Constituent Order in Functional Grammar: Synchronic and Diachronic Perspectives [Reprint 2012 ed.] 9783110875836, 9783110133899

Citation preview

Constituent Order in Functional Grammar

Functional Grammar Series This series comprises m o n o g r a p h s a n d collections w r i t t e n in t h e f r a m e w o r k of Functional G r a m m a r . T h e a i m is to seek explanations for a w i d e variety of linguistic p h e n o m e n a , both language-specific a n d cross-linguistic, in t e r m s of t h e conditions under w h i c h a n d t h e purposes for w h i c h l a n g u a g e is used. Editors: A . M a c h t e l t Bolkestein S i m o n C. Dik Casper de G root J. Lachlan M a c k e n z i e

General address: Institute for General Linguistics Functional Grammar Spuistraat 210 NL-1012 VT Amsterdam The Netherlands Other books in this series: 1. A.M. Bolkestein, C. de Groot and J.L. Mackenzie (eds.) Syntax and Pragmatics in Functional Grammar 2. A. M. Bolkestein, C. de Groot and J.L. Mackenzie (eds.) Predicates and Terms in Funtional Grammar 3. Michael Hannay English Existentiels in Functional Grammar 4. Josine A. Lalieman Dutch Language Proficiency of Turkish Children born in the Netherlands 5. Jan Nuyts and Georges de Schutter (eds.) Getting one's Words into Line 6. Johan van der Auwera and Louis Goossens (eds.) Ins and Outs of the Predication 7. Judith Junger Predicate Formation in the Verbal System of Modern Hebrew 8. Ahmed Moutaouakil Pragmatic Functions in a Functional Grammar of Arabic 9. Simon C. Dik The Theory of Functional Grammar 10. John H. Connolly and Simon C. Dik (eds.) Functional Grammar and the Computer 11. Casper de Groot Predicate Structure in a Functional Grammar of Hungarian 12. HansWeigand Linguistically Motivated Principles of Knowledge Base Systems 13. Mike Hannay and Elseline Vester (eds) Working with Functional Grammar: Descriptive and Computational Applications Other studies on Functional Grammar include S.C. Dik, Functional Grammar (1978), T. Hoekstra et al. (eds.). Perspectives on Functional Grammar (1981). All published by FORIS PUBLICATIONS.

Constituent Order in Functional Grammar: Synchronic and Diachronic Perspectives John H. Connolly

FORIS PUBLICATIONS Berlin · New York 1991

Foris Publications Berlin · New York (formerly Foris Publications, Dordrecht) is a Division of Walter de Gruyter & Co., Berlin.

© Printed on acid-free paper which falls within in the guidelines of de ANSI to ensure permanence and durability.

Library of Congress Cataloging in Publication Data Connolly, John, H., 1953Constituent order in functional grammar : synchronic and diachronic perspectives / John H. Connolly. p. cm. - (Functional grammar series : 14) Includes index. Includes bibliographical references (p. ) and indexes. ISBN 3-11-013389-X (cloth : acid-free paper) I. Order (Grammar) 2. Functionalism (Linguistics) 3. English language-Syntax. I. Title. II. Series. P299.073C64 1991 91-30409 415-dc20 CIP Die Deutsche Bibliothek Cataloging in Publication Data Connolly, John H.: Constituent order in functional grammar : synchronic and diachronic perspectives / John H. Connolly. - Berlin ; New York : Foris Pubi., 1991 (Functional grammar series ; 14) ISBN 3-11-013389-X NE: GT

® Copyright 1991 by Walter de Gruyter & Co., D-1000 Berlin 30 All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printing: ICG Printing, Dordrecht Printed in The Netherlands.

To Margaret, Elizabeth, Edward and Sarah

Contents

Chapter 1:

Introduction

1

Chapter 2:

Constituent orden structure and function 3 2.1 Fundamentals of positional syntax 3 2.1.1 Linear structure 3 2.1.2 The three facets of positional syntax 4 2.2 Factors affecting the order of constituents 9 2.2.1 Non-stylistic factors 9 2.2.2 Stylistic factors 10 2.2.3 Marked and unmarked orders 16 2.3 Constituent order in relation to moiphology and phonology 19 2.3.1 Inflectional morphology 19 2.3.2 Non-segmental phonology 22 2.4 Approaches to the description of positional syntax ...24 2.4.1 The Functional Grammar approach 24 2.4.2 Comparison with other approaches 30 2.5 Constituent order typology 32 2.5.1 Typology and classification 32 2.5.2 Typology and universale 39 Notes 42

Chapter 3:

Synchronic functional description 3.1 The appeal of the functional approach 3.2 Towards a more adequate framework 3.2.1 Syntactic functions and the expression component 3.2.2 Semantic functions and the predicationformation component 3.2.3 Syntactic templates and placement rules 3.2.4 The status of functional patterns 3.2.5 Computer-aided rule formulation 3.2.6 Summary 3.3 Clause-rank constituents 3.3.1 Major elements

45 45 45 46 49 50 55 57 60 60 60

viii Constituent Order in Functional Grammar 3.3.2

Minor elements 3.3.2.1 Placement of individual elements 3.3.2.2 Sequencing of co-occurring elements in particular positions Phrase-rank constituents 3.4.1 The noun phrase 3.4.2 The verb phrase 3.4.3 Other types of endocentric phrase 3.4.4 The prepositional phrase Coordinate and appositive structures Order of the various sets of rules Closing remarks

70 .. 73

Diachronic functional description 4.1 Some theoretical issues 4.1.1 Variation and change 4.1.2 Qualitative and quantitative change 4.1.3 Problems with dead languages 4.2 The functional description of constituent order change 4.2.1 Background 4.2.2 Qualitative change 4.2.2.1 Early Middle English 4.2.2.2 Qualitative changes within Middle English 4.2.3 Quantitative change 4.2.3.1 The identification of quantitative changes 4.2.3.2 Quantitative changes within Middle English 4.3 Morphosyntactic change 4.4 The explanation of change Notes

131 131 131 133 134

158 162 166 169

Conclusion

173

3.4

3.5 3.6 3.7 Notes Chapter 4:

Chapter 5:

91 101 101 110 117 120 121 125 127 128

135 135 138 138 148 152 152

References

175

Index of Names

193

Index of Subjects

199

Acknowledgements

It is my pleasure to take this opportunity of expressing my thanks to those who have helped and encouraged me in the research towards and in the writing of the present book. Simon Dik has provided much inspiration and encouragement from the outset, and I am grateful to him and to the other editors of the Functional Grammar Series for their constructive comments on the original typescript of this volume. With regard to the computational side of the work which the book embodies, the support of Ernest Edmonds and other Prolog-expert colleagues, past and present, has been greatly appreciated. So, too, on the linguistic side, has that of David Crystal, whose advice o v a the years has been most valuable. Finally, the encouragement, interest and understanding of my wife and family has been of especial assistance to me at all times.

Chapter 1

Introduction

For many years the dominant emphasis within linguistics may fairly be said to have been on the structure of language rather than on its function as a means of communication. Nevertheless, the past decade has seen a perceptible growth of interest in the functional view of language and an increase in awareness of the importance of taking communicative factors into account when describing and explaining linguistic structure. The fundamental principles of the functional approach to language are well summarised in Dik (1989b:4-7). One of the most important of these tenets is that because communication involves the actual use of language in real situations, adopting a functional approach entails regarding pragmatics as the over-arching framework to which other aspects of linguistics must be related. According to this view, then, semantics subserves pragmatics, and syntax in turn subserves semantics. The grammatical model which Dik advances to incorporate this principle is known as Functional Grammar (FG). The purpose of the present book is to explore the description of Constituent Ordering (CO) within the FG framework. The aim is to show how it is possible to achieve a comprehensive description of CO and of CO change which takes properly into account not only the formal or structural properties of ordering but also the part which CO plays in linguistic communication. The first step will be to survey past work on CO. Subsequent chapters will then be devoted in turn to the synchronic description of CO and to the problem of accounting for CO change from the functional point of view. The above will involve modifying and extending Dik's original proposals for the treatment of CO in FG, while remaining true to the basic principles on which the functional approach is founded.

Chapter 2

Constituent Order: Structure and Function

2.1 FUNDAMENTALS OF POSITIONAL SYNTAX

Any framework of grammatical description, whether functional in orientation or not, must be able to cope with the formal properties of CO;1 otherwise it cannot possibly be descriptively adequate. It is, therefore, appropriate to begin with an outline of these properties, in order to appreciate the purely syntactic dimension of what has to be dealt with by the CO mechanism of FG. The area of grammar concerned with CO has been referred to by previous writers, e.g. Strang (1970), as positional syntax. This term will likewise be employed here, and is to be understood as covering the ordering of words, phrases and clauses within the sentence. The present section is thus devoted to the basic principles of positional syntax. 2.1.1 Linear structure Order is a relational concept, and the order of co-occurring grammatical units is generally recognised in linguistics as one of the most fundamental syntactic relationships (see, e.g., Robins 1980:170). The order of constituents is thus seen as an aspect of the structure of the sentence, and is often referred to as its linear structure (see, e.g., Lyons 1968:209). The linear structure of any sentence containing more than one word can be described in terms of a series of one or more elementary statements of the form X < Y (i.e. X precedes Y), Y< Z, etc., where X,Y and Ζ are either grammatical constituents or parts of constituents (cf. 2.1.2 below). Linear structure is usually contrasted with hierarchical structure. Whereas the former is based on precedence relationships, the latter is to be described in terms of either the relationship of dominance or that of dependency (see, e.g., Matthews 1981:72). Moreover, the two types of structure are logically independent. Thus, it is possible to find pairs of sentences with the same constituent structure but different linear structures, for instance (la) and (lb), or with the same linear structure but different hierarchical structures, for example (lb) and (lc):

4

Constituent Order in Functional Grammar (1) (a)

Finish

the

road

off

Finish

off

the

road

Finish

off

the

road

Clearly, both types of structure need to be specified if the differences between (la), (lb) and (lc) are to be describable. 2.1.2 The three facets ofpositional syntax Although it is possible to describe the linear structure of any sentence in terms of precedence relationships, matters are far less straightforward when it comes to writing rules which actually place constituents in the appropriate order. A wellknown example of the kind of problem that arises concerns the placement of the finite verb in declarative main clauses in English as compared with German (cf., e.g., Fourquet 1938:20). In English the crucial consideration is the position of the verb relative to its sister elements (i.e. those elements, such as the subject, with which it co-occurs), whereas in German what matters is that the finite verb should normally be placed second in the clause. Plainly, rather different types of rule are involved here, and the whole issue calls for some detailed discussion at this point. The basic consideration is that positional syntax has three different facets: relative position, proximity of position and absolute position. When we formulate a statement of the form X < Y, we are making a claim as to the position of X and Y relative to one another, and the rules pertaining to the position of the subject relative to the finite verb in English declarative main clauses are of this type. The German rule, however, is not of the same kind, as we shall see in a moment. First, though, the syntax of relative position needs to be looked at moie

Constituent Order: Structure and Function

5

closely. Within the syntax of relative position, a basic distinction may be drawn between positionally distinct and positionally non-distinct constituents. Two constituents are positionally distinct if neither interrupts the other. Thus, in (2a) the subject the man and die predicator2 has arrived are positionally distinct, whereas in (2b) they are not: (2) (a) The man has arrived. (b) Has the man arrived? In (2b) the subject is flanked by the predicator, i.e. it interrupts the predicator but is not interrupted by it (For this to happen it is, of course, necessary for the predicator to be discontinuous.) In this way it is possible for parts of constituents, as well as whole constituents, to be involved in order relationships with other constituents (cf. 2.1.1 above). Again, in (3) the subject (the man who brings the milk ) and the predicator are positionally non-distinct, but this time the two constituents overlap, i.e. they each interrupt the other (and must both, therefore, be discontinuous). (3) Has the man arrived who brings the milk? The occurrence of positionally non-distinct constituents represents an important complication in the formulation of the rules of positional syntax. Clearly, the notion of relative position is closely related to that of order. Indeed, the two terms can be used synonymously in certain contexts; thus, to state the order in which X and Y occur is equivalent to stating their relative position, and viceversa. The term sequence, too, may be employed in the same sense. However, the term position can also be applied to an individual place within a linear array, whereas order cannot; and the term sequence, unlike the other two terms just mentioned, may refer alternatively to a linearly ordered set of co-occuirent elements. For a mathematical linguistic treatment of the order relationship, see Levin (1971). Lyons (1968:77-9) identifies three logical possibilities for the ordering of two positionally distinct elements X and Y: (4) (a)

XY is grammatical but the reverse order YX is ungrammatical. Example: (i) They arrived. (ii) * Arrived they. (b) XY and YX are both grammatical, but the two are not equivalent. Example: (i) Cats dislike dogs. (ii) Dogs dislike cats. (c) XY and YX are both grammatical, and the two are equivalent. Example: (i) He is now free.

6

Constituent Order in Functional Grammar (ii)

He is free now.

He asserts that in case (4b), X and Y stand in a 'sequential syntagmatic relationship', since they are grammatically contrastive in these circumstances, but that the above relationship does not obtain in (4a) or (4c), where X and Y are not in contrast. By 'not in contrast' Lyons appears to mean non-distinct in terms of the expression of cognitive meaning. However, as we shall see in the next section, there are functional, and hence grammatically relevant, contrasts of other kinds besides. The view will be taken here, therefore, that Lyons has circumscribed the notion of the sequential syntagmatic relationship too narrowly, and case (4c) will be taken as applying only to instances of totally free variati«!. From the mathematical point of view, the (ordered) sequences XY and YX may be regarded as the two possible permutations (or arrangements) of the unordered combination {X,Y}.3 Given a combination of Ν elements, the number of logically possible permutations is equal to N! (i.e. Ν factorial, where N! is the product of the positive integers up to and including N). Thus, for instance, if Ν = 2, as in the present example, the number of permutations is also 2. If Ν = 3, then the number of permutations rises to 6 (viz. XYZ, XZY, YXZ, YZX, ZXY, ZYX), and so on. In actual languages, of course, the number of grammatically permissible combinations may be less than the logically possible number. Nevertheless, we can still say that if Ν represents the number of elements capable of co-occurring in a particular structure, and if Ρ represents the number of grammatically permissible permutations of those elements, then 1 < Ρ á N!. Haas (1973:108, 1974:533-4) has applied the notion of grammatical dependency to what we are here calling the syntax of relative position. He identifies three types of dependency, one of which is 'dependency for position*. For Haas, two constituents are dependent on one another for position if they cannot be moved relative to each other. Thus, in (Sa) the predicator and complement are mutually dependent for position since they cannot be transposed, whereas in (5b) the predicator and the vocative element can have their order reversed and are therefore not positionally dependent on each other. (5) (a) Be quick, (b) Look, David. The second facet of positional syntax is proximity of position. Here, the most important distinction is between elements which are contiguous and those which are non-contiguous. Positionally distinct elements are contiguous if and only if they stand immediately next to each other, like the subject and predicator in (6a) or (6b): (6) (a) Have you a moment? (b) I have an idea.

Constituent Order: Structure and Function

7

Positionally non-distinct elements are contiguous if at least part of one stands immediately next to at least part of the other, as in (7): (7) Has that person really finished who seemed to be intent on going on until midnight? In this sentence, the subject that person who seemed to be intent on going on until midnight and the predicator hasfinished overlap; they are contiguous because part of the subject (that person) and part of the predicator (has) stand immediately next to one another. In cases where the two constituents, or parts of constituents, are non-contiguous, their distance from each other can be reckoned quite straightforwardly in terms of the number of intervening elements or parts of elements. A further distinction may be drawn between adjacent and non-adjacent constituents. Consider the following sentences: (8) (a) (b)

He would have liked one. Would he have liked one?

In (8a) the constituents of the predicator, would have and liked, are arranged in such a way that would and have are contiguous, and likewise have and liked. In (8b), however, would and have are non-contiguous. Yet despite these differences, in both cases it is have rath«* than liked which is the nearest constituent of the predicator to would. This fact may be expressed by saying that in both (8a) and (8b), would and have are adjacent members of the same structure, two constituents being adjacent if and only if they are part of the same structure and no other constituent of that structure intervenes between them. It will be seen that a discontinuous structure is a syntactic unit in which at least one pair of adjacent constituents is non-contiguous. Proximity of position and relative position are logically independent of one another. Thus,giventhreefreelypermutableconstituents {X,Y,Z},X may precede Y and the two be contiguous (e.g. in the sequence XYZ) or non-contiguous (e.g. in the sequence XZY); or if Y precedes X, again the two may be either contiguous (as in YXZ) or non-contiguous (as in YZX). Despite this logical independence, however, it may at times be convenient to combine both facets within a single descriptive statement, such as 'X immediately precedes Y \ The third facet of positional syntax is absolute position. Whereas relative position and proximity of position are relationships between specified members of the same structure, the absolute position of a constituent is defined in terms of its relationship to the structure of which it forms part, without reference to any specific sister-element of that structure. One way of stating the absolute position of an element is to give its ordinal number in the sequence, starting either from the beginning (i.e. first, second, etc.) or from the end (i.e. last, last-but-one, etc.). This is the kind of statement that is required for the description of the placement of the

8

Constituent Order in Functional Grammar

finite verb in German. However, in some instances it is not possible to be sufficiently precise to make statements of absolute position in ordinal terms, in which case one has to be content with less specific expressions such as 'near the beginning (or end) of'. Non-ordinal statements of this latter kind are necessary when describing tendencies such as the one whereby 'heavy' (i.e. long and/or structurally complex) elements tend, in languages generally, to be placed near the beginning or end of the clause, rather than in the middle. When the ordinal method of specifying absolute position is used, a slight complication arises in connection with discontinuous constituents. Consider the following German sentence: (9) Haben Sie ihn gesehen? have you him seen 'Have you seen him?' The predicator in this sentence haben gesehen is discontinuous, the finite verb haben standing first, the subject Sie second, the object ihn third and the non-finite verb gesehen fourth (and last). The complication is that in such circumstances the number of absolute positions exceeds the total number of complete elements, which in this sentence is three: the subject, the predicator and the object. Absolute position, as a descriptive category, is independent of relative position and proximity of position, inasmuch as it is possible to make a statement about absolute position (such as the one pertaining to the place of the German finite verb) without making any direct stipulation in respect of the other two categories. However, it is clear firstly that if we specify the absolute ordinal position of two or more constituents, we thereby indicate their relative position, and secondly that contiguous elements will occupy successive positions on the ordinal scale. On the other hand, statements relating to the relative position or proximity of elements carry no specific implications for their absolute position. In many (and perhaps all) languages, certain types of constituent are more constrained in their ordering possibilities than others. For example, in English the order of the elements subject, predicator and object is more rigid than that of most kinds of adverbial. Elements which are comparatively unrestricted in their placementare often described as positionally mobile (see, e.g., Robins 1980: ISO). Strictly speaking, however, mobility should be related explicitly to individual facets of positional syntax. Thus, mobility in respect of relative position implies permutability, mobility in terms of contiguity means separability, and mobility in respect of absolute position implies that the element concerned is not confined to a particular position in the superordinate structure, however constrained it may or may not be in relation to its placement with reference to its sister-elements.

Constituent Order: Structure and Function

9

2.2 FACTORS AFFECTING THE ORDER OF CONSTITUENTS

CO is influenced by a variety of factors. The purpose of the present section is to summarise the most important of these4 and to classify them under a small number of fairly broad headings. The major distinction to be drawn is between stylistic and non-stylistic factors, these two broad categories then being divided into various sub-categories. The classification is probably not watertight, but it is preferable to an unstructured list of factors. 2.2.1 Non-stylistic factors Certain types of factor that influence the order of constituents are not the subject of stylistic choice, and are naturally termed non-stylistic factors. The main subtypes are: (10) (a) 0?) (c)

Purely syntactic Primarily syntactic Primarily semantic

An example of a purely syntactic factor is found in the convention whereby in English the determiner always precedes the noun it modifies.5 Here there is no possibility of the reverse order (cf. (4a) above), so that the relative position of the two words is purely a matter of the linear-structural aspect of syntax. The second type of non-sty listic factor, the primarily syntactic, is exemplified in the rule for German whereby normally the finite verb occupies second position in declarative main clauses but final position in subordinate clauses. The basic determinant of the choice of position here is the syntactic consideration of whether or not the clause is grammatically dependent on another clause or phrase, rather than any directly semantic factor. Nevertheless, there is an indirect involvement of the semantics, in that subordinate clauses differ from main clauses at this level insofar as the latter, though not the former, generally have some particular semantic relationship with either the predicator of a superordinate clause or the head of a superordinate phrase. Thus, in (11) the subordinate clause has the semantic role of agent6 in relation to the predicator of the main clause: (11) Wer seine erste Sinfonie hörte, lobte ihn. who his first symphony heard praised him 'Whoever heard his first symphony praised him.' The main clause, however, has no such semantic role relationship to the predicator of any other clause. Often, however, the cognitive meaning of the sentence depends directly upon the choice of linear syntactic order. In such cases, we speak of a primarily semantic

10 Constituent Order in Functional Grammar factor being at woik. Examples include the relative position of the two noun phrases in (12a) and (12b) or that of the subject and finite verb in (13a) and (13b): (12) (a) (b) (13) (a) (b)

The driver frightened the passengers, The passengers frightened the driver. Will you go tomorrow? You will go tomorrow.

The cognitive meaning is similarly affected by the placement of the phrase the surgeon in (14a) and (14b),7 though here the crucial factor is not the relative position but the contiguity of the apposed elements, since his father precedes the surgeon in both examples: (14) (a) (b)

His father, the surgeon, telephoned Mr Smith, His father telephoned Mr Smith, the surgeon.

2.2.2 Stylistic factors The remaining factors that influence CO may be termed stylistic, in that the choices of ordering which they bring about do not affect the cognitive meaning of the sentence, but are, nevertheless, significant in the pragmatic process of communication. These stylistic factors can be subdivided into the organisational, which relate to the sequencing of the elements of the message when constructing a text, and the aesthetic, which have to do with the achievement of a pleasing effect from the point of view of the audience. The organisational factors concern: (a) (b) (c) (d) (e) (f) (g)

Topic-comment structure Distribution of information Emphasis Empathy Composition of constituents Cohesiveness Connectivity

There is a vast literature on these organisational factors and their effects on CO, going back at least as far as Weil (1887). Here we shall simply note the main points; for a fuller review, see Allerton (1978), Brown & Yule (1983:ch.5). A distinction is commonly drawn between topic and comment (Hockett 1958:191) or theme and rheme (Halliday 1967:212). The topic or theme represents what the sentence is about, and the comment or rheme is what is said about it. It is natural, therefore, for the former to precede the latter. For example, in (16a) the topic, the man just mentioned, precedes the comment, spoke for an

Constituent Order: Structure and Function

11

hour, and again in (16b) the topic, someone, precedes the comment, coughed. (16) (a) (b)

The man just mentioned spoke for an hour, Someone coughed.

A distinct, though related, dichotomy is that between what Halliday (1967:204) calls given information and new information, where the former, unlike the latter, is presented by the speaker as being recoverable from the preceding text. For example, the man just mentioned in (16a) represents given information, whereas the remainder of the sentence conveys new information, inasmuch as the reference of someone is not assumed to be identifiable from what has gone before. (16b) thus also illustrates the non-equivalence of the topic-comment distinction and the given-new dichotomy, since (16b) contains no given information but does have a topic (cf., for instance, Allerton 1978:157, Dik 1989b:267). The distribution of given and new information within a sentence is such that the former tends to precede the latter, as in the above examples. This is only a tendency, however, and the reverse order is quite possible, for example in (17), where the given item is placed at the end. (17) Someone jeered the man just mentioned. The thematic organisation of the sentence has been much studied by linguists of the Prague School, who use Mathesius' (1929, 1975) term Functional Sentence Perspective (FSP) to refer to this area of syntax, which is intended to cover both topic-comment structure and the distribution of given-new information (without always distinguishing sharply between the two).8·9 The central concept within the study of FSP is communicative dynamism (CD). This is a gradience on which the constituents of a sentence may be ranked in terms of their contribution to the development of the message, such that elements which represent given information and/or constitute the topic (or point of departure) of the sentence carry a low degree of CD, while those which convey new information bear a high degree of CD. Typically, a sentence may be analysed into its 'theme', 'transition' and 'rheme', these categories representing increasingly high ranks on the CD scale. The element bearing the lowest degree of CD in a particular sentence is its 'theme proper' and that carrying the highest degree is its 'rheme proper'; see, for example, Firbas (1959a:42). With regard to CO, elements tend to be sequenced in terms of increasing CD, especially in languages in which the order of elements is comparatively free. Not all linguists agree, however, that FSP constitutes a gradience. For example, Chafe (1976:33) argues that the evidence for its scalarity is unconvincing and that all that is needed is a binary distinction between given and new information. What does appear to be generally agreed, though, is that only a small number of discrete categories are required for the description of the relevant facts in respect

12 Constituent Order in Functional Grammar of particular languages. As regards the recognition of thematic as opposed to non-thematic material in actual sentences, Sgall, Hajicová & Benesová (1973:ch.3) propose a 'question test'. At its simplest, this test operates in the following manner. Any statement can be regarded as an appropriate response to a particular set of questions. Given, then, an acceptable question-response pair, the elements in the response which must necessarily appear in the question are thematic. (Sgall, Hajicová & Benesová actually use the terms 'topic' and 'comment' rather than 'theme' and 'rheme', but this does not affect the point at issue.) To take an example, (18c) is a reasonable response to (18a) but not (18b). (18) (a) (b) (c)

What has Christopher seen? Who has seen an eclipse? Christopher has seen an ECLIPSE.

(18a) and (18c) thus form an acceptable question-response pair. Now, since the elements Christopher has seen in (18c) must be included in (18a), these elements constitute the thematic part of (18c). Attempts have been made to incorporate pragmatic concepts such as topiccomment structure and given/new information-distribution into generative grammar, though the terminology for the relevant concepts varies from author to author, with Chomsky (1972), Jackendoff (1972:ch.6) and other proponents ofTransformational Generative Grammar (TGG) employing the terms 'presupposition' and 'focus'. 10 Linguists working within generative frameworks other than TGG have likewise made use of such pragmatic concepts; see, for instance, Dahl (1969, 1974b,c), Fawcett (1980), Hajicová & Sgall (1975), Sgall, Hajicová & Benesová (op. cit.) and Takami (1988). Within FG, in particular, there has been much discussion of the notions of 'topic' and 'focus'; see, for instance, Bolkestein et al. (1981), de Groot (1983), de Jong (1983), Dik (1978a, 1980a,1989b:ch. 13), Dik et al. (1980), Hannay (1983,1985a,b, 1990) and Mackenzie & Keizer (1990). Topic-comment structure and given/new information-distribution have also been treated from the point of view of discourse; see, for instance, Clark & Haviland (1977), Davison (1984), Enkvist (1978,1985), Erteschick-Shir (1988), Hetzron (1975), Kurzon (1988), Prince (1978, 1981, 1984), Reinhart (1981), Sandulescu (1976), Sgall (1973, 1974, 1975a,b), Smyth, Prideaux & Hogan (1979), Sperber & Wilson (1986), Taglicht (1984), van Dijk (1977) and Werth (1979, 1984). These studies are, naturally, concerned with the organisation of coherent text in such a way as to facilitate effective interpersonal communication. This requires the speaker or writer to arrange the text in such a way that the audience can make the necessary connection between successive utterances, and this, of course, often results in the familiar given-new and topic-comment sequencing. In the case of narrative, there is also a tendency to describe events in the order in which they happened, and in some circumstances the sequencing of clauses is actually

Constituent Order: Structure and Function

13

crucial to the cognitive meaning, as can be seen by comparing pairs such as (19a) and (19b). (19) (a) (b)

We crossed the bridge and turned right, We turned right and crossed the bridge.

For further discussion, see Brown & Yule (1983:ch.4), Givón (1988), Gundel (1988), Levinson (1983:ch.3) and Vincent (1979:5-6). Another consideration that has long been regarded as relevant to CO is emphasis. Sweet (1903:3) points out that a word may be emphasised by placing it in any abnormal position; cf., similarly, Charleston (1960:142). However, sentence-initial and sentence-final position are particularly favoured for the placement of emphatic elements. Emphasis, of course, is a very general term and, as with the term 'focus', may be used either in a contrastive or in a non-contrastive sense.11 For example, in (20a) the word torrential is most naturally interpreted as bearing a non-contrastive emphasis, since the opposite expression, slight downpour, is rather unlikely.12 In (20b), however, the most natural interpretation of red is that it contrasts with some implicit alternative, probably white. (20) (a) (b)

It was a TORRENTIAL downpour, I preferred the RED wine.

As far as CO is concerned, though, either form of emphatic element qualifies for highlighting by means of exceptional placement, as, for instance, in (21a) and (21b). (21) (a) (b)

A TORRENTIAL downpour it was. The RED wine I preferred.

Emphasis may also be achieved with the help of other devices, such as clefting or extraposition (see, for example, Jacobs & Rosenbaum 1967:chs.6,21, Kiefer 1970, Quirk et al. 1972:ch. 14), and these, too will have an effect on the order of elements. The use of clefting or extraposition does, of course, involve the employment of special constructions, and the resultant emphatic ordering may be considered to be a contingent consequence of choosing the constructions concerned. It may also be noted that extraposition is sometimes obligatory. Compare (22a), which exhibits extraposition, with the ungrammatical (22b), which does not: (22) (a) (b)

Was it clear to you that he intended to quit? "'Was that he intended to quit clear to you?

A further factor in the sequencing of constituents, especially in languages with a comparatively fixed order of elements, is what Kuno (1975:321,1976:432,1986:

14 Constituent Order in Functional Grammar 206) and Kuno & Kaburaki (1977:647-78) call the surface structure empathy hierarchy, whereby the speaker finds it easiest to identify with the referent of the subject, next easiest to empathise with that of the object and most difficult to identify with that of the agent in the passive construction. The pairing of noun phrases with these different syntactic functions in the production of sentences is thus affected by considerations of empathy, and insofar as these syntactic functions are associated with particular relative positions, the empathy factor exerts an indirect influence on the order of elements. Essentially the same phenomenon as empathy is described in Dik (1989b:ch.l0) in terms of the subject representing the primary vantage point and the (direct) object representing the secondary vantage point from which a state-ofaffairs is presented.13 It is thus possible to account in such terms for the different relative positions of the direct and indirect objects in (23a) and (23b). (23) (a) (b)

I gave a sweet to my son. I gave my son a sweet.

Although the immediate explanation of the ordering of the two objects in this pair of sentences is their structural type (noun phrase versus prepositional phrase), the choice of structural type can in tum be accounted for in terms of the allocation of the function of secondary vantage point to the direct object in (23a) but to the indirect object in (23b). The internal structure of a constituent can also have a more direct effect on its placement, since the composition of syntactic units in terms of length and structural complexity is a further factor influencing their linear ordering. Long and/or structurally complex constituents, especially whole (subordinate) clauses, aie, as Jespcrsen (1949:57) notes, usually placed either at the beginning or at the end of the clause of which they form part.14 Final position is particularly favoured for such constituents (cf., for instance, Quirk et al. 1972:943, Dryer 1980). Although these are only tendencies, it can happen that a failure to conform to them results in a sentence which is not optimally easy to interpret. An example is seen in (24a), where the complement, a single adjective, has been placed after the syntactically complex object. Compare (24b), where the order is reversed. (24) (a) (b)

I considered the piece of music which my friend thought quite good bad. I considered bad the piece of music which my friend thought quite good.

Another factor which helps in the ease of interpretation of sentences is cohesiveness, the principle whereby items that belong together are placed together. Sometimes this principle is not just stylistic but is of significance for the cognitive meaning of sentences; see, for example, (14a) and (14b) in 2.2.1 above. At other times,

Constituent Order: Structure and Function

15

however, it operates in relation to style. For example, the infelicity of (25) results from placing the coordinated subjects so far apart. (25) David would like to go to the Isle of Wight, and Ann. A further consideration in the ordering of elements is connectivity with the surrounding text. To some extent this, too, is beyond the scope of stylistic choice, for instance when the position of conjunctions or relative pronouns is fixed. However, it can also be a matter of style, as can be seen in (26a), where the placement of the conjunctive adverbials in clause-final position results in a less natural effect than is found in (26b), where the adverbials concerned are placed clause-initially and thus between the parts of the text which they serve to link. (26) (a) (b)

He looked in the cupboard; he searched in the drawer then; he tried the filing cabinet next He looked in the cupboard; then he searched in the drawer, next he tried the filing cabinet.

The placement of connective material early in the clause is a further aspect of the influence upon text organisation of the pragmatic requirement to arrange utterances in a way which furthers effective interpersonal communication. As for the aesthetic factors, the two main ones affecting prose are: (27) (a) (b)

Euphony Balance

The judicious placement of constituents can make for euphony in the avoidance of cacophonous juxtapositions; compare (28a), especially when spoken aloud, with the aesthetically preferable (28b). It can also, if desired, be directed towards rhythmic regularity, as in (29b) compared with (29a). (28) (a) (b) (29) (a) (b)

Hand-picked, cryptic clues, Cryptic, hand-picked clues. Dark, sober attire, Sober, dark attire.

Euphony involves the actual pronunciation of utterances. Balance, on the other hand, operates only at the higher levels of structure. The clearest examples are to be found in parallelisms, such as (30), and in chiastic constructions like (31), which depend for their effect upon order-reversals within successive syntactic units. (30) In go the ingredients, and out comes the mixture. (31) He is likely to win. To qualify he is certain.

16 Constituent Order in Functional Grammar Verse, of course, is traditionally subject to two additional constraints: (32) (a) (b)

Metre End-rhyme

Clearly, these factors, too, can readily exert an influence upon the order of words. We have now put together a fairly comprehensive, classified listof the factors which influence CO. Before closing the present section, however, it should be pointed out that it is not possible to employ these factors to account unequivocally for the order of elements within absolutely every sentence. The interaction among them is complex, and sometimes they conflict At other times they reinforce each other, but then it becomes impossible to ascribe a particular order to any single one of them. They do not, of course, cover instances of free variation, nor should it be thought that speakers always carefully select the optimal ordering, as determined by these factors, on each and every occasion. Nevertheless, taken together the above list of factors give a reasonable idea of the influences to which CO is subject Before leaving the subject of the influences on CO, it is necessary to comment on the status of the stylistic factors in relation to grammatical description. Chomsky (1965:126-27) suggests that these factors are a matter of mereperformance. However, his view is unsatisfactory, because unless one refers to to such factors, one cannot account for a number of systematic structural differences between sentences, and one is thus preventedfromarriving at adequate syntactic descriptions of languages. See further Contreras (1976). 2.2.3 Marked and unmarked orders It was noted in 2.2.2 above that the influence of stylistic factors can result in atypical orders, as in the following example: (33) The LAST one I liked best. Such atypical orders are usually described as marked, in contradistinction to the corresponding normal or unmarked sequences. Marked and unmarked orders can be characterised either statistically or non-statistically. According to the nonstatistical view, an unmarked sequence is a default ordering, which is used unless there is some particular reason (such as emphasis) to depart from i t in which case the marked sequence is selected instead. This is the approach taken by, for instance, Bacquet (1962:27). According to the statistical view, on the other hand, a sequence is unmarked if it is the most frequently occurring order among a set of alternatives, the remaining alternatives being marked. This approach is adopted by, for example, Werth (1970:32). When we need to refer to the specifically statistical sense of 'unmarked' and 'marked', we may use the terms dominant and exceptional respectively, while the terms basic and special may be employed to

Constituent Order: Structure and Function

17

refer to the respective non-statistical senses. (See Siewierska 1988:8 for a comparable use of terminology.) It is also necessary to acknowledge the possibility that more than one alternative ordering may be unmarked, in which case we may speak either of jointnormal orders (if we do not wish to commit ourselves to either the statistical or the non-statistical approach) or else of joint-dominant or joint-basic orders (if we require a specifically statistical or non-statistical term, respectively). To describe alternative orders as joint-dominant is to imply that they have the same frequencyof-occurrcnce as one another, whereas to characterise them as joint-basic is to imply that they are in free variation, with neither having any special function not shared by the other(s). How are the statistical and non-statistical senses of the terms 'marked' and 'unmarked' related to one another? Some authors, including Bacquet, have viewed markedness in the non-statistical sense as being quite independent of quantitative considerations, and have thus been prepared to countenance the idea of a basic order which is, nevertheless, exceptional in the statistical sense. This point of view, however, is untenable as a general principle, since it ignores the functionality of marked orders. The functional effects of marked oiderings, such as theemphasising of a constituent by placing it in an unusual position, often rely on the statistically exceptional status of their occurrence, and in cases of this kind the exceptional order has to be identified with the special rather than the basic order; cf. Dik (1980b: 156). It must be concluded, therefore, that the two senses of 'marked' and 'unmarked' are not independent but are related. An element which is marked in one sense will frequently be marked in the other sense too. (Cf. Moravcsik & Werth 1986:3, who contend that such correlations are at the core of markedness theory.) The next question that must be addressed is how one establishes whether a particular order is either marked or unmarked. When these terms are used in the statistical sense, the answer is straightforward. One simply obtains a suitable sample and determines the dominant and exceptional sequences through a process of counting. The only real problem arises in relation to joint-dominant orderings. Suppose, in an ideal case, that two alternative orders both have precisely the same frequency-of-occurrence, namely 50%. The snag is that there is no guarantee that this 50:50 distribution will be exactly reflected in any given sample, and one might well find that the sample frequencies turned out to be (say) 55% and 45%, so that one would then be faced with the decision as to whether the two alternative orderings were or were not joint-dominant. The key to making such a decision lies in the sample size, which is conventionally referred to as N. For example, if Ν = 100, then we can be fairly confident that when the true proportion of both of the alternative orderings is 50%, the proportion in the sample will fall within the range 40% to 60%, i.e. 50 ±10%. If the frequencies of the two orders calculated from the sample had lain outside this range, then the situation would not appear to have been one of joint-dominance, but when they are 55% and 45%, joint dominance does, indeed, appear to be a plausible interpretation.

18 Constituent Order in Functional Grammar In the example just given, the error (E) was 10%. Assuming still that the true proportion of either ordering is 50%, then E can be calculated according to the following formula: (34) E = 2[V(2500/N)] Thus, if we want to reduce E, we need to increase the sample size N. For instance, if Ν is increased to 300, then E is reduced to approximately 6%. If we wish to calculate the appropriate sample size for a given value of E, then we can use the formula: (35) Ν = 10,000/E2 Thus, forinstance, to reduce the range to 50±5%, so thatE=5, Nhas to be increased to 400. Establishing whether a particular ordering is marked or unmarked in the nonstatistical sense, on the other hand, is a matter not of data analysis but of knowledge capture. Speakers of a language tend to be able to judge whether or not a particular ordering is unusual, on the basis of their familiarity with their language and how it is used. Such judgments are not normally based on precise quantitative analysis but on one aspect of their language-related knowledge, and it is this that we try to capture when we describe particular orderings as either basic or special. There is no automatic means of capturing such knowledge, and there are cases where speakers' judgments are uncertain, but the same is, of course, true of other aspects of language-related knowledge as well. In the present book, it will be the policy to describe as marked only those orderings which the author is reasonably certain merit that label, and to treat borderline cases as unmarked. A further theoretical point concerning markedness is that the status of a sequence as marked or unmarked is relative rather than absolute in nature. For example, the sequence 'subject predicator object ' (SPO) in English is unmarked relative to most types of clause, but relative to interrogative clauses the SPO order seen in (36) is marked, since the wh-element which it contains is not situated at the beginning of the clause. (36) He borrowed what? On the other hand, if echo-questions like (36) are treated as a distinct type of clause, then the order manifested in the present example must be considered as unmarked in relation to that specific clause-type! Clearly, such matters need to be handled with care when describing the positional syntax of a language. The notion of marked and unmarked orders is strongly connected with the organisational factors that influence CO. The arrangements whereby topic precedes comment and given precedes new information generally constitute

Constituent Order: Structure and Function

19

unmarked orders, while marked sequences are often employed for the sake of emphasis. Moreover, although the choice of subject in a transitive clause is primarily a matter of empathy, it can also be affected by the desire to maintain the natural, unmarked progression from given to new information. Compare, for example, (37a) and (37b). (37) (a) (b)

The aforementioned lawyer was interviewing a client. The aforementioned client was being interviewed by a lawyer.

In both these sentences the subject represents given information while the other nominal constituent conveys new information, so that the choice of subject makes the unmarked given-new progression achievable without the need to resort to emphatic orderings like the following alternative to (35b): (38) The aforementioned client a lawyer was interviewing. The placement of given information near the beginning of the clause also tends to serve the interests of connectivity, and when this is accomplished by means of an unmarked order, the conditions are most favourable for a natural stylistic effect. For further discussion of issues relevant to those raised in the present paragraph, see Chafe (1970:ch.l5), Contreras (1976:passim), Givón (1979:ch.2), Halliday (1967,1968,1985:ch.3), Kirkwood (1969), Lyons (1966:212-13,1977:500-11), Thompson (1978) and Waugh (1976,1977:39-55).

2.3 CONSTITUENT ORDER IN RELATION TO MORPHOLOGY AND PHONOLOGY

There are important connections between positional syntax and certain lower-level aspects of linguistic structure, particularly inflectional morphology and nonsegmental phonology. These relationships will now be summarised. 2.3.1 Inflectional morphology It is possible to distinguish two main types of relationship between positional syntax and inflectional morphology. The first is where the morphology depends on the linear arrangement of constituents. An example is found in French,1® where the past participle, when used verbally, agrees with the direct object if the latter precedes, as in (39a), but not if it follows, as in (39b). (39) (a) (b)

Combien de cartes postales avez-vous écrites? How many post-cards (fem. pl.) have you written (fem. pl.) J'ai écrit douze cartes postales. I have written (uninfected) twelve post-cards (fem. pl.)

20

Constituent Order in Functional Grammar

Another example concerns contact mutations in Welsh, where the initial consonant of one word is changed when that word is immediately preceded by one of a specific group of other words.16 To take one instance of this phenomenon, a word will undergo the soft mutation if immediately preceded by a personal pronoun. Thus in (40b), where the personal pronoun chi (you) precedes the word for 'clock* (normally cloc ), the latter undergoes the soft mutation and appears as gloc. In (40b), however, where the order is different, cloc remains unmutated. (40) (a)

(b)

Oes gyda chi gloc? Is with you clock 'Have you a clock?' Oes cloc gyda chi?

The second main type of relationship between positional syntax and inflectional morphology is that in which the one does not depend on the other, but both are influenced by some third factor. The obvious example here is the marking of syntactic function (especially subject versus object) by means of position and/or morphological case. Some languages rely very heavily on one rather than the other of these devices, but others, for instance German, employ both. In languages such as German, therefore, four possibilities exist for the differentiation of subject and object: (41) (a) (b) (c) (d)

Differentiation by position alone Differentiation by case alone Differentiation by both of these devices Differentiation by neither of these devices

The first of these possibilities is exemplified in (42). (42) Dieses Mädchen photographierte das Gebäude. This girl photographed the building Here, the subject precedes the object, and this is the only syntactic indication of the respective functions of these items, since from the inflectional point of view both are undifferentiated between nominative and accusative case. In (43a-c), however, we see the opposite state of affairs, as described by (41b). (43) (a)

(b)

Diesen Palast hatte das Mädchen photographiert. This (acc.) palace had the girl photographed 'The girl had photographed this palace.' Dieses Gebäude hatte der Mann photographiert. This building had the (nom.) man photographed 'The man had photographed this building.'

Constituent Order: Structure and Function (c)

21

Diesen Palast hatte der Mann photographiert This (acc.) palace had the (nom.) man photographed 'The man had photographed this palace.'

In these examples either the subject or the object or both can be seen to be casemarked. As regards the CO, however, the object has in each instance been placed initially for connectivity or emphasis and, as a result, precedes rather than follows the object. These syntactic functions are therefore not indicated by their positions, since their characteristic order has been reversed. The third possibility, (41 c), is illustrated in (44), where the subject and object are differentiated by means of both position and morphology: (44) Der Mann photographierte die Frau. The (nom.) man photographed the lady Finally, the fourth possibility, (4 Id), is exemplified in (45a-c), where neither order nor case-inflection serves as a cue to the differentiation of subject and object (45) (a)

(b) (c)

Das Mädchen, das die Frau photographierte The girl who/whom the lady photographed 'The girl who photographed the lady' or 'The girl whom the lady photographed' Das Mädchen, das die Männer liebten The girl whom the men loved Das Mädchen, das das Gebäude photographierte The girl who the building photographed 'The girl who photographed the building'

From the point of view of Functional Linguistics, examples of this type are rather interesting, since they illustrate the interaction between the systemic and pragmatic aspects of language. (45b) and (45c) are not ambiguous, despite the fact that in neither of them is the subject or object identified through either of the abovementioned devices. In (45b) the subject is recoverable on the basis of concord with the plural verb, while in (45c) the hearer is able to draw on the semantic cue that it is the animate rather than the inanimate entity that has to be the subject of the verb photographierte. (45a), on the other hand, is ambiguous and can be disambiguated only with reference to some context. For example, if it had been made clear in the preceding discourse that the woman had taken a photograph of the girl, rather than vice versa, then the subject ofphotographierte in (45a) would be straightforwardly recoverable. What these examples illustrate is that firstly, the language system does not al way s provide sufficient cues for the unambiguous recovery of grammatical relations; secondly, because of the operation of pragmatic factors it is not always necessary for it to do so; and thirdly, when formal cues are provided, these are not

22

Constituent Order in Functional Grammar

always the ones whose primary purpose is to signal what they are actually employed to recover. On the other hand, there will always be contexts where the recoverability of a syntactic function does require its being signalled by one of the two standard devices. If the language concerned, like German, makes provision for both, then although there will sometimes be redundant double-marking, as in (43c) or (44), on the other hand the speaker will have all the more resources for signalling syntactic functions. Moreover, when it is possible to employ casemarking or to rely on some other cue for the differentiation of subject and object, then linear order can be used for some other purpose, such as emphasis. In intransitive constructions, of course, the scope for ambiguity is correspondingly less, so that here again there is room for more flexibility in the ordering of constituents. The interconnection of case-inflections and positional syntax in signalling syntactic functions is sometimes referred to under the title of morphosyntactic relations (see, for instance, Shannon 1964:55, Shores 1971b). Morphosyntactic relations are important with regard to the functional explanation of CO change, and we shall therefore return to the topic in 4.3 and 4.4 below. 2.3.2

Non-segmentalphonology

It was noted in 2.2.2 above that the factors influencing CO included emphasis and the distribution of information. These two factors are also relevant to nonsegmental phonology, so that we again find interconnections between the two levels of structure.17 The most convenient framework for summarising the main points here is that provided by Halliday (1967:202-11). Halliday observes that when the speaker constructs a text in order to convey some message, he selects certain elements of the text for special prominence within that message. These elements are termed points of information focus, and they may serve as the basis for the analysis of the text into a succession of information units, Each information unit contains just one primary point of information focus, and sometimes also one secondary. These points of information focus always represent new information. The information unit corresponds at the phonological level to the tone group. Every tone group contains an obligatory tonic segment, optionally preceded by a pretonic segment. The initial syllable within the tonic segment is the tonic syllable and is characterised by nuclear pitch-movement. Each point of information focus is realised as a tonic segment. In the unmarked case, the tonic syllable coincides with the accented syllable of the last lexical (as opposed to grammatical) item in the tone group, and the constituent of which it is a part is signalled as representing new information. In the marked case, however, the tonic syllable falls elsewhere; the element to which it belongs is again shown to be informationally new and (unlike in the marked case) the remainder of the unit is necessarily signalled as representing given information. Quirk et al. (1972:938-

Constituent Order: Structure and Function

23

39) refer to these two cases as end focus and contrastive focus respectively. In the following examples the tonic syllable is written in capital letters and the information unit happens to be co-extensive with the whole sentence. (46) (a) (b) (c) (d)

The noise just referred to STOPPED. Someone has just LEFT. PAUL won the prize. David wants THOSE.

(46a) and (46b) exemplify end focus. In (46a) the verbal constituent is new but the subject is given, whereas in (46b) the subject is new as well, though the predicator is still the point of information focus. In (46c), however, we have contrastive focus, since the tonic falls on a non-final lexical item; Paul represents new information and the rest of the sentence given information. (46d), too, exemplifies contrastive focus, since the tonic falls on a grammatical item; the fact that it is in final position is irrelevant, since it is non-lexical. Here, those is informationally new, therefore, and David wants is given. The interaction between positional syntax and the distribution of information relates to the choice of final constituent within the information unit in the situation where the speaker wishes to select the unmarked option of end focus. Compare, for example, the following pairs of sentences: (47) (a) (b) (48) (a) •(b) (49) (a) (b)

I will gladly water the FLOWERS. I will water the flowers GLADly. They found the walk to the summit of Snowdon most Tiring. They found most tiring the walk to the summit of SNOWdon. We gave the voucher to TOM. We gave Tom the VOUcher.

End focus is preserved in all these examples, and this entails adjusting the sequence of constituents in order that the requisite element should stand as the point of that unmarked focus. In (47a) and (47b) advantage has been taken of the positional mobility of the adverbial gladly.1* In (48b), however, the postponement of the object the walk to the summit of Snowdon has resulted in a marked order, even though the distribution of information remains unmarked. In (49a) and (49b), on the other hand, the necessity of a marked ordering has been obviated by an internal structural adjustment to the indirect object, which is represented by a prepositional phrase in (49a) but a noun phrase in (49b). The distribution of information is related to the notion of emphasis, in that points of information focus are more emphatic than the remainder of the text, and contrastive focus is a matter of degree, with different degrees of emphasis being possible within a particular assignment of focus. Various prosodie devices are available to contribute to the signalling of emphasis, such as dynamic stress and

24

Constituent Order in Functional Grammar

pitch range, and these may, clearly, be used to apply strong emphasis to a word even in the informationally unmarked case of end focus, especially in a sentence like (48b) which is marked in terms of the order of its syntactic elements. This last point illustrates the fact that emphasis is a broader notion than can be captured in terms of the distribution of information focus, and that its interaction with CO depends not only upon whether the point of information focus is the element containing the last open-class item within the information unit but also upon whether that element is occupying its normal place within the clause.

2.4 APPROACHES TO THE DESCRIPTION OF POSITIONAL SYNTAX

Within different frameworks of grammatical description, different means are employed for dealing with CO. In this section we shall first outline the linearisation mechanism used in FG and then compare it with the ordering techniques incorporated into other approaches. As far as FG is concerned, we shall describe the ordering procedure as advanced by Dik (1978a, 1980b, 1983b, 1989b). Improvements to this mechanism will be proposed in Chapter 3 below. 2.4.1 The Functional Grammar approach The generation of the sentences in FG is essentially a two-part process. The first part of the process involves the construction of abstract logico-grammatical structures known as predications. For instance, the predication underlying (SO) is (51).19 (50) David sent the pretty card to Ann. (51) PASTsend^dlx,: D a v i d ^ ) ) ^ ^ ( d i x , : c a r d a r p r e t t y ^ ) ^ (dlxj: AnnN(x3))RecFoc This predication consists of the verbal predicate send together with three terms, (dlXji Davidj/Xj)), (dlx2: card^x,)) and ( d ^ : Ann^Xj)). The predicate is marked for its syntactic class, namely V(erb), and is preceded by an operator, PAST, which specifies the tense. As for the terms, each of these contains a variable, such as Xj, preceded by one or more operators and followed by one or more restrictors. Each of the terms in example (51) contains two operators: d, which stands for 'definite', and 1, which means 'singular'. With regard to the restrictors, the purpose of the latter is to specify the content of the element concerned. Examples include card^x^ and pretty^x,). Terms are marked in respect of up to three types of function; these will here be termed semantic-relational, perspectival and pragmatic. Semantic-relational categories include such functions as agent (Ag), goal (Go) and recipient (Ree), all of which are exemplified in (51). (The term semantic-relational is preferable to Dik's semantic on the grounds that the latter

Constituent Order: Structure and Function

25

is over-general.) The perspectival functions are subject (Subj) and object (Obj), used in the empathy-related sense of the chosen vantage point(s) or perspecti ve(s) explained in 2.2.2 above, and not in the conventional syntactic sense; hence the avoidance of Dik's term syntactic functions in the present context As for the pragmatic functions, Dik' s original proposal recognises four of these: topic, focus, theme and tail. Topic (Top) carries the same sense as in 2.2.2 above. Focus (Foe) appertains to the informationally most important or most salient item(s) in a particular context. Thus, in (50), David is the topic and Ann the focus. The theme marks out the universe of discourse to which the ensuing predication relates; for example, this wine fulfils theme function in (52). (52) This wine, I think it is delicious. Dik thus uses theme in a different sense from previous authors, to refer to what is sometimes called a left-dislocated constituent (cf., for example, Ross 1971, Emonds 1976, Harris 1984a). As for right-dislocated constituents, such as that delay in (53), these are said in FG to have the function of tail. (53) It was a nuisance, that delay. The tail thus represents an 'afterthought', whose purpose is to clarify or modify whatever precedes. More recently, Hannay (1985b) has proposed an additional pragmatic category, called sub-topic, for elements that are inferable from the discourse without having been directly evoked, for example through explicit mention. Dik (1989b:ch.l3) incorporates this category into a subdivision of the notion of topic, in which he distinguishes sub-topic from given topic, resumed topic and new topic. The term given topic refers to an entity already introduced into the discourse and at the forefront of the interlocutors' consciousness at the point of utterance; resumed topic relates to an entity reintroduced into the discourse after some significant interval; and new topic refers to a new discourse entity. In addition, Dik divides focus into new focus, referring to completely new information, and contrastive focus, relating to an entity whose salience is attributable to contrast rather than novelty. Hannay (1990:5) and Mackenzie & Keizer (1990:17), however, are dissatisfied with this set of subcategorisations, contending that new topic is really a type of focus. Partly because of the controversy surrounding these recent proposals, we shall operate in this book with the original system, leaving for future research the incorporation of whatever modified scheme may eventually emerge from the current debate. The second part of the sentence generation process in FG involves converting predications into surface realisations known as linguistic expressions. This is accomplished by means of expression rules, whose purpose is to ensure that the various elements contain all the correct words in the proper morphological form

26

Constituent Order in Functional Grammar

and that all the constituents appear in the requisite order, and also to assign appropriate accentual and intonational features. The set of expression rules is known as the expression component of the grammar, as distinct from the predication formation component, which generates the underlying predications themselves. Those expression rules which are responsible for the ordering of constituents are called placement rules, and in Dik's original proposal (1978a, 1980c) these work in conjunction with functional patterns, such as (54). (54)

P1SVO

This functional pattern is a schema which represents the ordering possibilities of the subject, verbal element (or predicator) and object within the English declarative clause. Normally, the subject is placed in the position marked S in the functional pattern, the verbal element in position V and the object in position O. However, if a constituent is assigned the functions of topic and focus, then it will (under certain circumstances) be placed not in its regular position (or pattern position), but in the special position PI, located at the beginning of the clause; see further Dik (1978b, 1979b). The placement rules required to achieve these effects are: (55) Topic, Focus Subject Predicator Object

-> -> -> ->

PI S V O

(A slightly modified linearisation mechanism has been proposed subsequently, but for convenience of exposition, discussion of this will be postponed until Chapter 3.) The functional pattern in (54) applies only to what may be termed the clauseproper; in other words, it excludes any left- or right-dislocated constituents. In order to encompass these as well, the schema needs to be extended as follows:20 (56) P2, PI S V O, P3 In (56), in which the clause-proper has been delimited by commas, there are two new special positions: P2 for the theme and P3 for the tail. Two additional placement rules are needed to go with these: (57) Theme Tail

-> ->

P2 P3

The functional pattern for the clause-proper also needs extending to cope with indirect objects and complements. If S and O are interpreted in their perspectival

Constituent Order : Structure and Function

27

sense, then the requisite pattern is simply: (58) P2, PI S V Ο Χ, P3 where X represents either a complement or a prepositional direct object or a direct object following a non-prepositional indirect object. Examples of these three possibilities are as follows: (59) (a) 0)) (c)

We (Subj) found him (Obj) interesting (X = complement). We (Subj) lent a book (Obj) to her (X = prepositional indirect object). We (Subj) lent her (Obj) a book (X = non-prepositional indirect object).

If, on the other hand, we wish to spell out the relevant syntactic information, more explicitly within the extended functional pattern, then the latter may be restated in the following form: (60) P2.P1S v r i O ( * P r e p P ) I X n C,P3 IDO IO ( = PrepP) J Here, DO represents the pattern position for the direct object, 10 for the indirect object and C for the complement. (Although the category indirect object is not recognised in standard FG, it will nevertheless be employed in the present work in its traditional sense; cf. Connolly 1983, and see further 3.2.1 below.) This extension will cope with structures such as those in (61), which involve the assignment of elements to the pattern positions indicated: (61) (a) (b) (c)

We showed them our passes. We wrote a letter to Richard. We thought it odd.

S V IO DO (IO * PrepP) S V DO 10 GO = PrepP) S V DO C

Further complications arise when we consider interrogative and optative clauses. In clauses of these types it is necessary to distinguish the pattern position of the finite verb (Vf) from that of the non-finite part of the verbal element (Vi), and to make use of an alternative functional pattern: (62) P2, PI Vf S Vi flO ( * PrepP) D O ! C, P3 IDO IO ( = PrepP) J An example of a polar interrogative conforming to this pattern is seen in (63) and an optative in (64).

28

Constituent Order in Functional Grammar (63) Have you given them any money? (64) May you enjoy many years of happiness!

Vf S Vi IO DO Vf S Vi DO

The same functional pattern applies also to wh-questions, apart from echo questions (see Quirk et al. 1972:408-9). However, in wh-questions the interrogative element is a PI-constituent; that is, it is regularly placed in PI rather than in its pattern position. An example is seen in the following: (65) Whose pyjamas are you hiding?

DO ( = P1 -constituent) Vf S Vi

Another feature of Dik's approach to CO is the importance which he attaches to the factor which in 2.2.2 above was termed the composition of constituents. This factor has also been given prominence in the past by other writers, such as Reszkiewicz (1966:13,115). Dik, however, goes further than previous authors in singling it out as a key cross-linguistic principle, known as the language-independent preferred order of constituents (LIPOC). An essential part of LIPOC is the following scale (see Dik 1983c:273,1989b:351): (66) Clitic < Pronoun < Noun Phrase < Adpositional Phrase < Subordinate Clause What is encapsulated in (66) is the tendency for structurally complex constituents to be placed later than structurally simpler constituents, especially when the latter have a low inherent capacity for sentence stress. Thus, pronouns tend to be placed relatively early, especially if they are clitic. Other noun phrases tend to occur later, while adpositional phrases tend to be placed later still. Finally, subordinate clauses, the most complex of the items on the scale, gravitate towards the end of the superordinate structure. In addition to (66), LIPOC includes the following tenets (see Dik 1980b:23): (67) (a) (b) (c)

For any category X , X < P X For any category Χ, X < X & X For any categories X and Y, X < X(Y)

These state that, all else being equal, (a) adding a preposition or postposition to a constituent will make it more likely to occur later than it otherwise would have; (b) conjoining two constituents will make them likely to appear later than either singly; and (c) a constituent, X, will be likely to occur later if it contains an embedded constituent, Y, than it would have otherwise. LIPOC does not, however, apply to constituents which have to be placed in special positions. The description just given of the treatment of CO within FG has been aimed mainly at the placement of clause-rank elements (i.e. direct constituents of the clause, such as subject and object). Moreover it has concentrated upon the major

Constituent Order: Structure and Function

29

elements of the clause (i.e. subject, predicator, object and complement) rather than the minor elements (i.e. adverbials, vocatives, coordinators and subordinators). However, as will be seen in Chapter 3, the same basic principles are applicable to minor clause-rank elements and also to phrase-rank elements (such as modifiers and heads within noun phrases). It should also be noted that in FG, maje«· elements are sometimes referred to as nuclear elements, which are distinguished from satellite elements (the latter being, in effect, adverbials). Returning to the FG linearisation mechanism itself, it may now be seen how the latter copes with the various formal properties of CO. In fact, all three facets of positional syntax (see 2.1.2 above) can be handled quite straightforwardly by means of rules operating in conjunction with functional patterns. With regard to relative position, if constituent X is to precede constituent Y, this is achieved by constructing a functional pattern in which the pattern position of X precedes that of Y, and by formulating rules to assign X and Y to those positions. If X and Y are to be contiguous, then their pattern positions must similarly be contiguous within the appropriate functional pattern. Turning to absolute position, let us come back to the example of the verb-second rule in German. A suitable functional pattern for German declarative main clauses is: (68) P2, PI Vf S f DO ΙΟΊ C Vi, P3 llODOJ In order to ensure that the finite verb appears in second position it is necessary to have a placement rule which obligatorily assigns the syntactic subject to PI, even when it carries no special pragmatic function such as topic, unless PI has already been Glied by another constituent. (Note that this implies that the placement rules are ordered.) If the subject is not placed in PI, then it is allocated instead to its pattern position. In cases of absolute positional tendencies which cannot be reduced to simple ordinal terms, the rules have to specify for the element concerned a choice of positions in the appropriate area of the functional pattern. As for the problem of discontinuity, this is resolved through the inclusion in the functional pattern of separate places for the different parts of the constituent concerned; an example is the distinction between the positions Vf and Vi in (68), which handles the discontinuity of the predicator that is often found in German main clauses. The FG linearisation mechanism is also able in principle to accommodate most of the different factors which influence CO (see 2.2 above), thanks to the richness of the information contained in the predication. The rules which take account of the purely or primarily syntactic factors are able to draw upon the specifications of syntactic class included therein, and those which relate to the primarily semantic factors can make reference to the specifications of semanticrelational function. However, it appears that some further syntactic-functional information is required in addition to that which is provided by the predication itself. We shall return to this matter in 3.2.1 below.

30

Constituent Order in Functional Grammar

With regard to the organisational factors, topic-comment structure and the distribution of new information can be taken into account via the sensitivity of the placement rules to the pragmatic functions of topic and focus, respectively, as also can empathy through their sensitivity to the perspectival functions. Emphatic orders can be traced to the interaction between (a) the allocation of the pragmatic functions and (b) the expression rules responsible for the choice of syntactic structure, forexamplethecleftconstruction (see Dikl980b:ch.l0). Thecomposition of constituents as a factor in CO emerges in FG under the guise of LIPOC, as already noted, though its precise effect in a given language has to be stated by means of appropriately formulated placement rules. Cohesiveness can be handled by formulating the placement rules in such a way that they produce contiguous rather than non-contiguous arrangements where appropriate, while connectivity can be achieved through the placement of connective elements in a position at or close to the relevant boundary of the functional pattern concerned. The aesthetic factors are more difficult to build into the placement rules in any systematic way. However, it may well be considered that they lie outside the scope of what can reasonably be expected to be incorporated into die grammar, and that it is sufficient for the latter to specify the available choices. In particular, the grammar should not be designed to generate verse. The aesthetic factors will thus not be built into the linearisation mechanism described in the next chapter. To sum up, it appears that in general the FG linearisation mechanism is well fitted to handle most structural and functional aspects of CO. Nevertherless, as already implied, there are some deficiencies in the original proposals. These will be discussed in Chapter 3, together with possible remedies. 2.4.2 Comparison with other approaches There are various ways of incorporating the ordering of elements into the overall design of a generative grammar.21 The approach adopted by Chomsky (1957) for TGG was to utilise a concatenation process whereby constituents are introduced into the syntactic structure in a particular order by the phrase structure rules, though not necessarily in the sequence in which they appear in the final surface structure of the sentence. One of the purposes of the transformational rules was thus to reorder the underlying syntactic structure so that the appropriate surface order might be achieved. TGG has undergone a number of major changes between the original version and the current GB (Government and Binding) framework, but movement rules have remained a feature throughout. Such rules are now, however, distributed among several components of the grammar; see Chomsky (1986:18). It is well known that while Chomsky's basic idea of constructing explicit generative grammars has been very widely accepted, many different proposals have been made with regard to the details of the actual generative apparatus. In some of the alternative frameworks that have been put forward, the CO mechanism works in much the same way as Chomsky's. This is the case, for instance, in certain

Constituent Order: Structure and Function

31

types of dependency grammar (see, for example, Hays 1964, Robinson 1970, Goralcikova 1973 and Vater 1975)22 and in the Montague-style grammars espoused by, for example, Lewis (1976) and von Stechow (1979). However, many other authors have departed in some way from Chomsky's approach. One alternative that has been suggested, notably by Staal (1967), Sanders (1970,1975a,b), Boas (1975), Anderson (1977)23 and Hudson (1979), is that the underlying structure be unordered.24·25 This has the advantage that in languages with relatively free CO, the grammar does not generate a string with the elements in a fixed order, only for the ordering to be undone by means of 'scrambling' rules (cf., for example, Ross 1967:43) or 'stylistic' rules (cf. for instance, Chomsky 1965:126-27, Chomsky & Lasnik 1977, Emonds 1980). Scrambling rules are also criticised by Peterson (1979), but the latter adopts the alternative strategy of, in effect, generating all possible orders, and then filtering out the ungrammatical ones by means of 'phrase structure ordering conditions', reminiscent of the 'surface structure constraints' suggested earlier by Perlmutter (1970). More radically, Edmondson (1986) takes the view that sequentiality in language is motivated psychophysical^ rather than linguistically, and that underlying order in grammar is therefore a concept which has no theoretical foundation at all. Other grammatical frameworks have been advanced which are less like TGG in their overall structure. These cope with CO in a variety of ways. Some do not use rules at all to produce the correct CO. These include Tagmemic Grammar and Role & Reference Grammar, which employ formulae for this purpose (see, respectively, Cook 1969:19 and Foley & van Valin 1984:81), and Stratificational Grammar and Augmented Transition Networks, which make use of graphical representations (see, respectively, Lockwood 1972 and Kaplan 1972).26 Others do handle CO by means of rules (or, in the case of categorial-type grammars, recursive definitions), but differ from TGG in that once a constituent has been placed in a particular position it cannot then be moved. S uch frameworks include Generalised Phrase Structure Grammar (see Gazdar et al. 1985), Systemic Grammar (see Berry 1977, Butler 1985a), Daughter-dependency Grammar (see Hudson 1976) and its successor Word Grammar (see Hudson 1984), Co-representational Grammar (see Kac 1977:26-29), Lexical Functional Grammar (see Kaplan & Bresnan 1982), certain versions of Categorial Grammar, such as those of Ades & Steedman ( 1982) and Flynn (1985), Surface Compositional Grammar (see Hausser 1984), Residential Generative Grammar (see Binkert 1984), Robinson & Moulton's (1985) Syntax Crystal,27 Lexicase (see Starosta 1988), Relational Grammar (see Blake 1990) and, of course, FG. One advantage of dispensing with movement rules is that it avoids sterile debates such as whether and how the surface order differs from the underlying order (see 2.5.1 below), or whether X and Y are transposed by moving Y to the left of X or X to the right of Y (see Schwarz 1972,1975). It also obviates problems such as that identified by Li & Thompson (1976a: 178-79) in relation to Mandarin Chinese, where a constituent might appear to have been moved into initial position, but cannot be said to have any 'home' position from which it might

32

Constituent Order in Functional Grammar

have been moved. In addition, the absence of movement rules makes it easier to develop computerised procedures for analysing sentences within theFG framework; see Jansen (1989:68). FG differs, however, from almost all other frameworks in incorporating a linearisation mechanism that involves the insertion of constituents into a pattern schema. The only other framework in which a similar procedure is used is Fawcett's version of Systemic Grammar, where elements are placed in 'starting structures' (see Fawcett 1980:47-50) which are comparable in many respects to the functional patterns of FG. Such differences as there are between these two linearisation mechanisms seem to be attributable to the differences in the overall structure of FG and Systemic Grammar rather than to the CO apparatus itself. It would be difficult to argue that any of the above approaches was grossly superior or inferior to the others. On the other hand, as has already been indicated, there are certain advantages in dispensing with movement rules, as is done in FG and some of the other frameworks mentioned above. Of course, these advantages are genuine only if the FG-type mechanism is capable of handling the full range of different structures whose elements require ordering, and this has not yet been established. In Chapter 3, therefore, it will be our purpose to show that the FG linearisation mechanism can be modified and extended so that it is, indeed, adequate in this respect.

2.5 CONSTITUENT ORDER TYPOLOGY

During the last twenty years or so the aspectof positional syntax which has received the most attention from linguists is that of CO typology. A considerable literature has arisen on this topic, not all of it from authors particularly well versed in the subtleties of CO syntax, with the result that some contributions are rather more rich in insight than others. In this section we shall summarise the main points from this literature as far as they relate to synchronic syntax (deferring discussion of the diachronic dimension until 4.4 below), as well as making mention of some of the more traditional concerns of CO typology. The survey will, of course, refer to contributions from workers within the FG community. 25.1 Typology and classification Various kinds of criteria have been used in the past for the typological classification of languages. Perhaps the best known, traditional method of categorisation is the morphologically-based division into analytic, inflecting, agglutinating and polysynthetic languages; for a good discussion see Comrie (1981:39-49). In view of the close relationship between inflectional morphology and positional syntax (see 2.3.1 above), this classification is to some extent relevant to CO typology, as the morphosyntactic relations are necessarily different in a highly inflecting

Constituent Order: Structure and Function

33

language compared with a highly analytic language. The most common objection to such a classification, however, is that the categories represent extremes of continua, with no language providing an example of a pure type. More directly related to CO is another traditional typological continuum, namely that between languages in which the order of elements is comparatively rigid and languages in which it is relatively free. However, it is important to remember that, as implied in 2.1.2 above, the same language may show greater freedom in one area of positional syntax than in another. Firstly, the freedom may not be the same in different types of structure. In Latin, for instance, the order of clause-rank elements is fairly free, whereas the order of adposition and adpositional complement within adpositional phrases is much more rigid. Even within structures of the same type, moreover, there can be varying degrees of freedom of ordering. For example, in English the order of nuclear elements is comparatively rigid, while adverbials tend to be positionally mobile. Secondly, the amount of freedom is not necessarily the same in respect of the three different facets of positional syntax (see 2.1.2 above). It is possible to have considerable constraints on relative position combined with comparative freedom of absolute position; compare, for instance, the balance of these two types of constraint in the placement of the nuclear clause-rank constituents in English and German. Contiguity is a different matter again. Even in languages where the relative and absolute positioning of clause-rank elements is fairly free, there may be a tendency for the constituents of a particular element to be placed together, as is the case, for example, in Old English. When we speak of 'free' CO, therefore, we must be careful to state what exactly we are referring to. In the present book, the term free-order languages w ill be used to describe languages in which, in particular, the relative position of clauserank nuclear elements is comparatively free, and fixed-order languages to describe languages which are relatively rigid in this specific respect. The other main method that has been employed in the classification of languages in terms of CO typology involves assigning them labels such as SOV or SVO. Such labels are intended to reflect the conventional order of elements in declarative main clauses in the language concerned. Of the six logically possible permutations, it has been noted by Greenberg (1966) and others that the overwhelming majority of languages are either SOV, SVO or VSO, these being the three patterns in which S precedes O. SOV languages include Amharic, Basque, Hindi, Japanese and Persian, SVO languages are exemplified by English, Finnish, French, Swedish and Vietnamese, and VSO languages include Arabic, Breton, Chinook, Hebrew and Welsh. However, the remaining three patterns are also attested. VOS is exemplified by Cakchiquel, Gilbertese, Huastec, Malagasy and Tzeltal (see Keenan 1976,1978b, Pullum 1977,1981), OVS by Apalai, Basarano, Hianacota, Hixkaryana and Panare, and OS V by Apurinâ, Nadëb and Urubú (see Deibyshire & Pullum 1979, Pullum 1981). Shorthand labels of the kind just mentioned are often very handy, and indeed

34

Constituent Order in Functional Grammar

we shall from time to time have cause to use them in this book. However, they do represent oversimplifications. To begin with, they obscure any variation in patterning from one type of clause to another, such as that found in German between main clauses (SVO) and subordinate clauses (SOV),28 and they do not highlight the difference between SVO resulting from a verb-second rule of absolute position and SVO produced by rules based solely on relative position. Furthermore, they omit important detail. They do not make any mention of the position of the complement (cf., for example, Kohonen 1978:26), and they also fail to distinguish the finite from the non-finite components of the verb phrase, even where this is relevant. For example, Welsh is described as a VSO language, but in the very common periphrastic verbal construction the normal order is Vf S Vi O,® as for instance in (69). (69) Mae'r bachgen yn prynu afalau. Is the boy buying apples 'The boy is buying apples.' Again, shorthand labels do not show alternative unmarked orderings such as those affecting the direct and indirect objects in Welsh or French. In French, pronominal objects precede the verbal element (except in imperative affirmative clauses), while non-pronominal objects precede it; thus, compare (70a) and (70b): (70) (a)

(b)

Nous la vendons We it sell 'We sell iL' Nous vendons la laine. We sell the wool

In Welsh, the pronominal object of a periphrastic verb is treated as the possessor in a genitive construction.30 The pronoun thus occurs before the non-finite verb, where it appears in the genitive case, and is repeated post-verbally in the unmarked case-form: (71) Rydyn ni'n eich clywed chi Are we your hearing you 'We hear you.' Labels such as SVO or SOV have been criticised by Lehmann (1973b, 1974: ΜΙ 5) on the grounds that in some languages the subject is grammatically optional. He therefore prefers a simple division between OV and VO languages. However, these labels are even more gross than those which they are intended to replace. Moreover, in languages such as Italian where the subject is relatively omissible, the object is not always obligatory either; compare the examples in (72):

Constituent Order: Structure and Function (72) (a) (b) (c) (d)

35

L' orchestra suonerà la musica The orchestra will play the music L'orchestra suonerà. Suonerà la musica. It will play the music Suonerà.

Pursuing Lehmann's argument to its logical conclusion, we should have to omit reference in our label for Italian not only to S but also to O. This approach would have the singularly unhelpful effectof reducing many of the languages of the world simply to ' V languages'! A further disadvantage of employing such general labels as VO is, as has been pointed out by Hawkins (1979) and Mallinson & Blake (1981:379), that it obscures the real differences between S VO and VSO languages, which are conflated if the position of S is disregarded. Another relevant consideration here is the fact that in some languages the order of clause-rank elements is more strongly determined by topic-comment structure than others in which the subject-predicate structure forms the main basis of linear ordering. Li & Thompson (1976b) term languages of the first kind topicprominent and languages of the second kind subject-prominent. They suggest, furthermore, that this distinction can be used as the basis for an alternative typology, whereby four different language types may be recognised. These are (i) languages which are purely subject-prominent, such as English; (ii) languages that are purely topic-prominent, like Chinese; (iii) languages which are both subjectprominent and topic-prominent, such as Japanese; and (iv) languages that are neither subject-prominent nor topic-prominent, like Tagalog. Abraham & de Meij (1986) extend Li & Thompson's ideas to include, in addition, focus-prominent languages, such as Hungarian. Accounting properly for typological differences is one of the goals of the FG framework; it is part of what Dik (1989b: 14) terms 'typological adequacy'. It is appropriate, therefore, that FG makes available a means of comparing languages in terms of their unmarked orderings which does not suffer, at least to the same degree, from the drawbacks of oversimplification that characterise the shorthand labels discussed above. In FG, languages can be compared by means of their functional patterns. Consider, for example, the functional patterns relating to declarative main clauses in Welsh (73a),31 English (60), repeated here as (73b), and Hindi (73c): (73) (a) (b) (c)

P2,P1 Vf S O ( = Pron) Vi DO | I O J , P3 P2.P1S V fIO(*PrepP)DOi C,P3 IDO IO ( = PrepP)/ P2, PI, S IO DO C V, P3

36

Constituent Order in Functional Grammar

(Example sentences are given below.) If necessary, shorthand labels can be read off these patterns, subject to the qualifications outlined above, but the functional patterns enable more detailed comparisons to be made, such as whether the finite and non-finite constituents of the predicator, or structurally different types of object, are treated separately. A further means of comparing languages is in terms of the variety of functional patterns which they allow. For example, English requires (74) as well as (73b) in order to accommodate questions as well as statements: (74) P2, PI Vf S Vi flO ( Φ PrepP) D01 C IDO IO ( = PrepP)J The interrogative sentences in (75) correspond to the pattern in (74), while those in (76) correspond to (73b). (75) (a) (b) (76) (a) (b)

Has he fed the cat? Vf S Vi DO What did he say? PI Vf S Vi You will see him? S V DO You did what? S V DO

In Welsh, however, pattern (73a) will suffice for both declarative main clauses, as in (77) and interrogatives, as in (78): (77) (a)

(b)

(c)

(78) (a)

Mae'r athrawes yn darllen stori i'r plant Vf S Vi DO IO Is the teacher reading story to the children 'The teacher is reading a story to the children.' Mae'r pwyllgor yn neud Iorwerth yn gadeirydd. Vf S Vi DO C Is the committee making Iorwerth chairman 'The committee are making Iorwerth chairman.' Rydw i'n eich nabod chi. Vf S O Vf DO Am I your knowing you Ί know you.' Ydych chi'n talu'r bil? Vf S Vi DO Are you paying the bill?

Constituent Order: Structure and Function (b)

37

Beth ydych chi'n canu? PI Vf S Vi What are you singing?

Similarly, in Hindi the same functional pattern will accommodate both declarative and interrogative main clauses:3233 (79) (a)

(b)

(c)

Suriil us aurat ko kitäb dikhä rahä hai. S IO DO V Sunil that woman to book showing is 'Sunil is showing the book to that woman.' Suriil kis ko kitäb dikhä rahä hai? S IO DO V Sunil whom to book showing is 'To whom is Sunil showing the book?' Kya Suriil lambä hai? PI S C V Question Sunil tall is marker 'Is Sunil tall?'

English is, of course, by no means unique in displaying more than one functional pattern. For instance, German, too, requires more than one pattern because of the different characteristics of main and subordinate clauses. There is no problem in FG about allowing variation of this kind, since it can be viewed functionally as a signal of the distinction between different clause types. The quest for a single underlying order for a particular language thus makes no sense in FG, and perhaps this is just as well. In TGG, English is generally assumed to have an underlying order of S VO (see, for instance, Chomsky 1965:ch.2). However, it has alternatively been suggested that English has an underlying order of SOV (Ross 1973; see, however, Anderson & Dahl 1974) or VSO (McCawley 1970; see, however, Newmeyer 1971, Berman 1974, Lehmann 1978a, and in the end McCawley 1978).34 Such conflicts cannot arise in FG. Comparisons of languages based on pattern positions within functional patterns, like comparisons based on shorthand labels, tend to emphasise the relative rather than the absolute position of elements. However, languages can also be compared in terms of the latter facet of positional syntax. One way of doing this involves comparing functional patterns in respect of the special positions which they contain, since these special positions are generally absolute in character. The languages mentioned so far in this section have had three special positions: PI at the beginning of the clause-proper, P2 at the beginning of the whole clause, and P3 at the end of the whole clause. In Hungarian, however, according to de Groot (1980), there are four:

38

Constituent Order in Functional Grammar (80) P2, PI PO V X, P3

PI is for the topic and PO for the element bearing focus, which normally occupies immediate preverbal position in the clause-proper. (No pattern positions are provided for S and O, as these functions are not assigned in Hungarian, which, as noted above, is a focus-prominent language.) A further means of comparison based essentially on absolute position is proposed by Dik (1980b: 152), who distinguishes the following: (81) (a) (b) (c) (d)

VI languages V2 languages V2s or strong V2 languages V3 languages

V1 languages and both types of V2 language have a functional pattern for clausesproper which begins: (82) P l V ( f ) S . . . In VI languages, such as Welsh, PI is usually unfilled, so that the (finite) verb is placed first. In V2 languages PI is normally filled, so that the (finite) verb tends to be placed second; Dik claims that Old High German constitutes an example of such a language. If the tendency for the (finite) verb to occupy second place is very pronounced, as in German, then the language is of the strong V2 type. V3 languages, however, have a functional pattern for clauses-proper which begins: (83) P l S V ( f ) . . . In such languages, for instance English, the (finite) verb appears in third position among the nuclear elements if PI is occupied by any constituent other than the subject. The title 'V3* is, therefore, a little loose. The main criticism of Dik's treatment of CO typology is, however, that it does not cope adequately with languages in which the object normally precedes the subject. Dik (1978a: 175) proposes the following language-independent patternschema, which we may abbreviate as LIPS: (84) P2, PI (V) S (V) O (V), P3 This is supposed to be a kind of blueprint for the functional patterns of particular languages, such that verbal constituents may occur in principle in any of the places indicated, but the subject always precedes the object. Dik tries to maintain this proposal by arguing that apparent counterexamples in fact lack rules of object assignment, this being possible because in FG 'object' is essentially a meaning-

Constituent Order: Structure and Function

39

related category. However, it is possible for elements to behave grammatically like objects, for example in determining relevant morphological markings and realising the semantic-relational function of goal in non-passive transitive sentences (see Derbyshire 1977, Derbyshire APullum 1981),even in languages which lack object assignment in Dik's sense. In order to accommodate such languages within a typological classification, a slightly different approach is needed. This will be outlined in 3.2.4 below. 25.2 Typology and universals Greenberg (1964,1966) observes that certain positional characteristics seem to cluster together in languages. For example, there is a strong statistical tendency on the one hand for VSO languages to be prepositional, but on the other hand for SOV languages to be postpositional. Such observations have given rise to a great deal of further research on the subject of CO universals. Vennemann (1973) proposes a general principle, known as 'natural serialisation', to account for positional correlations of the kind just mentioned; see also Bartsch & Vennemann (1973:131 -39), Vennemann (1974a), Anderson (1976), Vennemann & Harlow (1977). The principle states that languages tend to place all operators (i.e. dependent elements) either to the left or to the right of the operands on which they depend. Thus, for instance, in a language in which the object (operator) precedes the verb (operand), the adjectival or genitive modifier (operator) also precedes the noun (operand), while the adposition (operand) duly follows the noun phrase (operator). However, Hawkins (1980:198) points out that there are an alarmingly high number of exceptions to the sets of correspondences predicted by the principle. Lehmann (1973a, 1978c) advances the principle that modifiers occur on the opposite side of an element from that element's primary concomitant. For example, if the primary concomitant of a verbal element is taken to be its object, then according to this principle, verbal modifiers should precede the veib in VO languages but follow it in an OV language. Again, however, it is not difficult to find exceptions. For instance, in both English and Welsh the object follows the verb, yet the negative verbal modifier is not pre-verbal, as can be seen in (85a) and (85b). (85) (a) (b)

She does not eat meat Chaiff e ddimamser. Will-have he not time 'He will not have time.'

In Hindi, on the other hand, both object and negative verbal modifier precede the verb, as in (86).

40

Constituent Order in Functional Grammar (86) Suriii kitäbem nahlm parhnä. Sunil books not read 'Sunil does not read books.'

Keenan (1978a) puts forward what he calls the 'serialisation principle', the 'dissimilation principle' and the 'subject front principle' in the hope of capturing appropriate generalisations about CO typology. According to the first of these principles, different 'functional expressions' (e.g. adjectives or articles) which take the same class of 'argument expressions' (in this instance nouns) tend to be placed on the same side of the latter expressions. The second principle states that functional expressions such as adpositions, which take as arguments determined noun phrases,35 and functional expressions such as adjectives, which take undetermined noun phrases, tend to be placed on the opposite side of their argument expressions. According to the third principle, the subject precedes the other major constituents. However, these principles are not absolute, and among the six logical orders of the elements S, V and O, it is only SOV that is consistent with all of them. A study of CO variation within languages of a given typological class, such as S VO or SOV, is found in Steele (1978), who puts forward two general principles. The first states (in its strong form) that if in the basic order the verb is neither initial nor final, then variant orders in which the verb is initial or final tend to be avoided. So, too, according to the second of the two principles, do variant orders in which the object precedes the subject. However, these principles were formulated at the costof confining their application to verbal elements together with nominal (rather than pronominal) subjects and objects within declarative main clauses. A major re-examination of CO typology has been undertaken by Hawkins (1979, 1980, 1982, 1983, 1984, 1985).36 Hawkins proposes a small number of multi-term implicational universale, together with a distributional generalisation known as the principle of 'cross-category harmony' (CCH). An example of his implicational universale is the statement that if a language is of the SOV type, and if it also has the order adjective-noun, then it will also have the order genitive-noun. The principle of CCH concerns the different type of operator-operand pairs that occur in languages. It states that the greater the number of such pairs in which the relative position of operator and operand is the same, the greater the number of exemplifying languages there are. This differs from Vennemann's natural serialisation principle in that it accommodates integrally those languages which are not totally consistent in their placement of operator relative to operand. As such, it is likely to prove a more robust typological principle than any based on exceptionless implicational universals, since it is not only very wide-ranging but also less liable to be overthrown as a result of extending the sample to include hitherto relatively unstudied languages.37 Hawkins (1983:ch.3) also proposes three principles aimed at accounting for the implicational universals. The first is the 'heaviness serialisation principle',

Constituent Order: Structure and Function

41

which arranges heavier noun modifiers to the right of lighter ones. The second is the 'mobility principle', which states that adjectives, demonstratives and numerals are potentially moie mobile relative to their heads than are genitive or relativeclause modifiers. The third is the 'mobility and heaviness interaction principle', whose purpose is to resolve any conflicts between the first two principles. In general, the mobility principle outweighs the heaviness serialisation principle, but the latter overrides the former in cases where relative clauses are involved. Hawkins further suggests that these principles have their origin in the more general properties of syntactic structure, language history and psycholinguistics. Maxwell (1984) puts forward a single principle aimed at accounting for cross-linguistic generalisations relating to CO. Termedthe 'functor-sensehierarchy (FUSE-Η) ordering principle', it states in essence that elements which function semantically as links between two other constituents are likely to be positioned between them. However, Hawkins (1988a) makes a strong case against the possibility of accounting for universels by means of any one single principle. Another linguist who has advanced explanatory principles in the field of CO is Tomlin (1986). The first of Tomlin's proposals consists in the 'theme first principle', which embodies the well-known tendency for thematic material to be placed early. The second is the 'verb-object bonding principle', whereby the verb and object are more closely bound together than the verb and subject. The third is the 'animated first principle', according to which the highest noun phrase on a scale from human to inanimate tends to precede the other noun phrases with which it co-occurs. Tomlin claims (1986:126) that the relative frequency of languages of a given CO type is associated with the number of his principles to which they conform. A relationship between CO typology and morphological patterning has been noted by Nichols (1986:79-83) in her study of head-marking and dependentmarking languages. (The distinction at stake here is essentially whether a language indicates grammatical relations via the morphology of the head of a construction or that of the dependent member of the construction.) Nichols finds that languages of the head-marking type favour verb-initial ordering, while languages of the dependent-marking type prefer other CO patterns. Within the FG framework, Dik (1983c) sets down a number of principles relating to the normal order of constituents in the world's languages. Naturally, these principles echo some of those put forward by workers in other frameworks. In general, they refer to the tendencies for (i) languages to exhibit consistency in the relative position of what others call operator and operand; (ii) subject to precede object; (iii) elements bearing pragmatic functions to be placed early in the clause; (iv) elements to be placed in order of increasing complexity; and (v) 'relators' (such as adpositions and subordinators) to be positioned between their relata. (See further Kahrel 1985, Limburg 1985, Rijkhoff 1986.) Dik (1989b:352-5) adds some further principles relating to the positioning of other types of grammatical word, which he claims is influenced partly by their relationship of such words to

42

Constituent Order in Functional Grammar

the heads of the terms in which they occur38 and partly by idiosyncratic factors. It is possible, however, for the various principles to conflict, and it is not known whether there are rules governing the outcome in such circumstances. There is no doubt, therefore, that a good deal of work remains to be done in this area. As stated earlier, we shall be returning to the topic of CO typology in 4.4 below, when we deal with syntactic change. First, however, we must consider in detail the synchronic description of positional syntax in FG.

NOTES 1. The terni 'constituent order' has been preferred here to the frequently-used expression 'word order' on grounds of accuracy, since the present book is concerned with the positioning not only of words within phrases but also of larger structures within clauses and sentences. 2. The temi 'predicator' refers to the verbal element of the clause; this nomenclature is used by, for example, Halliday (1961:257). 3. This principle is discussed in relation to Tagmemic formulae by Cook (1969:160-92). 4. Previous attempts to enumerate the relevant factors include Jespersen (1949:53-61), Dover (1960:2-3), Allan (1987) and Siewierska (1988:ch.2). Hie present section draws upon this earlier work, but does not follow it in detail, being based instead on Connolly (1977:ch.2). 5. Expressions like Henry the Eighth do not constitute counter-examples to this rule, since the Eighth is a noun phrase (albeit an elliptical one) in its own right. 6. Unless otherwise stated, Dik's (1978a:ch.4) semantic categories will be used for descriptive purposes. 7. Cf. also Trnka (1968). 8. Writers on FSP include Benes (1968), Danes (1964,1967,1974a), Dvoíákoví (1964), Fiibas (1957, 1959a,b, 1961,1962, 1964a,b, 1965,1966,1970, 1971, 1974, 1975a,b, 1980, 1982, 1987), Hajicová (1984), Halliday (1974), Kuno (1972), Lipka (1977), Sgall (1969,1974), Svoboda (1968), Uhlffová (1974) and Vachek (1966:ch.5). 9. Bolinger's (1952) concept of 'linear modification' bears comparison with FSP, in that it refers to a progressive reduction, during the course of the sentence, of the 'semantic range' of what has preceded. 10. 'Presupposition' and 'focus' correspond to a large extent with 'given information' and 'new information', respectively. Note, however, that Chomsky (1965:221) employs the terms 'topic' and 'comment', though regarding these as marginal to the concerns of grammatical theory. 11. In fact, there is an obvious overlap between the terms 'focus' and 'emphasis'. 12. The expression slight downpour may be possible in an ironic sense, but would still not be contrastive even then. 13. Cf. also Eitel (1977:146-47). 14. See also Green (1980) for discussion of related issues. 15. English is not a good source of examples for this type of phenomenon. However, Gil (1982:132) suggests that English pronominal morphology is to some extent sensitive to purely positional factors. 16. For a summary of the Welsh mutations, see Rhys Jones (1977:326-41). These mutations give rise to inflectional alternants inasmuch as the latter are restricted in their grammatical distribution to a greater extent than would be expected on grounds of their part-of-speech (sub)classification alone. 17. This does not mean, however, that prosodies should not be clearly separated from CO, as has been claimed by Seiler (1962). 18. For further discussion of intonation in relation to the position of adverbials, see Schubiger (1946, 1964). 19. In recent work on FG, especially Hengeveld (1988) and Dik (1989b), clauses have been viewed as multi-layered structures, an idea derived from the work of Foley & van Valin (1984). However,

Constituent

Order: Structure and

Function

43

this innovation does not seem directly to affect the description of CO, and it will therefore not be further discussed in the present book. 20. De Schutter (1985) proposes a further special position, PO, as the final place within the functional pattern for the clause-proper. This proposal has some merit as a possible device for dealing with elements placed in clause-final position for emphasis, though it does not appear to be essential at least for the description of English CO, and has not been employed in the present book. See also Dik (1989b:365). 21. For an earlier comparative survey, see Vincent (1979). However, the latter does not coverall the main function ally-orientated approaches. An in-depth comparison of CO in FG, Lexical Functional Grammar, Generalised Phrase Structure Grammar and Word Grammar is presented in Siewierska (1988). 22. Also Garde (1977), except that the latter achieves initial concatenation though a single, general rule based on the direction of dependency between the elements being sequenced. 23. Anderson's version of Case Grammar differs on this point (among others) from that of Fillmore (1968). 24. On the other hand, certain authors, such as Bach (1975) and Werth (1979), have defended linear ordering in underlying structure. It may also be noted that Koutsoudas (1981) finds fault with the set of linearisation rules proposed by Sanders (among others). 25. In the context of X-bar theory, Falk (1983) suggests that constituency rules and ordering rules should be distinguished. This again appears to imply an unordered underlying structure. 26. Although Augmented Transition Networks are represented as diagrams rather than rules, when actually implemented on computers they do make use of operations analogous to movement rules; see Kaplan (1972:87). 27. The Syntax Crystal is an abstract device for mapping scope and dependency relations on to linear sequences. 28. See further Dik (1978a:183-4). 29. From the structural point of view, the Welsh Vi and O form a unit which in some ways resembles a nominalisation; cf. Awbery (1976:ch.l). However, since we are here concerned with functional patterns, Vi and O can reasonably be treated separately in that (i) they are functionally distinct and (ii) they functionally parallel Vi and O in other languages sufficiently to warrant a similar description for present purposes. 30. This is an illustration of the general point made by Myhill (1985) that the structural and pragmatic characteristics of elements such as subject are not identical across languages. 31. Cf. Connolly (1983:257). 32. There is a construction in Hindi with the verb cihie (see McGregor 1972:77-80) in which the orderisDOS V;thus, mujhe ekdaftar c&hte QïL 'me (acc.) a book want',i.e. Ί want a book'). Itwill be assumed for the present that this order can be handled by placing the direct object in Ρ1, rather than requiring a separate functional pattern. 33. The author is grateful to Mrs Surindera Singh for her help in the construction of the examples in (79). 34. Heinämäki (1976) even allows for the possibility of a mixture of two different underlying orders at the same time when a language is in a state of CO change. 35. Roughly, this means definite noun phrases. 36. Hawkins (1985) is a reply to criticisms levelled by Coopmans (1984). 37. For some recent discussion of CCH, see Dryer (1988), Hawkins (1988b). 38. See also Rijkhoff (1990) on this point

Chapter 3

Synchronic Functional Description

3.1 THE APPEAL OF THE FUNCTIONAL APPROACH

It is clear from the survey in 2.2 that CO is influenced by a wide range of factors, and that the great majority of these are, in one way or another, functional1 in nature. All the stylistic factors relate to either the thematic or the aesthetic functions of language, while the non-stylistic factors include both the semantic functions of clauses or constituents and their syntactic functional status. Indeed, it is difficult to find any influence on CO which is not in some way related to functional considerations. Even purely syntactic conventions, such as the rule concerning the position of determiners relative to nouns in English, are not free of functional qualification. For example, the rule just quoted applies only to determiner-noun pairs which stand in the functional relationship of modifier-head; it does not state that all determiners precede all nouns within the same sentence. In view of the above, the appeal of the functional approach to CO is manifest. Any non-functional alternative would necessarily be extremely impoverished. The attraction of the FG framework in particular is the way that, in principle, as indicated in 2.4.1, it allows the CO rules to be made sensitive to all the different types of pragmatic, semantic and syntactic influences which can reasonably be expected to be encompassed by a grammar. This is why it is worth elaborating a comprehensive CO mechanism for FG, a task to which the remainder of the present chapter is devoted.

3.2 TOWARDS A MORE ADEQUATE FRAMEWORK

Before turning our attention to the linearisation mechanism itself, it is necessary to give some preliminary consideration to the overall structure of the grammar. Two matters in particular need to be raised. One is the question, to be discussed in 3.2.1, of whether the expression component ought to be expanded to take into account a broad range of syntactic functions, rather than merely subject and object. The other concerns the representation of sentential functions such as statement or question; this will be dealt with in 3.2.2. Attention will then be focused upon the linearisation mechanism itself, and a proposal will be outlined in 3.2.3 which accommodates both free-order and fixed-order languages. This will lead to an

46

Constituent Order in Functional Grammar

altered status for functional patterns, as will be shown in 3.2.4. The proposed revised mechanism also lends itself to testing by means of a computer, the benefits of which procedure are indicated in 3.2.5. The result of all these proposals, which will be summarised in 3.2.6, will be a more adequate framework for the treatment of CO in FG. The proposed revisions of the FG linearisation mechanism were first outlined in Connolly (1983). However, as Gebruers (1984) points out in a review, they had at that stage not been fully worked out. The present chapter is, therefore, devoted to a comprehensive account of the operation of the modified mechanism. 3.2.1 Syntactic functions and the expression component The expression component of FG, as outlined in Dik (1978a:20-23, 157-58, 1989b:ch. 14), consists of several discrete groups of rules, or subcomponents. The rules within each subcomponent are ordered, as are the subcomponents themselves. The various subcomponents deal with the following phenomena: (1) (a) (b) (c) (d)

The basic form of constituents The order of constituents Adjustments to the form of constituents Prosody

The purpose of (la) is both to introduce certain grammatical words, such as prepositions and auxiliaries, and to assign case markings. The placement rules ( lb) have to follow (la) in order that they may specify the position of the said grammatical words as well as the items already present in the predication. However, as we saw in 2.3.1 above, some morphological phenomena (such as the contact mutations in Welsh) depend on the order of constituents. These have to be accommodated by means of the adjustment rules (lc). Finally, stress and intonation are specified (Id). The design of the expression component is well suited to handling the interactions between positional syntax, inflectional morphology andnon-segmental phonology that were described in 2.3 above. There is, however, oneproblem. This resides in the fact that CO rules appear to be sensitive to syntactic function; for instance, adverbials in English have very different positional characteristics from subjects, objects or complements. In current FG, however, syntactic functions are not specified, except in the case of subject and object, and we therefore have to consider whether adequate placement rules can be written without making reference to syntactic functions in general. If they cannot, then the expression component will have to be expanded to include a group of rules which assign a complete set of syntactic functions to constituents. Such a subcomponent would occupy a place between (la) and (lb), so that it could allocate functions to any items introduced by the former, while also providing an input to the placement rules. The question

Synchronic Functional Description

47

of whether or not it is necessary to add a new subcomponent to FG is obviously an important one. It is to this issue, therefore, that we must now turn. A distinction must first be drawn between Functional Linguistics, as a theory of language, and FG as a generative grammar. Any comprehensive functional theory of language must, by definition, account for all types of function at all the different levels, including the syntactic, and provide for an analytic as well as a generative framework. Syntactic analysis, in both traditional and modern linguistics, has involved and continues to involve the identification of syntactic functional elements, and functional categories are also used routinely in various areas of applied linguistics, such as speech therapy (see, for instance, Crystal, Garman & Fletcher 1989:passim) and natural language processing by computer (see, for example, Winograd 1983:216-22). Indeed, it is in this analytic spirit that, for instance, Chomsky (1965:70) is able to refer to categories like 'subject', or Dik (1978a:59,1989b: 117) to speak of 'modifier' and 'head', even though the terms in question do not play a direct role in the generative mechanisms that they have proposed. The real issue, then, is not whether the whole range of syntactic functions should be recognised within Functional Linguistics (since it is difficult to see how they could not be), but whether they should actually be specified as part of the process of generating a sentence in FG. This turns out to be quite a difficult question to answer. However, before discussing the problems involved, it is first necessary to draw a further distinction. We must differentiate between the strictly syntactic and the semantic aspects of the terms 'subject' and 'object'. Generally in linguistics these two terms are employed in a syntactic sense which, as with syntactic functional terms in general, designates their grammatical relationship to other elements in the same construction; cf. Longacre (1965:65), Touratier (1977:31). In FG, however, as noted in 2.4.1 above, they are used in an essentially semantic sense, to represent different vantage points or perspectives in the presentation of a state of affairs. Syntactically defined functional elements, by contrast, currently play no part in the generative apparatus of FG. Clearly, it is more parsimonious to dispense (if possible) with syntactic functions in the usual sense of the term. Accordingly, it may be argued that the present FG framework is the more economical and elegant for their absence. On the other hand, a grammar which specifies the whole range of syntactic functions is more complete than one which does not.2 Thus, it is possible to adduce a theoretical motivation either in favour of or against the inclusion of purely syntactically defined functional elements. From the empirical point of view, it is again difficult to find decisive evidence on either side of this question. To appreciate the problems involved, let us consider two types of structure where the issue arises, namely impersonal and coordinate constructions. In Modern English it is normal for declarative and interrogative clauses to have subjects, so that zero-valent verbs, such as rain, occur with the impersonal

48

Constituent Order in Functional Grammar

subject it, which is obligatory in clauses of the said types. From the syntactic point of view, therefore, the following sentences are possible: (2) (a) (b)

It is raining, He is writing.

In FG, however, only in (2b) is the subject function actually assigned, as the element concerned is present in the underlying predication in this sentence, whereas the same is not true of (2a). The parallelism of the functional syntactic structure between (2a) and (2b) is thus not accounted for in current FG, and their identity in respect of CO and verbal concord is presented as a mere coincidence.3 On the other hand, if we opt to include purely syntactic functions, we are likely to find that in some instances this leads to a more complicated description than would otherwise be required. For example, if direct and indirect objects are distinguished, a more elaborate set of placement rules will be required to handle their ordering in English than if we deal purely with the semantically-oriented functions Object and Recipient of conventional FG. In English coordinate constructions, the conjoining of nouns and interrogative pronouns is possible in certain circumstances, as in (3): (3) The interconnected tunnel and what? In this elliptical echo-question, the noun tunnel and the pronoun what are coordinated, and one of the essential corollaries of this is that they have a similar syntactic-functional status (in this case, head of noun phrase). Now, inFG the noun tunnel would be derived from a restrictor in the underlying term, whereas what would be derived from an operator (see Dik 1978a:62). The term underlying (3) would thus be something like (4): (4) ({dix,: tunnel,^,) & Qx.: entityN(Xj)}: interconnectedΑ(χ^ρ) The restrictor entityN(x.) is deleted by the expression rules to give the correct realisation. The result, however, is that tunnel and what are nowhere assigned the same syntactic function, so that FG lacks the means of showing any connection between their functional equivalence and their coordination. Note that what in (4) cannot be taken as standing for a whole nominal predicate, since this would imply that (4) could be paraphrased as the interconnected tunnel and the interconnected what?, which is, clearly, not the case. Accordingly, the evidence here seems to favour the inclusion rather than the exclusion of purely syntactic functional information. A further consideration relates to the typological classification of languages. Given two languages which are syntactically (say) SVO, but where one has (semantic) subject and object assignment while the other does not, the languages

Synchronic Functional Description

49

concerned cannot be considered as instances of the same CO type within the FG framework (cf. Dik 1980b: 132). The fact that they cannot be given the same categorisation serves to obscure a significant similarity between them. It would seem, then, that there are certain advantages in including syntactically defined functional elements within FG-based descriptions. This is not to say that such elements have been shown conclusively to be indispensable, but on the other hand such empirical evidence as has been adduced above does point in favour of their inclusion. Furthermore, it must be emphasised that parsimony should not be pursued at the expense of omitting significant information from linguistic descriptions, and the author considers that because a broad range of syntactic functional categories are used in real and important applications of grammatical description, those categories ought to be defined as legitimate aspects of sentence structure; and the clearest and most explicit way of achieving this is to incorporate them into the generative mechanism. We shall thus proceed on the basis that the expression component consists of groups of rules dealing with all of the following: (5) (a) (b) (c) (c) (e)

The basic form of constituents The assignment of syntactic functional categorisations The order of constituents Adjustments to the form of constituents Prosody

The exact format of the rules in (Sb) will not be discussed here, as they are beyond the scope of this book. In what follows, however, it will be assumed that they are part of the grammar,4 and that placement rules may make reference to any syntactic function whenever this proves either necessary or the simplest option. 3.2.2 Semantic functions and the predication-formation component As far as the predication-formation component is concerned, few changes are necessary for the purpose of facilitating an adequate CO mechanism. However, there is one piece of information which needs to be included in predications and was not originally provided by the FG predication generation apparatus, though it is included in more recent work (for instance, Dik 1986a, 1986b, 1989a,b), namely a direct specification of sentential function. The means of representing this function used by Dik in the publications just mentioned is to employ an operator referring to the whole clause, as exemplified in (6) and (7). (6) (a) (b) (7) (a) (b)

DECL {PERFpolishidlx,: D a v i d ^ x ^ ^ d l x , : c a r ^ ) ) ^ } David has polished the car. INT {PERF polishidlx,: DavidN(x1))A PI place (plconstit):plconstit(X), asserta(pos(X, pi)), dec.

(36) may be loosely translated as the following instruction to the computer. Your goal is to place the PI -constituent, and you will achieve it if you accomplish a series of sub-goals. These are: firstly that you find there is, in fact, some element (X) in this clause which is a PI-constituent; secondly that you add to your store of knowledge the assertion that the position of X is to be PI; and thirdly that you decrease by one the number of elements remaining to be placed. Thus, if the clause contains a PI-constituent, it will be placed in PI, by means of the sub-instruction 'asserta(pos(X,pl))\ whereas if the clause does not contain such a constituent, PI will, naturally, not be filled by it. As may be gathered from this example, the encoding of a set of placement rules involves embedding the actual positioning instructions within a considerably larger and more complex framework, since the running of the program needs to be controlled by means of a number of instructions and sub-instructions. Writing such a program is, therefore, a fairly complicated and time-consuming process. The reward, however, is a set of rules which, when translated back into English, can be demonstrated to produce the output claimed of them. Although this does not guarantee their empirical correctness, it does represent an important step in the direction of descriptive adequacy. Computer-aided rule testing has been used in FG before, as the rules contained in Dik (1979b) were tested by means of a program written in ALGOL 68 by Kwee Tjoe Liong. It is believed, however, that the work described in Connolly (1986) was thefirstpublished attempt to employ PROLOG for the testing of FG placement rules. PROLOG is now being used by several others engaged in computational FG; see van der Korst (1987), Bakker (1989), Dignum (1989), Gatward (1989), Samuelsdorff (1989).9·10 The use of PROLOG for the above purpose has had a further benefit. In the placement rules proposed in Connolly (1983) the formulation of the rule-conditions was less disciplined than it might have been, in that four different types of annotation wereemployedasseemedfit,namely 'obligatory', 'unmarked', 'marked* and 'impermissible'; and the term 'otherwise' was used whenever convenient. However, the PROLOG implementation forced theabandonment of the 'otherwise' device and a reduction of the number of annotations. 'Obligatory' had to be dispensed with, and we are now left with just two binary options: 'permissible' and

60

Constituent Order in Functional Grammar

'impermissible', and if permissible then 'marked' or 'unmarked'. Thus, the computer-aided testing of the rules has led in more than one way to a real increase in scientific rigour. 3.2.6 Summary It is appropriate at this point to summarise the proposals that have been advanced in the present section. Firstly, the expression component should contain a subcomponent whose purpose is to assign syntactic functional categories to the constituents of clauses and phrases. Secondly, the predication formation component should include a means of specifying the full range of sentential functions. This is necessary prerequisite to the formulation of placement rules for different types of sentence. Thirdly, the placement rules should operate in conjunction with syntactic templates rather than functional patterns, the latter being retained but now with a derived status rather than as part of the placement apparatus itself. This will enable the linearisation mechanism to cope with free-order as well as fixed-order languages, without losing the benefits which the functional patterns confer. Fourthly, the placement rules should, where appropriate, be qualified by means of conditions. This is indispensable to the writing of rules which account for the data, and it also provides a means of differentiating marked and unmarked orderings. Finally, computer-aided testing of placement rules is to be recommended as it leads to an otherwise unattainable degree of rigour in the formulation of such rules. The overall result of these proposals is to provide FG with more adequate framework for the treatment of CO.

3.3 CLAUSE-RANK CONSTITUENTS

In 3.2.3 above it was shown that the revised linearisation mechanism may be applied in a satisfactory manner to the clause-rank CO of a free-order language like Latin. The purpose of the present section is to show that it is equally capable of providing a functional description of clause-rank CO in a fixed-order language. Present-day English will be used as the language of exemplification. It is not intended to give a complete description of English CO, as this is not possible in the present state of knowledge (and probably never will be).11 The aim, rather, is to demonstrate how the revised mechanism can accommodate the various kinds of phenomena involved. 3.3.1 Major elements Let us consider, first of all, the placement of major elements in English clausesproper containing no more than one element of any particular syntactic category. For this purpose the following syntactic template is necessary:

61

Synchronic Functional Description (37)

P1N1N2N3N4N5

The rules which operate in conjunction with this are designed first of all to place in PI any constituent which belongs there, and then to allocate the remaining constituents to the basic positions. The latter process proceeds along the following lines. The subject is placed in N2, and any objects or complements in N4 or N5. Both the finite and non-finite parts of the predicator are placed in N3, except in questions and in those declarative clauses where inversion is possible or necessary; in such cases the finite part is inserted into N1. The rules themselves are presented in (38). Discrepancies between these and the rules in Connolly (1983), which they supersede, are due in large part to the extra rigour in formulation imposed by computer-aided testing. With regard to the rule-conditions, it should be noted that the potential ambiguities associated with the connectives and or or have been resolved by the use of indentation. This has been done in preference to the use of bracketing conventions, which tend to be less readable. On a point of terminology, it should be noted that the expression verb phrase (which appears in the following and in several subsequent sets of rules) is used in this book to refer just to the verbal complex, and is not intended to include the non-subject arguments of the verbal predicate. (38)

(a)

(b)

Pl-constit

-> PI

Foe

-> PI

(if the clause is not an echoquestion) (if PI is vacant and either Foe and Top coincide or the clause contains more than one Foe. MKD if the element placed in PI is not Subj)

For constituents not placed in PI: Subj -> N2 IObj -> N4 (if IObj is not prepositional or MKD if IObj is prepositional and the clause contains a non-pronominal DObj which bears Foe and is not in PI) or N5 (if IObj is prepositional) (if N4 is vacant) DObj -> N4

Constituent Order in Functional Grammar or N5

Comp

-> N4 or N5

Vf

-> NI

(if the verb is divalent and N5 does not contain IObj and PI contains neither IObj nor Comp and either the clause does not contain a Comp or the clause contains a Comp and a non-pronominal DObj which bears Foe. MKD if the clause contains a Comp) (if N4 is vacant) (if the verb is trivalent and NS is vacant and PI contains neither IObj nor DObj) (if the clause is a non-echo wh-interrogative or the clause is a polar interrogative or the clause is an optative or the clause is bare conditional or PI contains an adverbial or PI contains a degree Comp or PI contains a negative or degree DObj or prepositional IObj and the verb phrase contains an auxiliary or PI contains an inverting adjective or PI contains a quoted DObj and the verb phrase contains no auxiliaries or PI contains Vn with Top function and both Subj and Vn carry Foe and Subj is non-pronominal

63

Synchronic Functional Description

or N3

Vn

and Vf is be or MKD the clause contains a quoted DObj not placed in PI and the verb phrase contains no auxiliaries) (if the clause is not bare conditional and the clause is not a whinterrogative other than an echo-question and the clause is not optative and PI does not contain an adverbial and it is not the case that: PI contains a negative DObj or prepositional IObj and the verb phrase contains an auxiliary. MKD if the clause is polar interrogative)

-> N3

The rules and their output will now be explained in detail. Let us again begin with those clauses where PI isunfilled. In declarative mainclausesofthiskind,theSubj is placed in N2 and the finite part of the Pred (Vf) is put in N3, as also is the nonfinite part of the Pred (Vn) if there is one. Thus the following patterns are generated: (39) (a) (b)

They cheered, They had cheered.

The placement of objects and complements depends partly on the valency of the construction, in that position NS is made available only if the construction is trivalent (i.e. contains two objects or else an object plus a complement). In bivalent constructions, such as those in (40), the DObj or Comp is placed in N4. (40) (a) (b)

They applauded him. They were hot.

When a DObj and IObj co-occur, the latter is placed either in N4 or N5, as appropriate, and the DObj is then inserted into whichever of these positions

64

Constituent Order in Functional Grammar

remains vacant. If the IObj is notaprepositional phrase it is put into N4, as in (4 la); otherwise its normal position is N5, as in (41b). However, a prepositional IObj can also be placed in N4, as in (41c), provided that the DObj is non-pronominal and bears focus. These two last-mentioned conditions, however, rule out the sentences in (42). (41) (a) (b) (c) (42) (a) (b)

We showed our friends the picture of the farm. We showed the picture of the farm to our friends. We showed to our friends the picture of the FARM. MKD *We showed to our FRIENDS the picture of the farm, *We showed to our friends it.

In constructions with an IObj plus a Comp, such as (43), the former, being nonprepositional, is placed in N4 and the latter in N5. (43) She made him a good wife. Where DObj co-occurs with a Comp, the unmarked position of the former is N4, as in (44a), but N5 is possible as a marked option provided that the DObj bears focus and is not pronominal, so that (44b) is allowed but (44c) is not (44) (a) (b) (c)

She found the long-term option preferable. She found preferable the LONG-term option. MKD *She found preferable it.

The placement of constituents in imperative clauses and in echo-questions follows the same pattern as in statements. This produces orderings such as: (45) (a) (b)

Hand your ticket to the guard, Hand what to the guard?

In polar interrogative clauses, however, the unmarked position of Vf is in Nl, though a marked ordering with Vf in N3 is also permissible: (46) (a) 0>)

Was it enjoyable? It was enjoyable? MKD

(It is assumed that the appropriate intonation is ensured by means of later expression rules.) In optative clauses like (47a), and also in conditional clauses like (47b) which lack subordinato», and which will be referred to here as bare conditionals, the placement of Vf in Nl is obligatory: (47) (a)

May it flourish!

65

Synchronic Functional Description (b)

Had it been snowing...

To complete the explanation of the rules, we must now consider those clauses in which PI is occupied. In relative clauses and in wh-questions (excluding echoquestions) PI is filled by the wh-element: (48) (a) (b) (c) (d)

... who advised you ... ... whom you advised ... Who saw you? Whom did you see?

In the case of the questions, though not the relative clauses, Vf is placed in Ν1. This does not, of course, lead to improper inversion in (48c), since the subject is in PI rather than N2. Subordinato^, too, are PI -constituents (see Dik 1989b:360) and are placed accordingly: (49) Before you arrived... Moreover, certain adverbials may act as Pl-constituents if the clause contains no other constituents of the latter type;12 examples include seldom and rarely. Such items are accompanied by inversion, and their presence thus causes Vf to be placed inNl: (50) Seldom had he been so shocked. Note that because the inversion is obligatory in such circumstances, it is necessary to block the insertion of Vf into N3 by means of a condition on the 'Vf -> N3' rule; otherwise the non-inverted order would also have been generated as an alternative. If PI has not been filled by a PI-constituent, it may be occupied under certain conditions by an element bearing the pragmatic function of Foe, resulting in a marked order unless the constituent concerned is Subj. This placement of focused elements in PI occurs under two circumstances. Firstly, it happens if the element also carries the function of Top. In English, as Hannay (1983:220) points out, Top function or Foe function alone is not sufficient to bring an element into initial position. However, it is contended here that if an element fulfils both these functions simultaneously, then it does occur at the beginning of the clause: (51) This RED wine I prefer. MKD Secondly, if the clause contains two elements with Foe function (cf. Dik et al. 1980:43), then one or other of these is placed in PI: (52) (a)

A VASE I gave to ANN.

MKD

66

Constituent Order in Functional Grammar (b)

To ANN I gave a VASE.

MKD

The placement of a focus-bearing element in Ρ1 often has an effect on the operation of the remaining rules. Several of the rules relating to N4 and N5 make reference to the contents of PI. For example, DObj may not be placed in N5 if PI contains either IObj or Comp. The motivation of this restriction is the desire to avoid redundancy. Sentence (55b), for instance, has its IObj in PI and its DObj in N4. Had it not been for the restriction just mentioned, however, the same order could have been redundantly generated by placing the DObj in N5. The rules relating to N1 and N3, too, are affected by the presence of focused elements in PI. These allow inversion if PI contains (i) a Comp expressing degree, as in (53), or (ii) a DObj or prepositional IObj which either has a negative meaning, as in (54a,b), or expresses degree, as in (55a,b), or (iii) an inverting adjective such as happy, as in (56), or (iv) a DObj which is a direct quotation, as in (57a). However, the peimissibility of the inversion is in some cases dependent on the composition of the verb phrase. It is possible with a quoted object only if there are no auxiliaries (compare (57a) and (57b)), while, on the other hand, the presence of such items is essential for inversion following negative or degree objects, hence the ungrammaticality of, for instance, (54c). (53) So complete was his victory... MKD (54) (a) Not a second had he lost. MKD (b) To none of his friends had he sent letters of that kind. MKD (c) *Not a second lost he. (55) (a) Such a commotion had he caused. MKD (b) Only to such a greatly valued customer would he have given that large reduction. MKD (56) Happy are such people. MKD (57) (a) 'It was an accident', said he. MKD (b) ""It was an accident,' had he said. (c) Said he, 'It was an accident.' MKD It should be noted that, as illustrated in (57c), inversion is possible with a quoted DObj even if the latter is not in PI. Usually, inversion is optional after a focused element in PI. Consequently, the alternative, uninverted order is generated alongside the inverted order in such circumstances. For instance, the rules allow for (58) as well as (53) and (59) alongside (56). (58) So complete his victory was ... MKD (59) Happy such people are. MKD Such uninverted orders are generated because nothing in the conditions on the ' V

Synchronic Functional Description

67

-> N3' rules serves to block them. There is, however, an exception. If a negative DObj or IObj is placed in PI within a clause whose verb phrase contains an auxiliary, then the inversion is obligatory and the placement of Vf in N3 therefore forbidden. Hence, for example, (60) is not generated alongside (54b). (60) *Not a second he had lost A problem is presented by sentences such as (61): (61) Drunk all the wine he had! MKD where Vn carries the functions of both Foe and Top and is then placed in PI, but is followed immediately by DObj, with Subj and Vf being postponed to the end. A convenient way to deal with this kind of situation is to establish the following convention: (62) If Vn functions as both Top and Foe, the nuclear elements of the clause are to be split, prior to the placement of its constituents, into two groups: (i) Subj and Vf; (ii) the residue, comprising Vn plus any other nuclear elements. The rule 'Foe -> P I ' is then to be taken as applying to the whole residue, which is thus inserted into PI, and Subj and Vf are then placed by the succeeding rules in the normal way. The residue can now be ordered internally by simply passing it through the same set of rules, i.e. those in (38). The result is a sequence such as that illustrated in (61). If Vf is the auxiliary be and the subject is non-pronominal, and if the subject as well as Vn carries Foe, then inversion is possible, as in (63): (63) Leading the field is our old friend. MKD A further complication is occasioned by the possibility of placing a wh-element at the beginning of a main clause instead of the subordinate clause to which its syntactic function relates. An example is seen in (64), where the element that is the object not of the main clause but of the subordinate clause, whose predicator is to say: (64) What do you want me to say? Quirk et al. (1972:397-8), in their discussion of this construction, term such a whelement as 'pushdown Q-element'. Further examples are given in (65-68):

68

Constituent Order in Functional Grammar (65) (a) What would you like me to become? (b) How long do you want me to wait? (66) What a time we expected we would have! (67) This is the book which I said you could borrow. (68) What do you want me to ask them to say?

These examples illustrate three conditions which, it appears, have to be met before the construction in question may occur. These, which will here be called the pushdown conditions, are as follows. Firstly, the pushdown element must be a DObj, as in (64) or (66-68), a Comp, as in (65a), or an adverbial, as in (65b). Secondly, the element must relate to a subordinate clause which either functions as the DObj of the clause into which the element concerned is raised,13 as in (6467), or is embedded within such a clausal direct object, as in (68). Thirdly, the clause into which the pushdown element is raised must be either a wh-question, as in (64), (65) or (68), a wh-exclamation, as in (66), or a relative clause, as in (67). Given these conditions, we can now handle the above instruction by introducing the following convention: (69) When the pushdown conditions are satisfied, a constituent eligible for placement in PI is to be inserted into position PI not in its own clause but in the highest superordinate relative clause if there is one, or otherwise into the main clause. It was stated above that the rules in (38) and the syntactic template in (37) apply to clauses in which there is not more than one element of any particular syntactic category. Consequently, they need extending to cope with constructions of the type seen in (70) and (71). (70) (a) It is appropriate to congratulate him. (b) There are no winners. (71) We find it difficult to believe your message. (70a) and (70b) both exemplify constructions which have two subjects. In (70b), the pronoun It is the anticipatory subject, while the non-finite clause to congratulate him represents what will here be called the subject-proper. In (70b) the anticipatory subject is There and the subject-proper no winners. Sentence (71), on the other hand, contains only one subject, but it has two direct objects, the pronoun it and the non-finite clause to believe your message. The former will be referred to here as the anticipatory object and the latter as the object-proper. In order to accommodate the constructions mentioned in this paragraph, it is necessary to replace the syntactic template in (37) by that in (72): (72) PI N1 N2 N3 N4 N5 N6 N7

69

Synchronic Functional Description

The new positions, N6 and N7, are to be used for the placement of the subjectproper and object-proper. The latter can be placed in N6 and the former in N7. As for the placement of the anticipatory subject and object, we may adopt the following convention: (73) Anticipatory subjects are to be placed by means of the rules applying to subjects generally, and anticipatory objects by means of those applying to direct objects generally. In other words, the anticipatory elements are to be handled by the rules in (38). The extension of the syntactic template to the form shown in (72) brings a further benefit Consider the following sentences: (74) (a) (b)

Few men survived the course who had been ill-prepared. MKD We gave the people food who were hungry. MKD

Both these sentences contain examples of discontinuous postmodification. In (74a) the postmodifying clause who had been ill-prepared is separated from the rest of the subject to which it belongs (i.e. Few men ), while in (74b) a similar discontinuity affects the indirect object, the people... who were hungry. In order to generate such structures we have first to detach such postmodifiers from their heads and then place them in position N7. The first of these two stages will be dealt with in the section on the noun phrase (3.4.1 below), but the second can easily be incorporated into the same set of rules as place the subject-proper in N7 and the object-proper in N6: (75)

Subject-proper Detached modifier

-> ->

N7 N7

Object-proper

->

N6

(MKD if a constituent intervenes between the detached modifier and its head)

The reason for the condition on the second of these rules is as follows. If the postmodifier is separated from its head, as in (74a) and (74b), then a marked order results, and this has to be indicated by means of a rule-condition. However, it is possible for the modifier of an element to be detached, but for the element concerned then to be placed in such a position by the rules in (41) that it ends up as the last nuclear element of the clause. If this happens, the result is the same as though the modifier had never been detached in the first instance, and the order is thus not marked. For example, in the sentence (76)

We gave food to the people who were hungry. MKD

70

Constituent Order in Functional Grammar

it makes no difference whether the postmodifier of the IObj was detached and placed in N7 or whether the modifier was not detached and the whole IObj was placed in N5. This situation is, of course, an instance of neutralisation, whereby the distribution between the marked and unmarked placement of the postmodifier is nullified.14 Another point which must be explained is the motivation for having two positions, N6 and N7, in the extended syntactic template, rather than just one extra position, N6. The reason for this is that it is possible for the same clause to contain both an object-proper and a detached modifier. For example, in the following sentence (77) Those people found it hard to cope with the strike who lacked their own transport. MKD we find not only the object-proper, to cope with the strike, but also the detached postmodifier of the subject, who lacked their own transport. The correct ordering is achieved by inserting the former into N6 and the latter into N7. As was made clear at the beginning of the present section, the rules given in (38) and (75) apply to the clause-proper, and thus exclude elements carrying the function of theme or tail. These last-mentioned elements are placed within the appropriate clause by means of the rules proposed by Dik himself, which are presented in 2.4.1 above. 3.3.2 Minor elements The minor elements of the clause, it will be recalled from 2.4.1 above, are the adverbial (A), vocative (Voc), coordinator (Coord) and subordinator (Sub). The placement of subordinators can be handled by treating them as Pl-constituents, as suggested in 3.3.1 above. Coordinators may have either a clause-rank or phraserank function, and will be dealt with in 3.4 below. For the present, therefore, we shall confine our attention to adverbials and vocatives. Previous woik on adverbial positioning, such as Ellinger (1904), Western (1906, 1908:18-49), Fijn van Draat (1921), Poutsma (1928:427-59), Curme (1931:ch.l6), Buyssens (1933), Zandvoort (1962:281-87), SJacobson (1964, 1970,1975,1978),Koefoed(1967), Hartvigsen (1969), Leech & Svartvik (1975:197202), Enkvist (1976) and Hicks (1976), indicates that the rules governing this area of syntax are very difficult to formulate with precision. A complete generative description of the placement of adverbials is, therefore, unattainable in the present state of knowledge. What follows below, then, is an attempt to show in principle how the phenomenon may be handled in FG, aiming for approximately the level of detail found in Leech & Svartvik (1975), which acts as our main source here. Even so, it will be considerably more comprehensive than is typical of generative treatments of adverbial order.

Synchronic Functional Description

71

The placement of the adverbials and vocatives in FG may be dealt with by means of rules and syntactic templates comparable to those which we have already encountered. The requisite template is to be formed by adding a series of extra positions to (72), so as to allow for the minor elements to be placed either between pairs of major elements or at the beginning or end of the clause: (78) MO PI Ml N1 M2 N2 M3 N3 M4 N4 M5 N5 M6 N6 M7 N7 M8 As can be seen, the new positional slots are labelled MO to M8 ('M* for 'minor'), and are arranged in such a way that each N-slot has the same index-number as the preceding M-slot. In addition to the M-slots, two other positions in (78) are available for the placement of minor elements. These are PI and N3. As already noted, PI is used for adverbials like seldom which cause invasion, while N3 is the position into which minor elements must be inserted if they are to appear within the verb phrase, as in (79): (79) I have already hurried. The ordering of items which occur within the verb phrase will be dealt within in 3.4.2 below. The placement of adverbials depends on (among other things) their semantic function. Dik (1978a:50) presents a comprehensive semantic classification of satellites, and this has been employed, with certain modifications, as the basis for the present treatment The following categories and groupings are recognised here: (80) (a)

Temporal type (Temp): (i) Time (Tm), e.g. at midnight. (ii) Duration (Dn), e.g. for an hour. (b) Locative type (Loc): (i) Place (PI), e.g. in Abergynolwyn. (ii) Comitative (Cm), e.g. in his friends' company. (c) Process type (Proc): (i) Manner (Mn), e.g. slowly. (ii) Instrument (Is), e.g. with a hammer. (iii) Attendant Circumstance (Ac), e.g . without looking. (d) Agentive type (Ag), found in passive clauses only: (i) Agent, e.g. [watched] by a neighbour. (ii) Position, e.g. [retained] by the lender. (iii) Force, e.g. [demolished] by a storm. (e) Relative type (Rei): (i) Degree (Dg), e.g. to a large extent.

72

Constituent Order in Functional Grammar (f)

(g)

(h)

(i)

(ii) Comparison (Cp), e.g. than you. Rationale type (Rat): (i) Cause (Ca), e.g. because of his personality. (ii) Result (Rs), e.g. with predicable consequences. (iii) Purpose (Pr), e.g. for his own ends. (iv) Beneficiary (Bn), e.g. for our sake. Qualification type (Qual): (i) Condition (Cd), e.g. if in doubt. (ii) Concession (Cs), e.g. despite this. Special adjunct types: (i) Negative/restrictive (Nr), e.g. seldom. (ii) Focusing (Fc), e.g. only. Peripheral type (Peri): (i) Disjunct (Dt), e.g. probably. (ii) Conjunct (Ct), e.g. therefore.

The chief differences between Dik's classification and the one used here are as follows. Dik's category of 'quality' has been excluded, as its sense is not made clear. 'Reason' has been subsumed under 'cause', 'frequency' under 'time', and 'location','source', 'direction' and'path'under'place', since the distinctions in question do not seem to be important in the present context. The term 'attendant circumstances' is taken from Curine (1931:259) as being slightly clearer than just 'circumstance'. The traditional categories of 'degree' and'comparison' havebeen added, as have 'disjunct' and 'conjunct'. (The terminology in (8Qg,h) is borrowed from Quirk et al. 1972:ch.8, as is the term 'process' adverbial.) A further addition is the category of 'agentive' type; although such elements are nuclear rather than satellite in status, syntactically they constitute adverbials in passive clauses (see Quirk et al. 1972:346). The grouping of individual categories of adverbial into types also differs from Dik's, again because it is better suited to our current purpose. In this connection two points need special mentioned. Firstly, following Quirk et al. (ibid.), all non-peripheral adverbials will be referred to as adjuncts. Secondly, the reason for identifying special adjunct types must be explained. Negative/ restrictive adverbials like seldom need to be singled out because of their peculiar capacity forplacementin PI (see 3.3.1 above), while focusing adjuncts such asonly are subject to different positional conventions from other adjuncts, so that they, too, must be treated separately here. Another factor influencing the placement of adverbials is their structure, and in particular whether they are clausal (i.e. comprise or contain a clause) or nonclausal. With the exception of comparatives, clausal adverbials tend to have much the same positional characteristics whatever their semantic function, whereas this is not the case for non-clausal adverbials. In the rules that follow, therefore, clausal adverbials (excluding comparatives) are treated as a single class, and where

Synchronic Functional Description

73

adverbials are referred to in terms of their semantic function, it is to be understood that allusion is being made to the non-clausal variety Comparative clauses, however, are handled separately from other clausal adverbials. 332.1 Placement of individual elements The rules for placing adverbials and vocatives are divided into several groups. The first group, which operate in conjunction with syntactic template (78), deal with the placement of non-focusing adverbials (i.e. all adverbials except focusing adjuncts) and vocatives, and employ the abbreviation E for 'Element to be placed': (81) (a) E -> PI

(b) E -> MO

or Ml

(if PI is vacant and either E is Nr or E is Dg and is introduced by so or E is so with 'additional' meaning or E is Tm or PI and the verb is intransitive and the verb phrase contains no auxiliaries and the subject is non-pronominal) (if PI is occupied by a PI-constituent and either E is Tm orLoc or Ca orBn orQual orCt or Voc or clausal; MKD if E is Bn; or PI is occupied by an element other than a PI-constituent or a focus-bearing major element and E is PI or Ca orQual or Peri or Voc or clausal) (if Ν1 is occupied and either E is Tm orLoc oris or Ac or Ca

74 Constituent Order in Functional Grammar and PI does not contain a Pl-constit; MKD if E is Is or Ac; or E is Qual or Voc or clausal; MKD if PI contains a Pl-constit; or E is Bn and PI is vacant; MKD if this condition is met; or E is Ct orEisDt and PI contains an adverbial) orM2 (if N2 is occupied and either the clause is polar interrogative and E is Ct or the clause is not interrogative and E is Tm or Rat or Qual or Peri or Voc or clausal or E is Loc and does not bear Foe or E is Dn and is prepositional or the clause is affirmative and E is Dn and is a noun phrase or an adverbial phrase; MKD if the latter; or E is Proc or Dg; MKD if E is Is or Ag; or E is PI and bears Foe) or M3 (if N3 is occupied and eitherE is Tm or Nr, MKD if E is a noun phrase or is prepositional; or E is Dg; MKD if the clause is copular; or E is Dn and the clause is not copular; MKD if E is not an adverbial phrase; or E is Dn and the clause is copular and E is a simple adverb; MKD if this condition is met; or E is Mn and the clause is not copular;

Synchronic Functional Description

or N3

or M4

75

MKD if E is prepositional; or E is Is or Ac and the clause is not copular; MKD if these conditions are met; or E is Cm or Rat orQual or clausal; MKD if this condition is met; or E is Peri or Voc) (if N3 contains more than one word and either E is Temp; MKD if E is a noun phrase or prepositional; or E is Dg or Nr or Peri or E is Mn and the clause is not copular; MKD if E is prepositional; or E is Is or Ac and the clause is not copular or E is Loe or Rat orQual or Voc or clausal; MKD if this condition is met) (if N4 is occupied and there is not a pronominal DObj and there is not a non-prepositional IObj and either E is Temp or Cm orProc orDg or Rat orQual or Peri and the clause is not copular; MKD if these conditions are met; or E is PI and the clause is not copular and the verb is not trivalent; MKD if these conditions are met; or E is Temp and the clause is copular;

Constituent Order in Functional Grammar

or M5

or M6

M K D i f E i s a noun phrase or prepositional; or E is Nr or Peri and the clause is copular or E is Loc or Dg or Rat or Qual or Voc or clausal and the clause is copular; MKD if these conditions are met) (if N5 is occupied and either E is Tm or Mn; MKD if N5 does not contain IObj; or E is Ca orPr or Qual orNr or Voc; MKD if this condition is met; or EisDn or Cm orDg orRs or Bn and N5 does not contain Comp; MKD if these conditions are met; or E is PI and N5 does not contain Comp and there is not a combination of either Comp in N4 and DObj inN5 or a prepositional IObj in N4 and DObj in N5; MKD if these conditions are met; or E is Peri; MKD if N5 does not contain Comp) (if N6 is occupied and either E is Tm or Ct or E is Dn orDg

77

Synchronic Functional Description

or M7

or M8

or Ca orPr or Bn or Cd or Voc; MKD if this condition is met) (if N7 is occupied and either E is Tm orCt or E is Dn orDg or Ca orPr orRs orQual orDt or Voc or clausal; MKD if this condition is met) (if either N7 is vacant and E is Temp or Loc orProc or Ag or Rei or Rat orQual orNr or Peri or Voc or clausal or clausal Cp; MKD if E is Dg and an adverbial phrase or E is Nr; or N7 is occupied and E is Temp or Dg or Ca orRs orQual or Peri or Voc or clausal Cp;

78 Constituent Order in Functional Grammar MKD if E is Dg) Rule (81a) is to be taken as applying after (38a) but before (38b) and (75). (81b), on the other hand, applies after (38b) and (75). The purpose of (8 la) is to place certain kinds of adverbial in PI, provided that this position is not already occupied by a major element. If it succeeds, then the 'Vf -> ΝΓ rule in (38b) will subsequently bring about the requisite inversion of subject and finite verb. PI may be filled by a negative/restrictive adverbial like seldom, as in (82a), a degree adverbial introduced by the word so, as in (82b), or so on its own, acting as an adverbial with meaning of 'in addition', as in (82c). Adverbials of time or place can also be inserted intoPl if the verb phrase consists of a single intransitive verb and the subject is non-pronominal, as in (83a) and (83b). (82) (a) (b) (c) (83) (a) (b)

Seldom had he been so surprised. So thorough had his investigations been... So will he. Then came the election, There is Margaret.

Rule (81b) is very complex, as there are so many positions besides PI that adverbials and vocatives may occupy. These will now be considered in turn. Position MO is available to adverbials or vocatives if PI is occupied. If PI contains a Pl-constituent, then MO may receive an adverbial of time, as in (84a), a locative adverbial (84b), or an adverbial of cause (84c) or beneficiary (84d); in the lastmentioned case, a marked ordering is produced. Alternatively, MO may receive an adverbial of qualification (84c), a conjunct (84f), a vocative (84g) or a clausal adverbial (84h). (84) (a) (b) (c) (Φ (e) (f) (g) (h)

At present who is the chairman of the committee? In England who is the best-known sportswoman? Because of this, who could advise us? For Fred's benefit who can be expected to stand down? MKD Despite this who volunteered to go? Therefore who is likely to refuse? Jim, where have you been? When you return, who willrememberyou?

IfΡ1 is occupied by an element other than a Ρ1 -constituent or a focus-bearing major element, then MO may receive an adverbial of place (85a), cause (85b) or qualification (85c), a peripheral adverbial (85d), a vocative (85e) or a clausal adverbial (85f).

Synchronic Functional Description (85) (a) (b) (c) (d) (e) (f)

79

In this firm seldom does anyone receive rapid promotion. Because of this, seldom did she go out alone. Despite this, seldom did he refuse a challenge. Therefore, seldom is there anything else to do. Jim, seldom have I been so pleased. Although he was willing, seldom was he asked to speak out

Position Ml is available if N1 is occupied (by the finite verb). If PI does not contain a PI -constituent, then Ml may receive an adverbial of time (86a), a locative adverbial (86b), or an adverbial of instrument (86c) or attendant circumstances (86d), the latter two options giving rise to marked orderings. (86) (a) (b) (c) (d)

Tomorrow will you bring your books? In this country seldom do people discuss their personal finances. With a chisel could you open it? MKD Without looking can you guess what it is? MKD

Whatever the contents (if any) of PI, an adverbial of qualification, a vocative or a clausal adverbial may be placed in Ml. Examples are seen in (87), (88) and (89), respectively. However, as the (b) examples illustrate, a marked ordering results if PI contains a Pl-constituenL (87) (a) (b) (88) (a) (b) (89) (a) (b)

In that case will you come? Who, in that case, is the culprit? MKD Joe, is it time to go? Who, Joe, is the secretary? MKD If you find out, will you let me know? Why, if you knew, didn't you tell me? MKD

If PI is vacant, an adverbial of the beneficiary category may be placed in Ml, though this is again a marked order: (90) For his benefit would you do it? MKD Peripheral adveibials, too, may be placed in Ml, as can be seen with the conjuncts in (91). However, in the case of disjuncts, as in (92), this is possible only if PI contains an adverbial. (91) (a) Who, therefore, is to be invited? (b) Seldom, therefore, has such a sight been seen. (92) Seldom, perhaps, has such a sight been seen. Position M2 can be filled if N2 is occupied (by the subject). In polar interrogative

80

Constituent Order in Functional Grammar

clauses, a conjunct may be placed here: (93) Did, therefore, the contractors agree? Otherwise, M2 is available only in clauses that are of neither the polar interrogative nor the wh-interrogative kind. In non-interrogative clauses, M2 may receive an adverbial of time (94a), rationale (94b), or qualification (94c), a peripheral (94d), a vocative (94e) or a clausal adverbial (94f). (94) (a) (b) (c) (d) (e) (f)

Immediately he protested. For that reason she refused. In spite of the problems he continued. Consequently she departed. Jim, you are late. When you come, you will see our new pictures.

With regard to other types of adverbial, the question of whether the clause is affirmative or negative becomes important. In general, the placement of adverbials of duration (95), process (96) or degree (97) in M2 is permitted only in affirmative clauses, and the same is also true of focus-bearing adverbials of place (98). If the adverbial is one of instrument (96b) or agent (96c) or if it is one of duration realised by an adverbial phrase (95b), then a marked ordering results. (95) (a) All the time they remained cheerful, (b) Always he would ask after her. MKD (96) (a) In this way he prevailed upon them. (b) With an axe he chopped the wood. MKD (c) By his father he was given a watch. MKD (97) To a large extent he succeeded. (98) In the CUPBOARD he found it. Adverbials of duration represented by aprepositional phrase may, however, appear in M2 in either affirmative or negative clauses: (99)

(a) (b)

For a while she slept. For a while she could not get to sleep.

The same is true of adverbials of location that do not bear focus: (100)

(a) (b)

In the cupboard he found some food, In the cupboard he did not find any food.

If N3 is occupied (by the finite and/or non-finite verb), then a minor element may

Synchronic Functional Description

81

be inserted into M3. The element concerned may be an adverbial of time ( 101 ) or a negative/restrictive adverbial (102), though a marked ordering will result if it is represented other than by an adverbial phrase (101b, c, 102c). (101)

(102)

(a) (b) (c) (a) (b)

He soon agreed to comply. He this week agreed to comply. MKD He at some stage agreed to comply. MKD He seldom obliged. He under no circumstances obliged. MKD

Alternatively, the element involved may be an adverbial of degree (103) though this will give rise to a marked ordering if the clause is copular (103b). (103)

(a) (b)

He in large measure accepted the argument. He in large measure was the person responsible. MKD

An adverbial of duration, too, may be placed in M3. If the clause is non-copular, as in ( 104), then there is no further restriction, except that a marked order will result unless the element is represented by a adverbial phrase: (104)

(a) (b)

He temporarily halted. They for a time inhabited the forest MKD

If, however, the clause is copular, an adverbial of duration may be inserted into M3 only if it is realised by a single adverb, and even then the resulting pattern will be marked: (105)

He temporarily was absent MKD

An adverbial of manner may also be placed in M3, as long as the clause is not copular (106), but the ordering will be marked if it is represented by a prepositional phrase (106b). (106)

(a) (b)

She easily coped. She with ease coped. MKD

Adverbials of instrument (107a) or attendant circumstances (107b) may likewise be inserted into M3 in a non-copular clause. A marked ordering will result in either case. (107)

(a) (b)

He, with a stick, tapped the window. MKD He, without looking round, dashed out MKD

82 Constituent Order in Functional Grammar Whether the clause is copular or not, it is possible for M3 to receive a comitative adverbial (108a), or an adverbial of rationale (108b) or qualification (108c), a clausal adverbial (108d), a peripheral adverbial (108e) or a vocative (108f). Once again, marked orderings result, except in the case of peripherals and vocatives. (108)

(a) We, in our friends' company, heard the sad news. MKD (b) We for that reason refused. MKD (c) We in that case must come. MKD (d) He, before he leaves, should tidy up. MKD (e) He perhaps should have refrained. (f) You, Jim, ought to write.

In clauses where the predicator contains more than one word, minor elements may be place in N3, as mentioned above. Temporal adverbials may be inserted here (109), as may adverbials of degree (110a), negative/restrictive adverbials (110b) or peripheral adverbials (110c). In the case of temporals, the ordering will be marked if the constituent is a noun phrase (109b) or a prepositional phrase (109c). (109)

(110)

(a) (b) (c) (a) (b) (c)

We have now seen everything. We have this week seen everything. MKD We have in a very short time seen everything. MKD We have to a large extent been satisfied. They have seldom been challenged. She will possibly have remembered.

If the clause is non-copular, adverbials of manner (111), instrument (112a) or attendant circumstance (112b) may also be placed in N3. A marked ordering will result in the case of a manner adverbial realised by a prepositional phrase (11 lb). (111) (112)

(a) (b) (a) (b)

He can readily be consulted. He can with difficulty be consulted. MKD Hehas,throughhardwork,madeacommendablecontribution. He can, without realising, offend people.

Whether or not the clause is copular, locative adverbials (113a), adverbials of rationale (113b) or qualification (113c), vocatives (113d) or clausal adverbials (113e) are similarly permitted in N3, but produce marked orderings. (113)

(a) (b)

You can, in London, obtain such things. MKD They have, for a secret purpose, concealed information. MKD (c) He will, despite his misgivings, complete the job. MKD (d) You should, Joe, be aiming to win. MKD

Synchronic Functional Description (e)

83

You can, when you come, buy plenty of goods. MKD

If N4 is occupied (by an object or complement), then M4 is available, provided that the clause contains neither a pronominal direct object nor a non-prepositional indirect object. If the clause is not copular, a marked ordering may be produced by inserting into M4 a temporal adverbial (114a), a comitative adverbial (114b), an adverbial of process (114c), degree (114d), rationale (114e) or qualification (114f) or a peripheral adverbial (114g). (114)

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

He had finished now his first assignment MKD He spent with us an enjoyable weekend. MKD He revealed in some ways his worst points to us. MKD She showed to some extent her capabilities. MKD He may disclose, with predicable results, this news to the media. MKD She will retain in that case a good deal of respect. MKD I'll send perhaps a small amount to you. MKD

The same applies to an adverbial of place, provided that the verb is not trivalent: (115)

We placed on his desk the complete file. MKD

If the clause is copular, then M4 may receive a temporal adverbial ( 116), a negative/ restrictive adverbial (117a) or a peripheral adverbial (117b). In the case of the temporal, the resultant order will be marked if the adverbial is a noun phrase ( 116b) or a prepositional phrase (116c). (116)

(117)

(a) (b) (c) (a) (b)

She is now much happier. She was, last week, much happier. MKD She is for the moment much happier. MKD We were seldom alone, He was certainly peevish.

A marked order will also result from inserting a locative adverbial (118a), an adverbial of degree (118b), rationale (118c) or qualification (118d), a vocative (118e) or a clausal adverbial (118f) into M4 in a copular clause. (118)

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

She was at school a very able mathematician. MKD He is, to a degree, correct. MKD He was for that reason a fearsome opponent MKD He is, in that case, reasonably welcome. MKD You are, my friend, somewhat mistaken. MKD You are, when you say such things, very provocative. MKD

84

Constituent Order in Functional Grammar

Position M5 is available if N5 is occupied (again by an object or complement). An adverbial of time (119) or manner (120) may be placed in this position, though a marked order will result, as in (119b), (120b), if N5 does not contain an indirect object. (119) (120)

(a) (b) (a) (b)

We shall give our advice now to the clients. We shall give the clients now our advice. MKD We shall read this interesting report carefully to them. We shall read them carefully this interesting report. MKD

Marked orderings will also result from inserting into MS an adverbial of cause (121a), purpose (121b), or qualification (121c), a negative/restrictive adverbial (121d) or a vocative (121e). (121)

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

We showed them for some reason the wrong chart. MKD Heboughtus,forouredification,someweightytomes. MKD We found her, in spite of everything, an indispensable help. MKD I will lend them under no circumstances that new radio. MKD I consider this, friends, a great honour. MKD

Moreover, if N5 does not contain a complement, then it is possible to produce marked orderings by placing in MS an adverbial of duration (122a), a comitative adverbial (122b), or an adverbial of degree (122c), result (122d) or beneficiary (122e). (122)

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

We lent them temporarily some warm blankets. MKD He bought his wife, in my presence, a huge bouquet MKD They asked us to some extent the same questions. MKD They showed her, with unexpected results, the latest designs. MKD We lent them, for the baby's benefit, a fan heater. MKD

In clauses where NS does not contain a complement it is also possible to insert an adverbial of place into MS, provided that there is not a combination of a direct object in NS with either a complement or a prepositional indirect object in N4. Such a placement is, therefore, possible in an example like (123a), whether the direct object precedes the indirect, or ( 123b), where the indirect precedes but is not prepositional. In either case, however, a marked ordering results. (123)

(a) (b)

He handed the pencil, in the waiting room, to his accomplice. MKD He passed me under the table a bottle of wine. MKD

Synchronic Functional Description

85

A peripheral, too, may be placed in M5, as in ( 124). This will give rise to a marked order if, as in (124b), N5 does not contain a complement. (124)

(a) (b)

We found it, therefore, quite a task. We gave a lift, therefore, to them. MKD

In clauses containing both an anticipatory object and an object-proper, the fact that the latter will have been placed in N6 makes position M6 available. Here may appear an adverbial of time (125a) or a conjunct (125b). So too may an adverbial of duration (126a), degree (126b), cause (126c), purpose (126d), beneficiary (126e) or condition (1260, ora vocative (126g), though in these instances a marked ordering will result (125) (126)

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

They find it pleasant nowadays working at home. They find it pleasant, therefore, working at home. They found it convenient, temporarily, working at home. MKD They find it pleasant, to some extent, working at home. MKD They find it pleasant, for that reason, working at home. MKD They will find it useful, for such purposes, working at home. MKD He found it tolerable, for their sake, working at home. MKD They will find it pleasant, in that case, working at home. MKD You will find it pleasant, Jim, working at home. MKD

If the clause contains a subject-proper or the detached postmodifier of a major element, then the placement of this in N7 makes position M7 available. Again, an adverbial of time (127a) or a conjunct (127b) may be inserted here without a marked ordering resulting. A marked order is, however, produced if M7 receives an adverbial of duration (128a), degree (128b), cause (128c), purpose (128d), result (128e) or qualification (1280, a disjunct (128g), a vocative (128h) or a clausal adverbial (128i). (127) (128)

(a) (b) (a) (b) (c) (d)

It is appropriate now to let the matter rest. It is appropriate, therefore, to let the matter rest. It has been appropriate, for some time, to be very cautious. MKD It was possible, to a large extent, to satisfy demand. MKD It made those people happiest, for obvious reasons, who wished to stay. MKD Those officials were available, to a large extent, who were new to the job. MKD

86

Constituent Order in Functional Grammar (e) (f) (g) (h) (i)

It was appropriate, for the purpose of finding out more, to listen to the radio. MKD It was appropriate, despite the counter-arguments, to proceed. MKD It was appropriate, probably, to abandon the project. MKD It made those people angriest, Jim, who were unable to protest. MKD It was appropriate, when he entered, to give him our attention. MKD

Whether N7 is occupied is also relevant to the placement of adverbiale in M8. If it is, then M8 may receive an adverbial of time (129a), degree (129b), cause (129c), result (129d) or qualification (129e), a peripheral (1290, a vocative (129g) or a clausal adverbial of comparison (129h). In the case of an adverbial of degree a marked ordering results if N7 is occupied. (129)

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

It was necessary to wait, in those days. It was appropriate to go, to some extent. MKD It was appropriate to wait, because of the weather. It became dangerous to stay, with predictable results. It was good to attend, despite our initial reservations. It was reasonable to object, therefore, It is good to see you, Margaret It was better to agree than to start arguing.

If N7 is not occupied, however, then other types of adverbial as well may be placed in M8. These include a locative (130a) or process adverbial (130b), an agent (130c), a non-clausal adverbial of comparison (130d), an adverbial of purpose (130e) or beneficiary (130f), or a negative/restrictive adverbial (130g). The placement of the last-mentioned type of adverbial in M8 will, however, produce a marked ordering. If an adverbial of degree ( 131) is inserted into M8 when N7 is unoccupied, then a marked ordering will result only if the adverbial is realised by an adverbial phrase, as in (131b). (130)

(131)

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

They came here. They told their tale amusingly. They were seen by everyone. They are larger than that. He is cunning for his own ends. He will do it for her sake. He visited her seldom. MKD He is right to some extent.

Synchronic Functional Description (b)

87

He missed the bus, nearly. MKD

The rules in (81) are designed to operate on one minor element at a time, and define the range of positions in which that particular element may be placed within the clause in which it occurs. If the clause contains more than one such element, then complications arise. The fact is that more than one element may be allocated to the same position. When this happens, they need to be ordered relative to one another within that position. This matter will be dealt with later. The second complication is that if two adveibials of the same semantic category co-occur, and if one has a more general and the other a more particular meaning, then their placement is restricted in that normally both must be inserted into the same positional slot. For instance, in ( 132) the temporal adverbials at seven o' clock and on Friday stand in a general-particular relationship, the latter being the more general in meaning. They therefore occupy the same position (in this instance M8). (132)

I awoke at seven o'clock on Friday.

However, it is also possible for the more general one to be placed early in the clause (i.e. in MO, Ml or M2). This is what happened in (133), where on Friday has been allocated to M2. (133)

On Friday I awoke at seven o'clock.

These observations necessitate that it be stipulated that the rules in (81) are subject to the following convention: (134)

If the element being placed is an adverbial A t and if it stands in a general-particular relationship with an adverbial A^ that has already been placed, then either A t must be placed in the same position as Aj or, if Aj is in a position to the right of N2, then as an alternative, Aj may be placed in MO, Ml or M2 as appropriate.

As was made clear earlier, a full account of adverbial positioning would require a more detailed set of rules than those presented here. Greater attention would, for example, have to be paid to the peculiarities of particular adverbs, such as ever, which are subject to special ordering restrictions. These peculiarities would be handled by refining the rule-conditions accordingly. However, that must be left for further research, and we must now pass on to another issue, namely the placement of focusing adjuncts. Focusing adjuncts, such as only, merely or just, require a different set of placement rules compared with other adverbials. The rules in question, which again apply in conjunction with syntactic template (78), and are ordered after those in (81), are the following:

Constituent Order in Functional Grammar (135)

A -> MO (if focused element is in PI) or Ml (if either focused element is in PI and A can follow focused element; MKD if these conditions are met; or focused element is in Nl) or M2 (if either focused element is in Nl and A can follow focused element; MKD if these conditions are met; or focused element is in N2) or M3 (if either focused element is in N2 and A can follow focused element; MKD if these conditions are met; or focused element is in N3 and there is not a finite auxiliary in N3 or focused element is in N4 and there is not a finite auxiliary in N3 or focused element is in N5 and there is not a finite auxiliary in N3 or focused element is in N6 and there is not a finite auxiliary in N3 or focused element is in N7 and there is not a finite auxiliary in N3) or N3 (if either focused element is in N3 and there is a finite auxiliary in N3 or focused element is in N4 and there is a finite auxiliary in N3 or focused element is in N5 and there is a finite auxiliary in N3 or focused element is in N6 and there is a finite auxiliary in N3 or focused element is in N7 and there is a finite auxiliary in N3) or M4 (if focused element is in N4) or M5 (if either focused element is in N4 and A can follow focused element; MKD if these conditions are met; or focused element is in N5) or M6 (if either focused element is in NS and A can follow focused element; MKD if these conditions are met; or focused element is in N6) or M7 (if either focused element is in N6 and A can follow focused element;

Synchronic Functional Description

89

MKD if these conditions are met; or focused element is in N7) or M8 (if focused element is in N7; MKD if this condition is met) (136)

A -> or or or

MO Ml M2 M3

or N3

or or or or or

M4 M5 M6 M7 M8

(if focused element is in MO) (if focused element is in Ml) (if focused element is in M2) (if either focused element is in M3 or focused element is in M4 and there is not a finite auxiliary in N3 or focused element is in MS and there is not a finite auxiliary in N3 or focused element is in M6 and there is not a finite auxiliary in N3 or focused element is in M7 and there is not a finite auxiliary in N3 or focused element is in M8 and there is not a finite auxiliary in N3) (if either focused element is in M4 and there is a finite auxiliary in N3 or focused element is in M5 and there is a finite auxiliary in N3 or focused element is in M6 and there is a finite auxiliary in N3 or focused element is in M7 and there is a finite auxiliary in N3 or focused element is in M8 and there is a finite auxiliary in N3) (if focused element is in M4) (if focused element is in M5) (if focused element is in M4) (if focused element is in M7) (if focused element is in M8)

Rule(135) applies if the element focused by the adjunct occupies either PI or one of the N-positions. If the focused element is in PI, N1 or N2, then the focusing adjunct may be placed in the M-position either immediately before it or after it, except in the case of certain adveibials, e.g. just, which cannot follow the element to which they apply15 and may therefore be placed only in the M-position immediately to the left of that element. Focusing adjuncts like only, on the other hand, can follow as well as precede the elements to which they apply, though a marked ordering results if they follow. To take a concrete example of the

90

Constituent Order in Functional Grammar

application of (135), consider a clause in which the subject (in N2) is focused by the adjunct only. In such circumstances the rule will generate the following alternative orders: (137)

(a) (b)

Only the president survived. The president, only, survived. MKD

If, however, the subject had been focused by the adjunct just, then only one ordering would have been generated: (138) Just the president survived. If the focused element is in N3, then the adjunct is placed in M3 unless there is a finite auxiliary also in N3, in which case the adjunct is itself placed in N3 as well. For example, suppose that in the following two sentences, the verb touched is focused by the adjunct only: (139)

(a) (b)

He only touched the sculpture. He had only touched the sculpture.

In (139a) the adjunct has been placed in M3, as there is no finite auxiliary in N3. In (139b), however, N3 does contain a finite auxiliary, and the adjunct has, therefore, also been placed in the N3 slot (The internal order of items within N3 will be dealt with in 3.4.2 below.) If the focused element is in N4-N7, then there are various possibilities. One is for the adjunct to be placed in the M-positions immediately to the left of the focused element. Alternatively, as long as it is capable of following the focused element, it may be inserted into the M-position immediately to the right of the latter. A further possibility is for it to be placed either in M3 or N3, depending once again on whether there is a finite or auxiliary in N3. All these possibilities are illustrated in turn in (140), where the adjunct only is focusing the object one sweet, which has been placed in N4: (140)

(a) (b) (c) (d)

I had wanted only one sweet. (Focusing adjunct in M4) I had wanted one sweet only. MKD (Focusing Adjunctin M5) I only wanted one sweet. (Focusing Adjunct in M3) I had only wanted one sweet. (Focusing Adjunct in N3)

Rule (136) applies when a focused element is in one of the M-positions. In such cases it is always possible for the focusing adjunct to be assigned to the same position as the focused minor element. However, if the focused element is in any of the positions M4-8, then there is the additional possibility of the adjunct being placed in M3 or N3, in accordance with the same principles as in Rule (135). For

91

Synchronic Functional Description

example, in the following sentences the focused element is the adverbial recently, which has been placed in M8, and the focusing adjunct is only: (141)

(a) (b) (c)

He had arrived only recently. He only arrived recently. He had only arrived recently.

(Focusing adjunct in M8) (Focusing adjunct in M3) (Focusing adjunct in N3)

The internal ordering of elements within M8 will be dealt with by means of the rules to be presented in the next sub-section. 3.3.2.2 Sequencing of co-occurring elements in particular positions When more than one minor element has been allocated to the same positional slot, such as M8, they have then to be ordered relative to one another within that slot through the operation of further sets of rules. The first set is given in (143), and the associated syntactic template is presented in (142). The symbol E here represents a vocative or an adjunct other than a focusing adjunct (142) (143)

XF1AF2 Detached extension Focusing adjunct

E

-> X -> Fl (if it focuses a major element to the left) or F2 (if it focuses a major element to the right) -> A

Leaving aside for a moment the idea of a 'detached extension', we see that the rules in (143) distinguish between focusing adjuncts on the one hand and the remaining types of adverbials, together with the vocatives, on the other. Focusing adjuncts precede these other types of minor element if they focus on a major element to their left. For instance, in ( 144) the focusing adjunct only and the process adverbial with great care have both been allocated to M8, but because only focuses on the preceding major element (the object), it is placed before with great care. (144)

Open this cupboard only, with great care. MKD

By contrast, if a focusing adjunct applies to a major element to its right, then it immediately precedes that element and thus follows any other minor element sharing the same slot For example, in ( 145) the temporal adverbial yesterday and the focusing adjunctjust have both been inserted into M2, but becausejust applies to the following major element (the subject), it is placed after yesterday. (145)

Yesterday just one member was ill.

92

Constituent Order in Functional Grammar

As for the 'Detached Extension', this is a product of the verb phrase rules to be described in 3.4.2 below. An extension16 is a preposition or adverb that follows the verb in expressions like look at or open up, and in certain circumstances it is possible for an extension to be detached from its verb and placed later in the clause. For example, in (146) it follows the object: (146)

They opened the garage up.

Such detached extensions occupy the first position within the slot in which they are placed, as can be seen in the following examples, where only focuses on the garage: (147)

They opened the garage up, only, with great care. MKD

As Leech and Svartvik (1975:201) observe, it is not normal for more than two adverbials to be placed in the same position, except at the end of the clause. We therefore add the following convention: (148) Except in respect of M8, if the rules in (143) place more than two adverbials within the same positional slot, then the resulting ordering is marked. For instance, the occurrence of three adverbials in M3 in (149) produces a marked ordering: (149)

He slowly in the garden yesterday worked out the solution to his problem. MKD

In this last example, the three adverbials are all non-focusing adjuncts, which will, therefore, all have been placed in position A in (142). The ordering of such adjuncts within that positional slot must now be considered. For this purpose, we require syntactic template (150) and the rules in (151). (150) Al A2 ... Ai (151) Process adverbial

Locative adverbial

-> Al or A2 or A3 etc. -> Al

or A2

(if vacant; MKD if it precedes a nonclausal process adverbial) (if vacant; MKD if either it precedes a non-

93

Synchronic Functional Description

clausal process adverbial or it is non-clausal and a clausal adverbial precedes it in the template) or A3

Temporal adverbial

etc. -> Al

or A2

or A3

(if vacant; MKD if either it precedes a nonclausal process adverbial or it is non-clausal and a clausal adverbial precedes it in the template) (if vacant; MKD if it precedes a nonclausal process adverbial or a non-clausal locative adverbial) (if vacant; MKD if either it precedes a nonclausal process adverbial or a non-clausal locative adverbial or it is non-clausal and a clausal adverbial precedes it in the template) (if vacant; MKD if either it precedes a nonclausal process adverbial or a non-clausal locative adverbial or it is non-clausal and a clausal

Constituent Order in Functional Grammar adverbial precedes it in the template) Other adverbial

etc. -> Al

or A2

or A3

(if vacant; MKD if it precedes a nonclausal process adverbial or a non-clausal locative adverbial or a non-clausal temporal adverbial) (if vacant; MKD if either it precedes a non-clausal process adverbial or a non-clausal locative adverbial or a non-clausal temporal adverbial or it is non-clausal and a clausal adverbial precedes it in the template) (if vacant; MKD if either it precedes a non-clausal process adverbial or a non-clausal locative adverbial or a non-clausal temporal adverbial or it is non-clausal and a clausal adverbial precedes it in the template)

95

Synchronic Functional Description etc.

Template (150) differs from those encountered so far, in that it is elastic. That is, instead of containing a fixed number of positions, it is deemed to contain as may positions as required, given the total number (f ) of elements to be placed, the minimum being two. S imilarly, t limits the range of alternatives within each rule. As it is difficult to formulate absolute constraints on the relative order of adjuncts within a single positional slot, the rules in (151) are designed to allow for any permutation, but to stipulate that a marked ordering results under certain conditions. These, too, are hard to determine accurately, but they may be stated in a highly provisional way as follows. Firstly, a marked ordering occurs if a locative adverbial precedes one of process unless the latter is clausal. Hence, (152b) is marked, while (152a) is not (since the process adverbial here precedes the locative), nor is (152c) either (since the process adverbial, although second, is clausal). (152)

(a) (b) (c)

Things happen quickly here. Things happen here quickly. MKD He escaped from prison by bribing the warder who was in charge of him.

Secondly, a marked ordering results if a temporal adverbial precedes a non-clausal process or locative adverbial. Thus, (153b,d) are marked, while (153a,c) are not (153)

(a) (b) (c) (d)

They treated her with respect for a time. They treated her for a time with respect MKD They live in Preston now. They live now in Preston. MKD

Thirdly, a marked ordering is produced if any other type of adjunct precedes a nonclausal adverbial of any of the three types just mentioned. Hence, (154b,c,d) are marked, while (154a) is not, since the adverbial of rationale is the last of the four. (154)

(a) (b) (c) (d)

He walked slowly to work that day because of his knee. He won the race, despite this, with ease. MKD He drove the car, for her benefit, to the garage. MKD He hid there, because of his anxiety, for several hours. MKD

Fourthly, whatever the semantic category of theadverbials involved, the placement of a clausal adverbial to the left of a non-clausal one within the same positional slot results in a marked sequence, as in (155).

96 Constituent Order in Functional Grammar ( 155) She went to the house where she knew her friend lived, immediately. MKD The condition 'if vacant' in the rules in (151) is to be interpreted as meaning 'if not already occupied by an adjunct of the same broad semantic classification as denoted by the terms on the left hand sides of the rules'. It is, however, possible that more than one adjunct of the same category will be placed into a single slot within the template (150). Accordingly, further rules are needed in order to sequence such elements, which are here represented as E l t E2, etc. These are presented in (157) and operate in conjunction with a further elastic template: (156) XI X2 ... Xf (157) E1 -> XI or X2 or X3 etc. E2 -> XI (if vacant; MKD if either process adverbials are being placed and a prepositional phrase is positioned to the left of an adverbial phrase or an adverbial of more general meaning precedes one of more specific meaning and either the more general one is clausal or neither is clausal) or X2 (if vacant; MKD if either process adverbials are being placed and a prepositional phrase is positioned to the left of an adverbial phrase or an adverbial of more general meaning precedes one of more specific meaning and either the more general one is clausal or neither is clausal) or X3 (if vacant; MKD if either process adverbials are being placed and a prepositional phrase is positioned to the left of an adverbial phrase

Synchronic Functional Description

97 or an adverbial of more general meaning precedes one of more specific meaning and either the more general one is clausal or neither is clausal)

etc. E3 -> XI (if vacant; MKD if either process adverbials are being placed and a prepositional phrase is positioned to the left of an adverbial phrase or an adverbial of more general meaning precedes one of more specific meaning and either the more general one is clausal or neither is clausal) or X2 (if vacant; MKD if either process adverbials are being placed and a prepositional phrase is positioned to the left of an adverbial phrase or an adverbial of more general meaning precedes one of more specific meaning and either the more general one is clausal or neither is clausal) or X3 (if vacant; MKD if either process adverbials are being placed and a prepositional phrase is positioned to the left of an adverbial phrase or an adverbial of more general meaning precedes one of more specific meaning and either the more general one is clausal or neither is clausal) etc. These rules allow any permutation of the adverbials concerned, but again make

98

Constituent Order in Functional Grammar

stipulations concerning markedness. Firstly, if process adverbials are being placed, then a marked ordering results where a prepositional phrase precedes an adverbial phrase within the same positional slot Thus, for instance, (158b) is marked, while (158a) is not. (158)

(a) (b)

He painted slowly with his brush. He painted with his brush slowly. MKD

Secondly, if two adverbials of any semantic category stand in a general-particular relationship (cf. 3.3.2.1 above) and the more general precedes the more specific, then a marked ordering is produced either if the one with the more general meaning is clausal or if neither is clausal. Consider the following examples: (159)

(a) (b) (c) (d)

She stayed in a hotel in London. She stayed in London in a hotel. MKD She stayed in the city which she liked best, in a hotel. MKD She stayed in London in a hotel which I had previously recommended to her.

Sentence (159a) is unmarked because in it the more specific expression in a hotel precedes the more general in London. In the three other categories, however, the more general expression comes first. (159b) is marked because neither adverbial is clausal. So too is ( 159c), where the more general expression, in the city which he liked best is clausal, while (159d) is unmarked, since in this case the more specific expression has clausal status. So far we have said little about the placement of vocatives that have been allocated to the same position as other minor elements by the rules in (81). In fact, it does not seem that there are any strong constraints on this aspect of CO, with the result that the rule in question is very simple. The rule for placing a vocative relative to other adjuncts in the same positional slot is given in (161), and requires the following template: (160) (161)

a Xb Voc -> X

The purpose of rule (161) is to take a vocative which has been placed in position A in template (142) and place it immediately before or after any adjunct which has been inserted into the same position in the same template, whether or not that adjunct has been further processed by means of the rules in ( 151 ) or ( 157). Such adjuncts are denoted by a and b in template (160), though one (but not both) of a and ft may be the slot boundary. In either event, α andò must be contiguous prior to the application of rule (161). Suppose, to begin with, that just one adjunct has been placed in the same

Synchronic Functional Description

99

positional slot as the vocative. To take a concrete example, let the vocative be Jim and the adjunct quickly. In that case rule (161) may either treat α as the left-hand boundary of the slot and b as quickly, yielding the order illustrated in (162a), or else it may treat a as quickly and è as the right-hand boundary, yielding the order seen in (162b). (162)

(a) (b)

Run, Jim, quickly. Run quickly, Jim.

Now let us take a more complicated example. Suppose that the vocative needs to be placed relative to three adjuncts, quickly, to the shed and if you can. In this instance four alternative orders are generated: (163)

(a) (b) (c) (d)

Run, Jim, quickly to the shed if you can. Run quickly, Jim, to the shed if you can. Run quickly to the shed, Jim, if you can. Run quickly to the shed if you can, Jim.

These four different orders are obtained by letting a and b have the following values. For (163a) α is the left-hand boundary and b is quickly; for (163b) α is quickly and b is to the shed; for (163c) a is to the shed and b is if you can; and for (163d) α is if you can andò is the right-hand boundary. A further topic not so far mentioned is the placement of peripheral adverbials relative to other minor elements which have been allocated to the same position by the rules in (81). Here the procedure is very similar to that for the placement of vocatives which has just been described. The rule required for the purpose is (164), which operates in association with a template identical to (160). (164)

Peripheral adverbial -> X

In this case, the variables α and 6 in the template may either denote slot boundaries (as before) or else they may denote any adverbial or vocative in the same positional slot as the peripheral to be placed. (These other elements may or may not have been further processed by means of the rules already described in this section). However, neither α nor è may stand for a detached extension. To take a concrete example, suppose that the peripheral adverbial therefore needs to be placed relative to the adverbials to London and every day. By letting α andò assume different values, the various alternatives can be generated, as may be seen, for instance, in the context of the following sentences. (165)

(a) (b) (c)

He travelled, therefore, to London every day. He travelled to London, therefore, every day. He travelled every day to London, therefore.

100 Constituent Order in Functional Grammar In (165a) a represents the left-hand boundary of the slot and b denotes to London; in (165b) a and b denote the two adjuncts; and in (165c) a denotes every day and b the right-hand boundary. Finally, we must deal with the placement of focusing adjuncts which apply to minor rather than major elements and thus are not handled by the rules in (143). For this purpose we require rule (167), which operates in conjunction with the following template: (166) (167)

A

aFlbCdFle -> FI (MKD if ò is not null) or F2 (if A can follow focused element; MKD if this condition is met)

The symbols in the template have the following meanings. Fl and F2 are the positions into which the focusing adjunct may be inserted; C is the focused constituent; a and e may each denote either the appropriate slot boundary or a minor element; b may denote a peripheral or may be null; and d may denote a peripheral or a vocative or it may be null. Rule (167) will place the focusing adjunct either immediately to the left of the focused element φ taking a null value) or immediately before any peripheral preceding that element. These alternatives may be seen in the following examples, where the adjunct only focuses on the process adverbial through good luck, and where both of these have already been placed in M8 along with the peripheral adverbial presumably. (168)

(a) (b)

He survived, presumably, only through good luck. He survived only, presumably, through good luck. MKD

If the focusing adjunct is one which, like only, can also follow the focused element, then rule (167) will generate the appropriate alternative or alternatives, labelling the latter as marked. Examples are seen in ( 169) and ( 170) where only focuses on the temporal adverbial today. (169) (170)

(a) (b) (a) (b)

They are available today only, therefore. MKD They are available today, therefore, only. MKD They are available today only, Fred. MKD They are available today, Fred, only. MKD

In (169a) and (170a), d is null; in (169b) it stands for a peripheral; and in (170b) it represents a vocative. We have now assembled a fairly comprehensive battery of rules and templates for the placement of adverbials. As stated above, the rules need to be refined through further research, but they will serve to demonstrate how the placement of adverbials may be handled in principle by FG. Two particular

Synchronic Functional Description

101

features of this treatment are worth drawing attention to at this point. The first is that the process has been tackled in a step-by-step manner. To have attempted to perform the whole task with a single set of rules would have led to quite intractable complexity. The second feature is that a good deal of functional as well as formal information has been incorporated into the rule-conditions. Indeed, both types of information have proved indispensable. Similar remarks also apply to the placement of phrase-rank elements, and it is to this subject that we now turn our attention.

3.4 PHRASE-RANK CONSTITUENTS

The rule and template approach to CO is applicable to the placement of items within phrases in much the same way as to the positioning of elements within clauses. In what follows, we shall demonstrate this in relation to all the different phrase-types of English. 3.4.1 The noun phrase The noun phrase (NP) is one of the most complex of the English phrase-types as far as CO is concerned. The head, which is normally a noun or a pronoun, can be preceded and/or followed by modifiers of several different types. These may be either words, phrases or subordinate clauses, and even in the case of single words, a variety of different parts of speech are possible: (171) (a)

(b) (c) (d)

Determinatives, which include: (i) Predeterminers (ii) Determiners (iii) Postdeterminers Adjectives Nouns Adverbs

The terms for the different types of determinative are used in the sense of Quirk et al. (1985:ch.5). Determiners include the definite and indefinite article, the demonstratives this and that, and various other items with a similar distribution. Predeterminers are words like all or half, which can precede determiners, while postdeterminers are words (mostly numerals, whether cardinal or ordinal) that follow any determiner but precede any adjectives within the same phrase. Of these different parts of speech it is undoubtedly the adjectives that are the most difficult to handle with regard to CO. Various writers have made attempts to describe the conventions of adjective ordering in English; see, for instance, Teyssier (1968), Martin (1969,1970), Danks & Glucksberg (1971), Martin & Ferb (1973), Quirk et al. (1972:922-7), Hetzron (1978). However, no-one has so far

102 Constituent Order in Functional Grammar succeeded in formulating a definitive set of rules governing the positioning of adjectives, any more than has been done for adverbials. As the treatment of the matter in Quirk et al. (1972:922-7) is probably the most comprehensive, this will be the basis of the following further refinement. In order to describe the positioning of adjectives, it is first necessary to classify them semantically. The following categorisation is to be used here: (172) Adjectives of: (a) Age, e.g. old (b) Colour, e.g. green (c) Provenance, e.g. French (d) Nature, e.g. social (e) Other, e.g. beautiful In respect of adjectives of nature, it is to denominal adjectives of this category, such as social, rather than to non-denominal ones, such as peculiar, that reference is made in the placement rules below. Rules for the placement of items within the NP are presented in (174). Because the range of items involved is so large, a fairly long syntactic template (173) is required to go with them. The positions within the template are labelled 'W' (for 'word', as it is typically words that are the constituents of phrases). (173) (174)

W1W2W3 W4 W5 W6 W7 W8 W9 W10 Head -> W7 Predet -> W1 Det -> W2 Cardinal -> W4 Ordinal -> W5 Adj -> W8 (if either it is of the obligatorily postposed type or the head is a pronoun ending in -body, -one or -thing ) or W6 (if it is not of the obligatorily postposed type and the head is not a pronoun ending in -body, -one or -thing ) or W8 (if the adverb is enough or else ) Adv -> W1 (if the adverb is rather or

103

Synchronic Functional Description or W3 Ν PrepP Sub clause

-> or -> -> or

quite ) (if the adverb is not enough, else, rather or quite) (if Ν is in the common case) (if Ν is in the genitive case)

W6 W2 W9 W10 Detached (if the NP contains only one sub clause and the NP represents a clause-rank constituent and the sub clause contains Foe and the clause contains no subject-proper)

As can be seen, the above rules have the affect of placing all determinatives in positions Wl-2 or W4-5, all premodifying adjectives in position W6, the head of the NP in W7, any postmodifying adjectives in W8, and any prepositional phrases or subordinate clauses in W8 or W9 respectively. Among the determinatives, any predeterminer is placed in W1, any determiner in W2, any cardinal numeral in W4 and any ordinal numeral in WS, producing sequences such as the following: (175)

All the four first prizes.

If there is a genitive-case noun instead of a determiner, this will take the latter's position in W2: (176)

All Jim's four first prizes.

It is true, of course, that orderings like (177) are also possible: (177)

All Jim's first four prizes.

However, these will be dealt with in 3.4.3 below. Most adjectives precede the head of the NP and are therefore assigned to W6, resulting in orderings such as: (178)

Four good prizes.

Common-case17 nouns are also placed in W6:

104

Constituent Order in Functional Grammar (179) Four Newport housewives.

Position W8 is used for the placement of obligatorily postmodifying adjectives such as galore, as in (180). (180)

Bottles galore.

It also receives adjectives modifying pronouns ending in -body, -one or -thing, as in the next examples: (181)

(a) (b) (c)

Somebody kind. Anyone new. Nothing big.

It sometimes happens that an adverb acts as a modifier within a NP. Examples are the adverbs rather anáquite in(182),íAe/i in(183),ande«