-place predicates. (‘the sequences a and c, as bracketed by b and d, respectively, are R-equivalent.’) D27. ┌EQ2(R)ab┐ for ┌(Ec)(Ed)(EQ4(R)acbd)┐. (‘the sequences a and b are R-equivalent.’)

3. A SYSTEM OF SYNTACTIC ANALYSIS The general procedure, stated in § 1, towards which we have been aiming, can be constructed within our system as follows. We construct an

15

‘environment-included’ will always be used in the sense of proper inclusion.

Systems of Syntactic Analysis

21

indefinite series of similarity relations (‘Sn’)16 such that the even similarities (‘S2n’) hold between sequences, and the odd similarities between terms of these sequences. ‘S0’ holds between sequences that conform term by term. D28. ‘S0abcd’ for ‘EQ4(CON)abcd’. D29. ‘S0ab’ for ‘EQ2(CON)ab’.

Thus ‘S0’ is simply conformity.17 ‘S1’ holds between terms which occur in contexts related by ‘S0’. ‘S2’ will then hold between sequences which are S1-equivalent, ‘S3’ between terms which occur in contexts related by ‘S2’, etc. In the example (1) in § 1, ‘a’ and ‘c’ would be related by ‘S1’ (as would ‘d’ and ‘f’), ‘a..d’ and ‘c..f’ by ‘S2’, and ‘x’ and ‘y’ by ‘S3’. We can give the general definition of ‘S2n’ as follows. D30. ┌S2nabcd┐ for ┌EQ4(S2n-1)abcd┐ D31. ┌S2nab┐ for ┌EQ2(S2n-1)ab┐

(n > 0). (n > 0).

The four-place predicate will be designated by ‘‘S2n(4)’’, the two-place, by ‘‘S2n(2)’’. The following will also be useful. D32. ┌Σ2nab┐ for ┌(x)(y)(XRxa.XRyb. ⊃ S2naxby) ┐

(n ≧ 0).

It remains to define ‘S2n+1’. We will see that there are various ways of constructing this definition, and that the kind of language for which the system may be adequate is directly determined by the choice teken here. The basic notion behind the definiens for ‘S2n+1ab’ in the first system to be considered, henceforth system I, is 16

17

In the definitions themselves, the variables ‘m’, ‘n’, etc., must be taken as syntactic variables ranging over numerals; elsewhere (including range specification) it is convenient to take them as numerical variables, ranging over numbers. It thus appears that ‘CON’ as explained and axiomatized above could have been defined from a simple conformity relation among atoms. The same is true of ‘PRE’. This conformity relation could, in turn, be defined as the ancestral of a non-transitive matching relation, in a way analogous to that demonstrated in Structure, pp. 234-235. (Added November 19, 1952.) These reductions would in fact increase the complexity of the basis in the sense of Structure, pp. 59-85, because the predicate formed (in calculating complexity) by compounding ‘EQL’ and ‘CON’ would have two segments rather than one under this revision, since ‘CON’ would now hold only of atoms. However, under a more recent formulation of the notion of simplicity (N. GOODMAN, New notes on simplicity, this JOURNAL, vol. 17 (1952), pp. 189-191) the two bases would be of equal simplicity.

22

Noam CHOMSKY (3)

(Ea1)..(Ear)(S2naa1.Ka1a2.S2na2a3...S2nar-2ar-1.Kar-1ar.S2narb).

To ensure transitivity this must be strengthened slightly, replacing ‘S2n’ by ‘Σ2n’ at each end. Hence as the actual definition in system I we take the following. D33. ┌S2n+1ab┐ for ┌(Ea1)....(Ear)(Σ2naa1.Ka1a2.S2na2a3.Ka3a4. S2na4a5...S2nar-2ar-1.Kar-1ar.Σ2narb) ┐, where n ≧ 0, r being any multiple of 4 (i.e., r＝4i, i ≧ 1).

Thus for each n, ‘S2n+1ab’ is introduced as an abbreviation which stands, simultaneously and ambiguously, for each of infinitely many expressions, one for each multiple of 4. For each n, the definiens for ‘S2n+1ab’ may be understood as an infinite disjunction of terms, each term of the disjunction being of the form given schematically on the right of D33, there being, for each integer i, a term with 4i quantified variables. See, however, end of § 3. We can now define ‘same syntactic category’. D34. ‘SSCab’ for ┌Snab┐, for some n ≧ 0. (‘a and b are in the same syntactic category.’)

Since ‘S2nabcd ⊃ S2nac’ follows directly, we see that any two elements related by any one of the sequence of similarities are in the same syntactic category. The following theorems give several pertinent characteristics of system I. T1. T2. T3. T4. T5. T6. T7. T8. T9. T10. T11. T12. T13.

EIa ≡ .(Ey)(Ew)(CONaw.Kyw).CIa. ‘S2n+1’ is symmetrical and transitive. ‘S2n(4)’ is symmetrical and transitive. ‘S2n(2)’ is reflexive and symmetrical. S2nabcd ⊃ S2n+2abcd S0abcd.EItrmab. ⊃ S2abcd. S2nab ⊃ S2n+2ab S2n+1ab ⊃ S2n+3ab EIa.EIb. ⊃ .S2nab ⊃ S2n+1ab EIa ⊃ S2n+1aa. EItrmab ⊃ S2nabab. S2n+1ab ⊃ .EIa.EIb. ‘SSC’ is an equivalence taken over the ‘EI’.

(n ≧ 1). (n ≧ 0). (n ≧ 0). (n ≧ 0).

restricted field of inscriptions satisfying

Systems of Syntactic Analysis

23

The distinguishing feature for each system of analysis will be the restricted field partitioned by ‘SSC’ as defined in that system. From T1 and T13 we see that the restricted field associated with system I is the set of continuous inscriptions conforming to environment-included inscriptions. As the syntactic categories for a given language we may take the equivalence classes of the restricted field augmented by the addition of such inscriptions as bear ‘SSC’ to some member of that class, but bear ‘SSC’ to no member of any other class. This notion of an ‘extended category’ allows of several interpretations. Suppose that a1, .. ,an are inscriptions satisfying ‘EI’, and that A1, . . ,An are the corresponding equivalence classes. Thus (4)

Ai = xÎ (EIx.SSCxai).

If we require merely that the syntactic categories be disjoint, we may define the syntactic categories Ā1, .. ,Ān as (5)

Āi = xÎ ( (Et)(tє Ai.SSCхt).(y)(SSCxy.EIy. ⊃ yєAi)).

If we require further that no member of a syntactic category bear ‘SSC’ to any member of any other, then we may take them as (6)

Āi = Ai∪xÎ ((Et)(tєAi.SSCxt).(y)(z)(SSCxy.SSCyz.EIz. ⊃ zєAi)).

In either case we can state in non-class terms the definition of ‘same extended category’ (‘SEC’). Thus along the lines of (5) we have (7)

‘SECab’ for ‘(Et)(Eu)(EIt.EIu.SSCat.SSCbu.(x)(SSCax ∨ SSCbx.EIx. ⊃ SSCxt))’.

From (7) we can prove that ‘SEC’ is transitive. We see further that two inscriptions can be in the same extended category though not related by ‘SSC’. The considerations of the next section do not depend on a decision as to the preferability of (5) or (6), or some third formulation, since there we will be concerned with general applicability of a system to all languages of a given kind, and thus will limit our attention to the restrictied fields corresponding to these kinds of language. It is of interest to note that although we can define ‘same syntactic category’ and even ‘same extended category’ within the system, we cannot give an adequate definition of the term ‘syntactic category’ itself. A syntactic category cannot be considered a mere sum of inscriptions in the same

24

Noam CHOMSKY

syntactic category, since a given inscription may contain as part of itself inscriptions belonging to various syntactic categories, even the category to which it itself belongs.18 This may not be serious if ‘syntactic rule’ (a term in the metalanguage to these systems, since syntactic rules are composed of expressions denoting syntactic categories, or, non-univocally, their members) can be defined directly in terms of ‘SEC’.19 From the fact that the last few definitions were presented only schematically, because of the appearance of numerical variables, it is clear that we have defined ‘same syntactic category’ in a nominalistic system only in the sense that these constructions may be viewed as a form of grammatical systems. For any given finite amount of textual material, we can construct a definition of ‘same syntactic category’ by giving a finite realization of the ‘rules’ for the construction of definitions laid down in this form of systems. For each n, the definiens of D33 will be a disjunction of a finite number of terms, and there will be only finitely many n’s for which ‘S2n+1’ need be defined. 4. ALTERNATIVE SYSTEMS. ADEQUACY Suppose that we have a classification of languages on some structural basis, and a set of criteria of adequacy which must be met by the concept ‘same syntactic category’ as defined in a system of analysis. We will say that such a system is applicable to a given kind of language if the defined term meets the criteria of adequacy when applied to any language of this kind. Two systems are equivalent for a given kind of language if they are each applicable to this kind of language and if they yield exactly the same syntactic categories for the constituents20 of the language when applied to any language of this kind. Two systems are equivalent if they are equivalent for all kinds of language to which either is applicable. System B is an extension of System A if it is equivalent to system A for all kinds of language to which system A is not applicable, and is applicable to some kind of language to which system A is not applicable. We can initiate a very limited investigation into these questions by making a simple classification of languages into kind 1, with only continuous constituents, and kind 2, with at least some discontinuous 18 19

20

As in the so-called endocentric constructions, e.g., ‘poor John,’ which belongs to the same category as ‘John.’ See L. BLOOMFIELD, Language, p. 194. It seems that this can be done by means of the devices developed by R. M. MARTIN and J. H. WOODGER, Towards an inscriptional semantics, this JOURNAL, vol. 16 (1951), pp. 191-203. We will call the immediate constituents, their immediate constituents, etc., down to ultimate constituents, simply the constituents of the language.

Systems of Syntactic Analysis

25

constituents, considering here only languages whose constituents conform to sequences properly included in environments. As a criterion of adequacy we will here take the simple (and in general, obviously insufficient) condition that the syntactic categories be a partitioning of the constituents of the language. Thus system I is applicable to languages of kind 1 only. We can now construct systems which are extensions of system I, and systems applicable to both kinds of language although not extensions of system I. Consider system II, constructed from system I by replacing D33 by (3), § 3, as D332.21 Thus D332 differs from D33 only in that it has ‘S2n’ in both places in which D33 has ‘Σ2n’. We can establish T14. EIa.EIb. ⊃ .SSCab ≡ SSC2ab. T15. ‘SSC2’ is an equivalence over the restricted field of inscriptions, continuous or discontinuous, conforming to inscriptions properly included in environments.

Thus by relaxing this one restriction on ‘S2n+1’ we derive an extension of system I. Although ‘S22n+1’ is not generally transitive, we see that it is transitive in the interesting cases. We may now attempt further and more interesting simplifications. Consider system III, based on D333. D333. ┌S32n+1 ab┐ for ┌(Ea1)(Ea2)(Ea3)(Ea4)(Σ32n aa1.Ka1a2. (n ≧ 0). S32n a2a3.Ka3a4.Σ32n a4b) 3

This is just D33 with i = l. Since ‘S2n+1’ is not transitive, a basic problem will be to show that SSC3ac, where (8)

Σ32n aa1. Ka1a2. S32n a2a3.Ka3a4.Σ32n a4b, and Σ32n bb1.Kb1b2.S32n b2b3.Kb3b4.Σ32n b4c.

Basically, the line of proof will be as follows. We show that (i) for some m, S a b , Σ32m a2a3, and Σ32m b2b3. It follows that (ii) S32m+1 a2b3, by D333, and further, that (iii) S32m+2 a2b3. But now we can show that (iv) Σ32m+2 aa1 and 3 Σ2m+2 b4c, hence, with (iii) and ‘Ka1a2. Kb3b4’ from (8), we show, by D333, that (v) S32m+3 ac. But suppose that a2 is discontinuous. We derive (iii) from (ii) by considering a2 and b3 as single terms (i.e., instead of (iii) we actually have 3 2m 4 1

21

The systems constructed in this section will keep the symbolism of system I (as well as the numbering of definitions and theorems), but with numerical superscripts, ‘2’ for system II, etc. The symbols of system I appear without superscripts. Obviously, Kab ≡ K2ab, EIa ≡ EI2a, etc. Superscripts will ordinarily be dropped in such cases.

26

Noam CHOMSKY

proved the stronger ‘Σ32m+2 a2b3’). Thus we must permit discontinuous sequences to be terms in system III. This can be effected most simply by revising D17, dropping the requirement that a divisor must include the atom on the extreme right of each maximal continuous segment. D173. ‘DV3ab’ for ‘(Er) (BEGra. SEGrb. (x)(XRxb ⊃ SEGxr))’.

But again if a2 is discontinuous, since ‘S2n’ holds only between comparable (hence equally discontinuous) sequences, to assert (iii) we must revise D22 to permit a continuous and a discontinuous sequence to be comparable3 in this special case. D223. ‘CMP3 abcd’ for ‘E3dv abcd. (x)(y)(BEGxa.BEGyc.E3dv xbyd. ⊃ Edcxy).∨.XRba.XRdc’.

System III is then identical with system I except that D17, D22, and D33 are replaced, respectively, by D173, D223, and D333. We can now establish T13. EI3a ≡ (Ey)(Ew)(CONaw.Kyw). T16. EI3a.EI3b.EI3c. ⊃ :S3n ab.S3n bc. ⊃ SSC3ac. T17. ‘SSC3’ is an equivalence taken over the restricted field of inscriptions satisfying ‘EI3’.

Thus system III is applicable to both kinds of language. Before investigating the relation between systems II and III, we consider system IV, bearing the same relation to system III as did II to I, i.e., differing from III only in that D333 is replaced by D334, where D334 is just D333 with ‘Σ2n’ replaced by ‘S2n’ throughout. Systems III and IV are equivalent, thus T18. EI3a.EI3b. ⊃ .SSC3ab ≡ SSC4ab.

Relating systems I and II with system III we have the following, as the strongest such theorems. T19. EIa.EIb. ⊃ .SSCab ⊃ SSC3ab. T20. EI3a.EI3b. ⊃ .SSC2ab ⊃ SSC3ab.

Thus system III is not an extension of system I. This is illustrated in the following four-sentence text, in which primed and unprimed terms conform.

Systems of Syntactic Analysis (9)

27

‘azb’, ‘cz′d’, ‘a′rb′x’, ‘c′sd′x′’. 2

3

On the basis of the first two sentences we can assert that ‘S1 ’ and ‘S1 ’ hold between ‘a .. b’ and ‘c .. d’ (and trivially, between ‘x’ and ‘x′’). Hence 3 ‘S2 ’ holds between ‘a′ .. b′x’ (bracketed into ‘a′ .. b′’ and ‘x′’) and ‘c′ .. d′x′’ 3 (bracketed into ‘c′ .. d′’ and ‘x′’). Hence ‘S3 ’ and ‘SSC3’ hold between ‘r’ and 2 ‘s’. But ‘S2 ’ cannot hold between ‘a′ . . b′x’ and ‘c′ . . d′x′’. bracketed in this way, since discontinuous inscriptions cannot be terms. Hence neither ‘SSC’ nor ‘SSC2’ holds between ‘r’ and ‘s’. Of course we would naturally be inclined to say that a text like (9) belongs to a language with discontinuous constituents. This consideration suggests that it might be of interest to invert the procedure of this section and to attempt to determine the structural classification of languages formally in terms of the results given by application to them of various systems. Thus the distinguishing characteristic of a language of kind 1, as opposed to kind 2, would seem to be that its continuous inscriptions are partitioned by system I in exactly the same way as by system III. Although system III failed to be an extension of system I we can construct a system equivalent to system II even with the essential simplification of D33 which characterized system III. System V is the same as system III except that D303 and D313 are replaced as follows. D305. ┌S52n abcd┐ for ┌EQ54 (S52n-1 )abcd.XRba∨(x)(y)(SP5xabycd ⊃ . (n ≧ 1). CIx.CIy) ┐ 5 5 ┌ 5 ┐ ┌ ┐ D31 . S2n ab for (Ec)(Ed)(S2n acbd) . T21. EI5a.EI5b. ⊃ .SSC2ab ≡ SSC5ab.

This is the required theorem, since ‘EI5’ (or equivalently, ‘EI3’) is the condition that defines the restricted field of ‘SSC2’. Consider system VI, identical with system I except that D33 and D34 are replaced as follows. D336. ┌S62n+1 ab┐ for ┌(Ea1) .. (Ear)(CONaa1.Ka1a2.S62n a2a3.Ka3a4. CONa4a5.Ka5a6.S62n a6a7.Ka7a8.CONa8a9 ... CONar-4ar-3.Kar-3ar-2.S62n ar-2ar-1.Kar-1ar. CONarb)┐ , where n ≧ 0, r =4i, i ≧ 1. 6 6 ┌ 6 D34 . ‘SSC ab’ for S2n+1 ab┐, for some n ≧ 0. T22. SSC6ab ⊃ SSC2ab.

Since system VI partitions the inscriptions satisfying ‘EI3’, it gives a subpartitioning of that given by system II. That a significantly different

28

Noam CHOMSKY

notion of ‘syntactic category’ is involved here can be seen from the fact that T96 does not hold. It was this that necessitated the change of D34. Finally, consider system VII, differing from system I only in that D33 is replaced by D337. D337. ┌S72n+1 ab┐ for ┌(Ea1) ... (Ear)(CONaa1.Ka1a2.S72n a2a3.Ka3a4. S72n a4a5 ... S72n ar-2ar-1.Kar-1ar.CONarb) ┐, where n ≧ 0, r =4i, i ≧ 1. T23. EI3a. EI3b. ⊃ .SSC2ab ≡ SSC7ab.

These systems fall into two major descriptive groups. Group I, with an infinitely long definition of ‘S2n+1’ for each n, contains systems I, II, VI, and VII. Group II, with a finite definition for each n, contains systems III, IV, and V. On the basis of equivalence of systems, we have a cross-classification into system VI alone, and of the remainder, system I and its extensions II, V, and VII on the one hand, and systems III and IV on the other. Although the systems of group II sacrifice the simplicity of the associated theorems of group I, they have several redeeming features. Since for every pair of elements in a syntactic category we can determine the lowest n such that ‘Sn’ holds between them, the limitation, for each n, of ‘S2n+1’ results in a much finer subdivision of the syntactic categories into nested subcategories of increasingly divergent distribution, as reflected in increasing n. Furthermore it is clear that because of the restriction to continuous terms, the extensions of system I, although applicable in our terms, can not give interesting results in general for languages of kind 2, as is clearly shown by the failure of system II in the above analysis of (9). The preceding discussion probably does not exhaust the constructional possibilities within the limits of this general program, but until a more refined set of criteria of adequacy is given, and along with it, a refined analysis of kinds of language, it seems premature to enter into an exhaustive analysis of all possible systems of this general form. Furthermore, there are certain peculiar special features of these systems. Thus if CONaa′, CONbb′, etc., we see that from ‘ax’ and ‘ya′’ (where a = a′, etc.), we can derive ‘S1xy’, in fact, ‘S1x′y′’; but we cannot do so from such pairs as ‘abxc’ and ‘a′yb′c′’, etc. It appears as well that even though system III is applicable to languages of kind 2, it cannot give interesting results for certain languages of this kind because of the limitation of terms of a sequence to through inscriptions. In both cases attempts at more general solutions lead to problems which are difficult to meet in the absence of clearer notions of adequacy.

Systems of Syntactic Analysis

29

REFERENCES Bar-Hillel, Y. 1950. “On Syntactic Categories”. Journal of Symbolic Logic 15. 1-16. Bloomfield, L. 1933. Language. New York: Henry Holt & Co. Goodman, N. 1951. The Structure of Appearance. Cambridge: Harvard University Press. 1952. “New Notes on Simplicity”. Journal of Symbolic Logic 17. 189-191. and W. V. Quine. 1947. “Steps Towards a Constructive Nominalism”. Journal of Symbolic Logic 12. 105-122. Harris, Z. S. 1951. Methods in Structural Linguistics. Chicago: Chicago University Press. 1952. “Discourse analysis”. Language 28. 1-30 Leonard, H. S. and N. Goodman. 1940. “The Calculus of Individuals and Its Uses”. Journal of Symbolic Logic 5. 45-55. Martin, R. M. and J. H. Woodger. 1951. “Towards an Inscriptional Semantics”. Journal of Symbolic Logic 16. 191-203. Quine, W. V. 1943. “Notes on Existence and Necessity”. Journal of philosophy. 40, 113-127. Wells, R. S. 1947. “Immediate Constituents”. Language 23. 81-117.

30

Noam CHOMSKY

Some Methodological Remarks on Generative Grammar* Noam CHOMSKY 1. These remarks are prompted by several critical discussions1 of recent work in descriptive syntax and syntactic theory carried out within the framework of generative grammar.2 I will try to clarify several issues that have, apparently, been the source of some misunderstanding and confusion as to the purposes and assumptions of this work, and to indicate how it can perhaps be extended into certain important but neglected areas.3 It is useful, at the outset, to make a distinction between data and facts. The linguist’s data consist of certain observations about the form and use of utterances. The facts of linguistic structure that he hopes to discover go well beyond these observations. A grammar of a particular language is, in effect, an hypothesis about the principles of sentence formation in this language. It represents a factual claim concerning the rules that underlie the data that have been collected. We judge the truth or falsity of this hypothesis by considering how well the grammar succeeds in organizing the data, how satisfying an explanation it provides for a wealth of empirical observations, how far-reaching are its generalizations, how successfully it accomodates new data. An enormous amount of data is available to any native speaker; the deeper facts of linguistic structure, however, remain hidden to him. Similarly, a theory of linguistic structure is an hypothesis about linguistic universals. It asserts that the grammars of all human languages are constructed on such-and-such a specified plan. Such a theory should *

1

2 3

Reprinted by permission from Word, 17: 2. 219-239 (August, 1961). The Center for Research on Language and Culture, Senshu University would like to express our thanks to Professor Noam Chomsky and the International Linguistic Association. Hill, “Grammaticality,” Word XVII (1961) (referred to henceforth as [G]); Jakobson, “Boas’ view of grammatical meaning,” The Anthropology of Franz Boas, American Anthropologist (1959) (henceforth, [BVGM]); Bolinger, “Linguistic Science and Linguistic Engineering,” Word XVI (1960) (henceforth, [LSLE]). As exemplified, in particular, in my Syntactic Structures, The Hague, 1957 (henceforth [SS]). Much of what I say here is a paraphrase of earlier remarks, e.g., in [SS] and in “A transformational approach to syntax,” Third Texas Conference on Problems in the Analysis of English, A. A. Hill, editor, 1962 (henceforth, [TAS]).

32

Noam CHOMSKY

explicitly characterize the form of grammars (the schema for grammatical description) and should provide a method for selecting among alternative grammars, each compatible with the available data from some language (and each making a specific set of factual claims concerning this language). A theory of this sort can be verified in an indirect, but quite satisfactory way. We can ask whether, for each language, the most highly valued grammar of the specified form is confirmed in the sense described in the preceding paragraph. A linguist who is exclusively concerned with linguistic data and not with the facts of language, in this broader sense, has severely limited the scope of his research.4 A grammatical description that gives nothing more than “a compact one-one representation of the stock of utterances in the corpus” 5 can be challenged only to the extent that the observations it summarizes or rearranges are defective. It claims little, and its interest is correspondingly limited. In [SS] I was concerned with grammars that make a much stronger factual claim than this. A generative grammar, in the sense of [SS], is not a large collection of neatly organized examples, supplemented with comments about these examples and hints as to how to construct similar ones. Nor is it a discussion of efficient and compact notations (e.g., inventories of phonemes, morphemes, categories or construction types) in terms of which the utterances of a corpus can be represented. A generative grammar is a system of explicit rules that assign to each sequence of phones, whether of the observed corpus or not, a structural description that contains all information about how this sequence of phones is represented on each of the 4

5

In his [LSLE], Bolinger distinguishes between data-oriented and model-oriented approaches to linguistic description, the implication being that ‘model-oriented’ studies are less concerned with the facts of language as such. The essential difference between the two approaches that Bolinger describes seems to me, rather, that his ‘data-oriented approach’ limits its concern to a narrow subset of more immediately accessible facts; it does not attempt the precise formulation of deeper underlying generalizations. It should be remembered, in this connection, that the most partial and limited set of explicitly formulated recursive (iterative) rules is, literally, far more extensive in coverage than the most immense collection of examples assembled and arranged by the data-oriented linguist. We return to this matter below. At one point (p. 366) in his Methods in Structural Linguistics (Chicago, 1951), Harris describes this as “the over-all purpose of work in descriptive linguistics.” Later, however, he remarks (p. 372) that “the work of analysis leads right up to the statements which enable anyone to synthesize or predict utterances in the language.” It is this set of statements that I am calling the grammar. I am not convinced that the last-quoted remark is really justified (that is, that any known procedures do provide an adequate grammar), but this is another question. At the moment I am only concerned to make the difference in aim explicit.

Some Methodological Remarks on Generative Grammar

33

several linguistic levels—in particular, information as to whether this sequence of phones is a properly formed or grammatical sentence6 and if not, in what respects it deviates from well-formedness.7 In particular, then, this grammar distinguishes a class of perfectly well-formed (fully grammatical) sentences. It is designed, in other words, to meet what Hockett has proposed as the basic test of significance for a grammar, namely, that it “generate any number of utterances in the language, above and beyond those observed in advance by the analyst—new utterances most, if not all of which will pass the test of casual acceptance by a native speaker.”8 To Hockett’s remark I should only like to add that a grammar must not merely specify an infinite class of properly formed utterances, but must also assign to each sequence of phones a structural description that provides a basis for explaining how this utterance is used and understood—that provides the structural information without which it is impossible to undertake this further task in a serious way. Three theories of generative grammar are proposed for comparison and study in [SS]. The first (based on the notion of finite state Markov process or finite automaton) was shown to be too weak. The second was a formalized and somewhat generalized version of the theory of immediate constituents, reinterpreted as a theory of generative grammar. The third was a transformational model. I tried to show, by considering several kinds of linguistic data, that the immediate constituent theory was inadequate, and that the transformational theory seemed much more promising. Additional work that has appeared since then9 seems to me to give support to this claim, 6

7

8

9

In grammars of the type studied in [SS], part of the structural description of a phone sequence is its representation on the phrase structure level by a set of abstract strings that can be presented, equivalently, as a tree with labelled nodes. See [SS], and references there, for details. The fully grammatical sentences generated by the grammar are the ones represented by trees headed by a single node labelled S, where S is the distinguished initial symbol of the underlying constituent structure grammar. Transformational rules also play a part in assigning a phrase structure representation, in that theory, although the question exactly how they do so has not yet been satisfactorily answered. For discussion of this see my Logical Structure of Linguistic Theory (mimeographed, 1955, henceforth, [LSLT]), of which [SS] is essentially an excerpt, and “On the notion ‘rule of grammar’,” Proceedings of the Symposia on Applied Mathematics, XIII (1961). That is, sentences not generated directly can still be assigned a degree of grammaticalness, and a structural description which will indicate how they are related to directly generated sentences. A chapter of [LSLT] is devoted to this question, which is mentioned only in passing in [SS] (footnotes 2 and 7 of chapter 5). We return to this below in §5. “Two Models of Grammatical Description,” Word X (1954), reprinted in Joos, ed., Readings in Linguistics (Washington, 1957). See also Hockett’s “A Note on ‘Structure’,” International Journal of American Linguistics XIV (1948), also reprinted in Joos’ Readings. E.g., [TAS] (which is, again, largely an excerpt from [LSLT]) and Lees’ Grammar of English Nominalizations (Baltimore, 1960), henceforth referred to as [GEN].

34

Noam CHOMSKY

which, I should like to emphasize again, is a factual claim concerning English and, more generally, the structure of natural languages— a claim that can be proven false in several ways; e.g., by the construction of grammars of a different form that are more successful in handling data of the kind that originally motivated the theory, by a proof that any grammar of this form must fail, for some natural language, because of certain formal features of utterances in this language, etc. 2. The investigations of generative grammar described in [SS] were motivated in part by an interest in the problem of accounting for the ability of a speaker to produce and understand an indefinite number of new sentences (or for that matter, to recognize them as properly formed, or as deviating from well-formedness in one or another respect) ([SS], p. 15), a task that he performs regularly with great facility. A generative grammar can be regarded as an attempt to characterize certain aspects of this ability, and a particular theory of generative grammar is a proposal concerning its general and universal features.10 If we hope to make significant progress towards these goals, the grammars we construct must be sufficiently precise so that we can deduce from them consequences that can be confronted directly with linguistic data of a rich and varied sort. This requirement is rarely met by either traditional or modern studies,11 though this fact may be obscured by the informality of linguistic descriptions, whether traditional or modern. The study of precise generative grammars with explicit consequences (and the general form of such grammars) is in its infancy. Traditional grammars rely on the intelligence and linguistic intuition of the reader to provide the structural descriptions of items that do not appear in the 10

11

A theory of generative grammars can be regarded quite naturally as a proposal concerning certain fundamental and specific skills that the child brings to language learning. See my “Explanatory Models in Linguistics,” Proceedings of the International Congress on Logic, Methodology and Philosophy of Science (August 1960), 1962, P. Suppes, editor (henceforth [EML]); for some further discussion. With some very limited exceptions, e.g., inflectional paradigms of the traditional sort. Given a list of lexical items, a paradigm gives explicit instructions for constructing a finite set of inflected forms. But it is clear that this is a relatively minor success. It is on the level of syntax, of course, that the intuitive character of grammatical descriptions, and the defects of such descriptions, are most apparent, since it is here that the data cannot be enumerated in a finite list. However, I think that the preoccupation of modern linguistics with inventories rather than rules has had the effect of removing even significant aspects of phonology from the domain of linguistic research; for example, questions of the kind mentioned further below. For further observations on this see [EML], and Halle and Chomsky, Sound Pattern of English (1968, henceforth, [SPE]).

Some Methodological Remarks on Generative Grammar

35

collection of examples. There is little doubt that deep features of grammatical structure are unexpressed in a grammar addressed to an intelligent human who, in some totally unknown way, acquires the ability to produce and understand new sentences, having absorbed the information presented to him in a compendious traditional grammar. The study of generative grammars is, however, a natural outgrowth of traditional descriptive linguistics. Modern linguistics has, typically, been concerned with the much narrower problem of constructing several inventories of elements in terms of which utterances can be represented, and has given little attention to the rules that generate utterances with structural descriptions.12 3. Clearly if our goals are as described in §2, we will bring to bear data of many kinds in attempting to determine the validity of particular generative grammars or, more seriously, particular theories of generative grammar. Thus in [SS], [TAS], and elsewhere, I have been concerned with such data as (1)

(a) (b) (c) (d) (e) (f)

phonetic transcriptions; judgements of conformity of utterance tokens; judgments of well-formedness; ambiguity that can be traced to structural origins; judgments of sameness or difference of sentence type; judgments concerning the propriety of particular classifications or segmentations;

and so on.13 By (a), I refer to the kind of data given by impressionistic phonetics of the usual kind, e.g., transcriptions of stress contours, etc. 12

13

In [TAS] I noted a few exceptions to this. In particular, Harris’ morpheme-to-utterance rules can be reinterpreted as a model for generative grammar, and Hockett’s Markov process model was explicitly designed as a sentence-generating grammar. To avoid a possible misunderstanding, notice that I am suggesting that data of these kinds are relevant to determining the validity of grammars and linguistic theory, but not that these data should be utilized in the construction of or choice among grammars, insofar as this is specified by the theory of linguistic structure. The difference is fundamental. See [SS] and [EML] for further discussion. Thus we might try to develop a general theory of linguistic structure powerful enough to lead us from certain data about a language to a generative grammar of this language, limiting the data, in this context, to (a), (b), (c), that is, to the kind of data that we might regard as being available to the child who learns language. This might be a reasonable decision for someone whose motivation in studying generative grammars is to construct some sort of model for language learning. But such a person would still concern himself with such data as (d), (e), (f) in evaluating the success with which his theory of language learning duplicates the performance of the human.

36

Noam CHOMSKY

Category (b) includes the intuitive sameness-difference judgments that are fundamental to descriptive phonology, e.g., the observation that in my normal speech, latter and ladder are phonemically identical (i.e., tokens of them are) but writer and rider are phonemically distinct. To illustrate (c), I gave such examples as (2) (3) (4) (5)

colorless green ideas sleep furiously furiously sleep ideas green colorless the child seems sleeping the book seems interesting

and several others. Thus I am as certain of the fact that (2) and (5) are well-formed (in a sense in which (3) and (4) are not) as I am of particular impressionistic phonetic transcriptions (e.g., of the various light house keeper’s) or of particular judgments of sameness and difference. In connection with (d), I was concerned with such forms as flying planes can be dangerous, I found the boy studying in the library, etc. Examples of type (e) are of many kinds. Thus I discussed the problem of finding a grammatical explanation for the fact that did you eat the apple, who ate the apple, what did you eat, etc. are all naturally regarded as belonging to a special sentence type (as distinct from you ate the apple, etc.), and the problem of explaining the intuitively obvious difference between the police suspected the man behind the bar, . . . questioned the man behind closed doors, . . . put the man behind bars. Many other questions of the same kind were also raised, e.g., why smôking mén seems to mean “men who are on fire” rather than “men who smoke.” Concerning the propriety of segmentation and classification, it would be absurd not to use the fact that (6)

[it ended late ([iDêndid+léyt], [D]=alveolar flap)

obviously contains at least four morphemes in evaluating a theory of morphemic analysis, or not to use the fact that boy, horse, justice are, in a certain clear sense, words of a different type from compel, retain, bring, etc., in determining the success of a theory of word classes. Considerations of all of these kinds are familiar in traditional grammar. Considerations of kinds (a), (b), (f) are explicit in modern structural linguistics. Thus the methodological studies of American descriptivists are motivated by information of type (f), presumed known. If someone were to propose a theory of morphemic segmentation that led to the conclusion that (6) contains two morphemes, this consequence would, presumably, be regarded as grounds for rejecting the theory. The same is true of proposed

Some Methodological Remarks on Generative Grammar

37

methods (e.g., substitution procedures) for finding classes, for defining the phoneme, etc. These methodological studies (and the controversies to which they give rise) become utterly unintelligible if we do not make the perfectly justified assumption that they are an attempt to characterize precisely something that is, in part, intuitively given. The attempts to define ‘phoneme’, ‘morpheme’, etc. presuppose a set of clear cases of applicability or non-applicability of these terms, with reference to which the adequacy of the proposed definition can be tested. The same is true of an attempt to define ‘well-formed (grammatical) English sentence’ (that is, the attempt to construct an English grammar), or the attempt to define ‘well-formed (grammatical) sentence in an arbitrary language L’. In terms of the preceding discussion, these are attempts to discover and state the facts that underlie and account for the observed data. A refusal to use data such as (1) would eliminate linguistics as a discipline, just as surely as a refusal to consider what a subject senses or perceives would destroy psychophysics. In both cases, we are trying (though in very different ways) to find a basis for intuitive judgements. In both cases, furthermore, the difficulty of obtaining reliable and relevant reports is quite apparent. It seems to me perfectly plain that an enormous amount of data of this sort is available to the native speaker (though it does not follow that it is easy to elicit such data). It is pointless to refuse to make use of such data in evaluating the success of a particular attempt to formulate the rules of some grammar (i.e., the rules that the native speaker must somehow have internalized when he has achieved the ability to produce and understand new sentences) or the principles that underlie grammatical structure in general. It is data of the kinds mentioned above, it seems to me, that we are trying to characterize precisely, to account for and reduce to general underlying principles, when we study grammatical structure. A generative grammar can be thought of as a theory of the language that attempts to characterize and explain data of this sort. This point of view is discussed at length in [SS] (e.g., §§6, 8), [EML] and elsewhere, but I will illustrate what I have in mind by a few examples. In the case of data of type (la), a generative grammar will attempt to exhibit general rules that underlie particular phonetic observations. It will try to show how the stress contours of English phrases and words are related to and determined by their immediate constituent structure and segmental phonemes;14 it will offer an account of vowel reduction (as in the second 14

Cf. Chomsky, Halle, Lukoff, “On Accent and Juncture in English,” For Roman Jakobson (The Hague, 1956). A new version of this material, simplified, generalized, and incorporated into a much more general study of phonological processes, will appear in [SPE]. It is described in part in [EML] and (in an earlier version), in my contribution to the IVth Texas conference (1959).

38

Noam CHOMSKY

syllable of compensation, but not condensation, in my speech) in terms of IC and syllable structure (cf. [EML] and [SPE]); it will try to explain why such items as telegraph, aristocrat (in the contexts #—#, —+y, —+ic), etc., have a particular set of variants, instead of others, by showing that this is a result of application of independently motivated rules (cf. [EML] and [SPE]); and so on, in innumerable similar cases. A generative grammar will be judged successful—that is, it will be judged to have brought to light the underlying facts of linguistic structure—to the extent that it shows how a relatively small number of simple and systematic rules account for a substantial quantity of such data. Cf. §1, above. Similarly, in the case of data of the kind (1c), the generative grammar will attempt to exhibit interconnections among apparently independent judgments as to well-formedness and to state general rules from which a large variety of such judgments (e.g., those mentioned above (6)) can be deduced as consequences. Cf. [SS], [TAS], [GEN], for many examples. There is a second and independent way of approaching the study of data of kind (1). We can try to discover objective experimental techniques for characterizing these intuitive observations. Thus in the case of impressionistic phonetic observations of type (1a), we can try to discover objective acoustic or articulatory correlates to what the linguist hears as stress level, vowel quality, and so on. And in the case of data of kind (1c), we can try to construct behavioral tests that will correspond, in their results, to the intuitive judgments of well-formedness.15 Here, then, are two approaches to the data of linguistics. We can attempt to characterize the data by a generative grammar or by experimental tests. The two approaches are not alternatives. Obviously, both are justified and we should like to pursue them in such a way that they give convergent results. It is important to realize that each approach complements the other and that neither presupposes the other. Whichever approach we are pursuing, the criterion of success is provided by the data that give rise to the investigation;

15

Linguists have in general not been much concerned with this approach to linguistic data. Thus even in the case of (1b), where a reasonably good test has been suggested (namely, the paired-utterance test proposed by Harris, op. cit., pp. 32f.), there is not a single paper in the linguistic literature devoted to an objective evaluation of its results, despite the fact that the relation of conformity (sameness-difference) among utterance tokens has been regarded, since the publication of Bloomfield’s Postulates, as the cornerstone of structural linguistics. There is no doubt that this test will often fail to give the right results, in the form in which it is usually described. Some ways in which it can be improved have been suggested (e.g., in [SS], p. 96), but have not been systematically investigated.

Some Methodological Remarks on Generative Grammar

39

in this case, such data as (1).16 Suppose, for example, that someone proposes that impressionistic judgments of stress level correspond to physical measures of intensity. He will quickly discover that ‘stress’ as so defined does not at all correspond to the impressionistic transcriptions that constitute the phonetic record (1a). In the face of just this evidence, it would be absurd for him to conclude that the transcriptions are incorrect, that phonology is built on sand, etc. This lack of correspondence may simply show that he has chosen the wrong test.17 Similarly, if the paired utterance test, administered in a certain way, fails to show that writer and rider are phonemically distinct in my speech, this is evidence that the proposed test requires revision. Or, if a test for grammaticalness (i.e., a test of (1c)) is proposed which fails to make the required distinction between (2), (5) on the one hand and (3), (4) on the other, we can conclude only that the test is poorly constructed. In other words, we evaluate the success and relevance of an operational test (just as we evaluate the success of a generative grammar) by asking how well it corresponds to the given data. There is no difficulty in constructing a bad grammar. It is just as easy to construct a bad test. There is no interest in either. 4. I would like to turn now to some of the criticisms of [SS]. Consider first Hill’s [G]. What Hill shows is that several tests that he has tried are quite bad. I do not see the importance of these negative results, nor do I see any reason 16

17

To some extent, of course, theoretical investigations or operational tests may lead us to reject some of the original data as irrelevant, or as probably incorrect or contaminated. We can expect that one result of systematic study, of either kind, will be to improve our understanding of what are the natural bounds of the discipline or what phenomena are relevant to deepening our investigation of linguistic structure. Continued failure of attempts to find a test might, ultimately, shake our confidence in the phonetic record, but it is not easy to say just when this would be justified. Thus we might discover that what the phonetician hears as stress level is determined in part not by the acoustic event but by an ideal stress pattern that is assigned to the utterance by the rules of a generative grammar—that is, the phonetician may be “hearing” something that is not present in the sound wave in full detail, but that is implied by the grammatical structure of the sound wave that he has understood in a certain way. I think, in fact, that this is not at all out of the question, in the case of stress levels. Cf. [SPE]. Similarly, a continued failure to elicit the judgment that (2), (5) are distinct in some significant way from (3), (4) might ultimately shake our confidence in this particular judgment, at least for certain speakers. It is, I suppose, conceivable that some speakers of English are unaware of the difference between (2), (5) and similar examples, on the one hand, and (3), (4), (or “tall the man cigar the smoked black,” with whatever intonation —cf. [G], footnote 4), etc., on the other. Such a person, if he exists, is likely to be about as successful in studying syntax as a deaf man would be as a phonetician.

40

Noam CHOMSKY

to find them at all surprising.18 Thus if we ask a speaker the question “which of these sentences are grammatical?” giving him no idea how this question is to be interpreted, we will naturally expect the responses to be of no use in characterizing data of type (1c). Similarly, if we were to ask a speaker “what do you hear?” his responses would probably be of little phonetic relevance (they would have little relation to (1a)); or if we ask “were these utterances the same or different?” not indicating how we intend this question, we would expect the result to correlate poorly with the classification of utterance tokens as phonemically identical or distinct (cf. (1b)). Lack of correspondence, in such cases, demonstrates nothing about the validity of such observations as (1a), (1b), (1c). It merely indicates, as we should have expected in advance, that more subtle means will be necessary if we are to obtain relevant information about the speaker’s awareness.19 Hill does suggest one test which might yield some useful data. In this test, two sets of utterances are presented to a subject who is asked, essentially, whether the same formal relation holds in the two sets. Hill regards this test 18

19

Hill seems to have the impression that I have suggested these tests as general criteria of grammaticalness or degree of grammaticalness. But this is not the case. In [SS], I made it clear at the outset (p. 13) that I was not proposing any general behavioral test. I did assert (p. 16, as an aside, in the context of a discussion of order of approximation) that a speaker of English will treat (3), but not (2), as a sequence of unrelated words, assigning phrase final intonation to each, and that he will be able to recall (2) more easily than (3), to learn it more quickly, etc., despite the fact that they are equally new to him. I was careful not to suggest that these remarks offer a general criterion for grammaticalness—they would, for example, no doubt fail in the case of (4), (5). Later, on pp. 35-36, I made some further remarks on phonetic accompaniments to deviation from grammaticalness, again suggesting no general operational criterion. In fact, I know of none. There is only one apparent conflict between [SS] and Hill’s results, namely, he has found that his informants did not assign phrase final intonation to each word in (3), as I did. This lack of correlation is not at all surprising. In such quasi-operational tests as these, the instructions, the setting and so on can have a significant effect on what the subject does. It is for this reason, among others, that neither my comments nor his can be taken seriously as the description of an experimental procedure. Talk of “statistical reliability” ([G], p. 7) in connection with such informal suggestions as these betrays a misconception as to what is a serious scientific experiment. In any event, the question how the particular sequence of words (3) is read (nothing else is at stake) is surely too insignificant to justify a sustained discussion. This is noted explicitly by Maclay and Sleator in their “Responses to Language: Judgements of Grammaticalness,” International Journal of American Linguistics XXX (1960), 281. Their tests apparently gave somewhat better and more relevant results than the one that Hill describes. In his “Language, Purposive Behavior and Evolution” (mimeographed, 1958), Eric Lenneberg describes a test of grammaticalness applied to (2) and randomly chosen rearrangements of the words of (2) which, given to M.I.T. undergraduate language students with appropriate instructions (there stated), gave the expected results unequivocally.

Some Methodological Remarks on Generative Grammar

41

as “more structural” than a direct test of grammaticalness, and considers it somehow more relevant to the “transformationist.” 20 Insofar as I can reconstruct the reasoning that leads to this conclusion, it seems to me to be based on two confusions. The first has to do with the nature of transformational relations. Thus some of the sets that Hill presented in his test contain sentences that are transformationally related to one another, but others do not. If Hill considers this test particularly relevant to transformational grammar, it may be in part because he has misinterpreted the technical use of the term “grammatical transformation” in [SS], [TAS], [GEN], and elsewhere, taking it to refer to any formally statable relation among sentences. I do not think it has ever been used in this sense. Secondly, there is apparent in Hill’s paper a confusion about the relation between the terms “generative grammar” and “transformational grammar.” Hill seems to identify the two. Similarly, Bolinger regards Jakobson’s comments about grammaticalness (to which I return below) as having some particular relevance to “transformationalists” ([LSLE], p. 377). In fact, neither Hill’s nor Jakobson’s comments have any direct bearing on the transformational model for generative grammar. They apply rather to generative grammar as such, that is, to the attempt to construct precise grammars with explicit consequences as to well-formedness of utterances. Grammaticalness is no specific concern of those who happen to think that the particular transformational theory proposed in [SS], [GEN] and elsewhere is the most promising specific characterization of generative grammars. It is, rather, the concern of anyone who is interested in syntax in the traditional sense, or in the problems discussed above in §2. One final set of criticisms of transformational grammar presented in [G] also requires some clarification. Hill is under the impression that those of us who are exploring transformational models for generative grammar believe that sentences are primarily to be regarded as strings of printed words, and that we reject the view that “language is basically a system of sound, only secondarily one of writing.” This, he believes may “explain the hostility of many structuralists.” In [SS], [GEN], and elsewhere, a grammar is regarded as consisting of a syntactic component and a morphophonemic component. The former generates an infinite class of strings of minimal syntactically functioning units. Each such string, furthermore, automatically receives a 20

It should be observed that this test, as presented, is not yet very useful. Thus it does not tell us which of the two sentences “the child is sleeping,” “the child seems sleeping” is to be rejected as deviating from well-formedness. Furthermore, it does not seem to be generalizable to such distinctions as (2)-(3). However, it can perhaps be refined, and it should be considered as one of a battery of tests that may ultimately succeed in giving a characterization of grammaticalness.

42

Noam CHOMSKY

structural description. The morphophonemic component converts each item given as output by the syntactic component into, ultimately, the phonetic representation of an utterance. Both [SS] and [GEN] are devoted primarily to the syntactic component. Other studies have been primarily devoted to the morphophonemic component of a generative transformational grammar.21 The fact that the output of a generative grammar is a representation of speech is made quite clear in [SS] (e.g., p. 46) and [GEN], however.22 Jakobson, in his [BVGM], also indicates his dissatisfaction with the use to which such examples as (6) (7)

colorless green ideas sleep furiously furiously sleep ideas green colorless

were put in [SS]. This criticism (which is cited with approval in Bolinger’s [LSLE]) I find quite curious and puzzling. What Jakobson says about these examples, and about the question of grammaticalness and degree of grammaticalness in general, seems to me in full support of the position maintained in [SS]—in fact, it is in part no more than a paraphrase, in a slightly different terminology. It seems to me, therefore, that Jakobson (and Bolinger) must have completely misinterpreted the point of view expressed in [SS]. In [SS] examples (6) and (7) were used to illustrate the futility of (a) a search for a semantically based definition of “grammaticalness,” or (b) an attempt to define grammaticalness in terms of order of approximation. Thus (6), but not (7), is clearly a well-formed sentence (or, at least, more like one 21

22

In particular, Halle’s Sound Pattern of Russian (The Hague, 1960). This is also the central topic in my “Transformational Basis for Syntax,” to appear in the Proceedings of the IVth Texas conference, 1959, A. A. Hill, editor; and it is discussed in a section of [TAS]. What led Hill to conclude that transformational grammar is concerned primarily with writing rather than speech was apparently my remark in [SS] that “in the context ‘I saw a fragile—’, the words ‘whale’ and ‘of’ may have equal (i.e., zero) frequency in the past linguistic experience of a speaker who will immediately recognize that one of these substitutions, but not the other, gives a grammatical sentence” (p. 16). The point in that context, was to show that judgment of well-formedness is not based on statistical order of approximation. This is rather obvious, and hardly deserves more than this casual comment. It is true that the phrasing of this observation makes use of an appeal to the literacy of the audience. It would have been possible to make the same point, more elaborately, without this appeal. That is, I could have stated that in the context Σ— (where Σ stands for a phonetic transcription of “I saw a fragile”, with full intonation, etc.), /wéyl/ will be interpreted as the familiar lexical item whale, / əv / will be rejected with stress 4, and, with stress 1, either rejected or treated as a new, unfamiliar lexical item, or as the quotation form of. I suppose that one could add even further qualifications, but there are other places where preciseness and rigor are likely to be more rewarding.

Some Methodological Remarks on Generative Grammar

43

than (7)), although (a) neither (6) nor (7) seems to be “meaningful” in any sense independent of grammatical structure and (b) the two are not differentiated in terms of statistical order of approximation. The same examples could be used to illustrate the impossibility of (c) defining “grammaticalness” in terms of a frame of “grammatical morphemes,” since (8) has the same frame as (6) and (9) the same frame as (7) (note that it would be begging the question to argue that the morphemes ly, etc., are different in (6) and (8)) (8) (9)

harmless seem dogs young friendly (—less — —s — —ly) friendly young dogs seem harmless (—ly — —s — —less).

What Jakobson points out is that we can recognize the grammatical structure of (6) (but not (7)), and that “these grammatical relations create a meaningful sentence.” He also observes that only (6) could appear in a poem,23 that we know how to question it, that we can recognize in it certain metaphors, etc. These remarks are precisely in support of the argument in [SS] that (6) but not (7) is grammatical, and that the basis for whatever meaningfulness we can assign to it is its independently recognizable grammatical structure. The fact that the independently recognized “grammatical relations create a meaningful sentence” is perfectly consistent with the claim in [SS] that grammaticalness is not attributed to an utterance by virtue of the fact that the utterance is recognized to be meaningful. To argue against this claim, one would have to maintain that there is an absolute semantic property of “meaningfulness” that can be assigned to utterances quite independently of any consideration of their grammatical structure—a property that can be shown to hold of (6) but not (7). It is difficult to believe that anyone would seriously uphold this view. From consideration of such examples as (6), (7) (and many others of different kinds—cf. [SS], §9) it seems evident that perception of grammatical 23

Hill, on the contrary, considers that (7) but not (6) could be a line of modern poetry ([G], p. 4). Both Hill and Jakobson refer to poems written to prove these (opposing) points. In fact, the question whether a sequence of words might appear in a poem is entirely beside the point, since it is perfectly plain that deviation from well-formedness is not only tolerable, in prose or poetry, but can even be effectively used as a literary device. We return to this below. It is by no means a novel observation. Cf., e.g., Empson, Seven Types of Ambiguity (Meridian), p. 34: “The demands of metre allow the poet to say something which is not normal colloquial English, so that the reader thinks of the various colloquial forms which are near to it, and puts them together; weighting their probabilities in proportion to their nearness. It is for such reasons as this that poetry can be more compact, while seeming to be less precise, than prose. It is for these reasons, too, among others, that an insensitivity in a poet to the contemporary style of speaking . . . is so disastrous . . .”

44

Noam CHOMSKY

relations in a sentence does not depend on a prior and independent identification of semantic properties, and that the study of grammatical structure seems to be, in fact, quite independent of meaning. Furthermore, as noted in [SS], p. 101, “it seems that the study of meaning is fraught with so many difficulties even after the linguistic meaning-bearing elements and their relations are specified that any attempt to study meaning independently of such specification is out of the question.24 To put it differently, given the instrument language and its formal devices, we can and should investigate their semantic function (as, e.g., in R. Jakobson, “Beitrag zur allgemeinen Kasuslehre,” TCLP (1936)); but we cannot, apparently, find semantic absolutes, known in advance of grammar, that can be used to determine the objects of grammar in any way.”

Nothing in Jakobson’s remarks gives any reason to revise these conclusions. He is saying that having recognized a sentence as grammatical and having identified its grammatical relations, a speaker can proceed to interpret it, determining how these formal elements and constructions are functioning, in this case. It will, in general, be easier to invent some sort of an interpretation for a grammatical than for an ungrammatical sequence, as Jakobson observes. This is basically the point of view expressed in [SS].25 In this connection, we can turn to Bolinger’s discussion of what he regards as the semantic basis for transformational grammar ([LSLE], p. 377). He asserts that “knowing the meaning of seem, we can predict with some assurance that the structure *He is seeming is not likely to occur.” Under one interpretation, this remark is perfectly true. In just the same way we could say that knowing the meaning of seem, we can predict that this element will not appear in the context I saw—boy, or knowing the meaning of in, we can

24 25

This is taken for granted in investigations that are actually concerned with meaning. Cf., e.g, Ziff, Semantic Analysis (Ithaca, 1960). The temporal implications in this description should not be taken too literally. However, I think that a reasonable perceptual model would have the property that grammatical structure is identified independently of any semantic consideration, and a reasonable model for language learning would have the parallel property. This is a large topic in itself, and I will not attempt to pursue it here. I would merely like to emphasize that those who regard semantics as providing the basis, in some sense, for grammar, may not have realized how extreme and implausible is their claim, if stated quite openly. The claim must be that there are semantic absolutes, identifiable in noises independently of the grammatical structure (in particular, the lexical items) assigned to them. Perhaps some weaker claim than this is intended, but if so, it has never, to my knowledge, been clearly formulated.

Some Methodological Remarks on Generative Grammar

45

predict the ungrammaticalness of I saw in boy, etc.26 The comment is useful, however, only if we are presented with some characterization of the meaning of seem, in, etc., stated without reference to their grammatical function, from which this function can be predicted.27 In the absence of this, Bolinger’s statement reduces to the observation, misleadingly expressed, that seem does not appear in the progressive. In general, study of the continuing controversy over the reliance of grammar on meaning seems to me to show that the issue is only verbal—that no substantive claim is being made by those who claim that such reliance is essential. In apparent conflict with [SS], Jakobson also argues that the notion “degree of grammaticalness” is untenable. Here too, however, I can find only a terminological issue. Jakobson recognizes ([BVMG), final paragraph) that there is some sort of scale of “obliteration” of “syntactic forms and the relational concepts which they carry,” and he would, of course, insist on the importance of the distinction between literal and figurative usage. It is precisely such observations as these that motivate the study of deviation from grammaticalness to which he expresses his objection. Clearly the only question is whether the study of these problems is a natural outgrowth of grammatical investigations—that is, the question is to what extent considerations of the kind that are fundamental to the study of principles of sentence formation can provide some systematic account of such phenomena. I think that to some extent they can, and the investigation of degrees of grammaticalness (or, if one prefers, the scale of obliteration of syntactic forms . . .) is an attempt to show how a simple generalization of some familiar notions can give some insight into these problems. 5. Since the point has been widely misunderstood, I would like to emphasize that I am using the terms ‘grammatical’ and ‘degree of grammaticalness’ in a technical sense (which is, however, not unrelated to the ordinary one). In particular, when a sentence is referred to as semi-grammatical or as deviating from some grammatical regularity, there is no implication that this sentence is being “censored” ([BVGM], p. 144) or 26

27

The phrase “likely to occur” is often used in place of some such technical term as “grammatical.” Its use gives an appearance of objectivity which is, however, quite misleading. In the literal sense of these words, practically nothing can be predicted about what is likely to occur in speech. It should be emphasized again that in studies devoted to meaning, the argument is usually reversed. Thus it is common to find the claim that knowing, seeing, etc., are not processes supported by the observation that know, see (like seem) do not naturally occur in the progressive, in their usual senses.

46

Noam CHOMSKY

ruled out, or that its use is being forbidden. Nor, so far as I can see, are there any “ontological” considerations involved ([LSLE], p. 377, [BVGM], p. 144), except insofar as these are reflected in grammatical categories and subcategories. Use of a sentence that is in some way semi-grammatical is no more to be censured than use of a transform that is remote from the kernel. In both cases, what we are attempting to do is to develop a more refined analysis of sentence structure that will be able to support more sophisticated study of the use and interpretation of utterances. There are circumstances in which the use of grammatically deviant sentences is very much in place. Consider, e.g., such phrases as Dylan Thomas’ “a grief ago,”28 or Veblen’s ironic“perform leisure.” In such cases, and innumerable others, a striking effect is achieved precisely by means of a departure from a grammatical regularity. Given a grammatically deviant utterance, we attempt to impose an interpretation on it, exploiting whatever features of grammatical structure it preserves and whatever analogies we can construct with perfectly well-formed utterances. We do not, in this way, impose an interpretation on a perfectly grammatical utterance (it is precisely for this reason that a well-chosen deviant utterance may be richer and more effective—cf. footnote 23). Linguists, when presented with examples of semi-grammatical, deviant utterances, often respond by contriving possible interpretations in constructed contexts, concluding that the examples do not illustrate departure from grammatical regularities. This line of argument completely misses the point. It blurs an important distinction between a class of utterances that need no analogic or imposed interpretation, and others that can receive an interpretation by virtue of their relations to properly selected members of this class. Thus, e.g., when Jakobson observes ([BVGM], p. 144) that “golf plays John” can be a perfectly perspicuous utterance, he is quite correct. But when he concludes that it is therefore as fully in accord with the grammatical rules of English as “John plays golf,” he is insisting on much too narrow an interpretation of the notion “grammatical rule”—an interpretation that makes it impossible to mark the fundamental distinction between the two phrases. The former is a perspicuous utterance precisely because of the series of steps that we must take in interpreting it—a series of steps that is initiated by the recognition that this phrase deviates from a certain grammatical rule of English, in this case, a selectional rule that determines the grammatical categories of the subject and object of the verb “play.” No such steps are necessary in the case of the nondeviant (and uninteresting) “John plays golf.” 28

One of the examples analyzed in Ziff’s interesting study of the problem of deviation from grammaticalness, “On Understanding ‘Understanding’ ” (mimeographed, 1960).

Some Methodological Remarks on Generative Grammar

47

I am not, of course, suggesting that every difficult, interesting or figurative expression is semi-grammatical (or conversely). The important question, as always, is to what extent significant aspects of the use and understanding of utterances can be illuminated by refining and generalizing the notions of grammar. In the cases just mentioned, and many others, I think that they can. If this is true, it would be arbitrary and pointless to insist that the theory of grammatical structure be restricted to the study of such relatively superficial matters as agreement, inflectionally marked categories, and so on.29 In short, it seems to me no more justifiable to ignore the distinctions of subcategory that give the series “John plays golf,” “golf plays John,” “John plays and,” than to ignore the rather similar distinctions between seeing a man in the flesh, in an abstract painting, and in an inkblot. The fact that we can impose an interpretation in the second case and sometimes even the third, using whatever cues are present, does not obliterate the distinction between these three strata. Examples such as these provide a motive for the study of degrees of grammaticalness. Thus in addition to such data as (1), we can try to account for the observation that such phrases as (10) are not as extreme in their violation of grammatical rules as (11), though they do not conform to the rules of the language as strictly as (12): (10) a grief ago; perform leisure; golf plays John; colorless green ideas sleep furiously; misery loves company; John frightens sincerity; what did you do to the book, understand it? (11) a the ago; perform compel; golf plays aggressive; furiously sleep ideas green colorless; abundant loves company; John sincerity frightens; what did you do to the book, justice it? (12) a year ago; perform the task; John plays golf; revolutionary new ideas appear infrequently; John loves company; sincerity frightens John; what did you do to the book, bite it? 29

Notice that if we do, arbitrarily, limit the study of grammar in this way, we cannot even account for the difference between (6) and (9), on the one hand, and (7) and (8), on the other, since this difference can be expressed only in terms of categories that are established in terms of syntactic considerations that go well beyond inflection. But if we distinguish (6) from (9) by rules involving such syntactic categories as Adjective, Noun, etc., we can just as well distinguish “John plays golf” from “golf plays John” by rules involving such syntactic subcategories as Animate Noun, etc. These are simply a refinement of familiar categories. I do not see any fundamental difference between them. No general procedure has ever been offered for isolating such categories as Noun, Adjective, etc., that would not equally well apply to such subcategories as are necessary to make finer distinctions. I return to this below.

48

Noam CHOMSKY

Here too, we can find innumerable relatively clear cases, and we can attempt to express these distinctions in a generative grammar (and, more importantly, we can try to find some basis for them through the study of generative grammar). The question then arises: by what mechanism can a grammar assign to an arbitrary phone sequence a structural description that indicates its degree of grammaticalness, the degree of its deviation from grammatical regularities, and the manner of its deviation (cf. §1). This is a natural question to ask within the framework of §2. Suppose that we have a grammar that generates an infinite set of utterances with structural descriptions. Let us call the units in terms of which these utterances are represented by the neutral term formatives (following a suggestion of Bolinger’s). Suppose, in addition, that we have an m-level hierarchy of categories of formatives with the following structure. On level one we have a single category denoted C11, the category of all formatives. On level two, we have categories labelled C21, . . ., C2n2. On level three, we have categories C31 , . . ., C3n3, where n3 > n2, and so on, until we reach the mth level with categories Cm1, . . ., Cmnm (1 < n2 < . . . < nm). On each level, the categories are exhaustive in the sense that each formative belongs to at least one, perhaps more (in the case of grammatical homonymy). We might also require that each level be a refinement of the preceding one i.e., a classification into subcategories of the categories of the preceding level. Let us assume, furthermore, that the mth level categories are the smallest categories that appear in the rules of the generative grammar. That is, the members of Cm1 are mutually substitutable in the set of generated utterances. Many of them may contain just a single formative. For concreteness, think of the formatives as English words.30 Suppose we have a three-level hierarchy. Then C11 is the class of all words. Let C21 = Nouns, C22 = Verbs, C23 = Adjectives, C24 = everything else. Let C31, . . ., C3j be subcategories of Verbs (pure transitives, those with inanimate objects, etc.); subcategories of Nouns, and so on. Every sequence of words can now be represented by the sequence of first level, second level, third level categories to which these words belong. Thus “misery loves company” is represented C11 C11 C11 on level one, C21 C22 C21 (i.e., NVN) on level two, Nabstr VkNabstr on level three (where these are the appropriate C3i’s). One of the selectional rules of the generative grammars (i.e., in the transformational model of [SS], one of the context-restricted constituent structure rules) will specify that Vk occurs only with animate subjects. Thus “misery loves company” will not be generated by the grammar, though “John loves company” will. However, 30

This is merely an illustrative example.

Some Methodological Remarks on Generative Grammar

49

“misery loves company” has a level two representation in common with a generated utterance, namely, NVN. We therefore call it semi-grammatical, on level two. “Abundant loves company,” on the other hand, has only a level one representation in common with a generated utterance, and is therefore labelled completely ungrammatical. Without going into details, it is obvious how, in a similar way, a degree of grammaticalness can be assigned to any sequence of formatives when the generative grammar is supplemented by a hierarchy of categories. The degree of grammaticalness is a measure of the remoteness of an utterance from the generated set of perfectly well-formed sentences, and the common representing category sequence will indicate in what respects the utterance in question is deviant.31 The more narrowly the mth level categories circumscribe the generated language (i.e., the more detailed the specification of selectional restrictions) the more elaborate will be the stratification of utterances into degrees of grammaticalness. No utterances are “lost” as we refine a grammatical description by noting more detailed restrictions on occurrence in natural sentences. By adding a refinement to the hierarchy of categories, we simply subdivide the same utterances into more degrees of grammaticalness, thus increasing the power of the grammar to mark distinctions among utterances.32 Thus a generative grammar supplemented with a hierarchy of categories can assign a degree of grammaticalness to each sequence of formatives. If we could show how a hierarchy of categories can be derived from a generative grammar, then the latter alone would assign degree of grammaticalness. There are, in fact, several ways in which this might be possible. Notice, first, that a transformational grammar will have such symbols as Noun, Adjective, . . .(in addition to much narrower subcategories) at intermediate levels of representation, even if it is designed to generate only a narrow class of highly grammatical sentences, since these larger categories will simplify 31

32

We can represent only one “dimension” of deviation from grammaticalness in this way. There are others. Cf., e.g., [SS], §5, footnote 2. In obvious ways, we could give a more refined stratification of utterances by considering their parts, but I will not go into this. What is the natural point where continued refinement of the category hierarchy should come to an end? This is not obvious. As the grammatical rules become more detailed, we may find that grammar is converging with what has been called logical grammar. That is, we seem to be studying small overlapping categories of formatives, where each category can be characterized by what we can now (given the grammar) recognize as a semantic feature of some sort. If this turns out to be true in some interesting sense when the problem is studied more seriously, so much the better. This will show that the study of principles of sentence formation does lead to increasingly deeper insights into the use and understanding of utterances, as it is continually refined.

50

Noam CHOMSKY

the descriptions of the domains of transformational rules. Thus we can expect to find a hierarchy of categories embedded within the constituent structure rules of the transformational grammar. This might be the appropriate hierarchy, or a step towards its construction.33 We might approach the question of projecting a hierarchy of categories from a set of utterances in a different way, by defining “optimal k-category analysis,” for arbitrary k. Suppose, for simplicity, that we have a corpus of sentences all of the same length. Let C1, .., Ck be (perhaps overlapping) categories that give an exhaustive classification of the formatives appearing in the corpus. Each sentence is now represented by at least one category sequence. Each such category sequence, in turn, is the representation of many sequences of formatives, in particular, of many that may not be in the original corpus. Thus a choice of k categories extends the corpus to a set of sentences that are not distinguishable, in terms of these categories, from sentences of the corpus. It is natural to define the optimal k-category analysis as that which extends the corpus the least, i.e., which best reflects substitutability relations within the corpus. Given, for each k, the optimal k-category analysis, we might select the optimal k-category analysis as a level of the hierarchy if it offers a considerable improvement over the optimal k-l-category analysis, but is not much worse than the optimal k+l-category analysis (this could be made precise, in various ways). It is easy to see that there are circumstances under which the optimal k-category analysis might contain overlapping classes (homonyms)34 It is also easy to drop the restriction that all sentences be of the same length, and that the

33 34

This possibility was suggested by some remarks of R. B. Lees. In general, it is to be expected that overlapping of categories will lead to an extension of the set of generated sentences, since categories will now be larger. Therefore, in general an analysis with disjoint categories will be preferred, by the evaluation procedure suggested above, over an analysis with an equal number of overlapping categories. Suppose, however, that the overlap includes true homonyms—suppose, e.g., that the categories N and V are allowed to overlap in such elements as /riyd/ (read, reed), etc, We now have two ways of representing the sentences read the book (namely, VTN or NTN), the reed looks tall (TNVA or TVVA), and so on, instead of just one (e.g., VTN and TVVA, if /riyd/ is assigned to V). We can select, in each case, the representation which is required, on independent grounds, by other sentences, i.e., VTN and TNVA, in this example. In this way we can reduce the number of generated sentences by allowing categories to overlap. Overlapping of categories will be permitted, then, when the gain that can be achieved in this way more than compensates for the loss resulting from the fact that categories are larger. We might inquire then whether homonyms can be defined as elements that are in the overlaps in the optimal set of categories on some level. Some evidence in favor of this assumption is presented in [LSLT].

Some Methodological Remarks on Generative Grammar

51

corpus be finite. Such suggestions as these, when made precise,35 offer an alternative way in which the generative grammar itself may impose degrees of grammaticalness on utterances that are not directly generated, through the intermediary of the category hierarchy projected from the set of generated sentences. This suggestion is schematic and no doubt very much oversimplified. Nevertheless, such an approach, as this to the problem of defining syntactic categories has many suggestive features, and offers some important advantages over the alternatives (e.g., substitution procedures)36 that have been described in the literature (cf. [LSLT] for a detailed discussion—in particular, it allows for the possibility of setting up a hierarchy of categories and subcategories and for a principled and general solution to the problem of recognizing homonyms). I mention it here to indicate one way in which the further investigation of deviation from grammaticalness might be systematically pursued. Massachusetts Institute of Technology Cambridge 39, Mass. REFERENCES Bloomfield, L. 1926. “A Set of Postulates for the Science of Language”. Language 2. 153-164. (Reprinted in International Journal of American Linguistics 15. 195-202, 1949; in Readings in Linguistics. Ed. by M. Joos, 19-25, Washington: American Council of Learned Societies, 1957; and in A Leonard Bloomfield Anthology. Ed. by C. F. Hockett, 128-140, 1970, Bloomington & London: Indiana University Press.) Bolinger, D. 1960. “Linguistic Science and Linguistic Engineering”. Word 16. 374-391. Chomsky, N. 1955. Logical Structure of Linguistic Theory. Mimeographed. (Published New York: Plenum Press, 1975) 1957. Syntactic Structures. The Hague: Mouton. 1958. “A Transformational Approach to Syntax”. In Proceedings of the Third Texas Conference on Problems of Linguistic Analysis in 35

36

This approach to degrees of grammaticalness was described in more detail in [LSLT]. It was presented, with some supporting empirical evidence, in a Linguistic Institute lecture in Chicago in 1954, and again in the discussions of the IVth Texas conference, 1959. It is often proposed that categories be defined in terms of particular sets of inflectional morphemes, but unless some general method is given for selecting the relevant sets (none has ever been proposed, to my knowledge), such definitions are completely ad hoc, and simply avoid the problem of discovering the basis for categorization.

52

Noam CHOMSKY

English on May 9-12, 1958. Ed. by A. A. Hill, 124-158. Texas, 1962. (Reprinted in Structure of Language. Ed. by J. A. Fodor and J. J. Katz. New York: Prentice-Hall, 1964; reprinted as “Une Conception Transformationelle de la Syntaxe”. Language 4 (December 4, 1966): 39-80; Reprinted in Classics in Linguistics. Ed. by D. E. Hayden, E. P. Alworth and G. Tate, 337-371. New York: Philosophical Library, 1967.) 1959. “The Transformational Basis of Syntax”. Paper presented at IVth University of Texas Symposium on Syntax, June 1959, Unpublished. 1961. “On the notion ‘rule of grammar’”. In Proceedings of Symposia in Applied Mathematics 12: Structure of Language and its Mathematical Aspects. Ed. by R. Jakobson, 6-24. Providence, RI: American Mathematical Society. (Reprinted in The Structure of Language. Ed. by J. A. Fodor and J. J. Katz. New York: Prentice-Hall, 1964.) 1962. “Explanatory Models in Linguistics”. In Logic, Methodology and Philosophy of Science: Proceedings of the 1960 International Congress. Ed. by E. Nagel, P. Suppes, and A. Tarski, 528-550. Stanford: Stanford University Press. (Reprinted in Portuguese. In Fundamentos Metodológicos Da Lingüística 1, Concepções Gerias da Teoria Lingüística, M. Dascal (org.), Editora Parma, 61-93. São Baulo, Brasil: 1978.) and M. Halle. 1968. The Sound Pattern of English. New York: Holt, Rinehart and Winston. , M. Halle and F. Lukoff. 1956. “On Accent and Juncture in English”. In For Roman Jakobson. Ed. by M. Halle, H. Lunt and H. MacLean, 65-80. The Hague: Mouton. Empson, W. 1957. Seven Types of Ambiguity. New York: Meridian. Halle, M. 1960. Sound Pattern of Russian. The Hague: Mouton. Harris, Z. S. 1951. Methods in Structural Linguistics. Chicago: Chicago University Press. Hill, A. A. 1961. “Grammaticality”. Word 17. 1-10. Hockett, C. F. 1948. “A Note on ‘Structure’”. International Journal of American Linguistics 14. 269-271. (Reprinted in Readings in Linguistics. Ed. by M. Joos, 279-280. Washington: American Council of Learned Societies, 1957.) 1954. “Two Models of Grammatical Description”. Word 10. 210-234. (Reprinted in Readings in Linguistics. Ed. by M. Joos, 386-399. Washington: American Council of Learned Societies, 1957.) Jakobson, R. 1936. “Beitrag zur allgemeinen Kasuslehre”. TCLP (Travaux du Cercle Linguistique de Prague) VI, 240-288. (Reprinted in Selected

Some Methodological Remarks on Generative Grammar

53

Writings II, 23-71. Ed. by R. Jakobson. 1971. The Hague: Mouton. Translated 1984 as: Contribution to the General Theory of Case: general meangs of the Russian cases. In Roman Jakobson. Russian and Slavic Grammar: Studies 1931-1981. Ed. by L.R. Waugh & M. Halle, 59-103. Berlin: Mouton de Gruyter) 1959. “Boas’ View of Grammatical Meaning”. In The Anthropology of Franz Boas: essays on the centennial of his birth. Ed. by W. Goldschmidt; sponsoring editors, A.L. Kroeber, R. H. Lowie. San Francisco: The American Anthropological Association. Lees, R. B. 1960. The Grammar of English Nominalizations. Supplement to International Journal of American Linguistics 26. Baltimore. Lenneberg, E. 1958. “Language, Purposive Behavior and Evolution”. Mimeographed. (Published as “Language, Evolution, and Purposive Behavior”. In Culture in History Essays in Honor of Paul Radin. Ed. by S. Diamond, 869-893. New York: Columbia University Press, 1960.) Maclay, H. and M. D. Sleator. 1960. “Responses to Language: Judgements of Grammaticalness”. International Journal of American Linguistics 26, 275-282. Ziff, P. 1960a. Semantic Analysis. Ithaca, New York: Cornell University Press. 1960b. “On Understanding ‘Understanding Utterances’ ”. Mimeographed. Reprinted in Readings in Philosophy of Language. Ed. by J. J. Katz and J. A. Fodor. New York: Prentice-Hall, 1963.

54

Noam CHOMSKY

Knowledge of Language: Its Elements and Origins* BY N. CHOMSKY, CORRESPONDING F.B.A. Department of Linguistics and Philosophy, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, U.S.A. My approach to the study of language is based on the assumption that knowledge of language can be properly characterized by means of a generative grammar, i.e. a system of rules and principles that assigns structural descriptions to linguistic expressions. On this view, the basic concepts are those of ‘grammar’ and ‘knowledge of grammar’. The concepts of ‘language’ and ‘knowledge of language’ are derivative: they involve a higher level of abstraction from psychological mechanisms and raise additional (though not necessarily important) problems. Of central concern, from this point of view, will be to determine the biological endowment that makes it possible for a grammar of the required sort to develop in human beings provided that they are exposed to some appropriate body of experience. This biological endowment may be regarded as a function that maps a body of experience into a particular grammar. The function itself is commonly referred to as universal grammar (u.g.) and can be expressed, in part, as a system of principles that determine the class of accessible particular grammars and their properties. Recent work suggests that u.g. consists, on the one hand, of a theory of so-called core grammar and, on the other, of a theory of permissible extensions and modifications of core grammar. Given the intricate internal structure of u.g., it can account for the superficially highly diverse grammars and languages that do in fact exist. Thus, what appear to be quite different systems of knowledge may arise from relatively little experience. A number of subsystems of u.g. have now been explored, each with its distinctive properties and possibilities of variation. Some current proposals concerning these systems are sketched, and some consequences considered with regard to the nature and acquisition of cognitive systems (including systems of knowledge) more generally. *

Reprinted by permission from Philosophical Transactions of the Royal Society of London. B 295. 223-234 (1981). The Center for Research on Language and Culture, Senshu University would like to express our thanks to Professor Noam Chomsky and the Royal Sociey of London.

56

Noam CHOMSKY

GRAMMAR AND UNIVERSAL GRAMMAR Three questions arise at once when we consider the notion ‘knowledge of language’: (I)What is the nature of this knowledge? (II) How is it acquired? (III) How is it put to use? The first question has a certain conceptual priority; we can hope to gain some insight into acquisition or use of knowledge only to the extent that we have some understanding of what it is that is acquired or used. A standard, and I think basically correct, approach to the first question is to assume that a person who knows a language has internalized a grammar, a system of rules and principles that assigns structural descriptions to linguistic expressions. If so, then the central notions to be developed are the correlative notions ‘grammar’ and ‘structural description’, and the central cognitive relation is the one that holds between the person and the grammar. I shall refer to this relation as ‘(tacit) knowledge’, thus saying that a person knows his grammar and knows the rules and principles of his grammar, which of course does not imply that he has propositional knowledge that these are the rules and principles of his grammar. The linguist may develop such propositional knowledge, but that is quite a different matter. A person who knows English has attained a certain mental state, different from that of someone who knows Japanese. Abstracting from possible individual differences, there is some innate mental state common to the species that provides the basis for acquisition of knowledge of grammar, a characteristic that distinguishes humans from birds or apes. One proposal, which I think is basically correct, is that this innate endowment consists of a system of principles, each with certain possibilities of parametric variation, and that acquisition of knowledge of grammar with all that it entails is, in part, a matter of setting these parameters one way or another on the basis of presented experience. Let us say that this process yields a ‘core grammar’. Then an actual grammar, representing full knowledge of language, consists of a core grammar extended to a periphery that incorporates more idiosyncratic (marked) elements, also constrained by innate principles but in somewhat different ways. There are familiar idealizations implicit in this rough account, but they are, I think, quite legitimate ones. One component of the human mind–brain, then, is a genetically determined initial configuration, which we may call ‘the initial state of the language faculty’. It is characterized by a theory of principles and parameters and a theory of markedness, which permits the extension of core grammar to a full grammar. I shall refer to this complex as ‘a theory of universal grammar (u.g.)’, and shall say that u.g. is a component of the initial state. Under the triggering and shaping effect of experience, the initial state is transformed to a more or less steady state incorporating the attained grammar.

Knowledge of Language: Its Elements and Origins

57

My use of mentalistic terminology involves nothing beyond the assumption that u.g. and grammar are elements of the initial and steady state, respectively, characterized at some appropriate level of abstraction; in short, that this characterization expresses conditions satisfied in some way by physical mechanisms of the brain. In this usage, we would, for example, refer to the proposals of David Marr and his colleagues concerning visual mechanisms as ‘mentalistic’, for example the rigidity principle that permits determination of structure from motion, or the theory of visual representation in terms of axes of elongation, etc. (Marr & Nishihara 1978; Ullman 1979). Another familiar idealization is the assumption that the transition from initial to steady state can be regarded as instantaneous, i.e. that the actual interstate transitions have at most a negligible effect on the state attained. This is surely questionable, but so far as is known it yields an extremely good first approximation, and I shall adopt it here. We may then think of u.g. as a function mapping presented experience to steady state attained, through the means of fixing parameters of u.g. and adding a marked periphery. U.G. AS AN EXPLANATORY THEORY Approaching our initial questions in this way, the second question is in part answered by the theory of u.g. Furthermore, this account of growth of knowledge can be interpreted as a model of explanation. There is a vast array of facts about knowledge of language that we would like to explain. Consider, for example, the sentences (1), (2). (1) (2)

everyone wants John to like him everyone wants him to be liked by John

The sentence (1) can have the interpretation expressed in (3), whereas (2) cannot have the corresponding interpretation expressed in (4). (3) (4)

for every person x, x wants John to like x for every person x, x wants x to be liked by John

Thus, even though ‘John likes him’ and ‘he is liked by John’ are virtual synonyms, insertion of the corresponding infinitives as the clausal complement of the verb want yields expressions that differ in their range of meaning. The language-user’s knowledge of these facts is propositional knowledge that so-and-so. Such examples of propositional knowledge extend beyond the bounds of well-formed English sentences. It is, for example, a property of my dialect of English that the verb want does not take a that-clause complement, e.g.

58

Noam CHOMSKY

the sentence (5) is not well-formed. (5)

everyone wants that John like him

Nevertheless, I know that this expression, were it well-formed, could have the meaning expressed in (6), just as (7) – in contrast to (2) – could have the meaning expressed in (8). (6) (7) (8)

for every person x, x wants that John like x everyone wants that he be liked by John for every person x, x wants that x be liked by John

In this case, interchange of synonyms (‘John like him’, ‘he be liked by John’) does preserve meaning. How can we explain these facts? At one level, we can explain them by postulating a grammar that yields these consequences, taking it to be a component of the steady state attained. At a deeper level, we can explain these facts by postulating a u.g. as an element of the initial state with the following property: given experience sufficient for language acquisition, this u.g. determines a grammar that in turn yields these consequences. A theory of u.g. with this property is a genuine explanatory theory. In the case illustrated, explanatory adequacy in this sense can be partly achieved. I shall not formulate the relevant principles of u.g. here; let us refer to them as ‘the binding theory’. There has been a good deal of work in the past 10 years on binding theory, and principles have been formulated that are quite simple and natural and that have explanatory force over a fairly broad empirical domain (see Chomsky 1981 a, b). The binding theory entails that certain configurations are ‘opaque’ in the sense that an anaphor such as each other must be bound within them. An element is bound if it has an antecedent that c-commands it, where in the simplest case a category A c-commands B if A is an immediate constituent of a category that contains B but A does not contain B; then B is in the c-command domain of A. The opaque configurations include the subject of a finite clause and the minimal c-command domain of the subject of any clause. A pronominal such as he must be free exactly where an anaphor must be bound. By using indices in the obvious way to express the binding relation, typical examples of the opacity conditions are the following. (9) Johni expects [Billj to like himk] (10) Johni expects [himj to win] (11) Johni expects [that hej will win]

Knowledge of Language: Its Elements and Origins

59

The brackets bound embedded clauses; finite in (11) and infinitival in (9), (10). In (9), him is in the minimal c-command domain of the subject Bill, and since it must be free in this domain, k ≠ j. But k may = i. In (10), j ≠ i or him will be bound in the minimal c-command domain of the subject John. In (11), j may = i since in any event he is free as subject of a finite clause. Anaphors satisfy the same conditions, with ‘free’ replaced by ‘bound’. Coindexing may be interpreted here as intended coreference. Thus in (9) we may understand him as referring to John but not to Bill. It is easy to show that the examples discussed earlier fall under the opacity conditions, where we interpret a pronoun coindexed with a variable as an occurrence of this variable (see Higginbotham 1980). Another consequence of the binding theory is that names cannot be bound by pronouns. Thus we can have the binding represented in (12) but not (13). (12) the woman hei married expected [us to like Johni] (13) hei expected [us to like Johni]

In (12), he does not c-command John, so John is free, though he is its antecedent. But in (13) he does c-command John so that the coindexing violates the binding theory. We can account for the propositional knowledge that these sentences have the interpretations indicated on the assumption that the binding theory is an innate property (possibly parametrized) and that experience provides sufficient evidence to determine that he is a pronoun. Many other cases are similar. Consideration of the binding theory provides extensive evidence concerning the form of mental representations of linguistic expressions. Consider, for example, sentence (14). (14) who did Johni expect to like himk

Here him may be understood to refer to John, so that k may = i. This fact is compatible with the binding theory only if him is in the minimal c-command domain of the subject of an embedded clausal complement of expect. No such subject is overtly expressed, but we can account for our knowledge of the facts on the assumption that such a subject is mentally represented; that is, at the level of syntactic representation at which the binding theory applies, the representation of (14) must be (15). (15) whoj did Johni expect [xj to like himk]

60

Noam CHOMSKY

Here x is the variable bound by the quasi-quantifier who. Since this mentally represented variable is the subject of the embedded clause, the binding relation between John and him satisfies the binding theory if k = i. The representation (15) is quite natural; with k = i, we may assume it to correspond to the logical form (16). (16) for which person x, Johni expected [x to like himi]

The clause now has the form of (9), with k = i. Note that we cannot set k = j in (15), thus replacing him in (16) by x, or the binding theory will be violated, just as him cannot refer to John in ‘John likes him’, ‘I expect John to like him’, etc. The example (14) illustrates the fact that a subject may be a mentally represented empty element, lacking phonetic content. There is, incidentally, evidence that such empty elements, though lacking phonetic content, may affect the phonetic form of utterances (see Chomsky 1980, 1981 b). Thus we have independent phonetic evidence that such empty elements are mentally represented; the phonetic rules actually ‘see’ them. Other examples show that a direct object may also be a mentally represented empty element. Consider the sentence (17). (17) I wonder whom he expected us to like

We cannot understand (17) as having the meaning expressed in (18). (18) I wonder for which person x, x expected us to like x

In contrast, (18) can be the logical form corresponding to (19). (19) I wonder who expected us to like him

These facts follow if we assume the syntactic representations of (17) and (19) to be (20) and (21), respectively. (20) I wonder [for which person x, he expected [us to like x]] (21) I wonder [for which person x, x expected [us to like him]]

The embedded complement of wonder in (20) has essentially the form of (13), so that the variable x cannot be bound by he just as John cannot be bound by he in (13), accounting for the fact that (18) is not the interpretation of (17). No binding principle prevents (18) from being the interpretation of

Knowledge of Language: Its Elements and Origins

61

(19) or (21), however. In this case, the embedded complement of wonder has the form of (9). Note that these examples provide evidence that variables behave in the manner of names, with regard to the binding theory. These examples illustrate the fact that either a subject or an object may be a mentally represented empty element. More complex cases show that both may be empty elements, as we would expect. Consider the sentences (22), (23). (22) John is too stubborn to talk to Bill (23) John is too stubborn to talk to

We understand these sentences as (24), (25), respectively. (24) Johni is so stubborn that hei will not talk to Bill (25) Johni is so stubborn that one cannot talk to himi

What is particularly interesting about these examples is that the understood subject of the transitive verb talk to is interpreted differently in the two cases: it is taken to refer to John in (22) but to some arbitrary person in (23). Yet these sentences differ only in that the object of talk to is overtly expressed in (22), while it is missing in (23). These curious facts also follow from the binding theory, if we assume that the ‘understood subject’ and ‘understood object’ are in fact mentally represented, as in (26) and (27), corresponding respectively to (22) and (23). (26) John is too stubborn [PRO to talk to Bill] (27) Johni is too stubborn [PROj to talk to Xk]

What I have represented as ‘PRO’ is to be understood as an abstract pronominal, i.e. a pronoun lacking phonetic content. The binding theory permits PRO to be bound by John in both (26) and (27), and another subtheory, the theory of control, requires this binding in (26). Turning to (27), the binding theory prevents binding of X by PRO: thus k ≠ j, exactly as in (9). Another consequence of the binding theory, which there is no space to explain here, is that X in (27) cannot be free PRO; since k ≠ j, it must be that k = i since X cannot be free. Therefore j ≠ i. Since there is no other antecedent for PRO, it must be interpreted as referring to some arbitrary person, just as in (28), mentally represented as in (29). (28) it is unclear how to solve the problem (to help oneself) (29) it is unclear [how PRO to solve the problem (to help oneself)]

62

Noam CHOMSKY

Note that the reflexive option again entails that an antecedent must be present, namely, the empty pronominal PRO. The careful reader should now be asking why PRO in (26) is permitted to be bound by John, even though it is a pronominal in the minimal c-command domain of this subject. This fact too follows from the binding theory, on the quite natural assumption that PRO, as distinct from an overt pronoun, is a pronominal anaphor. For details, and discussion of assumptions that are tacit here, see Chomsky (1981 b). Further examples provide additional support for this analysis. Consider (30) and (31). (30) John is too clever to expect us to catch Bill (31) John is too clever to expect us to catch

In (30), we understand the subject of expect to be John; in (31), to be some arbitrary person. These facts again follow if we take the mental representations to be (32) and (33), respectively. (32) John is too clever [PRO to expect [us to catch Bill]] (33) John is too clever [PRO to expect [us to catch X]]

In (32), PRO can be coindexed with John. Turning to (33), as noted earlier, the binding theory requires that X be distinct from free PRO. Example (33), in fact, provides direct evidence that X ≠ PRO. For if X were PRO, then there would be no reason why it should not be coindexed with the PRO subject of expect, yielding an instance of the same configuration as (9), with k = i. But this interpretation is impossible, so X cannot be PRO. The only alternative is that X is a variable; for independent evidence in support of this conclusion, see Chomsky (1977, 1980, 1981 a, b). But this variable lacks an appropriate binding operator, so it must be assigned an antecedent. By the binding theory, its antecedent cannot be us or the PRO subject of expect. Therefore it must be John. Exactly as in case (27), it follows that PRO cannot be coindexed with John, and must therefore be arbitrary in interpretation. The very simple principles of the binding theory that account for the behaviour of overt elements as in (9) and (11) also explain the properties of these considerably more complex examples, on the assumption that empty elements appear in mental representation where they are ‘understood’. This fact provides evidence that such empty elements do appear in mental representations. As we have seen, these empty elements fall into two distinct types: variables, which behave in the manner of names, and PRO, which is a pronominal anaphor. There is in fact a third type, so-called ‘trace’, which is a

Knowledge of Language: Its Elements and Origins

63

pure anaphor (see Chomsky 1981 b). There is extensive further evidence that empty elements appear in syntactic representations exactly where they are ‘understood’ in the intuitive sense, and that these elements have quite specific properties and fall under three distinct types: anaphor, pronominal anaphor, and name-like variable. The evidence is similar to what I have just sketched: on these assumptions, a wide variety of empirical facts can be explained by the binding theory, which is independently motivated for overt elements, and other subtheories of u.g. Furthermore, the positions in which these various types of empty elements may appear are determined by the interaction of these subtheories. Finally, slight modifications in the parameters of the subtheories yields a substantially different array of data, which is attested in other languages. For example, a slight change in the parameter associated with one morphological rule yields the quite different properties of the so-called ‘pro-drop languages’ such as Italian and Spanish (see Chomsky (1981 b) and, for a slightly different approach, Rizzi (1980)). These are the kinds of results that we hope to attain in the study of u.g. THE PROJECTION PRINCIPLE Examples of the sort just sketched suggest a very strong and quite natural principle, which I shall call ‘the projection principle’, as a component of u.g. The principle states that syntactic structure is projected from lexical properties in the sense that the argument structure of lexical items is represented explicitly at each syntactic level. Thus, the verbs hit, help and talk to, as a lexical property, take an object to which they assign a certain semantic role and a subject to which they assign a different semantic role. By the projection principle, at every syntactic level there must appear a subject and an object in the appropriate structural configuration. These appear overtly in the sentence ‘John hits Bill’ and they appear as empty elements in (26) and (27). I cannot give a precise formulation here (see Chomsky 1981 b), but this is the intuitive idea. The projection principle has a wide range of consequences. Assuming it to be a component of u.g., i.e. of innate endowment, a child who is ‘learning English’ somehow discovers the lexical properties of the verbs hit, help and talk to and then knows without further evidence that the examples just discussed are to be understood as indicated. The projection principle thus vastly facilitates the task of what is misleadingly called ‘language learning’ – a better term would be ‘growth of grammar’. Equivalently, it has substantial explanatory force, as in the examples indicated. One virtual consequence of the projection principle is what has been called ‘trace theory’, which itself has substantial explanatory scope. Turning briefly to question (III), – how is knowledge put to use? – the

64

Noam CHOMSKY

projection principle also has suggestive consequences. Equipped with this principle, and knowing lexical entries, a person who hears examples of the sort illustrated can at once construct the abstract representations that underlie them, yielding the required interpretations, given the binding theory and other subtheories of u.g. Internal to the theory of grammar, the projection principle has the consequence that the complex apparatus of phrase structure rules can be eliminated, apart from language-specific parameters such as order of major categories. At the same time, it seems that the transformational component can be restricted to a single rule of core grammar – namely, the rule ‘move any category anywhere’ – with a few simple parameters. The grammar can therefore be quite ‘small’, containing few and simple rules, a property that itself has significant consequences for parsing and acquisition. The single transformational rule of core grammar, which has broad scope, can be characterized abstractly in terms of several distinctive properties: it relates an empty element to an antecedent in a position assigned no semantic role; the empty element must be ‘governed’ in a sense that generalizes the traditional notion of government; and a strong locality principle holds of the antecedent–empty element relation. The empty pronominal PRO, in contrast, has quite different properties, which follow from the binding theory. Let us explore some further consequences of the projection principle. Consider the sentence (34). (34) the men asked me [how they could help each other]

Clearly, the bracketed phrase is clausal; it has the semantic properties of a clause and is the domain of wh-movement, which applies to clauses only. The verb ask takes an object and a clausal complement, so the projection principle is satisfied by the bracketing indicated. Consider now the sentence (35). (35) the men asked me [how to help each other]

Clearly, (35) is the counterpart to (34), with an infinitival rather than a finite clausal complement to ask, these being the two possible forms of a clause. Again, the semantic content is propositional and the complement is the domain of wh-movement. By the projection principle, the structure must be as in (36). (36) the men asked me [how PRO to help each other]

Knowledge of Language: Its Elements and Origins

65

The binding theory is satisfied with PRO as antecedent of each other. The surface form of (36) is derived by trivial rules, and an appropriate representation for logical form can also be derived in a straightforward way (see Higginbotham 1981). Examining the distribution of the empty pronominal PRO, we discover some interesting properties. First, its distribution is virtually complementary to that of overt anaphors; there is (virtually) no context in which both can appear. This property is illustrated by interchange of PRO and each other in (36), which yields the non-sentence (37). (37) the men asked me how each other to help PRO

This expression does not mean that each of the men asked me how the other is to help him, as the interpretive principles for the reciprocal imply, though this is a perfectly sensible interpretation of some sentence. Secondly, we find that at a somewhat more abstract level of structure, the empty pronominal shares a good part of the distribution of anaphors, which is not surprising, given that it functions as a pronominal anaphor as already noted. For example, both can occur as subjects of infinitival clauses or of noun phrases, as in (36), (38) and (39), but not as subjects of tensed clauses as in (40). (38) the men prefer [PRO helping each other] (39) (i) the men would prefer [for each other to win] (ii) the men would prefer [each other’s books] (40) (i) the men preferred [that PRO win the race] (ii) the men preferred [that each other win the race]

Note that although PRO and each other share the distribution illustrated at an abstract level of structure, these examples still illustrate the complementary distribution in actual sentences; thus PRO and each other cannot be interchanged in the well-formed examples. There is, in fact, a restricted class of contexts in which both PRO and each other can appear, but it is reasonable to assume that this fact reflects an idiosyncratic marked property of English, and that at a more abstract level corresponding to core grammar the complementary distribution is not virtual but complete. As usual, the unanalysed phenomena that we observe tell us little in themselves. Assuming this, we have to explain why PRO and each other are in complementary distribution in the class of actual sentential contexts, while they share fundamental distributional properties at a more abstract level. A theory of u.g. that does not yield this conclusion is plainly missing

66

Noam CHOMSKY

something important. The natural conclusion is that these facts reflect the interaction of two subsystems of u.g., one of which determines the general distribution of anaphors (both each other and PRO), and the second of which distinguishes overt from empty elements. In fact, the binding theory yields the first of these results, as is fairly evident from the examples already discussed; and the theory of abstract case (in the sense of J.-R. Vergnaud) yields the second (see Chomsky 1981 a, b). This is a typical example of the modularity of structure of the language faculty, an internal counterpart to the modularity of the total system of cognitive structure. The latter assumption is, in my opinion, more controversial than it should be; it appears to be increasingly well supported as we learn more about particular cognitive systems, as well as quite natural. There is little reason to expect to find anything analogous to the rigidity principle in the theory of language, or anything analogous to the binding theory or the projection principle in the theory of vision. And despite much talk of generalized learning mechanisms, general principles of organism–environment interaction, and the like, I am unaware of any proposal substantive enough to bear investigation. On the basis of what is now known or plausibly surmised, there seems to be good reason to adopt the position that John Marshall has called ‘the new organology’, and to search, as he has suggested, for the neural correlates to this structural organization of mind (see Marshall 1980). As the examples cited suggest, empty elements have provided an important probe into the properties of u.g. They are interesting from another point of view as well. Evidently, the grammar that is attained in the steady state results from the interaction of experience and initial state. It is difficult to believe that the intricate properties of empty elements are learned on the basis of direct evidence; in fact, one has no evidence bearing directly on their nature and properties. Therefore, it is reasonable to suppose that these properties reflect intrinsic properties of the mind–brain. If our interest is in the nature of the human mind–brain, rather than in phenomena of the external environment, then these elements have a particular fascination. THE GROWTH OF KNOWLEDGE Let us assume now that a successful theory of u.g. can be constructed along the lines indicated and think about the possible consequences. Note first that in the examples cited, and innumerable others like them, there is strong reason to believe that what we know, we know without relevant evidence. It is quite unreasonable to suppose that everyone who knows these facts has been given a sufficient basis of evidence to ground this knowledge. In many cases, it is difficult even to imagine what evidence might suffice for

Knowledge of Language: Its Elements and Origins

67

this purpose. There is no reason to believe that such propositional knowledge – for example, our knowledge that (31) means (33) – can be attained from evidence available to each speaker of the language by some generally valid procedure of induction or confirmation. Attention to the facts quickly convinces us that such an assumption is about as plausible as the belief that there exists some comparable general account for the fact that the human embryo grows a mammalian rather than an insect eye, or that children undergo puberty at a certain age, or that they employ the rigidity principle to determine structure from motion. In all such cases, it seems highly likely that innate properties account for the resulting state, a state of knowledge in the case that we are considering, a state of knowledge that yields specific examples of propositional knowledge. We can do better than mere speculation; there is, in fact, a reasonably successful empirical theory that gives substance to this proposal, namely the theory that postulates the projection principle, the binding theory, the theories of government, case, control, locality and others, along with certain parameters, as a component of the initial state. In contrast, no proposal of even minimal plausibility exists to account for facts of the sort illustrated in terms of some sort of ‘learning’ or ‘confirmation’, very much as with puberty or the rigidity principle. Certainly what I have just suggested could be correct; indeed, there is fairly good reason to suppose that something of the sort is correct. Since it could be correct, it cannot be that propositional knowledge must have adequate grounds or justification or warrant in anything like the sense assumed in modern epistemology. This cannot, in short, be a conceptual requirement associated with the concept ‘knowing that’. The reason is that we have clear examples of such knowledge that may violate (and apparently do violate) this requirement. Therefore, it is a contingent question of fact whether specific examples of knowing-that in other domains are grounded or rather simply ‘grow in the mind’, as appears to be so with respect to much of linguistic knowledge. Suppose, for example, that we observe a moving object passing behind a barrier and we know that it will emerge at such and such a point. Surely there are such examples of propositional knowledge. Is this justified knowledge, or is it a consequence of some innate system of principles perhaps triggered by experience? Or suppose that we observe a plane figure perpendicular to the line of sight and when this figure rotates until it is parallel to the line of sight, we see it as a rotating plane figure. That is, we know that it is a rigid rotating plane figure and not (as it might have been) a plane figure shrinking to a line. Is this a case of justified knowledge, or is it a consequence of something like the rigidity principle? These are questions of fact, and in these and many similar cases it seems likely that the fact is that

68

Noam CHOMSKY

the knowledge is not grounded. If some organism had different built-in principles, distinct from the binding theory or the rigidity principle and so on, it would simply have different beliefs and even a different domain of knowledge, including propositional knowledge, on the same evidence. It seems to me reasonable to speculate that a substantial part of our knowledge about language, about the behaviour of objects in three-dimensional space, about other people, and the like, is knowledge that grows in the mind on the basis of a system of innate principles, rather than knowledge that is grounded in experience; though the latter category too exists no doubt, in domains where built-in structure is inadequate: are all ravens black: are there free quarks? etc. At least, these are factual questions, not resolvable on a priori grounds. If so, then our general approach to questions of the nature of knowledge and belief requires considerable rethinking, in a fundamental way. LANGUAGE AND GRAMMAR; RULES AND PRINCIPLES I shall conclude with a few observations on the ways in which ideas about language have evolved within the past several decades, at least in the work I have been considering here. I think that there have been two significant shifts of focus with regard to the way in which the major problems are conceived. The first is the shift of focus from language to grammar, which took place about 25 years ago. For structural linguistics and its predecessors, the object of inquiry was language, and correspondingly, the closest analogue to u.g. was the theory of general properties of many or all languages. The approach that I have been discussing took a radically different stance. The focus of inquiry is grammar; language is a derivative and possibly uninteresting notion. The shift of focus was sometimes obscured in introductory expositions or in mathematical linguistics, but it was clear in the earliest work on generative grammar. I think that this shift was quite proper. Contrary to what is commonly believed, the notion ‘language’, if it is even coherent, is at a much higher level of abstraction from actual mechanisms than grammar. Grammars exist in the world, as components of steady states attained. As for languages, one may perhaps think of them as determined by grammars, or in some other way, but in any event they are clearly at a further remove from real mechanisms of the brain than the grammars represented in these mechanisms. Correspondingly, the theory of u.g. is not the study of general properties of language, but rather u.g. is a postulated component of genetic endowment, on a par with the properties that determine that the embryo will grow arms rather than wings, use the rigidity principle, undergo puberty, etc. – all, of course, under appropriate external conditions. Once this change of focus is adopted, this

Knowledge of Language: Its Elements and Origins

69

part of linguistics becomes part of psychology, and ultimately biology. It is a subdomain delimited by the uninteresting characteristic that its practitioners, for the moment, happen to rely primarily on certain kinds of evidence, such as those illustrated earlier, because such evidence appears to be most useful in advancing their inquiry into properties of the initial and steady state of the language faculty. There is every reason to hope that this artificially delimited discipline will disappear as other kinds of evidence become available that bear on the questions with which it has been concerned. The second shift of focus is more subtle, and its significance is just beginning to become clear. Early theories of generative grammar permitted a wide range of possible grammars in the interest of attaining descriptive adequacy, and the associated theory of u.g. was correspondingly limited, though far from vacuous, in explanatory power. Much of the work of the past 20 years has been devoted to restricting the class of available grammars so that explanatory adequacy is enhanced, as it becomes possible to determine why the steady state takes the specific form that it does. The general approach has been to limit the variety of possible rules while formulating general principles, with a few parameters, that restrict the ways in which rules may apply. As this work has progressed, the focus of attention has gradually shifted from rule systems to systems of principles such as the binding theory and the projection principle. In fact, current theories of transformational generative grammar are so restrictive that they permit only a finite number of grammars in principle, apart from the lexicon. Much of the debate over the character of rule systems proves to be near vacuous, e.g. the question of whether a certain rule is a transformational rule or a rule of interpretation with exactly the specific properties of transformational rules. I say ‘near vacuous’ because there may be empirical differences, but if they exist, and they may, they are rather subtle. In contrast, questions concerning the subsystems of principles often have substantial empirical content. One consequence of this second shift of focus is that most of the results of mathematical linguistics, which in any event have been seriously misinterpreted, become empirically virtually or completely empty, since they deal with properties of infinite classes of grammars. The same is true of the mathematical theory of learnability in so far as its results depend on the presumed infinity of the class of grammars. The question of generative capacity, for example, has little meaning for finite sets of grammars. The specific detailed structure of the elements of these sets now has an overwhelming effect on any results that might be obtained in studies of generative capacity or learnability. Similar remarks hold for the theory of parsing, though for somewhat different reasons, as has been pointed out by Berwick & Weinberg (1981). The mathematical results concern parsability in

70

Noam CHOMSKY

the limit, as sentences grow longer. But any empirically meaningful results will deal with parsability of ‘short sentences’, whether we are considering parsing of functional explanation for language evolution. Thus, the standard result for parsing of context-free languages shows that parsing time is proportional to the square of the size of the grammar and the cube of the length of the input sentence. The first factor, however large, is fixed, and disappears in the limit as sentences become longer. But for ‘short sentences’ of the sort that are actually used, the first factor may be overwhelming and the second insignificant. The real empirical content of existing results, then, may well be that grammars are preferred if they are not too complex in their rule structure. If parsability is a factor in language evolution, we would expect it to prefer ‘short grammars’ – such as transformational grammars based on the projection principle or the binding theory – and to care very little about parsability in the limit or even parsability for ‘long sentences’ of a sort rarely encountered in ordinary life. Such principles as the projection principle make strong empirical predictions, and have a large effect on simplifying grammar and probably on the proper formulation of questions of parsing and learnability. It also becomes very important to determine in just what respects parametric variation is permitted. In a theory of u.g. with a fairly rich structure, a small change in a single parameter may lead to substantial difference in the resulting ‘language’, as effects proliferate through the system. And a few such changes may yield languages that look very different from one another, though they are basically cast in the same mould. For similar reasons, species may seem highly diverse as a result of modifications of regulatory circuits and the like in a common biochemical system. These are the kinds of result that we hope to obtain in the study of language, or better, the study of grammar, the real object. Qualitatively speaking, languages appear to be highly diverse, yet there is strong reason to believe that the initial state is highly restrictive, so that an intricate system of knowledge, with quite subtle and specific properties, is acquired on fragmentary evidence. A theory of principles and parameters with simple subtheories and a fairly rich structure of interactions appears to be what is required to provide the actual observed results. Such theories are now, for the first time, becoming available. This is a new development, and one that I believe to be very exciting. REFERENCES Berwick, R. & Weinberg, A. 1981. Parsing efficiency, computational complexity, and the evaluation of grammatical theories. M.I.T. mimeograph. Chomsky, N. 1977. On wh-movement. In Formal syntax ed. P. Culicover, T. Wasow & A. Akmajian), pp. 71-132. London: Academic Press.

Knowledge of Language: Its Elements and Origins

71

1980. Rules and representations. New York: Columbia University Press. 1981a. Principles and parameters in syntactic theory. In Explanation in linguistics (ed. N. Hornstein & D. Lightfoot) London: Longmans. (1981) 1981b. Lectures on government and binding: the Pisa Lectures. Amsterdam : Foris. Higginbotham, J. 1980. Pronouns and bound variables. Ling. Inquiry 11, 679-708. 1981. Reciprocal interpretation. J. Ling. Res. 1, 97-117. Marr, D. & Nishihara, K. 1978. Visual information processing: artificial intelligence and the sensorium of sight.Technol. Rev. 81, 2-23. Marshall, J. 1980. The new organology. Behav. Brain Sci. 3, 23-25. Rizzi, L, 1980. Negation, wh-movement and the null subject parameter. Università della Calabria - M.I.T. mimeograph. Ullman, S. 1979. The interpretation of visual motion. Cambridge: M.I.T. Press.

72

Noam CHOMSKY

2. Current Issues in Language Descriptions

Germanic Passive Constructions John Ole ASKEDAL 1. INTRODUCTION In what follows I shall present a proposal for comparative classification of passive constructions in modern standard Germanic languages with regard to the areal distribution of their defining features in the individual languages and general typological characteristics of the languages in question. The languages dealt with are the following: the Insular Scandinavian languages Icelandic and Faroese; the modern Mainland Scandinavian languages Danish, Swedish and Norwegian, including both standard varieties Riksmål/Bokmål and Nynorsk; the Continental Germanic languages Modern West Frisian, Dutch and German; and English (cf. Harbert 2007: 13–20).1 1.1 A morphosyntactic definition of passive constructions Traditionally, the active is considered the primary or unmarked voice category, and the passive the secondary or marked voice category. Accordingly, the passive is commonly defined in terms of the ‘changes’ or systematic differences that distinguish it from the active. 2 The defining morphosyntactic features needed to describe Germanic passives are listed in (1): (1)

1

2 3

Morphosyntactic characteristics of passive constructions in comparison with the active a. In the argument domain:3 i Subject demotion: SuACT → Agent-PPPASS ii Non-subject promotion: Non-subject (DO > IO > NP in PO > NP in Adv-PP)ACT → SuPASS iii Agent-PP deletion: Agent-PPPASS → Ø

Dialects of these standard languages, the fairly recent standard language Luxembourgish and the contemporary non-European Germanic diaspora consisting of Pennsylvania German, Yiddish, and Afrikaans are not taken into consideration for reasons of space. This is often taken to mean that the passive is in some sense ‘derived’ from the active but, as is well known, this is not a theoretically necessary conclusion. I use the terminology of one-time “relational grammar” (Johnson 1977). This implies no further theoretical commitment on my part.

76

John Ole ASKEDAL b.

In the predicate domain: passive verb morphology4

Agent-PP deletion can be considered a special case of subject demotion. The category scale in (1a ii) corresponds to the non-subject part of the syntactic “noun phrase accessibility hierarchy” of Keenan & Comrie (1987: 6). The categories in question have different morphosyntactic properties and behave differently with regard to passive formation in the individual languages (cf. below). In particular, modern Germanic languages represent one of the two different types of case systems in (2) (cf. Askedal 2009: 10–11; Harbert 2007: 104–105): (2)

Germanic case systems i ‘NP-case languages’: German, Icelandic, Faroese • in non-pronominal and pronominal NPs: nominative, accusative, dative; genitive (rare in Faroese5) ii ‘Non-NP-case languages’: English, Dutch, Frisian, Mainland Scandinavian (Danish, Swedish, Norwegian) • non-pronominal NPs: no case distinctions in the non-genitive domain; agglutinating s-morpheme as morphosyntactic subordination marker in NPs • pronominal NPs: instances of a non-oblique vs. oblique opposition in all languages; to a varying extent s-forms and other ‘genitive’ forms in subordinate position in NPs

1.2 Construction types: ‘personal’ and ‘impersonal’ passive Subject demotion is a structural characteristic common to all passive constructions but non-subject promotion is not (as indicated by the order in which the two conversions appear in (1a)). Passive constructions showing both subject demotion and non-subject promotion are called ‘personal’ and those showing only subject demotion ‘impersonal’ (cf. Keenan 2007: 345–348). Cf. (3): (3) Personal passive Impersonal passive

Subject demotion + +

Non-subject promotion + –

From this perspective, the less common type – the impersonal passive – is the conceptually simpler construction. The individual Germanic languages 4 5

In morphological descriptions of the passive, passive morphology takes pride of place, cf. for instance Keenan & Dryer (2007: 328, 332–339). Cf. Thráinsson et al. (2004: 248–252, 433–435, 456–457).

Germanic Passive Constructions

77

differ with regard to the extent to which, and the specific conditions under which, the two types of passive construction occur. 1.3 Passive verb morphology In Germanic, passive morphology in the verbal domain always accompanies the conversions, i.e. the theoretically possible type in (4b) is non-existent: (4) a. *b.

Subject demotion + +

Non-subject promotion ± ±

Passive morphology + –

This does not imply that passive morphology is morphosyntactically uniform. In Germanic, passive morphology manifests itself either analytically as an auxiliary or synthetically as a suffix. The term ‘auxiliary’ presupposes some degree of grammaticalization and this prompts the question of the existence of both fully desemanticized and grammaticalized passive auxiliaries and paradigmatically related, perhaps less grammaticalized auxiliary-like verbs that, in addition to their structural function, carry some sort of specific meaning, such as aspectual or modal meaning (cf. section 3.3 and n. 20). 2. THE ARGUMENT DOMAIN: THE CONVERSIONS 2.1 Personal passive: Subject demotion with non-subject promotion Non-subject promotion is the constitutive characteristic of personal passive constructions. The Germanic languages show considerable variation with regard to the availability of different non-subject categories for non-subject promotion. The Mainland Scandinavian –NP-case languages allow for a wide range of non-subject promotion options comprising direct and indirect objects (5) and the NP constituent of prepositional objects (6) as well as of adverbial complements (7):6

6

(5)

a.

→

b.

Regjeringen tildelte henne en høy orden. (NO) government-DEF awarded her a high decoration ‘The government awarded her a high-ranking decoration.’ En høy orden ble tildelt henne (av regjeringen). a high decoration was awarded her (by government-DEF) ‘A high-ranking decoration was awarded her (by the government).’

Cf. Faarlund et al. (1997: 841–845) on Norwegian, Hansen (1967: 51–52) on Danish, and Teleman et al. (1999: 366–370) on Swedish.

78

John Ole ASKEDAL Or:

c.

(6)

a.

→

b.

(7)

a.

→

b.

Hun ble tildelt en høy orden (av regjeringen). she was awarded a high decoration by government-DEF ‘She was awarded a high-ranking decoration (by the government).’ Noen hadde tuklet med PCen. (NO) someone had tampered with PC-DEF ‘Someone had tampered with the PC.’ PCen var blitt tuklet med. PC-DEF was been tampered with _ ‘The PC had been tampered with.’ Noen hadde sovet i sengen ganske nylig. (NO) someone had slept in bed-DEF quite recently ‘Someone slept in the bed fairly recently.’ Sengen var blitt sovet i ganske nylig. bed-DEF was been slept in _ quite recently ‘The bed had been slept in fairly recently.’

In Mainland Scandinavian, a fair number of cases are found, especially in set phrases, where only the the indirect object allows for non-subject promotion. Cf. for instance (8): (8)

a.

→

b.

Not: c. Or:

d.

De nektet henne adgang. they refused her-IO.OBL admission-DO Hun ble nektet adgang. she-SU was refused admission *Henne ble adgang nektet. her-OBL was admission-SU refused. *Henne ble nektet adgang. her-OBL was refused admission-DO

English has basically the same options (cf. Quirk et al. 1985: 57–58, 163, 727–728). Cf. (9)–(11): (9) → Or: (10) → (11) →

a. b. c. a. b. a. b.

Someone sent him/John a copy of the letter. (EN) A copy of the letter was sent him/John. He/*him/John was sent a copy of the letter. Someone had laughed at her. She had been laughed at _. Someone had slept in the bed. The bed had been slept in _.

In Dutch, however, indirect objects are not in general promoted to

Germanic Passive Constructions

79

subject as easily as in Mainland Scandinavian and English (cf. Haeseryn et al. 1997: 1323, 1406–1412). Cf. (12): (12) a.

b.

c.

De politie bood de overvallers een vrijgeleide aan. the police offered the robbers a safe-conduct to ‘The police offered the robbers safe conduct.’ De overvallers werd door de politie vrijgeleide angebooden. the robbers-PL was-SG by the police safe-conduct to-offered ‘The robbers were offered safe conduct by the police.’ *De overvallers werden door de politie vrijgeleide angebooden. the robbers-PL were-PL by the police safe-conduct to-offered ‘The robbers were offered safe conduct by the police.’

Subject promotion of the indirect object is facilitated by a direct object in the shape of a subordinate or infinitival clause (13)–(14), a direct object that is part of a set phrase (or is only weakly referential) (15), or deletion of the direct object (16): (13) a.

→

b.

Or:

c.

(14) a.

→

b.

Or:

c.

Zij maakten hem wijs dat hij gek was. (DU) they made him wise that he mad was ‘They told him that he was mad.’ Hem werd wijs gemaakt dat hij gek was. him-OBL was wise made that he mad was ‘He was told that he was mad.’ Hij werd wijs gemaakt dat hij gek was. he-SU was wise made that he mad was Na afloop verzocht men de aanwezigen/hun de zaal zo spoedig mogelijk te verlaten. (DU) at end asked one the present-PL / them-OBL.PL the hall so soon possible to leave ‘At the end they asked those present / them to leave the room as soon as possible.’ Na afloop werd de aanwezigen / hun verzocht de zaal zo spoedig mogelijk te verlaten. at end was-SG the present-PL / them-OBL.PL asked the hall so soon possible to leave ‘At the end those present / they were asked to leave the room as soon as possible.’ Na afloop werden de aanwezigen / ze verzocht de zaal zo spoedig mogelijk te verlaten. at end were-PL the present-PL / they-SU.PL asked the hall so soon possible to leave

80

John Ole ASKEDAL (15) a.

→

b.

Or:

c.

(16) a.

→

b.

Or:

c.

Men wees ons telkens de deur. (DU) one showed us-OBL.1PL always the door ‘They always showed us the door.’ Ons werd telkens de deur gewezen. us-OBL.1PL was-SG always the door shown ‘We were always shown the door.’ We werden telkens de deur gewezen. we-SU.PL were-PL always the door-sg shown Op ons bellen maakten zij ons open. (DU) upon our ringing made they us-OBL.1PL open ‘After we had sounded the doorbell, they opened the door to us.’ Op ons bellen werd ons opengedaan. upon our ringing was-SG us-OBL.1PL open-done Op ons bellen werden we opengedaan. upon our ringing were-PL we-SU.1PL open-done

Former dative-governing verbs with one object argument behave inconsistently. Whereas the passive of ‘help’ retains the former direct dative object, resulting in an impersonal passive (17b), it is made the subject of the personal passive of ‘obey’ (18c): (17) a.

→

b.

Not: c. (18) a.

Not:→b. But:→c.

Helpt men u al? (DU) helps one you-OBL.2PL already ‘Do they help you already?’ Wordt u al geholpen? is-SG you- OBL.2PL already helped ‘Are you being helped already?’ *Worden je-SU.2PL al geholpen? are you already helped De jongeren gehoorzamen tegenwoordig niet meer de ouderen. (DU) the younger-PL obey today not more the older-PL ‘In our days, the young people don’t obey their elders any more.’ *De ouderen wordt tegenwoordig (door de jongeren) niet meer gehoorzaamd. the older-PL is-SG today (by the younger-PL) no more obeyed De ouderen werden tegenwoordig (door de jongeren) niet meer gehoorzaamd. the older-PL are-PL today (by the younger-PL) no more obeyed ‘Today, older people are not being obeyed by their juniors any more.’

In comparison with Mainland Scandinavian and English, the Dutch state of affairs indicates a transitional stage where a former general case-based restriction (of the kind still operative in German; cf. below) has

Germanic Passive Constructions

81

been replaced by less general lexical rules. To my knowledge, the West Frisian passive has not been made the subject of such detailed scrutiny as the Dutch passive but it appears that only direct objects are promoted to subject in the passive (cf. Tiersma 1985: 110–111). Cf. e.g. (19): (19) a. b.

Boukje jout de appel oan har mem. (FR) Boukje gives the apple to her mother De appel wurdt fan Boukje oan har mem jûn. the apple is from Boukje to her mother given ‘The apple is given to her mother by Boukje.’

Formerly case-marked indirect dative objects are not in general accessible to promotion. In such cases an impersonal passive with retained oblique object is used. Cf. (20): (20) a.

→

b.

Men fertelt my dat ... (FR) one tells me that … ‘They tell me that ...’ My wurdt (der) ferteld dat ... me is-3SG (there) told that ... ‘I am being told that ...’ (lit.: ‘Me is being told that ...’)

The +NP-case languages German and Icelandic behave in a syntactically conservative fashion. German non-subject promotion is subject to an absolute accusative requirement (cf. Duden 2005: 953): a direct object in the accusative is made subject; all other sentence elements, including dative objects of all kinds and NPs in PPs, remain as they are in the active. Cf. (21): (21) a.

→

b.

Or:

c.

Jemand hat vermutlich dem König die Sklaven geschenkt. (GE) someone has presumably the king-IO.DAT the slaves-DO.ACC given ‘Someone has presumably given the king the slaves.’ Dem König sind die Sklaven vermutlich geschenkt worden. the king-DAT are-PL the slaves-SU.PL.NOM presumably given been ‘The king has presumably been given the slaves.’ Die Sklaven sind vermutlich dem König geschenkt worden. the slaves-SU.PL.NOM presumably are-PL the king-DAT given been ‘The slaves have presumably been given to the king.’

In the singular case of both the indirect and the direct object being in the accusative in the active, the passive construction normally adheres to the

82

John Ole ASKEDAL

prototypical pattern in (21), cf. (22), but the accusative indirect object may be promoted when the direct object carries no distinctive case marking, cf. (23):7 (22) a.

b.

(23) a.

→

b.

Or:

c.

Sie lehrte ihn viele russische Wörter. she taught him-IO.ACC many Russian words-DO.PL.ACC ‘She taught him many Russian words.’ Ihm wurden viele russische Wörter gelehrt. him-DAT were-PL many Russian words-SU.PL.NOM taught ‘He was taught many Russian words.’ Sie lehrte ihn die russische Sprache. she taught him-IO.ACC the Russian language-DO.ACC/NOM ‘She taught him the Russian language.’ Ihm wurde die russische Sprache gelehrt. him-DAT was the Russian language-SU.NOM taught ‘He was taught the Russian language.’ Er wurde die russische Sprache gelehrt. he-SU.NOM was the Russian language-ACC/NOM taught

In Icelandic, non-subject promotion obeys an equally strict accusative requirement (cf. Thráinsson 2007: 152–153, 160, 162, 238–242). Cf. (24): (24) a.

→

b.

Or:

c.

Einhver hefur líklega gefið konunginum þrælana. (IC) someone has presumably given king-IO.DAT.DEF slaves-DO.PL.ACC.DEF ‘Someone has presumably given the king the slaves.’ Konunginum hafa líklega verið gefnir þrælarnir. king-sg.DAT.DEF have presumably been given-PART.PL.NOM slaves-SU.PLNOM.DEF Þrælarnir hafa líklega verið gefnir konunginum. slaves-SU.PL.NOM. DEF have presumably been given-PART.PL.NOM king- SG.DAT. DEF

Due to the accusative requirement on non-subject promotion, Icelandic also follows German in not allowing extraction of NP out of PP in the passive (Thráinsson 2007: 153), cf. (25a–b), in spite of the fact that extraction of NP out of PP is possible in active clauses for purposes of topicalization, cf. (25d); hence an impersonal passive is used, cf. (25c):

7

On passive constructions based on active valency patterns with two accusatives, cf. Askedal (1980) and Plank (1987).

Germanic Passive Constructions (25) a. Not:→b.

But: c. d.

83

Einhver hefur sofið í þessu rúmi. (IC) someone has slept in this bed *Þetta rúmið hefur verið sofið í _. this bed-NOM has been sleept in _ ‘This bed has been slept in.’ Það hefur verið sofið í þessu rúmið. it has been slept in this bed Þessu rúmi hefur einhver sofið í _. this bed-DAT has someone slept in _ ‘This bed someone has slept in.’

The basic passivization rules of the third NP-case-language Faroese correspond to those of German and Icelandic. A couple of differences are to be noted that indicate a certain weakening of the syntactic role of case marking (cf. Thráinsson et al. 2004: 266–274; Thráinsson 2007: 185). In the passive of ditransitive constructions with a direct accusative and an indirect dative object, the accusative object is usually made subject and the indirect dative object is retained as such, as in Icelandic and German. Cf. (26): (26) a.

→

b.

Hann seldi bóndanum kúnna. (FA) he sold farmer-IO.SG.DAT.DEF cow-DO.SG.ACC.DEF ‘He sold the cow to the farmer.’ Kúgvin varð seld bóndanum. cow-SU.SG.NOM.DEF was sold farmer-SG.DAT.DEF ‘The cow was sold to the farmer.’

In two-place constructions with a dative as the only object, the traditional accusative requirement is still valid with a number of verbs but promotion of the dative also occurs and is probably the more common pattern in modern Faroese (Thráinsson et al. 2004: 266–270). Cf. (27) with ‘to thank’ vs. (28) with ‘to help’: (27) a. b. (28) a. →

b.

Teir takkaðu honum. (FA) they thanked him-DAT Honum varð takkað. him-DAT was thanked Teir hjálptu honum. (FA) they helped him-DAT Hann varð hjálptur. he-SU.NOM was helped

84

John Ole ASKEDAL

This state of affairs in NP-case Faroese is reminiscent of the vacillation found in cases like (13)-(16) in the non-NP-case language Dutch, suggesting that even in NP-case Faroese, a formerly general morphological requirement is yielding to lexical rules.8 Like Icelandic, Faroese has extraction out of PP (‘preposition-stranding’; cf. Thráinsson et al. 2004: 290, 292), but no extraction with non-subject promotion to nominative subject. From a typological point of view, it is noteworthy that neither the verb-final nor the NP-case languages allow for extraction out of PP with promotion, this being an option in non-verb-final and non-NP-case languages only. The range of non-subject promotion options in the present-day individual Germanic languages seem to form a kind of typological cline. Consider (29): (29) The Modern Germanic non-subject promotion cline i NO, SW, DA, EN: DO, IO, NP in PP (PO and Adv) ii DU: “Accusative” DO (“Dative” DO, IO) iii FA: Accusative DO (Dative DO) iv FR: DO v GE, IC: Accusative DO

> > > >

The languages in question may be taken to represent several stages of a stepwise development: The NP-case languages German and Icelandic are the most conservative and retain the ancient case requirement that subjects of passive clauses correspond to the accusative object of corresponding active clauses. In Faroese, this restriction has in part been loosened. The non-NP-case language West Frisian shows remnants of the accusative case requirement even though the relevant morphological case opposition has been lost. Presumably, the Frisian state of affairs can be described as the transfer of a formerly morphology-based passive subject formation requirement to the lexicon. Dutch is developing, perhaps even more strongly, in the same direction; while not abandoning all syntactic reflexes of the old case requirement, Dutch seems to represent a further approximation towards the state of affairs in English and Mainland Scandinavian, where the case requirement on non-subject promotion has been lost altogether.

8

One may note that Icelandic has a far larger variety of three-place valency patterns, reflecting the corresponding variety of Old Norse, than Faroese, where considerable simplification of the overall system has taken place. Cf. Thráinsson (2007: 173–178, 216–222, 228–232), Thráinsson et al. (2004: 261–265, 431–433).

Germanic Passive Constructions

85

2.2 Impersonal passive: Subject demotion without non-subject promotion With the exception of English, all modern Germanic languages possess impersonal passive constructions. Impersonal passives are used under the following conditions: (i) lack of a non-subject to which non-subject promotion can apply; or (ii) non-activation of non-subject promotion even though the valency conditions for its application are fulfilled. In the former case, the extent to which the impersonal passive is used depends on the categorial scope of non-subject promotion to subject or, inversely, the extent to which non-subject promotion is subject to restrictions. For instance, when non-subject promotion is subject to a strict accusative case requirement, as in German and Icelandic, there is correspondingly less scope for personal passives. On the other hand, the non-NP-case and non-verb-final languages – Mainland Scandinavian and English – make no distinction between former accusative and dative objects with regard to non-subject promotion, yielding in these cases personal passives. As noted, the non-NP-case language Dutch, however, in this respect behaves in a fashion similar to Faroese. Cf. for instance (30)–(31) and (32), which have personal passive parallels in English and the Mainland Scandinavian languages (cf. the English glosses): (30) a.

→

b.

(31) a.

→

b.

(32) a.

→

b.

Man half ihm sofort. (GE) one helped him-DAT immediately ‘They helped him immediately.’ Ihm wurde sofort geholfen. him-DAT was immediately helped ‘He was helped immediately.’ Die Leute arbeiten seit Jahrzehnten an diesem Wörterbuch. (GE) the people work since decades on this dictionary ‘These people have been working on this dictionary for decades.’ An diesem Wörterbuch wird seit Jahrzehnten gearbeitet. on this dictionary is since decades worked ‘This dictionary has been worked on for decades.’ Einhver hjálpaði strákunum með heimaverkefnið. (IC) someone helped boy-SG.DAT. DEF with homework-DEF ‘Someone helped the boy with the homework.’ Strákunum var hjálpað með heimaverkefnið. boy-SG.DAT.DEF was helped with homework-DEF ‘The boy was helped with the homework.’

As noted, the NP-case language Faroese shows a certain amount of

86

John Ole ASKEDAL

vacillation between impersonal and personal passive constructions with verbs governing a single dative object, cf. (27)–(28). All the languages, with the exception of English, have impersonal passives in those cases where there are no candidates for promotion. Cf. e.g. (33)–(35): (33) a.

→

b.

(34) a.

b.

(35) a.

→

b.

Margir hafa áreiðanlega dansað þá. (IC) many have certainly danced then ‘Many people were certainly dancing then.’ Það hefur áreiðanlega verið dansað þá. it has certainly been danced then ‘There was certainly dancing going on then.’ Gestern arbeiteten die Leute bis spät in die Nacht hinein. (GE) yesterday worked the people until late in the night into ‘Yesterday people were working until late in the night.’ Gestern wurde bis spät in die Nacht hinein gearbeitet. yesterday was until late in the night into worked ‘Yesterday there was work going on until late in the night.’ I går arbeidet man til langt på natt. (NO) yesterday worked one until long at night ‘Yesterday they worked until late in the night.’ I går ble det arbeidet til langt på natt. yesterday was it worked until long at night ‘Yesterday there was work going on until late in the night.’

The status of second scenario – (ii) non-activation of structurally possible non-subject promotion – varies in the individual languages. In the languages having a strict accusative case requirement, accusative object promotion is as a rule mandatory. Cf. e.g. German (36):9 (36) a.

9

Man hat gestern einen neuen Fehler entdeckt. (GE) one has yesterday a new error-ACC.SG.MASC discovered ‘Yesterday they discovered another error.’

The following containing the locution Karten spielen ‘to play cards’ is a maginal exception: (i) a. Man spielte damals häufig Karten. one played then often cards-ACC.PL ‘At that time they often played cards.’ → b. Damals wurde häufig Karten gespielt. then was-SG often cards-ACC.PL played ‘At that time cards were often played.’

Germanic Passive Constructions →

b.

Not: c.

87

Gestern wurde ein neuer Fehler entdeckt. yesterday was a new error-NOM.SG.MASC discovered ‘Yesterday another error was discovered.’ *Gestern wurde einen neuen Fehler entdeckt. yesterday was a new error-ACC.SG.MASC discovered

Whereas the accusative requirement operates across the board in German and Icelandic, its operation in Faroese is not equally general, cf. (27)–(28). The following example (37) with the locution ynskja eina góðu ferð ‘to wish (someone) a good journey’ shows retention of both indirect dative and direct accusative object (Thráinsson et al. 2004: 273–274): (37) a.

→

b.

Teir ynsktu honum eina góðu ferð. (FA) they wished him-DAT a good journey-ACC.SG.FEM ‘They wished him a good voyage.’ Honum varð ynskt eina góðu ferð. him-DATIVE was wished a good journey-ACC.SG.FEM ‘He was wished a good voyage.’

The second scenario plays a systematic, typologically conditioned role as a syntactic means of presentation in a number of languages. In all the languages there is a tendency both in active and passive constructions to place a – mostly indefinite – NP in a late position in the clause for reasons of rhematic focusing. In non-NP-case languages, the question arises whether such NPs are in fact syntactic subjects or objects. In non-NP-case languages with verb agreement, this question is, at least on the face of it, decided in favour of subject and, concomitantly, personal passive. For this reason, the indefinite NP in the following Dutch (38b) and West Frisian (39b) examples may be identified as syntactic subject because the finite verb agrees with it, even though these sentences also may contain er and der, respectively, as a formal subject or subject slot filler:10 (38) a.

→

10

b.

Gisteren heft men veel prijzen uitgedeeld. (DU) yesterday has one many prizes awarded ‘Yesterday they awarded many prizes.’ Er werden gisteren veel prijzen uitgedeeld. there were yesterday many prizes awarded ‘There were many prizes awarded yesterday.’

On Dutch er- and West Frisian der-presentatives cf. Haeseryn et al. (1997: 467–471), Tiersma (1985: 111).

88

John Ole ASKEDAL Or:

c.

(39) a.

→

b.

Or:

c.

Gisteren werden (er) veel prijzen uitgedeeld. yesterday were (there) many prizes awarded ‘Yesterday many prizes were awarded.’ Moarn wurdt men twa hinders ferkoche. (FR) tomorrow will one two horses sell ‘Tomorrow they will sell two horses.’ Der wurden moarn twa hinders ferkocht. there are-PL tomorrow two horses sold ‘There will be two horses sold tomorrow.’ Moarn wurden (der) twa hinders ferkocht. tomorrow are-PL (there) two horses sold ‘Tomorrow two horses will be sold.’

In corresponding English there-constructions (cf. Castillo 2009), agreement operates in the same way as in functionally corresponding Dutch er- and West Frisian der-constructions. Cf. e.g. (40): (40) a. b. c.

... because they arrested three people at noon. (EN) ... because there were three people arrested at noon. *... because there were arrested three people at noon.

Such English there-passive constructions seem, however, to occur far less frequently than their active counterparts in English and their passive counterparts in Dutch, Frisian or modern Mainland Scandinavian (on the latter cf. below). In English the question arises whether to place the rhematic NP before or after the non-finite lexical verb; whereas a sentence like (40c) is commonly considered ungrammatical (cf. Castillo 2009: 93), (41c) for some reason seems to sound better than (41b) as passive counterpart of (41a): (41) a. → b. Or: c.

During the following week, they held several splendid banquets. (EN) During the following week, there were several splendid banquets held. During the following week, there were held several splendid banquets.

Nor is the relationship between the sentences in (42) altogether straightforward: (42) a. b. c. d.

Someone had stolen a whole box (from him). (EN) Someone had a whole box stolen (from him). There has been a whole box stolen (from him). There has been stolen a whole box (from him).

Germanic Passive Constructions

89

First, in (42a) the subject someone denotes the thief, whereas in (42b) it denotes the person suffering the theft. Thus (42c–d) can be considered passive counterparts of (42a) but not of (42b). In terms of usage acceptability (42c) seems preferable to (42d). On closer inspection, in (42c) a whole box stolen would seem to be analyzable as a small-clause construction, where the non-finite form stolen following the NP a whole box is the predicate of the small clause, and the small clause as such appears to occupy subject position. In (42d) the indefinite element a whole box appears in what is the normal object position but this sentence appears to be lower on the usage acceptability scale than (42c) (cf. (40c)). With regard to subject demotion and non-subject promotion, (42d) would be a case of impersonal passive with subject demotion only, but with insertion of the formal subject there, whereas (42c) would show both subject demotion and non-subject promotion and in addition insertion of there. This creates a dilemma concerning there. If there is analyzed as a subject element, as is commonly done, the small clause analysis proffered above would entail that the more acceptable sentence (42c) in fact contains two subjects, which is a syntactic outrage. On the other hand, the less acceptable sentence (42d) would from a syntactic point of view be more normal, having there in subject and the indefinite NP a whole box in non-subject position. A possible way out would be to assume a ‘squish’ analysis of there (cf. Ross 1974), meaning that there would be more or less of a subject, i.e. would differ more or less from the subject prototype in accordance with its syntactic environment. Mainland Scandinavian represents another, more common and at least superficially more internally consistent case. For instance, in Norwegian presentational passive sentences like (43b–c) a sentence element that could also be the subject of a normal passive sentence without the formal subject det (cf. (43d)) occupies the post-verb position, which is the normal position of a direct object (cf. Faarlund et al. 1997: 720): (43) a.

→

b.

Or:

c.

Vs.

d.

De utgav mange gode bøker / de beste bøkene i denne perioden. (NO) they published many good books / the best books in this period ‘They published many good books / the best books during this period.’ Det ble utgitt mange gode bøker / *de beste bøkene i denne perioden. it was published many good books / the best books in this period ‘There were published many good books during this period.’ I denne perioden ble *(det) utgitt mange gode bøker / *de beste bøkene. in this period was (it) published many good books / the best books ‘During this period, there were published many good books.’ Mange gode bøker/De beste bøkene ble utgitt i denne perioden. many good books / the best books were published in this period ‘Many good books / The best books were published during this period.’

90

John Ole ASKEDAL

Sentences such as (43b–c) are amenable to a syntactic analysis in terms of, first, subject demotion without non-subject promotion and, second, addition of a merely formal subject det. In contrast to corresponding elements in subject position, such post-verbal indefinite NPs are subject to the same indefiniteness requirement as the syntactically corresponding so-called “potential subjects” in active clauses like (44b–c). The formal subject is in general mandatory in both passive sentences like (43b–c) and active sentences like (44b–c). Cf.: (44) a.

b.

Or:

c.

Mange gode bøker / De beste bøkene var kommet i denne perioden. (NO) many good books / the best books were come in this period ‘Many good books / The best books had appeared during this period.’ Det var kommet mange gode bøker / *de beste bøkene i denne perioden. (NO) it was come many good books / the best books in this period ‘Many good books / The best books had appeared during this period.’ I denne perioden var *(det) kommet mange gode bøker / *de beste bøkene. in this period had (it) come many good books / the best books ‘During this period many good books had appeared.’

Both in passive and active det-sentences, syntactic object properties can be identified in the indefinite NP (cf. Faarlund et al. 1997: 833–837, 845–847).11 Consider for instance the way VP-pronominalization works in (45a–b), encompassing in the same way the indefinite post-verbal sentence element: (45) a.

b.

Var det blitt utgitt mange gode bøker? – Ja, det var det. (NO) was it been published many good books? – yes, it was it ‘Had many good books been published. – ‘Yes, they had.’ Var det kommet mange gode bøker? – Ja, det var det. was it come many good books? – yes it was it ‘Had many good books appeared? – Yes, they had.’

Danish and Swedish behave in basically the same way as Norwegian in these respects,12 but two minor differences may be noted. Differently from Norwegian and Swedish, the desemanticized local adverb der ‘there’, not the neuter personal pronoun det ‘it’ is used as a formal subject in Danish 11

12

As noted by Faarlund et al. (ibid.), indefinite NPs in active clauses have both object and subject properties, underscoring the object nature of indefinite NPs in passive det-constructions. Cf. for instance Teleman et al. (1999: 371–376), Hansen (1967: 171–172), Heltoft & Falster Jakobsen (1996).

Germanic Passive Constructions

91

presentational sentences.13 In Swedish, the formal subject, although commonly used, appears to be omissible to a somewhat greater extent than in Danish and Norwegian (cf. Askedal 1997: 202). The syntactic behaviour of the formal subject and the indefinite NP in presentational sentences strongly indicates that Mainland Scandinavian has, in its basic typological makeup, passed on to a topological, or ‘configurational’, syntax where position is a grammaticalized coding device for the syntactic relation categories subject and object.14 Germanic examples can thus be found for the following subtypes (i)-(iii) of the impersonal passive: (46) The impersonal passive (IP) in Modern Germanic i NO, SW, DA, (FA): IP of intransitive and transitive verbs ii a. GE, IC; b. DU, FR: IP of intransitive verbs iii EN: no IP iv — IP of transitive (but not intransitive) verbs

With the exception of Faroese and German examples like (37b) and (i) in n. 8, impersonal passives of transitive verbs are restricted to presentational constructions with a retained object. From the non-existence subtype (iv) in (46) it follows that the existence of impersonal passives implies the existence of personal passives.15 2.3 Occurrence or non-occurrence of formal subject in the impersonal passive As indicated at the end of the preceding section, Germanic languages vary as to the form and use of expletive elements (formal subjects and the like; cf. Harbert 2007: 224–234). In Germanic two historical pathways for the formation of expletives exist: (i) desemanticization of a neut. sg. personal pronoun ‘it’ (German, Dutch, Frisian, English) or ‘that’ (Icelandic, Faroese, Mainland Scandinavian); (ii) desemantization of the local adverb ‘there’. Icelandic, Faroese, Norwegian, Swedish and German opt for the former 13

14

15

Danish has det ‘it’ in lexical impersonals (cf. 2.3 below) and thus makes a distinction between lexical and constructional impersonals that is not made in the most common varieties of Norwegian and Swedish. Cf. e.g. Askedal (1993) on Norwegian and Heltoft (2001) on Danish. As noted for instance in Askedal (2001, with references), this syntactic restructuring process to all appearances predates the demise of morphological case marking. But it should be noted that English comes close to representing this universally rare type when verb agreement is disregarded as a criterion in constructions like (40b), (41b–c) and (42c–d).

92

John Ole ASKEDAL

pathway both in lexical and constructional impersonals (e.g. es regnet ‘it’s raining’ and es kamen drei Leute ‘there came three people’), thus having a single expletive element, whereas the North Sea Germanic languages Danish, Dutch, Frisian and English opt for the neut. sg. pathway in lexical and the local adverb pathway in constructional impersonals (it’s raining vs. there came three people), thus showing expletive variation. 16 In addition, the Germanic languages also vary with regard to the linear distribution of expletives, both in lexical and constructional impersonals including impersonal passives.17 (i) German and Icelandic have obligatory ‘it’-expletive in first position in assertive main clauses but no clause-internal expletives (cf. Zifonun et al. 1997: 1793–1796; Thráinsson 2007: 310, 334): (47) a.

b.

(48) a.

b.

Gestern abend wurde (*es) bestimmt lange getanzt. (GE) yesterday evening was (it) certainly long danced ‘Yesterday evening there was certainly dancing going on for a long time.’ Es wurde gestern abend bestimmt lange getantzt. it was yesterday evening certainly long danced ‘There was certainly dancing going on for a long time yesterday evening.’ Þá hefur (*það) áreiðanlega verið dansað. (IC) then has (it) certainly been danced ‘Then there was certainly dancing going on.’ Það hefur áreiðanlega verið dansað þá. it has certainly been danced then ‘There was certainly dancing going on then.’

(ii) Dutch and West Frisian have obligatory ‘there’-expletive in first position in assertive main clauses; ‘there’-expletives also occur clause-internally, where they may even be obligatory, at least in Dutch (cf. Haeseryn et al. 1997: 477; Tiersma 1985: 111): (49) a.

16

17

Meestal zingt men. (DU) mostly sings one ‘One mostly sings.’

Duden (2005: 830–832), Faarlund et al. (1997: 678–683, 830–833, 845–847), Haeseryn et al. (1997: 258–259, 467–477, 1137–1138), Hansen (1967, vol. 3: 49–52), Teleman et al. (1999, vol. 4: 53–64), Thráinsson (2007: 309–313, 333–334), Thráinsson et al. (2004: 281–287, 436), Tiersma (1985: 107, 105). In the following, only the use of expletives in impersonal passive constructions is illustrated.

Germanic Passive Constructions b.

(50) a.

b.

(51) a.

b.

93

Meestal wordt er gezongen. mostly is there sung ‘There is mostly singing gong on.’ Er wordt hier geklopt. (DU) there is here knocked ‘Someone is knocking on the door here.’ Hier wordt (er) geklopt. here is (there) knocked ‘Here someone is knocking on the door.’ Hjir waard (der) altyd dûnse. (FR) here was (there) always danced ‘Here there was always dancing going on.’ Der waard hjir altyd dûnse. there was here always danced ‘There was always dancing going on here.’

(iv) Faroese has obligatory ‘it’-epletive in first position in assertive main clauses and largely optional ‘it’-expletive clause-internally (Thráinsson et al. 2004: 285–287; 436): (52) a.

b.

c.

d.

Tað hefur eitt par dansað í havanum í gjárkvøldid. (FA) it has a pair danced in garden-DEF in yesterday-evening-DEF ‘There was a couple dancing in the garden last night.’ Tað varð dansað í havanum í gjárkvøldid. it was danced in garden-DEF in yesterday-evening-DEF ‘There was dancing going on in the garden last night.’ Í gjárkvøldid varð tað dansað í havanum. in yesterday-evening-DEF was it danced in garden-DEF ‘Last night there was dancing going on in the garden.’ Í havanum varð nógv dansað í gjárkvøldid. in garden-DEF was much danced in yesterday-evening-DEF ‘In the garden there was a lot of dancing going on last night.’

(iv) Danish has obligatory ‘there’-expletive in first position in assertive main clauses and clause-internally (Hansen 1967, vol. 3: 50–51): (53) a.

Der dansedes altid til langt på nat i det hus. (DA) it danced-PASS always until late in night in that house ‘There was always dancing going on until late in the night in that house.’

94

John Ole ASKEDAL b.

I det hus dansedes *(der) altid til langt på nat. in that house danced-PASS there always into late in the night ‘In that house there was always dancing going on until late in the night.’

(v) Norwegian and Swedish have obligatory ‘it’-expletive in first position in assertive main clauses and, mostly, clause-internally (Faarlund et al. 1997: 845–847; Teleman et al. 1999, vol. 4: 57): (54) a.

b.

Det ble danset til langt på natt i det huset. (NO) it was danced till long on night in that house-DEF ‘There was always dancing going on until late in the night in that house.’ I det huset ble *(det) danset til langt på natt. in that house-DEF was it danced till long on night ‘In that house there was always dancing going until late in the night.’

The above comparison may be summarized thus (56): (55) Expletives in Modern Germanic impersonal passives i GE, IC ‘it’-expletive as topologically conditioned slot-filler clause-initially ii DU, FR ‘there’-expletive clause-initially and, inconsistently, clause-internally; functional ambivalence between formal subject and toplogically conditioned slot-filling function iii FA ‘it’-expletive clause-initially and, inconsistently, clauseinternally; functional ambivalence between formal subject and slot-filling function iv EN, DA ‘there’-expletive as formal subject both clause-initially and clause-internally v NO, SW ‘it’-expletive as formal subject both clause-initially and clause-internally

The non-NP-case languages show greater variation with regard to the use of expletives. Dutch, West Frisian and Faroese seem to represent an intermediate stage between options (i) on the one hand, and (iii–iv) on the other. 3. THE PREDICATE DOMAIN: PASSIVE VERB MORPHOLOGY In present-day Germanic both analytic and synthetic passive morphology are found. The analytic type, involving passive auxiliaries, is older and more widespread, being used in all the languages, whereas the synthetic, suffixal type is a characteristic of North Germanic (with the exception of Icelandic,

Germanic Passive Constructions

95

cf. section 3.2). The analytic constructions come in two inflectional varieties, either with a participle showing agreement with the subject of the sentence or with a non-agreeing, so-called ‘supine’ form. 3.1 Analytic (periphrastic) passive morphology Within the Germanic area, there is a certain amount of variation as far as passive verb morphology is concerned. The continental West Germanic verb-final languages behave uniformly, having analytic, non-agreeing participle 18 constructions throughout, as do also English, Danish and Norwegian Riksmål/Bokmål. Cf. (56)–(60): (56) a. →

b.

(57) a.

→

b.

(58) a. →

b.

(59) → (60) →

a. b. a. b.

Sie entdeckten die Fehler sofort. (GE) they discovered the errors at-once Die Fehler wurden sofort entdeckt. the errors were at once discovered-SUP ‘The errors were discovered at once.’ Ze bouwen het huis. (DU) they build the house ‘They are building the house.’ Het huis wordt gebouwd. the house is built-SUP ‘The house is being built.’ Sy seagen ús. (FR) they saw us Wy warden sjoen. we were seen-SUP They built a new house. (EN) A new house was built. De byggede et nyt hus. (cf. (59a)) (DA) Et nyt hus blev bygget. (cf. (59b))

In a Northern area encompassing Icelandic, Faroese, Norwegian Nynorsk, and Swedish, the historically originary agreeing participle construction is still used. Cf. (61): (61) a.

18

Vegirnir voru gerðir fyrir hestvagna. (IC) roads-SU.MASC.PL.DEF were done-PART.MASC.PL for horse-carriages ‘The roads were built for horse carriages.’

Rather: supine; cf. Bech (1955: 12–20) on German.

96

John Ole ASKEDAL b.

c.

d.

Teir vóru sviknir av øllum. (FA) they-SU.MASC.PL were betrayed-PART.MASC.PL by all ‘They were betrayed by all.’ Dei vart drepne i ulykka. (NO-NN) they-SU.PL were killed-PART.PL in accident-DEF ‘They were killed in the accident.’ Av sagans personer är båda kungasönerna ganska fritt behandlade. (SW) of saga-DEF-GENITIVE persons are both king-sons-PL quite freely treated-PL ‘Of the persons in the saga, the sons of the king are described quite freely.’

In the latter group of languages, the participles conform to the agreement rules pertaining to predicative adjectives. In Danish and Norwegian, which have non-agreeing participles in the passive, predicative adjectives and analytic passive constructions behave differently. Cf. (62): (62) a.

b.

Leilighetene ble likevel dyre. (NO) apartments-SU.PL.DEF became after-all expensive-ADJ.PL ‘The apartments became expensive after all.’ Leilighetene ble likevel bygget. apartments-SU.PL.DEF became after-all built-SUP ‘The apartments were built after all.’

It is tempting to view this neutralization of agreement in the verb constructions, contrasting with non-neutralization in predicative adjectives, as a sign of grammaticalization. 3.2 Synthetic (suffixal) passive morphology The origin of North Germanic synthetic passive morphology is to be found in an enclitic reflexive pronoun that has been further grammaticalized as a passive suffix (cf. Harbert 2007:327–329). In all the languages in question, examples of lexicalization are found alongside with grammaticalized usage. It is a moot question how far the functional grammaticalization has gone in the individual languages. It is commonly agreed that Icelandic has not yet developed past the lexical stage (cf. e.g. Anderson 1990 and the discussion summarized in Harbert 2007: 322–325). From an areal point of view one may note that the morphological generalization of the synthetic passive across different verb forms has progressed further in the more easterly languages than in the more westerly ones. Cf. (63)–(67):

Germanic Passive Constructions

97

(63) Dörren måste öppnas / öppnas / öppnades / har öppnats långsamt. (SW) the door must open-INF.PASS / open-PRES.PASS / open-PRET.PASS / has open-SUP.PASS slowly ‘The door must be opened / is being opened / was opened / has been opened slowly.’ (64) a. Døren skal åbnes / åbnes / åbnedes langsomt. (DA) the door must open-INF.PASS / open-PRES.PASS / open-PRET.PASS slowly ‘The door must be opened / is being opened / was opened slowly.’ b. Pengene blev stjålet/*stjales. the money was stolen-SUP.PASS ‘The money was stolen.’ (65) a. Døren skal åpnes / åpnes langsomt. (NO-RM/BM) door-DEF shall open-INF.PASS / open-PRES.PASS slowly ‘The door has to be opened / is being opened slowly.’ b. Lyden hørtes klart og tydelig. sound-DEF hear-PAST.PASS clarly and distinctly ‘The sound was heard clearly and distinctly.’ (66) a. Døra skal opnast langsamt. (NO-NN) door-DEF shall open-INF.PASS slowly ‘The door has to be opened slowly.’ b. Bråket høyrdest lang veg. noise-DEF hear-PAST.PASS long way ‘The noise was audible at a considerable distance.’ (67) a. Her skal eitt hús byggjast. (FA) here shall a house build-INF.PASS ‘Here a house will be built.’ b. Nógv fekst fyri fiskin. something received-PAST.PASS for fish-THE ‘They got a lot of money for the fish.’ c. Ikki eitt livandi kendist. not a living knew-PAST.PASS ‘There was no indication of a living thing.’

Norwegian Nynorsk (66b) and Faroese (67c) are examples of special cases of lexicalization. It should be noted that the analytic and the synthetic passive are not in all cases full semantic equivalents. For instance, Norwegian (68a) may be interpreted as a general rule, request or command, whereas (68b) sounds more like future-oriented promise:

98

John Ole ASKEDAL (68) a.

b.

Barna skal hentes klokken tre. (NO-RM/BM) children-DEF shall fetch-INF.PASS clock-DEF three ‘The children have to be picked up at three o’clock.’ Barna skal bli hentet klokken tre. children-DEF shall be fetched clock-DEF three ‘The children will be picked up at three o’clock.’

The table in (69) illustrates the geographical cline that probably reflects the origin of the grammaticalization of the reflexive as a passive suffix in the eastern part of Scandinavia, indicating a relationship with corresponding Baltic and Slavic developments:19 (69) Distribution of the passive -s(t)-suffix in Mainland and Insular Scandinavian infinitive present past supine i SW + + + + ii DA + + +/– – iii NO-RM/BM + + – (+) – iv NO-NN + – (+) – (+) – v FA + – (+) – (+) – vi IC – – – –

3.3 Auxiliaries in passive periphrases with agreeing participles and non-agreeing supines The most common passive auxiliaries are formally identical with copula verbs and represent different grammaticalization stages of such verbs. For instance, German has three copula verbs that contrast with each other with regard to stative (statal), or unmarked, transformative (ingressive) and intransformative (continuative) aspectual meaning. Cf. (70): (70) Er ist / wird / bleibt krank. (GE) he is-STAT / becomes-TRANSF / remains-INTRANSF ill

The same verbs reappear in passive constructions but in this function, werden ‘become’ has been desemanticized, yielding a new opposition between aspectually neutral werden and markedly statal sein ‘be’ in the sense that its transformative (ingressive) meaning has been lost, while bleiben remains semantically in place. Cf. (71):

19

Cf. Tesnière (1966: 250), Dahl & Koptevskaja-Tamm (1972: 17–24).

Germanic Passive Constructions (71) a.

b.

c.

99

Das bis vor wenigen Jahren noch nicht erschlossene Gebiet wird jetzt besiedelt. (GE) the until before few years still not opened area is now populated ‘The area that just a few years ago had not been made accessible is now being populated.’ Das früher unzugängliche Gebiet ist jetzt besiedelt. the earlier inaccessible area is now populated ‘The area that was formerly inaccessible is now populated.’ Dieses Gebiet soll auch künftig besiedelt bleiben. this area shall also in the future populated remain ‘This area will remain populated even in the future.’

In consequence, the markedness relationship between werden and sein ‘be’ has been reversed. Cf. (72): (72) sein-STAT werden-NEUTRAL

< werden-TRANSF < sein-STAT

< bleiben-INTRANSF → < bleiben-INTRANSF

The semantic development in (73) has taken place in all three Continental Germanic languages. Mainland Scandinavian has borrowed its most widespread passive auxiliary –Danish blive, Swedish bli(va), Norwegian bli – from Low German (Markey 1969). It started out as the Low German equivalent of High German intransformative bleiben ‘remain, stay’, and then acquired an ingressive (transformative) meaning variant, presumably before it was borrowed into Scandinavian; in Scandinavian it finally suffered desemanticization as a passive auxiliary along the same lines as German werden etc. (cf. Nübling 2006: 186–188). All three meanings or functions exist in the modern Scandinavian languages. Cf. for instance (73): (73) a. b.

Han ble hjemme fordi han ble syk. (NO) he stayed-COP.INTRANSF because he became-COP.TRANSF ill Han ble behandlet godt. he was-AUX.PASS treated well

The semantic and functional shift from ingressive copula to passive auxiliary caused a markedness reversal between desemanticized ‘become’ and stative ‘be’ that parallels the one in Continental Germanic, cf. (71a–b) and (74a–b). Icelandic and English represent a state of affairs where ‘be’ does service both as an aspectually unmarked ‘dynamic’ and as a statal passive

100

John Ole ASKEDAL

auxiliary (74c):20 (74) Modern Germanic passive auxiliaries unmarked a. GE werden DU worden FR wurde b. DA blive NO-RM/BM: bli NO-NN verta/bli SW bli(va) FA verða/blíva c. IC vera EN be

statal sein zijn wêze være være vera vara vera

A sort of reflex of the older state of affairs represented by English and Icelandic is found in those languages that need not, or do not, insert the participle of the passive auxiliary to express perfect passive meaning (cf. Harbert 2007: 318–319). Cf. (75)–(80): (75) → (76) →

a. b. a. b.

(77) a. → b.

20

They have changed this now. (EN) This is changed (has been changed) now. Teir hafa broytt hetta nú. (cf. (75a)) (FA) Hetta er broytt (er vorðið/blívið broytt) nú. this is changed (is been / becomes been) now (cf. (75b)) De har forandret dette nå. (cf. (75a)) (NO-RM/BM) Dette er forandret (er/har blitt forandret) nå. this is changed (is/has been changed) now (cf. (76b))

In addition to the verbs mentioned here, a number of other verbs are used in the same conversion patterns and hence should be considered passive auxiliaries from a syntactic point of view. To these belong for instance English get and German gehören with (non-agreeing) participle (supine), Icelandic verða ‘become’ with agreeing participle, German sein ‘be’ with the zu-Infinitive, which have parallels in Dutch, Frisian, Mainland Scandinavian and Faroese, and similar constructions with bleiben ‘remain’, scheinen ‘seem’ and gehen ‘go’. i.e. ‘be possible’ with the zu-infinitive. Norwegian has so-called “double passives” with two passive participles (supines) and Norwegian and Swedish have “raising passives” without active equivalents. Furthermore, German and to a certain extent Dutch and Mainland Scandinavian have developed a “dative” or “recipient passive” characterized by indirect object conversion, supplementing the regular passive with non-subject conversion of varying categorial scope in the individual languages. Cf. for instance Harbert (2007: 320, 321–322, 326–327), Duden (2005: 554–562), Haeseryn et al. (1997: 962–963, 1420–1422), Tiersma (1985: 127–128), Faarlund et al. (1997: 847–853, 1029–1032), Hansen (1967, vol. 3: 91–92).

Germanic Passive Constructions (78) a.

→

b.

(79) a.

b.

(80) a.

→

b.

101

Sy hawwe ús sjoen. (FR) they have us seen ‘They have seen us.’ Wy binne sjoen (wurden). we are seen (been) ‘We are seen / have been seen.’ Þeir hafa gert vegina fyrir hestvagna. (IC) they have made roads-DEF for horse-carriages ‘They have built the road for horse carriages.’ Vegirnir eru gerðir fyrir hestvagna. roads-DEF are made for horse-carriages ‘The roads are built / have been built for horse carriages.’ Ze hebben het huis gebouwd. (DU) they have the house built ‘They have built the house.’ Het huis is gebouwd. the house is built ‘The house is built / has been built.’

3.4 Areal aspects of passive morphology in Germanic The following table (81) summarizes the areal distribution of supine, participle and passive s(t)-suffix passive formations in modern Germanic: (81) Distribution of the supine, participle and passive s(t)-suffix formations in the modern Germanic languages supine participle s(t)-suffix i GE + – – DU + – – FR + – – EN + – – ii DA + – + NO-RM/BM + – + iii NO-NN (+) + + iv SW – + + FA – + + v IC + – –

In Norwegian Nynorsk, the use of the participle is basic to the system but there is a fair amount of morphological neutralization and morphologically distinctive supine forms are often used even in cases where distinctive agreeing participles are available (cf. Faarlund et al. 1997: 473,

102

John Ole ASKEDAL

492–507, 518–520). Concerning the categorial and areal distribution of the North Germanic synthetic passive see (69). 4. SUMMARY AND TYPOLOGICAL PERSPECTIVES The following (82) is a somewhat simplified summary of main characteristics of the individual systems of passive characteristics in the languages we have been dealing with: (82) Summary of typological and passive characteristics OV Case IP expSu D/IO-c i GE + + + – – ii IC – + + – – iii FA – + + – + iv DU + – + + + v FR + – + + – vi EN – – – + + vii DA – – + + + NO-RM/BM – – + + + viii NO-NN – – + + + SW – – + + +

PP-c – – – – – + + + + +

Sup + – – + + + + + – –

sPM – – + – – – + + + +

The present-day Germanic standard languages show considerable variation in the domain of passive constructions. Within the not overly comprehensive frame of reference provided by the criteria in (82), altogether 10 languages and language varieties form as much as eight different groups; when more criteria are taken into consideration, such as for instance the form of the expletive, the degree to which synthetic passive morphology is used, or the additional passive morphology options in note 20, even the two groups (vii) and (viii), each of which comprise two different languages, fall apart. 4.1 Typological aspects of Germanic passive verb morphology Within the domain of passive verb morphology, there are three main groups of languages: (i) Icelandic and English are the two most conservative languages, sticking to ‘be’, which has been there there from the very early beginnings of the Germanic analytic passive. (ii) German, Dutch and Frisian have as aspectually neutral passive auxiliary a desemanticized transformative (ingressive) copula ‘become’ of indigenous origin. (iii) The Scandinavian languages, with the exception Icelandic, have as aspectually neutral passive auxiliary a desemanticized transformative

Germanic Passive Constructions

103

(ingressive) copula ‘become’ of foreign provenance; Faroese also has desemanticized ’become’ of indigenous origin. In addition, these languages have synthetic passive morphology, but its distribution in the individual languages form an unmistakable geographical cline, reflecting its area of origin, presumably in a language contact situation. In a Northern area comprising Swedish, Norwegian Nynorsk, and Faroese, the non-finite form of the analytic passive is an agreeing participle, as in the historically basic copula construction.21 In a far more extensive “southern” area comprising all the other languages, non-agreeing, supine forms reign supreme. From a typological point of view, one may note that analytic, periphrastic passive constructions are in keeping with a long-term development tendency in Germanic from syntheticity to analyticity. In German, the loss of agreement in participles, i.e. transition to non-agreeing supine constructions, may be seen as a kind of phrase-internal economization of basically the same kind that manifests itself in the mono- or biflexional morphological group inflection of German NPs (cf. Werner 1979; Ronneberger-Sibold 1997). The retention of participle agreement is typologically natural in Icelandic and Faroese, given the general morphological profile of these two languages. In Swedish and Norwegian Nynorsk, on the other hand, the longevity of participle agreement can be considered a case of morphosyntactic persistence (Hopper 1990). 4.2 Typological aspects of Germanic conversion differences In the domain of conversions, characteristic differences exist in both the personal and the impersonal passive. These differences appear to be related to general typological differences of the kind presented in the adaptation of Greenberg’s ‘Universal 41’ (Greenberg 1966) to modern Germanic given in (83) (cf. Askedal 2009: 11): (83) Verb order–case marking correlations in Modern Germanic i GE +V-final, +NP-case ii DU, FR +V-final, –NP-case iii EN, DA, SW, NO –V-final, –NP-case iv IC, FA –V-final, +NP-case

In the OV and/or case languages – German, Icelandic, Faroese, Dutch, West Frisian – non-subject promotion is to a varying extent subject to 21

The largest class of weak verbs in Nynorsk does not allow agreement in past participles. In Swedish and Faroese, agreement in past participles is more pervasive.

104

John Ole ASKEDAL

restrictions. None of these languages allow promotion out of PP. In Dutch and West Frisian the main tendency is to restrict non-subject promotion to direct objects even though this is not motivated by morphological restrictions due to surface case marking. Exceptions are found in Dutch, where some indirect objects are promoted. On the other hand, the morphologically motivated restriction of Icelandic and German against promoting dative objects is in certain instances dispensed with in Faroese. The other languages – practically VO languages without NP-case marking – allow non-subject promotion out of all the positions in the category hierarchy in (1a, iii). It appears that the syntax of the OV languages Dutch and Frisian has not yet liberated itself from the restrictions once caused by case morphology; since the case morphology in question no longer exists, it appears likely that former morphological and lexical restrictions now to a certain extent persist as lexical restrictions without a morphological basis. In Dutch and Frisian, the lexicon is thus more syntactically conservative than the inflectional morphology. In Faroese, on the other hand, the promotion of dative objects to subject indicates that syntax is more progressive than morphology. This can be seen as an approximation to the freer use of non-subject promotion in the non-NP-case Mainland Scandinavian languages. A couple of decades ago the concept of ‘configurationality’ was introduced into morphosyntactic theory (cf. e.g. Faarlund 1990). In typological parlance this may translate into ‘topological syntax’, meaning that syntactic relations and rules are systematically related to fixed positions in linear structure as a result of ‘grammaticalization of word order’. Arguably, this is exactly what has happened in Mainland Scandinavian, as has been pointed out by several authors (Diderichsen 1941; Heltoft 2001). In Mainland Scandinavian, this development is seen in connection with the demise of case-marking morphology. There are, however, clear indications that Insular Scandinavian is part of the same development. This is evident from the notorious so-called ‘oblique subjects’ of Icelandic (Thráinsson 2007: 159–167) and Faroese (Barnes 1986; Thráinsson et al. 2004: 252–257) that have for the last couple of decades been an important part of syntactic thinking about these languages. ‘Oblique subjects’ are NPs in the dative, accusative or, marginally, genitive that occupy the same position in the clause and possess the same syntactic control properties as nominative subjects. From a comparative perspective, it is a part of the picture that the Icelandic nominative and oblique subjects have the same linear distribution as Mainland Scandinavian non-case-marked subjects. In addition, the ordering of indirect and directs objects with respect to each other and to the verbal predicate is also basically the same in Insular as well as Mainland Scandinavian.

Germanic Passive Constructions

105

Both Icelandic and Faroese have oblique subjects, but only Faroese allows for promotion of a dative object to subject. This indicates that Faroese has progressed somewhat farther on the path towards a topological, ‘configurational’ syntax than Icelandic. Tightly structured VPs and, concomitantly, ‘topological syntax’ is more likely to develop within a general VO rather than an OV structure, probably because the former provides better topological demarcation possibilities for the syntactically important subject category than the latter. In this context, we may again refer to the curious fact that in Dutch and West Frisian, certain formerly dative-marked non-subjects remain as such in the passive although they are located at the beginning of the clause, in a position that is otherwise typical of subjects. On the other hand, modern Mainland Scandinavian has impersonal passive constructions with an expletive subject and a retained direct object that, with regard to syntactic rule properties, continues to behave like a true object. This is of course also understandable from a topological syntactic point of view. Within the ‘configurationality’ conception, non-occurrence of expletive subjects was considered a defining property of a non-configurational syntactic system. In this regard, German behaves according to expectations, but the ‘oblique subjects’ of Icelandic and Faroese rather point in the direction of ‘configurationality’ or ‘topological syntax’. Among the non-NPcase-marking languages, Mainland Scandinavian has the formal subjects we expect it to have. Dutch and West Frisian show occasional clause internal expletive subjects and thus seem to occupy an intermediate position between German and Mainland Scandinavian that corresponds to their Janus-like typological character as non-NP-case marking OV languages. The modern Mainland Scandinavian languages represent a configurational syntactic type characterized by, among other things, the properties listed in (84): (84) Mainland Scandinavian configurationality i categorially wide operational domain of non-subject promotion ii impersonal passive iii obligatory expletive subjects in the impersonal passive and in lexical impersonal constructions iv indefinite VP-internal NPs with object-like syntactic rule properties

English represents another variety of the general configurational type, displaying the properties in (85):

106

John Ole ASKEDAL (85) English configurationality i categorially wide operational domain of non-subject promotion ii no impersonal passive iii obligatory expletive subjects in lexical impersonal constructions iv VP-internal NPs in constructions with the expletive there, where the finite verb agrees with the indefinite NP

A general conclusion is that the present-day Germanic languages vary with regard to the structural characteristics of their passive constructions and that the variation is non-random in the sense that it may be correlated with general typological properties of the individual languages. REFERENCES Anderson, S. R. 1990. “The Grammar of Icelandic Verbs in -st”. Modern Icelandic Syntax [Syntax and Semantics 24]. J. Maling & A. Zaenen (eds.). San Diego: Academic Press. 235–273. Askedal, J. O. 1980. “Über das Passiv von Verben mit zwei Akkusativergänzungen im Deutschen”. Festschrift für Gunnar Bech. Zum 60. Geburtstag am 23. März 1980 [Kopenhagener Beiträge zur Germanistischen Linguistik, Sonderband 1]. M. Dyhr, K. Hyldgaard-Jensen & J. Olsen (eds.). Kopenhagen: Institut für Germanische Philologie, Universiät Kopenhagen. 1–18. 1993. “Configurationality in language typology and diachronic syntax: Evidence from Germanic”. Norsk Lingvistisk Tidsskrift 11. 125–134. 1997. “Typological and Semiotic Aspects of Certain Morphosyntactic Differences Between Norwegian and Swedish”. Interdisciplinary Journal for Germanic Linguistics and Semiotic Analysis 2. 187–220. 2001. “‘Oblique subjects’, structural and lexical case marking: Some thoughts on case assignment in North Germanic and German”. Grammatical Relations in Change [Studies in Language Companion series 56]. J. T. Faarlund (ed.). Amsterdam and Philadelphia: Benjamins, 65–97. 2009. “Some General Evolutionary and Typological Characteristics of the Germanic Languages”. Germanic Languages and Linguistic Universals [The Development of the Anglo-Saxon Language and Linguistic Universals 1]. J. O. Askedal, I. Roberts, T. Matsushita & H. Hasegawa (eds.). 2009. Amsterdam and Philadelphia: Benjamins. 7–56. Barnes, M. 1986. “Subject, nominative and oblique case in Faroese”. Scripta Islandica 38. 13–46.

Germanic Passive Constructions

107

Bech, G. 1955. Studien über das deutsche verbum infinitum. Vol. 1. Kopenhagen: Munksgaard. Castillo, C. 2009. “On There-Passive Participle Constructions”. Atlantis. Journal of the Spanish Association of Anglo-American Studies 31. 91–110. Dahl, Ö. & M. Koptjevskaja-Tamm. 1992. Language typology around the Baltic sea: A problem inventory [Papers from the Institute of Linguistics, University of Stockholm 61]. Stockholm: University of Stockholm, Institute of Linguistics. Diderichsen, P. 1941. Sætningsbygningen i Skaanske Lov. Fremstillet som Grundlag for en rationel dansk Syntaks. København: Munksgaard. Duden. 2005. Die Grammatik [Duden Band 4]. Dudenredaktion (eds.). Mannheim etc.: Dudenverlag. Faarlund, J. T. 1990. Syntactic Change: Toward a Theory of Historical Syntax. The Hague: Mouton. , S. Lie, & K. I. Vannebo. 1997. Norsk referansegrammatikk. Oslo: Universitetsforlaget. Greenberg, J. H. 1966. “Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements.” Universals of Language. 2nd Ed. J. H. Greenberg (ed.). Cambridge, MA and London: M.I.T. Press. 1966. 73–113. Haeseryn, W., K. Romijn, G. Geerts, J. de Rooij, & M. C. van den Toorn. 1997. Algemene Nederlandse Spraakkunst. Vol. 1–2. Groningen: Martijnus Nijhoff, and Deurne: Wolters Plantyn. Hansen, A. 1967. Moderne Dansk. III: Sprogbeskrivelse. Udgivet af Det Danske Sprog- og Litteraturselskab. København: Grafisk Forlag. Harbert, W. 2007. The Germanic Languages [Cambridge Language Surveys]. Cambridge: Cambridge University Press. Heltoft, L. 2001. “Recasting Danish subjects: Case system, word order and subject development”. Grammatical Relations in Change [Studies in Language Companion Series 56]. J. T. Faarlund (ed.). Amsterdam and Philadelphia: Benjamins, 171–204. & L. F. Jakobsen. 1996. “Danish passives and subject positions as a mood system”. Content, Expression and Structure: Studies in Danish functional grammar [Studies in Language Companion Series 29]. Elisabeth Engberg-Pedersen, M. Fortescue, Harder, L. H. & L. F. Jakobsen (eds.). Amsterdam: John Benjamins. 199–234. Hopper, P. J. 1990. “Principles of Grammaticalization: Towards a Diachronic Typology”. Language Typology 1987: Systematic Balance in Language. Papers from the Linguistic Typology Symposium, Berkeley, 1–3 December 1987 [Current Issues in Linguistic Theory 67]. W. P. Lehmann (ed.). 1990. Amsterdam and Philadelphia: Benjamins. 157–170.

108

John Ole ASKEDAL

Johnson, D. E. 1977. “Relational Constraints on Grammars”. Relational Grammar [Syntax and Semantics 8]. P. Cole & J. M. Sadock (eds.). New York: Academic Press. 151–178. Keenan, E. L. & M. S. Dryer. 2007. “Passive in the World’s Languages.” Language Typology and Syntactic Description. Volume 1: Clause Structure. 2nd ed. T. Shopen (ed.). Cambridge: Cambridge University Press. 325–361. & B. Comrie. 1987. “Noun Phrase Accessibility and Universal Grammar”. Universal Grammar: 15 Essays. E. L. Keenan. London etc.: Croom Helm. 3–45. Markey, T. L. 1969. The Verbs varda and bliva in Scandinavian: With Special Emphasis on Swedish [Studia philologiae Scandinaviae Upsaliensia 7]. Uppsala: Uppsala universitet, and Stockholm: Almqvist & Wiksell. Nübling, D. 2006. “Auf Umwegen zum Passivauxiliar – Die Grammatikalisierungspfade von geben, werden, kommen und bleiben im Luxemburgischen, Deutschen und Schwedischen”. Perspektiven einer linguistischen Luxemburgistik: Studien zu Diachronie und Synchronie. C. Moulin & D. Nübling (eds.). Heidelberg: Winter. 171–201. Plank, F. 1987. “Direkte indirekte Objekte, oder: was uns lehren lehrt”. Leuvense Bijdragen 73. 37–61. Quirk, R., S. Greenbaum, G. Leech, & J. Svartvik. 1991. A Comprehensive Grammar of the English Language. Tenth Impression. London and New York: Longman. Ronneberger-Sibold, E. 1997. “Typology and the diachronic evolution of German morphosyntax”. Linguistic Reconstruction and Typology. J. Fisiak (ed.). Berlin and New York: Mouton de Gruyter, 313–335. Ross, J. R. 1974. “There, there, (there, (there, ...))”. Proceedings of the Tenth Regional Meeting of the Chicago Linguistic Society. M. LaGaly, R. Fox, A. Bruch et al. Chicago: Chicago Linguistic Society, University of Chicago. 569–587. Teleman, U., S. Hellberg, E. Andersson et al. 1999. Svenska Akademiens grammatik. Vol. 1–4. Stockholm: Norstedts. Tesnière, L. 1966. Éléments de syntaxe structurale. 2. Éd. Paris: Klincksieck. Thráinsson, H. 2007. The Syntax of Icelandic [Cambridge Syntax Guides]. Cambridge: Cambridge University Press. , H. P. Petersen, J. í L. Jacobsen & Z. S. Hansen. 2004. Faroese: An overview and reference grammar. Tórshavn: Føroya Fróðskaparfelag. Tiersma, P. M. 1985. Frisian Reference Grammar. Dordrecht, Holland and Cinnaminson, USA: Foris. Werner, O. 1979. “Kongruenz wird zu Diskontinuität im Deutschen”. Festschrift für Oswald Szemerenyi on the Occasion of his 65th Birthday.

Germanic Passive Constructions

109

B. Broganyi (ed.). Amsterdam: Benjamins, 959–988. Zifonun, G., L. Hoffmann, B. Strecker et al. 1997. Grammatik der deutschen Sprache. Vol. 1–3 [Schriften des Instituts für deutsche Sprache 7.1–7.3]. Berlin and New York: Walter de Gruyter. ABBREVIATIONS Languages DA Danish DU Dutch EN English FA Faroese FR (Modern West) Frisian GE German IC Icelandic NO Norwegian NO-NN Norwegian Nynorsk NO-RM/BM Norwegian Riksmål/Bokmål SW Swedish Grammatical categories ACC accusative ACT active (voice) Adv-PP adverbial PP AUX auxiliary Case case-marking in non-pronominal NPs COP copula (verb) DAT dative DEF definite DO, DO direct object expSu expletive subject FEM feminine indirect object IO, IO IP impersonal passive INF infinitive INTRANSF intransformative (continuative) MASC masculine NOM nominative neut. neuter NP noun phrase OBL oblique (non-subject) case OV V-final order

110

John Ole ASKEDAL

V-final

(agreeing) participle passive (voice) plural prepositional object prepositional phrase present preterit singular synthetic (suffixal) passive morphology stative (statal) subject (non-agreeing) supine transformative (ingressive) verb-final

1, 2, 3

1st, 2nd, 3rd person

PART PASS PL

PO PP PRES PRET SG

sPM STAT

Su, SU SUP TRANSF

Prosodic Constraints on Old English Alliteration Yasuaki FUJIWARA 0. INTRODUCTION Alliteration is generally defined as repetition of homophonous syllable-initial phonemes, as in house and home, cash and carry, tea for two (Bussman 1996: 16). However, this definition is insufficient, for it seems to presuppose only maximum stresses on the word level as far as the examples are concerned. Indeed, we sometimes find words like mílkman, Máidment, lívely, néwness with reduced stresses in the second syllables, to say nothing of words like bláckbìrd, lífelìne, píckpòcket with subsidiary stresses on the second elements. The definition in question cannot distinguish fortuitous coincidence of sounds from intentional ones, and admits otherwise wrongly recognized vocalic alliteration. A precise definition of alliteration is crucial for metrical works such as OE poems, for most of them are composed based on the strict metrical rules. Indeed, we sometimes come across cases in Old English poetry where we question whether given words or phrases receive enough degree of stress for alliteration. For example, it is quite uncertain whether we may regard words like lóf-læ ¯ can ‘praise’, lúf-līce ‘kindly’, mídd-mest ‘in the very center’ with unstressed suffixes as instances of double alliteration. Hence, we will address ourselves to investigate the relationship between the degree of stress and alliteration and clarify what motivates alliteration in OE poetry. 1. DISTINCTION BETWEEN COMPOUND WORDS AND COMPLEX WORDS 1.1 Definition of Compound Words and Complex Words In this paper, compound words are defined as two or more independent words joined together as one, characterized by a trochaic stress pattern, i.e., primary stress followed by allegedly secondary one (Halle and Keyser (1971: 87-97)). Hence, not only words like éald-wìta ‘priest’, nórþan-wìnd ` tan ‘forsake’, ‘north wind’, hrē´ow-cèarig ‘sad’ but also words like ā´ n-for-l¯æ hónd-ge-wìnn ‘combat’, ín-wit-ðànc ‘evil thought’ are regarded as compound words, for the first or second elements (e.g., for-læ ¯´ tan, ge-wínn, and ín-wìt) are as a whole independent words although they include bound forms.

112

Yasuaki FUJIWARA

On the other hand, words like ún-rī`m ‘countless number’, mís-ge-hỳgd ´ n-nes ‘purity’ together with words ‘evil thought’, fē´ond-scipe ‘enmity’, cl¯æ like ofer-wádan ‘wade across’, under-gā´ n ‘undergo’, or ymb-hríngan ‘surround’ fall under the group of complex words, for one of their elements is made up of a bound form. Hence, complex words are characterized by either iambic or trochaic stress pattern. Thus, compound words and complex words in Old English are contrastive both morphologically and phonologically. 1.2 The Relationship between Reduced Stresses and Alliteration A unique trochaic stress pattern of OE compound words is made more prominent by virtue of alliteration, because it is usually realized on their first elements with primary stress. The same holds good for OE complex words like ún-rī`m with trochaic stress pattern, for alliteration is regularly realized on such prefixes but not on the stem syllables. This may lead to an assumption that OE alliteration is confined to the syllables with maximum lexical stresses, while secondary stresses on the second elements of OE compound and complex words are not relevant to alliteration. However, we sometimes find instances like cwéalm-cùma ‘murderous comer’, wóruld-wèla ‘worldly possession’, mís-mìcel ‘of varying sizes’, sín-sòrg ‘of constant sorrow’ in which two pretonic consonants of both elements of OE compound or complex words are identical and hence double alliteration seems to be realized. Hence, we must offer a definite reason for such type of alliteration if we claim secondary lexical stresses actually also contribute to alliteration. On the other hand, there are many phrases of the type “an adjective + a noun” or “a noun + an adjective” like grìm gū´ ð-plega ‘fierce attack’, ` ger ‘curtain beautiful>beautiful curtain’ in which adjectives seem flē´oh-net fæ to form double alliteration together with nouns. Because such cases reduce the stresses on adjectives when they modify noun phrases, we must offer an obvious reason for the realization of reduced phrasal stresses. 2. CONDITIONS FOR DOUBLE ALLITERATION As the degree of lexical stress may be different between the second elements of compound or complex words and adjectives modifying nouns as a result of stress subordination, we will first examine the principles of stress subordination. Take noun phrases of the type “an adjective + a noun” in Old English for instance. Old English abounds in these types of noun phrases in the same way as Present-day English in which phrases like blàck bírd, blàck bóard, dàncing gírl show a clear contrast in meaning and stress pattern with corresponding compound words like bláck-bìrd, bláck-bòard, dáncing-

Prosodic Constraints on Old English Alliteration

113

gìrl. Hence, OE noun phrases and compound words made of an adjective and a noun offer the same type of contrast. 2.1 Stress Subordination in OE Phrases First, we will investigate the examples of (1) taken from Beowulf. Hereafter, in the examples alliterating consonants are indicated in italics; “a” and “b” after numerals stand for the first and the second half-line, respectively, of a quoted long-line; and the boundary of elements is hyphenated for the sake of clarity. In noun phrases like (1a) composed of an adjective plus a noun, it is regularly adjectives that take part in alliteration both in a- and b-verses, while in noun phrases like (1b) with reverse word order, it is always nouns that take part in alliteration. In addition, in verses like (1c), regardless of the order of an adjective and a noun, both words of the phrase are relevant to alliteration. Hence, we may naturally infer from these examples the indisputable fact that stress subordination, though it may really occur in the same way as in Present-day English, does not have any effect on the realization of alliteration. That is, weakened stresses on adjectives are still relevant to alliteration. (1)

a. b. c.

snotere ceorlas ‘wise men’ (416b), dēoran sweorde ‘with bold sword’, (561a), wunden gold ‘twisted gold’ (1193b) duguð unly¯tel ‘no little tried warriors’ (498a), lēoht unfæger ‘terrible light’ (727b), byrnan sīde ‘broad corselet’ (1291a) þes hearda hēap ‘this strong band’ (432a), bēah-sele beorhta ‘the bright ring-hall’ (1177a), geō-sceaft grimme ‘grim fate’ (1234a)

2.2 Stress Subordination between the Two Elements of OE Compound Words Another question is involved in the possibility of alliteration of the weakened second elements of compound words. We will examine the instances of (2) taken from Andreas. All instances of (2a) are composed of a noun plus a noun and those of (2b) an adjective plus a noun. Regardless of the difference of lexical categories of the two compound words, the first elements always take part in alliteration. Hence, at least as far as these instances are concerned, it is natural to assume that the stresses on the first elements are stronger than those on the second ones. However, in the compound words of (2c), double alliteration seems to be present in both elements. Hence, we will have to offer a strong reason if we deny an accidental coincidence of the two pretonic consonants and admit double alliteration between the two elements in the proper sense of the term.

114

Yasuaki FUJIWARA (2)

a. b. c.

brégo-stō` l ‘princely stool’ (209a), héofon-càndel ‘lamp of the sky, i.e. the sun’ (243b), hólm-wèarde ‘ocean guardian’ (359b) fā´ mig-hèals ‘with foaming prow’ (497a), hē´ ah-fæ ` dera ‘patriarch’ (791a), ´ fēor-wègas ‘remote journey’ (928a) néaro-nē` dum ‘cruel bond’ (102a), héofon-hā` lig ‘of heavenly holiness’ (728a), wél-wànge ‘field of slaughter’ (1226a)

3. DISTRIBUTION OF DOUBLE ALLITERATION BETWEEN THE TWO ELEMENTS OF COMPOUND WORDS 3.1 Groups and Frequencies of Half-Lines in Beowulf As the point at issue has been fully discussed in Fujiwara (1990: 183-192), a brief outline of the investigation will suffice here. In this analysis, we took all long-lines from Beowulf that consist of two to four words with primary linguistic stresses and divided them into four groups according to the number and distribution of secondary stresses in half-lines: Group A (′ ′ |′ ′ ), Group B (′ ′ |′ ` ), Group C (′ ` |′ ′ ), and Group D (′ ` |′ ` ). A vertical line “|” in the groups indicates the boundary of the two half-lines in a given long-line. Next, we investigated the frequency of instances of each of the four groups, of each of the three alliterative types, i.e., double alliteration, single alliteration, and extra alliteration. As a result, we found 1,170 instances of long-lines in Group A in which double alliteration is realized in 617 lines (=55.74%), as shown in (3a), while single alliteration occurs in 490 lines (=44.26%), as shown in (3b). As for the extra alliteration, it is found in only 29 lines (i. e. 29/490=5.92%), as indicated in (3c). Hereafter, a single slash “/” indicates a boundary between the two half-lines, and double slashes “//” the end of a long-line. (3)

a. b. c.

f y¯r ond fæstor / sē þæ ¯ m fēonde ætwand // “he who escaped the fiend (kept himself) farther off and more secure” (143) Bēowulf maþelode, / bearn Ecgþēowes // “Beowulf, son of Ecgtheow, spoke” (529) sigora Waldend, / þæt hē hyne sylfne gewræc // “wielder of victories, (granted him) that he might avenge himself ” (2875)

On the other hand, in the instances of Group B like (4a, b) and also in those of Group C like (4c) in which either one of the two elements includes secondary stress, the relation between secondary stress and alliteration becomes very clear. For example, 125 lines in Beowulf belong to Group B. However, the frequency of double alliteration and single alliteration is nearly the same as in Group A, that is, double alliteration is realized in 77 instances

Prosodic Constraints on Old English Alliteration

115

of the total 125 lines (=61.6%), while single alliteration is found in 48 lines (=38.4%). One possible assumption for this result is because elements with secondary stresses are not included in a-verses. In fact, of 48 instances, we find only one example of (4c) in which secondary stress seems to be relevant to extra alliteration. We may therefore regard this as an accidental coincidence of the two pretonic consonants. (4)

a.

b. c.

Þā wæs eft swā æ ¯ r / ellen-rōfum// “Then again, as before, (a feast) was (prepared) for the courageous warriors” (1787) æt nīða ge-hwām / ny¯d-ge-steallan // “in any struggle comrades in need” (882) ¯ r æt hy¯ðe stōd / hringed-stefna // Þæ “There, at the harbor, the ship with a ringed stem stood” (32)

Next, we will investigate the relationship between secondary stress and alliteration in lines of Group C where secondary stress appears in one element in a-verses. Among the total 452 instances of this group, double alliteration is found in only 21 cases (=4.65%), as shown in (5a). In inverse proportion to this, 431 examples (=95.35%) of single alliteration outweigh double alliteration in number, as indicated in (5b). On the other hand, extra alliteration is found in only 35 examples (i.e. 35/452=8.1%), as shown in (5c). Thus, it is evident that in Group C secondary stress in a-verses as a substitute for primary stress causes a hindrance to double alliteration in Group C in the same way as in Group B. (5)

a. b. c.

geond wīd-wegas / wuldor scēawian // “throughout the wide ways to see the wonder” (840) ¯ ded // of feor-wegum / frætwa ge-læ “from distant lands jewels (were) brought” (37) æfter cear-wælmum / cōlran weorðað // “on account of welling sorrow becomes cooler” (2066)

Finally, we will take up Group D where secondary stress replaces primary stress in both half-lines. The group at issue exhibits only 27 instances, and their metrical significance can readily be apparent, for double alliteration is realized in only 3 cases (=11.11%), as shown in (6a). Conversely, there is a high frequency of single alliteration, 24 examples (=88.89%), as shown in (6b), which include three examples of extra alliteration (i.e. 3/24=12.5%), as shown in (6c).

116

Yasuaki FUJIWARA (6)

a. b. c.

heard-hicgende / hilde-mecgas // “the brave-minded warriors” (799) Heaðo-Scilfingas / heals-gebedda // “the dear consort of the warlike Scylfing” (63) Hwæt, wē Gār-Dena / in gēar-dagum // “Lo! We (have heard) of the Spear-Danes in days of yore” (1)

3.2 Summary of the Analysis and Residual Problems Summarizing the results of the preceding analyses of the relationship between the degree of stress and alliteration, we can argue that the frequency of double alliteration and extra alliteration is greatly influenced by the degree of stress; that is, the rate of these two types of alliteration decreases sharply when secondary stress is substituted for primary stress. Then, a question arises on how we should explain a number of examples such as (6a) in which secondary stresses seem to motivate behind alliteration. Hence, if we are to disregard such instances as mere exceptions on account of their accidental coincidence of the two pretonic consonants between the elements of a given compound word, we will have to offer a clear argument for it. Below is a summary of the claim made by Fujiwara (1990: 190-1) on such instances. There are 19 consonant phonemes (i.e. /p b t d k g tR dY m n r l f S s R h w j/) which occupy pretonic positions in OE words. To these may well be added a glottal stop [>] for the instances without pretonic initial consonants which are wrongly referred to as vocalic alliteration (Fujiwara 1990: 64). Hence, there are 20 consonants in all that are relevant to alliteration in OE poetry. As no graphic device is known for indicating the glottal stop in question, stressed vowels immediately following this consonant will be indicated in this paper in italics for convenience in this article. Accordingly, the probability of any two pretonic consonants to be identical in Old English poetry is in the ratio of one to twenty, or a frequency of 5%. OE alliteration is apparently a result of the poets’ efforts to adjust the quality of two or three pretonic consonants of words or phrases in each given line. Hence, identical pretonic consonants occurring in a frequency less than 5% cannot be a consequence of a poet’s effort but a mere fortuitous coincidence that may well occur in prose works, too. Hence, such identical pretonic consonants are not qualified for alliteration in the true sense of the term. The frequency of double alliteration and extra alliteration, too, as shown in (4), (5) and (6), in which secondary stress seems to be involved in alliteration, hardly exceeds accidental possibility. Hence, at least as far as the results of the preceding analyses are concerned, we may conclude that secondary stress is irrelevant to alliteration as far as Old English is concerned.

Prosodic Constraints on Old English Alliteration

117

4. DISTRIBUTION OF COMPOUND WORDS IN DOUBLE ALLITERATION 4.1 Problems of Fujiwara (1990) Although the conclusion arrived at in the previous section is based on sufficient data and persuasive argument, the analysis of Fujiwara (1990) is defective in that it overlooked metrical differences between a- and b-verses, especially the peculiarity of distribution of the three types of alliteration; that is, double alliteration is confined to a-verses, while single alliteration is found both in a-verses and b-verses. Second, the frequency of double alliteration greatly exceeds that of single alliteration. Third, extra alliteration presupposes single alliteration and is not an independent alliterative type on its own right. Thus, the following investigation will be focused on the distribution of alliterating compound words in a- and b-verses. 4.2 The Frequency and Distribution of Compound Words First, we will start our investigation by extracting as many compound words as possible from OE poems, which is however not exhaustive due to lack of time and space. We will pick up 12 OE poems listed in (7), which are selected at random irrespective of their scale or theme lest any great deviation may result in the data to be obtained. As the boundary between compound words and phrases may sometimes be opaque, we will basically follow the judgement of the boundary made by the editor of each text. Proper nouns are excluded from analysis because their internal morphemic structure is not always explicit and the boundary of elements is sometimes vague. (7)

a. b. c. d. e. f. g.

Andreas (771 instances in 1722 lines: 499 instances in a-verse, 272 instances in b-verse) Beowulf (1471 instances in 3182 lines: 1028 instances in a-verse, 443 instances in b-verse) Cadmon’s Hymn (6 instances in 9 lines: 2 instances in a-verse, 4 instances in b-verse) Deor (8 instances in 42 lines: 7 instances in a-verse, 1 instance in b-verse) The Dream of the Rood (43 instances in 156 lines: 27 instances in a-verse, 16 instances in b-verse) Elene (475 instances in 1321 lines: 325 instances in a-verse, 150 instances in b-verse) The Fight at Finnsburg (16 instances in 48 lines: 8 instances in a-verse, 8 instances in b-verse)

118

Yasuaki FUJIWARA h. i. j. k. l.

Juliana (241 instances in 731 lines: 167 instances in a-verse, 74 instances in b-verse) The Phoenix (204 instances in 677 lines: 133 instances in a-verse, 71 instances in b-verse) The Riming Poem (36 instances in 87 lines: 22 instances in a-verse, 14 instances in b-verse) Waldere (15 instances in 63 lines: 11 instances in a-verse, 4 instances in b-verse) Widsið (23 instances in 143 lines: 15 instances in a-verse, 8 instances in b-verse)

What can be deduced from the frequency and distribution of compound words in (7) is that compound words are used in the ratio of one to three lines on average, and that the frequency of their occurrence in a-verses is about twice as often as that in b-verses. Especially, the exceedingly high frequency of compound words in a-verses will surely be an important clue for us to know the relaionship between compound words and alliteration, because double alliteration is allowed only in a-verses. Hence, our investigation will naturally be focused on the distribution of all compound words in 12 OE poems. 4.3 Distribution of Alliterating Compound Words in Andreas First, we will extract all compound words used in Andreas whose pretonic consonants of the first and second element are identical, as listed in (8). Surprisingly, all 23 instances at issue occur in a-verses, and no single instance is attested in b-verses. (8)

bān-ge-brec ‘breaking of bones’ (1442a), clūstor-cleofan ‘prison-house’ (1021a), efen-ealdum ‘of like age’ (553a), ferð-ge-fēonde ‘joyful in heart’ (915a, 1584a), forht-ferð ‘terrified at heart’ (1549a, 1596a), geōmor-gidd ‘song of sorrow’ (1548a), hand-hrine ‘touch of the hand’ (1000a), heofon-hālig ‘of heavenly holiness’ (728a), heofon-hwealfe ‘vault of the sky’ (545a, 1402a), lago-lāde ‘water-way’ (314a), lyft-ge-lāc ‘motion through the air’ (827a, 1552a), mōd-ge-mynd ‘memory’ (688a), nearo-nēdum ‘cruel bondage’ (102a), sār-slegum ‘painful blow’ (1275a), waruð-ge-win ‘moving surf ’ (439a), wæl-wulfas ‘wolf of slaughter’ (149a), wel-wange ‘field of slaughter’ (1226a), winter-ge-worpum ‘snow-fall’ (1256a), wundor-worca ‘miracle’ (705a)

4.4 Distribution of Alliterating Compound Words in Beowulf As an accidental coincidence may lurk in the examples in (8), other examples will have to be taken from several other poems. Hence, we first

Prosodic Constraints on Old English Alliteration

119

picked up 1471 compound words from Beowulf and found that there are 28 instances of 24 compound words out of the totality in which consonants of the two elements are identical, as shown in (9). (9)

bearn-ge-byrdo ‘child-bearing’ (946a), bry¯d-būre ‘woman’s room’ (921a), cwealm-cuman ‘murderous comer’ (792a), dēað-dæge ‘death-day’ (187a, 885a), eall-īrenne ‘all-iron’ (2338a), fela-fricgende ‘learning much’ (2106a), fen-freoðo ‘fen-refuge’ (851a), ferhð-frecan ‘bold in mind’ (1146a), geō-sceaft-gāsta ‘fated spirit’ (1266a), giōmor-gyd ‘mournful song’ (3150a), gold-gyfan ‘gold-giver’ (2652a), gryre-geatwum ‘warlike trappings’ (324a), gryre-gieste ‘terrible stranger’ (2560a), heard-hicgende ‘brave-minded’ (394a, 799a), heoro-hōcyhtum ‘sword-hooked’ (1438a), hilde-hlæmmum ‘battle-crash’ (2201a, 2351a, 2544a), ¯ -sīðe ‘sea-journey’ (1149a), s(c)yn-scaþa mil-ge-mearces ‘mile-mark’ (1362a), sæ ‘hostile demon’ (707a), swāt-swaðu ‘blood-track’ (2946a), þēod-þrēaum ‘great calamity’ (178a), wīd-wegas ‘distant regions’ (840a, 1704a), wīg-ge-weorþad ‘distinguished in battle’ (1783a), wīg-weorþunga ‘honour to idols’ (176a)

Several interesting facts can be brought to light from the examples in (9). First, all 28 instances of compound words in (9) are confined to a-verses in exactly the same way as in Andreas. Hence, it is most likely that the strict restriction on the distribution of the compound words in question in (8) and (9) is not merely accidental nor resulted from the poets’ preference, but is based on some linguistic or poetic property concerning such compound words that was known to the poets in those days. Specifically, it seems that the poets regarded not only the primary stresses on the first elements but the subsidiary linguistic stresses on the second elements of compound words as properly involved in alliteration. Accordingly, we may assume that double alliteration realized in the two elements of compound words was as legitimate as that in the two primary linguistic stresses of two separate words. Second, considering the size of the texts of Andreas and Beowulf, the ratio of occurrence of alliterating compound words is comparatively low in Beowulf. We regard this interesting feature of the poem as a reflection on the part of the poet that he did not favor so much this type of compound words as the Andreas poet did. Third, because most alliterating compound words in (8) and (9) do not overlap in the two poems, it is likely that the poets resorted to “nonce words” in addition to easily available noun phrases composed of an adjective and a noun in order to realize double alliteration. Fourth, in case of the compound word geō-sceaft-gāsta composed of

120

Yasuaki FUJIWARA

three words and supposed to have an internal structure [[geō+sceaft]gāsta], it has the stress pattern 1-3-2, the degree of stress being indicated by numerical figures 1, 2, 3 in the order of strength. The vowel of sceaft undergoes stress subordination twice and its stress is reduced to tertiary, while the stem vowel of gāsta is subject to stress subordination just once and hence retains secondary stress which seems to be strong enough for alliteration. On the other hand, in case of in-wit-wrāsne [[in-wit]wrāsne] ‘cruel bondage’ used twice in Andreas (63b, 946b) with the same stress pattern 1-3-2 as geō-sceaft-gāsta, it seems that double alliteration cannot be realized between wit and wrāsne, for the stress on wit is weakened to tertiary by the twice application of stress subordination. Hence, our assumption that secondary stress of compound words is metrically strong enough for alliteration while tertiary is not is supported by the fact that the two instances of in-wit-wrāsne are confined to b-verses in which double alliteration is not allowed. 4.5 Distribution of Alliterating Compound Words in Other Poems As a further confirmation of our hypothesis, we can exemplify instances of alliterating compound words from several OE poems other than Andreas and Beowulf, as shown in (10). According to our expectations, all instances of compound words in (10) are found only in a-verses. Most of these are “nonce words”, which means the poets tried to combine two words into one in order to realize double alliteration instead of adding adjectives immediately before nouns. Also worthy of note is the fact that in The Riming Poem (which is well known as a most artful one in all OE poems and many compound words are used as nonce words), there is not a single instance of compound words in which double alliteration is realized. Instead, double alliteration in the poem is found only in the phrases made up of two independent words. However, at present we cannot explain the reason why these peculiarities are found in the poem. (10) a.

b. c.

d.

Elene (7 instances): fyrn-ge-flit ‘old stroke’ (903a), gold-gimmas ‘gold-gem’ (1113a), heolstor-hofu ‘dark dwelling’ (763a), lyft-lācende ‘floating in the air’ (795a), mōd-ge-mynd ‘memory’ (381a, 839a), wundor-wyrd ‘wonderful event’ (1070a) The Dream of the Rood (1 instance): dōm-dæge ‘Doomsday’ (105a) Juliana (5 instaces): lyft-lācende ‘playing in the air’ (281a), sār-slege ‘painful blow’ (341a, 547a), sweord-slege ‘sword-stroke’ (671a), wēohweorðinga ‘worshipping of idols’ (180a) Phoenix (8 instances): blēo-brygdum ‘variety of colours’ (292a), dēað-dene ¯ red ‘terrified’ (525a), heofum-hrōfre ‘valley of death’ (416a), forht-ā-f æ

Prosodic Constraints on Old English Alliteration

e.

121

‘vault of heaven’ (173a), will-wonge ‘pleasure place’ (89a), winter- ge-wæ ¯dum ‘garment of winter’ (250a), winter-ge-weorp ‘winter storm’ (57a), woruldwelan ‘worldly possession’ (480a) Widsið (1 instance): efen-eald ‘of the same age’ (40a)

5. COMPLEX WORDS AND ALLITERATION Next, as for the investigation of the relationship between the degree of stress and alliteration in complex words it must be based on the following three points. First, as OE prefixes are relatively few in number, pretonic consonants that occupy the initial positions of such prefixes and participate in alliteration are also limited in number. Second, two groups of prefixes must be discriminated, that is, unstressed prefixes such as be-, for-, ge- and those such as ed-, mis- and un- with primary stress when they are added to words except verbs. Third, denominal verbs retain primary stress on the prefixes inherited from their corresponding nouns as in ánd-wỳrdan ‘answer’, fúl-tùmian ‘assist’ and ónd-swàrian ‘answer’. The examples of prefixed words in (11) are taken at random from OE poems in which two pretonic consonants of prefixes and stems are identical. These instances will give us a clue to the way how the poets elaborated in adjusting two pretonic consonants. These examples, too, are qualified as double alliteration for their regular occurrence in a-verses in the same way as alliterating compound words. As a result of investigating all occurrences of the complex words in the whole body of OE poems with the aid of Bessinger and Smith (1978), it turned out that all of them are nonce words except un-ēaðe which is used once in The Metrical Psalms of Paris Psalter. This may well be excluded from consideration for its peculiarity, since it is composed on completely different rules from those of other OE poems. (11) mis-micelra ‘of more varying size’ (Exodus 373a), sin-sorgna ‘of constant ¯ dum ‘ceaseless piece, huge gulp’ sorrows’ (The Wife’s Lament 45a), syn-snæ (Beowulf 743a), un-earge ‘brave’ (The Battle of Maldon 206a), un-ēaðe ‘difficult’ (Andreas 205a), un-orne ‘worn-out, old’ (The Battle of Maldon 256a)

In the same way, we sometimes come across the instances of derived words such as lēof-līc ‘dear’ (Andreas 1446b, Beowulf 2603a) in which the initial consonants of the suffix are identical to the pretonic consonants of root syllables. However, we will claim that double alliteration is not realized in such cases, for they appear not only in a-verse but also in b-verses where double alliteration is not permitted.

122

Yasuaki FUJIWARA

6. CONCLUSION AND RESIDUALS Summarizing the foregoing investigations, we can argue that secondary stresses on the second elements of compound words are qualified for alliteration in the same way as primary stresses on the first elements, while tertiary or zero stresses are irrelevant to alliteration. This is because the frequency of double alliteration between the two elements of compound words is far beyond an accidental one and instances of such compound words are confined to a-verses where double alliteration is allowed. Although it seems that syllables with primary or secondary stresses of compound words or noun phrases may be involed in extra alliteration, this type of alliteration itself is nothing but a result of fortuitous coincidence of the two pretonic consonants. The results of the present analysis may not only claim modification to the metrical framework of Fujiwara (1990) but also shed light on the investigation of OE word formation, especially on the distinction between a word and a phrase, and on the motivation behind OE nonce words. It is also made clear from the preceeding analyses that the difference of stress degrees between the adjectives comprising noun phrases and the second elements of compound words may not be so great because both of them are relevant to double alliteration. Although our assumption concerning the distinction of phrases and compound words in Old English may be justified, editors of OE poetical texts are sometimes divided with respect to the categorization of two given words, i.e., whether they are compound words or phases. Indeed, it may well be likely that there was not a clear distinction between the two when such texts were originally composed. To illustrate the situation under consideration, we will take up all words or phrases in which the editors’ interpretation splits on compiling Genesis A. For instance, Doane (1978) is rather faithful to the original manuscript, while Krapp (1931) follows the framework of traditional metrical theory and tries to make some modifications to the manuscripts on such principles. Doane regards cases in (12a) as compound words while Krapp regards them as phrases. On the contrary, in (12b) Doane regards them as phrases while Krapp regards them as words. Some words may be changed into completely different words from the original ones as a result of emendation like wendeð sæ ¯ > Wendel-sæ ¯ . Thus, even if all clues such as contexts, lexical forms and lexical meanings are resorted to, differences of editorial interpretation may arise as to whether given two words are a compound word or a phrase, as shown in (12).

Prosodic Constraints on Old English Alliteration

123

(12) Lines 117a 870b 934b a. 1481a 1618a 1821b 1905b 2519a 1797a 2046b b. 2211a 2482a 2840b

Doane (1978) græs-un-grēne ‘un-green, as with grass’ forð-gān ‘go forth’ swatig-hlēor ‘sweaty face’ salwed-bord ‘darkened board’ ful-frēolīce ‘very noble’ hēa-byrig ‘high city’ eall-tela ‘quite well’ hēa-burh ‘high city’ self cyning ‘the very king’ folce ge-trume ‘people’s company’ wendeð sæ ¯ ‘the sea runs’ þīne þearfende ‘being in need of food’ hēah stēap-reced ‘high building’

Krapp (1931) græse un-grēne ‘not green with grass’ forð gān ‘go forth’ swatig hlēor ‘sweaty face’ salwed bord ‘darkened board’ ful frēolīce ‘very noble’ hēa byrig ‘high city’ eall tela ‘quite well’ hēa burg ‘high city’ self-cyning ‘the very king’ folc-ge-trume ‘army’ Wendel-sæ ¯ ‘Mediterranean Sea’ wine-þearfende ‘friendless’ hēah-stēap reced ‘very high hall’

REFERENCES Bessinger, J. B. and P. H. Smith. eds. 1978. A Concordance to The Anglo-Saxon Poetic Records. Ithaca and London: Cornell University Press. Blake, N. F. ed. 1964. The Phoenix. Manchester: Manchester University Press. Brooks, K. R. ed. 1961. Andreas and The Fates of the Apostles. London: Oxford University Press. Bussmann, H. ed. 1966. Routledge Dictionary of Language and Linguistics. London and New York: Routledge. Cook, A. S. ed. 1904. Judith: An Old English Epic Fragment. (The Belles=Lettres Series, Section I, Vol. 7) (repr. 1972) New York: AMS. Doane, A. N. ed. 1978. Genesis A: A New Edition. Madison and London: University of Wisconsin Press. Fry, D. K. ed. 1974. Finnsburg Fragment and Episode. London: Methuen. Fujiwara, Y. 1990. Koeishi Inritsu Kenkyuu (Studies in Old English Metrics). Hiroshima: Keisuisha. 1991. “Koeigo hund no Gengotokusei ni tsuite” Gendai Eigogaku no Shoso (Aspects of Modern English Linguistics). ed. by S. Chiba, et al. Tokyo: Kaitakusha. Gordon, E.V. ed. 1957. The Battle of Maldon. London: Methuen. Halle, M. and S. J. Keyser. 1971. English Stress: Its Form and Its Growth,

124

Yasuaki FUJIWARA

and Its Role in Verse. New York and London: Harper & Row. Kent, C. W. ed. 1889. Elene: An Old English Poem. (repr. 1973) New York: AMS. Klaeber, F. ed. 1950. Beowulf and The Fight at Finnsburg. Bosoton: Heath. Krapp, G. P. and E. V. Dobbie. eds. 1931-42. The Anglo-Saxon Poetic Records. 6 vols. New York: Columbia University Press. Leslie, R. F. ed. 1988. Three Old English Elegies. Exeter: University of Exeter Press. Malone, K. ed. 1977. Deor. Exeter: University of Exeter Press. Strunk, W. ed. 1904. Juliana. (The Belles=Lettres Series, Section I, Vol. 8) (repr. 1972) New York: AMS. Swanton, M. ed. 1987. The Dream of the Rood. Exeter: University of Exeter Press. Whitelock, D. ed. 1967. Sweet’s Anglo-Saxon Reader in Prose and Verse. Oxford: Clarendon Press. Zettersten, A. ed. 1979. Waldere. Manchester: Manchester University Press.

The Historical Role of Genitives in the Emergence of DP* Kazuyo MIYAMAE 0. INTRODUCTION The main aim of this paper is to investigate the historical development of the English nominal construction in terms of emergence and establishment of a new functional category (FC), D. I will show that many factors and motives in the preceding period contributed to this drastic syntactic change in the latter half of the 15th century. In course of this discussion, I will give special focus to genitive constructions, which I assume to be the ‘loci’ of the change. My basic standpoint is the same as Osawa (2000, etc.) in that the English nominal construction, which is now generally assumed to be the projection of D, was just NP in the earlier stage, and later developed into DP when a new category D was introduced into the language. This drastic change was motivated, Osawa claims, when the morphological case (m-case) was decayed and some other device came to be required to take over that role. It is a manifestation of the general drift from a morphology-oriented language to a syntax-oriented language. Although I believe her claim is correct, it does not seem fully persuasive or specific, because the concrete steps of the historical change are not shown yet. Is it really the case that only decay of m-case was so decisive to motivate such a radical syntactic change? From this point of view, in this paper I will try to give a more realistic illustration of when and how the new category D was introduced into the grammar. The very first phase may be the leveling of m-case, but some subsidiary factors must have been at work, too, when D was actually realized and established. More concretely, I will show that the direct incentive to create a new node was standardization/cliticization of the genitive marker, and furthermore, a structural need to develop ‘hierarchy’ in the nominal structure encouraged the change. My point is, thus, that D must have been *

This is a revised version of my paper “The Historical Development of the English Nominal Structure from NP to DP,” published in Senshu Jinbun Ronshu Vol. 75. I would like to thank Susan Pintzuk and Anthony Warner for their valuable comments on the earlier versions.

126

Kazuyo MIYAMAE

realized only when some cooperative motives pushed and dragged the structure to the same direction, and with a rather lengthy period of competition. The paper is organized as follows: in Section 1, I will briefly introduce my basic assumptions about FCs and the inner structure of the DP. On the observation that the genitive constructions may provide some clue to the structural change of the NP/DP, the historical facts of various genitive constructions will be outlined in Section 2. Then, in Section 3, I will present my analysis of the historical development of the English nominal structure from NP to DP. Section 4 concludes the discussion. 1. THEORETICAL BACKGROUND 1.1 Emergence of FCs My discussion starts by assuming that the FCs are not universally installed into the human languages, but subject to parametric variation. Unlike the more widely accepted ‘Structure Uniform Hypothesis’ where all languages are assumed to have the fixed set of FCs, I claim that the number and the kind of FCs, and consequently the basic clause structure, can differ from language to language, or stage to stage. This hypothesis was first proposed by Fukui (1986) in order to explain the difference between English and Japanese,1 and was supported in the discussion of the first language acquisition. Radford (1990), for example, observes that child grammars have no determiners, no nominative/objective contrast, nor subject requirements, and concludes that these phenomena reflect the total absence of FCs in child languages. Then, some historical syntacticians provided further reinforcements to the hypothesis. Gelderen (1993) claims that some drastic changes in the late 14th c., such as the appearance of do-support, modals, split infinitives, pro-infinitives and accusative-cum-infinitive constructions, etc., should be related to the rise of the new functional category T (I in her term). This syntactic change, she explains, was triggered by the loss of morphological endings. Similarly, Kiparsky (1995) argues that Proto-Indo-European had no FCs at all, and syntactically embedded clauses were not possible until the FC C was introduced into the language. This discussion was naturally expanded to the nominal level. Osawa (2003a) points out that historical facts bear quite surprising similarities to child language in respect of the usage of articles, and claims that it should be attributed to the absence of the FC D in both child and early languages. It was not until D appeared in the course of historical development, she 1

Later he restates the same claim in the Minimalists’ terminology in Fukui & Sakai (2003).

The Historical Role of Genitives in the Emergence of DP

127

explains, that the modern usage of article was established.2 Furthermore, she claims that the emergence of D was motivated by the decay of m-case; in other words, the rise of all FCs should be considered as the manifestations of the universal ‘grand design’ of languages from morphology to syntax. The author of this paper basically supports her analysis. Needless to say, the claim that a certain language has no FCs does not mean that it lacks grammatical elements; it means, instead, that grammatical features do not project automatically into syntactic categories. The ways to express grammatical elements can differ from language to language, and from stage to stage. In this sense, the explanation that leveling of the morphology triggered the rise of new FCs seems to be quite plausible. Still, there remain some questions for which I do not feel completely satisfied by this explanation. Is it really adequate to attribute such drastic syntactic changes only to the loss of morphological endings? First of all, it is not easy to define the ‘richness’ of the morphology. On what evidence can we say that a certain language (or stage) has a rich morphology? The ‘richness’ of the morphology is always a relative concept. It is true that the verbal endings in the 14th c. were ‘leveled’ in comparison with those in OE, but they were far ‘richer’ than those in PDE. We are not sure on what point we can judge that the morphology was lost so badly as to motivate an alternative device. Another thing we should keep in mind is the chronological gap between the cause (loss of morphology) and its effect (the rise of the new FC). The second half of the 14th c. is generally recognized as a crucial turning point for the English sentence structure, and therefore the new FC T was supposed to emerge in that period. In fact, however, the verbal inflectional system was decayed to a considerable extent already in the OE stage. Doesn’t this chronological gap seem a little too large to prove a direct causal connection? The time lag between the erosion of m-case and the establishment of articles is the case, too. It seems, therefore, that there must have been some ‘bridges’ to connect the two events. Although the very first phase to motivate new FCs may be the leveling of inflectional endings, some other subsidiary factors must also have been at work in the process where they actually came to be realized and established. We have to clarify those factors. In order to clarify those factors, I will take a closer look at the inner structure of NP/DP in the English history in the following section.

2

For similar analyses, see Abraham (1997) and Philippi (1997).

128

Kazuyo MIYAMAE

1.2 The Inner Structure of DP: Genitives as ‘Locus’ of the Change Since Abney (1987), nominal phrases in PDE are generally assumed to be projections of a functional head D, just as verbal structures are those of T. One of its advantages is that we can explain some characteristics of DPs and TPs in a parallel way. Consider (1), for example. Just as the verb analyze is a two-place predicate, the related noun analysis has two arguments in the same position, and theta-marking seems to be carried out by the rather similar mechanism in TP and DP, as roughly shown in (2).3 (1)

a. b.

(2)

a.

Mary analyzed the data. Mary’s analysis of the data [Agent] [Theme] TP Mary

T’ T

VP

analyzed

V’ V

NP

b.

DP Mary

D’ D ’s

3

the data

NP

N’ N

PP

analysis

of the data

The exact status of the category D is, however, far less clear than T; although the head position D is generally considered to be filled by a ‘determiner’, which is traditionally split into some subcategories such as definite article the, indefinite article a/an, demonstratives this/that/these/those, and quantifiers every/each/some/all/many, there is no agreement among researchers on the positions of these respective elements. In this paper, I limit the head determiner only to articles (definite and indefinite) and the possessive marker -’s, though I cannot spare pages for the detailed discussion.

The Historical Role of Genitives in the Emergence of DP

129

It should be noted, however, that the two systems do not allow fully parallel explanation. Unlike TPs, DPs do not necessarily require overt elements in the Spec position (and in that case the head position D must be filled with an article a/the), for instance. Moreover, the Agent appears as nominative and the Theme as accusative in TP, whereas in DP they appear as genitive and a PP, respectively. Interestingly, Weerman (1997) observes that the complements of Ns have to be PPs in the languages that lack m-case system, while they appear as the genitive form if the language has rich morphology. Consider (3) and (4). He also remarks that Middle Dutch genitive sharply declined and was replaced with the post-nominal van ‘of’ phrase in the 14th century. (3)

Modern Dutch het huis van de buren “the house of the neighbours”

(4)

Middle Dutch derste der werelt the-first the world-Gen “the first of the world”

This remark of Weerman’s is quite consistent with the historical change in English; Fries (1940:74) already noticed that originally the genitive elements in the post-nominal position were expressed exclusively by m-case, but they were replaced by of-phrases in the period from early 13th to 14th c. It implies, then, that the two structures in (2) were linearly shown as (5) in the period when the rich m-case was preserved. (5)

a. b.

(verbal structure) (nominal structure)

N[Nom] … V … N[Acc] N[Gen] … V … N[Gen]

To compare the two, the nominal structure (5b) seems semantically more ambiguous than its verbal counterpart (5a), for the different theta roles were expressed by the same m-case marking. In spite of the fact that the word order was relatively free those days, there was no means to differentiate Agent and Theme (or other theta roles) at all. Such a construction must have been syntactically unstable, too. In the later stage, one of the genitive nouns in (5b) was replaced by of-phrase. This change was definitely different from the simple erosion of the genitive case. If the decay of m-case was the cause of the replacement, there is no explaining that the first genitive noun remains to be used until PDE.

130

Kazuyo MIYAMAE

Given that (5b) was untenable both semantically and syntactically, we should rather call it a ‘remedial’ change. In other words, the structural ‘defect’ of (5b), not the loss of m-case, was responsible for the change of the inner structure of the nominal structure. My point is, thus, that the genitive forms should be investigated as the ‘1ocus’ of the change in question.4 The peculiarity of the genitive case has already drawn the attention of some researchers. Abraham (1997), for instance, considers that the choice of a different m-case for the object of a verb could lead to different semantic interpretations, i.e. definite or indefinite, and that the loss of genitive case as the verbal object correlates with the rise of the ‘definite’ articles. Philippi (1997) makes a similar argument as well. Their claim is especially intriguing in that they relate the rise of ‘definite’ articles to the loss of ‘genitive verbal object’ in particular, not to the loss of m-case in general. Based on the discussion so far, we will examine the earlier data of various genitive constructions in the following section, assuming that they hold the ‘key’ to explaining the historical development of the English nominal structure. 2. HISTORICAL CHANGE OF GENITIVE CONSTRUCTIONS In this section we will outline the historical change of various genitive constructions in English. Primary data cited here owe much to Nakao (1972) and Ono and Nakao (1980), if not particularly mentioned. 2.1 Inflectional Genitive The inflectional marking -es (~ys~is), originated from the OE genitive ending of the masculine and neutral nouns with strong declension, was expanded to all kinds of nouns including proper nouns from EME, and came to be the dominant genitive marker by the 14th c. at the latest (Mustanoja (1960:71)). It should be noted that the genitive case inflection was exceptionally well retained compared with other case markers. According to Allen (1997:115-6), irregular genitive forms, too, persisted into the 15th c., showing that the genitive case was still learned by language learners at that period. 4

I have to admit that the discussion here lacks any factual evidence. In fact, Anthony Warmer comments (personal communication) that such structures as (5b), which contain two thematically different genitive arguments, are rarely found in the earlier data, and accordingly there was no confusion that could trigger the change. Nevertheless, I suppose that my discussion is not vacuous; if Warner’s remark is correct, the OE syntax did not have the way to express two thematically different arguments in the same noun phrase, which implies that the nominal structure in this period was poor, or deficient, to express semantic relationships adequately. This ‘defect’ could be a powerful incentive to motivate the rise of D.

The Historical Role of Genitives in the Emergence of DP

131

The first example of the modern usage with apostrophe, -’s, appeared in 1466 (in divers men’s manners, PL), but it was after 1650 that the usage was standardized (Nakao (1972:151)). This form can be regarded as the only survivor in PDE of the earlier rich m-case. 2.2 Of-Genitive The inflectional genitive was dramatically taken over by the analytic type (of-genitive). The analytic type began to appear already in OE, but it was only marginal before the 12th c. Its frequency explosively increased in the 13th c., and became overwhelming in the 14th c. Chart (1)

(Thomas (1931), cited by Rosenbach (2002:179))

Interestingly enough, however, it did not happen that the inflectional genitive was completely extinguished. It is notable that Rosenbach (2002) claims that the inflectional genitive (s-genitive in her term) became more productive again in 1400-1630, as shown in Chart (2). She considers that the possessive marker -’s has changed from a fully fledged inflection in OE into a clitic in ModE, and this transition period, late ME, correlates precisely with the period in which s-genitive is becoming more frequent again.

132

Kazuyo MIYAMAE

Chart (2)

(Rosenbach (2002:186))

2.3 Gen+H vs H+Gen The Gen(itive)+H(ead) order (as in (6a)) and the H+Gen order (6b) were competing in OE. From the first half of the 11th c., however, the former became more frequent gradually. The substitution proceeded very rapidly, and by the mid 13th c. the prenominal genitive replaced the postnominal type completely. (6)

a.

b.

þæs cyninges þegnas the-Gen king-Gen thanes “the king’s thanes” (ChronA 755) heafod ealra haligra manna head all-Gen holy-Gen men-Gen “the head (leader) of all saints” (ÆCHom ii. 14. 23)

Chart (3)

(Fries (1940), cited by Nakao (1972:392))

The Historical Role of Genitives in the Emergence of DP

133

2.4 Split Genitive When more than two genitive nouns modify the head noun, OE allowed two word orders, that is, (7a) Gen1+Gen2+H and (7b) Gen1+H+Gen2 (OE split type). Although the first type came increasingly frequent during ME, the split type was more common in OE, and survived for quite a long time. Mustanoja (1960:78) notes that the split genitive is common all through ME, and Jespersen (1909:293) calls this splitting ‘the universal practice up to the end of the 15th c.’ (7)

a.

b.

Gen1+Gen2+H Ælfredes cyninges godsunu “King Alfred’s godson” Davið kinges kinn “King David’s kin” Gen1+H+Gen2 (OE split type) Ælfredes godsune cyninges “King Alfred’s godson” the kynges metynge Pharao “the king Pharao’s dream”

(ChronA 890) (Orm)

(ChronE 890) (Ch, BD 282)

The post-nominal genitive of the split genitive construction, Gen2, lost its inflectional ending after LOE, as seen in the last example in (7b), and was gradually replaced by the of-genitive, which began to appear in the early 12th c. This new split type is often referred to as the ME split genitive construction. (8)

Gen1+H+of-N (ME split type) þes eorles dohter of Angeow “the earl of Anjou’s daughter” This kynges sone of Troie “this king of Troy’s son” The archbishop’s Grace of York “the archbishop of York’s grace”

(ChronE 1127) (Ch, TC, III, 1715) (Sh, 1 Henry IV, III. ii. 119)

Anderson & Lightfoot (2002), however, comment that the ME split type with of-genitive was not a replacement of the OE split type (Gen1+H+Gen2). They observe that the OE split type was strictly restricted to the contexts where the postnominal genitives are in the appositional relation or the conjunct of the first (prenominal) genitives, but the ME split type belonged to neither of them. The ME split genitives were characteristic in that postnominal of-phrase was name of a place which belongs semantically to

134

Kazuyo MIYAMAE

the prenominal genitive element. Unlike those in OE, the postnominal genitive phrase did not have any direct thematic relations with the head noun. Compare (8) with (7b). On the other hand, the inflected version of the ME examples such as *þæs cyninges godsune Frances (‘the king’s godson of France’) could not be found in the OE texts. 2.5 Group Genitive The ME split genitive type discussed above was a short-lived custom. Already in the 15th century it began to give way to another genitive construction where two genitives are regrouped and placed before the head noun, that is, [N1 of N2]Gen+H type. The first example of this group genitive is found in Chaucer’s prose, but it was still quite rare at this period. According to Allen (1997:120), there are only 3 examples in Chaucer’s works, all of which are of the form god of Xs. This type of construction became increasingly frequent during and after the 15th c. (9)

[N1 of N2]Gen+H grete god of Loves name “the great god of love’s name” þe kyng of Fraunces men “the king of France’s men”

(Ch HF 1489) (POLYCH VIII, 349. 380) (cited by Allen (1997:121))

In these examples, the morphological ending -(e)s attached to the word immediately preceding the head noun stands for the genitive case of the whole pronominal phrase. The appearance and standardization of the modern usage with apostrophe -’s was a practice of about a century later, from the latter half of the 15th c. to the 17th c. Allen states that the early examples of the group genitives were most easily used when the possessor NP could be regarded as the ‘title’, and none of the ME examples she found involve any preposition other than of. She assumes that the group genitive construction was triggered by the shift from the OE genitive inflection into the ModE clitic -(e)s; namely, the group genitive developed when the old inflectional genitive ending -(e)s was generalized to all noun classes and reanalyzed as a clitic. 2.6 Separate Genitive The separate genitive construction, in which the independent form his appears after the possessor NP as in (10), came into use around 1250, and was frequent especially from the 15th c. to the end of the 17th c. It was sometimes used with feminine or plural nouns, but the form was always his

The Historical Role of Genitives in the Emergence of DP

135

or its variant ys, totally irrespective of the gender of preceding nouns, as (11) shows. (10) Winchestre his town “the town of Winchester” (Lay Brut B 19630) God ys blessiyng “God’s blessing” (PL[(ed). Davis] 333, 1) (11) And on Mary Magdalene ys day he dyde “and he died on Mary Magdalene’s day” (Gregory Chron. P. 90. 20) ... not borrowed of other men his lippes “not borrowed from other men’s lips” (Ascham b. 1515, Tox. A. p.5. 23) (cited by Allen (1997:118))

Allen notes that the separate genitive seems to have undergone a change around the end of the 16th c.: examples of her and their came to be used with feminine/plural nouns.5 (12) Lucilla hir company “Lucilla’s company” (Lyly, Anatomy of Wit, p. 237. 8, 1578) Canterbury and Chillingworth their book “Canterbury and Chillingworth’s book” (Verney Memoirs II. 222. 11, letter ofc. 1645) (cited by Allen (1997:123, 116))

The time lag between the appearance of his type and her/their type is clearly due to the fact that this construction originally arose from the phonetic similarity between the genitive marker es~is and the possessive pronoun his. Allen remarks that the genitive marking gradually came to be recognized as a clitic, and the spread of the group genitive led to an increase in the practice of writing the genitive separately. Then, the separate genitive, which was an orthographical variant of OE inflectional genitive in the beginning, was interpreted as the pronoun his, which in turn enabled other genitive pronouns her/their occur in the same context. 2.7 Absolute Genitive The absolute genitive construction refers to the structure in which the head noun noting church/house/shop is deleted. Mustanoja (1960:74) states 5

Although OED and Mustanoja (1960) give ‘earlier examples’ of this type, Allen claims that both of them are too disputable to label as the separate genitive. See Allen (1997:125-8).

136

Kazuyo MIYAMAE

that such a construction did not appear in OE texts, and the earliest example was found in the end of the 13th c. (13) at seint poules “at St. Pau1’s” at my moders “at my mother’s”

(SLeg 109, 91) (PL II 3611-2) [Carstensen]

After the 15th c. at the latest, the absolute genitive came to be used as general custom, which can be seen from the expressions such as the house of St Johns/the Church of St Peters that came to be used after 1400. 2.8 Double Genitive The double genitive construction, in which both of the preposition of and genitive forms appeared as in (14), was another important ME innovation. According to Allen (2002), the full-search of the on-line corpus provides no evidence of double genitive during the OE period. She declares that the double genitive constructions with indefinite article or quantifiers came into use from the middle of the 14th c., and extended to demonstratives (and marginally to definite determiners) in the 15th c. (14) a. b.

c. d.

a man o þair “a man of theirs” (CM(Vsp, c 1350) 7465) Oon Maximus, that was an officer Of the prefectes “one Maximus, who was an officer Of the prefect’s” (Ch CT G 369) Or thou gettis this stede of myne “before you get this place of mine” (Ipom. A 7747) under the fellys of Thomas Bettsons “under Thomas Bettson’s sheepskins” (Cely 131. 46 (1481 William Cely)) (cited by Allen (2002:26-31))

At the earliest stage, Allen comments, such examples with possessive pronouns as (14a) were more numerous than examples with nouns as (14b). Furthermore, the double genitive with a definite determiner such as (14d) never really flourished. 2.9 Summary of the Historical Facts The rises and falls of the English genitive constructions outlined so far can be roughly shown as Table 1. In the ‘RISE’ column, the terms in brackets stand for rare or sporadic occurrence, whereas the terms in block type stand

The Historical Role of Genitives in the Emergence of DP

137

for more frequent, or established, occurrence. The items in block letters in the ‘DECLINE’ column, similarly, represent that they became completely obsolete. Table 1.

Rises and falls of various genitive constructions (1) DECLINE

RISE

12c H+Gen inflectional genitive

[of-genitive] [ME split genitive (Gen1+H+of-N)]

13c

ME split genitive (Gen+H+of-N) [separate genitive (his-type)] [absolute genitive]

14c OE split genitive (Gen1+H+Gen2)

of-genitive generalization of -(e)s [group genitive] [double genitive w/indefinite art]

15c OE split genitive (Genl +H+Gen2) ME split genitive (Gen+H+of-N)

group genitive separate genitive (his-type) double genitive w/indefinite art [double genitive w/demonstrative and definite art] [-’s form] [inflectional genitive-26]

16c

separate genitive (her/their-type)

17c double genitive w/ definite art separate genitive (his- type and her/their-type) inflectional genitive-2

-’s form

And Table 2 tries to visualize the same historical transition; solid and broken lines illustrate established and unstable usage, respectively.

6

The term ‘inflectional genitive-2’ means the second rise of inflectional genitive, following Rosenbach’s observation (s-genitive in her term).

138

Kazuyo MIYAMAE

Table 2.

Rises and falls of various genitive constructions (2) 11c

12c

13c

14c

15c

16c

17c

18c

19c

inflectional genitive of-genitive OE split genitive (Gen1+H+Gen2) ME split genitive (Gen+H+of-N) group genitive separate genitive (his-type) separate genitive (her/their-type) asolute genitive double genitive (w/indef art) double genitive (w/demon) double genitive (w/def art) use of apostrophe (-’s)

Tables 1 and 2 clearly illustrate that the 15th c. is the crucial period when most radical changes occurred; both OE and ME split genitive constructions were lost from the language altogether, while group genitive and double genitive (with an indefinite article) were established as the full-fledged structures. At the same time the modern practice of using apostrophe first appeared, and the inflectional genitive, which once lost its power completely, became productive again in this period. Furthermore, it is noteworthy that separate genitive (his), which was destined to be short-lived, was quite frequent in this period as if it worked as a link between the two stages. We cannot possibly consider that these changes were independent of each other. Instead, it is far more plausible that all of them were closely related to each other, and consequently to one drastic change at the deeper level. The first change may have been a slight fluctuation in pronunciation or meaning. When it triggers other fluctuations one after another, however, they proceeded in the same direction to bring about a new syntactic structure as a whole. Thus, we claim that all the historical events discussed in this section worked together to introduce a new FC D, and that ‘conspiracy’ changed the English nominal structure completely in the latter half of the 15th c.

The Historical Role of Genitives in the Emergence of DP

139

3. ANALYSIS: HOW THE CATEGORY D WAS INTRODUCED INTO ENGLISH In this section, I will try to show the whole ‘scenario’ of the historical facts discussed in the previous section, relating to when and how D was introduced into the English grammar. My main point here is that two important factors were involved in this process: namely, (i) generalization and cliticization of the genitive case morphology, and (ii) structural change from parataxis to hypotaxis. 3.1 Generalization and Cliticization of the Genitive Morphology First of all, the ‘1oss’ of the m-case is not a good description of the situation, at least of the genitive. It is true that the inflectional genitive became far less frequent and the genitive object in the verbal structure was absorbed into the neutral objective case in the ME period, but it does not necessarily mean that genitive case marker was lost. In fact, the genitive marker –es (~ys~is) was expanded to all kinds of nouns from EME and is retained as -’s until today. It is very important that this dominant form happened to be phonetically similar to the weak form of a different grammatical word, his. When the -(e)s form became the standard and productive genitive marker in EME, other m-case markings had already decayed to a considerable extent. Accordingly, it is quite probable for children who did not have m-case any longer to misunderstand the ‘relic’ of the genitive marker in the adult grammar as the possessive pronoun his, which was phonetically and semantically similar. In short, the ‘standardization’, not ‘loss’, of the genitive morphology, and some accidental ‘coincidence’ played an important role in the following syntactic change. Possessive pronouns, on the other hand, co-occurred with other determiners (in the PDE terms) quite freely in ME, as in (15). In those examples, possessive and demonstrative pronouns were able to modify the same head noun, just as adjectives. They show that the category D and DP structure in the PDE sense were not developed yet. Such distribution of possessive pronouns, too, may have further promoted confusion with the inflectional endings. (15) this my present testament thees oure letters þe same your suppliant

(PL [(ed.) Davis] 194,4) (ibid. III 340)[Carstensen]) (ibid. [(ed.) Davis] 488, 20)

Thus, children’s misinterpretation brought the change (16):

140

Kazuyo MIYAMAE (16) a.

NP N God-es

b.

NP

N

N

blessing

God

-(e)s(~ys~is)~his

N blessing

One might be dubious about the structure (16b), since it does not show a well-formed binary structure nor the exact category for the variants -(e)s(~ys~is)~his. What I want to represent in (16b), however, is not the individual grammar, but the state of the whole speech community of this period. In the 13th-14th c., the earlier genitive marker was confused with the independent word his, and this confusion enabled the earlier bound morpheme to have an independent syntactic position. The structure may seem somewhat vague, but it should be regarded as the transitional stage toward another syntactic change. I assume that this new ‘unclear’ slot is the predecessor of the PDE functional head D, which hosts not only determiner but also the possessive marker -’s, as in (17). (17)

DP N

God

D’ D

NP

-’s

blessing

The biggest advantage of my assumption illustrated as (16) and (17) is that it can explain the historical facts of some genitive constructions in a consistent way. Remember that the separate genitive, such as God his blessing ‘God’s blessing’ came into use around 1250 when the generalization of -(e)s was in progress, and was used most frequently from the 15th c. when the standardization was completed. Around the end of the 16th c., the separate genitive expanded to her/their. It strongly suggests that the slot for a clitic in (16b) was so well-established as to allow the phonetically different possessive pronouns to appear. In the late 17th c., this wrong expansion came to be obsolete with their ‘authentic’ forerunner his, when the modern usage -’s came to be a rule. It implies that the syntactic status of -’s was full-fledged by

The Historical Role of Genitives in the Emergence of DP

141

this period so that the confusion with possessive pronouns could not happen any more.7 Furthermore, the group genitive as the grete god of Loues name also came into use in the late 14th c., when the separate genitive became popular. It was not a mere coincidence. As Allen (1997) claims, the group genitive became possible only when the old inflectional genitive ending was reanalyzed as a clitic. This remark is quite consistent with the explanation shown above. Here, it should be noted that I do not mean that the genitive marker -es changed to the possessive pronoun his and then to -’s. The change in question was just -es>-’s; the pronoun his was ‘supporting’ this change from the outside, because of its phonetic/semantic/syntactic similarity. 3.2 Structural Change from Paratxis to Hypotaxis Before discussing group genitive further, let us have a closer look at two kinds of split genitive constructions, ‘Gen1+H+Gen2’ and ‘Gen+H+of-N’, which disappeared in the 15th c. giving way to group genitive. According to Anderson and Lightfoot (2002), the OE split genitive is analyzed as (18). They assume that the head noun assigns the same case and the same thematic role both to the left and the right, based on the fact that the OE grammar allowed both Gen+H and H+Gen orders.8 (18) (=Anderson and Lightfoot’s (8.16)) DP Ælfredes

D’ D

NP Godsune

cyninges

This scheme, however, does not seem very persuasive. Above all, it violates some universal hypotheses which have been presented and widely supported so far, such as theta-criterion and Uniformity Theta-Assignment Hypothesis (UTAH). It is true that both of the two word orders Gen+H and H+Gen were possible and consequently any theta-role assignment could 7 8

For a more detailed discussion on the historical role of separate genitive, see Miyamae (2005). According to Anderson and Lightfoot (2002:177, fn. 5), Allen (2002) argues that cyninges is adjunct to godsune rather than a complement. They say that this raises interesting questions, but they do not discuss it in detail.

142

Kazuyo MIYAMAE

have worked in both directions in OE, but allowing bi-directional case-marking mechanism in one and the same structure is another thing. What structure did the OE split construction have, then? As is shown in 2.4, the OE split genitive was characteristic in the respect that both of the two genitive nouns had the thematic relation with the head noun and these ‘split’ elements had to be conjunct or in apposition. This restriction seems to indicate a rather primitive setup of OE split construction, which is far more flat than (18), i.e. (19). (19)

NP NP

(and)

NP

Gen1

N

Gen2

(N)

Ælfredes

godsune

cyninges

godsune

Here I presume that the OE split type was derived from the conjoined NPs where the repetition of the head noun of the identical reference was avoided. The conjunction ‘and’ actually appeared between the two elements in some examples such as Sodoma land 7 gomorra ‘the Sodomites’ and the Gomoreans’ land.9 The modern DP analysis cannot provide any tenable structure for the ME split construction as in wives tale of Bath, either. The tentative DP structures for this example would be (20a) or (20b). (20) a.

DP wives

b. D’

D

9

DP DP

NP

wives

N

PP

tale

of Bath

PP D’

D

of Bath N tale

This construction reminds me of another split construction common in OE, such as Hengest feng to rice 7 Æsc his sunu ‘Hengest and his son Æsc succeeded to the kingdom’ (ChronA 455). Traditionally this construction is analyzed as the postposing of the conjoined element, but Sasao(=Miyamae) (1981) proposes a parallel analysis to (19); namely, it is derived from the coordinate structure which contains the identical elements (Hengest feng to rice 7 Æsc his sunu feng to rice).

The Historical Role of Genitives in the Emergence of DP

143

(20a) is a possible expansion of Anderson and Lightfoot’s scheme (18), and (20b) is the adjunction analysis in which the postnominal of-phrase is assumed to be adjoined to the preceding DP. Neither structure, however, predicts the correct thematic relation between wives and of Bath. The third possibility to defend the DP analysis would be (21), where wives of Bath is originated in the DP-Spec position and later of Bath is ‘extracted’ and ‘postposed’ into the complement PP position. (21)

DP Spec

D’

NP

D

NP

N

PP

N

wives

of Bath

tale

PP

Although (21) is more reasonable than (20a, b) from the point of semantic relationship between elements, it is also untenable when we consider that the structure (21) itself, namely, the group genitive construction, occurred later than the ‘derived’ (=split) structure in the history. Hence the discussion naturally brings us to the flat conjunct hypothesis again. The prototype of the ME split genitive is assumed as (22): (22) NP2

NPl Gen

H

H

PP

wives

tale

wife

of Bath

In this case the structure may be rather fuzzy in some respects; two conjuncts are not parallel in a strict sense, and it is uncertain whether the two NPs are really ‘conjoined’ or not. Nevertheless, I consider that this analysis is still of great advantage to others, mainly because (22) can explain why two split constructions were taken over by group genitive at the same time in a consistent way. In my analysis, the historical process from OE/ME split genitive to

144

Kazuyo MIYAMAE

group genitive can be roughly illustrated as (23) and (24), respectively. (23) OE split genitive > group genitive a. NP0 NP1

NP2 N

NGen

b.

NGen

NGen N

NGen

NP N N

N

Alfred

king

NGen

d.

-es

N godson

Alfred’s king’s

Alfred’s godson king’s godson

c.

NP1

N

DP Spec

D’

godson

D -’s

(24) ME split genitive > group genitive a. NP NP1

b.

NP2

NP1 NGen

N

NGen

N

NGen

PP

NGen

PP

wife’s

tale

wife

of Bath

wife’s

of Bath

c.

NP N

NP

-es

N

PP

wife

of Bath

d. N tale

tale

DP Spec

D’ D

NP

-’s

What happened in (23) and (24) was essentially the same process. First, the genitive element in NP2 of the initial state (a) was adjoined to the preceding genitive NGen in NP1. The incentive of this movement is perhaps to

The Historical Role of Genitives in the Emergence of DP

145

‘regroup’ two arguments of the same thematic role as one syntactic constituent. The topmost NP0 was pruned since it did not have branching any more, as in (b). Around the same time the earlier m-case marker was cliticized, producing a different node as in (c). The stage (c) can be justified by the existence of such examples as David kinges kinn ‘King David’s kin’ or þe kyng of Fraunces men ‘the king of France’s men’, where only the last word of the genitive elements had the -es ending. This clitic -es was a purely grammatical element which was irrelevant to lexical meaning, and in the next stage (d) it was developed into the new functional category/node D. Consequently, -’s became the functional head D which marks Case of the preceding NP, and the whole NP became the projection of D. Since both split genitive constructions completely disappeared in the mid 15th c., I assume that this new syntactic array was established by this period.10 The process from (c) to (d) was not a very simple step. The very first incentive to create D could be the decay of m-case in general, which had begun at the earliest stage of the English history. When m-case gradually decayed, the language lost the way to specify the definiteness or referential role of nouns, and an alternative way to express it came to be required. The genitive case marker, however, was not lost but cliticized, and created a new position by the 13th-14th c. It would be probable that this new node was established as a purely functional category/node to host the definiteness or referential role of nouns. As for the establishment of the binary DP structure, the analogy from other FCs or some general tendency toward the uniform syntactic structure must have been involved as an important incentive. In addition, I want to emphasize here that this change was also driven by the necessity to create ‘hierarchy’ within the nominal structure. The flat structure (a) was evidently insufficient to express complicated information unambiguously. The emergence of D brought another layer to the nominal structure and allowed the appearance various kinds of syntactic embedding, just as the introduction of C and T did so in the clausal level. For example, after the emergence of D the language has gradually expanded the contexts where the group genitive can appear, as in (25). This expansion was possible because -’s can mark the whole preceding DP which is located in the Spec position. 10

Anthony Warmer remarks (personal communication) that the structure (24b) was not attested in the historical data, which might harm my story. I do not think that this ‘missing link’ is a very serious defect, since I assume that there was not such a big time lag between (24b) and (24c).

146

Kazuyo MIYAMAE (25) [the man in the car]’s ears (Quirk et a1. (1985:§17.119)) [people who hurry]’s ideas (Zwicky (1987:141)) It’s not mine, it’s [a person who went to Cambridge and got a first in engineering]’s. (all examples cited by Rosenbach (2002:8-9))

Hoekstra (1995), too, argues that the primary motivation for functional categories is the insufficient number of positions provided by a simple phrase structure configuration. Thus, the English nominal structure created a new position which had not existed at the earlier stage, established it in the course of time, and enabled the language to express more complicated grammatical relations which had been impossible in the earlier ‘simple phrase structure configuration.’ 4. SUMMARY AND CONCLUDING REMARKS In this paper I showed that the English nominal structure developed from NP to DP via emergence of the functional category D, giving special focus to the genitive constructions. The emergence of D should not be attributed only to the loss of morphology as often claimed, but should be related also to other various factors which ‘pushed’ and ‘dragged’ the change. Based on the historical overview of genitive constructions (summarized as Table 1 and 2), I indicated the 15th c. as the critical period for the nominal structure, and argued that it reflects the drastic change from NP to DP in the deeper level, which was accomplished in that period. I discussed that historical facts of separate genitive and group genitive constructions in particular represent the path, in which minor fluctuations in the phonetic or semantic levels led to this syntactic change, taking a quite lengthy period. My goal is to give a whole picture of the process of emergence of D, but needless to say, this study is only the first step. Many more possible factors remained to be discussed; above all, absolute genitive and double genitive constructions which cannot be dealt with here could give further light upon the subject. The historical investigation of articles/demonstratives, too, will be indispensable to clarify the picture completely. Still, I hope that the discussion in this paper will provide some corroboration and new perspective on the emergence of D. REFERENCES Abney, S. 1987. The English Noun Phrase in its Sentential Aspect. Doctoral dissertation, MIT. Abraham, W. 1997. “The Interdependence of Case, Aspect and Referentiality

The Historical Role of Genitives in the Emergence of DP

147

in the History of German: the Case of the Verbal Genitive”. Parameters of Morphosyntactic Change, Kemenade & Vincent 1997. 62-93. Allen, C. 1997. “The Origin of the ‘Group Genitive’ in English”. Transactions of the Philological Society 95:1. 11-131. 2002. “On the Development of a friend of mine”. English Historical Syntax and Morphology, Fanego & Lopez-Couso & Perez-Guerra 2002. 23-41. Anderson, S. & D.W. Lightfoot. 2002. The Language Organ: Linguistics as Cognitive Physiology. Cambridge: Cambridge University Press. Fries, C. C. 1940. American English Grammar. New York: Appleton. Fukui, N. 1986. A Theory of Category Projection and Its Applications. Doctoral dissertation, MIT. & H. Sakai. 2003. “The Visibility Guideline for Functional Categories: Verb Raising in Japanese and Related Issues”. Lingua 113. 321-375. Gelderen, E. van. 1993. The Rise of Functional Categories. Amsterdam: Benjamins. 2002. “The Role of Person and Position in Old English”. Pathways of Change, Fischer & Rosenbach & Stein 1993. 187-207. Giusti, G. 1995. “A Unified Structural Representation of (Abstract) Case and Article”. Studies in Comparative Germanic Syntax, Haider & Vikner (eds.) 1990. 77-93. 2001. “The Birth of a Functional Category: From Latin ILLE to the Romance Article and Personal Pronoun”. Current Studies in Italian Syntax, Cinque & Salvi (eds.) 2001. 157-172. Hoekstra, T. 1995. “The Function of Functional Categories”. Glot International 1(2).3-6. Mustanoja, T. F. 1960. A Middle English Syntax. Helsinki: Société Néophilologique. Kiparsky, P. 1997. “The Rise of Positional Licensing”. Parameters of Morphosyntactic Change, Kemenade & Vincent (eds.) 1997. 460-494. Miyamae, K. 1997. “Why Verb-Second was Lost in English: the Rise of Functional Categories”. Studies in English Linguistics, Ukaji, et a1. (eds.) 1997. 1194-1210. 2004. “The Historical Development of the English Nominal Structure: from NP to DP”. Senshu Jinbun Ronshu 75. 263-297. 2005. “Kinou-hannchu D wa Naze Shutugen shitaka—His-Zokkaku no Shiteki Yakuwari wo Megutte [Why the Functional Category D was Introduced—In Terms of the Historical Role of His-Genitive]”. Tsuda Journal of Language and Culture 20. 86-98. Jespersen, O. 1909-49. Modern English Grammar. 7 vols. London &

148

Kazuyo MIYAMAE

Copenhagen: Munksgaard and Allen & Unwin. Nakao, T. 1972. Eigo-shi II [The History of English II]. Tokyo: Taishukan. Ono, S. & T. Nakao. 1980. Eigo-shi I [The History of English I]. Tokyo: Taishukan. Osawa, F. 2000. “The Historical Emergence of DP in English”. English Linguistics 17 (1).51-79. 2003a. “Syntactic Parallels between Ontogeny and Phylogeny”. Lingua 113.3-47. 2003b. “The Rise of IPs in the History of English”. Historical Linguistics 2001, Blake & Burridge (eds). 2003. 321-338. 2003c. “The Development of Complementation: from Nominalization to Embedding”. Paper presented at The 16th International Conference on Historical Linguistics, University of Copenhagen. 2007. “The Emergence of DP from a Perspective of Ontogeny and Phylogeny: Correlation between DP, TP and Aspect in Old English and First Language Acquisition”. Nominal Determination, Stark, Leiss & Abraham (eds). 2007. 311–337. Philippi, J. 1997. “The Rise of the Article in the Germanic Languages”. Parameters of Morphosyntactic Change, Kemenade & Vincent (eds.) 1997. 62-93. Radford, A. 1990. Syntactic Theory and the Acquisition of Syntax. Oxford: Blackwell. Rosenbach, A. 2002. Genitive Variation in English: Conceptual Factors in Synchronic and Diachronic Studies. Berlin: Mouton de Gruyter. Sasao(=Miyamae), K. 1981. “Some Consideration on Reduced Coordinate Structures”. Studies in English Linguistics 9.145-57. Weerman, F. 1997. “On the Relation between Morphological and Syntactic Case”. Parameters of Morphosyntactic Change, Kemenade & Vincent (eds.) 1997. 427-459.

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose Akinobu TANI 0. INTRODUCTION Word pairs (WPs) such as time and tide were one of the most important rhetorical ornaments frequently employed by medieval writers, which has long been recognized by critics of medieval English literature. Chaucer, for example, employed this rhetorical ornament to stylize “curial prose” (Burnley, 1986). Therefore, Tani (forthcoming b) has analyzed the use of the WPs in all the Chaucer’s prose texts, i.e. the Tale of Melibee (Mel), the Parson’s Tale (Pars), Boece (Bo) and a Treatise on the Astrolabe (Astro), in contrast with two additional prose texts as controls. Through the analyses of 1) the frequency of WPs and 2) the etymologies and orders of the members of the WPs, he found 1) that the frequency of WPs and the etymologies and orders of the members of the WPs are useful in distinguishing the texts, 2) that these texts are classified into three groups, Mel > Bo & Pars > Astro, 3) the use of the WPs in the Astro is substantially different from that in the other prose texts, and 4) the ratio of the types of WPs with different etymologies and orders of the members is approximately parallel among the prose texts except for the Asto. Though a large number of studies have very lightly touched on the WPs in Chaucer’s verse, actually no study has yet been done with special reference to the WPs in his verse. For one thing, most of the studies were directed toward those in the prose texts. Actually, they mainly deal with the WPs in the Mel and the Bo to examine the style of the Mel, or to examine Chaucer’s translation technique in either text. Thus, studies on the WPs in Chaucer’s verse are desiderata in comparison with those in the prose texts. Chaucer was clearly conscious of the difference in writing between verse and prose as is evident from the following comments: I speke in the prose, and lat him rymes make MLT(2) 96 But al shal passe that men prose or ryme Scog 41

Therefore, the comparison of the WPs between Chaucer’s verse and prose is expected to better characterize his use of the WPs in his verse texts.

150

Akinobu TANI

In addition, the poet’s verse has had a profound effect on later English poets. Since Tani (forthcoming b) discussed Chaucer’s use of the WPs in the prose, the emphasis in the present study is directed toward the WPs in the verse. To compare the results which Tani (forthcoming b) found concerning the WPs in all Chaucer’s prose texts, this study treats as the verse texts Fragment I of the Canterbury Tales (CT): i.e. the General Prologue (GP), the Knight’s Tale (Kn), the Miller’s Tale (Mil), the Reeve’s Tale (Rv) and the Cook’s Tale (Co). In addition to the comparison of the WPs between Chaucer’s verse and prose, this study also discusses whether there are any similarities and/or differences in the use of WPs among the verse texts contained in Fragment I, and, if any, what they are.1 .

1. PREVIOUS STUDIES As a study dealing with the WPs in Chaucer’s texts, Jespersen (1912) may be mentioned. He (1912:97) claimed, concerning Chaucer’s WPs, that they were used “to heighten or strengthen the effect of the style” because “the reader is evidently supposed to be equally familiar with” both native and French-loan members of the WPs. Yet we cannot be sure whether Jespersen’s claim is intended for the WPs in Chaucer’s verse or prose because all the seven examples of Chaucer are quoted from the GP of the CT, not from the prose texts. The studies related to the WPs in Chaucer’s verse include Elliot (1974) and Masui (1964). Elliot (1974:99) pointed out that alliterative WPs are the most common pattern of alliterative phrases which he claimed are “much more common in Chaucer’s English than is often realized”. Yet we cannot ascertain whether Elliot’s claim is valid or how frequent such alliterative WPs are used, because he does not give any numerical data whatsoever. Masui (1964) examined Chaucer’s rime technique in general, and dealt with rime phrases, or “phrases which fill up the lines to eke out the metre” (Masui, 1964:169). Such rime phrases include WPs as well. Masui (1964:169) stated that: Especially, phrases which denote ‘one and all, persons of all ranks’ are found at the end of a line. ‘Alle and some’ and ‘bothe moore and lesse’ are, among other phrases, frequently met with in Chaucer.

Masui (1964:173) also directed our attention to the WPs employed as 1

This study is a revised version of the paper read at the 1st international conference of SHELL (the Society of Historical English Language and Linguistics) held at Chiba University, Japan, in September 2005. All the quotations of Chaucer’s texts are from the Riverside edition (1987).

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

151

rime phrases which consist of “or”, “ne” and “nor”. He (1964:169-74) gave the examples of the WPs employed as rime phrases such as: alle and some, (bothe) moore and lessee, (bothe) yonge and olde, grete and smale, (bothe) fremed and tame, muche and/or lite, hye or lowe, lowe and hye, hoote and cold etc.

After pointing out the commonality of Chaucer’s rime phases with those in ME verse romances, Masui (1964:169) suggests the reason for such commonality, saying that: Chaucer may have been influenced by the conventional tendency of this day ‘which had been fostered by the custom of oral delivery’ in the Middle Ages (R. Crossby, Chaucer and the Custom of the Oral Delivery).

Masui made such statements on WPs as a kind of rime phrases, not on WPs per se. So what is not clear from his study is how many of the WPs are used as rime phrases, and how the other uses of the WPs are in Chaucer’s verse. Despite these studies, there is no study which takes into account, 1) the symmetry and/or asymmetry of the WPs between Chaucer’s verse and prose, and 2) the functions of the WPs among Chaucer’s verse texts. To answer these issues, this study examines the WPs in Fragment I of the CT in comparison with those in all Chaucer’s prose texts, by the following methods, i.e. 1) frequency of WP’s, 2) etymological make-ups of WP components and 3) repetition of WPs. 2. FREQUENCY OF THE WPs IN CHAUCER’S VERSE AND PROSE The frequency of WPs in the individual texts in Fragment I of the CT was examined in comparison with Chaucer’s prose texts. The occurrences of WPs and the normalized frequency per 1000 words and per 100 lines were calculated and summarized in the following table:

152

Akinobu TANI

Table 1. Frequency of WPs in Chaucer’s Verse and Prose Texts # of WPs Freq/1000Wds Freq/100 LL GP Kn Mil Rv Co Verse Tot.

130 334 90 54 12 620

19.6 19.4 15.6 14.1 15.4 18.1

15.2 14.8 12.1 11.5 12.2 14.0

Total wds 6646 17193 5759 3838 777 34213

Mel

415

24.5

16933

Pars

493

16

30742

Bo

1044

17.9

58379

Astro

106

7.1

14866

Prose Tot.

2058

17

Total lines 858 2250 746 470 98 4422

120920

The results demonstrate that the average frequency of WPs per 1000 words in the verse (18.1) is slightly higher than that in the prose (17). Average frequency shows, however, not so much discernible differences in the WPs between Chaucer’s verse and prose. The differences in average frequency among verse and prose, of course, cannot be taken to mean that Chaucer’s verse style is stylistically more elevated than his prose style. Therefore, the frequency data alone cannot be the basis for a definite conclusion. Average frequency is thought to be mainly concerned with quantitative aspects of the WPs. Thus, slight differences in frequency of the WPs between Chaucer’s verse and prose may mean that the differences are not quantitative, but qualitative. There is some variation in frequency among the individual verse texts, though it is not so great as that among prose texts. Tani (forthcoming b) has revealed that the difference of frequency among individual prose texts of Chaucer is attributed to the differences in genre. Therefore, such variation among the verse texts may be able to be identified as reflecting different genres. The issue of variation in frequency among the verse texts will be discussed later. 3. ETYMOLOGIES AND ORDERS OF WPs IN CHAUCER’S VERSE AND PROSE The WPs in Chaucer’s verse and prose were analyzed in terms of the etymologies and orders of the members of the WPs. The major types of the WPs in the verse are:

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

153

OE + OE: And gladly wolde he lerne and gladly teche. GP 308 For which Dyane wroghte hym care and wo. Kn 2072 Ful ofte he seide “Allas and weylawey”, Mil 3602 This millere into toun his doghter sende / For ale and breed, Rv 4137 OE + OF: To riden out, he loved chivalrie, Trouthe and honour, fredom and curteisie. GP 45-6 Among the goddes hye it is affermed, / And by eterne word writen and confermed, Kn 2349-50 And thus lith Alison and Nicholas, / In bisynesse of myrthe and of solas Mil 3653-4 Dauncen he koude so wel and jolily. Co 4370 OF + OE: Wo was his cook but if his sauce were / Poynaunt and sharp, and redy al his geere. GP 351-2 Ful prively two harneys hath he dight, / Bothe suffisaunt and mete to darreyne / The bataille in the feeld bitwix hem tweyne; Kn 1630-2 With compaignye, hym to disporte and pleye Mil 3660 And gan to crie “Harrow!” and “Weylaway!” Rv 4072 OF + OF: He koude songes make and wel endite, / Juste and eek daunce, / and weel purtreye and write. GP 95-7 For whom ye have this strif and jalousye Kn 1834 As clerkes ben ful subtile and ful queynte Mil 3275 In purpos was to maken hire his heir, / Bothe of his catel and his mesuage, Rv 3978-9

All the members of WPs from the verse and prose were put to analysis according to their etymologies and orders, based on the OED2. The results are summarized in the following table:

154

Akinobu TANI

Table 2. Etymologies of WP Components in Fragment I & Prose Texts 1st 2nd OE OE OE OF OF OE OF OF Others Total 1st

2nd

OE OE OE OF OF OE OF OF Others Total

GP Kn 49 (37.7%) 132 (39.5%) 27 (20.8%) 57 (17.1%) 19 (14.6%) 57 (17.1%) 30 (23.1%) 72 (21.6%) 5 ( 3.8%) 16 ( 4.8%)

Mil 47 (52.2%) 12 (13.3%) 14 (15.6%) 10 (11.1%) 8 ( 8.8%)

Rv 36 (66.7%) 8 (14.8%) 4 ( 7.4%) 2 ( 3.7%) 4 ( 7.4%)

130

334

90

54

Mel

Pars

Bo

Astro

90 (21.7%) 135 (27.4%) 105 (25.3%) 107 (21.7%) 72 (17.3%) 74 (15 %) 130 (31.3%) 151 (30.6%) 18 ( 4.3%) 26 ( 5.3%) 415 493

225 (21.6%) 245 (23.5%) 195 (18.7%) 353 (33.8%) 26 ( 2.5%) 1044

59 (55.7%) 7 ( 6.6%) 3 ( 2.8%) 36 (34 %) 1 ( 0.9%) 106

Co 7 (58.3%) 3 (25 %) 1 ( 8.3%) 1 ( 8.3%) 12

Total (Av.) 271 (43.7%) 107 (17.3%) 95 (15.3%) 115 (18.5%) 33 ( 5.3%) 620 Total 509 (24.7%) 464 (22.5%) 344 (16.7%) 670 (32.6%) 71 ( 3.5%) 2058

Table 2 shows that the most common etymology and order of members of WPs in the verse texts as a whole is OE + OE followed by OF + OF, then OE + OF and OF + OE. The percentage of OE + OE in the total of Fragment I amounts to 43.7%, while that of OF + OF 18.5%. In contrast, the most common etymology and order of members of WPs in the prose texts is OF + OF followed by OE + OE, then OE + OF and OF + OE. The percentage of OE + OE in the total of prose texts amounts to 24.7%, while that of OF + OF is 32.6%. The results reveal, therefore, that there is a stark contrast, in terms of etymologies and orders of members of WPs, between the verse and prose. Chaucer’s verse texts are in favor of the OE + OE type of WPs, while his prose texts the OF +OF type. Diagrammatically, the findings can be put as follows: OE + OE: verse > prose OF + OF: verse < prose

I think that the findings demonstrate the qualitative differences in the use of the WPs in Chaucer’s verse and prose because the etymologies and orders of WPs are thought to reflect, though indirectly, the nature of the texts in question. Such differences may be interpreted in terms of several causes: lexical, metrical, or functional. Actually, these causes are too intertwined to separate clearly. Functionally, the findings may be interpreted as reflecting the differences between verse and prose. That is to say, verse is constrained by meter and rime in contrast to prose. This means that, in the verse, every word should be properly put into the metrical and riming structure of a line.

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

155

Lexically, this means that words of OF origin are generally longer and therefore harder to handle and fit into metrical structure of a line, while native words are shorter and therefore easier to fit into the metrical structure. It follows that the higher frequency of OE + OE WPs might be explained as deriving from such metrical facility as well. This issue is also connected with rime in that, as Masui (1964:169) pointed out, some WPs are placed in the riming position. Furthermore, as Masui pointed out, stylistically or generically speaking, the predominance of the OE + OE type in the verse might be thought to come from the phraseological affinity of stock phrases with ME verse romances, many of which are made of native words only. I will come back to this problem later. 4. REPETITION OF THE SAME WPs IN CHAUCER’S VERSE AND PROSE Here I examine the repetition of an identical WP except for the conjunction in the individual texts because Chaucer’s individual texts in the verse and prose respectively are complete by themselves so that it is pointless to examine the repetition of WPs in the entire verse or prose texts instead of that in the individual ones. The results of such repetition in Chaucer’ verse and prose are summarized in Table 3: Table 3.

Repetition of WPs in Chaucer’s Verse and Prose

Verse Texts Frequency of Repetition 11 6 3 2 Total # of All WPs Ratio of Repeated WPs Prose Texts Frequency of Repetition 13 10 6 5 4 3 2 Total # of All WPs Ratio of Repeated WPs

#

GP Tot.

3

6 6

130 4.62% #

Mel Total

4 16 3 9 37 74 44 99 407 24.32%

Kn Tot. 1 11 1 6 1 3 8 16 36 334 10.78%

#

#

Pars Total 1 0

1 0 2 0 2 32 5 45 22 1628 33 1705 481 19.96%

#

Mil Tot.

3

Bo Total

2 12 1 5 3 12 8 24 51 102 65 155 967 16.03%

Co Tot.

1 3

6 6

90 6.60% #

Rv #

Astro Total 1 1 1

Tot.

3 6 9

54 16.67% #

#

12

3 36 5 120 11 173 100 40%

0 12 0.00%

156

Akinobu TANI

It is quite notable that the repetition of WPs in the verse is very sparse in contrast to that in the prose. It is difficult to determine whether or not the repetition of WPs is intentional. As far as Tani (forthcoming a) shows, such repetition in Caxton’s Reynard cannot be said to be intentional rather than arbitrary: like the other lexical items, WPs related to the subject-matter may be repeated without any intention of emphasizing the idea. If this is the case, the repetition of WPs is expected to occur like that of single words. Much less repetition in the verse, however, might be interpreted as the use of the WPs being motivated by not semantic but the other reasons. Anyway, the findings of the repetition suggest that the functions of WPs in the verse are substantially different from those in the prose. The repetition in the Kn and the Rv is outstanding in Fragment I. First, the small number of WPs in the Rv may distort the high ratio of repeated WPs in the tale. In contrast, that in the Kn is remarkable even compared to the prose texts: i.e., the ration of repeated WPs is high, and two types of WPs are repeated six and eleven times respectively. The two types of repeated WPs in the Kn are to and fro (six times) and up and down (eleven times). Even these WPs cannot be said to contribute semantically to the contexts where they occur. Actually, all the examples of the two kinds of WPs are used as rime words except for one example of up and down: And with that word Arcite gan espye / Wher as this lady romed to and fro (:so) Kn 1112-3 This world nys but a thurghfare ful of wo, / And we been pilgrymes, passynge to and fro Kn 2847-8 Ther as this fresshe Emelye the shene / Was in hire walk, and romed up and doun (:Palamoun) Kn 1068-9 To telle yow al the descripsioun. / Depeynted been the walles up and doun Kn 2053-4

The type of WP up and doun in the riming position rimes with French loanwords ending in -oun except for the above example riming with a proper noun Palamoun and for one non-riming example. This strongly suggests that one of the major reasons for the use of the WPs in Chaucer’s verse derive from the use of them as rime phrases. This possibility is examined later. The sparse repetition of the same WPs in Chaucer’s verse, actually, exhibits a striking contrast with ME verse romances. Shimogasa (1997:72) made comments on the repetition of WPs or what he calls binomials in Le Morte Arthur, saying that “t[T]he repetition of binomials consisting of simple or plain words is ubiquitous throughout our romance as well as in other metrical or tail-rime romances”. We have already seen that, except for the two types of WPs in the Kn, Chaucer does not repeat so many WPs in his

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

157

verse. This is considered to be related to qualitative differences between Chaucer’s verse texts and ME verse romances. 5. MORPHOLOGY OF MEMBERS OF WPs IN CHAUCER’S VERSE AND PROSE Since the difference in the use of WPs between verse and prose was discerned in terms of the etymologies and orders of members of WPs, the members of WPs are analyzed in more detail. For the purpose, the morphology or form of members of the WPs is examined in order to explore what qualitative differences exist between Chaucer’s verse and prose. By morphology, I mean 1) word lengths of WP members and 2) stylistically elevated loanwords with distinctive Franco-Latin affixes. 5.1 Word Length of WP Members In order to examine the word lengths of WP members, I counted the number of syllables of members of WPs. The results are put in Table 4. Table 4. Word Length of WP Members 1st mem 2nd mem GP KN Mil Rv Co Verse

1.52 1.59 1.52 1.38 1.54 1.51

1.67 1.72 1.59 1.44 1.54 1.59

Mel Pars Bo Astro Prose

2.01 2 2.19 1.76 1.99

2.21 2.13 2.35 1.91 2.15

The results clearly show that the word lengths of the members of WPs are shorter in the verse than in the prose. In addition, the first members are shorter than the second ones, incidentally confirming the validity of Behaghel’s (1909:139) “Gesetz der wachsenden Glieder” or Behaghel’s law of increasing elements. According to the law, there is rhythmic “tendency to move from shorter to longer sentence elements” (Translation mine).

158

Akinobu TANI

5.2 Stylistically Elevated Affixes of Franco-Latin Loanwords Burnley (2000:237) states that a substantial body of Franco-Latin vocabulary “which . . . would for most people give an impression chiefly and simply of stylistic elevation” in Chaucer’s language can be distinguished by a set of affixes rather than mere etymology. Such affixes include -al, per-, in-, -able, -ible, -aunt, -yn, -ive, -ous, -ate, -ion , -uncy, -ent, and -ency, all of which make the word structure complex and the affixed words longer and more outstanding. Some of the WPs containing the members with such affixes are as follows: Verse: For it was of no superfluitee, / But of greet norissyng and digestible. GP 436-7 And shortly, turned was al up so doun / Bothe habit and eek disposicioun Kn 1377-8 This is youre ende and youre conclusioun. Kn 1869 The vertu expulsif, or animal Kn 2749 Prose: Of this comth poverte and destruccioun, bothe of spiritueel and temporeel thynges. Pars 722 for I putte me hoolly in youre disposicioun and ordinaunce. Mel 1727 For wit and ymaginacioun seyn that that that is sensible or ymaginable, it ne mai nat ben universel Bo5pr5.47-8

In order to ascertain the difference in the use of WPs between the verse and the prose, I searched for the members of the WPs with such affixes. The results are summarized in Table 5: Table 5. WPs with Stylistically elevated Members Texts # (%) 1st mem 2nd mem GP Kn Mil Rv Co

4 ( 3 %) 13 ( 3.9%) 1 0 0

5

4 10 1

Total WPs 130 334 90 54 12

Vers tot.

18 ( 2.9%)

5

14

620

Mel Pars Bo Astro

71 (17.1%) 82 (16.6%) 144 (13.8%) 18 (17 %)

35 28 68 5

43 61 92 14

415 493 1044 106

Prose tot.

318 (15.5%)

139

210

2058

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

159

The examination of this type of affixation shows the different nature of the WPs between the verse and prose more clearly than that of word length. That is, the OF loanwords employed in the verse are normally shorter, thus more fused into the contemporary language of Chaucer’s time than those in the prose, while those employed in the prose are generally more learned words. Such learned words contribute not only to the stylistically elevated impression, but also semantically to the contents of the text. These affixes are means to make words more outstanding. Then, if Chaucer had really wanted to make a text look more stylistically elevated, he could have taken the option of using WPs with such affixes in the verse as well. He did not, in fact, choose such an option. Though I do not deny the possibility that Chaucer wanted “to heighten or strengthen the effect of the style” in the verse as Jespersen (1912) claimed, it seems better to think of some other reasons for Chaucer’s use of WPs in the verse. The use of stylistically elevated members in the verse are almost confined to the GP and the Kn, which will be examined in section 7.3. 6. WPs IN RIME POSITION IN CHAUCER’S VERSE As we have seen that the use of WPs in the verse seems to be motivated other than semantic reasons, here I examine the possibility that the WPs are used as rime phrase. Though Masui (1964) suggested such a possibility, it is necessary to examine how many of the WPs are used as rime phrases in Chaucer’s verse in order to pin down the function of WPs in such texts. Before examining the possibility of WPs as rime phrase, it should be pointed out that almost all the WPs in Chaucer’s verse texts are used in either the riming or non-riming positions. In other words, very few WPs are used in enjambment or “t[T]he separation of even closely connected elements of a sentence by the conclusion of the metrical line” (Brink 1901:226). I found four exceptional WPs in enjambment such as: I putte me in thy proteccioun, / Dyane, and in thy disposicioun. Kn 2363-4 But oonly for the feere thus hath she cried, /And weep that it was pitee for to heere. Kn 2344-5

The scant use of WPs in enjambment appears to make clear Chaucer’s awareness of WPs as a closely connected group, not just as a free combining syntactic structure. Even in the examples of the WPs in enjambment above, however, the first members are put in the riming position with the conjunction and put normally first in the next line. Incidentally, one noteworthy feature concerning the above examples is that these examples come only from the Kn, which shows its great difference from the other tales

160

Akinobu TANI

in Fragment I in the use of WPs. In contrast to such WPs in enjambment, almost all the WPs are either in the riming or non-riming position in Fragment I. Some WPs in the riming position are: Ne of his speche daungerous ne digne, / But in his techyng discreet and benygne. GP 517-8 As evere mote I drynke wyn or ale, (:tale) GP 832 Som wikke aspect or disposicioun / Of Saturne, by som constellacioun, Kn 1087 And hem vitailled, bothe trogh and tubbe, (:jubbe) Mil 3627 And ful of hoker and of bisemare. (:spare) Rv 3965

In contrast, there are WPs in the non-riming position as well: Wo was his cook but if his sauce were / Poynaunt and sharp . . . GP 351-2 For nature hath nat taken his bigynnyng / Of no partie or cantel of a thyng, Kn 3007-8 And she was proud, and peert as is a pye. Rv 3950

The number of the WPs in the riming and non-riming positions was counted and summarized in Table 6: Table 6.

Ratio of Riming and Non-Riming WPs Riming WP Non-riming WP % of Riming WPs

Total

GP Kn Mil Rv Co

89 256 72 31 9

41 74 18 23 3

68.5% 76.6% 80 % 57.4% 75.0%

130 334 90 54 12

Total

457

159

74.2%

616

The overall results show that some 75 % of the WPs in Fragment I are used in the riming positions. This I take to mean that one of the major reasons for the use of WPs in Chaucer’s verse is to use them as a riming device. Though Masui (1964) noted the use of some WPs as rime phrase, it was not clear how many of the WPs are employed as rime phrase. The above findings made clear the point. In section 5.2, no mention was made as to whether the WPs with stylistically elevated Franco-Latin members are in the riming or non-riming position. Here such WPs were examined in connection with rime. Of all the eighteen examples of the WPs in the verse containing members with such

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

161

Franco-Latin affixes, one is used in enjambment and another is used in the non-riming position, while the others (sixteen examples) are used in the riming position: And shortly, turned was al up so doun / Bothe habit and eek disposicioun Kn 1377-78 Hath stablissed in this wrecched world adoun / Certeyne dayes and duracioun Kn 2995-6 That lord is now of Thebes the citee, / Fulfild of ire and of iniquitee Kn 940-1

This fact, therefore, makes it more valid to interpret the use of WPs containing members with stylistically elevated Franco-Latin affixes as being partly motivated by the desire to put such WPs in the riming position. In addition, among sixteen examples of such WPs, no less than eight examples have the affix -ioun: conclusion (1), duracioun (1), composicioun (1), correccioun (1), corrupcioun (1), commissioun (1) and dispossicioun (2). This means that even the selection of the members with Franco-Latin affixes in the riming position seems to be constrained by the requirement of necessary rimes. In short, the use of the WPs containing the members with Franco-Latin affixes is also interconnected with putting WPs in the riming position, which corroborates the above claim of mine. 7. WPs IN INDIVIDUAL VERSE TEXTS OF FRAGMENT I So far, the emphasis was put on the differentiation of the WPs between the verse and prose. Here the focus is placed on the differentiation of the WPs in the individual verse texts and on the characterization of the individual verse texts via the analyses of the WPs. This is based on the concept that WPs can be a means to distinguish texts as Tani (forthcoming b) shows. 7.1 Frequency of the WPs in Individual Verse Texts The normalized frequency of WPs is different among the verse texts as Table 1 shows. Among the verse texts, the frequency decreases from the GP to the Rv, that is, 19.6 to 14.1. The texts are placed in frequency order as: GP = Kn > Mil = Co > Rv. What matters most here is that the frequency is lower in the Mil and the Rv than in the GP and the Kn. The Co is too small in text size to make any objective judgments. The frequency differences may be connected to different styles of these texts. If what I have pointed out concerning the relation of the frequency of WPs with style is right, then the frequency difference may be reflecting the stylistic differences, showing the degree of stylistic elevation in the order of GP = Kn > Mil > Rv.

162

Akinobu TANI

7.2 Etymologies and Orders of Members of WPs in Individual Verse Texts Table 2 shows that there is a difference especially in the proportion of the OE + OE type and the OF + OF type among the verse texts, and that the frequency of the OE + OE type seems to be in inverse proportion to that of the OF + OF type. Such inverse proportion is found among the GP, the Kn, the Mil and the Rv. The Co shows quite different patterns probably because its short text distorts the numerical facts. In the GP, the percentage of the OE + OE type is 37.7 % while that of the OF +OF type 23.1% In the Kn, the percentage of the OE + OE type is 39.5 % while that of the OF + OF type 21.6%. In contrast, in the Rv, the percentage of the OE + OE type increases to 66.7 % while that of the OF + OF type decreases to 3.7%. In the Mil, the percentage of both OE + OE and OF + OF types (52.2% and 11.1% respectively) are halfway between those in the GP, the Kn and the Rv. The findings can be put diagrammatically as: OE + OE: GP = Kn < Mil < Rv OF + OF: GP = Kn > Mil > Rv

Such a difference in the etymology of the WPs among these texts may be actually related to styles or genres of these texts. In section 7.1, I pointed out that the frequency of WPs decreases from the GP and the Kn through the Mil to the Rv. This pattern in frequency shows, actually, remarkable parallel to the proportional differences in both OE + OE and OF + OF WPs among these verse texts. To make it clearer, I repeat the diagrammatical representation of the frequency of the WPs and the proportion of different etymological types of the WPs: Frequency: GP = Kn > Mil > Rv. OF + OF: GP = Kn > Mil > Rv OE + OE: GP = Kn < Mil < Rv

Given these patterns, I believe it safe to claim that such parallel between the frequency and the etymological proportion of WPs reflects some stylistic phenomena. Furthermore, I have already discussed that the frequency of WPs has much to do with stylistic elevation. Then the parallel between the frequency of WPs and the ratio of the OF + OF type in the verse implies that the OF + OF type is stylistically more elevated than the OE + OE type in Chaucer’s verse as well. Consequently, my examination of the WPs among Chaucer’s verse texts in terms of frequency and etymology revealed that Fragment I texts can be

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

163

stylistically classified into three groups: the GP and the KN vs. the Mil vs. the Rv. This classification of Fragment I texts, in fact, has close affinities with the findings of Muscatine (1957).2 He posited three styles in the Canterbury Tales, that is, conventionalism, naturalism and the mixed style. Two versions of conventionalism comprise the Kn and the Clerk’s Tale. The three versions of naturalism comprise the Rv, the Wife of Bath’s Prologue and the Canon’s Yeoman’s Tale. The mixed style comprises the Mil, the Merchants Tale and the Nuns Priests Tale. Therefore, Muscatine’s tripartite classification of style in the CT exactly coincides with my findings. Muscatine’s analysis and discussion is based on rhetoric. Speaking of conventionalism, he (1957:173) states that “one of the most general attributes of Chaucer’s conventionalism, and most convenient point of entry into the critical issues it raises, is rhetoric”. He (1957:173) also points out the existence of “a number of standard artes, textbooks of letter writing and of poetry” which are “primarily concerned with the amplified, ‘high style’ ”. Given this point, the frequent employment of WPs in the Kn should be said to be natural because WPs are known to be one of the important medieval rhetorical ornaments or embellishments. At the same time, conventionalism means conforming to the convention. The Kn is a courtly romance as is usually acknowledged, then the tale should be in accordance with the necessary forms of the genre, one of which is the use of WPs. This reason is somewhat different from Masui’s (1964:169) claim that the use of rime phrases may be motivated by oral formulaic reasons. The Kn is different in the uses of WPs other than in frequency and etymology, as already pointed out: the repetition of WPs and the use of stylistically elevated affixes. In contrast to the other verse texts, the Kn contains two kinds of WPs repeated six and eleven times respectively: to and fro and up and down. These WPs do not contribute semantically to the contexts as I pointed out, but such repetition in combination with the high frequency of the WPs, like the other rhetorical devices, contribute to the pace of the tale which is “deliberately slow and majestic” as Muscatine (1956:177) discusses, because WPs slow down the pace of tales and contribute to the audience’s expectation of different genres in the verse. The Kn is unique as well in that it contains the members of WPs with stylistically elevated affixes, though limited in number as seen in section 5.2. In contrast, naturalism is free from such formal constraints of 2

Fisher (1973) advocates three styles of the Kn, the Mil and the Rv more clearly than Muscatine.

164

Akinobu TANI

conventionalism. Cooper (1989:115-6) states that the Rv “stresses the speed of narrative” and is simple in language. This is reflected by the spare use of WPs in the tale because WPs normally slow down the flow of speech as I said. Furthermore, Cooper (1989:116) adds that the tale contains “decisively lowest proportion of Romance-derived words of all the tales”, which is also seen in the lowest proportion (3.7%) of the OF + OF type. The proportion is almost equal to the non-use of that type of WPs. I assessed the styles of these verse texts through the analyses of the WPs in Fragment I independently from Muscatine’s contention. The correspondence corroborates, therefore, my findings. 7.3 Morphology of Members of WPs in Individual Verse Texts First, the word length of the members of WPs is considered in the individual verse texts. As Table 5 shows, the word length of both the first and second members of the WPs decreases from the Kn (1.59-1.72) through the Mil (1.52-1.59) to the Rv (1.38-1.44): that is, Kn > Mil > Rv. This tendency among the Kn, the Mil and the Rv is again parallel to the tripartite grouping of Fragment I texts. This reinforces, therefore, what I claimed concerning the relationship between the WPs and style in the last section. In addition, the shortness of the members in the Rv even among the verse tales is noteworthy in that it also reflects the style of the tale. Next, the WPs with stylistically elevated Franco-Latin affixes are considered. As already claimed, the use of the WPs with such Franco-Latin affixes is partly motivated by the desire to put such WPs in the riming position. As Table 6 shows, the Kn like the GP contains a handful of such members in stark contrast to the Rv with no such members, and to the Mil with just one such member. This contrasts the Kn, a text of conventionalism, to the Rv, a text of naturalism, and the Mil, a mixed style. The only WP with such an affix in the Rv is, however, very suggestive: And forth he gooth, jolif and amorous, (: hous) Rv 3355

The member amorous contains the affix -ous, making the WP more outstanding, though the word amorous, first attested in 1303 according to the OED2, is not as outstanding as the other members with stylistically elevated affixes. Notwithstanding, in the texture of the Rv with “decisively lowest proportion of Romance-derived words of all the tales” (Cooper, 1989:116), the word does stand conspicuous. This means that the member resonates more, helping characterize and impress the image of the person designated by he or Absolon. Therefore, there is rhyme and reason for this use of

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

165

amorous in the Rv, which is quite different from the use of the WPs with stylistically elevated Franco-Latin affixes in the other verse texts. 7.4 WPs in the Riming or Non-riming Position in Individual Verse Texts Section 6 showed the ratio of riming to non-riming WPs in the verse texts to be approximately 72:25, and the use of WPs as riming device to be one of the major reasons for using WPs in Chaucer’s verse. The difference among the individual verse texts is examined here. Table 6 shows that the Rv uses the least proportion (57.4%) of WPs in the riming position in contrast to the average of 74.7% among the texts in Fragment I. This contrasts the Rv with the other texts. Riming words echo well in the audience’s ears so that they are important with aesthetic and mnemonic functions. Therefore, the riming members of WPs in the individual verse texts in Fragment I were etymologically investigated in order to examine the nature of the WPs in individual verse texts. The results are summarized in Table 7: Table 7. Etymologies of Riming members of WPs Etym. of riming members OE

OF

Others

Total

GP KN Mil Rv Co

45 (50.1%) 139 (54.3%) 50 (69.4%) 23 (74.2%) 5 (55.6%)

43 (48.3%) 106 (41.4%) 19 (26.3%) 6 (19.4%) 4 (44.4%)

1 11 3 2

89 256 72 31 9

Total

262 (57.3%)

178 (33.3%)

17

457

The results show 1) the preponderance of riming OE members in the Mil and the Rv, and 2) the balanced use of OE and OF members in the riming position in the other verse texts. This is thought to reflect the different uses of WP in these two texts among the verse texts. The preponderance of OE members in the riming position in the Mil and the Rv may be linked to the styles of these tales, i.e. mixed style and naturalism respectively. That is to say, fewer OF members in the riming position in the two texts may help them sound more churlish, which fits the tone of both tales. 7.5 Alliterative WPs in Individual Verse Texts Since Elliot (1974) mentions WPs as the major pattern of alliterative phrases, alliterative WPs were searched and counted. Some examples are:

166

Akinobu TANI Withoute bake mete was nevere his hous, / Of fissh and flessh, and that so plentevous GP 343-4 “I, wrecche, which that wepe and wayle thus, Kn 931 she koude skippe and make game, / As any kyde or calf folwynge his dame. Mil 3260-1 Or hoord of apples leyd in hey or heeth. Mil 3262 Til in a dych they caughte hym atte laste. / Wery and weet, as beest is in the reyn Rv 4106-7

As the examples show, the majority of the alliterative WPs are of the OE + OE type, which totals thirty-nine or accounts for 62 percent. The results of such alliterative WPs are summarized in Table 8: Table 8. Alliterative WPs Allit WPs Total of WPs

%

GP Kn Mil Rv Co

11 29 16 7 0

130 344 90 54 12

8.5% 8.4% 17.8% 13.0% 0.0%

Verse

63

630

10.0%

Mel Pars Bo Astro

34 42 75 9

415 493 1044 106

8.2 8.5 7.1 8.5

Prose

160

2058

8.5

It is not certain whether or not the frequency of alliterative WPs in the CT is high from the results alone. In this connection, Koskenniemi (1968) found that the average frequency of alliterative WPs in all her data of Old and Early Middle English texts is 38%. Koskenniemi’s findings are based on religious texts mostly in alliterative prose tradition, and it is difficult to compare her results with the present ones. But Koskenniemi’s figure gives an approximate concept concerning the ratio of the alliterative WPs in Chaucer. Table 8 shows 1) that the percentage of alliterative WPs are a little higher in the verse than in the prose, and 2) among the verse texts, the ratio of alliterative WPs in the Mil and the Rv is half as much again as in the rest of Fragment I texts. In contrast, the use of alliterative WPs is much constrained in the GP and the Kn. The pattern of use of alliteration is parallel to the other tendencies of WPs except for the Mil. I interpret this to be

The Word Pairs in Chaucer’s Verse in Comparison with Those in His Prose

167

Chaucer’s conscious stylistic manipulation. The frequent use of alliterative WPs helps debase the styles of the Mil and the Rv given Chaucer’s denouncing reference to alliteration which says that “I kan nat geeste ‘rum, ram, ruf ’, by lettre” (ParsPro 43). 8. CONCLUSION The present study has shown that what are termed as WPs under the same designation are employed with a noteworthy difference between Chaucer’s verse and prose: i.e., the function of the WPs in the verse is more for rime than for semantics in stark contrast to that in the prose. The results reveal Chaucer’s awareness of rhetorical potential of WPs and his dexterity in using WPs as well. Furthermore, I have found, from the analyses of the WPs, that the WPs show differences in the individual verse texts of Fragment I, and that the verse texts in Fragment I can be classified into three groups: GP, Kn :: Mil :: Rv. This classification is corroborated by the tripartite styles advocated by Muscatine (1957) and Fisher (1973). Furthermore, this tripartite classification of the texts was also discerned by the morphology of the members of WPs and the use of alliterative WPs. REFERENCES Behaghel, O. 1909. “Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern”. Indogermanische Forschungen 25. 110-142. Benson, L. D., et al., eds. 1987. The Riverside Chaucer. Boston: Houghton. Brink, B. ten. 1901. The Language and Metre of Chaucer. 2nd ed. Rev. F. Kluge, trans. M. B. Smith. London: Macmillan. Burnley, D. 1986. “Curial Prose in England”. Speculum 61. 593-614. Cooper, H. 1989. The Canterbury Tales. Oxford: Oxford UP. Elliott, R. W.V. 1974. Chaucer’s English. London: Andre Deutsch. Fisher, J. H. 1973. “The Three Styles of Fragment I of the Canterbury Tales”. The Chaucer Review, Vol. 8, No. 2. 119-127. Jespersen, O. 1912. Growth and Structure of the English Language. Leipzig: B. G. Teubner. Koskenniemi, I. 1968. Repetitive Word Pairs in Old and Early Middle English Prose. Turku: Turun Yliopisto. Masui, M. 1964. The Structure of Chaucer’s Rime Words: An Exploration into the Poetic Language of Chaucer. Tokyo: Kenkyusha. Muscatine, C. 1957. Chaucer and the French Tradition: A Study in Style and Meaning. Berkeley: University of California Press. Shimogasa, T. 1997. “Binomial Expressions in Le Morte Arthur”. Bulletin of

168

Akinobu TANI

the Faculty of International Studies, Yamaguchi Prefectural University 3. 59-74. Tani, A. Forthcoming a. “Word Pairs or Doublets in Caxton’s History of Reynard the Fox: Rampant and Tedious?” Medieval English Literature —Torches from the Ancient World—. Oxford: Peter Lang. Forthcoming b. “The Word Pairs or Doublets in Chaucer’s Prose Texts”.

A Short Note on Movement and Control in the English Noun Phrase Junji HAMAMATSU 0. INTRODUCTION In this paper, I shall attempt to offer an account of the alternation between (1a) and (1b): (1)

a. b.

John’s attempt to gain attention (failed) The attempt of John to gain attention (failed)

The Agent DP, realised as subject in (1a), seems to be ‘demoted’ in (1b). To the best of my knowledge, the literature presents no satisfactory analysis of the alternation. In what follows, I shall argue that the movement approach to control à la Hornstein (1998, 1999) offers a neat explanation for the alternation between (1a) and (1b), augmented by the discussion of nP structure put forward by Carstens (2000) and Radford (2000). The organisation of this paper is as follows: Section 1 makes clear the problem that arises with respect to the Agent that appears to be ‘demoted’ within the noun phrase. Section 2 pinpoints the cause of the alternation. Section 3 deals with a problem raised by the non-occurrence of raising in the DP. Section 4 concludes the discussion. 1. A DEMOTED AGENT: THE PROBLEM At first blush, the most straightforward way to analyse (2a) is to assume that it involves control of PRO in the specifier of the embedded TP, as schematised in (2b): (2)

a. b.

John’s refusal to respond John’s refusal [TP PRO to respond]

Obviously, this is in parallel with the analysis of (3a) as (3b): (3)

a. b.

John refused to respond John refused [TP PRO to respond]

170

Junji HAMAMATSU

However, I will call into question the structure in (2b), given that (2b) is parallel to (4) in meaning:1 (4)

the refusal of John to respond

In (4), the Agent appears in the postnominal position. In analysing (4), we would have to regard the of-phrase as an adjunct controlling PRO in the specifier of the embedded TP. (5) illustrates this: (5)

DP D

NP

the

NP

PP

N

TP

refusal

PRO to ...

of John

This means that the of-PP is accepted as similar to the by-phrase in (6): (6)

the refusal by John PRO to respond

It is mistaken, however, to assimilate the Agent of-PP to the by-PP. The data in (7) through (9) conspire to indicate that the latter is an adjunct, whereas the former is not: (7) (8)

1

a. b. a. b.

the refusal to respond by John *the refusal to respond of John the refusal by the court on Oct. 3 to reconsider the decision the refusal on Oct. 3 by the court to reconsider the decision

Williams (1987) notes that the affinity between (2a) and (4) is only superficial, maintaining that the noun and the of-PP together form an appositive to the to-infinitive. He bases his argument on the gap between (i) and (ii): (i) The decision of the committee was to leave (ii) *The decision by the committee was to leave (Ibid.: 368) In his theory the Agent can only be demoted unless there is no Theme represented in the noun phrase. The contrast does not seem to be so general, however. Observe that replacing decision with refusal results in degradability: (iii) *The refusal of/by the committee was to accept the policy

A Short Note on Movement and Control in the English Noun Phrase (9)

a. b.

171

the refusal of the court on Oct. 3 to reconsider the decision *the refusal on Oct. 3 of the court to reconsider the decision

As shown by the pair in (7), the by-phrase can follow the to-infinitive, while the of-phrase cannot. On the other hand, the by-phrase can either precede or follow the adjunct PP in (8), but the of-phrase has to precede one in (9). 2. THE NOMINAL ARCHITECTURE AND THE THEORY OF CONTROL Essentially following Carstens (2000) and Radford (2000), let us assume that a ‘light noun’ heads ‘NP shell’ structure and take a nominal suffix to constitute n. The example in (10a) would form the structure in (10b) at some point in its derivation: (10) a. b.

John’s refusal of the offer nP DP John

n' n -al

NP N

PP

refus-

of the offer

The derived noun refusal is split into the base N and the suffix n, with the former raising to the latter. The N assigns Theme and Goal to the DP in [Spec, NP] and to the PP in N’s complement position, respectively. On the other hand, the n assigns an external θ−role to John in [Spec, nP]. If we assume that both D and n can assign genitive Case in the English DP, D and n assign genitive Case to the DP in [Spec, nP] and to the PP in N’s complement position, respectively. This is schematised in (11): (11) [DP [D' ‘s [nP John [n' -al [NP [N' refus- [PP of the offer]]]]]]]

Suppose further that D has an EPP feature in (11). This makes John raise to [Spec, DP], resulting in the sequence in (10a). Now consider the structure of (2a), repeated here as (12a). If the above line of analysis is correct, it would be drawn as (12b):

172

Junji HAMAMATSU (12) a. b.

John’s refusal to respond nP DP John

n' n -al

NP N refus-

TP DP PRO

T' T

νP

to

respond

John is given an external θ−role from the n'. The DP in [Spec, nP] controls PRO in the specifier position of the TP that is N’s complement. However, the instance in (2b), reproduced as (13), does not seem to fit the structure in (12b): (13) the refusal of John to respond

If we respect the similarity in meaning between (12a) and (13), John should be assigned an external θ−role from the light noun in (13) in the same manner as in (12a). Obviously, though, the Agent DP in (13) is located lower than [Spec, nP]. The only place for John in (13), it seems, is [Spec, TP] in the infinitival clausal complement. Since John is also an external argument of respond, it could originate in [Spec, νP] below the TP, moving to [Spec, TP] afterwards. This would be illustrated as (14): (14) [nP [n' -al [NP refus- [TP John [T' to [νP John respond]]]]]]

Thus, the movement of John to [Spec, TP] has driven PRO away. The verbal counterpart of (14) would be (15): (15) [νP [ν' ν [VP refused [TP John [T' to [νP John respond]]]]]]

Interestingly, this is what Hornstein (1998, 1999) proposed for control phenomena. If his theory is correct, the structure in (15) will end up with (16), which is the sentential analogue of (12a).

A Short Note on Movement and Control in the English Noun Phrase

173

(16) John refused to respond (= 3a)

Specifically, he argued that control involves movement, which would dispense with the need for PRO. His important assumptions include (17i-iii), which are adopted from Hornstein (1999: 78): (17) i. θ−roles are features on verbs. ii. A DP receives a θ−role by checking a θ−feature of a verbal/predicative phrase that it merges with. iii. There is no upper bound on the number of θ−roles a chain can have.

In (16) John originates in [Spec, νP], where it gets the first θ−role from respond. The DP then moves to the specifier position of the embedded TP. This creates the control effect. Next, it moves to the specifier of the upper νP, in which it receives the second θ−role from refused and has its Case checked by T. The derivation is schematised in (18): (18) [TP T [νP John ν [VP refused [TP John to [νP John ν [VP respond]]]]]

We could generalise this derivation to (14). Remember that n has an option to assign a Case, as it does in (11). Then John is assigned a genitive Case in [Spec, TP] in (14), which is realised as of. The derivation does not stop here, because n’s θ−feature remains to be checked. This means that the DP should further move to [Spec, nP], receiving a θ−role by checking n’s θ−feature. Thus, we get the structure in (19):2 (19) [nP John [n' -al [NP refus- [TP John [T' to [νP John respond]]]]]

With the merger of D and the nP, we get the sequence in (20). This does not reflect a correct surface word order, however: (20) *the (of) John refusal to respond

It is known, however, that there are more functional projections within the noun phrase. Also, movement of N to such functional heads has been proposed (Ritter 1988; Bernstein 1993; Cinque 1995). If we assume that some functional head F takes the nP as a complement and N moves to F, we will get the structure in (21):

2

For ease of exposition, I omitted in (19) the of that precedes John.

174

Junji HAMAMATSU (21) [DP the [FP refusal [nP of John [n' refusal [NP refus- [TP John [T' to [νP John respond]]]]]

Let us now return to the derivation of the sentence in (18). This is not the end of the derivation: John further moves to [Spec, TP] and checks T’s EPP feature. This results in the structure in (22): (22) [TP John [T' T [νP John ν [VP refused [TP John to [νP John ν [VP respond]]]]]]]

Other things being equal, its nominal counterpart in (14), reproduced here as (23), may well undergo the same derivation: (23) [nP [n' -al [NP refus- [TP John [T' to [νP John respond]]]]]]

John moves to [Spec, nP] to check n’s θ−feature, thus receiving an external θ−role. Suppose that instead of n, D has a Case feature and an EPP feature. Then the Agent DP receives a Case and moves to D’s specifier position, as illustrated in (24): (24) [DP John [D' ‘s [nP John -al [NP refus- [TP John to [νP John ν [VP respond]]]]]]

Notice that this derivation is completely in parallel with (22). To recapitalise, our line of approach accounts for the similarity between (2a) and (4), reproduced here as (25a) and (25b), respectively: (25) a. b.

John’s refusal to respond the refusal of John to respond

This has been made possible by assuming that they share the structure in (26) at some point in their derivation: (26) [DP D [nP John [n' -al [NP refus- [TP John [T' to [νP John respond]]]]]]]

The difference in the position of John, on the other hand, has been attributed to the presence/absence of an EPP feature in D and that of a Case feature in n. 3. ON THE IMPOSSIBILITY OF RAISING IN THE NOUN PHRASE In the last section, we capitalised on Hornstein’s (ibid.) premise that control in the DP reduces to movement of a nominal from the subject position of the embedded TP to the matrix [Spec, TP]. The gap between

A Short Note on Movement and Control in the English Noun Phrase

175

(27a) and (27b), however, poses a problem to the movement approach to control in the nominal: (27) a. b.

John’s refusal to respond *John’s appearance to be successful

Since Chomsky (1972), it has been well known that raising is prohibited in the noun phrase, though it is possible in its sentential counterpart in (28): (28) John appears to be successful

If control involves movement as in (29a), the question arises as to what hampers raising in (29b): (29) a. b.

[TP John(’s) refusal [TP John to respond]] *[TP John(’s) appearance [TP John to be successful]]

In (29a) John moves from the specifier of the embedded TP to that of the main TP. Obviously, the lack of movement in (29b) as contrasted with (29a) and (28) needs some explanation. I argue that the gap in grammaticality between (29a) and (29b) originates from the categorical status of the clausal complements. It is uncontroversial to take the embedded clauses in (29b) and (28) as TPs. By contrast, there is room for argument about the categorical status of the control predicate in (29a). Consider the example in (30): (30) Washington’s desire [for Israel to reach an agreement with the Palestinians]

The appearance of for in (30) points to the CP status of the clausal complement. This in turn suggests the possibility that the to-infinitival complement in (29a), albeit the absence of an overt complementiser, also constitutes a CP. In this connection, the bracketed part in (31a) has been argued to form a CP, by the analogy of (31b) (Bresnan 1972; Chomsky 1981; Snyder and Rothstein 1992; Boskovic 1997): (31) a. b.

John wants [Mary to dance] John wants very much [CP for Mary to dance]

This assumption is supported by the possibility of coordination in (32): (32) John wants [to sing] and [for Mary to dance]

176

Junji HAMAMATSU

Crucially, the same is true of N’s infinitival complements, as shown by the grammaticality of (33): (33) John’s refusal [to sing] and [for Mary to dance]

All this leads us to conclude that N cannot take a TP as its complement. Given that control predicates are uniformly CPs, nouns can select them as their complements. Since raising predicates always constitute TPs, by contrast, nouns cannot select them. The contrast is illustrated in (34a) and (34b): (34) a. b.

refusal [CP φ [TP John to respond]] *appearance [TP John to be successful]

More specifically, N’s clausal complement should involve full-fledged extended projections in the sense of Grimshaw (2005). This seems to hold true of N’s complements generally. Consider the gap between (35a) and (35b): (35) a. b.

John’s entry [PP into [DP the room]] *John’s entry [DP the room]

Suppose, following Grimshaw, that P and D together constitute N’s extended projections. The unacceptability of (35b) will then naturally originate from the lack of N’s topmost extended projection. All in all, the difficulties that the absence of raising within the noun phrase might seem to be present are only apparent. It reduces to the impossibility of nouns to take TP complements. 4. CONCLUSION Hopefully, this paper has established that the apparent demotion of the Agent boils down to the movement of the Agent from the embedded TP to the specifier of the nP. This necessitates the movement analysis of control, thus getting rid of PRO in the noun phrase. Also, the alternation between the prenominal Agent and its demoted analogue has been explicated in a natural way; they share the same structure, with the difference being attributed to D’s EPP and Case features. NOTE This paper grew out of discussions I had with Neil Smith over his 2005 paper during my sabbatical stay at UCL. This research was supported by a

A Short Note on Movement and Control in the English Noun Phrase

177

grant from Centre for Language and Culture at Senshu University. REFERENCES Bernstein, J. 1991. Topics in the Syntax of Nominal Structure across Romance, Doctoral dissertation, CUNY. Boskovic, Z. 1997. The Syntax of Nonfinite Complementation: An Economy Approach. Cambridge MA: MIT Press. Bresnan, J. 1972. Theory of Complementation in English Syntax, Doctoral dissertation, MIT. Carstens, V. 2000. “Concord in Minimalist Theory”. Linguistic Inquiry 31. 319-355. Chomsky, N. 1972. “Remarks on Nominalization”. Studies on Semantics in Generative Grammar. 11-61. The Hague: Mouton. 1981. Lectures on Government and Binding. Dordrecht: Foris Publications. Cinque, G. 1995. Italian Syntax and Universal Grammar. Cambridge: Cambridge University Press. Grimshaw, J. 2005. Words and Structure. Stanford: CSLI Publications. Hornstein, N. 1998. “Movement and Chains”. Syntax 2. 99-127. 1999. “Movement and Control”. Linguistic Inquiry 30. 69-96. Radford, A. 2000. “NP Shells”. Essex Research Reports in Linguistics 33. 2-20. Ritter, E. 1988. “A Head-Movement Approach to Construct-State Noun Phrases”. Linguistics 26. 909-929. Smith, N. 2005. The. Paper Presented at SOAS. 22 February. Snyder, W. and S. Rothstein 1992. “A Note on Contraction, Case, and Complementizers”. The Linguistic Review 9. 251-266. Williams, E. 1987. “English as an Ergative Language: The Theta Structure of Derived Nouns”. Papers from the Thirteenth Regional Meeting. Chicago: Chicago Linguistic Society. 366-375.

178

Junji HAMAMATSU

Coordinating and Subordinating Conjunctions in Spoken American English Yoko IYEIRI,1 Michiko YAGUCHI and Yasumasa BABA 0. INTRODUCTION The present paper discusses the employment of coordinating and subordinating conjunctions in spoken American English, paying special attention to their uses in different styles of professional English in conferences and meetings. The dataset investigated for this purpose is the Corpus of Spoken Professional American English (CSPAE),2 which includes utterances recorded in four different types of conferences and meetings in the 1990s, all in the United States. While conjunctions are commonly used functional items in English, their occurrence in different discourse settings is still to be explored (cf. Greenbaum and Nelson 1995). Some studies argue that written English is more complex than spoken English, providing subordinating conjunctions more frequently (e.g. O’Donnell 1974, Kroll 1977, Chafe 1982), whereas others take entirely the opposite position by saying that spoken English is no less complex than written English (e.g. Halliday 1979). Some scholars argue that subordinating conjunctions are frequently encountered in informational style of English (e.g. Biber 1988: 109), whereas others claim that it is in spoken English that a higher proportion of subordinating conjunctions is attested (Beaman 1984). Furthermore, there are some studies which hold that coordinating conjunctions are commonly attested in informal style of English (e.g. Leech and Svartvik 2002: 191). In this manner, there is still much room for the investigation of the employment of different conjunctions in different styles or settings of English. 1. METHOD 1.1 Data CSPAE includes transcripts of over two million spoken words recorded in four different types of settings in the United States during the period from

1 2

Correspondence to: Yoko Iyeiri, Faculty of Letters, Kyoto University, Yoshida-honmachi, Sakyo-ku, Kyoto 606-8501, Japan. We have used the CD-ROM version of the CSPAE released in 2000 (© Michael Barlow).

180

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

1994 to 1998.3 The four settings are: (1) press conferences held at the White House and other locations (White House); (2) faculty meetings at the University of North Carolina (Faculty Meetings); (3) national meetings on mathematics tests (Committee on Mathematics); and (4) national meetings on reading tests (Committee on Reading). Although the four settings are all formal in their own right, the language of the White House is considered to be the most formal for its obvious publicity, and this is confirmed by some previous studies dealing with CSPAE. Iyeiri, Yaguchi, and Okabe (2004a), for example, investigate the contrast between different from and different than in CSPAE, and find that the latter form (which is often stigmatized in current English) is the least frequent in the White House. Interestingly enough, we could find not a single example of different than among female White House spokespersons, which suggests possible gender differences of language use within the same setting. By contrast, male speakers of the White House press conferences do use different than, though less frequently than male speakers of the other settings. As far as the analysis of different from and different than is concerned, the setting of Reading is the least formal, where not only male but also female speakers freely employ different than or the stigmatized variant. Thus, one can assume the existence of a possible scale of formality with the White House at the most formal end and Reading at the other end. The positions of the Faculty Meetings and Mathematics are less clear, at least in terms of the occurrences of different than. We would, however, place the Faculty Meetings closer to the White House and Mathematics closer to Reading in our working hypothesis, since the Faculty is similar to the White House in that both include a kind of “display talk”. By contrast, the meetings of Mathematics and Reading are much more interactive in nature, and the language used there is closer to natural conversation.4 We will borrow the terms “expository talk” and “exploratory talk” from Holmes (1992: 134-135) and state that the former is characteristic of the White House and Faculty Meetings, while the latter is a feature of the Mathematics and Reading files of CSPAE. Since the speakers of exploratory 3

4

According to Barlow (2000: 4-5), CSPAE is a corpus of relatively unedited transcripts of data recorded at different conferences and meetings held in the United States. It includes descriptions like “hesitations”, “false starts”, and “laughter”, but the treatment of some non-verbal expressions may be inconsistent. For instance, mmm-hmm is most certainly missing in it. This has indeed been made clear by some existing studies. Iyeiri, Yaguchi, and Okabe (2005), for instance, observe that the discourse marker like is much more frequent in Mathematics and Reading than in the other two settings.

Coordinating and Subordinating Conjunctions in Spoken American English

181

talk of the Mathematics and Reading meetings are involved in discussions in the authentic sense, their language tends to be much more conversational than the language used by the speakers of expository talk. In the exploration below, we will use a Core Corpus extracted from CSPAE, since the entire corpus is far too large for an analysis of the present kind. The Core Corpus consists of about 520,000 words, still providing a fairly large number of conjunctions. The coordinating conjunction and, for example, numbers around 15,000 even in the Core Corpus. Considering the possibility that the gender of the speaker affects the language use, we have split the Core Corpus into male and female utterances in analysis. The files of the corpus, on which the present study is based, are listed in Table 1. Each file consists of about 15,000 words:5 Table 1. The files of the Core Corpus of CSPAE Male files White House WH94Cm, WH95Cm, WH96aCm, WH96bCm, WH97aCm, WH97bCm Faculty Meetings FM95Cm, FM96Cm, FM97-8Cm Mathematics CM597Cm, CM697Cm, CM797Cm, CM897Cm, CM8a97Cm Reading CR6a97Cm, CR6b97Cm, CR797Cm

Female files WH94Cf, WH95Cf, WH96Cf, WH97Cf FM95Cf, FM96Cf, FM97-8Cf CM597Cf, CM69Cf, CM797Cf, CM897Cf, CM8a97Cf CR6a97Cf, CR6b97Cf, CR797Cf

In this table, the last letters in the lower case of each file name indicate the sex of the speakers: ms for “males” and fs for “females”. The Cs before them simply stand for the “Core Corpus”, while the numbers before them, following the practice of the original CSPAE, indicate either the year or the month plus the year when the conferences and meetings took place. Accordingly, “WH94Cm,” for example, stands for “male utterances of the White House in 1994”, and “CR797Cf” for “female utterances of the Committee on Mathematics data in July 1997”. 1.2 The Conjunctions Analyzed in the Present Study The conjunctions investigated in the analysis below are: (1) and, (2) but, (3) or, (4) as, (5) if, (6) because, (7) when, (8) though (although), (9) 5

The Core Corpus of CSPAE does not include utterances whose speaker’s sex is unknown. Thus, the utterances in it are classified either into “male” or “female”.

182

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

whether, and (10) before. While the ten conjunctions are all relatively frequent in the files listed in Table 1, the coordinating conjunction and is particularly frequent in the data of the present study.6 Since it is often pointed out in existing studies that and combining clauses and the same item combining clausal constituents behave in different manners, we will make a separate analysis between them: and1 combines clauses and and2 constituents in the following sections. Biber et al. (1999: 81-83), for example, notice that and combining clauses, but not and combining constituents, is relatively common in conversation. Chafe and Danielwicz (1987: 101-102) contend the same. Although the same distinction is theoretically possible with but and or, we have decided not to make a separate analysis in their cases between those combining clauses and those combining constituents. This is partly because the frequencies of but and or are much lower than the frequency of and, and partly because previous studies do not point out any differences between the two types of but or or, at least to the same extent as in the case of and. As for the remaining items in the above list, it is necessary to make some comments on as and before in particular. They both have dual functions as a conjunction and a preposition. To eliminate confusion, the present paper deals only with the cases where the finite verb is available after these items. Hence, (1) in the following is counted, but (2) is not in the present study: (1) (2)

So, as we saw it, we’re in a period of change. (FM95Cm) We were two faculty members working together, and, we brought high school students in as actors for the PSA. (FM96Cf)

In total, we have investigated 26,714 examples of conjunctions in the present research: 7,767 examples of and1, 7, 659 examples of and2, 3,130 examples of but, 2,184 examples of if, 1,158 examples of because, 944 examples of when, 156 examples of though (although), 2,070 examples of or, 1,161 examples of as, 138 examples of before, and 347 examples of whether.

6

In the preliminary version of this paper, which we delivered at the 23rd Meeting of the Japan Association for English Corpus Linguistics (i.e. Iyeiri, Yaguchi, and Okabe 2004b), we examined the seven frequently occurring conjunctions and, but, or, as, if, because, and when. Although the paper made a contribution in its own right, the conclusion was slightly monotonous, perhaps because the selected conjunctions were all too common. We have accordingly decided in the present paper to deal with the additional conjunctions of though (although), whether, and before, which are less commonly attested in English.

Coordinating and Subordinating Conjunctions in Spoken American English

183

1.3 Principal Component Analysis (PCA) In the discussion of the following sections, we will use the statistical method of PCA to see the overall relationships among the different conjunctions and relationships among the different settings. PCA is one of the most widely used and major methods of multivariate analyses, main features of which Jolliffe (2004: 1) summarizes in the following passage: The central idea of principal component analysis (PCA) is to reduce the dimensionality of a data set consisting of a large number of interrelated variables, while retaining as much as possible of the variation present in the data set. This is achieved by transforming to a new set of variables, the principal components (PCs), which are uncorrelated, and which are ordered so that the first few retain most of the variation present in all of the original variables.

As this passage explains, PCA is a method to reorganize the original variables and to extract a new set of variables, called principal components, by utilizing the mutual correlations of the original variables. The newly obtained axes are mutually orthogonal in nature. Since “the first few” of the newly obtained variables or the principal components “retain most of the variation present in all of the original variables” as Jolliffe (2004) states, it is a common practice in analysis to concentrate upon them, discarding the rest. In this sense, PCA is a method to reduce dimensionality as pointed out by the above quotation. If the first two principal components are significantly more important than the others are, for example, the discussion can focus upon them. Certainly, one can easily visulaize two vectors, and in this sense, PCA is a method to enhance the visualization of data. It is possible to present them in the form of a graph of two dimensions, which is again a common practice of PCA. For further details of PCA, see Woods, Fletcher and Hughes (1986) and Jolliffe (2004) among others. In the application of PCA, we have used the statistical package of SPSS with the varimax rotation. The newly achieved principal components are to explain the relationships among the different conjunctions defined in the previous section, which are the original variables. They are also to explain the relationships among the files (i.e. WH94Cm, WH95Cf, etc.) of the Core Corpus of CSPAE. 2. OVERALL TENDENCIES The present section provides an overall survey of the result obtained from the application of PCA. In the analysis, we first of all normalized the data of and1, and2, but, or, as, if, because, when, though (although), whether,

184

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

and before by calculating their frequencies per 1,000 words in each file. For detailed figures of the normalized data, see Appendix 1. We then conducted PCA by using the correlation matrix,7 and obtained a new set of variables or principal components. In this process, we decided to eliminate CM897Cm from analysis, since its behaviour is clearly exceptional and its inclusion deforms the overall relationships of the remaining files in the way it is contrasted too strongly with the rest, diminishing the differences among the remaining files. For some unknown reason, the use of whether is exceptionally frequent in CM897Cm (see Appendix 1), which affects the entire analysis of PCA.8 For reference’s sake, we have quoted a few examples of whether used in CM897Cm below: (3) (4)

In fact, on each of the items you indicate whether or not you used a calculator or did not use a calculator to answer the items. (CM897Cm) I just was curious about whether there was any consideration of that side of the political issue as well. (CM897Cm)

These are not at all peculiar examples, but whether is simply more frequent in this particular file than in the others. PCA conducted in this manner has yielded a new set of 11 principal components, of which the first two have turned out to be of particular importance. The contribution ratio of the first is 47.0%, while that of the second is 14.2%. In other words, the first two principal components account for more than 60% of the total variance of the original data. Hence, the following discussion concentrates upon them. Figure 1 shows the scatter plot representing the relationships among the files on the two vectors, which are mutually orthogonal in relation, while Figure 2 displays the relationship among the different conjunctions under review:

7

8

One can use either the correlation matrix or the covariance matrix in the application of PCA. For the difference between these two methods of PCA, see Woods, Fletcher, and Hughes (1986: 287-290) among others. Since the use of whether in CM897Cm is not restricted to a particular speaker, this is not a matter of “idiosyncrasy”.

Coordinating and Subordinating Conjunctions in Spoken American English 2

CM597Cf CM8a97Cm FM97-8Cf

1

CM797Cm CM8a97Cf FM95Cm CM597Cm CR6a97Cm CM897Cf CM697Cm

FM97-8Cm

WH97Cf CR6b97Cm FM96Cm FM96Cf

0 WH96aCm

Principal Component 2

185

CR797Cf

CR6a97Cf CM697Cf CR797Cm CM797Cf

WH97aCm FM95Cf WH95Cm WH96bCm WH94Cf WH97bCm WH95Cf

-1

CR6b97Cf

WH94Cm -2

WH96Cf

-3 -2

-1

0

1

2

Principal Component 1 Figure 1.

The scatter plot of the files of the four setting (CM897Cm excluded)

1.0 though whether

or_ because if

.5

but when

Principal Component 2

and2 0.0

and1

-.5

before

as

-1.0 -1.0

-.5

0.0

.5

1.0

Principal Component 1 Figure 2.

The scatter plot of the conjunctions under analysis (CM897Cm excluded)

Although detailed descriptions of the two axes or the two principal components are left to the next section, we would like to make a brief survey of the two graphs.

186

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

Figure 1, which displays the relationships among the files of the four settings of CSPAE, suggests that the gender of the speaker determines the first principal component, since female files are inclined to be located rightwards and male files leftwards on this axis. Male and female files are particularly distant in the case of Mathematics and Reading, whereas the distance is not as large when it comes to the Faculty and the White House. Thus, one could probably assume that gender differences are not so prominent in the case of relatively formal meetings, to which the White House and the Faculty belong. Turning to the second axis of Figure 1, we notice that files of formal conferences and meetings are likely to have negative scores and that the less formal meetings, to which the Mathematics and Reading files belong, tend to yield positive scores. It is, therefore, most feasible that this dimension represents the level of formality of the language used in the files. Therefore, files with negative scores are likely to be formal and those with positive scores are inclined to be informal. Further detailed discussions will be given in the section that follows. Figure 2, which indicates the relationships among the conjunctions selected for analysis, is also suggestive in a number of respects. As mentioned in the Introduction, it is often the case in existing studies to discuss coordinating conjunctions as opposed to subordinating ones in relation to different styles of language. Some studies are, for example, of the contention that coordinating conjunctions, rather than subordinating ones, are commonly attested in informal style of English (e.g. Leech and Svartvik 2002: 191), while others propose the opposite view (e.g. Beaman 1984). As far as Figure 2 is concerned, however, coordinating and subordinating conjunctions are not necessarily distributed in contrastive manners.9 On the first axis or the first principal component, for example, the coordinating conjunctions or, but, and and1 are located closely to subordinating ones like because, if, and when. The same is true of the second axis. Furthermore, Figure 2 reveals that different subordinating conjunctions display different behaviours in relation to the two axes. Additional detailed discussions on the distribution of different conjunctions will be given in the next section, where we interpret the meanings of the two axes. 9

As a matter of fact, Beaman (1984: 79-80) concludes her paper with the following sentences and suggests that the relationship between coordinating and subordinating conjunctions is not so straightforward: “… it [i.e. her study] has shown that the basic assumption of many linguists in the past, i.e. that subordination implies complexity, is false. The evaluation of syntactic complexity is simply more complex than that”.

Coordinating and Subordinating Conjunctions in Spoken American English

187

3. THE FIRST AND SECOND PRINCIPAL COMPONENTS As stated above, the White House is considered to be the most formal of the four settings of CSPAE, followed by the Faculty and then by the meetings of Mathematics and Reading, and this scale of formality seems to be well reflected in the language used in them. Since this is a stable scale in discussing the relationship of the four settings of CSPAE, we had expected before the application of PCA that it would be represented by the first axis or the first principal component. The distribution of the files in Figure 1, however, suggests that it is in fact the second axis that is indexical of the degree of linguistic formality. In this graph, White House files are inclined to have negative scores and meeting files, especially those of Mathematics, tend to take the opposite position on the second axis. The Faculty files are likely to be intermediate in this respect. On the assumption that the second principal component shows the scale of linguistic formality, we will discuss it first, and then the first principal component. Evidently, the conjunctions which characterize the formal files of the White House are as and before. See Figure 2, where these conjunctions have negative scores on the second axis. Returning to the original transcripts of the White House, we realize that they indeed characterize the formal meetings of the White House, typically occurring in some “set” expressions. As, for instance, tends to occur in expressions like as you know and as you can imagine. The following are some illustrative examples of this usage: (5)

(6) (7)

…As you know, the President met yesterday with Speaker Foley and Majority Leader Mitchell to begin discussing the best strategy to move forward wit health care reform. (WH94Cf) …The President, as you have heard many times, also believes that we need to change welfare as we know it, and he’ll talk about that. (WH94Cf) …As you’ve read in the paper, I think this isn’t the most important thing we could be talking about right now, but John Swofford assures me that this is an issue that never goes away. (FM95Cf)

By the same token, the conjunction before tends to be attested in some formal phraseologies, as in: (8)

Now, let me say before I turn over the floor to Ambassador Kantor one additional thing. (WH95Cf) (9) Anything else before we close off? (WH96aCm) (10) I think, though, we have so much work to do in solving so many of the open investigations that before we can make any fair generalizations about all of these cases as a class we probably—I probably have said too much. (WH96bCm)

188

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

In this manner, the second principal component of the present analysis, which is most likely related to the degree of the formality of language, does not display a simple contrast between coordinating and subordinating conjunctions but reveals the existence of some conjunctions typically used in formal conferences. As mentioned above, the relationship between the use of coordinating and subordinating conjunctions and style is a controversial issue, which is perhaps because the matter is not as straightforward as one is inclined to imagine. Indeed, it cannot be resolved by the second principal component of the present study, since as and before are more significant factors in relation to the White House files on this axis, and as a result, almost all the other conjunctions, both coordinating and subordinating, tend to be associated with the remaining files. In the list of normalized frequencies in Appendix 1, we observe that the coordinating conjunction but tends to be more frequent in the female files of Mathematics and Reading than the female files of the White House and the Faculty. This, however, seems to be an issue which is more appropriately discussed in association with the first axis or the first principal component, which is treated in the discussion hereafter. A look at Figure 1 reveals that the first principal component seems to be obviously related to the gender of the speaker, at least in respect of the settings of Mathematics and Reading, whose files are widely spread on this axis. More specifically, a number of female files are unequivocally dragged rightwards, presenting positive scores on this axis, whereas male files are inclined to be located leftwards. Although it is not an easy task to tell exactly what kind of axis the second principal component is, the combination of Figures 1 and 2 shows that female speakers, who tend to have positive scores on this axis, are inclined to use subordinating conjunctions of time, reasons, and conditions, i.e. when, because, and if. This is possibly a characteristic feature of exploratory conversation, where arguments can be quite intricate, providing reasons, conditions, etc. to interact with other speakers. Where exploration is indeed necessary, it is most probably imperative for speakers to utilize complex sentences of various types to attain the maximal effect and to persuade other participants in conversation. In the expository talk of the White House and the Faculty, by contrast, the emphasis is more likely placed upon clarity and straightforwardness. It is relevant in this connection to mention that Ford and Thompson (1986: 354) point out the more frequent use of if in spoken English than in written English, suggesting that its use is more typical of natural conversation. Examples of when, because, and if include:

Coordinating and Subordinating Conjunctions in Spoken American English

189

(11) Gary, when you use the term “leverage” though, I think people are going to assume that there is some funding tied to it in some way or that— (CR6b97Cf) (12) If it’s not important to put in there, I would say take it out because I think it’s going to cause more problems from those who read things like that in ways that we don’t intend than the good that it would cause in terms of just giving people a general description of what readers are about. (CR797Cf) (13) Gloria, if I may from a classroom teacher’s point of view, critically diagnostic though, could you use it, the feedback as far as any kind of where the student is in terms of comprehension or their skills? (CR6b97Cf)

As far as CSPAE is concerned, this feature of exploration in arguments is most typically found with female speakers of Mathematics and Reading, hinting at the possible existence of gender differences in these settings,10 whereas the differences are not at all prominent in more formal settings like the White House and the Faculty. It is also noteworthy that female speakers of Mathematics and Reading are characterized by the coordinating conjunctions or, but, and and1 as well. The contrast between the two types of and on the first principal component is especially worthy of note. And combining clauses presents a positive score, while the same conjunction combining constituents gives a negative one on this axis. This again shows the possibility that the type of language used in the exploratory talk typically shown by the female speakers of interactive meetings has shared features with natural conversations. We are reminded of the argument by various previous studies that and1 or and combining clauses is likely to occur in colloquial English (Biber et al. 1999: 81). Thus, the relationship among different conjunctions is not so straightforward as to present a clear contrast between coordinating and subordinating conjunctions. In this connection, it is relevant to mention that but and and1, both of which have positive scores on the first axis, are commonly attested at the beginning of turns in CSPAE, which is also a characteristic feature of natural conversation.11 Examples include:

10 11

Holmes (1992: 135) also considers that exploratory interactions, which are “essentially collaborative”, tend to characterize female talk. Kroll (1977: 96) among others points out that the conjunction and is often attested in turn-initial position.

190

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA (14) Mr M.: Ms G.: Mr M.: Ms G.: Mr M.:

Well, I think we overlooked it. I’m not sure if we overlooked it. I thought that was in there. Okay. But if we did, we can double check. Right. So that’s the viewpoint at this juncture between the summary which we think is accurate and the text which needs to reflect it. Ms S.: And if anyone comes across any other or some that they think might be a possibility that we didn’t capture it quite the way we thought it should be from the summary from the last meeting to the draft, just raise that issue at any point. (CR797)

(15) Ms H.: I would also like to just respond from the student’s point of view. You have to think in terms of fourth grade students are very much social people. And they want to be included with their peers…. So I think sometimes, you have to look at the student’s point of view, too, and the whole situation because you can’t kind of do a blanket thing with the students. Ms A.: And certainly, from the parents’ perspectives, they want to know how their child (Inaudible). And they’re going to assess the same standards (Inaudible) because they are going to be competing for the same jobs in the same universities and for the same economic opportunities. (CR6b97)

This phenomenon may be interpreted in various ways from discoursal perspectives.12 One possible interpretation, though not the sole interpretation, is that speakers as in (14) and (15) initiate their turns as a “continuation” of the preceding utterance, showing, perhaps unconsciously, their empathy with the previous speaker or the group itself. Obviously, this does not necessarily mean that the speakers initiating their turns with and1 and but always support the previous speaker’s opinion. But is a conjunction to change the direction of the argument, and therefore its use is quite possibly related to the introduction of opposing views. Still, the speakers opening their utterances with coordinating conjunctions, whether consciously or unconsciously, place emphasis on the association between their utterances and the utterances 12

In relation to and in general, Beaman (1984: 60-61) notes that “not all instances of and in the spoken data constitute true coordination”. In other words, they may be used with discoursal functions.

Coordinating and Subordinating Conjunctions in Spoken American English

191

before. In this manner different utterances are coordinated as a single unit of conversation. As shown in Figures 1 and 2 (see Section 2), this is a characteristic feature of female speakers in the meetings of Mathematics and Reading. To visualize the point more clearly, we have counted how frequently the coordinating conjunctions and and but are used in turn-initial position in the different files of the Core Corpus of CSAPE. See the two graphs below, which display the frequencies (per 10,000 words) of and and but used as a turn initiator.13 30 and but

25 20 15 10 5 0

CR6b97Cm

CR797Cm

CR6b97Cf

CR797Cf

CR6a97Cm

CM8a97Cm

CM897Cm

CM797Cm

CM697Cm

CM597Cm

FM97_8Cm

FM96Cm

FM95Cm

WH97bCm

WH97aCm

WH96bCm

WH96aCm

WH95Cm

WH94Cm

Figure 3.

Turn-initial and and but in male files (frequency per 10,000 words)

30 and but

25 20 15 10 5 0

CR6a97Cf

CM8a97Cf

CM897Cf

CM797Cf

CM69Cf

CM597Cf

FM97_8Cf

FM96Cf

FM95Cf

WH97Cf

WH96Cf

WH95Cf

WH94Cf

Figure 4.

Turn-initial and and but in female files (frequency per 10,000 words)

Since and1 and but are only two of the 11 variables used in the multivariate analysis, their frequencies in Figures 3 and 4 are not directly reflected in the distribution of the files in Figure 1. In other words, files 13

We have counted all examples of and in turn-initial position, but in practice they all illustrate and1 or the conjunction and combining clauses.

192

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

presenting high frequencies of and1 and but in Figures 3 and 4 are not necessarily located rightwards on the first principal component of Figure 1. Still, the tendencies shown in Figures 3 and 4 are largely in accordance with what we have so far argued in the interpretation of Figures 1 and 2. The frequent use of turn-initial and and but is a characteristic feature of exploratory conversations as represented by the meetings of Mathematics and Reading, and is particularly noticeable with female speakers. By contrast, the gender difference is not at all noticeable in more formal conferences like the White House, where this particular use of and and but is not at all frequent to begin with. Finally, we would like to make a brief comment on the relationship between the first and the second principal components so far discussed. From the statistical perspective, these axes are orthogonal and unrelated. However, the comment by Holmes (1992: 134) that expository talk is inclined to be associated with formal style is in keeping with our intuition. This is reflected in Figure 1, where the distribution of the files on the first axis and that on the second axis are in some ways similar. In both cases, White House files are likely to be contrasted with many of the meeting files of Mathematics and Reading. There are, however, clear differences between the first and second axes in the above analysis. The most prominent is the issue of gender, which is relevant only to the first principal component, on which male and female files of Mathematics and Reading are quite distant. By contrast, the gender difference is not at all attested in the formal setting of the White House or perhaps of the Faculty. 4. CONCLUSIONS We have hitherto discussed how the conjunctions and, but, or, as, if, because, when, though (although), whether, and before are used in different files of the Core Corpus of CSPAE. Of the axes extracted from the PCA analysis based upon the correlation matrix, the first two turned out to be highly significant: the contribution ratio of the first was 47.0% ad that of the second was 14.2%. While we had expected the first principal component to represent the degree of formality, it was in fact the second axis that was relevant to it. In relation to the second axis, the White House files (which are considered to be the most formal of the four settings of CSPAE) present negative scores and are quite distinctively separated from the files of the other meetings, particularly those of Mathematics. This seems to be ascribable to the fact that the conjunctions as and before are characteristically used in the White House, particularly in formal set expressions like as you know and before we close off.

Coordinating and Subordinating Conjunctions in Spoken American English

193

On the other hand, the first principal component seems to be relevant in some measure to the gender of the speaker, since female files tend to present positive scores on this axis and male files negative scores. Analyzing the distribution of the conjunctions on this axis, we have concluded that it is perhaps an axis linked to the exploratory feature of conversation. Female speakers, particularly in settings where natural conversation is possible, tend to utilize various conjunctions, both coordinating and subordinating, and present rather complex sentences. The use of conjunctions like when, because, and if is characteristic of this type of language, i.e. exploratory talk. Turn-initial and and but also characterize this type of talk. Their use is distinctively more common in the meetings of Mathematics and Reading in general than in the White House and the Faculty, but it is particularly typical of female speakers of Mathematics and Reading. This is in a way to coordinate utterances given by different people to create a single unit of exploratory conversation, though of course this does not explain all relevant cases of this phenomenon. The gender difference is, however, not really available when the setting is more formal. Hence, the settings of the White House and the Faculty are quite neutral on the first principal component. Finally, we discussed the relationship between the two principal components. Statistically, there is no correlation between them, but we feel that exploratory talk is inclined to be a feature of informal style. This is reflected in the fact that the distribution of the different files of CSPAE on the first axis and that on the second axis are in some ways quite similar. In both cases, the White House files tend to be separated from the meeting files of Mathematics and Reading. There are, however, significant differences between the two. The most noteworthy is the fact that the gender of the speaker matters in the first principal component, which is concerned with exploratory features. This is especially the case with settings where natural conversation is possible and where exploratory devices are well exercised. By contrast, the difference is not really observed in formal settings like the White House.

194

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

Appendix 1. Frequencies of conjunctions in the Core Corpus of CSPAE (per 1,000 words) and1

and2

but

or

as

if

because

when

though

wheth -er

before

WH94Cm

17.54

13.64

4.66

2.46

2.97

2.46

1.02

1.53

0.17

0.25

0.51

WH95Cm

13.09

16.99

5.05

2.24

3.10

3.27

1.21

0.86

0.17

0.23

0.23

WH96aCm

11.57

16.87

4.81

1.64

3.71

1.83

0.61

1.04

0.43

0.37

0.30

WH96bCm

12.67

18.70

5.67

3.14

4.16

1.81

1.03

1.09

0.12

0.48

0.42

WH97aCm

12.56

13.60

6.28

1.71

4.13

2.09

1.65

1.32

0.22

0.66

0.22

WH97bCm

14.22

14.22

5.38

2.39

3.26

2.72

1.46

1.79

0.20

0.47

0.40

FM95Cm

18.84

15.94

8.09

5.50

1.75

5.56

2.60

2.90

0.48

0.72

0.18

FM96Cm

17.43

16.13

4.57

4.06

2.99

4.12

1.86

1.64

0.39

0.56

0.28

FM97-8Cm

10.12

19.07

4.12

3.53

2.24

3.53

2.00

0.88

0.41

0.35

0.47

CM597Cm

14.84

16.01

6.35

4.80

1.95

3.40

1.78

2.68

0.56

0.39

0.17

CM697Cm

16.98

14.17

4.93

4.88

2.18

5.34

2.18

1.43

0.34

0.52

0.06

CM797Cm

15.00

13.06

7.03

4.87

1.44

4.04

2.55

1.66

0.33

1.33

0.11

CM897Cm

9.63

12.99

6.98

5.19

1.24

5.36

3.03

2.22

0.60

2.22

0.27

CM8a97Cm

10.72

18.06

4.02

4.26

2.01

4.14

1.25

1.19

0.44

0.88

0.31

CR6a97Cm

16.40

16.29

7.04

5.39

1.99

5.11

2.50

1.48

0.34

0.79

0.23

CR6b97Cm

18.45

15.34

5.98

5.11

2.07

5.34

2.53

1.72

0.34

0.69

0.34

CR797Cm

17.86

13.08

7.38

3.80

1.34

6.61

1.90

2.18

0.21

0.70

0.28

WH94Cf

11.97

15.13

7.68

2.26

3.39

1.49

0.30

0.83

0.12

0.42

0.24

WH95Cf

15.74

14.56

5.44

2.54

2.43

3.55

1.66

1.36

0.06

0.30

0.30

WH96Cf

14.49

12.59

4.83

1.99

5.78

2.84

1.61

1.89

0.09

0.38

0.47

WH97Cf

14.87

14.47

5.22

3.62

2.75

2.34

1.74

1.74

0.20

0.94

0.13

FM95Cf

16.38

15.31

3.16

4.17

2.50

4.71

1.31

1.67

0.06

0.18

0.12

FM96Cf

17.66

15.27

3.56

4.48

1.59

4.23

1.29

1.23

0.25

0.74

0.43

FM97-8Cf

13.60

18.11

5.86

3.52

2.46

2.99

1.82

1.23

0.35

0.88

0.12

CM597Cf

14.77

8.85

8.72

6.96

0.59

6.96

5.40

2.60

0.65

0.59

0.13

CM697Cf

15.82

11.29

9.39

4.20

1.18

7.29

3.55

2.76

0.13

0.79

0.20

CM797Cf

15.42

13.42

6.87

4.73

0.78

5.64

4.02

3.31

0.13

0.45

0.32

CM897Cf

8.58

11.56

7.79

4.69

0.97

5.66

3.23

1.58

0.30

0.79

0.24

CM8a97Cf

17.61

17.49

4.34

4.28

1.59

4.40

3.24

2.75

0.49

0.73

0.12

CR6a97Cf

17.86

12.93

6.91

4.03

1.34

4.55

3.52

2.75

0.38

0.77

0.51

CR6b97Cf

20.56

11.46

6.67

4.97

1.82

5.82

3.40

2.24

0.06

0.49

0.24

CR797Cf

16.31

12.25

7.50

4.81

0.75

4.94

3.94

2.94

0.38

0.94

0.31

Files

Coordinating and Subordinating Conjunctions in Spoken American English

195

Appendix 2. Turn-initial and and but in the Core Corpus of CSPAE (per 10,000 words) and

but

and

but

WH94Cm

Files

0

0

WH94Cf

Files

0.6

1.8

WH95Cm

0.6

1.1

WH95Cf

0

0

WH96aCm

0

0

WH96Cf

0.9

0

WH96bCm

1.2

0

WH97aCm

0

0

WH97Cf

2.0

0

WH97bCm

1.3

0.7

FM95Cm

3.0

0

FM95Cf

3.6

1.8

FM96Cm

0

1.7

FM96Cf

3.1

1.2

FM97-8Cm

0

0

FM97-8Cf

1.2

0

CM597Cm

3.3

6.1

CM597Cf

22.1

23.4

CM697Cm

9.8

3.4

CM697Cf

18.4

19.0

CM797Cm

1.7

3.3

CM797Cf

10.4

4.5

CM897Cm

1.1

3.8

CM897Cf

7.9

12.2

CM8a97Cm

2.5

1.3

CM8a97Cf

1.8

1.2

CR6a97Cm

11.9

4.0

CR6a97Cf

8.3

8.3

CR6b97Cm

12.1

9.2

CR6b97Cf

24.9

10.3

CR797Cm

23.2

11.3

CR797Cf

18.8

7.5

REFERENCES Barlow, M. 2000. The Manual of The Corpus of Spoken Professional American English. Houston: Athelstan. Beaman, K. 1984. “Coordination and Subordination Revisited: Syntactic Complexity in Spoken and Written Narrative Discourse”, in Coherence in Spoken and Written Discourse, ed. D. Tannen, 45-78. Norwood, New Jersey: Ablex Publishing Corporation. Biber, D. 1988. Variation across Speech and Writing. Cambridge: Cambridge University Press. , S. Johansson, G. Leech, S. Conrad, and E. Finegan. 1999. Longman Grammar of Spoken and Written English. Harlow: Longman. Chafe, W. L. 1982. “Integration and Involvement in Speaking, Writing, and Oral Literature”, in Spoken and Written Language: Exploring Orality and Literacy, ed. D. Tannen, 35-53. Norwood, New Jersey: Ablex Publishing Corporation. and J. Danielewicz. 1987. “Properties of Spoken and Written Language”, in Comprehending Oral and Written Language, ed. R. Horowitz. and S. J. Samuels, 83-113. San Diego: Academic Press. Ford, C. E. and S. A. Thompson. 1986. “Conditionals in Discourse: A Text-based Study from English”, in On Conditionals, ed. E. C. Traugott,

196

Yoko IYEIRI, Michiko YAGUCHI and Yasumasa BABA

A. T. Meulen, J. S. Reilly, and C. A. Ferguson, 353-372. Cambridge: Cambridge University Press. Greenbaum, S. and G. Nelson. 1995. “Clause Relationships in Spoken and Written English”. Functions of Language 2.1-21. Halliday, M. A. K. 1979. “Differences between Spoken and Written Language: Some Implications for Literary Teaching”, in Communication through Reading: Proceedings of Fourth Australian Reading Conference, Brisbane, August, 25-27th, 1978, ed. G. Page, J. Elkins, and B. O’Connor, 37-52. Adelaide, SA: Australian Reading Association. Holmes, J. 1992. “Women’s Talk in Public Contexts”. Discourse and Society 3.131-150. Jolliffe, I. T. 2004. Principal Component Analysis. 2nd edition. New York: Springer. Iyeiri, Y., M. Yaguchi, and H. Okabe. 2004a. “To be different from or to be different than in Present-day American English?: A Study of Style and Gender Differences Using the Corpus of Spoken Professional American-English”. English Today 20(3).29-33. 2004b. “The Corpus of Spoken Professional American-English niokeru Buntai to Gender nikakawaru Mondai: Setsuzokushi no Bunseki kara”. (Style and Gender Distinctions in the Use of Conjunctions: An Analysis of the Corpus of Spoken Professional American-English.) Paper presented at the 23rd meeting of the Japan Association for English Corpus Studies, Kyoto, 24 April 2004. 2005. “Gender and Style: The Discourse Particle like in the Corpus of Spoken Professional American English”. English Corpus Studies 12. 37-51. Kroll, B. 1977. “Combining Ideas in Written and Spoken English: A Look at Subordination and Coordination”, in Discourse across Time and Space, ed. E. O. Keenan and T. L. Bennett, 69-108. Los Angeles: Department of Linguistics, University of Southern California. Leech, G. and J. Svartvik. 2002. A Communicative Grammar of English. 3rd edition. London: Pearson Education. O’Donnell, R. C. 1974. “Syntactic Differences between Speech and Writing”. American Speech 49.102-110. Woods, A., P. Fletcher, and A. Hughes. 1986. Statistics in Language Studies. Cambridge: Cambridge University Press.

Complement Capacities in German: Three Types of Complements Yasuhito HOSAKA 0. INTRODUCTION In this paper, we will deal with infinitival complements, which have different sizes and which we will refer to herein as “complement capacities”. There are verbs that take only infinitival complements and verbs that take both infinitival and tensed complements. Complements consist of a proposition, a temporal element and modality. Some complements have a low capacity, e.g. only a proposition, and others have greater capacity, for example, a proposition and a temporal element, which means that the verb takes either an infinitival or a tensed complement. If the complement includes a modality, it yields the greatest complement size. In this paper, we will show the complement capacities have correlations with the syntactic properties shown in passivization in German. The passivizability of verbs with infinitival complements is dependent on some conditions, one of which is a well-known control condition. This paper shows that the control condition in German is not as strong as it is in English and that the hierarchy of complement capacities plays a role in passivization in German. 1. PASSIVE IN GERMAN Chomsky (1986) suggests that movement is a “last resort” and NP-movement applies to an NP if and only if the NP fails to get Case in situ. If the movement is a “last resort”, there must be languages that have no NP-movement. We have reason to allow the NP-movement if the NP moves overtly to TP-Spec. The analysis of English gives us two reasons: (1)

a. b.

There is no theta role in the subject position. An NP may move to the TP-Spec position without violating the theta criterion. The NP in the object position has no Case and it must move to the TP-Spec position in order to get (nominative) Case.

As Hosaka (1985) argues, German has no NP-movement according to the discussions by Lenerz (1977), Höhle (1978), den Besten (1985) and therefore there don’t seem to be any convincing arguments for the existence

198

Yasuhito HOSAKA

of passive movement in German. See Haider (2005). If we regard (1b) as a language specific movement, e.g. English, then what can we say about (1a)? If (1a) is a possible option in UG, we must think that there exists no NP in the TP-Spec position. According to the Minimalist framework by Chomsky (1995), which shed another light on language analysis, we can interprete the observations in (1) as (2): (2)

a. b.

In terms of passivization, German has weak features which trigger covert movement. In terms of passivization, English has strong features which trigger overt (category) movement.

A covert movement means movement of features (gender, numeral, case) and not the movement of the phrasal category NP itself. The analysis leads us to propose two sentence types. We will look at two examples in English for better illustration: (3)

a. b.

[TP The bookNOM was [VP read [NP e ] ] ] [TP [NP e ] was [VP read [NP the bookNOM ] ] ]

In (3a), the NP the book is overtly moved to the TP-Spec position, while the NP the book in (3b) is not, but has the nominative Case. We can say that the features are covertly checked in (3b). If the analysis is progressing correctly, it implies that the word order of VP in German doesn’t change as the examples in (4) show: (4)

a.

b.

Es wurde KarlDAT der GarausNOM It was Karl “Someone did in Karl.” Es wird sich beeilt! It is himself hurried. “Hurry up!”

gemacht. done in

The sentence (4a) shows that the NP Garaus stays in situ, while the construction with the NP-movement is awkward: (5)

??Der GarausNOM wurde Karl gemacht. “Someone did in Karl.”

The contrast between (4a) and (5) show us that the movement to

Complement Capacities in German: Three Types of Complements

199

TP-Spec may not appear in German and this property leads us to some peripheral examples as in (4b). The sentence (4b) is an example of passivization of a reflexive verb sich beeilen (hurry up). It is well understood that a verb with a reflexive pronoun is not passivizable, because the reflexive pronoun is not c-commanded by the antecedent - as (6) shows: (6)

*Gestern wurde sich gelobt. Yesterday was himself admired “*Himself was admired yesterday.”

The notable example in (4b) is only possible if the construction is based on the passive formation in (3b); that is, the passivization is well formed without changing any order in VP. The word order of VP is also reserved in passivization. There are, however, unacceptable sentences with verbs which take to-infinitival complements, as (7) show: (7)

a.

b.

*Karl wurde von Maria versprochen, pünktlich zu kommen. Karl was by Mary promised on time to come “*Karl was promised by Mary to come on time.” *Von den Kerlen wurde geglaubt, eine neue Chance zu haben. By the guys was thought, a new chance to have “The guys thought themselves to have a new chance.”

The ungrammaticality of the sentences (7) might be reduced to “control”, because the corresponding English sentences are also unacceptable, although we have no corresponding example for (7b): (8)

a. b.

*John was promised by Mary to come on time. *It was tried to fix the sink.

English provides many verbs with infinitival complements which cannot be passivized if the subject (henceforth PRO) in the complement is controlled by the subject in the matrix sentence of the active sentence. We illustrate one example: (9)

Maryi promised John PROi to come on time.

In (9), the subject in the complement, PRO, must be coreferent with Mary. The subscription means the coreference relation. The subject in the matrix clause controls PRO. We assume that a structural condition, say

200

Yasuhito HOSAKA

c-command, plays a role in control.1 In (9), Mary c-commands PRO and the control relation is preserved, while Mary in (8a) doesn’t c-command PRO and the control relation is not obtained. We find no controller in (8b), and the resulting structure is ill formed. However, the constraint concerning control does not seem to be an important constraint in German. See (10): (10) Es wurde (von Karl) versucht, den Wagen zu reparieren. It was (by Karl) tried the car to repair “*It was tried by Karl to repair the car.”

We will determine in this paper why (10) is not entirely unacceptable and thereby present a more precise characterization of complements. The notion of “capacity” plays an important role and the hierarchy of “capacity” described here corresponds to the X-bar-hierarchy. 2. THREE TYPES OF COMPLEMENTS X-bar theory attempts to show that all languages share certain structural similarities and X can signify not only a lexical category such as N, A, V, P, but also a functional category such as v, D, T, C.2 We will show one example: (11)

CP CP-Spec

C’ C

TP TP-Spec

T’ T

VP NP

V’ V

1

2

NP

We will give a definition for c-command : (i) A c-commands B if and only if A does not dominate B and every X that dominates A also dominates B. We will not elaborate this definition in this paper. We ignore v, D in this paper because there are no relevant discussions concerning these categories.

Complement Capacities in German: Three Types of Complements

201

This configuration shows a possibility whereby the infinitival complement could consist of three phrases, i.e. VP, TP and CP. 3 VP constitutes the propositional part, and the tense information and agreement information are added in TP (Tense Phrase). CP (The Complementizer Phrase) represents the illocutionary layer, sentence type and modality, therefore serving as an interface between a proposition and the discourse or a higher clause.4 The projections can be separated by intervening adverbials: (12) The new theory may possibly have indeed been badly formulated.5

As (12) shows, adverbs or adverbials describe or add to the meaning of a verb, an adjective, another adverb, or a whole sentence. This structure analysis concerning adverbs is not new: The four layers are proposed by Kajita (1968) in terms of complement sentences. Nakajima (1982) independently elaborates the assumption concerning adverbs, movement rules, relatives, transportability, questions, negation etc. We can apply this assumption to the X-bar system in (11) and assume the following types of infinitival complements: (13) a.

b.

c.

Er versucht, PRO den Wagen zu reparieren. (VP-type) He is trying PRO the car to repair “He is trying to repair the car.” Sie bat ihn, PRO den Wagen zu reparieren. (TP-type) She asked him PRO the car to repair “She asked him to repair the car.” Die Männer hofften, PRO dort Bodenschätze zu finden. (CP-type) The men hoped PRO there natural resources to find “The men hoped that they would find natural resources there.”

The underlined parts are infinitival complements and the complements in (3) are classified into three types: CP-type, TP-type and VP-type. The verb versuchen (try) takes only an infinitival complement while the verb bitten 3

4

5

One might logically question why VP is only a projection of lexical category and the other phrases projections of functional categories. As we noted in footnote 2, we don’t discuss the functional category v, but the existence of v is proposed not only from grammar theory, but also from language acquisition, as Murasugi, Hashimoto & Fuji (2007) shows. According to Rizzi (1997), we can assume the following structure: (i) [ForceP [TopP [FocP [TopP [FinP [IP [VP ]]]]]]] CP is split into four phrases, but this proposal has no bearing on our further discussion. Quirk et al.(1985: 495)

202

Yasuhito HOSAKA

(ask) takes both an infinitival complement and a tensed complement:6 (14) a.

b.

Sie bat ihn, den Wagen zu reparieren. She asked him the car to repair “She asked him to repair the car.” Sie bat ihn, dass er den Wagen reparieren möge. She asked him that he the car repair might “She asked him if he might repair the car.”

The verb hoffen (hope) also takes both an inifinitival complement and a tensed complement, but its complement contains modality: (15) a.

b.

Er hoffte, dort Bodenschätze zu finden. He hoped there natural resources to find “He hoped to find natural resources there.” Er hoffte, dass er dort Bodenschätze finden könnte. He hoped that he there natural resources find could. “He hoped that he could find natural resources there.”

The sentence (15b) shows the modal verb können in the conjunctive mood. The conjunctive mood exists covertly in infinitival constructions of the verb hoffen (hope). We have a set of verbs with a conjunctive mood: 6

There are, however, some speakers who accept the following sentence: (i) Er versucht, dass er rechtzeitig ankommt. He tries that he in time arrives “He tries to arrive in time.” The verb versuchen here has the same function as the verb wollen: (ii) Er will, dass er rechtzeitig ankommt. He wants that he in time arrives “He wants to arrive in time.” The existence of (i) lets us assume that some operation produces Er versucht rechtzeitig anzukommen after applying the deletion of dass er in (i). But this is not the case. There is a restriction concerning sentences of versuchen und wollen with infinitival complements as the following sentences show: (iii) a. *Er versucht, dass er das Buch liest. He tries that he the book reads b. *Er will, dass er das Buch liest. He wants that he the book reads The speakers who accept sentences (i) and (ii) don’t accept sentence (iii). The sentences are bad because the subjects in complement clauses have an agent role. If the subordinated subject has an agent thematic role, we then have no tensed complement. If this observation is true, we have no tensed clause underlying (13a). The sentence (i) must be analyzed from another point of view.

Complement Capacities in German: Three Types of Complements

203

(16) glauben (believe), erklären (explain), wünschen (wish), annehmen (assume), ahnen (suspect), fürchten (fear) ...

These verbs take tensed complements without complementizers: (17) a.

b.

Er glaubt, dass Maria heute kommt. He thinks that Mary today is coming “He thinks that Mary is coming today.” Er glaubt, Maria kommt heute. He thinks Mary is coming today “He thinks that Mary is coming today.”

Such verbs are often called “non-factive verbs” and have the same property in nominalized forms:7 (18) a.

b.

c.

die Hoffnung, dort Bodenschätze zu finden the hope there natural resources to find “the hope of finding natural resources there.” die Hoffnung, dass er dort Bodenschätze finden könnte the hope that he there natural resources find could “the hope that he could find natural resources there.” die Hoffnung, er könnte dort Bodenschätze finden the hope he could there natural resources find “the hope (that) he could find natural resources there.”

The sentences (18b,c) show that the complement selection corresponds to the selection in (17). Non-factive verbs select their complementizers optionally and its nominalized forms do the same. So far, we have seen three types of complements. 2.1 CP-type As Höhle (1978) shows, some passive constructions are not acceptable:8 (19) a.

7 8

die Kerle leugneten, ein Zelt zu besitzen. the guys denied a tent to possess “the guys denied that they possessed a tent.”

See Kiparsky & Kiparsky (1970) in detail concerning factive and non-factive verbs. The analysis shown there can be applied to German. See Höhle (1987:190)

204

Yasuhito HOSAKA b.

(20) a.

b.

(21) a.

b.

(22) a.

b.

(23) a.

b.

*von den Kerlen wurde geleugnet, ein Zelt zu besitzen. by the guys was denied a tent to possess “It was denied by the guys that they possessed a tent.” die Kerle hofften, dort Bodenschätze zu finden. the guys hoped there natural resources to find “The guys hoped that they would find natural resources there.” *von den Kerlen wurde gehofft, dort Bodenschätze zu finden. by the guys was hoped there natural resources to find “It was hoped by the guys that they would find natural resources there.” die Kerle gaben zu, nicht bis 3 zählen zu können. the guys admitted not to 3 count to be able to “The guys admitted not to be able to count to 3.” *von den Kerlen wurde zugegeben, nicht bis 3 zählen zu können. by the guys was admitted not to 3 count to be able to “It was admitted by the guys that they were able to count to 3.” die Kerle erklärten, zu Verhandlungen bereit zu sein. the guys explained for negotiations ready to be “The guys explained that they were ready for negotiations.” *von den Kerlen wurde erklärt, zu Verhandlungen bereit zu sein. by the guys was explained, for negotiations ready to be “It was explained by the guys that they were ready for negotiations.” die Kerle verlangten, endlich Nahrung zu bekommen. the guys demanded finally food to get “The guys demanded that they finally get some food.” *von den Kerlen wurde verlangt, endlich Nahrung zu bekommen. by the guys was demanded at last food to get “*It was demanded by the guys that they got food at last.”

We can see that the matrix verbs in unacceptable sentences belong to the CP-type. The CP-type complement has the modality, which comes from the matrix verb. As Postal (1970) already pointed out, the modal element can be deleted if the tensed complement is converted to an infinitival complement. Let’s return to (20a). The sentence is cited again: (24) die Kerle hofften, dort Bodenschätze zu finden.

Complement Capacities in German: Three Types of Complements

205

The sentence is in fact acceptable, but somehow “incomplete”. Let’s see the corresponding tensed complement: (25) die Kerle hofften, dass sie dort Bodenschätze ( fänden / the guys hoped that they there natural resources (findSUBJUNCTIVE /finden würden /finden könnten). /find would /find could) “The guys hoped that they would find natural resources there.”

The above sentence shows that the complement contains a modal expression, which appears explicitly in the verb forms: fänden, würden, könnten. As far as a tensed complement is concerned, the verb form reveals a certain modality, which is deleted, in a corresponding infinitival one. We can say that this incompleteness of CP-type is so strong that the passivization of CP-type is unacceptable. Compare (20b) (repeated here as (26a)) with (26b): (26) a.

b.

*von den Kerlen wurde gehofft, dort Bodenschätze zu finden. by the guys was hoped there natural resources to find “It was hoped by the guys that they would find natural resources there.” von den Kerlen wurde gehofft, dort Bodenschätze finden zu können. by the guys was hoped there natural resources find to be able to “It was hoped by the guys that they could find natural resources there.”

The passive sentence (26b) is much better than (26a) and we can say the former is acceptable. The modal expression können supports the acceptability of (26b). We have discussed for the relation of infinitival complements of CP-type and passivization. Although German shows apparently free passivization, complement types are, however, sensitive to the operation. 2.2 TP-type Many verbs belong to the TP-type verbs: factive verbs (bedauern, missdeuten, ablehnen ...), verbs of control (versprechen, bitten, vorschlagen...) The factive verbs can be passivized: (27) a.

Es wurde von Karl bedauert, Maria mit einem Messer angegeriffen It was by Karl regretted Mary with a knife attacked zu haben. to have “It was regretted by Karl that he attacked Mary with a knife.”

206

Yasuhito HOSAKA b.

c.

Es wurde von Karl missdeutet, sehr bewundert zu It was by Karl misinterpreted so much admired to “It was misinterpreted by Karl that he was admired so much.” Es wurde von Karl abgelehnt, das auf diese Weise zu It was by Karl refused it in this way to “It was refused by Karl that he do it this way.”

werden. be machen. do

There seems to be no problem with passivization, but we have many unacceptable examples in the case of control verbs: (28) a.

b.

*Karl wurde von Maria versprochen, pünktlich zu kommen. by Mary promised on time to come. KarlDAT was “*Karl was promised by Mary to come on time.” *Wir wurden von Karl angefleht, endlich sein eigenes Zimmer we were by Karl begged finally his own room zu bekommen. to get “*We were begged by Karl that he get his own room.”

The ungrammaticality can be explained in the following manner: The subject assumed in a complement, PRO, must be c-commanded by its antecedent. Let’s take an example in English: (29) *Mary was promised by Johni PROi to arrive punctually.

The index “i” shows the coreference relation. In (29), PRO is c-commanded (or controlled) only by Mary and a desired antecedent, John, cannot c-command PRO in (29), resulting in ungrammaticality. This is also the case in German: (30) a.

b.

*Karl wurde von Mariai versprochen, PROi pünktlich zu kommen. PRO on time to come. KarlDAT was by Mary promised “*Karl was promised by Mary to come punctually.” **Wir wurden von Karl angefleht, PROi endlich sein eigenes Zimmer we were by Karl begged PRO finally his own room zu bekommen. to get “*We were begged by Karl that he get his own room.”

PRO in both sentences are syntactically controlled only Karl or wir (we) which is not a desired controller. We have seen the type of subject control,

Complement Capacities in German: Three Types of Complements

207

and we now consider the other type of control: (31) a.

b.

Karl versprach Mariai, PROi das Haus verlassen zu dürfen. Karl promised Mary PRO the house leave to be allowed to “Karl promised Mary to be allowed to leave the house.” Karl bat Mariai, PROi das Haus zu verlassen. Karl asked Mary PRO the house to leave “Karl asked Mary to leave the house.”

The two sentences are examples of object control: The object NP Maria controls PRO. Maria can control PRO even if the sentence is passivized: (32) a. b.

Mariai wurde von Karl versprochen, PROi das Haus verlassen zu dürfen. Mariai wurde von Karl gebeten, PROi das Haus zu verlassen.

In both sentences, PRO is controlled by Maria. We have seen that the c-commanding relation plays an important role in the case of the passive. The fact that there is also a control phenomenon in English leads us to stipulate that the relation of control and passivization has a universal character. Consider further the universal character of control relation: The control relation is not involved in CP-type in section 2.1. Moreover, the relation observed in TP-type verbs, including factive verbs, has nothing to do with control conditions. Some might argue that the factive verbs are subject control verbs and cannot be explained for the same reason because the sentences (27) are acceptable, contrary to the assumption that the subject control verb cannot be passivized. Compare (33a) with (33b): (33) a.

b.

*Maria wurde von Karli versprochen, PROi pünktlich Mary was by Karl promised PRO on time “*Karl was promised by Mary to come on time.” Es wurde von Karli versprochen, PROi pünktlich zu It was by Karl promised PRO on time to “*It was promised by Karl to come on time.”

zu kommen. to come. kommen. come.

The sentence (33b) shows that the sentence becomes more acceptable if the c-commanding antecedent Maria doesn’t exist. If there is no possible antecedent c-commanding PRO, then PRO can be coreferent with NP in the matrix clause. This stipulation seems to be descriptively adequate enough to explain the grammaticality of (27).

208

Yasuhito HOSAKA

2.3 VP-type In the proceeding sections we have pointed out that there are some conditions on passivization of verbs with infinitival complements in German. To explain the conditions, we introduced three sizes of complement, VP, TP, CP. The CP-type verbs are not easy to passivize, as shown in section 2.1, while the TP-type verbs (section 2.2) are passivizable with fulfillment of control conditions. In this section, let us turn our attention to the case of VP-type of infinitival complements. The VP-type verbs are verbs with no tensed complementizers as (34) shows: (34) beginnen (begin), versuchen (try), wagen (risk), aufhören (cease),...

The following passive sentences are not particularly elegant with regards to style, but they are acceptable: (35) a.

b.

c.

d.

Es wurde von Karl begonnen, die Aufgabe zu It was by Karl begun the work to “Doing the work was begun by Karl to do the work”. Es wurde von Karl versucht, den Wagen zu It was by Karl tried the car to “Repairing the car was tried by Karl.” Es wurde von Karl aufgehört, den Wagen zu It was by Karl ceased, the car to “Repairing the car was ceased by Karl.” Es wurde von Karl gewagt, ins Wasser zu It was by Karl risked into the water to “Jumping into the water was risked by Karl.”

machen. do reparieren. repair reparieren. repair springen. jump

As Evers (1975) persuasively shows, we have clause union variations in VP-type:9 (36) Karl hat den Wagen [zu reparieren versucht].

The verbs in square brackets are one complex verb if the clause union occurred. The formation could be verb-raising in the sense of Evers (1975) or verb-stacking as explained by Haider (2005). VP-type is the smallest complement in our proposal and the verb in the complement is easy to attach to the matrix verb. If we have a verbal complex, we can get (37): 9

Evers (1975) argues that verbs can have the clause union, but we will not discuss the clause union of TP-type and CP-type.

Complement Capacities in German: Three Types of Complements (37) Der Wagen wurde von Karl the car was by Karl “*The car was tried to repair.”

209

zu reparieren versucht. to repair tried

3. CONCLUSION AND REMAINING PROBLEMS In the above discussion, we dealt with infinitival complements which have different capacities according to the matrix verbs. As suggested in section 1, German passives can be more freely generated than English ones because there is no overt movement to TP-Spec. See the following examples: (38) a.

b.

Gestern wurde auf seine Frau gewartet. yesterday was for his wife waited “Someone waited for his wife yesterday.” Es wurde gestern getanzt. It was yesterday danced. “They danced yesterday.”

The sentences (38) are impersonal passives, which means that we have any passive constructions in German as far as the active subject has an agent role. In this vein, the passivization of verbs with zu-infinitival complements ought to be possible, but this is not the case, as we have seen in section 2. Let us turn to the VP-type: It has a low capacity as (39) shows: (39)

VP PRO

V’ V

NP

The structure in (39) is English VP. The order V-NP is inverted in German. The VP-type, however, is not the smallest type. See (40): (40) a.

b.

Wir haben das Buch zu lesen. we have the book to read. “We have to read the book.” Du brauchst nur den Tisch zu reservieren. you need only the table to reserve “You only need to reserve the table.”

210

Yasuhito HOSAKA

These verbs take infinitival complements with zu, but the complements seem to be not the same as complements of versuchen (try): (41) a.

b.

weil er versuchte, das Buch because he tried the book “because he tried to read the book,” *weil wir haben, das Buch because we have the book “because we have to read the book,”

zu to

lesen, read

zu to

lesen, read

Infinitival parts are extraposed in (41), and only (41b) is ill-formed. The difference between (41a) and (41b) seems to be reduced to the property of sentential complement: VP-type in (41a) is sentential, but the complement in (41b) is not. If this is correct, then we can assume the V’-type for quasi-modal verbs. V’-type contains no subject, e.g. PRO. We leave this V’-type open, because this type is not passivizable and needs further research from another point of view. REFERENCES Chomsky, N. 1986. Knowledge of language: its nature, origin, and use. New York, Tokyo: Praeger. 1995. The Minimalist Program. Cambridge, MA: MIT Press. den Besten, H. 1985. The Ergative Hypothesis and Free Word Order in Dutch and German. In Toman, J. (Ed.), Studies in German Grammar. Dordrecht: Foris. Eisenberg, P. 2006. Grundriss der deutschen Grammatik. Band 2: Der Satz. 3., durchgesehene Auflage, Stuttgart, Weimar: J.B. Metzler. Evers, A. 1975. The transformational cycle in Dutch and German. Bloomington: Indiana University Linguistics Club. Haider, H. 2005. How to turn German into Icelandic - and derive the OV-VO contrasts. Journal of Comparative Germanic Linguistics, 8, 1-53. Höhle, T. N. 1978. Lexikalistische Syntax: die Aktiv-Passiv-Relation und andere Infinitkonstruktionen im Deutschen, Vol. 67 of Linguistische Arbeiten. Niemeyer, Tübingen. Hosaka, Y. 1984. Das Passiv in der deutschen Gegenwartssprache. MA thesis, Tokyo Metropolitan University. Kajita, M. 1968. A Generative Transformational Study of Semi-auxiliaries in Present-day American English. Tokyo: Sanseido. Kiparsky, P. & C. Kiparsky 1970. Fact, In Bierwisch, M. and K. E. Heidolph (eds.) Progress in Linguistics, 143-173. Lenerz, J. 1977. Zur Abfolge nominaler Satzglieder im Deutschen, Studien

Complement Capacities in German: Three Types of Complements

211

zur deutschen Grammatik. TBL-Verlag Narr. A revision of the author’s thesis. München, 1976. Murasugi, K., T. Hashimoto & C. Fuji. 2007. VP-shell analysis for the acquisition of Japanese intransitive verbs, transitive verbs, and causatives. Linguistics, 45, 615-651. Nakajima, H. 1982. The V4 System and Bounding Category. Linguistic Analysis, 9, 341-378. Postal, P. M. 1970. On Coreferential Complement Subject Deletion. Linguistic Inquiry, 1, 439-500. Quirk, R., S. Greenbaum, G. Leech & J. Svartvik. 1985. A Comprehensive Grammar of the English Language, Longman, London. Rizzi, L. 1997. The fine structure of the left periphery. In: Haegeman, L. (Ed.), Elements of Grammar. Amsterdam: Kluwer, 281-33. Růžiča, R. 1999. Control in grammar and pragmatics. A cross-linguistic study. Amsterdam, Philadelphia, Pa.: J. Benjamins.

212

Yasuhito HOSAKA

Index of Names Abney, S. 128 Allen, C. 130, 134-136, 141 Anderson, S. 96, 133, 141, 143 Askedal, J. O. 76, 82, 91, 103 Bar-Hillel, Y. 13, 14 Barlow, M. 179, 180 Barnes, M. 104 Beaman, K. 179, 186, 190 Bech, G. 95 Behaghel, O. 157 Bernstein, J. 173 Berwick, R. 69 Bessinger, J. B. 121 Biber, D. 179, 182, 189 Blake, N. F. 123 Bloomfield, L. 24, 38 Bolinger, D. 31, 32, 41, 42, 44, 45, 48 Boskovic, Z. 175 Bresnan, J. 175 Brink, B. t. 159 Brooks, K. R. 123 Burnley, D. 149, 158 Bussman, H. 111 Carstens, V. 169, 171 Chafe, W. L. 179, 182 Chomsky, N. 31, 34, 37, 55, 58, 60, 62, 63, 66, 175, 197, 198 Cinque, G. 173 Comrie, B. 76 Cook, A. S. 123 Cooper, H. 164 den Besten, H. 197 Diderichsen, P. 104 Doane, A. N. 122, 123 Dobbie, E. V. 124 Elliot, R. W. V. 150, 165

Empson, W. 43 Evers, A. 208 Faarlund, J. T. 77, 89, 90, 92, 94, 100, 101, 104 Fisher, J. H. 163, 167 Ford, C. E. 187, 188 Fuji, C. 201 Fujiwara, Y. 114, 116, 117, 122 Fukui, N. 126 Gelderen, E. 126 Goodman, N. 13, 16, 21 Gordon, E. V. 124 Greenberg, J. H. 103 Grimshaw, J. 176 Haeseryn, W. 79, 87, 92, 100 Haider, H. 198, 208 Halle, M. 34, 37, 42, 111 Halliday, M. A. K. 179 Hansen, Aa. 77, 90, 92, 93, 100 Harbert, W. 75, 76, 91, 96, 100 Harris, Z. S. 13, 15, 32, 35, 38 Hashimoto, T. 201 Heltoft, L. 90, 91, 104 Higginbotham, J. 59, 65 Hill, A. A. 31, 39-43 Hiż, H. 13 Hockett, C. 33, 35 Hoekstra, T. 146 Höhle, T. 197, 203 Holmes, J. 180, 189, 192 Hopper, P. J. 103 Hornstein, N. 169, 172-174 Hosaka, Y. 197 Iyeiri, Y. 179, 180, 182 Jakobson, R. 31, 37, 41-46 Jespersen, O. 150, 159

214

Index of Names

Kajita, M. 201 Keenan, E. L. 76 Kent, C. W. 124 Keyser, S. J. 111 Kiparsky, C. 203, 210 Kiparsky, P. 126, 203, 210 Klaeber, F. 124 Koskenniemi, I. 166 Krapp, G. P. 122, 123 Kroll, B. 179, 189 Leech, G. 179, 186 Lees, R. B. 33, 50 Lenerz, J. 197 Lenneberg, E. 40 Leonard, H. S. 16 Leslie, R. F. 124 Lightfoot, D. W. 133, 141, 143 Lukoff, F. 37 Maclay, H. 40 Malone, K. 124 Markey, Th. L. 99 Marr, D. 57 Marshall, J. 66 Martin, R. M. 24 Masui, M. 151, 155, 159, 160, 163 Murasugi, K. 201 Muscatine, C. 163, 164, 167 Mustanoja, T. 130, 133, 135 Nakajima, H. 201 Nakao, T. 130-132 Nishihara, K. 57 Nübling, D. 99 O’Donnell, R. C. 179 Osawa, F. 125, 126 Postal, P. M. 204 Quine, W. V. 13

Quirk, R. 78, 201 Radford, A. 126, 169, 171 Ritter, E. 173 Rizzi, L. 63, 201 Rosenbach, A. 131, 132, 137, 146 Rothstein, S. 175 Shimogasa, T. 156 Sleator, M. D. 40 Smith, N. 176 Smith, P. H. 121 Snyder, W. 175 Strunk, W. 124 Suppes, P. 34 Svartvik, J. 179, 186 Swanton, M. 124 Tani, A. 149, 150, 152, 156, 161 Thomas, D. 46 Thompson, S. A. 188 Thráinsson, H. 76, 82-84, 87, 92, 93, 104 Tiersma, P. M. 81, 87, 92, 100 Ullman, S. 57 Veblen, T. 46 Vergnaud, J.-R. 66 Warner, A. 125, 130 Weerman, F. 129 Weinberg, A. 69 Wells, R. S. 18 Werner, O. 103 Whitelock, D. 124 Williams, E. 170 Woodger, J. H. 24 Yaguchi, M. 180, 182 Zettersten, A. 124 Ziff, P. 44, 46

Index of Subjects Absolon 164 absolute genitive 135-137, 146 abstract level 65 abstract pronominal 61 abstract representation 64 accidental coincidence 113, 115, 116, 118 accidental possibility 116 active (voice) 75, 81, 82, 84, 87, 88, 90, 100, 109 adequacy 13, 15, 24, 25, 28 adjective 112, 113, 119, 120, 122 affixes 157-159, 161, 163-165 agent-PP deletion 75, 76 (agreeing) participle 95, 96, 98, 100, 101, 103, 110 alliteration 111-122, 166, 167 alliterative 150, 165-167 alliterative phrases 150, 165 alliterative type 114, 117 ambiguity 35, 43 American descriptivists 36 analytic(ity) 94-97, 102, 103 anaphor 58, 59, 63, 65, 66 Andreas 113, 117-121 antecedent 58, 59, 61, 62, 64, 65, 199, 206, 207 antecedent-empty element relation 64 applicable 24-26, 28 arbitrary person 61, 62 argument structure 63 artes 163 aspectual(ity) 77, 98 assumption 112, 115, 120, 122 atom 16-21, 26 a-verse 115, 117-122

barrier 67 The Battle of Maldon 121 Behaghel’s law 157 Beowulf 113, 114, 117-121 binding operator 62 binding principle 60 binding relation 58, 60 binding theory 58-70 biochemical system 70 biological endowment 55 biology 69 Boece 149 bound 57-62 bound form 111, 112 boundary 113, 114, 117 brain 57, 68 built-in principle 68 b-verse 113, 117, 118, 120, 121 Cadmon’s Hymn 117 Canon’s Yeoman’s Tale 163 Canterbury Tales 150, 151, 163, 166 capacity 197, 200, 209 case/Case 66, 67, 197, 198 case marking 82, 83, 91, 103-105, 109 categories 32, 46-51 categorization of word 122 Caxton 156 c-command 58, 59, 199, 200, 206, 207 Chaucer 149-152, 154-160, 162, 163, 165-167 classification 36, 40, 48, 50 clausal 64 clausal complement 57, 64 clause 58, 60, 64 clause union 208 cliticization 125, 139

216

Index of Subjects

cognitive structure 66 cognitive system 55, 66 coindexing 59 complement capacities 197 complementary 65 complementary distribution 65 complex word 111, 112, 121 component 56, 58, 67, 68 component of u.g. 63 compound word 111-114, 116-122 CON 15, 16, 18, 20-22, 26-28 conceptual priority 56, 67 conceptual requirement 67 configurational(ity) 91, 104-106 confirmation 67 conjunctive mood 202 consonant phoneme 116 constituent structure rules 48, 50 construction types 32 contain 58, 64 context 122 context-free language 70 Continental Germanic 75, 99 continuative (aspectuality) 98, 109 control 61, 67, 169-176, 197, 199, 200, 205-208 controller 200, 206 conventionalism 163, 164 Cook’s Tale 150, 152-155, 157, 158, 160-162, 165, 166 coordinating conjunction 179, 181, 182, 186, 188-191, 193 copula 98, 99, 102, 103, 109 core grammar 55, 56, 64, 65 coreference 59 covert movement 198 CP-type 201, 203-205, 207, 208 CSPAE (The Corpus of Spoken Professional American English) 179-

181, 183, 186, 187, 189, 192-195 curial prose 149 Danish 75-77, 90-93, 95, 96, 99, 109 data 31-35, 37-40, 47 degree of stress 111, 116, 120, 121 denominal verb 121 Deor 117 derivative 55, 68 descriptive adequacy 69 descriptive phonology 36 direct evidence 62, 66 direct object 60 discontinuity 14, 16-18 discontinuous 17, 19, 24-27 disjunction 22, 24 distribution 114, 117-120 distributional property 65 domain of knowledge 68 double alliteration 111-122 double genitive 136-138, 146 DP 169-174, 176 The Dream of the Rood 117, 120 Dutch 75, 76, 78, 80, 81, 84, 85, 87, 88, 91, 92, 94, 100, 102-105, 109 each other 58, 64-66 editor 117, 122 editorial interpretation 122 EI 20, 22, 23, 25-28 Elene 117, 120 embedded clausal complement 59 embedded clause 59, 60 embedded complement 60, 61 emendation 122 emergence 125-127, 145, 146 empirical content 69, 70 empirical difference 69 empirical domain 58 empirical observations 31 empirical theory 67

Index of Subjects

empty element 60-64, 66 empty pronominal 62, 64, 65 English 34, 37, 39, 40, 43, 46, 48, 75, 76, 78-80, 84-86, 88, 91, 92, 95, 99, 100, 102, 105, 106, 109 enjambment 159-161 ENV 15-18 epistemology 67 EPP 171, 174, 176 EQL 16-18, 20, 21 equivalent 20, 21, 24 etymology 154, 158, 162, 163 Exodus 121 expect 58-60, 62 experience 55-59, 66-68 explanatory adequacy 58, 69 explanatory force 58, 63 explanatory power 69 expletive 91-94, 102, 105, 106, 109 exploratory talk 180, 189, 193 expository talk 180, 181, 188, 192 extended category 23 extended projection 176 extension 14, 24-28, 55, 56 external condition 68 external environment 66 extra alliteration 114-117, 122 factive verb 205, 207 facts 31, 32, 37, 38 Faroese 75, 76, 83-85, 87, 91, 93-95, 97, 100, 103-105, 109 The Fight at Finnsburg 117 finite automaton 33 finite clausal complement 64 finite clause 58, 59 finite sets of grammars 69 finite state Markov process 33 first half-line 113 formal subject 87, 89-91, 94, 105

217

formality 180, 186-188, 192 formatives 48-50 fragment evidence 70 Fragment I 150, 151, 154, 156, 160-167 Franco-Latin 157, 158, 160, 161, 164, 165 free PRO 61, 62 French 150, 156 frequency 114-118, 122, 149, 151, 152, 155, 161-163, 166 full grammar 56 functional category (FC) 125-127, 138, 145, 146 functional explanation 70 gender 180, 181, 186, 188, 189, 192, 193 general principle 66, 69 General Prologue 150, 152-155, 157-167 generalized learning mechanism 66 generative capacity 69 generative grammar 31-35, 37-39, 41, 42, 48, 49, 51, 55, 68, 69 genetic endowment 68 genitive 171, 173 genre(s) 152, 162, 163 German 75, 76, 80-87, 91, 92, 95, 98100, 102-105, 109 Germanic 75-77, 84, 85, 91, 92, 94-96, 99-103, 106 Gesetz der wachsenden Glieder 157 glottal stop 116 governed 64 government 64, 67 grammar 55-58, 64, 66, 68-70 grammatical 31, 33, 36, 37, 40, 42-46, 48, 49 grammatical description 32-34, 49 grammatical morphemes 43 grammatical relations 43, 44

218

Index of Subjects

grammatical rule 15, 46, 47, 49 grammatical structure 35, 37, 39, 43, 44, 46, 47 grammatical transformation 41 grammaticalization 77, 96, 98, 104 grammaticalness 33, 39-43, 45-49, 51 graphic device 116 group genitive 134, 135, 137, 138, 141, 143-146 growth of grammar 63 growth of knowledge 57, 66 help 61, 63-65 hierarchy 125, 145 high style 163 hit 63 homonyms 50, 51 human embryo 67 hypotaxis 139, 141 iambic stress pattern 112 IC 38 Icelandic 75, 76, 81-85, 87, 91, 92, 94-96, 99, 100, 102-105, 109 idealization 56, 57 idiosyncratic 65 idiosyncratic (marked) elements 56 immediate constituent 58 immediate constituent theory 33 impersonal passive 76, 80-82, 85, 86, 89, 91, 92, 94, 103, 105, 106, 109, 209 independent word 111, 120 indices 58 induction 67 infinitival 59, 64 infinitival clause 79 infinitival complement 197, 199, 201, 202, 204, 205, 208, 210 infinitive 57 infinity 69 inflectional genitive 130, 131, 134, 135,

137-139, 141 inflectional morphemes 51 inflectional paradigms 34 ingressive (aspectuality) 98, 99, 102, 103, 110 inifinitival clause 65 initial consonant 116, 121 initial state 56-58, 66, 67, 69, 70 innate endowment 56, 63 innate property 59, 67 innate system of principles 67 input sentence 70 inscriptions 15-17, 19, 22-28 insect eye 67 instantaneous 57 Insular Scandinavian 75, 98, 104 internal structure 120 internalize 56 interpretive principle 65 interstate transition 57 intonation 39, 40, 42 intransformative (aspectuality) 98, 99, 109 intricate system of knowledge 70 intrinsic property 66 introductory exposition 68 intuitive observations 38 Italian 63 judgments 35, 36, 38, 39, 42 Juliana 118, 120 Knight’s Tale 150, 152-156, 158-164, 166, 167 knowledge of grammar 55, 56 knowledge of language 55-57 language acquisition 58 language evolution 70 language faculty 56, 66, 69 language learning 63 language-specific parameter 64

Index of Subjects

last resort 197 learnability 69, 70 learning 63, 66, 67 letter writing 163 level of abstraction 55, 57, 68 level of syntactic representation 59 lexical category 113 lexical entry 64 lexical form 122 lexical item 63 lexical meaning 122 lexical property 63 lexical stress 112 lexicon 69 light noun 171, 172 linguistic expression 55, 56, 59 linguistic knowledge 67 linguistic levels 33 linguistic property 119 linguistic stress 114, 119 linguistic structure 31, 35, 38, 39 linguistic universals 31 loanwords 156-159 locality 64, 67 locus of change 128 logical form 60, 65 long-line 113, 114 Mainland Scandinavian 75-80, 84, 85, 88, 89, 91, 99, 100, 104, 105 mammalian 67 manuscript 122 marked periphery 57 marked property 65 markedness 56 mathematical linguistics 68, 69 matrix clause 199, 207 maximum lexical stress 112 mental representation 59, 62 mental state 56

219

mentalistic 57 mentalistic terminology 57 Merchants Tale 163 meter/metre 150, 154 metrical difference 117 metrical framework 122 The Metrical Psalms of Paris Psalter 121 metrical rules 111 metrical significance 115 metrical theory 122 Miller’s Tale 150, 152-155, 157, 158, 160-167 mind-brain 56, 66 minimal c-command domain 58, 59, 62 Minimalist framework 198 mixed style 163-165 mnemonic 165 modality 197, 201, 202, 204, 205 modern structural linguistics 36 modularity 66 morphemes 14, 16, 18, 32, 35-37, 43, 51 morphemic analysis 36 morphemic segmentation 36 morphemic structure 117 morphological case (m-case) 125, 127, 129-131, 139, 145 morphological rule 63 morphology 157, 164, 167 morphophonemic component 41, 42 Le Morte Arthur 156 motivation 122 name-like variable 63 narrative 164 natural languages 13, 15, 16 naturalism 163-165 new organology 66 nominal 171, 174-176 nominal structure 125, 126, 129, 130, 138, 145, 146

220

Index of Subjects

(non-agreeing) supine 95, 98, 100, 101, 103, 110 nonce word 119-122 non-factive verb 203 non-NP case (language) 76, 84, 85, 87, 94, 104, 105 non-riming position 159-161, 165 non-subject promotion 75-78, 81, 82, 84-86, 89, 90, 103-106 Norwegian 75-77, 89-91, 94-97, 99-101, 103, 109 noun 112, 113, 117, 119-121 noun phrase 65, 112, 113, 119, 122 noun phrase accessibility hierarchy 76 nP 169, 171-174, 176 NP-case (language) 76, 77, 81, 83-85, 103, 104 NP-movement 197, 198 Nuns Priests Tale 163 Nynorsk 75, 95, 97, 101, 103, 109 object control 207 oblique subject 104, 105 OE poem 111, 117, 118, 120, 121 OE poetry 111, 116 OED2 153, 164 of-genitive 131, 133, 137, 138 Old English 112, 116, 122 Old English poetry 111, 116 opacity condition 58, 59 opaque configuration 58 optimal k-category analysis 50 oral delivery 151 order of strength 120 organism 68 organisum-environment interaction 66 overt anaphor 65 overt movement 198, 209 overt pronoun 62 paired-utterance test 38, 39

parameter 56, 57, 63, 64, 67, 69, 70 parametric variation 56, 70 parametrized 59 parataxix 139 parsability 69, 70 parsing 64, 69, 70 parsing time 70 Parson’s Tale 149, 152, 154, 155, 157, 158, 166 particular grammar 55 particular language 31 passive auxiliary 77, 94, 98-100, 102 passive (voice) 75-77, 80-84, 87-90, 9498, 100-103, 105, 106, 110, 197-199, 203, 205, 207-209 passivization 197-199, 205-209 passivization of a reflexive verb 199 PCA (Principal Component Analysis) 183, 184, 187, 192 periphery 56, 57 periphrastic passive 95, 103 personal passive 76, 77, 80, 86, 87 The Phoenix 118 phonemes 32, 37 phonetic content 60, 61 phonetic form 60 phonetic rule 60 phonetic transcriptions 35, 36, 42 phonological processes 37 phonology 34, 36, 39 phrase structure rule 64 poetic property 119 poetical text 122 poets’ preference 119 possible rule 69 PRE 16-19, 21 prefix 112, 121 prefixed word 121 presentational passive sentence 89

Index of Subjects

present-day English 112, 113 pretonic consonant 112, 113, 115, 116, 118, 121, 122 pretonic position 116 primary stress 111, 112, 115, 116, 119, 121, 122 principles and parameters 56, 70 principles of sentence formation 31, 45, 49 PRO 61, 62, 65, 66, 169, 170, 172, 173, 176, 199-201, 206, 207, 209, 210 PRO subject 62 pro-drop language 63 projection principle 63, 64, 66, 67, 69, 70 pronominal anaphor 62, 63, 65 pronoun 59, 61, 62 proper noun 117 properly formed 33, 34 propositional 64 propositional knowledge 56, 57, 59, 67, 68 prose 149-159, 161, 166, 167 prose style 152 prose work 116 psychology 69 puberty 67, 68 pure anaphor 63 quasi-quantifier 60 reciprocal 65, 71 reduced phrasal stress 112 Reeve’s Tale 150, 152-158, 160-167 reflexive 62 reflexive pronoun 199 regulatory circuit 70 repetition 151, 155, 156, 163 restrictive 69, 70 resulting state 67 Reynard 156

221

rhetoric 163 rhetorical ornaments 149, 163 rigidity principle 57, 66-68 Riksmål/Bokmål 75, 95, 109 rime 150, 154-156, 159-161, 167 rime phrases 150, 151, 156, 159, 160, 163 The Riming Poem 118, 120 riming position 155, 156, 159-161, 164, 165 Riverside editon 150 romance(s) 151, 155-157, 163, 164 root syllable 121 rule of interpretation 69 rule structure 70 rule system 69 rules and principles 55, 56, 68 sameness-difference judgments 36 SEC 23, 24 second half-line 113 secondary stress 112, 114-116, 120, 122 segmentation 35, 36 semantic content 64 semantic property 43, 44, 64 semantic role 63, 64 semi-grammatical 45-47, 49 sentential complement 210 separate genitive 134, 135, 137, 138, 140, 141, 146 separate words 119 sequence of phones 32, 33 s-genitive 131, 137 SHELL (Society of Historical English Language and Linguistics) 150 short grammar 70 short sentence 70 sight 67 single alliteration 114, 115, 117 single parameter 70

222

Index of Subjects

Spanish 63 species 56, 70 specific property 63, 69, 70 split genitive 133, 134, 137, 138, 141-145 SSC 22, 23, 25-28 statal (passive) 98-100, 110 state of knowledge 67 stative (aspectuality) 98, 99, 110 steady state 56-58, 66, 68, 69 stem 120, 121 stem syllable 112 s(t)-passive 98, 101 stress 111-113, 120, 122 stress contours 35, 37 stress level 38, 39 stress pattern 39 stress subordination 112, 113, 120 stressed vowel 116 strong features 198 structural configuration 63 structural description 32-35, 42, 48 structural linguistics 32, 36, 38, 68 Structure Uniform Hypothesis 126 style 149, 150, 152, 159, 161-165, 167 subdomain 69 subject 58-63, 65 subject control 206, 207 subject demotion 75-77, 85, 89, 90 subordinate clause 79 subordinating conjunction 179, 186, 188, 189 subsets 15 subsidiary stress 111, 119 substitute for primary stress 115, 116 substitution procedures 37, 51 subsystem of u.g. 55, 66 subsystems of principles 69 subtheories of u.g. 63, 64 suffix 111, 121

suffixal (reflexive) passive 96, 98, 110 surface form 65 Swedish 75-77, 90, 91, 94, 95, 99, 100, 103, 109 syllable 111, 112, 121, 122, 157 syllable structure 38 synonym 57, 58 syntactic analysis 15, 20 syntactic categories 14, 15, 22-25, 28 syntactic classes 14, 15 syntactic component 41, 42 syntactic level 63 syntactic representation 59, 60, 63 syntactic structure 63 synthetic(ity) 94, 96, 97, 102, 103, 110 system of innate principles 68 systems of knowledge 55, 70 (tacit) knowledge 56, 57, 59, 63, 66-68 Tale of Melibee 149, 152, 154, 155, 157, 158, 166 talk to 61, 63 temporal element 197 tensed clause 65 terms 17-28 tertiary stress 122 theory of language 66 theory of u.g. 56-58, 65, 66, 68-70 theory of vision 66 there-passive 88 three-dimensional space 68 topological syntax 91, 104, 105 TP-Spec 197-200, 209 TP-type 201, 205, 207, 208 trace 62 trace theory 63 transcriptions 35, 39 transformational component 64 transformational grammar 41, 42, 44, 49, 50, 70

Index of Subjects

transformational models 33, 41, 48 transformational rule 33, 50, 64, 69 transformational theory 33, 41 transformationist 41 transformative (aspectuality) 98, 99, 102, 110 transitive verb 61 A Treatise on the Astrolabe 149, 152, 154, 155, 157, 158, 166 triggering 56 trochaic stress pattern 111, 112 type of alliteration 112, 122 understood object 61 understood subject 61 Uniformity Theta-Assignment Hypothesis (UTAH) 141 universal grammar (u.g.)/UG 55-58, 6366, 68-70, 198 utterance tokens 35, 38, 40 vacuous 69 variable 59-63 verbal complex 208 verb-final(ity) 84, 85, 95, 110 verb-stacking 208 verse 149-167

223

violate 59, 60, 67 visual mechanism 57 visual representation 57 vocalic alliteration 111, 116 vowel quality 38 vowel reduction 37 VP-type 201, 208-210 Waldere 118 weak features 198 well-formed 33, 36, 37, 42, 46, 49 well-formedness 33-35, 38, 41-43 (West) Frisian 75, 76, 81, 84, 87, 88, 91, 92, 94, 100, 102-105, 109 wh-movement 64 who 59, 60 Widsið 118, 121 Wife of Bath’s Prologue 163 The Wife’s Lament 121 wonder 60, 61 word classes 36 word formation 122 word length 157, 159, 164 word pairs (WPs) 149-167 zero stress 122 zu-infinitival complements 209

224

Index of Subjects

Editors & Contributors Editors Ian ROBERTS

Department of Linguistics, University of Cambridge

John Ole ASKEDAL

Department of Literature, Area Studies and European Languages, University of Oslo

Tomonori MATSUSHITA

School of Literature, Senshu University

Contributors Noam CHOMSKY

Institute Professor and Professor of Linguistics (Emeritus), Department of Linguistics and Philosophy, Massachusetts Institute of Technology

John Ole ASKEDAL

Department of Literature, Area Studies and European Languages, University of Oslo

Yasuaki FUJIWARA

College of Humanities, Seitoku University

Kazuyo MIYAMAE

School of Laws, Senshu University

Akinobu TANI

Department of Language Studies, Hyogo University of Teacher Education

Junji HAMAMATSU

School of Literature, Senshu University

Yoko IYEIRI

Graduate School of Letters, Kyoto University

Michiko YAGUCHI

Faculty of Foreign Studies, Setsunan University

Yasumasa BABA

The Institute of Statistical Mathematics

Yasuhito HOSAKA

Department of Language Sciences, Faculty of Social Sciences and Humanities, Tokyo Metropolitan University