ON VOICE IN THE ENGLISH VERB

JANUA LINGUARUM

STUDIA M E M O R I A E

NICOLAI VAN WIJK DEDICATA

edenda curat

C. H. VAN SCHOONEVELD I N D I A N A UNIVERSITY

S E R I E S P R A C T I C A

LXIII

1966

MOUTON & CO. THE HAGUE · PARIS

ON VOICE IN THE ENGLISH VERB

by

JAN SVARTVIK

UNIVERSITY OF GÖTEBORG

© 1966

M O U T O N & CO. THE HAGUE · PARIS

© Copyright Mouton & Co., Publishers, The Hague, The Netherlands.

No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publishers.

Second Printing 1985

ISBN 90 2790669 6

Printed on acid free paper

Printing: Kupijai & Prochnow, Berlin. - Binding: Dieter Mikolai, Berlin. Printed in Germany

To Gunilla, Agneta, and Jesper

PREFACE

Voice in grammar is an aspect of English syntax which in recent years has attracted considerable interest in discussions of linguistic theory. This book is not primarily intended as a theoretical contribution (which of course does not exclude the possi-bility that it might be used to such ends); it is a corpus-based discussion of some grammatical categories that seem relevant to problems connected with voice in English. There are, at the present time, diverse views on the value of a corpus. Here, let it suffice to mention two reasons for using a corpus in the present inquiry: firstly, one of the aims of this monograph is to describe the use of voice, in particular the passive, in some varieties of present-day English; secondly, it is maintained that corpus-studies will help to provide descriptively more adequate grammars. It is interesting to note that in his latest book (which unfortunately arrived too late to be discussed in the main body of this study) Noam Chomsky comments: 'Perhaps the day will come when the kinds of data that we now can obtain in abundance will be insufficient to resolve deeper questions concerning the structure of language' (Aspects of the Theory of Syntax, Cambridge, Mass., 1965, 21). However far away the day is when corpus-data are needed, it seems likely that, to get nearer our goal of under-standing language, we need a multi-pronged approach.

Although presented as a Ph. D. thesis to Uppsala University, this book is essentially a British product: it was conceived in Durham, and the work took place chiefly in London where, for four years, I had the benefit of doing research in the stimulating atmosphere of University College. · Most of the research was carried out under the auspices of the Survey of English Usage, and I owe a very special debt to its Director, Professor Randolph Quirk — for suggesting voice as a suitable field of study, for helping to make available the necessary technical facilities, for setting an excellent (albeit inimitable) example with his zest for work, and, above all, for providing compelling stimulus and acute criticism at all stages of the work.

In spite of my long absence from Sweden, I have had the benefit of close connexion with many scholars there. I am particularly indebted to the following Uppsala professors: to Erik Tengstrand for guiding my first steps in English studies; to H. W. Donner for constant support and encouragement; and to Johannes Söderlind for supervising the final stages of my work. All of them have contributed to producing a more finished product by making valuable suggestions.

VIII PREFACE

The often very laborious tasks involved in this work were made lighter by the happy camaraderie among Survey colleagues, who also contributed in many other ways. In particular, 1 am obliged to Mr Henry Carvell, who lavishly and patiently provided computational and statistical knowledge. Among other London colleagues who have helped by criticism, discussion, informant responses, etc. are Mrs Judith Carvell, Mr Derek Davy, Mr Norman Fairclough, Mr Sidney Greenbaum, Mrs Joan Huddleston, and Mr Geoffrey N. Leech. My thanks are further due to Dr Sven Jacobson for stimulating linguistic arguments and to Miss Berit Hallberg, Miss Ann Helm, and Mr Bengt Odenstedt for assistance with proof-reading.

The research behind this book required both man and machine, and 1 have depended heavily on the computational expertise of Mr J. C. Gower, Rothamsted Experimental Station, Harpenden, Herts., and Mr A. J. Colin, University of London Institute of Computer Science.

Finally, 1 want to thank my wife who has always taken an active interest in my work, manifested by typing cards, providing consolation after computer breakdowns, and in innumerable other ways.

Göteborg, February, 1966.

CONTENTS

LIST OF FIGURES XIU

LIST OF TABLES XV

1 INTRODUCTION 1

1.1 Views on voice 1 1.2 Mode of analysis 3 1.3 Using a corpus 5 1.4 Material 6 1.5 Aims 8

2 THE VERBAL G R O U P 10

2.1 Simple and complex groups 10 2.2 Four types of complex groups 10 2.3 Type a 12

2.31 Criteria 12 2.32 Matrix 13 2.33 Classification 14 2.34 Concord valency classes 16

2.4 Type b 17 2.5 Type c 17

2.51 Criteria 18 2.52 Matrix 18 2.53 Classification 19

2.6 Type d 19 2.7 Lexical verbs 19

2.71 Phrasal and prepositional verbs 19 2.72 Prepositional verbs and prepositional phrases 20 2.73 Compounds 21 2.74 Verbal and nonverbal bases 22

3 THE CLAUSE 2 5

3.1 Clause elements 25 3.2 Exponent classes of subject and complement 26

X CONTENTS

3.21 Form-class 26 3.22 Gender 27 3.23 Finitude 28 3.24 Person 28 3.25 Number 28 3.26 Modification 28 3.27 Coordination 28

3.3 Exponent classes of adjunct 29 3.31 Form-class 29 3.32 Position 30 3.33 Agent 30

3.4 Internal clause relation 31 3.5 External clause relation 31 3.6 Central and peripheral clause types 33 3.7 Extensive and intensive clauses 33

3.71 Copula 34 3.72 Gender/number class 34 3.73 Gender-selection 34 3.74 Voice transformation potential 35 3.75 Exponence 35 3.76 Deletion 35

3.8 Major clause types 36 3.9 Values of clause element C 37

4 VOICE DATA PROCESSING 3 9

4.1 Procedure 39 4.11 Corpus 40 4.12 Data 41 4.13 Input and output 41 4.14 Programs 41 4.15 Computation 43 4.16 Results 43

4.2 Type of finite verbal group 43 4.3 Exponent classes of subject and agent (S and Ag elements) 49

4.31 Subject gender 50 4.32 Agent gender 51 4.33 Subject form-class 52 4.34 Agent form-class 52 4.35 Coordination 52 4.36 Length of passive subject and agent 52

4.4 Exponent classes of complement (C element) 55 4.41 Complement distribution 55

CONTENTS XI

4.42 Complement form-class 55 4.43 Complement gender 57

4.5 Exponent classes of adjunct (A element) 57 4.51 Adjunct frequency 58 4.52 Adjunct position 58 4.53 Adjunct form-class 59

4.6 Clause types 62 4.7 External clause relation 65 4.8 Internal clause relation 69 4.9 Summary of the results 69

4.91 Structure and frequency 70 4.92 Style 70 4.93 Voicu 71

5 PASSIVE CLAUSE CLASSIFICATION : CRITERIA 7 2

5.1 Numerical taxonomy 72 5.2 The Taxonomic Passive Corpus 73 5.3 The criteria 74 5.4 Criteria relating to the clause 78

5.41 External clause relation 78 5.42 External subject relation 81 5.43 Internal clause relation 82 5.44 Active transformation potential 83 5.45 Permutation and transmutation 88

5.5 Criteria relating to the passive verbal group 91 5.51 The auxiliary 91 5.52 The lexical verb 95

5.6 Criteria relating to the subject 97 5.7 Criteria relating to the complement 101 5.8 Criteria relating to the adjunct 101

5.81 Adjuncts with agentive function 102 5.82 Adjuncts with nonagentive functions 109

6 PASSIVE CLAUSE CLASSIFICATION : COMPUTATION A N D RESULTS ILL

6.1 Preparation of input data I l l 6.2 The classification program 113 6.3 Output analysis 115

6.31 Taxonomic analysis 115 6.32 Statistical assessment and linguistic interpretation 116

6.4 The diagnostic key 132 6.41 Classes α and β 132 6.42 Class β/γ 133 6.43 Class γ 134

XII CONTENTS

6.44 Class δ 134 6.45 Class ε 135 6.46 Class ζ 137

6.5 The passive scale and voice relation 138

7 THE USE OF THE PASSIVE 139

7.1 Five passive clause types 139 7.2 Three major passive classes 141

7.21 Agentive passives 141 7.22 Quasi-agentive passives 147 7.23 Nonagentive passives 148

7.3 Verbal group structure 150 7.4 The use of the passive voice in the texts 152

8 CONCLUSIONS 156

8.1 The concept o f 'The passive scale' 156 8.2 Transformational and serial voice relations 159 8.3 The description of the English passive 162

APPENDICES: Lexical verbs of the agentive clauses in the Major Passive Corpus 167

BIBLIOGRAPHY 194

INDEX 197

LIST O F TABLES

2:1 Verbal group types and some of their combinations 11 2:2 Matrix for Type a 13 2:3 Concord valency classes for some Type ο auxiliaries 17 2:4 Matrix for Type c 18 2:5 Diagnostic frames for simple, phrasal, and prepositional verbs . . . . 21

3:1 Gender classes 28 3:2 Major clause types (constituent structure) 36 3:3 Major clause types (transformational structure) 36

4:1 Distribution of verbal group types in the Voice Corpus 44 4:2 Voice distribution for the texts of the Voice Corpus 46 4:3 Distribution of verbal group types in the Voice Corpus (active only) and

in the Minor Passive Corpus 47 4:4 Contingency table for Types c/C and djD in the Voice Corpus . . . . 46 4:5 Occurrences of Type a/A auxiliaries 49 4:6 Voice and subject gender in the Voice Corpus 50 4:7a-c Voice and subject gender in Texts Ml , M2, J1 51 4:8 Subject gender in the Minor Passive Corpus 51 4:9 Number of words in subjects and agents of the Minor Passive Corpus . 56 4:10 Distribution of complements 56 4:11 Form-class of simple complements (C1) 57 4:12 Adjunct items, sets, and positions 59 4:13 Major clause types in the Voice Corpus 62 4:14 External clause relation 67 4:15 Internal clause relation 69

6:1-30 Contingency tables for categories in the Taxonomic Passive Corpus 118-131

7:1 The use of clause types in agentive passives 143 7:2 The use of verbal group types in agentive passives 150 7:3 The use of agentive passives in 28 texts 153 7:4 The use of passives in eight text sets 155

LIST OF FIGURES

1:1 The material 7

2:1 Word elements and bases 23

4:1 Procedure of the voice data processing experiment 39 4:2 Verbal group types in the Voice Corpus 45 4:3 Verbal group types: active (Voice Corpus) and passive (Minor Passive

Corpus) 48 4:4 Subject-lengths in Class γ 53 4:5a, b Subject- and agent-lengths in Class α-β 54-5 4:6 A set distribution 60 4:7 Positions of A elements 61 4:8 Form-class of A elements 63 4:9 Major clause types in the Voice Corpus 64 4:10 C and A elements in Class γ 66 4:11 External clause relation 68

5:1 Criteria relating to the clause 75 5:2 Criteria relating to the passive verbal group 76 5:3 Criteria relating to the subject 77 5:4 Criteria relating to the complement 78 5:5 Criteria relating to the adjunct 79

6:1 Specimen of data input (OTUs 1-4) 113 6:2 Specimen of output: top part of the full similarity matrix 114 6:3 Non-proportional, shaded matrix representing mean similarities be-

tween and within the groups 115 6:4 Diagnostic key for finite passive clauses 117

7:1 Frequencies of the five passive clause types in the agentive classes. . . 142 7:2 Frequencies of the agentive passive classes in eight text sets . . . . . 154

8:1 Diagrammatic representation of some voice relationships 163

1

I N T R O D U C T I O N

1.1 V I E W S O N V O I C E

The study of the grammatical category voice has enjoyed considerable popularity over the last few years.1 In fact, voice has probably received greater attention from linguists than ever before in the history of English scholarship. Its sudden appearance in the grammatical limelight can be attributed largely to the advent of transforma-tional grammatical theory, where the active-passive relation has been used as a prime illustration of the supremacy of the transformational model.2 In his grammar, Chomsky derives passive sentences from kernel active sentences: 'For every sentence JVPj— V— NP2 we can have a corresponding sentence NPi-is-\-Ven—by-\-NP1.'

'Thus every sentence of the language will either belong to the kernel or will be derived from the strings underlying one or more kernel sentences by a sequence of one or more transformations.'3 The idea of representing the English active-passive relation in terms of transformations was not, however, revolutionary. Jespersen, for instance, had spoken of the 'turning'4 and Poutsma of the 'conversion'3 of the verb form from one voice to another; but it was only in the case of transformational theory that the

1 In this study the term 'voice' will be used only with reference to the category 'grammatical voice' or 'diathesis' in the verb. 2 See, for example, Chomsky 1957, 1962 and Lees 1957. Bach calls the passive 'a prototype of a transformational relation' (1964, 62). 3 Chomsky 1957, 43, 45. In later rules (1962, 140) the requirement 'transitive verb' (K ( ) was added. However, in the most recent versions o f T G rules, many transformations, such as passive, negation, question, etc., have been partly or wholly replaced by phrase structure or 'base' rules, since such transformational rules were found to be too strong in 'expressive power'. This modification of the theory 'excludes in principle certain kinds of derivational pattern that were permitted by the earlier version o f transformational theory, but never actually found' (Chomsky 1964b, 61). See the reformulation by Lees (1964) below. 4 Jespersen 1924, 164: 'what was the object ... in the active sentence is made into the subject, and what was the subject in the active sentence is expressed ... by means of a prepositional group, in English with by ( formerly of)...' 'We may express this in a formula, using the letter S for subject, Ο for object, V for verb, a for active, ρ for passive, and C for "converted subject":

S V « Ο S ν » C Jack loves Jill = Jill is loved by Jack,

thus Jack: S" = Cr> Jill: O " =-- S". '

; Poutsma 1926-9, 2.2.107.

2 INTRODUCTION

use of transformation was extended, formalised and systematically incorporated into a unified grammatical framework.

Chomsky took actives rather than passives as his kernel sentences, since this method would lead to less complexity than deriving actives from passives (Chomsky 1957, 79-80). The same unidirectional transformation was also favoured by Lees because 'passives are less central than actives' (1957, 388). In a more recent article, how-ever, Lees finds that Chomsky's rule 'is not only very a-typical of grammatical transformations in general - it also fails to provide for the correct constituent-structure of the resulting passive sentences, even though it does correctly serve to derive [the] passive from an underlying active sentence.' An English passive sentence 'contains two components which do not appear in the underlying active sentence from which it is derived. First, there is the special verbal auxiliary consisting of the morpheme be and the participial suffix for the following verb base; ... Second, there is a special "agentive" adverbial-like prepositional phrase in by plus a nominal [expression].' His remedy is to 'permit in the expansion of the underlying constituent trees preceding all grammatical transformation rules the optional selection in any transitive-verb sentence of a special Agentive constituent. If not chosen, the deriva-tion leads to active sentences; if chosen, the derivation leads obligatorily to passive sentences.' Hence, the passive transformation rule is no longer optional but becomes obligatory: 'it serves to permute in all trees containing the Agentive formative the transitive-verb object nominal with whatever precedes it, and it substitutes the subject nominal in [j/c] for the nominal object of by in the Agentive phrase' (Lees 1964,29-30).

Kruisinga had earlier pointed out that while 'it is usual to consider the passive as a kind of secondary form of the verb, a derivative form dependent upon the "active"', 'this treatment, though supported by tradition and convenience, does not really permit us to state the facts completely or correctly' (Kruisinga 1927-31, 2.1.335). Nevertheless, his own classification is based on the active voice. W. S. Allen felt that 'a great deal of harm has been done by teaching the passive voice as if it were merely another way of expressing a sentence in the active voice. Students are asked to put such sentences as: John likes girls, Henry can read English and French, etc. into the fantastic forms of Girls are liked by John, English and French can be read by Henry, etc. We ought to stress the fact that the passive voice has an important and special place in the language; most sentences that are good in the active voice are just grotesque curiosities when put into the passive voice' (Allen 1959, 290).

McKerrow adopted a more radical attitude: 'If we were now starting for the first time to construct a grammar of modern English, without knowledge of or reference to the classics, it might never occur to us to postulate a passive voice at all. It seems to me that it is questionable whether in spoken English of to-day there is really any such thing, and though, as a matter of convenience, it may be well to retain it in our grammars, 1 doubt whether it ought to occupy quite so prominent a position as it sometimes does' (McKerrow 1922, 163).

Some of the diverse points of view advanced in these quotations may be paraphrased

INTRODUCTION 3

and summarized as two statements that are largely contradictory in regard to voice relationship:

(a) There is a relation between the active and the passive voice, and it is therefore economical to consider one voice in terms of the other. Passives, being less central than actives, are then best derived from actives. Subject to certain conditions, such as the verb being transitive, there is a passive sentence corresponding to every active sentence.

(b) There are indications that there is not a one-to-one relation between the active and the passive voice. There seem to be considerable restrictions on the use of passive sentences generated in this way, and considering the passive simply as a derivative of the active will not yield a good grammatical description. In fact, the only reason for keeping the category 'passive' is that it has come down to us as part of our classical grammatical heritage, and there is little or no place for the passive in a description of present-day English.

1.2 MODE OF ANALYSIS

This book will discuss voice in present-day English from different standpoints. Most attention will be given to the passive, and we shall try to view it in its own right by reversing the customary procedure of analysis and making the passive our point of departure.

This mode of analysis presupposes that we know what is meant by the term 'pas-sive voice' in English. Any attempt to define its boundaries by reference to previous work will soon reveal that there is no agreement among grammarians as to what constitutes an English passive.® Numerous reasons might be given for this lack of agreement. The name is certainly partly responsible. Grammarians do not generally claim that the subject of a passive construction must necessarily be the 'sufferer' of the action;7 yet there must be some such requirement present in the minds of those grammarians who preoccupy themselves so much with concepts of 'action' and 'occurrence' as opposed to 'state' when they are setting up a definition of the passive

* There is no exhaustive treatment of the passive in present-day English. Apart from the works already mentioned, discussions can be found for instance in Curme 1931, Erades 1950, 1958/9, Francis 1958, Fries 1940,1959, van der Gaaf 1928,1929,1930, Hatcher 1949,1956, Hendriksen 1948, A. A. Hill 1958, L. A. Hill 1964, Hockett 1958, Jespersen 1909-49,1937, Joos 1964, Kirchner 1936-7, 1951, Koumari 1956, Kruisinga 1927, Mihailovic 1963, Nida 1960, Owen 1914, Palmer 1965, Strang 1962, Svartengren 1948. - We have chosen a formal synchronic approach applicable only to present-day English, and we shall therefore not discuss here many works which, while dealing with the passive in English, have little or no bearing on the present material and method of treatment. The following list contains a selection of such works dealing with older periods of English: Akerlund 1914, Brose 1939, Curme 1913, Frary 1929, Fröhlich 1951, Green 1913, 1914, Jud-Schmid 1956, Klingebiel 1937, Kurtz 1931, Meier 1953, Mustanoja 1960, Söderlind 1951, 1958, Turner 1962, Visser 1941-56. 7 See, however, the controversy between Meyer-Lübke (1925, 1926) and Vossler (1925). Joos calls the passive subject the 'victim' of the action (1964, 98).

4 INTRODUCTION

voice: 'The forms of the verb conjugated with to be and the past participle of the verb when it does not denote a state resulting from an action.'8

The weakness of any such definition is not only that it makes agreement difficult to reach because of reference to semantic criteria, but also that it excludes, apparently quite arbitrarily, so many related constructions. The line of argument taken here is that syntactic relationships can or should be expected to be multidimensional rather than binary and that, in order to find and state this network of relations, it is best to cast the net wide.

In this study, 'passive' will be considered as a technical term, used in a very wide sense, for a formally defined construction. In the primary analysis, all the following sentences will be considered 'passives':

The house was built by experts. The house was built of wood * His bills are paid. His bills are paid regularly every month. His bills are paid, so he owes nothing now.10

The snow was piled high by the wind. The snow was piled high by the door.11

The village was {appeared, lay, looked, seemed) quite deserted. He felt thoroughly disappointed. The door remained locked,12

Whatever their differences in meaning, all these sentences have one formal feature in common: they all have as verbs combinations of be (or auxiliaries commutable with be) and a past participle. For the purpose of our discussion, this will be our simple working definition of the passive in English.13

8 Scheurweghs 1959, 416. The following are also representative of this view: The 'participles of transitive verbs can form a close group with verbs of little independent meaning to express an occurrence or an action. The most important verb giving rise to such a purely verbal group is to be' (Kruisinga 1927-31, 2.1.305). 'The subject signals either "that which undergoes the action" or "that to or for which the action is performed" whenever the Class 2 word [i.e. the verb] to which the subject is bound is the function word be (in its various forms) or get, with so-called past participle' (Fries 1959, 180). A 'verbal group consisting of one of the forms of to be plus the past participle of a transitive verb may denote an action undergone by the subject of the sentence', with the modification: 'in cases where "an action undergone by the subject" seems a somewhat forced definition, the passive may be said to express what "happens" to the subject' (Zandvoort 1960, 53). » Francis 1958, 336. 10 Jespersen 1909-49, 4.98ff. 11 Hill 1958, 323. la Zandvoort 1960, 49. 13 It may be objected that this is a definition of, say, the "be + ^erf-construction' rather than the 'passive'. The answer to this criticism is that, firstly, the name is too clumsy and, secondly, little harm can be done by extending pro tem the domain of an established term to cover not only central constructions but also those on the periphery. Disagreement on the meaning of grammatical terminology is largely due to the fact that terms like 'active' and 'passive' are often applied indiscrimi-nately to notional and formal categories alike, or are used loosely with reference to languages with different voice systems, or are used without clearly separating diachronic and synchronic approaches.

INTRODUCTION 5

Voice will be regarded as a grammatical system in the verbal group with two terms: active and passive. The active term and the passive term are in formal binary opposi-tion and will be studied both at the rank of the finite verbal group and at the rank of the finite verbal clause, with regard to their internal relations as well as to their external relations. The verbal group operates at place V in clause structure.14

This approach, starting the analysis with a certain formally defined structure and ending up with a statement of its values, makes it necessary to exclude a concurrent treatment of structures that are formally dissimilar but semantically similar to the passive, as defined here. We do not want to argue that the converse procedure, i.e. conducting the analysis from meaning to form, is impossible, but only that it is more difficult to control in the case of voice. What is essential, however, is to keep the semantic and nonsemantic approaches separate as far as possible.

Form and meaning may or may not coincide: 'Not only do structures usually signal several different meanings but, what is more important, there is probably in present-day English no structural meaning that is not signalled by a variety of structures' (Fries 1959, 203). Meaning may arise out of a pattern of distribution in the language under description so that 'the grammatical "meanings" are determined by their inter-relations in the systems set up for that language' (Firth 1957, 22). 'it is this distribu-tional characteristic which above all others allows the investigator to discover a morphemic class in the first place, and to know that such classes are relevant to the language in question, whereas an attempt to make class meaning a basic starting criterion for determining the classes is fatal to any structural analysis' (Pike 1954-60, 1.106).

1.3 USING A CORPUS

This investigation consists of a series of corpus-based studies. Some of them are quite independent of each other, but they have all been designed to shed light on problems of voice in English. There are several reasons for using a corpus. One follows directly from our aim to describe the use of the passive voice in some co-existing varieties of present-day English (see below). Another follows from the state-ment quoted above that some passives automatically derived by rule from actives are 'grotesque curiosities'. Passive sentences recorded from actual speech and writing

On the other hand, a wholesale rejection of traditional terminology is of no advantage. Traditional labels like 'active' and 'passive', 'subject' and 'object', are useful, provided each is defined in relation to the particular formal system employed by a particular language at a particular time (cf. Vogt 1950, 137; Buyssens 1950,41; Zandvoort 1961). This is not the point at which we should try to justify our somewhat unorthodox definition of 'passive'. Rather, its value should emerge from the discussion. 14 When it is convenient to make a distinction between the active and passive verbal groups, they will be said to operate at places V and W, respectively, in clause structure (see Chapter 3). For the use of 'system', 'structure', 'group', 'rank', and 'term', see Halliday 1961. Halliday, however, recognizes a system of voice in his grammar only in the verbal group, not at clause rank (1964,14f.).

6 INTRODUCTION

are assumed to be, on the whole, normal and natural uses,15 and a corpus-based study should provide basic information about how and to what extent 'corpus-passives' differ from 'rule-generated passives', and from actives. This raises a theoretical problem: the value of 'data' for the construction of a grammar. If, in general, 'corpus-passives' turn out not to be in a one-to-one transformational relationship to actives, it will seriously weaken arguments in favour of deriving passives from actives, at least if our grammar is to be economical and relevant to the use of the language.

1.4 MATERIAL

The material has either been taken from the files of the Survey of English Usage or collected, as far as possible, on the principles of the Survey's method of text com-pilation.18 The material is intended to represent some coexisting varieties of educated present-day English, spoken and written. In this case, 'present-day English' means English produced between 1950 and 1964; these are of course only arbitrary dates of limitation. For practical purposes, it has been necessary to restrict the material to British English. (There is however little reason to expect that an extension of the material to cover the other major Standards would give significantly different results from those arrived at for British English.)

The material consists of a number of 'texts', or samples of spoken and written English which, with a few exceptions, are continuous stretches. In all, 28 texts varying in length and totalling some 323,000 words have been analysed. (Text lengths are stated in number of words, but are not intended to be anything more than approx-imations. Hyphenated items count as single words.) The texts of 5,000 words each were taken from the files of the Survey of English Usage, whereas the others were collected separately.

17 texts of 5,000 words = 85,000 words 6 texts of 15,000 words = 90,000 words 1 text of 28,000 words = 28,000 words 4 texts of 30,000 words = 120,000 words

Total: 28 texts = 323,000 words

The texts are ordered in groups denoted by capital letters (A-M) and numbered (Bl, 15 This is not, however, a generally held assumption. Bach, for example, maintains that 'real discourse - especially when spoken in a natural context - is always full of fits and starts and in-congruities (This form is found in Homer, don't we?). In other words, we cannot identify the set of grammatical sentence with the set of actually occurring sentence' (1964, 90). It is not possible to discuss here the justification for equating, in principle, 'occurring' and 'grammatical' without entering into the whole complex problem of grammaticality. Suffice it to say that the material so far analysed in the Survey of English Usage seems to support the line taken here. For an investigation into linguistic acceptability, see Quirk & Svartvik 1966. 16 See Quirk 1960, Crystal & Quirk 1964, Godfrey 1965.

INTRODUCTION 7

Present-day English

Spoken Conversation (surreptious recording)

Discussions (impromptu unscripted speech)

Written

Television advertising Delivery in spoken form Radio news

-TEXT A

-TEXT Β

-TEXT C

-TEXT D

Printed

Informative

Press

News Sports reports Editorials

Science

Popular Arts

Learned Science

Arts

Imaginative

Fig. 1:1. T h e material.

Plays

Novels

-TEXT Ε -TEXT F -TEXT G

-TEXT Η

-TEXT I

-TEXT J

-TEXT Κ

-TEXT L

-TEXT Μ

B2, B3, etc.) where there is more than one text in a group. Figures after stops (e.g. B1.6; F.26.5; 1.9; Ml. 16, etc.) denote internal file references for Texts A, B, C, E, dates (day and month) for Texts D, F, G, and page references elsewhere. The relation between the text groups is shown in Figure 1:1. The imbalance between the spoken and written material (40,000 as against 283,000 words) does not reflect the importance attached to either variety but rather the difficulty of compiling spontaneously-spoken material. First year of publication, if different from the year of the edition or im-pression used, is given in square brackets.

A . CONVERSATION (surreptitious recording): Text A: Conversation between two university teachers. 1963. 5,000 words.

B . DISCUSSIONS (impromptu, unscripted speech recorded from discussions on B.B.C. programmes): Text B1: Any Questions? 30.1.1959. 5,000 words. Text B2: Brains Trust. 14.11.1958. 5,000 words. Text B3: Whafs the Idea? 16.6.1961. 5,000 words. Text B4: Any Questions? 4.3.1958. 5,000 words. Text B5: Any Questions? 28.10.1958 and

Any Questions? 16.1.1959. Together 15,000 words. C . TELEVISION ADVERTISING:

Text C: 617 advertisements broadcast over the Independent Television Network for the first time 1.12.1960-31.5.1961. 255 different products were advertised in 7, 15, 30, 45, 60, or 75 sec. advertisements, totalling nearly 35,000 words and representing about 4 hours 30 minutes of continuous broadcasting. When 'repeats' are subtracted, the size of the text is over 28,000 words.

D . RADIO NEWS: Text D: B.B.C. Home Service news at 1 p.m. 18-20.3.1964. 5,000 words.

8 INTRODUCTION

Ε . PRESS NEWS: Text El : The Times. 9.3.1964. 5,000 words. Text E2: The Daily Express. 10.4.1964. 5,000 words.

F . SPORTS REPORTS: Text F: The Guardian. 23-31.5.1960. 30,000 words.

G . EDITORIALS: Text G: The Times, not including 'the Fourth Leader'. 1-17.2.1960. 30,000 words.

H . POPULAR SCIENCE: Text H: Robert Thomson, The Psychology of Thinking (= Pelican Books A 453) (1959), pp. 11-57. 15,000 words.

I . POPULAR ARTS: Text I: Simeon Potter, Language in the Modern World (= Pelican Books A 470) (1960), pp. 9-52. 15,000 words.

J . LEARNED SCIENCE: Text J1: J. Z. Young, The Life of Vertebrates (London, 1962 [1950]), pp. 83-106. 5,000 words. Text J2: G. E. Bacon, Neutron Diffraction (London, 1955), pp. 162-185. 5,000 words. Text J3: Η. Ν. V. Temperley, Changes of State, A Mathematical-Physical Assessment (London, 1956), pp. 22-34. 5,000 words. TextJ4: G.H.Williams, Homolytic Aromatic Substitution (London, 1960), pp. 27-39. 5,000 words. Text J5: 14 articles and letters to the editor on biochemistry, biology, genetics, metallurgy, meteorology, physics, physiology, etc. published in Nature, a weekly journal of science, vol. 186, 4721-2. 23 and 30.4.1960. 15,000 words.

K . LEARNED ARTS: Text K: P. F. Strawson, Individuals (London, 1961 [1959]), pp. 142-154. 5,000 words.

L . PLAYS: Text LI: Noel Coward, South Sea Bubble, a comedy (London, 1956), pp. 1-90. 15,000 words. Text L2: Graham Greene, The Complaisant Lover, a comedy (London, 1959), pp. 1-77. 15,000 words.

M . NOVELS: Text Ml: Rosamond Lehmann, The Echoing Grove (= Penguin Books 1262) (1958 [1953]), pp. 16-30. 5,000 words. Text M2: Malcolm Bradbury, Eating People is Wrong (London, 1959), pp. 90-107. 5,000 words. Text M3: Auberon Waugh, The Foxglove Saga (London, 1960), pp. 197-212. 5,000 words. Text M4: David Beaty, The Proving Flight (= Penguin Books 1318) (1958 [1956]), pp. 45-59. 5,000 words. Text M5: Angus Wilson, Anglo-Saxon Attitudes (= Penguin Books 1311) (1958 [1956]), pp. 11-89. 30,000 words. Text M6: Michael Innes, The Long Farewell (London, 1958), pp. 9-110. 30,000 words.

1.5 AIMS

The principal aims of this book are to study the values of the two terms within the voice system of present-day English, with particular emphasis on the passive and its

INTRODUCTION 9

affinities with the active; to set up a classification of finite passive clauses; to examine the uses and frequencies of occurrence of the passive in some coexisting varieties of contemporary English; and to consider ways of accounting for the production of passive sentences. The discussion will be divided into four parts:

(a) A general discussion in Chapters 2 and 3 of the finite verbal group and the structure of the rank next above in which it operates, i.e. the clause. Naturally, both types of unit will be considered particularly from the point of view of voice. We must exclude from our discussion nonfinite verbal groups and nonfinite clauses.

(b) A study in Chapter 4 of both active and passive clauses in a small sample. This sample (Corpus I, or the Voice Corpus) consisted of two novel extracts (Ml and M2) and one scientific text (Jl), together totalling some 15,000 words. In this corpus all finite verb clauses, i.e. both active and passive, were collected except equative with be as copula. (For a discussion of major clause types, see Section 3.8, pp. 36 f.) This sample was chosen to give a general picture of voice in English and to provide a background for the subsequent more detailed analysis of the passive.

Chapter 4 also includes a specific study of only passive clauses in a larger sample (Corpus II, or the Minor Passive Corpus), consisting of agentive17 finite passive clauses from eleven texts totalling some 55,000 words: Texts Bl, B2, and B3 (spoken); Texts Jl, J2, J3, and J4 (learned scientific); Texts Ml, M2, M3, and M4 (novels). This sample was chosen to provide more information on 'central' passive construc-tions than could be ascertained from Corpus I.

(c) An attempt in Chapters 5 and 6 to set up a classification of various types of passive clause, making use of multiple criteria and numerical taxonomy. A small sample (Corpus III, or the Taxonomic Passive Corpus), consisting of 128 finite passive clauses, was used for this experiment.

(d) A typological and quantitative examination in Chapter 7 of passive clauses in some varieties of present-day English. The material (Corpus IV, or the Major Passive Corpus) consisted of all finite passive clauses in 28 texts, i.e. the comprehensive collection of material as described above, totalling some 323,000 words.

11 'Agentive', in contradistinction to 'nonagentive', will be used for finite passive verb clauses which either have an agent (and hence are 'agentful') or are 'agentless' but may have extension with an agent. These terms are discussed in Sections 3.33, pp. 30f., and 4.11, p. 40. Agentive passives are considered central to the passive construction.

2

T H E V E R B A L G R O U P

2.1 SIMPLE AND COMPLEX GROUPS

The English finite verbal group can be divided into two major categories: simple and complex. The simple verbal group (Type 0) consists of a lexical verb without an auxiliary (for example help-helps-helped, run-runs-ran, hit-hits-hit, etc.). The complex verbal group, on the other hand, consists of an auxiliary plus some form of the lexical verb.1

2.2 FOUR TYPES OF COMPLEX GROUPS

It is convenient to divide the complex finite verbal group into four types: a, b, c, and d.

Type a ('modal'): λ + jto/ V, e.g. He may examine. Type b ('perfective'): χ + Ved, e.g. He has examined. Type c ('continuous'): χ + Ving, e.g. He is examining. Type d('passive'): χ 4- Ved, e.g. He is examined.

Telescoped into one another, combinations of a, b, c, and d can form complexes of two, three, or four types. With the system used here the combinations occur in alphabetical order, so that a cannot follow b, nor b follow c: ab, ac, acd, bed, cd, etc.

In Table 2 : 1 , where combinations with Type d (passive) are contrasted with those without d (active), the auxiliaries cited are may for Type a, have for Type b, and be for Types c and d. Present forms, alone or as first elements in combinations, are indicated by lower case; past tense forms by upper case. When no present/past distinction is relevant, lower case letters will be used. Imperative and subjunctive are not included in this classification. The table does not list all possible combinations, for instance groups such as may be about to be getting fed, but, as we shall see later, all such strings can be ultimately related to the basic types. Some of the combinations in the table are of course extremely rare. Highly complex verbal groups such as the following abed-structure have been heard in conversation, but there are no instances recorded in the 1 Among earlier treatments of the verbal group to which I am particularly indebted are Hill 1958, Olsson 1961, Quirk 1962, Strang 1962, and Twaddell 1963.

THE VERBAL GROUP 1 1

present material (see Section 7.3, pp. 150ff.): 'Bynow the new cook will have been being introduced to her duties for several weeks.'

TABLE 2:1

Verbal group types and some of their combinations

ACTIVE PASSIVE

0 eats d is eaten 0 ate D was eaten

a may eat ad may be eaten A might eat AD might be eaten

b has eaten bd has been eaten Β had eaten BD had been eaten

c is eating cd is being eaten C was eating CD was being eaten

ab may have eaten abd may have been eaten AB might have eaten A BD might have been eaten

ac may be eating acd may be being eaten AC might be eating ACD might be being eaten

be has been eating bed has been being eaten BC had been eating BCD had been being eaten

abc may have been eating abed may have been being eaten ABC might have been eating A BCD might have been being eaten

In describing linguistic categories we can usually point to a central core which is surrounded, as it were, by more or less peripheral subcategories. The exact nature of the relationship between the various categories and subcategories is often very intricate. This is certainly true of the complex verbal group in English. The mode of analysis we use here is naturally prompted by the purpose for which it is intended, i.e. a study of voice categories. The boundaries of the different types are not clear-cut, and it will be the object of this book to treat one of them, Type d, in some detail. It does not purport to deal exhaustively with Types a, b, and c.

It will, however, be necessary to try and define more closely the area occupied by the auxiliaries. 'Auxiliaries' are here considered a special class of the general gram-matical category 'verb', the other class being 'lexical' or 'full' verbs.

In order to arrive at a fairly comprehensive classification of Types a and c, a number of criteria have been used, which are set out at the head of columns in a matrix table.2

Some specimen verbs are entered in a column on the left of the table, and when a particular verb satisfies a particular criterion, this is indicated in the cells by ' + ' , otherwise by '—'. 'Inapplicable' is denoted by '/'.

2 For the use of matrices in linguistics, see for example Pike 1962, 1963.

12 THE VERBAL GROUP

2.3 TYPE a

2.31 CritericP

CRITERION 1. Γο-less infinitive, for example, He must do it, as opposed to He has to do it.

CRITERION 2 . Invariance for person and number in the present. Test-frame: I/youjhe χ V now. Can but not be going to, for example, will satisfy this criterion: I am/you are/he is going to do it.

CRITERION 3. Both the present and the past forms can be used, with modal distinc-tion, in a present sequence. A test-frame like I think I χ do it now at once will accomo-date for example either can or could, but only have/has to and not had to.

CRITERION 4. Only the position as first element is possible in verbal group structure, i.e. the verb cannot occur in the frame: He may χ V. This criterion is satisfied for instance by will but not by appear to.

CRITERION 5. Postverbal enclitic negation is possible. Do but not seem to, for example, will fit the frame: You x'nt V.

CRITERION 6. No Jo-periphrasis is possible in negations and questions, i.e. the verb cannot occur in the frame You do not χ V. This criterion is satisfied for example by will, but not by want to.

CRITERION 7. This cannot serve as a substitute for (to) V, for example, He appeared to drink beer and *He appeared this, as opposed to He attempted to drink beer and He attempted this.

CRITERION 8. Voice transformation is possible of the form: Ν χ χ (to) V N2 «-> Ν2 x (to) be Ved. Both those finite verb clauses which satisfy Criterion 7, and those which do not, may take voice transformation in the clause, but the transforms are different and must be distinguished here. The two types may be illustrated by the following pair of sentences, both symbolised as Ν χ χ to V Ν 2:

(i) He appeared to drink beer. (ii) He attempted to drink beer.

Sentence (i), unlike (ii), may admit of a passive transform with χ as Type a auxiliary in the passive verbal group and N2 as subject of the passive clause (JV2 χ to be Ved):

Beer appeared to be drunk (by everyone in the pub). *Beer attempted to be drunk (by everyone in the pub).4

CRITERION 9. No ίΑαί-clause is possible. This criterion excludes, for example, hope to: 3 The slots in the frames where the verbs are tested are indicated by x. * This should be distinguished from the passive transform This was attempted. Sentence (ii) does not satisfy Criterion 7, and, in consequence takes passive transformation with χ as the lexical verb in the passive verbal group and the factive substitute as subjects:

*This was appeared. This was attempted.

THE VERBAL GROUP 13

He hopes to do it and He hopes that he can do it; it includes, for example, seem to: He seems to do it but *He seems that he can do it.5

CRITERION 10. No Ν (to) V construction is possible. This criterion excludes, for example, like, which admits both He likes to do it and He likes you to do it; it includes, for example, attempt which admits only He attempts to do it but not *He attempts you to do it.

CRITERION 11. No Ving construction possible. This excludes, for example, like: He likes to do it and He likes doing it; it includes, for example, manage: He manages to do it but *He manages doing it.

2.32 Matrix

TABLE 2:2

Matrix for Type a

CLASS VERB

1 2 3 4 5 CRITERIA

6 7 8 9 10 11

el can + + + + + + + + + + + al do6 + — — + + + + / + + + <z3 had better'' + + + + ± + + + + + + a4 ought to8

- + + + + + + + + + + a5 used to8

— + — + + + + + + + + «6 be to — — — + + + + + + + + al be going to — — — — + + + + + + + c8 have to' — — — — + — + + + + + «9 seem to + + + + + alO start to ± + + —

a l l manage to + + + al2 learn to + + el3 expect to + a!4 like to — — — — — — — — + —

The fourteen verbs listed in the matrix (Table 2:2) are different in that no two are fully alike with respect to the selected eleven criteria. The poles of the scale are re-presented by can, which satisfies all the criteria, and by expect to and like to, which satisfy only one. It should be emphasized that the criteria were chosen fairly arbitrarily 5 Seem with proleptic it, as in It seems that he's leaving, is a different construction. • Do does not cooccur with the auxiliary be in the verbal group except in emphatic imperatives (such as Do be quiet!) and, consequently, does not cooccur in the lie-passive. Cf. however: Did he get caught? The second do in Did he do it? is a lexical verb. 7 The negative He hadn't better do it is perhaps less common than He had better not do it. 8 Ought to, but less certainly use to, with <fo-periphrasis (Did he ought to/use to do it?) are probably marginal to Standard English. • Phonological and prosodic features may provide additional criteria: both have and used 'have lost the juncture before to and consequently have undergone assimilation of the final consonant' (Hill 1958, 198): ['haefta, 'hsesta, 'justs]. At least in British English, both the enclitic and periphrastic negatives occur with have to.

14 THE VERBAL GROUP

from those that were felt to be relevant; a different choice of criteria would naturally alter the classification. The decisive factor here in choosing a classification must be the particular application for which it is intended. For a study of voice the obvious division would seem to be one that enables us to make comparative statements about the active and passive terms in the voice system. Hence we make our point of division the one on the scale which separates those verbs which may operate as χ in the active as well as in the passive verbal group from those which do not operate in this way. In the table this point occurs between seem to and start to. The verbs in Classes al-9 will be called 'auxiliaries', and those in Classes alO-14 'lexical verbs'. All auxiliaries, as the term is used here, should then in principle permit voice transformation, which may be stated in the following formula:

There are, however, considerable constraints in the use of auxiliaries which affect the possibilities of voice transformation, for instance the alternation between shall and will as future tense auxiliaries depending on the person of the subject. Such constraints are not considered here.

Class alO (start, begin etc.) has ± for Criterion 8, since these verbs may take double passive transformation, for example

Because of this and the fact that they do not satisfy Criterion 11, they are classed as lexical verbs. This division will allow us to make comparative statements with Type c.

Type a auxiliaries can be subclassed in various ways. The most useful division for our purpose seems to be one between those auxiliaries which can occupy the position of first element only and those auxiliaries which cannot do so (as in Criterion 4), since this has a bearing on the verbal group structure. Classes al-6, represented by the verbs can, do, had better, ought to, used to, and be to, are termed 'closed class auxi-liaries' (a'); and Classes al-9, represented by be going to, have to, seem to, 'open class auxiliaries' (a").

Type a is formed either by closed class auxiliaries plus infinitive with or without to, or by open class auxiliaries plus ίο-infinitive:

(Type a) V Ν! χ (Type b) Ved

(Type c) Ving

(Type ad) be N2*-+ N2X (Type bd) been Ved (by NT).

(Type cd) being

They began to use a new method. -» A new method began to be used. -> A new method was begun to be used.(!)

2.33 Classification

THE VERBAL GROUP 15

closed class auxiliary (a'): may eat, or, open class auxiliary (a"): has to eat, or by a combination, closed class + open class auxiliaries (a' + a"): may have to eat, or open class + open class auxiliaries (a" + a"): seems to be about to eat.

The combinatory types occur because the open class auxiliaries can themselves form substructured combinations with closed class auxiliaries, for example, may have been having to eat ([a' + b + c +] a" + V), or with members of its own class, for example, may seem to be about to eat ([a' + a" +] a" + V).w

The following list of auxiliary classes for Type a, set up on the basis of the criteria in the matrix, is not intended to be exhaustive. There are, besides collocational11

restrictions, considerable idiolectal, regional, and stylistic variations in the use of the verbal group, and no general agreement can be expected on this listing.

CLOSED CLASS AUXILIARIES OPEN CLASS AUXILIARIES

a' 1 can, could a"l be (un)able to may, might12 be about to shall, should be due to will, would be going to ΊΙ, V13 be sure to must be (uri)likely to dare11 have got to need14 a" 8 have to

a'2 do a" 9 appear to a'2 had/'d better/best happen to a' 4 ought to seem to a' 5 used to tend to a' 6 be to

10 'The large number of these auxiliaries and the multiplicity of ways in which they may be combined permit a very large repertory of verb structures to be built in English. From a historical point of view, it is interesting to note that many of them are of quite recent development. This seems to be an area of English grammar where change has been taking place rapidly. Indeed, change seems to be still going on, and may continue to for some time to come' (Francis 1958, 259-60). 11 'Collocation' is used to mean 'the habitual association of a word in a language with other particular words in sentences' (Robins 1964, 67). " Mayn't is, according to Hill, 'very rarely used in the United States, though it is more frequent in England.' Mightn't 'is commoner' (1958, 196). 13 The auxiliaries ΊΙ and V, which are themselves enclitic, cannot, however, take the enclitic negation: *He'dn't do it. 14 Dare and need are typically marginal between open and closed class auxiliaries, as shown by overlapping distribution and a profusion of alternative forms. Cf. for example He may dare do it and *He may need do it. In an experiment involving the selection of negative forms with dare and need, the following combinations were chosen (see Quirk & Svartvik 1966, 90ff.):

to 49 1 aux + neg + dare ^ 13 I 6 2

16 THE VERBAL GROUP

Lexical verbs, Classes a 10-14, may be arranged in valency classes. (Superscripts denote valency as indicated by ' in Columns 8, 9, 10, 11 in Table 2:2.)

Criterion 2, the formal variability criterion for person and number, can be further refined if we adopt the idea of valency classes.15 These classes are set up on the basis of concord between the subject and the finite verb. As concord-bearers there are three classes of subjects:

(x) First person singular: I. (.p) Second person singular and the entire plural:you, cats, thoughts, ... (z) Third person singular: Tom, cat, thought,...

The subject classes are labelled differently from Olsson's in order to avoid confusion with the terms used in the present treatment for describing the verbal group types. Distinction is also made between present and past forms, indicated by lower and upper case respectively. No account is taken here of the modal contrasts, which to some extent cut across the temporal contrasts.

The valency for a particular verb form may then be stated as VALENCY \ Y for were, VALENCY 1* for does, VALENCY 2X+Z for was, VALENCY 2 f o r do, VALENCY for can, VALENCY 3X+Y+Z for could, VALENCY 6χ+ν+ζ+x+r+z for must, and so on. Concord valency classes could be extended to all verbs, but it will be sufficient for the present purpose if we restrict ourselves to Type a auxiliaries. On the basis of Criterion 3 we may further divide auxiliaries into

(i) NON-TENSE-MARKED : auxiliaries with only one form, which may occur in both present and past tense sequence ( = Concord Valency Class I).

V-e care, fail, manage, serve, ... ja8,-9 decide, learn, resolve, ... F"8·11 begin, cease, continue, start, ... y-β, »,-io expect, promise, want, wish, ... Κ8-10·11 like, love, ...

2.34 Concord valency classes

+ 0 0

+ to 61

Need appears to be more common in negative constructions and need to in positive. Voice trans-formation with dare is doubtful. " See Olsson 1961, 29 ff.

THE VERBAL GROUP 17

(ii) SEMI-TENSE-MARKED: auxiliaries with two forms of which the past form is not restricted to past tense sequence ( = Concord Valency Class II).

(iii) FULLY TENSE-MARKED: auxiliaries with one or more forms which do not share the tense sequence properties of (i) and (ii) ( = Concord Valency Classes III-V).

This classification is illustrated in Table 2:3.

TABLE 2:3

Concord valency classes for some Type a auxiliaries

VERB

VALENCY VALENCY

AUXIL-

IARY

PRESENT PAST

CLASS CLASS X y ζ X Y Ζ

NO

N-T

EN

SE-

MA

RK

ED

I

fix+ll+z+X+Y+Z a'l a'l a'l a' 3 fl'4

must dare need had better ought to

SEM

I-T

EN

SE-

MA

RK

ED

II 3 x+v+z/l 3x+r+z

a'l a'l a'l a'l a'l

can may shall will ' / /

could might should would 'd

III 3x+r+z a' 5 used to

w VI Ζ Β

IV 2χ*«/ I'// jx +r+z

a'2 <f 8 a'9 a" 9 a'9 a" 9

do have to dare to need to seem to appear to

does has, 's to dares to needs to seems to appears to

did had, 'd to dared to needed to seemed to appeared to

u. V ΐ χ / l v ß z ( 2 x + v ) l l

2x+z/\y pjr+y+z) a'6, 7 am, Ίη | are, 're

(aren't)1' is, 's was were 1 was (were)17

2.4 TYPE b

There is one central auxiliary in Type b, viz. have. Other auxiliaries which can occupy the place χ in the χ + Ved combination and commute with have might also be classed as Type b auxiliaries. However, since there is no clear borderline between Types b and d in many such cases, we shall be interested, for the present purpose, in considering them in relation to Type d as a whole (see, for example, Section 5.44, pp. 83 ff.).

2.5 TYPE c

Type c has be as its central auxiliary, but there are other possible candidates, which form a continuum similar to that described for Type a (see Section 2.3).

" In negative-interrogative clauses, aren't I occurs besides am I not. 17 Valency 3 J r + K + z refers to the subjunctive forms: Ifljyoujhe were ill, ...

18 THE VERBAL GROUP

2.51 Criteria

CRITERION 1. Postverbal enclitic negation is possible, for example, He isn't doing his homework as opposed to *He startsnt doing his homework.

CRITERION 2. This cannot serve as a substitute for Ving, for example, He went on singing and * He went on this, as opposed to He began singing and He began this.

CRITERION 3. Voice transformation is possible: Nl χ Ving Ν 2«-» Ν 2 χ being Ved (by N^. He is doing the job and The job is being done, but He avoids doing the job and * The job avoids being done.

CRITERION 4. No transformation is possible of the form N1 χ Ving ->• Ving is what N1x \ He keeps walking but * Walking is what he keeps, as opposed to He likes walking and Walking is what he likes.

CRITERION 5. No (to) V construction is possible, i.e. no change from Type c to Type a: He keeps walking and *He keeps to walk, as opposed to He likes walking and He likes to walk.

2.52 Matrix

TABLE 2:4

Matrix for Type c

CRITERIA CLASS VERBS

CRITERIA

1 2 3 4 5

e l be18 + H - + 4

<•2 keep — - i - + - L

<•3 begin - - - ! • -

c 4 avoid — — — — - f .

c5 like _ _ _ —

Five verbs are listed in the matrix (Table 2:4). Reapplying the argument put forth in Section 2.32 for the classification of Type a, we call Classes cl-2 Type c 'auxiliaries' and Classes c3-5 'lexical verbs' on the grounds that the former, unlike the latter, may operate as χ in an active as well as in a corresponding passive verbal group. The reason for dividing Type a auxiliaries into two classes, according to whether they can occupy first place or not in verbal group structure, does not obtain in the case of Type c. We may, however, recognize the difference between Classes 1 and 2 and call them 'closed' and 'open' respectively, although there is only one formal criterion which separates them. Class c3 verbs are taken as lexical, since they do not so readily take voice transformation: He continues playing the record -* (?) The record continues being played; a Type a construction (continues to be played) is probably more likely.

' 6 The relation between He is to walk and He is walking is different from that between He likes to walk and He likes walking.

THE VERBAL GROUP 19

2.53 Classification

CLOSED CLASS AUXILIARY

c'l be OPEN CLASS AUXILIARIES

c"2 keep (on (and on)) go on (and on)

Lexical verbs, Classes c3-5, may be arranged in valency classes. (Superscripts denote valency as indicated by ' in Columns 4-5 in Table 2:4.)

Type d is the main subject of this study and, for the purpose of our discussion, the net will therefore be cast wide to include instances which are more or less removed from the construction centre (see Sections 1.2, 2.4). There are, for example, no specific requirements of subject value for membership of Type d: at this point of the analysis no account is taken of whether the grammatical subject is identical or not with the logical subject (see Section 5.45, pp. 89 f.).

Lexical verbs may consist of one or more words. One-word verbs may be, for example, simple words (dress); or complex words with bound stem + derivational suffix (stupefy), with prefix + bound stem (intervene), with free stem + derivational suffix (broaden) and with prefix + free stem which may be a simple word (untie, overflow, outgrow, offset) or a complex word (reconvert).20

Verbs of two or more words consist of the verb proper (the 'nucleus') and adverbs or adjectives or prepositions, with which they may form close syntactic units. We distinguish the following three classes (example-centres are given in italics):

PHRASAL VERBS (Vph): An experiment to test this theory was carried out on January 6, 1959. (J5.273)

" Stop has different meanings when followed by to V and Ving: cf. He stopped fighting I to fight and He started fighting!to fight. " Francis 1958, 206ff.

V * avoid, remember, ... V b begin, continue, stop, ...10

V *'-b like, enjoy, prefer, try, ...

2.6 TYPE d

2.7 LEXICAL VERBS

2.71 Phrasal and prepositional verbs

2 0 THE VERBAL GROUP

But in the last six months attention has to some extent been drawn away from these central themes to peripheral worries. (G.10.2) Some raw shelving had been run up at one end to stack additional books - ...

(M6.83) PREPOSITIONAL VERBS ( V p ) :

Indeed, with his own salary and his wife's private income, they were really very comfortably provided for. (M5.24) 'But in this country there are singularly few murders which are conceived of as deliberately incident to a robbery.' (M6.38) If a woman reads history for a degree, and then rejects a career for marriage, her mind is not to be thought of as thrown away unless we count the family an arena of no importance. (G.2.2)

PHRASAL-PREPOSITIONAL VERBS (Vph-p): He had in fact recovered for Packford some valuable documents which had been made off with by a rather specialised sort of burglar. (M6.10)

There are two problems of delimitation here. One is between phrasal and preposi-tional verbs, where two criteria have been used.21

Firstly, there is a difference in stress and intonation patterns. The particle com-ponent of the phrasal verb normally (i.e. in unmarked use) bears full stress (and, if at tone unit boundary, nuclear tone), whereas the preposition with the nonphrasal verb is normally unstressed and does not carry a nuclear tone:

(Vph) He was taken Ίη. (The particle has a falling nuclear tone.) (Vp) He was laughed -at. (The preposition has the 'tail' of the nuclear tone.22)

Secondly, the particle component of a phrasal verb has twofold positional privilege in that it often takes either prenominal or postnominal position, whereas it is restricted to postpronominal position. The diagnostic frames in Table 2:5 show that the two classes of verbs, represented by call on and call up, structure differently.

Verb nuclei and adjectives may have close collocation and behave syntactically like phrasal verbs, for example He pushed the door open. Adjectives in such superficially similar constructions, as in He considered the man good, are however treated as com-plements (C elements in clause structure; see Section 3.1, pp. 25 f.) for reasons that are clear in Table 2:5.

2.72 Prepositional verbs and prepositional phrases

The other problem occurs in active clauses where prepositional verbs plus nominale (Vp + N) have to be distinguished from simple verbs plus prepositional phrases as

21 This treatment draws heavily on Mitchell 1958. Cf. also Kennedy 1920, and Strang 1962, 156ff. aa For the intonation system used here and elsewhere in the book, see Crystal & Quirk 1964.

THE VERBAL GROUP 2 1

TABLE 2:5

Diagnostic frames for simple, phrasal, and prepositional verbs

PREPOSITIONAL VERB PHRASAL VERB

(+) They call on the man. (+ ) They call early on the man. (+) They call on him. (—) *They call the man on. (—) *They call him on. (+) The man they call on. ( + ) The man on whom they call.

(+ ) They call up the man. (—) *They call early up the man. ( - ) »They call up him.23

(+) They call the man up. (+) They call him up. (+) The man they call up. (—) *The man up whom they call.

CLOSE COLLOCATION BETWEEN VERB AND ADJECTIVE

SIMPLE VERB WITH ADJECTIVE AS COMPLEMENT

(+) They push open the door. (—) *They push open it. (+ ) They push the door open. (+) They push it open. (+) The door they push open. (—) 'The door open which they push. (—) *They push (that) the door is open.

(—) *They consider good the man.34

(—) *They consider good him. (+) They consider the man good. (+) They consider him good. (+) The man they consider good. (—) "The man good whom they consider. (+) They consider (that) the man is good.

adjuncts (V + pN; see Section 3.1). There is a scale of closeness and openness whose poles may be illustrated by the following examples:

(Vp + N) She sent for the coat. -* The coat was sent for. ( V + pN) She came with the coat. *The coat was come with.

An attempt has been made elsewhere to classify such 'prepositional strings' compre-hensively by the use of numerical taxonomy (see Carvell & Svartvik 1966). For the present special purpose the chief criterion will be the voice transformation test, which will help to distinguish prepositional verbs from verbs plus prepositional phrases.25

2.73 Compounds

The past participle ( V e d ) in the passive verbal group presents special problems, as compared with the active, since it may be morphologically isolated by compounding:

23 May possibly occur with emphatic (contrastive) stress on him, as in We rang up him, not his brother. " Possible only with a heavy object, as in They would consider brave any man who could volunteer for such a task. " The acceptability of some such passive transforms varies considerably depending for example on the amount of linguistic context. In a poll conducted in order to register informant reactions to this kind of potential active-passive transformations, it was found, for instance, that the minimal sentence The girl was turned to (from the active They turned to the girl) and the same sequence differently contextualized The Prime Minister was turned to for help by the people suffering from depression in the north-eastern industrial areas produced very different results.

22 THE VERBAL GROUP

unwanted, well-known, new-laid, etc. have no bases {to) *unwant, *well-know, *new-lay (cf. Marchand 1960, 54).| The distinction between compounds and noncompounds is not an absolute one, but one of degree. We shall consider some aspects which are relevant to our problem.

Compounds often have single stress on the first element, or double stress: [ ' - - ] in airborne, awe-struck, [ ' - - ] or [ '- ' -] in well-judged, etc. However, stress alone cannot be used as a criterion, particularly since it is connected with position in the clause, for example this well-known ['- -, '-'-] author but usually this author is well-known [ ' - ' - - ' - ] ·

It will for our purpose be convenient to let the term 'compound' embrace not only word elements (for example all-admired) but also the prefix un- (as in undressed). The ««-prefix has two values (see Jespersen 1909-49, 6.464 ff.):

(i) REVERSATIVE un- which joins a simple word verb so as to form a complex word verb, whose past participle form may occur in both the active and passive paradigms (Types b and d): He has/is undressed; also undo, unpack, unbind, unbutton, uncover, unfasten, etc.

(ii) NEGATIVE un- which joins an adjective, a noun, or a participle so as to form a compound word (for example unsympathetic, unease, unwanted). This prefix is a bound negative making the compound the antonym of its base (see Lyons 1963,61ff.):

This result may have been unexpected after Holland's fine golf on previous days, ... (F.26.5) -* This result may not have been expected ... -* [they] may not have expected this result ...

The difference between the two types may be diagnosed by commutation tests within the verbal group, for instance between Types d and b (see Section 2.1) :2e

Reversative prefix: She is/has undressed. Negative prefix: She is/*has unexpected.

2.74 Verbal and nonverbal bases

Like the prefix un-, -ed has two values: it is noi only a verbal inflexion but also a suffix. As a suffix it 'is now added without restriction to any sb. from which it is desired to form an adj. with the sense "possessing, provided with, characterized by" (something); e.g. in toothed, booted, wooded, moneyed, cultured, diseased, jaundiced,

" The difference between the two un- prefixes may be neutralized. The minimal structure She was undressed may have the following expansion and transform: The chorus girl was undressed every night before the show by the wardrobe mistress. -> The wardrobe mistress undressed the chorus girl every night before the show. But whether the sentence She was/appeared undressed every night on the stage should be analysed as She appeared denuded!stripped... or as She appeared not dressed... is irrelevant, since the origin of unVedis unknown here and the distinction, in any event, syntactically neutralized.

THE VERBAL GROUP 23

etc., and in parasynthetic derivatives, as dark-eyed, seven-hilled, leather-aproned, etc.' 'In mod. Eng., and even in ME., the form affords no means of distinguishing between the genuine examples of this suffix and those ppl. adjs. in -ED1 which are ultimately f. sbs. through unrecorded vbs.' (OED, s. v. -ed).

Three relevant ««-types will be analysed as follows (X stands for the first element): (i) A complex word consisting of reversative prefix + verbal base + inflexion

[(X + V) ed], for example undressed. (ii) A compound word consisting of negative prefix + verbal base + inflexion

[X(V + ed)], for example unexpected. (iii) A compound word consisting of negative prefix + nominal base + suffix

[X (N + ed)], for example unskilled. The distinction between verbal and nonverbal bases87 may be diagnosed by commu-

tation within the verbal group, in compounds with the first compound element removed:

(i) He was undressed: He had undressed: He was dressed: He had dressed (ii) He was unexpected: *He had unexpected: He was expected: He had expected (iii) He was unskilled: *He had unskilled: He was skilled: *He had skilled

Verbal base

(Ved) Noncompound

_—(V) simple —{Vpk) phrasal —(Vp) prepositional —(Vph-p) phrasal-prepositional

(X- Ved) Compound

—Ad] -η — —

—Adv --N --N - + Ved —un — -self --all -

Nonverbal base

{Ned) Noncompound

taken looked up looked upon got away with

widespread well-judged man-made unfrequented self-appointed all-admired

wooded

—Adj -i — — red-haired -Adv - down-hearted -N - Ned hunchbacked

(X-Ned) Compound —Sum— + Ned one-eyed —un — unskilled -seif - self-willed

Fig. 2:1. Word elements and bases.

37 'Verbal base' naturally refers to present-day English. The fact that some of the Ved forms have had a verbal base at some time in the history of the language is immaterial for a synchronic descrip-tion. Ashamed, for instance, which has no verbal base today, is the past participle of a verb to ashame (OE äsceamian), last recorded in 1826 by OED. In the case of alternative participial forms, both are considered as having the same base, although there are often different restrictions on their use, such as semantic (He was born in 1950 / air-borne), collocational (poverty-stricken / thunder-struck), or paradigmatic (he is drunk / a drunken sailor).

24 THE VERBAL GROUP

By definition (see Section 1.2), the type illustrated by unskilled cannot enter into the verbal group, but we shall eventually want to consider its relation to the voice system as a whole (see Sections 5.52, p. 95 and 6.46, pp. 137 f.). Although ««-compounds with verbal bases can have only restricted active-passive clause transformation potential (see Section 5.44, p. 86), their constituent structure bears, nevertheless, great resemblance to that of noncompounds with verbal bases. This is illustrated for instance by their ability to take agents:

All his investigations were totally ungoverned by the slightest awareness of the actual substance of the stuff he dealt with to such triumphant effect. (M6.11)

Figure 2 :1 exemplifies and displays diagrammatically the morphology of some com-pounds and noncompounds with both verbal and nonverbal bases.

3

THE CLAUSE

3.1 CLAUSE ELEMENTS

The finite verbal group operates at the place in structure called V in the rank next above the group, viz. the finite verb construction, or, as it will be convenient to refer to it here, the finite verb clause. At the highest level of abstraction the English clause elements are V, S, C, and A.1

The V/W element (the verbal group) has been dealt with in Chapter 2. In the fol-lowing sections we shall first discuss the other clause elements and then the major clause types which are formed from them.

S (the subject) in English may be defined as that element in the clause with which the finite verb is in grammatical concord (overt or covert).2 Since the finite verb is so often morphologically invariable (see Section 2.34, pp. 16f.), a more useful working definition, which holds good in most cases, is one based on position: S is that nominal clause element which in affirmative clauses normally occurs before and in questions immediately after the first element of the finite verbal group (cf. Strang 1962, 71). 5 may be zero, particularly in sequentially related clauses (He said good-bye and left), and it is normally so with imperatives [Go away !).3

C (the complement) is that element besides S in the clause which is typically mani-fested by the nominal group.

A (the adjunct) stands, broadly speaking, for the rest. C may be said to be that nominal element which is usually essential to complete the

basic clause structure,4 i.e. C is generally obligatory whereas A is optional. C and A cover more indeterminate territories than S and V. They may be diagrammatically

1 For the term 'rank' see Halliday 1961, 250ff. In respect of this unit there is unique solidarity between clause and group, and V is therefore used here both at the rank of the group for denoting the lexical verb and at the rank of the clause for the place in structure at which the verbal group operates. When a voice distinction at clause rank is needed, it will be convenient to use V for active and W for passive. 2 Unless stated otherwise, 'subject' will be taken to mean 'grammatical subject'. 3 The fact that imperatives may have S realized (You go away!), albeit with restricted exponence, makes a good case for considering imperatives as members of the finite verbal group. Another reason for doing so is the fact that they take do-periphrasis (Don't go away!). * See Quirk 1962, 190; and Firbas, who considers complements 'essential amplifications of the respective verbal meanings' (1959, 46).

26 THE CLAUSE

represented as two points, A being lower than C, on a scale of essentiality or indispen-sability of clause structure elements:

V s c A

most

least

essential clause elements

Within the last two elements there are, however, subscales which may be so extensive as to form functional overlap of the C and A elements. Prepositional phrases, for example, which at the primary degree of analytical delicacy are all taken as A ele-ments, may be like C elements in some respect, such as having voice transformation potential (see Section 2.72, pp. 20f.).8

Dispensability seems to be correlated with the number of possible element items: there can be only one S and one V element but either one or two C elements, whilst in theory at least, there can be an unlimited number of A elements. In the material here analysed, the maximal number of A items was four (see Sections 4.51, p. 58, and 5.82, p. 109).

3.2 EXPONENT CLASSES OF SUBJECT AND COMPLEMENT

The S and C elements can be treated together, although typically they have different exponence. It is convenient to use the symbol Ν for any nominal and nominalized element which operates at places S and C in clause structure, and also for the nominal part of prepositional phrases which operate at place A.6

3.21 Form-class

We distinguish between nominal groups and nominalizations. The latter comprise constituents other than nominal groups but which operate as such, for example finite and nonfinite verb clauses.

Nominal group heads have three subclasses: nouns, pronouns, and names. Of these only nouns regularly take determiners. Pronouns constitute a closed class, some of whose members have oblique forms other than genitive (he/him, they/them, etc.). Names may take the provided there is modification as well {the inimitable Jeeves, the London I know); they do not admit of number change: if singular, they cannot be pluralized (Agneta but *Agnetas); or, if plural, cannot be singularized (The Andes but *an Ande). The in The Andes or The Thames is considered part of the name since it 6 For the concept of 'delicacy', see Halliday 1961, 272f. Subscales 'constitute small continua within themselves* (Bolinger 1961a, 44). β Denoted ρ Ν, see Section 3.3. The classification here of the clause elements S, C, and A is, naturally, less refined than that of V. See further Sections 5.6-8 (pp. 97 ff.).

THE CLAUSE 27

'is not in determinative contrast with any entity, with *a Thames and *some Thames, for instance. It follows that the is neither the definite article nor a generic determina-tive' (S0rensen 1958, 150; cf. 133 ff.)·

Finite verb clauses are, for instance, ίΑαί-clauses, and conditional and relative clauses, which in the following examples operate as C:

It was only men, Emma considered, who believed in things; women recognised that being a woman was way of life enough. (M2.94) 'He asked me if I could look after him till he came out ...' (Ml. 18) Faintly she frowned, contemplating the area of her labours, seeing what should have been, what could be done. (Ml. 17)

Nonfinite verb clauses include, for example,

(1) To-full infinitive (to V): To pass among the wooden stalls of the market, ... was a pleasant sensation.

(M2.91) (2) Γο-less infinitive (V):

they watched the apparition wave up and down ... (Ml.23)

(3) Present participle (Ving): Ί found him cutting his hair in my drawing-room the other afternoon before the departmental tea-party.' (M2.104)

(4) Past participle (Ved): Ί don't want other people's humanity tied round my neck\ (M2.107)

There are other possible exponents of S and, particularly, of C, for example direct speech and adjectives:

'You live alone?' said Madeleine, rather awkwardly. (Ml. 19) /anything that makes the : army atträctive # . a is de/sirable and a good thing #

(B1.7)

3.22 Gender

Gender distinctions are as far as possible set up on the basis of formal criteria, i.e. patterning with substitution classes of personal and relative pronouns (see Table 3: l).7

In addition to this, the category 'animate* also embraces nouns which collocate as S with verbs like see or like (rat, snake, etc.). For the present purpose it has usually been considered sufficient to recognize a two-term gender system with animate (Nan) and inanimate (Nin) as terms. The former then includes the subterms 'personal' and 'common'.

7 Cf. Francis 1958, 250f„ Fries 1959, 120ff., and Strang 1962, 95.

28 THE CLAUSE

TABLE 3:1

Gender classes

ANIMATE INANIMATE

PERSONAL COMMON

-J you he she he/she

jthey _

+ who

Γ he jit she/it [who ] he/she/it \whichj

L i t / they J

['' 1 + which L they\

boy, girl, doctor, mathematicians, ...

horse, cow, baby, committee, ...

book, life, grammars, mathematics, ...

3.23 Finitude

Ν may be definite or indefinite. Nominal groups are taken as definite when having the definite article, a genitive, or a possessive or demonstrative pronoun (the/John s/myl this girl). Names and referential pronouns are definite. The latter include for example personal he but not interrogative who (/ wonder who's coming?), proleptic it (It was learnt that ...), or nonreferential you and one:

It was a showpiece of the unendurably modern - when you saw the modern like that, it looked so dated that you couldn't believe it. (M2.97)

3.24 Person

Person comprises the traditional categories first, second, and third person as exem-plified in Section 2.34.

3.25 Number

Number is divided into singular and plural according to pronoun patterning (see Table 3:1). In the material where this category was used, no 'common number' (The committe is/are ...) occurred.

3.26 Modification

Modification subsumes premodification (as in the little girl) and postmodification (the man over there).

3.27 Coordination

Simple head (for example you) is contrasted with coordinate (for example you and I).

THE CLAUSE 29

3.3 EXPONENT CLASSES OF ADJUNCT

Exponent classes of adjunct (clause element A) are multifarious, and 'there is a continuous spectrum of classes from those functioning as adjuncts to verbs to a host of other kinds of adjuncts - the verb-adjuncts are at one end of the spectrum and should be seen in relation to it as a whole' (Strang 1962, 161; see also Sections 4.5, pp. 57ff., and 5.8, pp. 101 flf.). Our chief interest here lies naturally in that end of the spectrum where we find the verb-adjuncts, and hence A elements which are taken to have no such function are not considered in this study.

Consequently, A does not embrace adjuncts of S, C, and A elements, or sequential relators and sentence adjuncts, such as accordingly, actually, however, in fact, in general, of course, perhaps, probably. Sentence adjuncts are often characterized by having (i) actual or possible individual intonation units, (ii) freedom of movement within the clause, particularly if occurring with individual intonation units, or (iii) no generic substitution class meaning8 as do most, but by no means all, adjuncts. (This is indicated by the possibility of using them in response to questions with where, when, how, etc., see Section 5.8.)

A, as the term is used here, does however embrace other types of adjuncts which are more or less marginal to the V element, for example finite and nonfinite verb clauses. Adverbial particles (up, down, etc.) are treated as part of the lexical verb (see Section 2.71, pp. 19f.). Adjuncts have been classified according to form-class and position in clause structure.

3.31 Form-class

A elements may be divided into the following form-classes:

(1) Adverbs, which include, for example, (la) central invariables ('He arrived now, at once, here', etc.), (lb) central variables ('He works fastest'He runs slower), (lc) de-adjectival ('He runs slowly'), (Id) other derivatives, compounds, etc. (adrift, homewards, nowhere, publicity-wise, etc.). (See Strang 1962, 162 f.)

(2) Prepositional phrases:

professors nibbling cheese straws peered over the tops of them to see what was happening. (M2.107)

(3) Finite verb clauses:

The sucker is bounded at the edges by a series of lips, which besides being sensory serve also to make a tight attachment when the lamprey sucks. (J1.88)

(4) Nonfinite verb clauses (with or without conjunction), where the verb may be

» Cf. Francis 1958, 287f., Strang 1962, 164f.

3 0 THE CLAUSE

(4a) Present participle (Ving):

She followed his course with a dreamy look, remarking that he was very obedient. (Ml. 19)

(4b) Past participle (Ved):

'Tell me, has it arrived?' she asked, well pleased with the success of her little gesture. (M3.212)

(4c) 7o-full infinitive (to V):

Ί heard you were coming with us to give a lecture on your findings to the New York Canford Institute.' (M4.57)

(5) Verbless constructions (with or without conjunction):

it is unique among vertebrates in that the follicles have no ducts; when ripe they rupture into the coelom, ... (J 1.96)

3.32 Position

There are three major places of A in relation to other clause elements: front-, mid-, and end-position. The definitions given here apply to the order S V(C), the most common order of clause elements in statements (for a detailed discussion of adjunct positions, see Jacobson 1964).

(F) Front-position: pre-S, i.e. A before any other clause element. after a while she stopped dead close to the church door ... (Ml.27)

(M) Mid-position: post-5 but pre-V, i.e. A between the subject and the lexical verb. No distinction is made here between different mid-positions.

Ί have lived where you never could have survived;' (M2.97) 'Keats, Keats, Keats', he said spitefully. Ί don't care if I never see another keat again.' (M2.101)

(E) End-position: post- V, i.e. after the lexical verb. If there is a C element present, it makes no difference if A occurs before or after the complement.

The dog was no longer sitting in the porch. (Ml.23)

3.33 Agent

The different functions of A (such as indicating place, manner, time, etc.) are not dealt with here, with one notable exception: the agent. 'Agent' is defined as that element in passive clause structure whose nominal part may operate as S in an active clause transform. The agent (denoted by Ag in contradistinction to Ad for nonagen-tive adjuncts) is usually a prepositional phrase with by, but there are, as we shall see,

THE CLAUSE 31

other prepositions with similar functions. This definition of 'agent' is based on potential syntactic function alone and involves no semantic requirement such as agent being 'actor' in contrast to subject being 'patient'.8

3.4 INTERNAL CLAUSE RELATION

Internal clause relation, i. e. linear ordering of clause elements, has been restricted to the cardinal elements SVC, since A position is recorded in the description of the A element (see Section 3.32). Among the major categories are the following (C elements are italicized in the illustrative examples):

(1) SV order.

(la) SV(C), which is the normal, unmarked order:

Raleigh smokes Socrates is wise. (K.143) As the Prime Minister left, M.P.s gave him a massive demonstration of personal loyalty. (E2.5)

(lb) CSV:

'Holliday knows what a glass of beer is, but champagne he does not understand in the slightest degree.' (E2.53)

(2) VS order.

(2a) VS(C), where the finite verb precedes S:

'Can't you fetch your gardenerΤ (Μ 1.26)

(2b) CVS:

' You mean he's mad', said Treece. (M2.104)

(2c) Discontinuous C:

She smokes, thought Madeleine, like a chimney. (Ml.21)

3.5 EXTERNAL CLAUSE RELATION

External clause relation comprises three main types: syntactically bound, sequentially related, and free clauses. These types, each with several subtypes, should be seen only as convenient points of division on a scale of external clause relationships. • The term 'agent* has been used in many different ways (see, for example, Green 1913, 1914; Bloomfield 1933, 371; Hendriksen 1948); Koumari (1956, 51) goes as far as to deny its justification as a category. The present use of 'agent' as a technical term, corresponds most closely to Jespersen's 'converted subject' (1909-49, 3.317ff.) and Sweet's 'inverted subject' (1955, 1.113). The agent will be further discussed in Section 5.81 (pp. 102ff.).

32 THE CLAUSE

(1) Syntactically bound clauses ('a' are immobile, 'b' are so with zero, and 'c' are usually mobile within the sentence).

(la) Relative clauses:

Model (c) was immediately discounted by the neutron diffraction data since there was no trace of the superlattice lines which would be required by the enlarged unit-cell. (J2.175)

(lb) Conjunctional that- or zero ί/ιαί-clauses:

Still less do I say that grammatical substantives in general are such expressions. (K.148)

(lc) Other conjunctional clauses (if, before, when, etc.):

When they reached cruising altitude, five minutes later, all the stars were already there to meet them. (M4.49)

(2) Sequentially related clauses.

(2a) Fixed relator {and, but, etc.):

'Your soul rests easy, but nothing's solved .' (M2.105)

(2b) Fixed relator + elliptic subject:

He half-rose and gave an expansive gesture with his hand, ... (M2.96)

(2c) Fixed relator - f - elliptic subject + elliptic auxiliary:

but everybody had done something to the garden and made a mess of //.(Ml. 17)

(2d) Mobile relator (however, moreover, etc.):

'Are there no men in this village? Can't you fetch your gardener?' 'He's seventy-three,' said Madeleine. ''Besides, he's gone to the football match.

I think everybody has.' (Ml.26)

(2e) Elliptic subject:

/after all :mathematics . s :is thinking# . /is ldgic# /is reasoning# (B2.7)

(2f) Tag questions:

'He's a sporting dog, isn't he?' (M1.24) (3) Free clauses or, strictly, 'more free than the previous types', including clauses which

(3a) have a clause as C element:

he wished that he could tempt her to say the same thing about him. (M2.101)

THE CLAUSE 33

(3b) have direct speech as C element:

Again she said: 'We'd better go away.' (Ml.24)

(3c) are followed by a sequentially related clause:

nothing was resolved and there were no firm rocks to settle on. (M2.91)

(3d) are 'fully free', i.e. other than 3a-c:10

What does one do with dustbins to make them look interesting? (M2.97)

3.6 CENTRAL AND PERIPHERAL CLAUSE TYPES

Having surveyed the individual clause elements, we shall now consider the major clause types which are formed from them. In dealing with the clause, we adopt the same approach as for the verbal group, i.e. we first set up central types, to which peripheral types can be related. For this reason we leave out prepositional phrases, and indeed A elements generally, because they constitute a scale, the points of which should ultimately be related to the central clause types.

A major dichotomous division is made between EXTENSIVE and INTENSIVE clauses.11

Intensive clauses (Type INT for active, int for passive) are those clauses between two of whose nominal elements there is a relation of coreference. In extensive clauses (Type EXT for active, ext for passive) no such SYSTEMIC coreference relation obtains. Both types may, however, have EXPONENTIAL coreference, in which case the intensive type displays three, and the extensive two, coreferent clause elements. Exponential coreference occurs particularly with (i) reciprocal pronouns, (ii) reflexive pronouns, and (iii) nominal groups with possessive premodification:

(i) Their eyes, like two advance posts of opposing armies, crossed each other ... (M4.48)

(ii) He pulled himself up with the little groan of protest allowed to the middle-aged and successful. (M4.48) He considered himself a good player.

(iii) He took his leave. 19 This point does not, however, represent the end of the scale of external clause relation. Other features, such as finitude in a clause element implying grammatical reference, for instance anaphoric the (cf. Smith 1964, 51), are also relevant, and will be considered in the taxonomic experiment in Chapter 5 (see Criteria S 51-2, pp. 99f., and Ag85-6,p. 107). The lack of analytical delicacy in the voice data processing discussed in Chapter 4, as compared with the taxonomic experiment, reflects the natural progression of this enquiry as well as computational limitations and a desire for a comprehensive initial approach. 11 The terms 'extensive' and 'intensive', but not the definitions given here, derive from Μ. A, K. Halliday. For 'exponence', see Halliday 1961, 268ff.

3.7 EXTENSIVE AND INTENSIVE CLAUSES

34 THE CLAUSE

The distinction between systemic and exponential will simplify the description at various points. The following seven subsections will illustrate the difference between extensive and intensive clauses.

3.71 Copula

Systemic coreference between two Ns may be demonstrated by the fact that they can be connected by a copula (a closed list of verbs with the lexically unmarked be as its central exponent, see Section 5.51, pp. 90ff.).

Type INT: He considered her a friend: She was a friend. Type EXT: He gave her an apple: *She was an apple.18

3.72 Gender/number class

Consequently, the Ns in Type IN Τ must belong to the same gender/number class, whereas there is no such constraint for Type EXT.13

Type INT: [He is a teenager. They are teenagers.

N&n/sg + N^an/sgl N-ycmjpl + N2an/pl J

He likes the car. Νχαη/sg Njn/sg

• + ·

cars. • + •

N2in/pl • + · the girl. • + • N^an/sg

They like girls. Νχαη/ρΙ N^anjpl

Type EXT:

Deviations from this will be considered as metaphors, for example: The computer is his best friend. N^n/sg + N^flnjsg. There is however an exponential exception to this criterion: it does not apply to

'divisible' Ns, i.e. nouns which do not require agreement across the copula (cf. Stockwell & Schachter 1962).

John's problem is headaches. The flies are a nuisance here. John and Mary are a couple /*a child/.

3.73 Gender-selection

In question transforms Type EXT clauses have what with Ν in and who with Nan, whereas in Type IN Τ clauses the wA-form is not gender-selective in case of the latter coreferent Ν: 1! It is structural ambiguity between Types EXT and INT which make possible such linguistic jokes as * You are a cab, Sir Γ in reply to the order: 'Call me a cab Γ 1S an — animate, in = inanimate, sg = singular, pi = plural. Braces indicate that any com-bination of pairs of sequences in the braces is possible (cf. Bach 1964,17-8). Subscripts are used for labelling elements in clause structure: Nl VNt->N, (V, etc. Superscripts denote number of element items: C' = two Cs, A*· = four As, etc.

THE CLAUSE 35

Type EXT: He gave her a book. J Who/m/ did he give a book /to/?] [What did he give her? J

Type INT: He considered her iWho/m/ did he consider a pretty girl? a pretty girl. [What did he consider her?

In addition, Type INT often takes question-adverb here:

He considered Bleak House a great novel.

iWhat did he consider Bleak House? How did he consider Bleak Hornel

3.74 Voice transformation potential

Types EXT and INT may have different potential in regard to number of transforms: He gave her a book.

Type EXT:

Type INT:

(N, V N2 N3) (Ν1 VN3pN2) (N2 WN3) (Nz W Ν2)

W ρ N2) (NI VN2 N3)

WNa)

He gave a book to her. She was given a book. A book was given her. A book was given to her. He considered her a pretty girl. She was considered a pretty girl.

3.75 Exponence

Type INT, furthermore, takes a range of exponents different from EXT. (In the N1 V Ν2 Ν3 structures below, Ν is used here for nominal groups and nominalizations as well as other complements.)

Noun) He considered her a friend. Adj) He considered her kind. Num) He made it three. V) He saw her leave. to V) He asked her to leave. to be Adj) He considered her to be kind. Ving) He saw her leaving. Ved) He found the paper neatly folded.

3.76 Deletion

Type IN Τ seems never to have N2 deletable without structural change: Type EXT: He gave her a book: He gave a book. Type INT: He considered her a nuisance: *He considered a nuisance.

(Some deletions do not come altogether naturally and may best be described as ellipses. Even so, the criterion seems valid since ellipsis is not possible with the Type INT example above.)

36 THE CLAUSE

3.8 MAJOR CLAUSE TYPES

Returning to the clause types, we may, in addition to Types EXT/ext and INT/int, distinguish clauses according to the number of cardinal structural elements other than SV (i.e. in declarative clauses normally postverbal C elements): none (0), one (C1), or two (C2). Tables 3:2-3 show the major clause types according to these two parameters. Table 3:2 represents the major clause types from the point of view of constituent structure. The same clause types may also be stated in terms of their voice relationships. In Table 3:3 the S and C labels have been replaced by Nx, N2, etc., in order to show the systemic voice transformation potential of the clauses. Transformation is however closely linked with exponence, and the potential shown in the table occurs only with certain Ν exponents. (Systemic coreference is indicated by ΓΤ, active verbal group by V, passive by W. Agents, which are usually optional clause elements, are given in round brdckets.)

TABLE 3:2

Major clause types (constituent structure)

TYPE ACTIVE VOICE PASSIVE VOICE

TYPE CODE STRUCTURE CODE STRUCTURE

EXTENSIVE

EXT-0 S V ext-0 S W (Ag) EXTENSIVE EXT-C SVC ext-C SWC (Ag) EXTENSIVE

EXT-CC i K Q C , ext-C SWC (Ag)

INTENSIVE INT-C s~v~c int-C SWC (Ag) INTENSIVE

INT-CC S V cTc, int-C SWC (Ag)

TABLE 3:3

Major clause types (transformational structure)

TYPE ACTIVE VOICE PASSIVE VOICE

TYPE CODE STRUCTURE CODE STRUCTURE

EXTENSIVE

EXT-0 Ν V

EXTENSIVE EXT-C ΛΤ, V Nt <—• ext-0 ΛΤ, W(p N,) EXTENSIVE

EXT-CC Νί V N% N, ext-C' ext-C"

Ν, W Nt (ρ Ν,) Nt W N3 (p Nr)

INTENSIVE INT-C ΛΓχ V Ν,

INTENSIVE INT-CC Nr V Ν*Ν* int-C Nt W Nt (pNJ

THE CLAUSE 37

(.EXT-0) (.EXT-C) (.EXT-CC) (.INT-C) (INT-CC) {ext-0) (ext-C') (ext-C") (,int-C)

Active intransitive : Active transitive : Active ditransitive :14

Active equative : Active factitive : Passive transitive : Passive ditransitive' : Passive ditransitive" : Passive factitive :

He departed. He liked her. He gave her a book. He was a tall man. He considered her a beauty. She was liked {by him). A book was given her (by him). She was given a book (by him).15

She was considered a beauty (by him).

As the names imply, the clause labels have been chosen with a view to indicating transformational rather than constituent structure, taking actives as bases.

3.9 VALUES OF CLAUSE ELEMENT C

Having established the highest degree of clause rank abstraction, we may go on to assign different values to C.

In extensive clauses all C elements are 0 (object). If there is more than one O, they are differentiated as O* (indirect object) and Od (direct object), the former being the Ο which generally admits a transform with a prepositional phrase (and reversal of the elements):

They gave him a book. -* They gave a book to him.

Oi is typically animate and the order of elements in the clause Ot + Od- Prepositional phrases as transforms of indirect objects in active (They gave a book to him) or passive clauses (A book was given to him) will not, however, be considered indirect objects. 'To call such expressions as to the boy an "indirect object" in the sentence the man gave the money to the boy, leads to confusion. The expression to the boy does express the same meaning as that of the indirect object, but this meaning is signalled by the function word to, not by the formal arrangement which constitutes the structure, "indirect object"'. The subject 'in the sentence the boy was given the money also expresses the same meaning as that of the indirect object, but we rightly call it "sub-ject", not "indirect object'" (Fries 1959, 185).

In intensive clauses, the latter of two systemically coreferent nominal elements will be called Ρ (predicative; cf. Jespersen 1933, 321. 'Predicate', on the other hand, is used for the entire predication). The former coreferent element is S in Types INT-C and int-C; in Type INT-CC it will be called 0/S. In this last clause type, C1 functions

14 'Ditransitive', and 'equative' (or 'copulative') are terms used in tagmemics (see, for example, Longacre 1964). For 'factitive', see Sledd 1959. 16 The two passive ditransitive transforms may be distinguished as 'primary' (ext-C) and 'secondary' (ext-C"). These terms are adapted from 'primary' and 'secondary conversion' (see Poutsma 1926-9, 2.2.124ff.).

38 THE CLAUSE

both as 0 of the previous V and S of C2, if this is a nonfinite verb clause (J saw him come)', verbless constructions ( / found her unhappy) may be considered as truncated verbful constructions (/ found her to be unhappy, etc.) which have equative relation between the nominal elements.

Element C has been defined above as that nominal element which is essential to complete the basic clause structure. The further distinction which is made here between Ρ and Ο is, again, one between two poles of a scale: Ρ elements are generally more essential than Ο elements. 'Copulas', which will be used for V in equative intensive clauses, 'constitute a point in the structure of language at which the CD [communicative dynamism] of the verb is at its weakest, and its need of a semantic amplification at its greatest. They seem to set up one end of a gamut, one end at which we find verbs with the lowest possible amount of CD. The other end of the gamut would be made up by verbs performing the function of rheme proper' (Firbas 1959, 46).

4

VOICE DATA PROCESSING

4.1 PROCEDURE

The processing of voice categories comprised the following steps (see also Figure 4:1):

Text Ml 1 Class α Minor CORPUS Text M2 | Voice Corpus Class β • Passive

Text J l J Class γ Corpus

DATA Linguistic categories abstracted from 2106 clauses in the corpus.

Φ INPUT Data categories encoded into tape code characters in Columns 1-7 and

lexical verbs in Column 8.

Φ Program 1 Program 2 Program 3 and Program 4 (inventory) (two-way correlation) (three-way correlations)

PROGRAMS Φ Φ Program 5 Program 6

(tabulation) (alphabetization)

Φ

OUTPUT Tape code characters listed or tabulated for Columns 1-7 and lexical verbs given in alphabetical order for Column 8.

Φ RESULTS Interpretation of the output stated in terms of linguistic categories.

Fig. 4:1. Procedure of the voice data processing experiment.

(a) Selecting the corpus, (b) Abstracting the linguistic data from the corpus, (c) Preparing the computer input from the data, (d) Writing the programs, (e) Processing by computer, (f) Converting the output into linguistic categories, (g) Analysing and presenting the results.

The input to the computer was not 'raw' linguistic data but carefully selected categories which had been abstracted from the corpus. This experiment might be described as 'linguistically sophisticated' but 'computationally trivial'.

40 VOICE DATA PROCESSING

4.11 Corpus

The analysis covered two sets of material described in Sections 1.4-5 (pp. 6 if.), Corpus I (the Voice Corpus) and Corpus II (the Minor Passive Corpus), and the data derived from them were processed by using the same computer programs though in some cases with different significances attributed to the characters.

The Voice Corpus consisted of three texts, two novels (Ml and M2) and one scientific text (Jl). This corpus was chosen to give a general picture of some voice categories and to provide a background for the more detailed subsequent analysis of the passive term. All finite verb clauses in these texts were analysed but, because of the restricted storage of the available computer, we excluded from computer processing the highly frequent equative clause type (INT-C; see Section 3.8, pp. 36f.) with be as exponent of V, as in

The mouth is a small opening above the tongue and leads into a large buccal cavity.

The high proportion of active clauses in this limited material made it less useful for a detailed analysis of the passive. For this purpose the Minor Passive Corpus was used for an analysis conducted on similar lines. Since the texts making up the Voice Corpus were also part of the Minor Passive Corpus, all detailed statements about the passive term can best be made on the basis of this passive corpus. It consisted of eleven texts, three spoken (Bl, B2, B3), four scientific texts (Jl, J2, J3, J4), and four novels (Ml, M2, M3, M4). In contrast to the Voice Corpus, the data for the Minor Passive Corpus were not processed text by text but were divided into three classes according to whether the agent was expressed ('agentful passives') or not ('agentless passives'), and according to the character of the agent. All the agentless passives were, however, 'agentive', i.e. admitted agent-extension. The agentful category was divided into one class with animate and into another with inanimate agents:

Class α = agentful with animate agent, Class β = agentful with inanimate agent, Class γ = agentless (but agentive). For practical purposes, all expressed agents were usually treated together (Class

α-β). The taxonomic experiment (see Chapters 5-6) later confirmed that this was also linguistically reasonable.

Excepting anacolutha (two cases in Text Ml and one case in Text M2), the two corpora yielded the following number of clause entries:

(J1.89)

THE VOICE CORPUS THE MINOR PASSIVE CORPUS

Text M l : 561 clauses Text M2: 734 clauses Text J l : 355 clauses

Class a : 68 clauses Class β: 85 clauses Class γ: 497 clauses

Total: 1650 clauses Total: 650 clauses

VOICE DATA PROCESSING 41

4.12 Data

The following linguistic categories were abstracted from the two corpora and encoded into eight columns as input (the maximal number of characters representing linguistic categories in each column are given in square brackets):

Column 1: Type of finite verbal group [71 characters]. Column 2: Exponent classes of the subject and the agent (S and Ag elements) [26 characters]. Column 3: Exponent class of the complement (C element) [63 characters]. Column 4: Exponent class of the adjunct (A element) [62 characters]. Column 5: Clause type [28 characters]. Column 6: External clause relation [30 characters]. Column 7: Order of clause elements [23 characters]. Column 8: Ultimate exponent of the lexical verb.

4.13 Input and output

Each set of data input was an array of eight columns and at most 1024 (210) rows. One set of input (Text Ml) began as follows:

B2KkKRG THINK Bl-RZNX GO OUT B5uA8PX WATCH B5u-8IX SEE

Entries in the first seven columns could be any of about 80 characters drawn from the Ferranti seven-track tape code. The entries in Column 8 were strings of upper case letters and spaces. Spaces were ignored in input; the maximum number of letters permitted was 16. These strings were the ultimate exponents of lexical verbs. They were not stored in the computing store, but in the backing store, and could be re-trieved when necessary. The input was read in by the four primary programs in the same way, one row at a time, backspacing and erasing being permitted. The output had to be on five-hole tape, and consequently many of the output characters were different from the corresponding input characters, for example 'a ' became Ά ' , ψ became Ή Γ , and, to avoid confusion with the numerals used in totals, Τ became *N1\ etc.

4.14 Programs1

Six programs were used, four primary (Programs 1-4) and two secondary (Programs 5-6). The primary programs had as input the data encoded in eight columns, whereas 1 I am greatly indebted to my colleague Η. T. Carvell and to A. J. T. Colin, University of London Institute of Computer Science, for their help in producing the programs. On the use of computers in linguistics, see for example Lamb 1961, Kuöera 1962, Garvin 1963.

42 VOICE DATA PROCESSING

the secondary programs had as input the output of the primary programs. Program 5 ('the Tabulation Program') was used for the output of Program 2, and Program 6 ('the Alphabetization Program') was used for the output of Programs 3 and 4, for copying headings and rearranging listed verbs in alphabetical order.

Program 1 produced an inventory of the characters in each of Columns 1-7. (It was realized afterwards that it would have been substantially more convenient to have the characters in each column listed in a pre-assigned and linguistically mean-ingful order, including those characters that did not occur in all sets of the data.)

Program 2 listed the correlations between all 21 pairs of Columns 1-7, the output order being governed by the input order. The output was tabulated in a linguistically convenient order by Program 5, which also produced row and column inventories and column totals, which were useful both for purposes of analysis and as a constant check on the output.

Program 3 had as object to list the lexical verbs of clauses with certain specified characteristics. For example, since the verbs with animate subjects were required, part of the output (after alphabetization by Program 6) began as follows:

COL2 - Nl , N2, N3, ADMIRE -* 2 ADVANCE AFFORD AGREE etc.

For the program, this meant that ADMIRE in Column 8 cooccurred with one or other of Nl, N2, N3 in Column 2 twice; that ADVANCE occurred similarly once, and so on. Since the codings were

Nl in Column 2 = pronominal animate subject, N2 in Column 2 = nominal animate subject, N3 in Column 2 = coordinate animate subject (of any form-class),

this gave the required information. It will be seen that, in effect, Program 3 needed an additional set of data, namely a specification of all the sets of characters (for example Nl, N2, N3) that were required, together with the appropriate columns. In practice, the additional data were incorporated into the program itself.

Program 4 was similar to Program 3 but a little more ambitious. It caused the computer to print what occurred in both Column 8 (lexical verbs) and a specified column of the first seven when any one of some given list of characters occurred in another given column. In fact there were a number of lists for Column 2 when the cooccurring characters in Column 3 were asked for and similarly a number for Column 3 when the cooccurring characters in Column 5 were asked for; and all these lists were incorporated into the program itself as a matter of convenience. The verbs were again arranged in alphabetical order by Program 6.

VOICE DATA PROCESSING 43

4.15 Computation

The computer used in these experiments was a Ferranti Mercury at the University of London Institute of Computer Science.2

4.16 Results

The analysis of the results was made difficult and laborious by the profusion of de-tailed output. For Programs 2 + 5 alone there were 105 tables, some of which had more than 2000 cells each. It might in similar circumstances be useful to have a program to select the significant correlations and, also, to have some method for 'conflation', i.e. forming larger groups from the small correlations. There is, however, a danger of losing interesting information if this is done before the general picture is known.

The presentation of the results is in no way complete, partly because this would require too much space, partly because a large portion of the results are peripheral or irrelevant to this study. Most of the present results are accounts of the major linguistic categories in Columns 1-7. They derive from the inventories produced by Program 1, except when active categories have been separated from passive in the Voice Corpus, in which case the output of Programs 2 + 5 was used. 'Actives' refer to the actives of Texts Ml, M2, and J1; 'passives', unless otherwise stated, to Classes α-β and γ of the Minor Passive Corpus.

Since the texts and classes have unequal numbers of clause entries, it will be helpful to use percentages for easy comparison. Percentages, given in round brackets under the absolute numbers in the tables, are rounded off to the nearest integer, with the single exception that nonoccurring categories are indicated by dots. Thus 1.7% is given as (2%), 0.3% as (0%), and no occurrences as (.); also, 0.5%, 1.5%, 2.5%, etc. are given as (1 %), (2%), (3%), etc.3

4.2 TYPE OF FINITE VERBAL GROUP

The categories have been outlined above in Chapter 2. 'Present' and 'past' refer to all first elements in verbal group structure; 'simple present' and 'simple past' to 0 and 0 respectively.

It seemed generally true that frequency of occurrence decreased with increasing

1 The size of the computing store was 2,048 words of 20 bits, of which half could be used as a store of 2,048 words of 10 bits. A larger computer would of course perform these tasks more conveniently. For comparison it may be mentioned that the size at present (1965) of the computing store of the Atlas at the Institute is over 32,000 words of 48 bits, and, in addition, Atlas is a very much faster computer than Mercury. * It was found convenient to use logarithmic graph paper for presenting some of the diagrams. This has been indicated by 'log.graph'. The paper has been used in a slightly unusual way, in that 0 (per cent or occurrences) has been plotted against the base line.

44 VOICE DATA PROCESSING

TABLE 4:1

Distribution of verbal group types in the Voice Corpus

STRUCTURE TYPE TEXT Ml t ex t M2 t ex t J1 t o t a l

Simple 0 /0 340 (61%)

434 (60%)

215 (61%)

989 (60%)

989 (60%)

1 element

a/A

b/B

c/C

d\D

94 (17%)

38 (7%)

9 (2%)

24 (4%)

163 (22%)

40 (5%)

34 (5%)

31 (4%)

17 (5%)

7 (2%)

1 (0%)

85 (24%)

274 (17%)

85 (5%)

44 (3%) 140

(8%)

543 (33%)

2 elements

abf AB

act AC

adjAD

bc/BC

bd/BD

cd/CD

8 (1%)

0 (.) 6

(1%) 3

(1%) 8

(1%) 0

(.)

9 (1%)

3 (0%)

7 (1%)

6 (1%)

1 (0%)

0 (.)

0 (·) 0

(·) 19

(5%) 0

(•) 8

(2%) 2

(1%)

17 (1%)

3 (0%)

32 (2%)

9 (1%)

17 (1%)

2 (0%)

80 (5%)

3 elements

abc/ABC

abd/ABD

acd/ACD

bed/BCD

0 (.) 2

(0%) 0

(·) 0 (.)

0 (.) 0

(.) 0

(.) 0

(.)

0 (·) 1

(0%) 0

(.) 0

(.)

0 (.) 3

(0%) 0

(.) 0

(.)

3 (0%)

4 elements abcd/ABCD 0 (.)

0 (.)

0 (.)

0 (.)

0 (·)

Other Imperative 29 (5%)

6 (1%)

0 (·)

35 (2%)

35 (2%)

Total 561 (100%)

734 (100%)

355 (100%)

1650 (100%)

structural complexity in the verbal group. Thus, the distribution between active and passive forms was uneven, since the passive, being Type d in the verbal group, has one more structural component than its corresponding active forms (cf. Types 0'.d,

VOICE DATA PROCESSING 45

6 0 ι

actives

passives

Fig. 4:2. Verbal group types in the Voice Corpus (Percentages based on 1650 occurrences for all three texts. Log. graph.).

46 VOICE DATA PROCESSING

a :ad, b\bd, etc. in Section 2.2, pp. 10f.). The number of occurrences for both voices in texts Ml, M2, and J1 are shown in Table 4:1 and Figure 4:2.

The distribution of active and passive forms varied considerably for the different texts of the Voice Corpus. Table 4:2 provides evidence that the distributional differ-ence between Ml and M2 as regards voice is real but very small. However, between either of the Μ texts and J1 the degree of association is high, and the evidence is extremely significant (ρ < 0.001, i.e. the chance of occurrence is less than 1/1000, as determined by the χ2 test; see Section 6.32, pp. 116ff.).

TABLE 4:2

Voice distribution for the texts of the Voice Corpus

TEXT M l TEXT M 2 TEXT J1 TOTAL

Active 521 695 240 1456 Active (93%) (95%) (68%) (88%)

Passive 40 39 115 194 Passive (7%) (5%) (32%) (12%)

Total 561 734 355 1650 Total (100%) (100%) (100%) (100%)

In the Voice and Minor Passive Corpora only Types 0, 0, a, A, b, B, c, C, d, D, ad, AD, bd, and BD, had frequencies of 2% or more in any one text or class. Types ab, AB, AC, be, BC, cd, CD, abd, ABD had less than 2% occurrences. No instances were recorded of the active types ac, abc, ABC, and their corresponding passives acd, abed, ABCD, nor of ACD, bed, or BCD (see Table 4:3 and Figure 4:3).

There was in the material a negative correlation between Type c/C and the passive, not only for cd/CD (both 1 % or less) but also for the combinations acd/ACD and bed/BCD (all nonoccurring). The increasing effect of Types a/A and b\B with the passive (see next paragraph) seems absent here. The number of cooccurring c/C and d/D was low: 2/192 is less than 56/376 (see Table 4:4). The Minor Passive Corpus had an even lower ratio for c/C + d/D as against ~ c/C + d/D (5/642).

TABLE 4:4

Contingency table for Types c/C and d/D in the Voice Corpus p = 6 x 10-8

c/C ~ c/C TOTAL

d\D 2 192 194

~ d/D 56 376 432

58 568 626

~ = 'not'.

VOICE DATA PROCESSING 4 7

TABLE 4:3

Distribution of verbal group types in the Voice Corpus (active only) and in the

Minor Passive Corpus

ACTIVE PASSIVE

TYPE TEXT TEXT TEXT TOTAL TYPE CLASS CLASS TOTAL M l M 2 J 1 α - β Y

0 4 7 1 3 6 2 1 3 3 9 6 d 4 9 1 5 6 2 0 5 ( 9 % ) ( 1 9 % ) ( 9 0 % ) ( 2 7 % ) ( 3 1 % ) ( 3 1 % ) ( 3 2 % )

0 2 9 3 2 9 8 2 5 9 3 D 4 4 1 2 4 168 ( 5 6 % ) ( 4 3 % ) ( 1 % ) ( 4 1 % ) ( 2 9 % ) ( 2 5 % ) ( 2 6 % )

a 4 9 8 7 15 151 ad 19 7 8 9 7 ( 9 % ) ( 1 3 % ) ( 6 % ) ( 1 2 % ) ( 1 2 % ) ( 1 6 % ) ( 1 5 % )

A 3 4 5 6 1 9 1 AD 5 5 5 6 0

a%) ( 8 % ) ( 0 % ) ( 6 % ) ( 3 % ) ( 1 1 % ) ( 9 % ) a-A* 11 2 0 1 3 2 ad-AD* 0 15 15

( 2 % ) ( 3 % ) ( 0 % ) ( 2 % ) ( · ) ( 3 % ) ( 2 % ) b 1 6 18 7 4 1 bd 2 9 4 3 7 2

( 3 % ) ( 3 % ) ( 3 % ) ( 3 % ) ( 1 9 % ) ( 9 % ) ( 1 1 % ) Β 2 2 2 2 0 4 4 BD 3 16 19

( 4 % ) ( 3 % ) (.) ( 3 % ) ( 2 % ) ( 3 % ) ( 3 % ) c 3 15 1 19 cd 1 3 4

( 1 % ) ( 2 % ) ( 0 % ) ( 1 % ) ( 1 % ) ( 1 % ) ( 1 % ) C 6 19 0 2 5 CD 0 1 1

( 1 % ) ( 3 % ) (.) ( 2 % ) (.) ( 0 % ) ( 0 % ) ab 2 0 0 2 abd 1 1 2

( 0 % ) ( · ) ( · ) ( 0 % ) ( 1 % ) ( 0 % ) ( 0 % ) AB 6 9 0 15 ABD 1 3 4

( 1 % ) ( 1 % ) (.) ( 1 % ) ( 1 % ) ( 1 % ) ( 1 % ) ac 0 0 0 0 acd 0 0 0

(.) (.) (.) (.) (.) (.) (.) AC 0 3 0 3 ACD 0 0 0

(.) ( 0 % ) (.) ( 0 % ) (.) (.) (.) be 3 4 0 7 bed 0 0 0

( 1 % ) ( 1 % ) (.) ( 0 % ) ( · ) ( · ) (.) BC 0 2 0 2 BCD 0 0 0

(.) ( 0 % ) (•) ( 0 % ) (.) (.) ( · ) abc 0 0 0 0 abed 0 0 0

(.) (.) (.) ( · ) (.) (.) ( · ) ABC 0 0 0 0 ABCD 0 0 0

(.) (.) (.) (.) (.) (.) (.) Imperative 2 9 6 0 3 5 Imperative 1 0 1

( 6 % ) ( 1 % ) ( · ) ( 2 % ) ( 1 % ) (.) ( 0 % ) Subjunctive 0 0 0 0 Subjunctive 0 2 2

( · ) (.) ( · ) ( · ) (.) ( 0 % ) ( 0 % )

Total 521 6 9 5 2 4 0 1 4 5 6 Total 153 4 9 7 6 5 0 Total ( 1 0 0 % ) ( 1 0 0 % ) ( 1 0 0 % ) ( 1 0 0 % )

Total ( 1 0 0 % ) ( 1 0 0 % ) ( 1 0 0 % )

* Hyphens between types denote non-tense-marked auxiliaries (see Section 2.34).

VOICE DATA PROCESSING

ACTIVE

Percentages based on 1456 occurrences.

PASSIVE Percentages based on 650 occurrences.

Fig. 4:3. Verbal group types: active (Voice Corpus) and passive (Minor Passive Corpus). Log. graph.

VOICE DATA PROCESSING 49

If we consider the passive and active verbal group types separately (i.e. with and without Type d j D, as shown in Table 2:1, p. 11), we find that the passive have distributions similar to the corresponding active. The exceptions were Types bd (19 % in Class α-β and 9% in γ), and ad/AD in Class γ (30%), all of which were more frequent in the passive material than the corresponding actives in the active material. In interpreting the figures for the passive corpus, it must be remembered that one might expect them to bear closer resemblance to those for the corresponding actives in Text J1 than in Texts Ml and M2, since the passive is particularly frequent in scientific texts, which yielded the majority of the clauses in the Minor Passive Corpus. This would help to explain, for instance, the greater number of Type d than Type D occurrences, since simple present was strikingly more common than simple past in Text J1 (213 as against 2 occurrences). In the case of the categories just mentioned there is, however, no reason for suspecting that the high passive frequencies were text-conditioned rather than voice-conditioned: the percentage for Type bd was 19 in Class α-β, com-pared with 3 for Type b in Text J l ; and it was 30 for Type ad/AD in Class γ, compared with 7 for Type a/A in Text Jl. In fact, in this case the similarity was greater with Texts Ml (18%) and M2 (23%). Discounting the auxiliary do, which cannot occur in the 6e-passive (except in imperatives of the type Don't be deceived into doing it!) the percentage for Type a/A was much higher in the passive than in the active: 26% (172 out of 650 cases) for Type ad/AD, compared with 19% (274 out of 1456 cases) for Type a/A. In view of the fact that the number of active clauses in the Voice Corpus was more than twice that of passive clauses in the Minor Passive Corpus, the following auxiliaries in Table 4:5 seem particularly frequent in the passive verbal group.

TABLE 4:5

Occurrences of Type a/A auxiliaries

AUXILIARY PASSIVE ACTIVE

can 46 32 may 22 8 could 25 20 have to

and other open class auxiliaries (see Section 2.33) 28 22

4.3 EXPONENT CLASSES OF SUBJECT A N D AGENT (5 and Ag elements)

Subject and agent categories, only some of which will be distinguished here, have been outlined in Sections 3.2 (pp. 26if.) and 3.33 (pp. 30f.). (In the code, the same .S characters were used for both actives and agentless passives in the Voice Corpus, but

50 VOICE DATA PROCESSING

the distribution of the different categories within the two voices could be obtained by correlating, for instance, Columns 2 and 5. Agentful passives were, however, given specific characters for different combinations of S and Ag categories.)

4.31 Subject gender

There were about twice as many animate as inanimate subjects in the Voice Corpus (1072 : 539). Taking the texts and voice terms separately, however, it appears that the proportions are drastically different. The frequency of inanimateness in the sub-ject displayed a certain parallelism with passive frequency in the texts:

Text Ml : 21 % inanimate subjects and 7% passives Text M2: 12% inanimate subjects and 5% passives Text J1: 97% inanimate subjects and 32% passives

The ratios for passive clause subjects indicate that, compared with active, they have a tendency towards inanimateness. When both text and voice factors combined, as in the case of Text Jl, which had no animate but 115 inanimate passive subjects, the result is particularly striking. The relative proportions of each gender remained constant for both voices in the three texts: M2 had most, Ml less, and Jl least animate subjects in both active and passive.

TABLE 4:6 Voice and subject gender in the Voice Corpus

(p < 10""; for p, see pp. 116ff.)

San Sin TOTAL

Active 1035 383 1418

Passive 37 156 193

Total 1072 539 1611

The association was exceedingly high between passive voice and inanimate subject as can be seen in Table 4:6 where all three texts in the Voice Corpus are brought together in a contingency table. This association appears even more remarkable in view of the fact that there is no significant or even remotely significant textual differ-ence between Jl on the one hand and Μ1-2 on the other in the way in which they associate San/Sin with active/passive (see Table 4:7a, b, c).

In the Minor Passive Corpus, more than f of the clauses had inanimate subjects (see Table 4:8).

VOICE DATA PROCESSING 51

TABLE 4:7

Voice and subject gender in Texts Ml, M2, J1

(a)

San Sin TOTAL

TEXT M l

Active 406 85 491

TEXT M l Passive 14 26 40 TEXT M l

Total 420 111 531

p< 10-1 1

(b)

San Sin TOTAL

TEXT M 2

Active 617 70 687

TEXT M 2 Passive 23 15 38 TEXT M 2

Total 640 85 725

p< 10- '

(C)

San Sin TOTAL

TEXT J 1

Active 12 228 240

TEXT J 1 Passive 0 115 115 TEXT J 1

Total 12 343 355

ρ = 0.008

4.32 Agent gender

In the Minor Passive Corpus there were more inanimate than animate expressed agents ( = Classes β and α respectively), and animate subjects were even more infre-quent in both classes (see Table 4:8).

TABLE 4:8

Subject gender in the Minor Passive Corpus

San Sin TOTAL

Class α (Agon)

9 (13%)

59 (87%)

68 (100%)

Class β (Ag in)

5 (6%)

80 (94%)

85 (100%)

Class γ

OO 0

0 400 (82%)

488 (100%)

Total 102 (16%)

539 (84%)

641 (100%)

52 VOICE DATA PROCESSING

4.33 Subject form-class

As was the case with gender, Texts Ml and M2 were very similar in respect of subject form-class. Both had f pronouns and J nouns as subjects with actives and passives taken together, whereas for Text Jl, these figures were reversed in the active and there were | nouns and \ pronouns in the passive. For Class γ the proportions were f pronouns and $ nouns. (Classes α and β were not fully distinguished with regard to form-class.) The distribution of pronominal and nominal subjects was roughly parallel in the active and passive clauses of Texts Ml, M2, and Jl .

4.34 Agent form-class

The predominant agent form-class was nouns. Only 3 out of 153 agents had pro-nominal exponence.

4.35 Coordination

Coordination was rare in the subject (8 cases in Text Ml, 5 in Text M2, 2 in Text Jl), but it was much more frequent in the agent. There were 39 cases there, 34 of which occurred in Class a ; i.e. more than half the agents in this class were coordinate.

4.36 Length of passive subject and agent

Since there was little variation in Classes α, β, and γ for the categories described in Column 7 (order of clause elements), this column was also used for registering the length of the subject, measured in terms of number of words. (Hyphenated items counted as single words.) Elements consisting of 1-10 words were given precise values, those consisting of 11 or more words had to be grouped together in five groups. The number of subject elements with more than 10 words was small: 18 in Class α-β and 26 in γ. The lengths of agent elements in Class α-β were counted manually. See Figures 4:4 and 4:5.

The average length of the subjects in agentless passive clauses was just over 3 words (arithmetical mean = 3.2), and almost 5 words (mean = 4.96) in agentful clauses. The mode, i.e. the most common word-length, was 1 word (45 % in Class γ, and 28% in α-β; the next most common, 2 words, had percentages of 18% and 19%, respectively). Agents had an average length of 8.5 words. The mode was 2 words. For both elements, the norm seems to be: decreasing frequency with increasing word-length. The figures for up to 10 words, covering 95% of the subjects in Class γ, and 88 % of the subjects and 73 % of the agents in α-β, are shown in Table 4:9 (p. 56). These figures suggest firstly, that there is considerable difference in weight, measured

VOICE DATA PROCESSING 53

2 3 4 5 6 7 8 9 1 0 11 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3 2 4 ^ 5 jjj

NUMBER OF WORDS ' — · — « " r §

Fig. 4:4. Subject-lengths in Class γ. (Log. graph.)

54 VOICE DATA PROCESSING

Fig. 4:5a. Subject-lengths in Class α-β. (Log. graph.)

in terms of word-length, between subjects and agents, in that the latter tend to be much longer (the ratio of 1-word subjects to 1-word agents was 5 : 1 ) ; secondly, that the subjects in agentful passive clauses are longer than subjects in agentless clauses.

VOICE DATA PROCESSING 55

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 NUMBER OF WORDS Η

Fig. 4:5b. Agent-lengths in Class α-β. (Log. graph.)

4.4 EXPONENT CLASSES OF COMPLEMENT (C element)

60 categories besides C absent and anacolutha were encoded: 29 C1 (simple comple-ment) and 32 C2 (double complement). For illustration of C elements, see Section 3.2 (pp. 26 ff.).

4.41 Complement Distribution

Table 4:10 shows the overall distribution of complements in both corpora. Text J1 differed from Ml and M2 in having fewer complements, in particular fewer C2.

4.42 Complement form-class

In active clauses exponents of simple complements were, in order of frequency: nouns (38%), direct speech (22%), pronouns (20%), finite verb clauses (9%), non-

56 VOICE DATA PROCESSING

TABLE 4:9 Number of words in subjects and agents of the Minor Passive Corpus

NUMBER OF CLASS γ CLASS α - B

NUMBER OF CLASS γ

WORDS SUBJECT SUBJECT AGENT

0 9 2 —

1 223 42 8 2 93 30 27 3 47 15 11 4 27 8 21 5 20 11 8 6 12 9 11 7 17 8 5 8 11 7 7 9 6 1 8

10 6 2 6

Total 471 135 112 Other 26 18 41

(i.e. over 10)

Total 497 153 153

Mean 3.2 words 5.0 words 8.5 words

Mode 1 word 1 word 2 words

TABLE 4:10 Distribution of complements

ACTIVE PASSIVE

TYPE TEXT M l TEXT M 2 TEXT J 1 CLASS α - β CLASS γ

C-less 179 130 105 138 405 (34%) (19%) (44%) (90%) (81%)

Simple Complement 307 531 134 15 92 (C) (59%) (76%) (56%) (10%) 09%) Double Complement 35 34 1 0 0 ( C ) (7%) (5%) (0%) (·) ( · )

TOTAL 521 695 240 153 497 (100%) (100%) (100%) (100%) (100%)

finite verb clauses (6%), and verbless constructions (5%). Texts Ml and M2 had different distributions from Text J1: in the latter nouns were about twice as frequent as in Ml and M2, and pronouns and speech were very much more rare, whereas verbless constructions were more common (see Table 4:11).

VOICE DATA PROCESSING 57

TABLE 4:11

Form-class of simple complements (C1)

ACTIVE PASSIVE

TEXT M l TEXT M 2 TEXT J l TOTAL CLASS α - β CLASS γ TOTAL

Noun 107 166 98 371 1 9 10 (34%) (31%) (73%) (38%) (7%) (10%) (9%)

Pronoun 76 116 2 194 0 1 1 (25%) (22%) (2%) (20%) (·) (1%) (1%)

Finite verb 23 62 6 91 3 29 32 clause (8%) (12%) (4%) (9%) (20%) (31%) (30%) Nonfinite verb 14 34 8 56 9 45 54 clause (5%) (6%) (6%) (6%) (60%) (49%) (51%) Verbless 18 16 17 51 2 8 10 construction (6%) (3%) (13%) (5%) (13%) (9%) (9%) Speech 69 137 3 209 0 0 0

(22%) (26%) (2%) (22%) (·) (·) (·)

Total 307 531 134 972 15 92 107 Total (100%) (100%) (100%) (100%) (100%) (100%) (100%)

In passive clauses the order of frequency was: nonfinite verb clauses (51%), finite verb clauses (30%), verbless constructions and nouns (each 9 %), and pronouns (1 %). There were no instances of speech. Nonfinite and finite verb clauses were particularly common, with 54 and 32 occurrences respectively, hence making up 80% of C1

exponents in the passive. The to-infinitive {to V) was the most frequent (50 occur-rences) among nonfinite verb clauses, and ί/ιαί-clauses among finite verb clauses. Active clauses in the Voice Corpus as a whole had more zero that- than ίΑαί-clauses as C (21 versus 45 occurrences), whereas Text J1 had only ίΑαί-clauses (6 that- versus no occurrences of zero that-). The passive material had predominantly /Aai-clauses (27 occurrences as compared with 3 zero ίΑαί-clauses).

4.43 Complement gender

Inanimate complements were by far the most frequent everywhere, and the predomi-nance was complete in the scientific text. In the active, Texts Ml and M2 had 33:274 (11:89 %) and 73:458 (14:86 %) respectively for animate:inanimate simple complement, as compared with 0:134 for Text Jl . Texts Ml and M2 were similar in this respect, but the distribution in these texts in favour of animateness, which was apparent in the case of subjects (see Section 4.31), did not obtain for the complement. More than $ of their complements were inanimate.

4.5 EXPONENT CLASSES OF ADJUNCT (A element)

The categories relating to the adjunct have been described in Section 3.3 (pp. 29ff.). Column 4 registered only Ad elements (i.e. nonagentive adjuncts), since Ag (agentive

58 VOICE DATA PROCESSING

adjuncts) in Classes α-β were encoded into Column 2 (see Sections 4.32ff.). In the presentation below, however, A comprises both the subclasses Ag and Ad. A elements are considered either in terms of Ά sets' or in terms of Ά items'. A sets refer to overall adjunct structure. Thus we have M-less' clauses, which have no A item, and M-fulT clauses, which have one or more A items. Hence A items refer to the number of individual As in each clause. A sets can be given as A1, A2, A3, A4, the superscript denoting the number of A items in the A set. Frequencies for Texts Ml, M2, and J1 refer only to their active clauses.

4.51 Adjunct frequency

Less than half the active clauses in Texts Ml and M2 had one or more adjuncts, whereas most of the active clauses in Text Jl, and most of the clauses in the Minor Passive Corpus, had A sets. Class α-β had of course by definition one A item (Ag) per clause; in addition, Ad sets occurred in half the agentful clauses (see Table 4:12 and Figure 4:6).

The most common A set type was A1, the next most common was A2, then followed A3, and finally A*. In the material, A4 was the maximum number of A items in any one clause. It was very rare: only seven occurrences in 2106 clauses. The frequency of A sets decreased with the increasing number of A items in the set. (Cf. the negative correlation between frequency and structural complexity in the verbal group men-tioned in Section 4.2.)

4.52 Adjunct Position (see Table 4\12 and Figure 4:7)

With regard to the three major adjunct positions in the clause, front, mid, and end (outlined in Section 3.32, p. 30), remarkable similarity obtained in both corpora except for Class α-β, whose clauses had fewer front- and more end-positions than the others. The agent is almost invariably end-placed. For Class γ and actives in Texts Ml, M2, Jl the approximate proportions were as follows:

Front-placed adjuncts: ^ Mid-placed adjuncts: ^ End-placed adjuncts: ^

It is interesting to note that the agentful class had considerably fewer Ad items than the agentless class. The proportions of Ad items in relation to clauses were 95 : 153 (62%) for Class α-β and 470 : 497 (95%) for Class γ. This would suggest that it is not entirely realistic to consider Ag as an optional, additional clause element which is different from Ad. Rather, the figures suggest that Ag and Ad are to some extent interdependent, in so far as the number of A items is remarkably constant, and that there may be restrictions on the number of As generally in any one position.

VOICE DATA PROCESSING 59

TABLE 4:12

Adjunct items, sets and positions

A SET TYPE

FRONT POSITION

MID POSITION

END POSITION

A ITEM TOTAL

A SET TOTAL

f

TEXT Ml

Λ-less A1

A%

A3

A*

35 25 5 0

8 5 1 1

140 58 21

3

0 183 88 27 4

284 (55%) 183 (35%) 44 (8%) 9 (2%) 1 (0%)

t Total 65 15 222 302 521 (100%)

4 §

I Β 8 υ s

TEXT M2

-less A1

A' A» A*

36 15 1 1

30 7 2 1

146 74 12 6

0 212

96 15 8

428 (62%) 212 (30%) 48 (7%)

5 (1%) 2 (0%)

4 §

I Β 8 υ s

Total 53 40 238 331 695 (100%)

4 §

I Β 8 υ s

TEXT J1

Λ-less A1

A1

A» A1

23 13 0 2

14 9 2 0

77 50 19 6

0 114 72 21

8

81 (34%) 114 (47%) 36 (15%) 7 (3%) 2 (1%)

Total 38 25 152 215 240 (100%)

s 8

CLASS α-β

Λ-less A1

A%

A· A»

0 8 5 1

0 23 7 0

74 97 30 3

0 74

128 42

4

0 (.) 74 (48%) 64 (42%) 14 (9%) 1 d % )

s 8

Total 14 30 204 248 153 (100%)

ί ß*

i CLASS

γ

Λ-less A1

A" A' A*

34 33 5 0

28 25 2 1

187 138 14 3

0 249 196 21 4

142 (29%) 249 (50%) 98 (20%) 7 (1%) 1 (0%)

Total 72 56 342 470 497 (100%)

4.53 Adjunct form-class

Adjunct form-classes were registered for the most frequent A set types, A1 and A2

(see Figure 4:8). The prepositional phrase was by far the most common form-class, accounting for just under half the A items in Texts Ml and M2, and for 70% in Text J1 and 75 % in the passive clause classes. Adverbs, on the other hand, were more frequent in the active than the passive (27 % : 19 %), and, in the active material, more common in the novels than the scientific texts. Nonfinite verb clauses as A were

60 VOICE DATA PROCESSING

60%

50%

40%

30% -

20% -

10% -

- A A1 A2 A3 A4 - A A1 A2A3 A4 - A A1 A2 A3 A4

Voice Corpus Class α-β Class γ

Fig. 4 :6 . A set distribution in per cent. (Log. graph.)

VOICE DATA PROCESSING

F M E F M E F M E Active Class α-β Class γ

Fig. 4:7. Positions of A elements in per cent. (Log. graph.)

F = front-position Μ = mid-position Ε = end-position

62 VOICE DATA PROCESSING

TABLE 4:13 Major clause types in the Voice Corpus

ACTIVE PASSIVE

TYPE TEXT TEXT TEXT TOTAL TYPE TEXT TEXT TEXT TOTAL Ml M2 J1 Ml M2 J1

Intransitive 159 115 103 377 (EXT-0) (28%) (16%) (29%) (23%)

Transitive 310 522 117 949 Transitive 29 34 114 177 (EXT-C) (55%) (71%) (33%) (58%) (ext-0) (5%) (5%) (32%) (11%)

Ditransitive 15 14 0 29 Ditransitive 3 1 0 4 (.EXT-CC) (3%) (2%) ( · ) (2%) (ext-C) (1%) (1%) (·) (0%)

Equative 17 22 18 57 (INT-C) (3%) ( 3 % ) (5%) (4%)

Factitive 20 22 2 44 Factitive 8 4 1 13 (INT-CC) (4%) ( 3 % ) (1%) ( 3 % ) (int-C) (2%) (1%) (0%) (1%)

521 695 240 1456 40 39 115 194 (93%) (95%) (68%) (88%) (7%) (5%) (32%) (12%)

three times as common in active as in passive. Figure 4:8 shows that the form-class distribution for A items was very similar in both Class α-β and Class γ. It suggests, for example, that the number of prepositional phrases in a passive clause is usually limited and fairly constant, regardless of whether there is an agent-phrase or not.

4.6 CLAUSE TYPES

The categories presented in Section 3.8 (pp. 36 f.) were further refined. The clause type active transitive (EXT-C), the largest one, was subdivided into 'fully transitive' and 'semitransitive'. Fully transitive clauses have nominal groups as complements and, normally, systemic voice transformation potential:

He likes her. -*• She is liked.

Semitransitives have no such potential or have some kind of restriction on voice transformation, such as exponential constraint, when they have nominalizations or reflexive/reciprocal pronouns as exponents of C:

He likes to read. *To read is liked. . . . , Γ *That he was leaving was said. 1

They said he was leaving. -» T [I t was said that he was leaving. J He likes himself. -» ""Himself is liked.

Table 4:13 and Figure 4:9 show the frequencies of the major clause types in the texts of the Voice Corpus. The totals were, in descending order:

VOICE DATA PROCESSING 63 FVC = finite verb clause

Fig. 4:8. Form-class of A elements in per cent. (Log. graph.)

(EXT-C) Active transitive 949 (58%) cEXT-0) Active intransitive 377 (23%) (ext-0) Passive transitive 177 (11%) (INT-C) Active equative 57 ( 4 % ) (INT-CC) Active factitive 44 ( 3%) (EXT-CC) Active ditransitive 29 ( 2%) Cint-C) Passive factitive 13 ( 1%) (ext-C) Passive ditransitive 4 ( 0 % ) Total 1650 (100%)

64

Active transitive, than in J l :

ditransitive,

VOICE DATA PROCESSING

and factitive were more common in Texts Ml and M2

Text Ml Text M2

(EXT-C) 55% 71% (.EXT-CC) 3% 2% (1NT-CC) 4% 3%

Fig. 4:9. Major clause types in the Voice Corpus. (Log. graph.)

Text J1

33% 0% 1%

The difference for active transitive was due chiefly to the disproportionately larger number of semitransitive to fully transitive clauses in the novels as compared with the scientific text. Speech as C was particularly frequent in Ml and M2. The number of fully transitive clauses was very similar in the three texts: 31%, 25%, and 29% respectively for Ml, M2, Jl. The two novel texts were much alike with regard to clause type distribution, but differed in the following two categories (Text Jl also given for comparison):

VOICE DATA PROCESSING 65

Text Ml Text M2 Text J1

{EXT-0) (EXT-C)

28% 55%

16% 71%

29% 33%

Passives, as mentioned in connection with the verbal group (Section 4.2, p. 46), were much more frequent in the scientific text than in the novels. The large number of passive clauses in Text J1 was, however, unevenly distributed in the subcategories. In fact, the difference in passive clause frequency among the Voice Corpus texts appeared only in the passive transitive (ext-0), which totalled 5% in Text Ml , 4% in M2, but 32% in J l .

For the Minor Passive Corpus it was technically possible to register not only passive clause type according to whether the clauses were C-full or C-less, but, in the case of Class γ, also the number of A items. (The exponents of C and A could be obtained by correlation of Column 5 with Columns 3 and 4.)

Classes α and γ showed similar proportions of C-less and C-full clauses, approx-imately 8:1. Class β, on the other hand, was almost entirely C-less.

Figure 4:10 shows the A set type cooccurring with C-less and C-full clauses in the large Class γ. A1 was by far the most common for C-less clauses, whereas C-full clauses usually had no adjuncts. This suggests that passive clauses normally have one clause element other than Sand V. This element may be either Cot A, less often both. The most complex A set type was A3, which was rare in C-less (2%) and nonoccurring in C-full clauses. For both elements taken together the occurrences were as follows:

For a description of the relevant categories see Section 3.5 (pp. 31 ff.). No distinction is made below between affirmative and interrogative clauses. Imperatives are included in free clauses.

Table 4:14 and Figure 4:11 represent the distribution of the clauses among the three main types of external clause relation: syntactically bound, sequentially related, and free clauses. Sequentially related clauses, which showed little variation in the different texts and classes, accounted for roughly Considering only actives in the Voice Corpus, the percentages were: Text Ml 25%, Class α-β 24%, Text Jl 20%, Class γ 21 %, Text M2 17%. The ratios of syntactically bound clauses were approx-imately YQ for Text Ml, \ for Texts M2, and Jl , and Class α-β, but \ for Class γ.

No CI A element (C-less, or Λ-less) 75 (15%) One C\A element (C1, or A1) 294 (59%) Two C/A elements (C1 + A\ or A2) 119 (24 %) Three C/A elements (C1 + A2, or A3) 9 ( 2 % )

4.7 EXTERNAL CLAUSE RELATION

66 VOICE DATA PROCESSING

240 -

220 -

200 -

180 -

160 -

140 -

120 -

100 -

Vi

8 I 80 -g ο υ ο

60 -

40 -

20 -

Fig. 4:10. C and A elements in Class γ. (Log. graph.)

VOICE DATA PROCESSING 67

Free clauses accounted for half or more in Texts Ml, M2, Jl, and Class α-β, but for less than ^ in Class γ.

These distributional differences in the material are interesting because they seem to cut across voice as well as text divisions, both of which have been found to be significant elsewhere. The numbers here would suggest that, in respect of external clause relation, agentful passives are more like actives than agentless passives. Half the agentful clauses were free, i.e. had little or no external relationship, whereas, in agentless passive clauses, less than ^ of the clauses were free and as many as half were syntactically bound. However, the difference between syntactically bound and sequentially related in Text Ml, on the one hand and in Texts M2 and Jl and Class α-β on the other, would seem to be textually conditioned.

TABLE 4:14 External clause relation

ACTIVE PASSIVE

TYPE TEXT TEXT TEXT CLASS CLASS TYPE Ml M2 Jl α-β Υ

(la) Relative clauses 19 55 44 23 98 (lb) Conjunctional that- or elliptic /fart-clauses 14 43 10 11 59 (lc) Other conjunctional clauses 25 87 11 5 89

(1) SYNTACTICALLY BOUND CLAUSES 58 185 65 39 246

(2a) Fixed relator 19 55 17 10 45 (2b) Fixed relator + elliptic subject 40 14 26 4 11 (2c) ± Fixed relator + elliptic subject + elliptic

auxiliary 8 15 0 5 7 (2d) Elliptic subject 24 2 0 2 2 (2e) Elliptic lexical verb 18 16 0 0 5 (2f) Mobile relator 17 12 6 15 35 (2g) Tag questions 2 2 0 0 1

(2) SEQUENTIALLY RELATED CLAUSES 128 116 49 36 106

(3a) Having clause as C 16 42 5 0 2 (3b) Having direct speech as C 74 137 0 0 2 (3c) Being direct speech 3 12 0 0 1 (3d) Followed by sequentially related clause 49 38 46 17 44 (3e) Imperatives 31 7 0 1 0 (3f) Fully free 162 158 75 60 96

(3) FREE CLAUSES 335 394 126 78 145

TOTAL 521 695 240 153 497

If we analyse the main types in greater detail, it appears that, in syntactically bound clauses, Class γ had the highest proportions in all the three subtypes relative, conjunc-tional that, and other conjunctional clauses.

68 VOICE DATA PROCESSING

In sequentially related clauses, Texts Ml and M2 showed affinities, for example 52 and 35 cases, respectively, for the categories ± fixed relator + elliptic subject + ellip-tic auxiliary, elliptic subject, elliptic lexical verb, and tag question, as compared with no occurrences at all in Text J1. On the other hand, more than half the clauses in the latter text had fixed relator + elliptic subject. Although similar in other respects, Text J1 and the passive classes differed here: the latter had only 11% such clauses. In the passive, the most frequent sequentially related clause type was the one with fixed relators for Class γ, and the one with mobile relators for Class α-β. The fre-quencies of the two subtypes, representing close and open connection respectively on the scale of external clause relation within the sequentially related type, reflect the proportions of the main types of bound and free clauses in the passive classes.

Fig. 4:11. External clause relation. (Occurence in per cent)

VOICE DATA PROCESSING 69

4.8 INTERNAL CLAUSE RELATION (See Section 3.4, p. 31, for examples of the categories)

This column emphasized stylistic differences. Only texts Ml and M2 had any note-worthy deviation from the unmarked direct SV order. Considering only these two cardinal clause elements, the ratios for SV and VS are given in Table 4:15.

4.9 SUMMARY OF THE RESULTS

In our selective inventory analysis of some voice categories of the total output, we have arrived at numerous results of which a few seem significant, some suggestive, and many trivial.

TABLE 4:15

Internal clause relation (Both actives and passives are included in the Voice Corpus figures)

ORDER TEXT Ml

TEXT M2

TEXT J1

CLASS α-β

CLASS

Y

SK 490 577 350 150 489 (87%) (79%) (99%) (98%) (98%)

VS 71 157 5 3 8 (13%) (21%) (1%) (2%) (2%)

Total 561 734 355 153 497

That a large part of the results should be 'trivial' was only to be expected with the selected programs, and will not be further commented on here. It may be worth emphasizing, however, that the difference between trivial and non-trivial results is far from obvious. A statement such as 'there are more inanimate subjects in scientific expository writing than in fictional as represented by novels' might well be called trivial. Precise numerical results become less trivial, however, when considered in a larger linguistic context, for example, in studying the style used in one field of scientific research as compared with that of other fields. There are good reasons for believing that the term 'scientific style' embraces diverse scientific styles. When it comes to comparing large numbers of less strikingly dissimilar texts, we can no longer be sub-jectively sure of the differences. Furthermore, although we appreciate that our scien-tific text has more inanimate subjects and more passive clauses than the novels, we are not as likely to know what is the relationship of gender and voice distribution. The information that only 12 out of 355 subjects were animate in Text J1 becomes more significant in the light of the additional statement that all those 12 subjects occurred in active clauses, the ratio of animate to inanimate subjects in passive clauses of Text J1 being 0:115.

Some results which seem interesting but which were inadequately dealt with in the present analysis cannot be used for further statements at this point. They include

70 VOICE DATA PROCESSING

voice transformation potential, passive subclasses, and the weight of clause elements, and will be considered in greater detail in the two subsequent chapters.

4.91 Structure and frequency

As for the seemingly significant results, we shall now try to make a few points, summarising some of the scattered remarks which were made earlier in this chapter.

In the following categories there appeared to exist a negative correlation between frequency of occurrence and structural complexity.

(a) The finite verbal group (Section 4.2). Type d/D was more frequent than Types bJB and c/C and should, with respect to frequency, be placed between a/A and b/B. As was stated in Section 2.2 (pp. 10 f.), the reason for the adopted system is structural order of the verbal group elements.

(b) Adjuncts (Section 4.51). Figure 4:6 shows a positively skewed distribution of the sets A1*, which is to say that, whenever an A set was present in a clause of any text or class, A1 was more frequent than A2, which was more frequent than A3, which was more frequent than A*.

(c) Complements (Section 4.41). In all active clauses (there are no C2 in passive clauses), C1 was much more frequent than C2.

(d) Subject length, i.e. if their length, measured in terms of number of words, is considered comparable to structural complexity (see Section 4.36). The number of words in expressed subjects varied from 1 to 37, the most common being 1-word subjects, both in agentful and agentless passive clauses. In all classes the distributions were, by and large, positively skewed (see Figures 4:4 and 4:5). This was also generally true of agent length, but less clearly so, the mode being 2 words and the diagram contour having peaks at 2, 4, 6, and 9 words.

4.92 Style

The choice of material turned out to be a rather interesting one. We have found cate-gories where variations in distribution were obviously stylistically conditioned, and it is sometimes striking how similar the two novel texts were to each other, on the one hand, and, on the other, how different they were from the scientific text. This was found to be the case for example with the following categories:

(a) Voice distribution, with more than four times as many passive clauses in Text J1 as in either Ml or M2 (Section 4.2).

(b) Subject gender. The proportions were 5:1 in Text Μ1, 9:1 in Text M2, but 1:19 in Text J1 for animate:inanimate subjects of active clauses (Section 4.31).

(c) Complement gender in terms of animate :inanimate was roughly 1:7 in active clauses for Texts Ml and M2 as compared with 0:134 for Text J1 (Section 4.43).

(d) Subject form-class. Texts Ml and M2 had f pronominal and ^ nominal sub-

VOICE DATA PROCESSING 71

jects, whereas in Text J1 the proportions were reversed in active clauses (Section 4.33). (e) Complement form-class. The preponderance of subject nouns in Text J1 ob-

tained also in the complement, whereas Texts Ml and M2 had proportionately more pronouns and speech (Section 4.42).

(f) Clause types. Particularly rare in Text J1 were active transitives (EXT-C), and there were no instances of ditransitives (EXT-CC) (Section 4.6).

(g) Internal clause relation. 99% of the clauses in Text J1 had direct SV order, whereas there was more variation in the Μ texts (Section 4.8).

(h) Adjunct set distribution. A1 was more frequent than Λ-less in Text Jl , but the proportions were reversed for the Μ texts (Section 4.51). There were however no obvious stylistic differences in the case of, for example, external clause relation (Section 4.7) and position of adjuncts (Section 4.52).

4.93 Voice

Since we shall henceforth concentrate on the passive voice, it might be interesting to see in what respect voice terms may be taken to bear a relationship to the distributions of other categories. The following are some observations which can be made on the basis of the analysed material:

(a) No parallel active-passive distribution obtained in the verbal group, since the passive, coterminous with Type d/D in the verbal group, has one more structural component than its corresponding active form (Section 4.2).

(b) There was a negative correlation between Type c/C and the passive (Section 4.2).

(c) Positive correlation existed between passive and Types b and a/A (Section 4.2). (d) There was a tendency for inanimateness in the subject of passive as compared

with active clauses (Section 4.31). (e) As passive clause complements there was a high proportion of finite and non-

finite clauses (80 % altogether) but fewer nouns and pronouns; more that- than zero that-clauses (Section 4.42).

(f) The frequency of adjuncts was approximately the same for active as for agentless passive clauses (Section 4.51).

(g) Agentful clauses had fewer front-posited adjuncts than agentless and active clauses (Section 4.52).

(h) A difference was noted in the external clause relation of, on the one hand, Classes α-β clauses which, like active, were usually free, and, on the other hand, Class γ clauses, which were mostly syntactically bound (Section 4.7).

(i) Agents were longer than subjects and more than half the animate agents were coordinate (Section 4.35). Since the nominal parts of agents are potential active clause subjects, this may well be an important factor in the choice of voice.

5

PASSIVE CLAUSE CLASSIFICATION: CRITERIA

The previous chapters have dealt with voice categories generally. From this point onwards we shall narrow the scope of our inquiry and concentrate on the passive voice. This does not mean that relationships between the two terms in the voice system will be disregarded, but only that the passive will be our point of departure. We shall first attempt to set up a classification of passive finite verb clauses and then consider some of its uses in diverse texts.

The construction which, up to now, we have broadly called the 'passive' embraces in fact a number of different subtypes. In the final analysis, it would clearly be un-satisfactory to treat together, indiscriminately, all clauses with a Type d element in the verbal group, thus ignoring different kinds of relationship within the clause. We have in fact already in Chapter 4 classified some passive clauses with a division into Classes α, β and γ. The foundation for this rudimentary classification of what may be called 'central passives' according to present/absent and animate/inanimate agent was however entirely subjective and its relevance is so far unproven. The following two chapters will be devoted to an analysis in some depth of a small representative sample in an attempt to establish a comprehensive basis for passive clause classifi-cation. This chapter will discuss the criteria used in preparing the input data for an experiment in numerical taxonomy. The next chapter will describe the coding, the output analysis, and the setting up of a diagnostic key.

5.1 NUMERICAL TAXONOMY

It is possible to state a number of desiderata for the classification of finite passive clauses. We shall mention only four here.1 Such a classification should be, for example:

(a) Objective and empirically verifiable. (b) Consistent with intuition. This is probably a basic requirement. (c) Predictive, i.e. partial or complete knowledge of the relationship of a given

1 For a more detailed discussion of linguistic classification and numerical taxonomy, see Carvell & Svartvik 1966. The principles outlined in that study are basic to the present experiment and will not be repeated here. The techniques are developed for example in Sokal &. Sneath 1963, which also gives further references.

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 73

clause to the classification should enable us to predict with some certainty many other facts about the clause.

(d) Comprehensive, i.e. accounting for all the data, including the unclear, am-biguous and marginal cases.

If a classification of passive clauses is to have all or most of these properties, it seems essential to use multiple criteria. This requires some technique for dealing with a large number of features, or characteristics attributed to each clause, and comparing all the clauses in terms of these features. A rigorous technique falls under the heading of numerical taxonomy. It will not be possible to discuss this technique here apart from saying that it is based on the use of a large number of unweighted criteria. Although we shall treat linguistically different types of criteria slightly differently, we shall not consider some as a priori more important than others. The idea behind this approach is that the important features are exactly those which will correlate highly with other features and hence the groups formed will in fact be those that are con-sistent with our intuition.

5.2 THE TAXONOMIC PASSIVE CORPUS

The classification program (CLASP, see Section 6.2, pp. 113ff.) classified 128 finite passive clauses (called OTUs = operational taxonomic units) in terms of 108 criteria. The size of the corpus used for this experiment, Corpus III (or the Taxonomic Passive Corpus, see Section 1.5, pp. 8 f.), was dictated by the restricted storage of the available computer for which the classification program had been written. It is also worth mentioning that the large number of categories in terms of which each clause was analysed posed also, of course, a practical, entirely human restriction. The 128 clauses, which constituted the sample, were proportionately, and randomly, selected from six groups into which the material had been provisionally ordered on subjective grounds. The reason for this procedure was that where the use of numerical taxonomy was expected to be most rewarding - and most interesting - was in the marginal, indeter-minate, and classificatorily difficult cases. Hence steps had to be taken to ensure that these clauses which are very often members of numerically small classes, should be adequately represented in the sample. The OTUs came from the following sources:

22 OTUs from B-texts (B1 = 11, B2 = 5, B3 = 6); 71 OTUs from J-texts (J1 = 22, J2 = 18, J3 = 7, J4 = 24); 35 OTUs from M-texts (Ml = 10, M2 = 6, M3 = 15, M4 = 4).

The clauses were randomly ordered and numbered from 1 to 128.2 They will here be labelled O T U 1', O T U 2', etc. in addition to their full normal text identification ('J 1.94', 'J2.169', etc.). * The reason for using random ordering was to avoid routine reactions by the informant. Pilot experiments had suggested that, when confronted with a sequence of similar linguistic items, the informant tends to give identical responses to the criteria. Also, informants develop a certain skill as the testing is going on, and this effect would be amplified if similar OTUs were together.

74 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

5.3 THE CRITERIA

The criteria can conveniently be divided into five categories according to the aspect of the clause to which they relate.

THE CLAUSE: 2 0 criteria ( C L 1 -20) , described in Section 5 .4 (pp. 7 8 if.) and Figure 5:1 .

THE PASSIVE VERBAL G R O U P (Clause element JV): 25 criteria ( W 21-45), described in Section 5.5 (pp. 91 ff.) and Figure 5:2.

THE SUBJECT (Clause element S): 24 criteria (S 46-59), described in Section 5.6 (pp. 97 ff.) and Figure 5:3.

THE COMPLEMENT (Clause element C ) : 3 criteria ( C 6 0 - 6 2 ) , described in Section 5 . 7

(p. 101) and Figure 5:4.

THE ADJUNCT (Clause element A): 4 6 criteria ( A 6 3 - 1 0 8 ) , described in Section 5 . 8

(pp. 101 if.) and Figure 5:5.

The criteria are of two main types: overt (or constituent) and covert (or transforma-tional) criteria. The first type refers to features which are actually present in the particular clause under analysis, for example expressed agent (Ag 75). The second type refers to potential features, which the informant attributes to the clause, for example potential agent extension (Ag 92). A male academic staff member gave his reactions as a native informant to the clauses tested by covert criteria. It would of course be preferable to have judgments from a number of informants, but this is difficult to handle in an experiment requiring so many responses (see Section 6.1, pp. 11 Iff.; for some problems connected with informant reaction tests, see for example Quirk & Svartvik 1966). In the diagrams, overt criteria are indicated by solid lines, and covert criteria by dotted lines.

Letters in round brackets (CL, W, S, C, and A, subsuming Ag and Ad) refer to the category to which a given criterion relates. Numbers (1-108) following the letters are identification numbers of the criteria. Numbers preceded by a colon denote alter-native responses to a given criterion. Inventories, 'concocted' items, and initials standing for names in the spoken material, are given in square brackets. Example-centres, i.e. those clauses or parts of clauses which are being illustrated, are given in italics. For the sake of clarity, clauses operating within illustrated passive clauses are usually not italicized. Optional elements are in round brackets. No punctuation is given with transforms.

The presentation of criteria for passive clauses and their illustration will not, how-ever, be restricted to those used in the taxonomic experiment. In the appropriate

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 75

(CL 1:1) Syntactically bound

(CL 1:2) ~ Syntactically bound

(CL 3:1) Sequentially related

(CL3:2) ~ Sequentially related

(CL 5:2) » Free

(CL2:2) Conjunctional (re) Ihat

(CL2:3) Temporal

(CL ?;4) Other conjunctional

(CL4:i) Fixed relator (CL 4:2) Mobile relator

(CL4:3) ± Relator + elliptic S + elliptic Mux

(CL4:4) Relator + elliptic S

(CL 4:3) hollowed by sequentially related clause

(CL 6:1) Parenthetic clauses, iJ-clauses

(CL 6:2) Other clause»

THE CLAUSE ( C V M I A C L 1 - 2 0 )

(CL 7:1) S coreferent with S of contiguous active clause

(CL 7:2) S ~ coreferent with S of contiguous active clause

(CL8:1) S coreferent with S of contiguous passive dame

(CL 8:2) S - coreferent with S of contiguous passive clause

(CL 9:1) S coreferent with Ο or Ρ of contiguous clause

(CL 9:2) S " coreferent with 0 or / of «on· tigoout cUuse

ORDER o r EUMENIS

(CL 10:1) iDvereop ± Hurt

(CL 10:2) ~ Invenioo ± then

{OAXUETYPSIN J ACTIVE TRANSFORM

(CL !1:I) Potential | active transformation '

CHANGE OF AND

TRANSfOR NATION JCTENTtAL

J (CL 12:0) "Extensive

I (CL 12:2)

l(CL 13:1) Intensive

J(CL13:0) *» Intensive I (CL 13:2)

!"(CL 15:1) Subject suppletion

J* {_(CL 15.·0) ~ Subject suppletioe"*"*""

I (CL 15:2)

I TRANSFOFTMANONATR COWNTAINT

(CL 11:2) ~ Potential active transformation

I(CL 1?:1) Potential raw™** j «*«* permutation ΌΤΕΜΠΑΙ.

(CL 19:0) ~ Potential active permntttioD

j(CL 20:1) Potential activetransmutation

TRANSMUTATION _

(CL2Q-.0) -"potential active transmutation

t r . , . . _ - - ——· |(CL 16:1) Sequential constraint (CL 14:1) Constraint — —

J(CL 16:0) ~ Sequential constraint (ciTl·?

_|(CL_17:1) Aspectual constraint J(CL 17:0) ~ Aspectual constraint (CM7:2) " ""

m f(CL18:l) Miscellaneous constraint!

^(CL 18:0) ~ Miscellaneous cmtraiote _ CCLUDQ ~ ~ " ~

(CL 14:0) ~ "constraint

I — I I I

L -(CL 14:2)

mtKnAL

Fig. 5:1. Criteria relating to the clause (CL 1-20).

7 6 PASSIVE C L A U S E C L A S S I F I C A T I O N : CRITERIA

1 ' 1 a! =

i t i l i y t Κ Ύ . ! | ! J i ! ; b i

« is · « ι ί ί ι s ; ι; ° ι — I•=. r Ή m s J R . R j S . S S ? ! ; . , * fc Ϊ fe ί . fel* 7 +

° Liu l-^ j w L i — j ί ί I

1 1

2 1 ] i i

? i ] I ! a

ί ο" a

ο ά ä

t £ £ eq 2* υ 1 I ί ι 8. (Η ί ϊ η a s ä

ί £ ϊ

(Ν £ • — '

α 3

I 73 x>

<Λ a

u

(N <o Μ Λ

PASSIVE CLAUSE CLASSIFICATION: CRITERIA

(S 46:0) Noun (S 47:1) Personal referential (Ag 81:0) (S 47:2) Nonreferential (S 46:1) Pronoun (S 47:3) Relative

FORM-CLASS (Ag 81:1) (S 47:4) Demonstrative (S 46:2) Name (Ag 81:2)

(S 47:0)

THE SUBJECT (CRITERIA S 46-59)

(S 46:-) Zero

(S 4«: 1) Animate (Ag 82:1)

(S48:0) Inanimate (Ag48:0) (S 49:1) Concrete (Ag 83:1) (S 49:0) Abstract (Ag 83:0) (S 50:1) Countable (Ag 84:1) (S 50:0) Mass

(S 51:1) Definite FIN1TUDE (Ag 85:1)

(S 51:0) Indefinite (Ag 85:0)

(S 53:1) First

PERSON (S 53:2) Second

(Ag 84:0) (S 52:0) Zero article (Ag 86:0)

(Ag 84:0)

(S 52:1) Definite article (Ag 86:1)

(Ag 84:0)

(S 52:2) Poss. pron./Genitive (Ag 86:2) (S 52:3) Demonstrative pron. (Ag 86:3) (S 52:4) Wft-pronoun (Ag 86:4) (S 52:5) Indefinite article (Ag 86:5)

(S 53:3) Third

(S 54:1) Singular (S 56:1) Premodified (Ag 89:1) NUMBER (Ag 87:1) PREMODI-

HCATION

(S 56:1) Premodified (Ag 89:1) NUMBER

(S 54:2) Plural (Ag 87:2) (S 55:1) Modified (Ag 88:1)

PREMODI-HCATION

(S 56:0) Unpremodified (Ag 89:0)

MODIFICATION

(S 54:2) Plural (Ag 87:2) (S 55:1) Modified (Ag 88:1)

PREMODI-HCATION

(S 57:1) Postmodified (Ag 90:1)

MODIFICATION (S 55 :0) Unmodif ied POSTMODI-

(S 57:1) Postmodified (Ag 90:1)

(Ag 88:0) n c A ™ N (S 57:0) Unpostmodified (Ag 90:0)

(S 58:1) Coordinated COORDINATION (Ag 91:1)

(S 58:0) Uncoordinated (Ag 91:0)

(S 59) Log (Number of words)

Fig. 5:3. Criteria relating to the subject (S 46-59), and some subcategories of the agent (see Fig. 5:5)

78 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

THE COMPLEMENT (C 60-62)

(C 60:1/1) C-full

Form-class

(C 61:1) to V

(C 61:2) Noun

(C 61:3) Adjective

(C 61:4) Ving

(C 61:0)

(C 62) Log (Number of words)

(C 60:0/0) C-less

Fig. 5:4. Criteria relating to the complement (C 60-2).

places it will be convenient to extend the discussion to other criteria and examples, partly in order to provide fuller illustration or to avoid duplication of the few OTU examples and partly in order to discuss features which for some reason such as non-occurrence or infrequency in the sample were not used in the experiment. Categories which have been introduced earlier, in Chapters 2-3, need only be exemplified here, whereas new categories will be discussed in some detail.

5.4 CRITERIA RELATING TO THE CLAUSE (Criteria CL 1-20, see Fig. 5:1)

5.41 External clause relation

The categories have been presented in Section 3.5 (pp. 31 ff.). (CL 1) Syntactically bound clauses [52 cases] comprised the following four sub-

classes. (CL 2:1) Relative clauses [23 cases].

... in order to obtain information concerning the relative rates at which various reaction sites in various aromatic nuclei are attacked by free radicals.

(OTU 25, J4.27)

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 79

12

1 ! = i 3

! ! δ

11 I» Λ IHM iS <

r j L-j-

L g

• c

Ii'

i « ig

i ' 's-

' s I u a ! 6 ä 1 'i ι·® .1 < I e s Ο ! 2 i i i i s s Js> 'S

I I | ! < ι I

! s I ο ! ε I g l ü I c |3 I I |3

. I J L I i i : • ο ι : i® Ig; I» jg» ί .<· Ι<Ί< I"

i l l j .2

j j u lU *c <•3

3 3 =5· ce υ •s

.a «>

S S .2 <

•β u ϊ1* S S C -Ό Ö I au ü g « t _ j S £ l s

g ^ fj gf s

έ·° =

"5. o. £ t

» Ü

~ Cl

ause

C U, -·

Fron

t

ϊ - M

id

β 1U c 3 S - . I

Τ- ο Is ο τ τ ο" c 3 S - . I ο ο ο ο 8 ε ο <|, J1 Ό < <

Τ5 < •ο < J •α < < <

(rt W i r e

Ö f

VI VI

, s &L& &

4 - =! UJ C Ο

? Τ1?-®1 Τ T s T

i < S i < * ! < l i s • V " Τ -

—

c tf; X · j j:

.<£ S-* ι

g § .2 i

Γ "Ι τ τ ι ι

J ι ^ is χ τ » I 8.

f ?

<3

I i

ι

= t ; - i=j , · V

isS|<. 'LsJ j i ' l i ' j s >·φζ > i s s j s . ι I Τ ! Τ ι I ι j. I

O C > Ζ o «

« S ä i i i i l l j s s· *

f i <Ö oi fc

i-ä I f

3 Η

e f Ζ >o ä <

8 0 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

(CL 2:2) Conjunctional that- and so that-clauses [12 cases].

... it is a necessary corollary of this scheme that these radicals are formed only in negligible amount. (OTU 52, J4.38)

Nevertheless, the recent work of Huisgen and Nakaten (1954) and Davies, Hey and Williams (1956) has shown that nitrodiphenyls can be obtained from this reaction, although in rather low yield. (OTU 71, J4.32)

(CL 2:3) Temporal clauses [8 cases].

... it follows that the unit cell found by X-rays is indeed the true unit cell when the hydrogen atoms also are considered. (OTU 56, J2.163)

(CL 2:4) Other conjunctional clauses [9 cases].

... benzoic and anthranilic acids do not take part in the Gomberg reaction unless they are first esterified, i.e. convertedfrom molecules which are ionised in aqueous alkaline media, to typical organic esters. (OTU 93, J4.29)

(CL 3) Sequentially related clauses [55 cases] comprised the following five sub-classes.

(CL 4:1) Clauses with fixed relator, such as and or but [17 cases].

The side walls are strong but the roof is composed only of a tough membranous fibro-cartilage. (OTU 44, J1.87)

(CL 4:2) Clauses with mobile relator, such as however or then [11 cases].

However, values of (-0·40±0·02) χ 1012 cm. and (0·57±0·08) χ 1012 cm. were deduced for b for hydrogen and deuterium respectively. (OTU 45, J2.165)

(CL 4:3) Clauses with relator -f- elliptic subject + elliptic auxiliary [5 cases].

... he scuttled sideways ... into the old linen cupboard, where he was cornered, wrapped in a blanket, and carried effortlessly by Trained Pig Percy-Scroop-Beauchamp to the bathroom. (OTU 92, M3.209)

thus the forebrain is connected with smell, midbrain with sight, hindbrain with acoustico-lateral and taste-bud systems. (OTU 112, J1.101)

(CL 4:4) Clauses with relator + elliptic subject [7 cases].

'No, I don't think we want him,' said Martin, but was disconcerted when his mother said: Oh , I am so glad.' (OTU 114, M3.204)

(CL 4:5) Clauses followed by sequentially related clauses [16 cases].

Martins rocking horse was discovered in an attic and installed in the day nursery, (OTU 98, M3.210)

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 81

(CL 5) Free clauses [28 cases] comprised the following two subclasses.

(CL 6:1) Fully free clauses, i.e. with no external relationship of the types recognized here in CL 1-4 [24 cases]. (Other external clause relationship, such as that denoted by determiners, will be analysed in connection with fmitude in the subject.)

In this equation V is the volume of the crystal which is being radiated. The value of ε usedfor calculating the integrated reflection pß is first correctedfor absorption. [New paragraph] The structure factors determined in this way were ...

(OTU 43, J2.171)

(CL 6:2) Clauses with proleptic it, or parenthetic clauses [4 cases].

It would have been absurdly affected to expect a child of three to need lavatory paper or towels, and in any case it would have embarrassed Lady Foxglove to have to ask for them. (OTU 23, M3.200)

But it isn't that at all; he doesn't lack quality, I feel convinced. (OTU 72, M2.105)

One of the suggestive results of the voice data processing (reported in the previous chapter) was the high proportion of passive clauses with close external relation, particularly relative clauses. This result was corroborated in the present analysis by the inventory totals: 52 syntactically bound, 55 sequentially related, and 28 free clau-ses. These numbers indicate that there were some overlapping cases, as in the fol-lowing example with a bound conjunctional that-clause followed by a sequentially related clause:

... the troops ... were so affected by the sights and the smells and the testimony to human depravity that many of them were taken ill and had to be sent home on leave. (OTU 67, M3.209)

5.42 External subject relation

The relationship of a passive clause to a contiguous clause may be denoted by other categories than those mentioned in CL 1-6. There may, for instance, be identity of subject, as in the following example with an active equative clause followed by a sequentially related passive clause:

'He's only six; and not allowed to play with rough children' (OTU 60, Μ 1.26)

Following up the suggestion of the possibility of sequential clause relation as a factor which might influence the choice of the passive voice, the subject-relation of the passive clauses to clause elements of contiguous clauses was analysed in terms of coreference. 'Contiguous' was taken to mean 'immediately preceding or immediately following'. 'Coreference' implied either complete exponential identity, as in the exam-

82 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

pie above (OTU 60, which was of course also entered as a case of CL 4:3), or simple coreference relation, as for example between pronouns and their antecedents. The following three types of external subject relation were recognized.

(CL 7) The subject of the passive clause was coreferent with the subject of a con-tiguous active clause [37 cases].

/they'd been at school # . /they'd been a /they'd been - a at the :depot # /they'd been: trained # - /they hadn't "really # . / ' known and'had life # as it /is to be had # (OTU 108, B2.23) This [a well-developed heart] lies behind the gills and can be considered as a portion of the sub-intestinal vessel, folded into an S-shape and divided into three chambers. (OTU 85, J1.91)

(CL 8) The subject of the passive clause was coreferent with the subject of a contiguous passive clause [19 cases].

Lead roofs could no longer be patched and would have to be renewed completely. (E1.50)

It [the photolytic decomposition of phenylazotriphenylmethane in benzene] has been investigated by Horner and Naumann (1954) and Huisgen and Nakaten (1954), and was found to involve a primary dissociation into phenyl and triphenylmethyl radicals and nitrogen, in the manner indicated in equation (8).

(OTU 69, J4.34)

(CL 9) The subject of the passive clause was coreferent with the object (including prepositional object) or predicative of a contiguous clause, active or passive [18 cases].

This is now being used instead of grain which is stored on British farms waiting for buyers. (E2.29)

5.43 Internal clause relation

(For categories, see Section 3.4, p. 31.) (CL 10) Inversion of S and W elements, with or without introductory there [3

cases]: [- in I fact in nineteen sixty # there were only jseven # . j hanged # - and as /long ago as nineteen thirty # there were only /six -hanged # ]

(OTU 77, B3.4; [m=low, narrow])3

3 For the present purpose, it has been necessary to depart, in some cases, from the typographic conventions given in Crystal & Quirk 1964. Stretches with modulation features, for example, have been denoted by bracketing: [[allegro]], etc.

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 8 3

iri'lcluded in this class of crimes# . "jare as müch# . l_j instances of ma:licious wounding of :one thug or 'young man byj < an" lother

(OTU 79, B3.15, [low,narrow; piano] < allegro)

5.44 Active transformation potential [107 cases]

(CL 11) A closer analysis of these 107 clauses out of the total 128 proved, however, that unrestricted passive-active transformation was possible in only 18 cases. 'Un-restricted' transformation implies transformations without constraints or with con-straints other than those recognized here and discussed later in this section. Hence, the term 'free active transformation is simply the reversal of the passive transforma-tion of transformational-generative grammars: 'If Sx is a grammatical sentence of the form NP1 — Aux — V — NP2, then the corresponding string of the form NP2

— Aux + be + en — V — by + NPX is also a grammatical sentence' (Chomsky 1957, 43; cf. Section 1.1, p. 1). In the present terminology, the active transformation formula may be stated as

N1 W ρ N2 -» N2 VNj,

The Ν before the verb is the subject in either voice, W is the passive verbal group, V the active verbal group, and ρ is the agent-preposition, usually by.

A distinction was made between different clause types in the active transforms (CL 12) Extensive clause type [104 cases] and (CL 13) Intensive clause type [15 cases]. Normally the clause types in the active

transforms are those corresponding to the passives as outlined in Section 3.8. Passive transitive (ext-0) active transitive (EXT-C), for example:

The relevant experimental data are summarised in the following conclusions. (OTU 87, J4.37)

-* [we] summarise the relevant experimental data in the following conclusions

Passive ditransitive (ext-C) -*• active ditransitive (EXT-CC), for example:

if I"we had been offered the choice of being killed or not that möment # we would have [/"got down # ] as /low as we could # into the /nearest shill hole #

(OTU 103, B2.19; [ staccato])

-* [someone] had offered us the choice of being killed or not

Passive factitive (int-C) active factitive (INT-CC), for example:

'[this bookshop] does cater for what's called the cultivated reading public - and for specialists.' (OTU 106, Μ 1.20) -* what [they] call the cultivated reading public

84 PASSIVE CLAUSE CLASSIFICATION '. CRITERIA

Some 'adjectival passives' as, for example, in

She was astonished, and said so. (OTU 113, Μ 1.23)

may take two active transforms, one extensive and one intensive:

-> [the dog's behaviour] astonished her / made her astonished

It is a property of this class of 'emotive verbs' that their Ved forms may operate as C2 in active clauses of type INT-CC in addition to operating as Via active clauses of type EXT-C. In the following example, transforms with two INT-CC types are given, the latter of which may be analysed as a reduction of the former equative clause.

but he was ... embarrassed by the implication of sentiment in his remark. (M5.25) -»· the implication embarrassed him -* the implication made him feel embarrassed -» the implication made him embarrassed

Although Ved is most commonly the exponent of C2 in active intensive types, V and Adj may also be considered in some cases:

I be/lieve in regard to :capital piinishment # [which /is of course a . a : the Ideath · sentence # - for /murder # ] - < that it /should be applied # to jail

->•... that [the courts] should apply it / make it apply / make it applicable to all

(CL 14) Transformational constraints. The analysis of the 107 clauses with potential active transformation (see CL 11) proved that only one in six had 'un-restricted' active transformation potential: as many as 88 clauses had some form of transformational constraint, which was specified under four headings: (CL 15) subject suppletion, (CL 16) sequential, (CL 17) aspectual, and (CL 18) miscellaneous constraints.

(CL 15) Subject suppletion. By far the most common constraint [75 cases] was the necessity, in the case of agentless passives, of supplying active transforms with subjects. Suitable active subjects were inferred from the context. Such 'concocted' elements are given in square brackets:

The baby was installed in the old nursery wing at Bidcombe, and a very expensive nanny was hired from London to look after it. (OTU 55, M3.198)

[Lady F.] hired a very expensive nanny from London to look after it

On the dorsal side is a single nasal opening, and behind this there is a gap in the pigment layers of the skin through which the third or pineal eye can be seen as a yellow spot. (OTU 86, J1.84)

murderers # (OTU 17, B3.54; [low]; <allegro>)

murderers

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 85

-*• through which [one] can see the third or pineal eye as a yellow spot

(CL 16) Sequential constraints [14 cases] were due to external clause relation as stated in CL 3 and CL 4. In the following example with elliptic passive subject the active transform requires the suppletion of an object. (An additional constraint not considered in the experiment is the unsatisfactory stylistic imbalance which is caused by the suppletion of a light end-placed pronoun; see CL 18 below.)

Isokinetic relationships of this kind are not uncommon in organic chemistry, especially in free radical reactions, and have been discussed recently by Leffler (1955). (OTU 42, J4.33) -* isokinetic relationships of this kind are not uncommon ..., and Leffler (1955) has recently discussed [them]

If the passive clause is followed by a sequentially related clause with zero subject, some suppletion will be necessary in the latter clause. The active transform of the example in CL 4:5 might be as follows, with two active clauses,

[they] discovered Martin s rocking horse in an attic and installed [it] in the day nursery

or, with an active clause followed by a passive clause,

... and [it] was installed in the day nursery

(CL 17) Aspectual constraint. In eight cases active transforms were only 'pos-sible', or at least 'preferable', with a change of aspect or mode in the verbal group. In the first example the change is from Type d to b, in the second from D to Β (see Section 2.2, pp. 10f.):

Madeleine walked forward and stood at a little distance, near enough to see the shape at Dinah's foot. 'Nearly dead,' said Dinah slowly. Ί think its back's broken: (OTU 53, Ml.27) -»· I think [the stroke] has broken (*breaks) its back

... Herring wept quietly and tearlessly all day long after her husband had told her that he had sent the child to an adoption society. She was alone in a dark world, and the only thing she had ever loved was removed from her.

(OTU 107, M3.197) -* [her husband] had removed from her the only thing she had ever loved

Note on aspectual constraint: The illustrated passive type looms large in most handbooks. It is in fact a moot point of classification whether it should be admitted at all as 'passive'.

Jespersen makes a distinction between two classes of verbs: CONCLUSIVE VERBS whose 'action is either confined to one single moment, e.g. catch, surprise, awake, leave, end, kill, or implies a final aim, e.g. make, bring about, adorn, construct, beat'; and NON-CONCLUSIVE VERBS 'denoting feelings, states of mind, etc.: the activity, if any such is implied, is not begun in or-

86 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

der to be finished. As examples we may mention love, hate, praise, blame, see, hear' (Jespersen, 1909-49, 4.92-3). This dichotomy underlies his divisions of passives into two types: 'With non-conclusive verbs... the participle is a predicative of being... and we may therefore speak of a PASSIVE OF BEING' (He is admired). 'With conclusive verbs the time-relation is not so simple: sometimes the participle is a predicative of being, sometimes one of becoming, and therefore we sometimes have a PASSIVE OF BEING, sometimes a PASSIVE OF BECOMING'; "... in such a sentence as His bills are paid sometimes the element of present represented by are, sometimes the element of perfect implied in the participle is predominant. The sentence thus may mean two things, either the present action as in His bills are paid regularly every month = He pays, or the (present) result of a past action as in His bills are paid, so he owes nothing now = He has paid' (Jespersen, op. cit., 98). Other terms that have been used include STATAL or STATIC PASSIVE and ACTIONAL or KINETIC PASSIVE (Curme 1931, 443 ff.); PERFECTIVE and IMPERFECTIVE PASSIVE (Turner 1962, 181).

A view differing from that of Jespersen and Curme is held by many grammarians, among them Kruisinga:4 the term "passive" 'is only applied to the verbal group when it expresses an occurrence or an action. Thus, we have a group-passive in The book is sold for 5 s., but when a second-hand bookseller informs a would-be purchaser that a book in the catalogue is sold, the group expresses a state or condition and is not a passive. It is a curious and naturally accidental result of this that the same verbal group comes to express two opposite, and mutually exclusive, meanings'.

With the comprehensive approach adopted in this study, the term 'passive' embraces both the actional and statal types.

(CL 18) Miscellaneous constraints [15 cases]. Several constraints which were either too infrequent or unclassifiable as a group come under this heading,

(a) Introductory there and proleptic it are dropped in the active.

there's /also included in this class # . a /lot of !other -things # . such as] (pro\ curing abörtion # ) -and . enjdangering a ship at Isea #

(OTU 115, B3.16; [allegro]; <Iargo; narrow and rhythmic» [the law] also includes in this class a lot of other things

It is seen that the hydrogen atom may be removed by one of the two processes (17d) or (17e), that is, either by a radical R'· or by the solvent, and that for each separate case the overall kinetic result is different. (OTU 7, J4.38) -* [we] see that the hydrogen atom may be removed

(b) Compounds with the negative uw-prefix (see Section 2.73, pp. 21 f.) may sometimes take active transformation if the bound negation is replaced by a free negation.

The details of the composition of the nerves of lampreys are still unknown, but there are hints of considerable deviations from this plan. (OTU 95, J 1.97) ->· [we] still do [not] know the details of the composition

(c) In certain passive factitive clauses a copula is optional as part of C but impossible or unlikely in the active transform: 4 Kruisinga 1925-31, 2:1.305. Cf. also Poutsma 1926-9, 2:2.98ff., Zandvoort 1960, 48ff., Francis 1958, 335 ff., Fries 1940, 189, Scheurweghs 1959, 156, 416.

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 87

Dusk had perceptibly deepened before Dinah was seen to be marching briskly up the slope again, the bier across her shoulder. (OTU 109, M1.29) -> before [they] saw Dinah (*to be) marching

(d) 'Adjectival passives', which were mentioned in CL 13, share many properties with active equative clauses with adjectives as C. They can for example take modifiers like very. Notice, however, that the active transitive transforms require 'verbal' modifiers.

I'm /very surprised actually # at /this . quotation of Riissell's # (B2.10) this quotation of Russell's surprises me very [much]

(e) As was noted in connection with OTU 42 illustrating CL 16 above, stylistic imbalance was not registered as a constraint in the present experiment. The following example, however, shows that when the weight of clause elements becomes too dis-proportionate the acceptability of the transform can be drastically reduced.

This has been neatly demonstrated by Grieve and Hey (1938), who showed that benzoic and anthranilic acids do not take part in the Gomberg reaction unless they are first esterified, i.e. converted from molecules which are ionised in aqueous alkaline media, to typical organic esters. (OTU 101, J4.29) -*• Grieve and Hey (1938), who showed that benzoic and anthranilic acids do not take part in the Gomberg reaction unless they are first esterified, i.e. con-verted from molecules which are ionised in aqueous alkaline media, to typical organic esters, have neatly demonstrated this.

(f) Other types of constraint can be seen in the following examples (the first trans-form is, however, not strictly synonymous with its passive original):

... she must have been motivated more by the desire for notoriety than Christian charity in adopting, at her age, a baby from an unknown home and giving it her name. (OTU 18, M3.197) -> the desire for notoriety [rather] than Christian charity must have motivated her

The variations in energy of activation are compensated by the variations in entropy of activation, ... (OTU 61, J4.33)

the variations in entropy of activation ... compensate [for] the variations in energy of activation

These 'follicles of Langerhans' were, appropriately enough, first seen by the discoverer of the islets in higher forms, ... (OTU 125, J1.90) -*• the discoverer of the islets in higher forms, appropriately enough, first saw these 'follicles of Langerhans'

The following two transforms, however, bring out the sentence focus better, and are generally more acceptable:

88 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

-»• the discoverer of the islets in higher forms was, appropriately enough, the first to see these 'follicles of Langerhans'

it was, appropriately enough, the discoverer of the islets in higher forms who first saw these 'follicles of Langerhans'

Voice transformation entails changes not only from passive to active verbal group but also in the function and position of the nominal clause elements, for example X is followed by Υ -* Yfollows X which may be represented by the following diagram (cf. Jespersen 1924, 164):

The preposition is dropped when the postprepositional nominal part of the agent becomes the subject in the active. In stating the transformational properties it is generally more convenient to mark the nominal elements by subscripted Ns, the first of which is the subject in either voice (see Sections 2.32, pp. 13f., and 3.8, pp. 36 f.):

In an attempt to describe relations obtaining among clause elements, another test similar to transformation suggests itself: a change of voice in the verbal group without inversion of the nominal elements; the result still being a grammatical clause, as in X is followed by Y -+ Xfollows Y. In order to distinguish this test from transformation we may call it PERMUTATION. In voice permutation the passive subject remains subject in the active clause, but the agent changes, not, as in voice transformation, to active subject but to object, with deletion of the preposition.

Indeed, the hydrogen atom is probably never actually free, but may be removed by the aroyloxy radical from the α-complex formed between Ar'· and ArH.

-* the hydrogen atom ... may remove the aroyloxy radical

As deduced from the X-ray results the nitrogen atom, taken as the origin, is surrounded by eight chlorine atoms in positions such as \ £), giving a CsCl

5.45 Permutation and transmutation

(OTU 59, J4.39)

structure. -» the nitrogen atom ... surrounds eight chlorine atoms

(OTU 19, J2.174)

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 89

Those two permutations are grammatical (but perhaps examples of 'ontological nonsense'; see Bolinger 1961b, 371), whereas the following is non-grammatical:

The first investigations of the chloride by neutron diffraction were made by Goldschmidt and Hurst (1951) ... (OTU 102, J2.174)

*the first investigations ... made Goldschmidt and Hurst

In agentless passive clauses the change under permutation is restricted to the verbal group:

[it's a consolation for the survivors] /rather than a igenuine consol-ation # for /those who were -.killed # (OTU 81, B2.20) -* for those who killed

He had a half-formed apprehension that Percy might pull his head off or some-thing if he were bitten too hard, ... (OTU 22, M3.209) -*• if he bit too hard

Permutation is however different from transformation in one respect other than absence of inversion: it involves semantic change. This may be expressed by saying that in transformation the grammatical subject changes but the logical subject remains the same, whereas in permutation the grammatical subject is the same and the logical subject changes.5 When some verbs are permuted, however, the result is a grammatical sentence with no appreciable change of meaning, i.e. the logical subject is felt to be the same, or nearly the same, in the active as in the passive. We shall call this test TRANSMUTATION.

Her less fashionable friends piously closed their ears to the malicious rumours which were being spread abroad, ... (OTU 16, M3.201) -> which were spreading abroad

yet/these countries # . [/as we're !both a:gr£ed#J - /have#. <9 an in /some cases an :equally serious or :even> "Imore serious crime situätion# than /we do in this !!country# (OTU 35, B3.46; [low], <low, confident, sure)) -» as we both agree

There is no trace of any paired fins, but the tail bears a median fin, which is expanded in front as a dorsal fin. (OTU 122, J 1.84) -» which expands in front as a dorsal fin

... the curves do not again separate above the critical point, but are continued as a single curve. (OTU 21, J3.31) ->• but continue as a single curve.

5 'Grammatical subject' is that element in the clause with which the finite verb is in grammatical concord (overt and covert) (see Section 3.1, pp. 25 f.). 'Logical subject' may be defined as 'the person or thing from which the predication is considered to originate' (Poutsma 1926-9, 2:2.7).

90 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

'Permutation' and 'transmutation' are introduced here as specific technical terms for describing certain voice relations. If the analysis is carried farther, such criteria help to distinguish subtypes with different voice relations, as in the following sentences (cf. Olsson 1961, 180, and Erades 1950):

(a) He worried. = He was/felt very worried (about it). (b) The door closed. = The door was closed (by someone). (c) The house is building. (But *The house builds.) = The house is being built. (d) The play reads well. (But *The play reads.) φ The play is read well.

The 'transmuted' sentence, as in The war began, has been called 'notional passive',* and its subject 'neutral' since it is 'involved but not responsible' as in They began the war. Long takes the extreme view that, since they 'are often used to predicate events and states of affairs rather than actions', all active forms are best called 'common', in contrast to 'passive' (1961, 113, 510). If one wants to let the semantic distinction be reflected in the names of the categories, it seems that a more consistent classification would be: 'active' for He began the meeting; 'common' for The meeting began; 'passive' for The meeting was begun.

Summarising, the three tests we have just described have the following typical features.

TRANSFORMATION:

[different grammatical S, same logical S

PERMUTATION:

[same grammatical 5","] different logical S J

TRANSMUTATION: same grammatical S, same logical S

] N1 W ρ JV2 He was bitten by the dog. N2V N1 -> The dog bit him.

N1 W ρ Ν2 He was bitten by the dog. -* N1V Ν2 ->• He bit the dog.

Nx W ρ Ν2 He was worried about the dog. -* Nx V ρ Ν2 -* He worried about the dog. Ν ι W ρ Ν 2 He was opposed to the idea. -» Νχν Ν2 He opposed the idea.

(CL 11) Potential active transformation [106 cases] seems to be by far the most important test for denoting grammatical relationships between passive and active clauses.

(CL 19) Potential active permutation [19 cases] draws attention to such properties as gender of clause elements (cf. Chomsky 1957, 78).

(CL 20) Potential active transmutation [19 cases] highlights the uses of verbs and the relationship of clause types.

* Cf. Curme 1931, 440, Jespersen 1924, 164ff.

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 91

5.5 CRITERIA RELATING TO THE PASSIVE VERBAL GROUP (Criteria W 21-45; see Figure 5:2, p. 76)

Two aspects of the verbal group were considered: auxiliaries and lexical verbs.

5.51 The auxiliary, tenses and types

The tense of the first element of the verbal group was analysed in terms of the follow-ing categories (cf. Section 2.34, pp. 16f.):

(W 21:1) Present tense [75 cases].

It appears that the energy relationships involved are so balanced that many reactions have a definite and observable probability dependent on the properties of the radical and the substrate. (OTU 105, J4.30)

(W 21:0) Past tense [51 cases].

She thought it had had a number of owners; luckily the building itself had never been touched·, but everybody had done something to the garden and made a mess of it. (OTU 4, M1.17)

(W 21:-) Non-tense-marked [2 cases].

<5i a a [/[world] !wlde#] . {/increase in crime# - /notably :jüvenile crime#} (must be /brought into this :matter as !well#> (OTU 76, B3.2; [forte]; {low}; <crescendo & allegro»

(W 22) Auxiliary Type a/A [21 cases] was further divided into three subclasses.

(W 23:l/0)7 Fully tense-marked [4 cases].

He was covered in snow and had virtually to be carried to the fireplace and stripped of his outer clothing. (OTU 12, M2.101)

(W 23:0/0) Semi-tense-marked [15 cases].

This difficulty can be avoided in two ways: (OTU 48, J3.30)

(W 23:0/1) Non-tense-marked [2 cases]. (See W 21 for illustration.)

(W 24) Auxiliary Type b/B [17 cases].

there's a :rather "igreäter tendency for people to commit crimes a:gäin#] -/äfter they 'dbeenfl0gged# than /when they hädn ' t# (OTU 116, B3.19; [allegro])

(W 25) Auxiliary Type c/C [2 cases]. The infrequency of this type with the passive

' This criterion was technically treated as two tests with the three alternatives marked 'I/O, 0/0,0/1, respectively.

92 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

(noted for the Minor Passive Corpus, see Section 4.2) was corroborated by this inventory.

In this equation V is the volume of the crystal which is being radiated. (OTU 82, J2.171)

Auxiliary Type d/D. All clauses had, by definition, type d/D auxiliaries; they will be further analysed here by means of overt and covert criteria.

Note on auxiliary type d/D (the passive auxiliary): There is general agreement that be is a passive auxiliary,8 but the status is less clear for other auxiliaries which occur with past participles.

'Get and become are now increasingly common as auxiliaries for the passive of becoming; with some verbs the distinction between them and be is particularly useful' (was/got married) ·, get 'has (or had) a decidedly more colloquial colouring'; grow 'in the same function as auxiliary of the passive of becoming is comparatively rare, apart from combinations like grow accustomed'; stand is 'particularly frequent in judicial expressions like stand condemned'; rest, as in rest assured 'sometimes is approximately a mere auxiliary of the passive' (Jespersen 1909-49, 4.108 ff.).

'When connected with a past participle to get is apt to lose its character of a copula and assume a function which differs little from that of to be as an auxiliary of the passive voice. The altered function of to get, of course, postulates a change in the grammatical function of the participle, which, from being mainly adjectival, becomes almost purely verbal ... Some such modification of function may also be traced in to become and to grow, but with these verbs the change is less pronounced, the participle being more retentive of its adjectival features, and the combination suggesting a gradual process'; 'certain verbs that more or less partake of the character of copulas, such as to feel, to stand, may be used to advantage when no passiveness is intended ... Also copulas of the third kind, i.e. such as are used to express the changing of a state into another ..., are not unfrequently combined with a past participle to form a construction that bears a strong resemblance to the passive voice. Naturally the participle is not so entirely devoid of adjectival characteristics in these com-binations as it is in a pure passive voice. Instances with to get are quite common, especially in colloquial style; to become and to grow being far less frequent in this function';'... the combinations with to get ... are hardly distinguishable from a purely passive construction with to be, [whereas those containing to become, and to grow] only vaguely express passive-ness: in other words to get, when connected with a past participle, has lost almost entirely its power of indicating incipient action and may, accordingly, be called an auxiliary of the passive voice, but to become and to grow show this force in only a slightly weakened form' (Poutsma 1926-9, 1.30; 2:2.99).

Get 'expresses the getting into a state or situation denoted by the participle; in other words, it has a mutative meaning, which distinguishes it from the ordinary passive.' 'The difference between to get and to become as auxiliaries of the passive may be expressed by the terms PERFECTIVE and DURATIVE' (Zandvoort 1960, 57).

Get 'in spite of constant efforts of grammarians and teachers of English, has not been abolished, but rather seems to be increasing in favor in the common idiom' (Frary 1929, 73).

Get 'occurs more frequently in vulgar English' (Fries 1940, 193).

8 But, as was pointed out in 'Note on aspectual constraint' under CL17 above (pp. 85f.), not even all combinations of be with past participles are generally agreed upon as 'passive'.

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 93

The ^/-passive 'seems to be increasing in frequency, though grammarians are at present not agreed as to its status' (Francis 1958, 335).

Hatcher argues against Jespersen who 'lists become (and even grow, stand, rest) along with get as a passive auxiliary - without, of course, attempting to define the restrictions on its use. These restrictions are such ... as absolutely to exclude it from consideration as a passive auxiliary: for become may never be used to refer to any passive action performed by a human agent (unless, of course, the participle is such as to invite an adjectival interpretation: this tradition became accepted but not the present became accepted; ... unlike get, become is found only in border-line cases' (got accustomed, drunk, lost, engaged). (Hatcher 1949,442f.).

Svartengren coins the term ACTIONAL-DURATTVE for the become-passive, which he considers as a combination of the STATAL PASSIVE and the ACTIONAL PASSIVE since it 'describes the action, i.e. the transformation and its lasting result' (During the first half of this century the tribe had become christianized). The phrase had become christianized 'must undoubtedly be regarded as a type of the passive, but by using the term ACTIONAL-DURATTVE, we admit that there may be cases when ... the durative aspect so far outweighs the actional-verbal that the past participle acquires an adjectival character ... and the expression can no longer be regarded as a passive' (Svartengren 1948, 275).

The confused picture presented by these extracts is accounted for, largely, by lack of formal criteria. Such concepts as 'pure passive voice' and 'adjectival interpretation' are hard to agree on. Adhering to our intention to take a comprehensive view, we shall initially admit as passive auxiliaries all verbs (other than Type bIB auxiliary have, see Section 2.4, p. 17) that combine with past participles. Since only be is frequent in this function, little can be said about the others on the basis of Corpus III; but it is necessary to state at once how they will be handled.

Although we have made a distinction between auxiliaries in verbal groups (Types a, b, c, and d) on the one hand and copulas in equative clauses on the other, it is essential in setting up a classification of auxiliaries to see these two kinds of verbs in relation to each other. The division proposed here for passive auxiliaries will there-fore apply also to copulas.

The basic division is into LEXICALLY UNMARKED AUXILIARIES (be) and LEXICALLY

MARKED AUXILIARIES (auxiliaries other than be). The latter have the two subclasses MUTATIVE and NONMUTATIVE. These terms have previously been used for the passive (see Strang 1962, 146), but they have been extended here to cover all lexically marked auxiliaries and copulas that are aspectually related. Be is formally unmarked but may have either mutative or nonmutative value. The following list will illustrate the two subclasses:

MUTATIVE

become (established) come (unstuck)

NONMUTATIVE

get (lost) go (sour) grow (tired) fall (sick) run (wild)

appear (delighted) feel (annoyed) lie (scattered) look (offended) remain (divided) rest (assured) seem (determined)

94 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

turn (sour) smell (sweet) etc. sound (surprised)

stand (transfixed) taste (better) etc.

(W 26) Lexically unmarked auxiliary [120 cases] and lexically marked auxiliary [8 cases].

(W 27:1) Mutative [3 cases of become].

The tubules become longer and twisted after hatching and may perhaps serve for salt-reabsorption. (OTU 1, J 1.95)

(W 27:0) Nonmutative [5 cases] (the auxiliaries were feel, look, remain, seem).

Then suddenly, a frown drew together the strong brows, but his eyes remained tight shut. (OTU 20, M4.47)

Ί know what you're thinking, Doctor.' The florid face looked wounded. (OTU 9, M4.56)

But still the stones seemed rocked, the unsterile mounds, reimpregnated, exhaled dust's fever; a breath, impure, of earthbound anguish. (OTU 126, Ml.30)

The difficulty of classifying clauses with Type dJD verbal group is largely due to the fact that be, its most common auxiliary, has the two values of passive clause auxiliary and equative clause copula, and the distinction between them is by no means clear-cut. In equative or near-equative clauses, be often commutes with other, lexically marked copulas, for example He / was / seemed / sounded / felt / looked happy and interested. In an attempt to plot some of the values of Type djD auxiliaries in the corpus, commutation tests were used with seem to be, seem, and feel. Acceptance varied considerably for the three tests as can be seen in the following table:

Auxiliary Yes Query No replacement

(W 28) seem to be 88 7 33 (W 29) seem 45 6 77 (W 30) feel 15 2 111

if you /look at our !'present royal fämily# - am you can /see the thing which is [seems to be] known in . genetics# as the /Habsburg :lip# (OTU 80, B2.50)

This change is [seems to be, seems] especially marked in the larva and is produced by variation in the amount of a pituitary secretion (p. 107). (OTU 84, J1.84)

'No, I don't think we want him,' said Martin, but was [seemed to be, seemed, felt] disconcerted when his mother said: Oh , I am so glad'.

(OTU 114, M3.204)

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 95

his/property was [*seemed to be, * seemed, *felt] ordered !not to be confiscated# but in /fact to be Isold # (OTU 65, Β 1.44)

5.52 The lexical verb

The morphological differences between the types represented by undressed, unexpected, and unskilled have been dealt with in Section 2.73-4 (pp. 21 ff.). In the taxonomic experiment a distinction was made between compounds and noncompounds and between verbal and nonverbal (i.e. nominal) bases.

(W 31) Compound [4 cases] and noncompound [124 cases]. The following is an example of a compound past participle:

These letters had been so pompous and ill-considered in tone, and so unrelated to effective action, that it was impossible for Emma to think of them without either annoyance or amusement. (OTU 40, M2.92)

(W32) Verbal base [126 cases] and nonverbal base [2 cases]. The following had nonverbal bases:

The skin is mmy-layered ... (OTU 14, J 1.84)

The muscle-fibres run longitudinally and they are striped, but of a somewhat peculiar fenestrated type. (OTU 91, J1.85)

Nonverbal bases were included in the sample because of their syntactic similarity with passives. What is said in this presentation referring to the verbal group does not, strictly speaking, include nonverbal bases, since nonverbal bases cannot by definition enter into the verbal group. For the sake of simplicity, however, they will be included here with verbal bases. There are verbs corresponding to the participles layered and striped, but the -ed forms here seem to have closer relation with nominal bases; compare

be many-layered and have many layers be striped and have stripes.

(W 33) Phrasal verbs. The lexical verbs were entered according to the classes set up in Section 2.71 (pp. 19f.). There were no prepositional verbs in this corpus, but three phrasal verbs, one of which was the following:

/I think [öi]. that Mr Dean :does -make the -army a-ttractive# or . there was a /possibility he might# . a /for .[ the people who are going to be :called ύρ#\

(OTU 128, B1.8; [allegro & piano])

(W 34) Qualification [7 cases]. Ved premodified by 'qualifiers', i.e., typically, ad-jective modifiers {highly, so, more, especially, etc.; cf. Roberts 1955):

96 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

It is recorded that when, the invading armies of the Allies reached Belsen, the troops who witnessed the horrors of the camp for the first time were so affected by the sights and the smells and the testimony to human depravity that many of them were taken ill and had to be sent home on leave. (OTU 50, M3.209)

(W 35) Coordination with an adjective and/or a prepositional phrase [5 cases].

Flat fields unearthly green, dotted with grazing cattle, stretched into the distance on the side they walked on; the other side was broken, hillocky, and patched with unkempt plantations of smouldering beech and hazel.

(OTUs 41 and 124, M1.18)

(W 36) Potential coordination with an adjective [29 'yes', 3 'query', 96 'no'].

Here, it turns out that energy is needed [and indispensable] to create the 'regions of misfit' between two domains, because these regions are of finite thickness.

(OTU 64, J3.27)

when ripe they [the follicles] rupture into the coelom, which becomes filled with [and full of] spermatozoa ... (OTU 83, J1.96)

(W 37) Potential coordination with a prepositional phrase [33'yes', 4 'query', 91 'no'].

when there's / "so many variables and the mathematician is beaten # [or in doubt/difficulties] . /then we :come !in # (OTU 90, B2.8) Lady Foxglove ... had somewhere inside her a spring of the toughest steel which was all the more disconcerting because it was hidden [and out of sight] for most of the time, ... (OTU 104, M3.206)

Potential premodification by quite, much, rather, very.

Yes Query No

(W 38) quite 33 2 93 (W 39) much 26 5 97 (W 40) rather 18 1 109 (W 41) very 17 2 109

When news of Lady Foxglove's latest excursion in good works reached those Christian circles in which she occupied so illustrious a place, opinion was [quite] divided as to whether it was her best work so far, or whether it was conclusive proof of her approaching insanity. (OTU 37, M3.197)

... the reaction of benzoyl peroxide with benzene -114C, which was [much] discussed in Section 3a, ... (OTU 49, J4.37)

She was [rather] astonished, and said so. (OTU 113, Μ 1.23)

he doesn't lack quality, I feel [very] convinced. (OTU 72, M2.105)

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 97

Potential modifying function of the past participle (Ved). The noun with which this test was first made was the subject (Nx). When this was not possible for exponential reasons, as in the case of personal pronouns, a noun was inferred from the context. It appeared that the participle was much more readily accepted as a noun modifier when it had some modification of its own. The participles were tested in both pre-and postmodifying functions, in either case with and without individual adjuncts (A). The following distribution was obtained:

Yes Query No

(W 42) Ved Ν 49 2 77 (W 43) A Ved Ν 101 1 26 (W 44) Ν Ved 41 3 84 (W 45) Ν Ved A 115 0 13

In the following example the participle could take any of the four functions:

Table 20 shows the good agreement obtained with these parameters for the dis-ordered structure: for comparison the theoretical values of the intensities of the decisive reflections (111), (211), and (221) for the ordered structure are also given. (OTU 111, J2.177)

the given values -*• the [previously] given values

the values given -> the values given [by the analyst].

Other participles were preferably modified in some way, for example:

The intimate mechanism of this substitution step cannot be elucidated by exami-nation of the reaction kinetics and products alone, and further discussion of it is deferred to the next chapter. (OTU 120, J4.39)

->· *the elucidated mechanism -*• the [carefully] elucidated mechanism

*the mechanism elucidated -» the mechanism elucidated [by examination]

The formation of these unsymmetrically substituted diaryls was regarded as an indication that ... (OTU 89, J4.32)

*the regarded formation -• *the [previously] regarded formation

*the formation regarded -» the formation regarded as an indication that ...

5.6 CRITERIA RELATING TO THE SUBJECT (Criteria S 46-59; see Figure 5:3, p. 77)

Most of the categories in terms of which the subject was analysed have been described

98 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

in Section 3.2 (pp. 26 ff.), and need only be exemplified here. Some new categories will also be introduced.

Form-class. There were no nominalizations operating as subject in this corpus. The nominal groups as S had the following types of head:

(S 46:0) Noun [74 cases].

When the observed background was corrected for multiple scattering and thermal effects the residual value agreed with this expected amount of diffuse scattering, within the experimental error. (OTU 2, J2.169)

(S 46:1) Pronoun [49 cases]. Examples will be given in connection with their specification below under S 47.

(S 46:2) Name [3 cases].

but I /mean . !the real real reason why [ro] a Burns is :so . a :worshipped# is /because of course he was a :self-made män# (OTU 39, B1.30)

(S 46:-) Zero [2 cases].

Thus a second-order transition of type Β is not, as [0] has been stated by some authors, impossible [ital. in text] to realise, ... (OTU 57, J3.30)

Specification of pronouns. (S 47:1) Personal referential pronoun [27 cases] (see, for example, OTUs 72 and

113 above under W 38-41). (S 47:2) Nonreferential pronoun : proleptic it [3 cases].

Now, it can be shown without difficulty that, for an assembly containing a finite number of molecules f should be differentiable any number of times with respect to T, ... (OTU 29, J3.23)

(S 47:3) Relative pronoun [14 cases].

... the radical Ar'... ejects a hydrogen atom which is incorporated in the pro-ducts N. (OTU 32, J4.27)

The experimental results for the various groups of lines which could be measured are shown in Fig. 62... (OTU 33, J2.168)

(S 47:4) Demonstrative pronoun [5 cases].

... Davies (1958) has shown that a small but appreciable amount of the ester ArCO.OAr'... is indeed formed, and this is taken as evidence of the fundamental correctness of the approach expressed by Scheme III. (OTU 68, J4.38)

Gender, subsuming animate/inanimate, concrete/abstract, count/mass heads of nominal groups.

(S 48:1) Animate [27 cases].

/those who lären't insäne# the /vast ma!jority#. are con/cerned with !some

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 99

personal quärrel# - or have been /driven to desperation by a :personal re:lä-t ion# . or /something of this klnd# (OTU 119, B3.59)

(S 48:0) Inanimate [99 cases].

The purified peroxide, which is generally crystalline, is added to the aromatic solvent, and the decomposition is effected by heating, usually to about 70-80°.

(OTU 73, J4.34) (S 49:1) Concrete [77 cases].

The olfactory capsule, imperfectly paired, is also almost detached from the cranium. (OTU 66, J1.87)

(S 49:0) Abstract [49 cases].

... these investigations were carried out in 1948 (although they were not declassi-fied and published until several years later) ... (OTU 74, J2.184) This water must be removed without losing salt; accordingly in most freshwater animals, including vertebrates, we find some system by which the separation can be achieved. (OTU 88, J1.93)

(S 50:1) Countable [109 cases].

... the phenolic esters formed thereby are easily removed by hydrolysis and ex-traction with alkali. (OTU 70, J4.35)

(S 50:0) Mass [17 cases].

The presence of a disordered structure was again substantiated by a marked hump in the diffuse background of the powder diffraction pattern. (OTU 97, J2.178)

(S 51) Finitude. Definite [110], indefinite [16], zero [2 cases]. Finitude (for cate-gories other than names and referential pronouns which were all taken to be definite) was specified as follows:

(S 52:1) Definite article [46 cases].

... the hydrogen atoms are not restricted to the 0-0 bonds as they are in the other three models. (OTU 27, J2.168)

(S 52:2) Possessive pronoun or 's genitive [5 cases].

The suprarenal tissue receives 'preganglionic' fibres from the spinal nerves. Its cells sometimes seem to be connected with each other by fibres like those of neurons ... (OTU 127, J 1.98)

(S 52:3) Demonstrative pronoun [7 cases].

Below the epidermis lies the dermis, a layer of bundles of collagen and elastin fibres, running mostly in a circular direction. This tissue is sharply marked off from a layer of subcutaneous tissue ... (OTU 118, J1.84)

100 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

(S 52:4) WA-pronoun [2 cases].

/what character, what :physical. body you häve# and [to /some extent] what •.character you poss6ss# . [to the extent that your] . /"liver< shall we say) moulds your chäracter#. /is inherited# (OTU 46, B2.44; [allegro]; (parenthetic))

(S 52:5) Indefinite article [3 cases]. A different explanation ... has been offered by Augood and Williams (1957).

(OTU 110, J4.37) (S 52:0) Zero article [11 cases].

Direct control of the spinal cord from the brain is obtained through a number of very large Müllers fibres, ... (OTU 13, J1.98)

Person. (S 53:1) First person [4 cases].

/you and I and "äll of iis# are /human b6ings# we re /bora in the same coun-try# (OTU 123, B1.15)

(S 53:2) Second person [O cases]. (S 53:3) Third person [122 cases] (including 2 zero subjects).

/most of the :gagslüsed# have al/reädy been iised# by the /two chäracters at the other end of the :täble# (OTU 100, B1.35)

(S 53:-) Zero [2 cases] (see S 46). (S 54) Number. Singular [80 cases], plural [46 cases]. (S 55) Modification [49 cases] was specified according to: (S 56) Premodification [33 cases].

An early attempt was made ... (OTU 96, J2.166)

... so far as the heavier elements are concerned. (OTU 28, J2.163)

(S 57) Postmodification [34 cases].

... the formation of diaryls was observed ... (OTU 121, J4.28)

19 subjects had both types of modification, for example:

... the phenolic esters formed thereby are easily removed ... (OTU 70, J4.35)

(S 58) Coordination [6 cases].

... the unit cell and atomic positions of the elements other than hydrogen, or light elements, are known from X-ray studies ... (OTU 63, J2.162)

... the stability of the transition state, and hence the activation energy for the dissociation, are influenced ... (OTU 26, J4.32)

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 101

(S 59) The number of words in the subject ranged from 0 to 27 with an average of words, the most frequent being 1 word.

(C 60) Presence of clause element C. There were only nine clauses with complements as defined in Section 3.1-2, pp. 25if. They were the following form-classes:

(C 61:1) Nonfinite verb clause with to-infinitive, active or passive (to V) [5 cases].

[the diencephalon] is known to be concerned, throughout the vertebrate series, with the integration of the internal activities of the animal. (OTU 94, J1.106)

(C 61:2) Noun [2 cases].

'[this bookshop] does cater for what's called the cultivated reading public - and for specialists.' (OTU 106, M1.20)

(C 61:3) Adjective [1 case].

... many of them were taken ill ... (OTU 67, M3.209; for context see CL 6:2)

(C 61:4) Nonfinite verb clause with present participle (Ving) [1 case].

... one was left not knowing whether Viola was less innocent or Tanya more inno-cent than each seemed. (OTU 38, M2.98)

(C 62) The number of words in the complement ranged from 1 to 21 with an aver-age of words.

Adjuncts were analysed in terms of 46 categories. The clause element A in passive clauses has a particular interest to a study of voice, since the nominal part of A in its function as agent is the potential subject of a corresponding active transform (see Section 3.33, pp. 30f.).

Firstly, a distinction was made between '/4-full' and 'Λ-less' clauses, i.e. according to whether the clauses had an A element present or not. If they were Λ-full, some functions of A were established by tests using potential question-forms with wh-pronouns and w/i-adverbs in Criteria A 66-74. Secondly, the values of the A element with agentive function (Ag) and nonagentive function (Ad) were analysed separately in Ag 76-98 and Ad 99-108 respectively.

5.7 CRITERIA RELATING TO THE COMPLEMENT (Criteria C 60-2; see Figure 5:4, p. 78)

5.8 CRITERIA RELATING TO THE ADJUNCT (Criteria A 63-108; see Fig. 5:5, p. 79)

102 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

The distribution in the clauses of Ag and Ad as sets was the following:

v4g-less but Ad-ΐυΜ [60 cases] /4g-less and ^4<i-less [34 cases] /4g-full and Ad-full [26 cases] Ag-full but y4<i-less [ 8 cases]

(A 64) The 94 Ά sets' were the following types: A1 [53 cases], A2 [35 cases], A3 [5 cases], A4 [1 case],

(A 66) Potential question-form with wA-pronoun [51 cases], which was specified as A 67 who/m/ [15 cases], and A 68 what [36 cases].

(A 69) Potential question-form with wA-adverb with or without preposition [66 cases], which was specified as (A 70) where [21 cases], (A 71) when [19 cases], (A 72) how [25 cases], (A 73) why [4 cases], (A 74) other (excluding agentive function) [7 cases].

... he was ... carried effortlessly by Trained Pig Percy-Scroop-Beauchamp to the bathroom. (OTU 92, M3.209; see CL 4:3 for context)

(A66) by whom (who ... by) (A72) how (A70) where/ ... to/

was he carried?

Max Müller was subsequently convinced of the stupidity of his hypothesis and recanted in no uncertain terms. (1-24)

(A67) of what (what ... of) ι was Max Müller convinced ...? (A71) when

5.81 Adjuncts with agentive function (Ag 75-98)

As a consequence of the function-based definition of 'agent' in this study (see Section 3.33), the application of the term will be considerably extended compared with pre-vious treatments. A distinction will be made between the two classes grammatically and lexically determined agents. The preposition introducing grammatically deter-mined agents is by, and they will therefore usually be referred to as 'BY-AGENTS'.

Lexically determined agents are introduced by a variety of prepositions and will be referred to as 'QUASI-AGENTS ' . Unlike by in fty-agents, these prepositions are not voice-conditioned, but are selected by collocation with particular verbs: be worried about, surprised at, interested in, annoyed with, etc. Cf. the corresponding nouns and their prepositions: worry about, surprise at, interest in, annoyance with. The distinction between by- and quasi-agents may hence be stated in terms of colligation (for Ay-agents) and collocation (for quasi-agents). (For the use of these terms, see for example Robins 1964, 67if., 234.) It will also be convenient to let quasi-agents subsume a type of agentive complements which are not introduced by prepositions. Being lexically

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 103

determined, quasi-agents may naturally display specific lexical meanings, such as cause or instrument. Consider the following series with (a) Z>j>-agent, (b-c) quasi-agents, and (d) non-agent:

(a) AGENCY: Abel was killed by Cain. -+ Cain killed Abel. (b) CAUSE: Abel was killed out of hate. -* Hate killed Abel. (c) INSTRUMENT: Abel was killed with a jawbone. -*• A jawbone killed Abel. (d) PLACE : Abel was killed outside Eden. *Eden killed Abel.

Agent, cause, and instrument functions are related, and the distinction can be more or less syntactically neutralized.9

However, in our case the difference between the by-agent (a) and the quasi-agents (b-c) becomes clear if they are all in turn made active subjects with the other two phrases retained as prepositional adjuncts:

(a) Cain killed Abel with a jawbone out of hate. (b) *A jawbone killed Abel by Cain out of hate. (c) *Hate killed Abel by Cain with a jawbone.

In an example like the following, however, there is neutralization in the active:

• ' . . . just as the instrumental and causal domains border hard on each other and partly pass into each other, so causality and authorship are distinctions but of degree and not of kind' . ' I t was from the instrumental of means with an active verb that the instrumental of agency with the passive had arisen, i.e. that to a great extent the personal agent owes its provenience to the material instrument with which a deed is accomplished' (Green 1914, 551, 524f.). Green sets up a series of gradation (his examples are Latin, but the gradient might well apply to English):

(a) 'material means, pure and simple' (b) 'personal instrument, intermediary' (c) 'personified means, often of causal tenor' (d) 'personal agent' In this connection it is interesting to note the considerable fluctuation in the development of

English in the use of agent-prepositions, 'first purh begins to lose ground, then fram follows suit, and then of gives way to by, which becomes the modern preposition of agency through the gradual supersession of all its competitors' (Green 1914, 552). Mustanoja cites six prepositions (by, from, mid, o f , through, with) with agent-function in Middle English and points out that 'unambiguous cases with by indicating the agent of a passive verb are rare until the end of the 14th century' (1960, 442).

Some prepositions other than by occurring with agent-like functions are listed in the handbooks for present-day English. Of 'is still in literary use, as a biblical, poetic, or stylistic archaism, or by association with other constructions, e.g. 'on the part o f ' '(admired / loved / hated / ordained of\ abandoned / desirted / forgotten / forsaken of; unseen / unowned of)' (OED. sv. of: V. 15). With, 'the

104 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

We shall now illustrate the two agent-classes with examples from the material. Some quasi-agents (2 g-i below) are peculiar in not being prepositional phrases but C elements consisting of a nonfinite construction (fo-infinitive) or a finite (zero) that-clause. The latter may however be regarded as truncated forms of prepositional phrases of the type 'preposition + the fact + iAai-clause' or some other abstract nominalization:

I was delighted jto hear I that he won. To hear that he won That he won The fact that he won delighted me. His winning Hearing that he won

(1) 57-AGENTS. (la) Animate by-agent:

Both Hungary's goals were scored by Albert, ... (lb) Inanimate Ay-agent:

The settlers have not been converted by the events of the past fortnight. (2) QUASI-AGENTS. (2a) About: We were all worried about this (2b) At: I'm a bit surprised at [N.N.] you know (2c) Before: Appelby was puzzled before this somewhat incoherent vein. (2d) In: ... he's passionately interested in pandas. (2e) To: A man prepared to go pretty high in the matter of buying a large library the

value of which could only be approximately known to him, ... (M6.115) (2f) With: You won't be bothered with me any more. (L2.74) (2g) To-infinitive: 'So I was surprised ... to learn that the poor chap had blown his

brains out. (M6.47) (2h) 77ia/-clause: I'm not surprised actually that Mr [X.] has got on well. with a a

American industrialists and big businessmen (B5.6) (2i) Zero that-clause: 'I'm surprised that Portway girl was so restless,...'' (M5.79)

(F.23)

(G.6)

(B5.3) (M6.27) (L2.62)

JANUS-AGENTS. Some prepositional phrases may have twofold potential function in active transforms: their nominal part may function as active clause subject, or the entire prepositional phrase may function as adjunct (usually instrumental, 'by means of') with some other nominal element as active clause subject. Such adjuncts which

typical preposition in Present English to denote instrumentality, is not unfrequently met with after a passive participle in a function which is difficult to distinguish from that of by or of': met with strong opposition, visited with a serious outbreak of typhoid fever (Poutsma 1926-9, 2:2.96f.). Jespersen gives struck, seized, bit, eaten with, etc. (Part 3.319). In has 'a function approximating to that of agency' in caught in the rain, covered in black leather (Poutsma, op. cit., 97). To is 'frequent' after known: Ά man's intimate friends ... are quite unknown to, or by, his wife', and Jespersen asks 'What is the difference?' (op. cit., 3.319).

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 105

permit two different active clause transforms, according to whether they are inter-preted as Ag or Ad, will be called 'JANUS-AGENTS' (Ag/Ad).10

Oil will be replaced by coal.— coal will replace oil [they] will replace oil by coal

How much has in fact been achieved by the mergers that have been made! (G.16)

Loss of inhibitory power by avenacin on enzymic hydrolysis was confirmed by growth experiments ... (J5.325)

'... for his land when some official decided to take it from him, does not allow him some compensation for a business that has been ruined by vacillation and muddle when some other official has decided to give it back.' (M5.78)

Technically the French will need more tests if they are to develop the nuclear device tested on Saturday into a weapon that can be carried in an aeroplane or into a warhead that can be delivered by a missile. (G.15)

In some cases there exist more complicated transformational possibilities, as for instance:

[a] collapse now threatens market prices through a rush by fanners a rush by farmers to sell... now threatens market prices with collapse farmers now threaten market prices with collapse through a rush to sell farmers rushing to sell ... now threaten market prices with collapse

most languages express 'to whisper' by an imita-tive verb in most languages [people] express 'to whisper' by an imitative verb an imitative verb expresses 'to whisper' in most languages

In discussing potential transforms here we can say nothing about the probability of their use. For the three active transforms of the last example, the probability of occurrence may be 'likely', 'less likely', and 'least likely' in that order. .ßy-phrases,

Market prices are now threat-ened with collapse through a rush by farmers to sell before the new crop starts in July.

(E2.29)

'to whisper' is expressed in most languages by an imita-tive verb ... (1.52)

10 This notion helps to distinguish John was frightened by the new methods which is Ag/Ad (Janus-agent) from The picture was painted by a new technique which is only Ad (nonagentive adjunct). (See Chomsky 1957, 89 and Lees 1957, 383.)

106 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

like other prepositional phrases, can of course also function exclusively as non-agentive adjuncts:

, ι—• [this] took him by surprise He was taken by surprise-] . ^ . . . '-> »surprise took him

, „ , . , ι-» *[they] went him by 3 o'clock He was gone by 3 ο clock-] μ *3o clock went him

The last example accepts transmutation (see CL 20), He went by 3 o'clock, which indicates its close relation to Type Β (see Section 2.4, p. 17).

Adjuncts with agentive function (Ag) were analysed with regard to both overt and covert criteria.

(Ag 75): see p. 101. (Ag 76) /lg-position in relation to Ved. It followed immediately after Ved in 23

out of 31 cases of ^g-full clauses.

(Ved + Ag): The upper surface of the brain is covered by an extensive vascular pad, the choroid plexus or tela choroidea ... (OTU 51, J1.101)

(Ved + Ad + Ag): The green air ... was ... smashed to atoms, glowing round the wispy edges of grey vapour-bright bits of lime-colour that grew smaller and smaller until they were stamped out into nothing by the cold blackness of continuous cloud.

(OTU 6, M4.45) (Ved + Ag + Ad):

This mechanism was supported by De Tar and Sagmanli (1950) in an attempt to explain the products of the reaction. (OTU 10, J4.37)

(Ag 77) 'Janus-agent' [5 cases]. In principle the whole field of organic chemistry is filled with problems which can be solved by neutron diffraction investigation of the position of hydrogen atoms.

(OTU 15, J2.162) ι—• which neutron diffraction investigation ... can solve

which [we] can solve by neutron diffraction investigation

(Ag 78) The number of agent words varied from 2 to 40, the average being about 8 words. It is perhaps worth noting that the curve for agent-length is much more irregular than the one for subject-length.

(Ag 79) Potential deletion occurred in 17 cases, plus 2 queries. This criterion proved to be difficult to apply, and hence was of little value in its present formulation. Nevertheless, it would seem true that, in the examples above, the agent in OTU 6 is more dispensable than those in OTUs 51 and 15.

(Ag 80) Specific exponents of the agent preposition: by [30 cases], under [1 case].

PASSIVE CLAUSE CLASSIFICATION : CRITERIA 107

The sacs form by evagination from the brain and remain connected with the dorsal epithalamic or habenular region of the between-brain by two stalks.

(OTU24, J 1.103)

Presumably the osmotic flow of water into the body is relieved by the pressure of the heart-beat forcing water out from the glomeruli into the coelomic fluid, whence it is removed by the funnels, with the aid of their cilia.

(OTU 5, J 1.95) The clause with under as 'quasi-agent' was the following:

'... on behalf of the National Society for the Prevention of Suffering in Helpless Children, I am empowered, under the Act of 1948, to make an inspection of the premises and circumstances of living of one Tarquin Foxglove, an adopted child of this address.' (OTU 99, M3.206) -»• the Act of 1948 empowers me to make an inspection

The nominal part of the agent, i.e. the potential active subject, was analysed in terms of the same categories as the passive subject (see S 46-59): form-class, gender, finitude, person, number, modification, and coordination (for other examples of this typical agent-feature, see for example OTU 69, p. 82 and OTU 101, p. 87). The categories need not be further illustrated here. Instead we shall compare some of the inventories for both subject and agent in Ag-full clauses. (One of the ^g-full clauses had zero S which accounts for the discrepancies between the totals of the 5 and Ag inventories.)

SUBJECT AGENT

Form-class: Noun 20 23 (Ag 81) Pronoun 10 0

Name 0 8

Gender: Animate 5 11 (Ag 82) Inanimate 25 20 (Ag 83) Concrete 17 19

Abstract 13 12 (Ag 84) Mass 3 10

Countable 27 21

Finitude: Definite 26 19 (Ag 85-6) Indefinite 4 12

Person: First 1 0 Second 0 0 Third 29 31

Number: Singular 17 13 (Ag 87) Plural 13 18

108 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

SUBJECT AGENT

Modification: 17 23 (Ag 88) (Ag 89) Premodification 12 12 (Ag 90) Postmodification 14 19

Coordination: 1 10 (Ag 91)

No. of words (average): 8 (Ag65)

(Ag 92) Extension with an agent was possible in 107 clauses, impossible in 17, and was queried in 4. 'Possible' includes the 31 cases of expressed agents (see Ag 75 above).

(Ag 93-94) The most common potential agent preposition was by (102 'yes', 2 'no', 6 'queries'). Other elicited prepositions were at, of, and under (see Ag 80 above).

She was astonished [at/by the dog's behaviour] (OTU 113, Μ 1.23) we're /born in the same country # [of English parents]u (OTU 123, Β 1.15)

Potential exponence of the nominal part of the agent was tested for the following four categories (suggested agents are given in square brackets).

(Ag 95) Potential animate agent [70 cases]. Knowing the value of R® we can determine the structure factor F, when extinction is negligible, by the equation (7.1) as has already been described [by the author] in Chapter III. (OTU 30, J2.171)

(Ag 96) Potential inanimate agent [53 cases]. Now that Martin was a full lieutenant he was entitled [by tradition/his position, etc.] to a Jhon [ital] by himself ... (OTU 75,M3.203)

(Ag 97) Potential concrete agent [75 cases]. See for example OTUs 34, 30, 24 above.

(Ag 98) Concrete agent in the second of two possible agent types [2 cases]; abstract only [15 cases]. Ag 98 was introduced as a very tentative criterion following the observation that when a second type of agent was possible, this agent or near-agent (it was usually a less likely agent than the first) was remarkably often abstract, as for instance in the following two examples (where the second agents are of the Janus-type):

11 'In modern use the connexion with bear is no longer felt; the phrase to be born has become virtually an intr. verb' (OED, .s.v. born). In the sense 'bring forth, give birth to' only borne is used with by and the name of the mother, whereas born is restricted to the passive, when it has no by-agent, and only in one sense: 'it has rather a neuter signification = "come into existence, sprung"' (OED, s.v. bear, 44). Because of the tenuous relation of born to bear (*English parents bear us), the o/-phrase with born, which was supplied in this example by the informant, has doubtful status, and indeed does not qualify as agent in the definition of Section 3.33 (p. 30f.).

PASSIVE CLAUSE CLASSIFICATION: CRITERIA 109

Faintly she frowned, contemplating the area of her labours, seeing what should have been, what could be done. (OTU 36, Μ 1.17) -»· what could be done [by her/by hard work] 'and if the truth is to be expressed I never shall [pass].' (OTU 54, M2.90) -* if the truth is to be expressed [by me/by (in) a few words]

5.82 Adjuncts with nonagentive functions (Ad 99-108)

(Ad 99) The 86 Ad-iuYi clauses (see the beginning of Section 5.8) were analysed in terms of the following categories:

(Ad 100) Number of Ad items in each clause (i.e. Ad set types, see Section 4.5, pp. 57f.):

Ad1 [64 cases], Ad2 [18 cases], Ad3 [3 cases], Ad* [1 case].

Form-class. (Ad 101) Closed class adverbs [11 cases], for example again, already, almost, first, still.

... the full mathematical details of the theory of non-equilibria, as will now be shown. (J5.301) Probably this reputation was always exaggerated. (G.6)

(Ad 102) Open class adverbs (usually formed by the suffix -ly added to an adjective) [7 cases], for example absurdly, effortlessly, intimately, recently, neatly.

He tempted Cartwright with two half volleys which were royally punished ... (F.28)

The invitation hadn't been very felicitously phrased, but was entirely cordial. (M6.13)

(Ad 103) Prepositional phrases [61 cases]. When the phrase had potential question-form with wA-pronoun (see A 66), the exponents of the preposition were stated in Ad 104. They were, in order of frequency, from [7 cases], with [6 cases], to [6 cases], for [4 cases], as [4 cases], of [3 cases], and into [1 case].

the /only way of deciding [/whether it's a powerful deterrent#]# - is by [/look-ing at 'what has happened#]. /in the leases to which it {:has in the päs t# been appVlled#} (OTU 78, B3.26, [spiky], {rhythmic}) ... the occurrence of an irreversible effect is so intimately connected with the mechanism of the transition itself that ... (OTU 31, J3.34) the [sei?] "sole offence# . for/which i t . :could be [applied#]

(OTU 58, B3.30; [low/creak]) The available evidence will be considered to lead to two interpretations and these will be used as a framework for discussion. (J5.275)

110 PASSIVE CLAUSE CLASSIFICATION: CRITERIA

Too much, however, should not be made of Griffins temporary embarrassment. (F.23)

..., until an ancient parish church was made into the choir of a new cathedral for Truro in the present century. (G-9)

(Ad 105) Finite or nonfinite verb clauses [12 cases].

Not until the sun was moving down over the blue hills of Donegal was the destiny of these two matches decided, ... (F.28) T o qualify for a record consideration round a curve the race must be made on a track not more than 440 yards in circumference and started on some part of the circumference.' (F.30)

Position (see Section 3.32, p. 30): (Ad 106) Front-position [14 cases]. (Ad 107) Mid-position [14 cases]. (Ad 108) End-position [65 cases]. The first of the following two examples has a closed class adverb in mid-position;

the second has a finite clause in front-position and, in end-position, a prepositional phrase plus a nonfinite clause whose subject is introduced by a preposition.

... this private enterprise Emperor would inevitably follow the same primrose path that had already been blazed by other British long-range airlines designed since the war. (OTU 11, M4.49)

... in the hydrogen distribution the average coherent cross-section per deuterium nucleus would be 4 nbb, but when there are twice as many possible sites as atoms this will be reduced to 2nbt>, with the remaining cross-section of 2nb'b, which amounts to 2.6 barns, appearing as diffuse background scattering.

(OTU 47, J2.169)

6

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

In Chapter 5 we introduced and illustrated a large number of criteria which were applied to a small number of randomly selected finite passive clauses. This chapter will describe the preparation of the input data, the operations of the classification program, the analysis of the output, and the setting up of a diagnostic key.

6.1 PREPARATION OF INPUT DATA

The 108 criteria were applied to the 128 clauses (called OTUs', short for 'operational taxonomic units'1), whose 'features' were entered into cells. Linguistic analysis and informant reaction were used to determine, respectively, the features to be assigned to overt and covert criteria. (For the two types of criteria, see Section 5.3, p. 74.) This meant that 128 χ 108 = 13,824 cells had to be filled in.

The performance of an experiment on such a scale is a laborious process: to extend the same detailed mode of analysis to a large corpus would be impracticable. What we must hope for is that the results of such an experiment, or preferably a series of such experiments, will give us sufficient insight into the problem to enable us to detect certain 'key features'. Such criteria, once we have found them, can then be economically applied to a very much larger corpus, and will thus provide a practical means of arriving at a good classification.

In order to achieve such a sophisticated goal as was outlined in Section 5.1, pp. 72 f. (that the classification should be, for example, objective and empirically verifiable, con-sistent with intuition, predictive, and comprehensive), it would be necessary to follow up pilot experiments on the lines proposed here with a series of related experiments, each one successively making use of the previous results. There are many reasons for such a procedure. The most important is that with a small corpus we have no guarantee that the OTUs will include all passive clause types. The most that one can expect is that a pilot experiment will provide a classification of only the most frequent classes. An attempt to avoid this undesirable effect was made in this experi-

1 Here, it will be possible to give only a condensed outline of the operations of the classification program and other technical aspects of the experiment. For a detailed discussion of relevant problems in numerical taxonomy and for further references, see Carvell & Svartvik 1966.

112 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

ment by means of a provisional ordering of the material (as mentioned in Section 5.2, p. 73).

The features were encoded in two ways, according to whether or not absence of a certain feature counted. Criteria in the input data will be referred to as having 'levels'. For example, Criterion CL11 had two levels: 'potential active transforma-tion' and 'no potential active transformation'. In the first class of criteria, absence of the same feature in two OTUs is disregarded for purposes of comparison. This class includes 'dichotomies' (in the input encoded as 1 and 0) and 'positives', or 'one-level qualitatives' (encoded as 0 and -), where 0 and - , respectively, denote absence. An example of such criteria is the modification of the subject and agent (S 55-7, Ag 88-90, pp. 100, 108).

In the second class of criteria, which may be called 'qualitatives' and 'quantitatives', absence of the same feature in two OTUs counts as a similarity. Qualitatives take levels ranging from 0 upwards (/, 2,3,4, etc.) referring to mutually exclusive features. An example of a 3-level qualitative is subject form-class (see Criterion S 46, p. 98), where 0 denoted noun, 1 pronoun, 2 name, and - zero. In quantitatives the levels are an ordered series in a linear scale. This test type was suitable for the encoding of element lengths, as in the subject (S 59, p. 101) and agent (Ag 78, p. 106). In order to obtain what was thought to be a realistic scale, the logarithms of the element lengths were entered in the data input.2

In principle, the first class of criteria was used for optional overt features, such as modification and adjunct position (Ad 106-8, p. 110). The second class of criteria, on the other hand, was used for obligatory overt features, such as element gender (animate/inanimate, concrete/abstract, countable/mass, see S 48-50, Ag 82-4, pp. 98 f.; 107), and for covert features, such as agent extension potential (Ag 92, p. 108) and potential active transformation (CL 11, pp. 83 if.).

The hierarchal relationship of some of the criteria (as is clear in Figures 5 :1-5, pp. 75 ff.) posed a special technical problem. In a tree with many sub-branches, the first branches, which represented the primary distinction, tended to be considered more and more similar as the number of sub-branches increased. In order to counteract this undesirable effect in hierarchal criteria, 'dummy' levels were entered. For Criterion S 47, p. 98, for example, which specified pronouns (personal referential, proleptic it, etc., all of which were entered as 1 in S 46), a dummy 0 was encoded for non pronouns ( = S 46:0, S 56:2, and S 46:-).3

An example of the input form is given in Figure 6:1, which shows the first four OTUs labelled by their finite lexical verbs: TWIST, CORRECT 1 (where 1 indicates that this is the first of two or more occurrences of this verb in the corpus), EXCITE, and

1 The base of the logarithms was irrelevant. The effect of this was, for example, that the similarity between elements of word lengths 1 and 2 was the same as that between elements of word lengths 2 and 4, or 5 and 10. ® Objections may be raised to this procedure on theoretical grounds. The problem has not, however, been solved, and there was little choice in the programming facilities available.

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 113

TWIST 0100 11110000 011111 110100 11 001123 —01 -1111— 22 1010-1 -1--0- 0 — 7 -0122 11 6 6 - - 1101

CORRECT 1 0010 10110000 011002 100000 1100— 001113

00000— 22 1101-1 -1—0- 30 11 (10) (16) - - 1101

11110000 111011 111001 11 111613 01111— 22 1010-1 -1 7-00 0 - - - 0000

1022

EXCITE 1000 — 1 0022

3260 022250 0— — 0 -00011 010

3260 — 0

3260

022203 -00011

022200 -00011

0 - 0

010

TOUCH 0100 — 0 0122 etc.

10110010 00000-0 -

11 (10)

011002 22

0

000000 1101-1

110-0-- 1 - 0 —

001113 0 — 7 -1101

3260 -0—

022250 -00001 010

Fig. 6:1. Specimen of data input (OTUs 1-4).

TOUCH. The clauses containing these verbs are cited on pp. 94,98,119, and 91 .respect-ively.

6.2 THE CLASSIFICATION PROGRAM

The classification program (CLASP), which had been written for the ICT Orion com-puter, classified the OTUs in terms of their resemblances, these being estimated on the basis of their common features. The two principal operations performed by CLASP are, very briefly, as follows:4 firstly, it calculates a coefficient of similarity between each pair of OTUs; secondly, it sorts the OTUs into a nested hierarchy of groups on the basis of the similarity coefficients. Furthermore, as a by-product of this sorting process, it produces a linear ordering of the OTUs, such that the OTUs of any group that has once been formed are never separated.

These operations produced the following output: (a) THE FULL SIMILARITY MATRIX. This is a subdiagonal matrix of the similarity

coefficients printed to three significant figures. Both the names and the numbers of the OTUs are printed in a column on the left hand side, against the corresponding rows; the numbers only are printed at the bottom, at the foot of the corresponding columns. The order chosen for the OTUs in both rows and columns is the order referred to above, i.e. that obtained as a by-product of the sorting-process. Figure 6:2 shows the top part of the full similarity matrix.

* The data were actually prepared for a program called F39 written for an Elliott 401 computer. Progress in this field is, however, extremely rapid, and by the time the data were ready for input a new classification program (CLASP) was available for the Orion, a much larger computer than the 401. In addition to larger storage, permitting as many as 400 OTUs and 256 criteria to be used for each OTU, CLASP provided many additional facilities, which were made use of. I am greatly indebted to J. C. Gower, Rothamsted Experimental Station, for his cooperation in this experiment.

114 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION A N D RESULTS

TWIST 1 CORRECT 1 2 70 .6 TOUCH 4 74 .9 85.5 INSTALL 34 76.5 87.4 94.9 DISCOVER 98 74.0 84.7 92.6 94.2 HIRE 55 74.0 85.7 88.8 93.5 89.5 SPREAD 16 72.1 76.3 84.4 82.8 84.3 78.9 DESCRIBE 30 73.6 83.1 87.0 87.0 87.2 89.2 89.6 AVOID 48 72.4 80.1 91.1 90.1 88.8 85.1 82.9 86.8 ACHIEVE 88 76.7 79.9 87.1 88.6 85.9 84.7 78.6 80.5 93.1 DISCUSS 2 49 67.1 82.6 80.8 83.8 84.6 81.2 91.1 88.2 80.3 78.8 PUBLISH 74 69.7 85.2 82.3 83.0 84.6 81.9 89.6 91.4 81.6 78.6 91.3 HANG 1 77 72.0 80.8 87.6 88.7 87.7 90.7 78.9 85.6 80.7 80.0 77.8 83.5 INCLUDE 1 79 77.2 74.9 81.2 83.0 84.5 82.7 79.4 82.7 82.5 86.5 77.5 79.6 88.0 - - - -SUMMARISE 87 78.9 83.5 88.3 90.0 90.4 87.2 85.2 91.7 88.5 87.0 81.7 86.9 87.5 87.4 GIVE 111 79.6 82.4 86.5 87.3 87.8 85.9 83.0 87.8 86.0 89.3 81.4 85.8 88.4 91.4 etc.

Fig. 6:2. Specimen of output: top part of the full similarity matrix (printed to three significant figures).

(b) ABBREVIATED SIMILARITY MATRICES. The full similarity matrix will normally be found to contain more detail than is realistic in view of the number of criteria, and more than is possible for the mind to comprehend, even after careful inspection. For this reason the program has the additional facility of producing abbreviated similarity matrices, which have coarseness as their characteristic feature. We may distinguish two types of such matrices: the 'basic type' which is printed to one significant figure only: i.e. using only the numbers 0-9; and the 'derived type', which is normally coarser than the basic type.

The linear ordering between individual OTUs, as well as between groups of OTUs, may be improved by the analyst. There are a number of options offered by CLASP,

some of which are in practice only applicable after the first output has been obtained. On inspection of our first output, the OTUs appeared to form six major groups, and they were rearranged accordingly. It is possible to define for the computer any number of OTU groups with the members of the groups ordered according to the analyst's preference. In one case this was done taking all the OTUs as an ordered group, and in another case dividing the OTUs into six groups (I-VI). Needless to say, this procedure cannot alter the similarity coefficients: it only provides the linguist with a more readily analysable output, and also gives him an opportunity to ask for the forms of abbreviated similarity matrices which are most realistic.

(c) T H E MEAN SIMILARITIES BETWEEN A N D WITHIN THE GROUPS are calculated and printed in the form of a small matrix. This is represented in Figure 6:3 as a shaded similarity matrix.

(d) T H E MOST TYPICAL MEMBER OF A GROUP is obtained by working out the mean similarity of each member of a particular group to every other member of the group. The OTUs are printed in descending order of mean similarity.

(e) G R O U P INVENTORIES can also be computed by CLASP . They are printed in the form of two-way tables which state the number of OTUs in each group which attains

PASSIVE CLAUSE CLASSIFICATION : COMPUTATION AND RESULTS 115

each level. In addition, the program gives the feature inventories for each individual OTU.

GROUP I

GROUP II

GROUP HI

GROUP IV

GROUP V

GROUP VI

Per cent similarity

70-79

60-69 S5 ϊ ί ί mm

Fig. 6:3. Non-proportional, shaded similarity matrix representing mean similarities between and within groups.

6.3 OUTPUT ANALYSIS

6.31 Taxonomic analysis

The abbreviated similarity matrices indicate that the OTUs fall into six groups of varying internal and external relationship.

At the top of the triangle there are two closely related groups (I and II) with an inter-group similarity of 70 per cent. Group I has an internal similarity of 73 per cent and contains 19 members (the first and last of which are OTUs 70 and 99, going downwards in the linear order of the matrix). Group II has a similarity of 79 per cent and contains 12 members (OTUs 102-119).

116 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

Together, Groups III and IV comprise more than half of all OTUs. Group III has 53 members (OTUs 34-38) and a similarity of 76 per cent;6 Group IV has 17 mem-bers (OTUs 8-7) and a similarity of 78 per cent. These two groups are closely associated: their mean similarity is 72 per cent, which is the highest between any two groups.

Group V is the smallest cluster, comprising only 6 members (OTUs 3-23), with an internal mean similarity of 75 per cent.

Finally, Group VI has 21 members (OTUs 117-44) and 76 per cent intra-group mean similarity.

The mean similarities between and within the six groups are represented by different shading in Figure 6:3. Groups III and IV are clearly central because of their fre-quency and, above all, because of their external relationships. They are related to Groups I and II as well as to Groups V and VI, whereas Groups I and II, on the one hand, and Groups V and VI, on the other, have little affinity. For some purposes, we may well be justified in conflating Groups III and IV, since they are so closely related to each other and so similarly related to other groups. Groups I and II, likewise, show great affinity. However, it is worth noting that Groups II and III are more similar than I and III, than II and IV, and, particularly, than I and IV. Groups V and VI are not highly related; Group V is more similar to III and IV than to VI; Group VI is more similar to III than to V.

With the exception of Group VI, the groups conform most closely to a cline dis-tribution, i.e. clusters which are serially related in a continuum.

6.32 Statistical assessment and linguistic interpretation

So far, the output analysis has been entirely taxonomic, since we have considered the groups without any reference to their contents. We shall now give the output a linguistic interpretation, basing our assessments, as far as possible, on considerations of statistical significance. However small some of the numbers may seem, they are nevertheless valid for statistical purposes, and the significances of their distributions may be established by means of contingency tables and the χ2 test.®

It is necessary to discuss briefly the nature of the statements we make about 2 x 2 contingency tables. The situation will be that we have four entries, as for example in Table 6:1, which shows the number of coordinate and non-coordinate subjects and agents (Criteria S 58 and Ag 91, p. 100,108) in the Taxonomic Passive Corpus, 31 clauses (i.e. just under one quarter) were agentful. Two clauses (one Λ^-full and one Λ^-less) had zero S which accounts for the discrepancies between the totals of the S and Ag inventories.

' Considering that a large group tends to have a lower internal mean similarity than a small group, the percentages for Group III, and for III/IV, are quite high. • For testing significances in tables each one of whose marginal totals does not exceed 40, we have used Finney et. al. 1963. Significances in tables with higher marginal totals have been established by means of the χ2 test, using Yates' correction, or where some entries were small, by means of Fisher's exact test (see, for example, Herdan 1964, 37ff.; Moroney 1956, 249ff.; Reichmann 1964, 335ff.

P A S S I V E C L A U S E C L A S S I F I C A T I O N : C O M P U T A T I O N A N D R E S U L T S 117

QO ca

ö 2 « Ί . 5 tn 3 <3

at ft

c υ bO cd

C υ

ν >

Vi tS 6 D _ ο Γ - 1 2 Γ "\> - Λ " _S> Κ 02- ε « S : .• ?! > - Ή ea- .E J> i i

Μ C > « 3 > μ cd Έ η s SS < C ο < eQ u

ι—( rt hJ ^ Λ t j a u ft

c υ

SP

* ·π

2 a . 2 < υ

υ cd

6 e

< cd

c υ βο cd υ cd ε

' c cd

r—ι *C/i t>

' w Μ Λ α

cd c

•α t o 3 CO CO < hj < u

cd ft

. « Μ > CO w ο g ε a w υ

cd α .

00 cd c ο

Ζ

α υ 60 cd

ft 3 n 3 ο σ ·

e α ao cd

£ c υ Ο

Ο-

Ι Ι Η Μ

< £

3 ~ 2 c ο ω Μ 2Ρ ο 5

Ζ £

ζ, g £Λ Η Β η « <2

.Ξ c — ι d «J

C Ο ω >

u Ο cd 3 4-» υ cd

•ts §

c -μ a> cd βρ

'S .2 £ 3 Ο 3 ο, σ1

c ο cd rt

•3 y

Ο 3 α , &

ο β .

ο

•κ C α Χ ο

AJ (ί-.

. ä c

a 2

υ >

cd ft

Ό C 3 ο

ΐ - Ρ ft CO Ρ < §

3 υ υ ü j

c ι - υ Ο 60 — Λ

cd — 3 cd υ cd Ο

Ζ

C ω Ο ft

Ε « Ο C/3 ft e

, ο

>

Ό C 3 Ο

A ft

§ Ι ζ 8

ω cd a Ξ

u 8 .δ 8 cn «a

cd i ä U, ft 73

1 1 8 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

TABLE 6:1 Criteria S 58, Ag 91

Feature

Clause element

Coordinate Non-coordinate Total

Subject 1 29 30

Agent 10 21 31

Total 11 50 61

On the basis of Table 6:1 we may hypothesize that there is a greater tendency to co-ordination in the agent than in the subject. It is unreasonable to hope for an estimate of how likely it is that this hypothesis is true. Instead we proceed indirectly: assuming provisionally that there is no connection, how unlikely is the table? The significance test used is based on the principle of the Null Hypothesis, which 'postulates that, as between two samples, there is no really significant difference and that such difference as occurs is attributable only to random sampling errors'.7 More precisely: we establish the chance - assuming the Null Hypothesis - that this table, or one even more biased, would have occurred. The situation we consider as given is provided by the row and column totals. In the example, this would give the following 12 alternatives, not all equally likely (indeed, A is more likely than L):

A Β C D L

There are two reasonable interpretations of 'or one even more biased'. One applies when we wish to test the bias in the form it has taken. This will be so if we have already suspected that, for example, coordination was more likely to occur in agents than subjects. In this case we shall be interested in the probability of one of A or Β on the Null Hypothesis. This is usually known as the single-tail case.

On the other hand, we may only wish to know whether the table is sufficiently remarkable to be worthy of investigation, in which case we shall have to consider bias in the opposite direction as well. Suppose the Null Hypothesis is true: at the 95 per cent level, we may then expect a deviation in one direction once in twenty times, and in the other direction once in twenty times. (To work at the 95 per cent level means, roughly, that we shall on average be wrong 5 per cent of the time.) Altogether, we may expect to be in error once in ten times if we test for significance at the 95 per cent level, using the single-tail test, and we shall therefore have to alter the significance level. Here, we shall use single-tail tests.

An example may make the situation clearer. Suppose a coin is tossed five times and comes down heads every time. How remarkable is this? If we know that the tosser is a cricket captain who has called 'heads', we should be justified in saying that the result is, on our terms, significant, for the chance of five heads, with an unbiased coin, is If, however, the tossing is done with no particular purpose in mind, we shall have to say that it is not so remarkable: all that has happened is that the same result has been obtained five times, and that five tails would have been equally surprising. The chance of five heads or

7 Reichmann 1964, 225.

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 119

five tails is of course not so small, being in fact The first situation we have described is the single-tail case; the second situation is the double-tail case.

The probability of the tables is given as p, assuming the Null Hypothesis. The terms used will be 'non-significant' for ρ > 0.05 (i.e. i ) , 'probably significant' for ρ < 0.05, 'significant' for ρ < 0.01 (i.e. and 'highly significant' for ρ < 0.001 (i.e. ^g).

Before we go on to discuss the characteristic features of the groups, it will be helpful to list their members. We shall first exemplify the most typical member of each group, and then list the other OTUs in order of their mean similarities with the other members of the group. (Owing to a computer printing error no value was obtained for OTU 62.) Percentages denote the mean similarity of that member to every other member of the group. In the exemplification of the most typical members, the entire OTUs are given in italics. In the list of OTUs, references are made for each OTU to place of citation in Chapter 5 (criterion and page). Most typical members:

Group I: Since aroyloxy radicals appear to be considerably less reactive entities than aryl radicals in aromatic substitution, the extent of aroyloxylation is generally small, and the phenolic esters formed thereby are easily removed by hydrolysis and extraction with alkali. (OTU 70, J4.35; 77 %)

Group II: The first investigations of the chloride by neutron diffraction were made by Goldschmidt and Hurst (1951) using the powder method for deuterated ammonium chloride. (OTU 102, J2.174; 83%)

Group III: The baby was installed in the old nursery wing at Bidcombe, and a very expensive nanny was hired from London to look after it.

(OTU 34, M3.198; 81%) Group IV: there had been "/no sig nificant increase# of "/any -kind# in /crime#

. and there was /nö . pr£ssure# . for the /reintrolduction of xorporal : punishment# . and "/no belief# that if it "/were introdüced# it would [/do anything about the .'crime • wave#]

(OTU 8, B3.46; [allegro]; 84%) Group V: '... Truly, what do we live for?' He became excited·, vodka splashed in

his glass; professors nibbling cheese straws peered over the tops of them to see what was happening. (OTU 3, M2.107; 80%)

Group VI: In later vertebrates the choroid extends only into the third and fourth ventricles. Presumably the vascular membranes of the brain are highly developed in lampreys because of the absence of cerebral blood vessels.

(OTU 117, Jl.101; 80%)

List of group members in order of internal group similarity. Forms with nominal bases (layered, striped), compounds (unknown), and participles with tenuous relation to verbal bases (bom, taken ill) are given in their forms of occurrence:

120 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

Group OTU OTU Mean Criterion Page number name similarity citation citation

70 REMOVE-3 77% Most typical member 120 ELUCIDATE 76% W 42 (p. 97) 73 EFFECT 76% S 48 (p. 99) 51 COVER 75% Ag 76 (p. 106) 5 REMOVE-1 75% Ag 80 (p. 107)

19 SURROUND 75% CL 19 (p. 88) 61 COMPENSATE 75% CL 18 (p. 87) 11 BLAZE 75% Ad 106-8 (p. I l l ) 97 SUBSTANTIATE 74% S 50 (P. 99) 25 ATTACK 74% CL 2 (p. 78) 59 REMOVE-2 73% CL 19 (p. 88) 6 STAMP OUT 72% Ag 76 (p. 106)

15 SOLVE 72% Ag 77 (p. 106) 127 CONNECT-4 71% S 52 (p. 99) 24 CONNECT-1 70% Ag 80 (p. 107) 26 INFLUENCE 69% S 58 (p. 100) 18 MOTIVATE 69% CL 18 (P. 87) 50 AFFECT-2 67% W 34 (p. 96) 99 EMPOWER 65% Ag 80 (p. 107)

102 MAKE-2 83% Most typical member 100 USE 82% S 53 (p. 100) 96 MAKE-1 82% S 56 (p. 100) 10 SUPPORT 82% Ag 76 (p. 106)

110 OFFER-2 81% S 52 (p. 100) 101 DEMONSTRATE 80% CL 18 (P. 87) 69 INVESTIGATE 79% CL 8 (p. 82)

125 SEE-3 78% CL 18 (p. 87) 42 DISCUSS-1 78% CL 16 (p. 85) 57 STATE 77% S 46 (P. 98) 92 CARRY 74% CL 4 (p. 80)

119 DRIVE 70% S 48 (PP. 98-9) 34 INSTALL 81% Most typical member

111 GIVE 81% W 42-5 (p. 97) 98 DISCOVER 80% CL 4 (p. 80) 87 SUMMARISE 80% CL 13 (P- 83) 30 DESCRIBE 80% Ag 95 (p. 108) 86 SEE-2 80% CL 15 (p. 84) 4 TOUCH 80% W21 (P- 91)

74 PUBLISH 80% S 49 (p. 99)

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 121

Group OTU OTU Mean Criterion Page number name similarity citation citation

III 2 CORRECT-1 80% S 46 (p. 98) III 88 ACHIEVE 80% S 49 (p. 99) III 48 AVOID 80% W 23 (p. 91) III 76 BRING 79% W 21 (P- 91) III 55 HIRE 79% CL 15 (p. 84) III 43 CORRECT-2 79% CL 6 (P- 81) III 89 REGARD 79% W42-5 (p. 97) III 71 OBTAIN-2 79% CL 2 (p. 80)

49 DISCUSS-2 79% W 39 (P- 96) III 63 KNOW-1 78% S 58 (p. 100) III 68 TAKE-1 78% S 47 (P- 98) III 79 INCLUDE-1 78% CL 10 (p. 83) III 52 FORM-1 78% CL 2 (p. 80) III 80 KNOW-2 78% W 28 (p. 94) III 45 DEDUCE 78% CL 4 (p. 80) III 121 OBSERVE 77% S 57 (p. 100) III 16 SPREAD 77% CL 20 (p. 89) III 77 HANG-1 77% CL 10 (p. 82) III 22 BITE 77% CL 19 (p. 89) III 13 OBTAIN-1 77% S 52 (p. 100) III 78 APPLY-3 76% Ad 103 (p. 109) III 64 NEED 76% W 36 (p. 96) III 27 RESTRICT 76% S 52 (p. 99) III 95 UNKNOWN 76% CL 18 (p. 86) III 118 MARK-2 76% S 52 (p. 99) III 85 CONSIDER-3 76% CL 7 (p. 82) III 115 INCLUDE-2 76% CL 18 (p. 86) III 58 APPLY-2 75% Ad 103 (P- 109) III 47 REDUCE 75% Ad 106, 108 (p. 110)

17 APPLY-1 75% CL 13 (p. 84) III 29 SHOW 74% S 47 (p. 98) III 93 CONVERT 74% CL 2 (P- 80) III 107 REMOVE-4 74% CL 17 (P- 85) III 83 FILL 73% W 36 (P- 96) III 75 ENTITLE 73% Ag 96 (p. 108) III 12 STRIP 73% W23 (P- 91) III 104 HIDE 73% W 37 (P- 96) III 90 BEAT 72% W 37 (p. 96) III 53 BREAK-2 71% CL 17 (P- 85)

122 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

rroup OTU OTU Mean Criterion Page number name similarity citation citation

111 1 TWIST 71% W 27 (P- 94) 111 94 KNOW-3 70% C 61 (p. ιοί) III 60 ALLOW 70% CL 7 (p. 81) III 103 OFFER-1 68% CL 13 (p. 83) III 105 BALANCE 67% W 21 (P· 91) III 38 LEAVE 67% C 61 (p. 101) IV 8 INTRODUCE 84% Most typical member IV 81 KILL 83% CL 19 (p. 89) IV 36 DO 82% Ag 98 (p. 109) IV 33 MEASURE 82% S 47 (P· 98) IV 82 RADIATE 81% W 25 (p. 92) IV 116 FLOG 81% W 24 (p. 91) IV 56 CONSIDER-2 81% CL 2 (p. 80) IV 54 EXPRESS 81% Ag 98 (p. 109) IV 46 INHERIT 80% S 52 (p. 100) IV 39 WORSHIP 80% S 46 (p. 98) IV 128 CALL-2 80% W 33 (p. 95) IV 109 SEE-3 79% CL 18 (p. 87) IV 106 CALL-1 77% C 61 (p. 101) IV 108 TRAIN 77% CL 7 (p. 82) IV 65 ORDER 76% W28-30 (p. 95) IV 7 SEE-1 74% CL 18 (p. 86) V 3 EXCITE 80% Most typical member V 113 ASTONISH 80% CL 13 (p. 84) V 72 CONVINCE 77% CL 6 (p. 81) V 9 WOUND 77% W27 (p. 94) V 114 DISCONCERT 76% CL 4 (p. 80) V 23 AFFECT-1 60% CL 6 (p. 81) VI 117 DEVELOP 80% Most typical member VI 84 MARK-1 80% W28-9 (p. 94) VI 66 DETACH 80% S 49 (p. 99) VI 21 CONTINUE 79% CL 20 (p. 89) VI 112 CONNECT-3 79% CL 4 (p. 80) VI 126 ROCK 78% W27 (p. 94) VI 14 LAYERED 78% W 32 (p. 95) VI 122 EXPAND 78% CL 20 (P· 89) VI 41 BREAK-1 77% W 35 (p. 96) VI 32 INCORPORATE 77% S 47 (p. 98) VI 91 STRIPED 77% W 32 (p. 95)

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 123

Group OTU OTU Mean Criterion Page number name similarity citation citation

VI 37 DIVIDE 76% W 38 (p. 96) VI 28 CONCERN 75% S 56 (p. 100) VI 31 CONNECT-2 75% Ad 103 (p. 109) VI 40 CONSIDER-1 75% W 31 (p. 95) VI 20 SHUT 75% W 27 (p. 94) VI 35 AGREE 73% CL 20 (p. 89) VI 124 PATCH 73% W 35 (P- 96) VI 123 BORN 70% S 53 (p. 100) VI 67 TAKEN ILL 69% CL 6 (p. 81) VI 44 COMPOSE 69% CL 4 (p. 80)

Some of the groups correspond very closely to our previous intuitive classification (see Section 4.11, p. 40): Group I is, roughly, coterminous with Class β, Group II with Class a, and Groups III and IV with Class γ. We shall now consider all six groups one by one as units. It will be convenient to say, for example, that 'Group II has animate agents' instead of 'the members of Group II have animate agents'.

Groups I, II, III, IV, and V have as their common distinctive feature potential active clause transformation (Criterion CL 11, pp. 83 f.). There is one exception: OTTJ 23 in Group V. It is clear from the list above that it is a misfit and should have been placed in Group VI. Not only is this OTU the least typical member of Group V, but the gap between it and the previous member (OTU 114) is as great as 16 per cent. This difference in internal group similarity is far greater than that between any other two adjacent OTUs in any group. For Groups I-IV, transformation is possible only into the extensive active clause type (CL 12); Group V usually admits of transformation into both extensive and intensive (CL 13) active clause type.8 Group VI (with OTU 124 as a doubtful exception) takes no active transformation. The first two groups have expressed Zry-agents (Ag 75): Group I has inanimate and Group II animate agents (Ag 82, p. 107). Groups III/IV may have agent extension (Ag 92, p. 108) with animate (Ag 95, pp. 108f.) or inanimate (Ag 96) agents; Group V usually takes only inanimate agents.

These will be our 'key criteria' since they are common to all, or almost all, the members of the groups which have been formed on the basis of the total set of criteria. Hence, these criteria may be used to define almost identical groups when we go outside the small sample used in the experiment. In addition, there are a number of other 8 'Extensive' is used here in contrast to 'intensive' for distinguishing, for example, This remark embarrassed him from This remark made him embarrassed as transforms of He was embarrassed by this remark. These terms are convenient but somewhat inaccurate, since a transformation of a passive factitive clause into an active factitive (for example He was called stupid [they] called him stupid), which is a transformation from one intensive clause to another, is here called 'extensive' (see Criteria CL 12-3, p. 83 f.).

124 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

TABLE 6:2 Criterion S 55, Groups I/II, IV (p = 0.0018)

S modified S non-modified Total

Groups I/II Group IV

18 2

13 15

31 17

Total 20 28 48

TABLE 6:3 Criterion S 55, Groups / / / / , III/IV (p = 0.011)

S modified S non-modified Total

Groups I/II Groups III/IV

18 22

13 48

31 70

Total 40 61 101

TABLE 6:4 Criteria CL 7-8, Group I (p = 0.009)

Coreferent with contiguous clause 5

Non-coreferent with contiguous clause S

Total

Contiguous active clause Contiguous passive clause

8

1

11

18

19

19

Total 9 29 38

TABLE 6:5 Criterion W 29, Groups I, II (p = 0.0004)

Commutation with seem

No commutation with seem

Total

Group I Group II

12 0

7 12

19 12

Total 12 19 31

features which are not common to as many class members but are frequently con-comitant with their key features (cf. what was said about 'predictive power' in Section 5.1, pp. 72f.).

Subjects of agentful clauses (Groups I/II) are, for example, more often modified than other subjects (Criterion S 55, p. 100; see Tables 6:2-3 for the relevant contin-gency tables).9 Group I subjects are frequently coreferent with subjects of active, but

* This particular application of contingency tables must be interpreted with some caution, and is further discussed in Carvell & Svartvik 1966, Section 7.

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 125

TABLE 6:6 Criterion CL 19, Groups I, II (p = 0.019)

Potential No potential Total permutation permutation

Total

Group I 11 7 18 Group II 2 10 12

Total 13 17 30

not of passive, contiguous clauses (CL 7-8, p. 82; Table 6:4). Unlike Group II, its passive auxiliaries commonly commute with seem (W 29, p. 94; Table 6:5), and potential permutation (CL 19, pp. 88 ff.) is more frequent than in Group II (Table 6:6) and, in fact, in any other group.

GROUPS I/II have a number of complementary characteristics associated with their key features (inanimate agent for Group I, animate agent for Group II). Unlike inanimate agents, animate agents are usually coordinate (Ag 91, p. 108; Table 6:7) and deletable (Ag 79, p. 106; Table 6 :8). Inanimate agents tend to be modified more often

TABLE 6:7 Criterion Ag 91, Groups I, II (p = 0.019)

Coordinate Non-coordinate Total

Ag inanimate 3 16 19 Ag animate 7 5 12

Total 10 21 31

TABLE 6:8 Criterion Ag 79, Groups I, II (p = 0.028)

Ag deletable Ag non-deletable Total

Ag inanimate 7 10 17 Ag animate 10 2 12

Total 17 12 29

than animate agents (Ag 88, p. 108; Table 6:9); furthermore, they tend to be indefinite (Ag 85, Table 6:10), and countable (Ag 84, Table 6:11). Animate agents, on the other hand, are more often definite and mass nouns. In order to find out whether there is, normally, association in nominal heads (as was, in fact, found to be the case between concrete and animate) between definite and mass or between indefinite and countable, the χ2 test was applied to the subject, the other nominal element in the same sample which was similarly analysed. Table 6:12 shows that there is no evidence of associa-tion between animate/inanimate gender and definite/indefinite in the subject through-out the corpus (p — 19/20), whereas the distribution for the same categories in the

126 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

TABLE 6:9 Criterion Ag 88, Groups I, II (p = 0.022)

Ag modified Ag non-modified Total

Ag inanimate 17 2 19 Ag animate 6 6 12

Total 23 8 31

TABLE 6:10 Criterion Ag 85, Groups I, II (p = 0.007)

Ag definite Ag indefinite Total

Ag inanimate 8 11 19 Ag animate 11 1 12

Total 19 12 31

TABLE 6:1 1 Criterion Ag 84, Groups I, II (p - 0.002)

Ag count Ag mass Total

Ag inanimate 17 2 19 Ag animate 4 8 12

Total 21 10 31

TABLE 6:12 Criteria S 48, S 51, Groups I-IV (non-significant)

S definite S indefinite Total

5 animate 23 4 27 5 inanimate 87 12 99

Total 110 16 126

agent has ρ — 0.007 (Table 6 :10). Similarly, Table 6:13 shows that the distribution of animate/inanimate and count/mass in the subject provides no evidence of associa-tion, whereas ρ = 0.002 for the same categories in the agent (Table 6:11). Tables 6:14-15, giving the two distributions expressed in percentages, show that the marginal totals are only moderately different; the entries are much more so. Compared with Group I (Table 6:16), and indeed all other groups, Group II clauses have notably often free external clause relation (CL 5, p. 81) and verbal groups with a Type b auxiliary (W 24, p. 91).

GROUPS I I I / I V have zero agents as their joint key feature: they may take agent

PASSIVE CLAUSE CLASSIFICATION : COMPUTATION AND RESULTS 1 2 7

TABLE 6:13 Criteria S 48, S 50, Groups I-VI (nonsignificant)

Scount S mass Total

S animate 25 2 27 5 inanimate 84 15 99

Total 109 17 126

TABLE 6:14 Criteria S 48, S 50, Groups I-VI

S count 5 mass Total

S animate 19.8% 1.6% 21.4% S inanimate 66.7% 11.9% 78.6%

Total 86.5% 13.5% 100%

TABLE 6:15 Criteria S 48, S 50, Groups I, II

Ag count Ag mass Total

Ag animate 12.9% 25.8% 38.7% Ag inanimate 54.8% 6.5% 61.3%

Total 67.7% 32.3% 100%

TABLE 6:16 Criterion W 24, Groups I, II (p = 0.007)

Type b auxiliary No Type b auxiliary Total

Group I 2 17 19 Group II 7 5 12

Total 9 22 31

extensions with by (Ag 92, Ag 93, p. 108) and subsequent active transformation (CL 11, pp. 83f.)· Apart from its key features, Group III has hardly any overall individual characteristics. Group IV, however, has the following additional features: its clauses never have adjuncts (Ad 99, p. 109), and their external clause relationship is almost invariably bound (CL 1, p. 78). In respect of these two features combined, Group IV is at variance with all the other groups. Furthermore, its subjects are often pronouns (S 46, p. 98), and just under half of them are coreferent with the object or predicative of a contiguous clause (CL 9, p. 82). The participles of these clauses, more often than those of other group clauses, can operate alone as nominal postmodifiers (W 44, p. 97);

128 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

TABLE 6:17 Criteria W 38, W 40, W 41, Groups IV, V (p = 0.00001)

Qualification No qualification Total

Group IV 0 17 17 Group V 6 0 6

Total 6 17 23

TABLE 6:18 Criterion W 30, Groups IV, V (p = 0.0002)

Commutation No commutation Total with feel with feel

Total

Group IV 0 17 17 Group V 5 1 6

Total 5 18 23

they do not admit of qualification with quite (W 38), rather (W 40), very (W 41) (Table 6:17). In this respect Group IV is like Group II; Groups I and III have some admissions of qualifiers, whereas Group VI has a scatter. Similarly, like Group II, the passive auxiliaries of Group IV do not commute with seem (W 29, p. 94). Here, Group I usually admits seem (12:7), Group III usually not (13:37 with 3 queries). Group IV clauses are generally shorter than clauses in other groups, and their functions are essentially those of postmodifying relative clause or clausal adjunct. In view of their similar relations to other groups (see Figure 6:3), Groups III and IV may be conflated, with Group IV as the 'marked' subgroup.

GROUP V has its typical features: the potential dual transformation into extensive and intensive active clauses (CL 12-3, pp. 83 f.); most of its Type d auxiliaries commute with seem (W 29), and with feel (W 30, Table 6:18); all its participles permit co-ordination with adjectives (W 36) and qualification (W 38, W 40, W 41, Table 6:17), but cannot usually operate alone as postnominal modifiers (W 44). This group differs from all other groups in that its external clause relation is normally free (CL 5, p. 81), and its subjects typically animate (S 48, p. 98). Subjects are, furthermore, frequently pronominal (S 46), and they are not coreferent with an object or predicative in a contiguous clause (CL 9).

GROUP VI does not usually admit of agent extension (Ag 92, p. 108). There are four exceptions, all of them doubtful cases: OTUs 40,44,123,124. Nor do they take direct active transformation (CL 11; one exception: OTU 124). The passive auxiliaries of this group normally commute with seem (W 29); it is interesting that its complex verbal group never has a Type a auxiliary (W 22, p. 91; Table 6:19). It has sometimes potential transmutation (CL 20, p. 90; Table 6:20) and potential coordination with adjectives (W 36). This is the only group besides Group V whose participles may often

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 129

operate alone as prenominal, but infrequently as postnominal, modifiers (W 42, W 44, Table 6:21).

TABLE 6:19 Criterion W 22, Groups III, VI (p = 0.02)

Type α No Type a Total auxiliary auxiliary

Total

Group III 11 42 53 Group VI 0 21 21

Total 11 63 74

TABLE 6:20 Criterion CL 20, Groups Ijll, VI (p == 0.05)

Potential No potential Total transmutation transmutation

Total

Groups I/II 2 29 31 Group VI 6 13 19

Total 8 42 50

In addition to the characteristics of the agentful groups (I/II) with regard to each other or to other groups, we may assess the probabilities of the subject and agent distributions within each group. On the basis of our sample, there is good statistical evidence to show that subjects and agents in agentful clauses differ from each other significantly in many respects.

Agents tend to be longer than subjects. In the Taxonomic Passive Corpus, subject lengths varied from 0 to 26 words, with a mean of 3.3 words and a mode of 1 word (in the Minor Passive Corpus the lengths were 4.96 words and 1 word, respectively); agent lengths varied from 2 to 40 words, with a mean of about 8 words, and a mode of 6 words (8.5 and 2 words, respectively, in Corpus II). Curves showing frequency and length of the two elements are quite different in that the subject curve decreases more or less geometrically from 1 word on, whereas the agent curve is irregular (cf. Section 4.36, pp. 52ff.).

As many as one third of the clauses in Corpus III and one half in Corpus II (see Section 4.35) have coordinate agents, whereas coordination is rare in subjects of any group (6 out of 128).

In agentful clauses, subjects tend to have definite and count nouns, and agents to have indefinite and mass nouns (Tables 6:22-24). Subjects are usually nouns; agents are usually nouns or names, but never pronouns in the sample (Tables 6:25-30). The tables indicate that the chances of having, for example, a name as subject and a pronoun as agent are very small indeed. The probability of Table 6:30 is 1/50,000!

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

TABLE 6:21 Criteria W 42, W 44, Group VI (p = 0.0009)

Possible Not possible Total

VedN 11 10 21 Ν Ved 1 19 20

Total 12 29 41

TABLE 6:22 Criteria S 50, Ag 84, Groups I, II (p = 0.03)

Mass Count Total

Subject 3 27 30 Agent 10 21 31

Total 13 48 61

TABLE 6:23 Criteria S 51, Ag 85, Groups I, II (p = 0.05)

Definite Indefinite Total

Subject 26 4 30 Agent 19 12 31

Total 45 16 61

TABLE 6:24 Criteria S 51, Ag 85, Group I (p = 0.0006)

Definite Indefinite Total

Subject 18 1 19 Agent 8 11 19

Total 26 12 38

TABLE 6:25 Criteria S 48, Ag 81, Group I (p = 0.01)

Pronoun Noun or name Total

Subject 6 13 19 Agent 0 19 19

Total 6 32 38

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

TABLE 6:26 Criteria S 48, Ag 81, Group II (p = 0.04)

Pronoun Noun or name Total

Subject 4 7 11 Agent 0 12 12

Total 4 19 23

TABLE 6:27 Criteria S 48, Ag 81, Groups / / / / (p = 0.0003)

Pronoun Noun or name Total

Subject 10 20 30 Agent 0 31 31

Total 10 51 61

TABLE 6:28 Criteria S 48, Ag 81, Groups / / / / (p = 0.003)

Noun or pronoun Name Total

Subject 30 0 30 Agent 23 8 31

Total 53 8 61

TABLE 6:29 Criteria S 48, Ag 81, Groups / / / / (p = 0.002)

Noun Pronoun Total

Subject 20 10 30 Agent 23 0 23

Total 43 10 53

TABLE 6:30 Criteria S 48, Ag 81, Groups I/II (p = 0.00002)

Pronoun Name Total

Subject 10 0 10 Agent 0 8 8

Total 10 8 18

132 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

It is of course impossible to give an estimate of the frequency, since there are no occurrences of pronominal agents in the Taxonomic Corpus.

6.4 THE DIAGNOSTIC KEY

The diagnostic key is meant to be an economical method, arrived at via the OTU classification, for obtaining a sufficiently detailed yet realistically coarse classification of new instances of passive clauses. It is conceivable that several different key features might give roughly the same classification. The diagnostic key proposed here should therefore be seen only as a first step towards a better key, which can evolve only from further experience.

Several factors must be taken into account when we set up a diagnostic key. Besides fulfilling the basic requirement of providing the means of a 'good' classification (as stated in Section 5.1, pp. 72 f.), a diagnostic key should be simple to use. The selection of key criteria is largely conditioned by practical convenience. One factor is, necessarily, the ease with which the criteria can be applied. A criterion which applies very rarely or tends to produce different responses from different informants, or from the same informant on different occasions, may not be very pernicious to the OTU classification, but it may not satisfactorily be used for diagnostic purposes.

We found that the groups in the output had a small number of individual charac-teristics (key features), which defined all, or almost all, the members of the groups. Hence, these criteria have a high information content, and it is natural that we should use them in the diagnostic key. Unlike the OTU criteria, which were unweighted (see Section 5.1), key criteria are weighted and ordered.

The output of the taxonomic experiment showed that the OTUs formed clusters on a scale. The diagnostic key proposed here is designed with a view to bringing out a cline. However, it has not been felt necessary or even desirable to follow the output order slavishly in constructing the key.

The proposed classification comprises six major classes (α, β, γ, δ, ε, ζ), each of which may admit of varying degrees of further breakdown.

6.41 Classes a and β

Class α and Class β, which are at one end of the scale, have close voice relation, with a systemic one-to-one correspondence between passive and active (see Section 3.7, p. 33). The active transformation rule can be applied without systemic restrictions to generate active transforms. It is important to emphasize the very considerable restrictions not subsumed under 'systemic' here: they include, for example, sequential constraints (see CL 16, p. 85), case alteration for pronouns with object forms (he.hirn, etc.), adjunct placement, and stylistic imbalance (see CL 18e, p. 87). The key features of

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 133

Class α and Class β are, respectively, expressed animate by-agent and expressed inanimate fry-agent. Since Class α corresponds to Group II and Class β to Group I, see Section 6.32 (pp. 125f.) for 'concomitant features'.

Class α ('animate agent passives'):

He was given this puppy by a farmer in the Welsh hills. (Ml. 18) Some work has also been done, notably by Broadhead and Pauson (1955), on the arylation of ferrocene and phenylferrocene with various diazo compounds. (J4.27) But more drastic action is to be demanded in Parliament by Mr. Fred Peart, Socialist Shadow Minister of Agriculture. (E2.30) if he'd been /introduced :properly and decently to :army life# he would /"not have been followed by . '.press photographers and :press reporters# (Bl.l)

Class β ('inanimate agent passives'):

/mäy I -say# we've been "/well rewarded# [by our /visit to IBognor NlRegis#J (B4.75; [narrow])

The removal of large quantities of water is an important problem in all freshwater animals and is facilitated by a high pressure in the kidneys. (J 1.92) ... French, which, as already observed, remained the first language until after Napoleon, when it was exceeded by German, which in its turn has recently been surpassed by Russian. (1.31) Pratt bowled rather longer than was justified by results, ... (F.23)

6.42 Class β/γ

One reason for reversing the output order (Class α = Group II, Class β = Group I) is that there is frequently structural indeterminacy between Class β and Class γ (which comprise clauses with potential fty-agents): prepositional fry-phrases with inanimate nominal parts are often 'Janus-agents' (see Section 5.81, pp. 102ff.), i.e they may be interpreted either as agents (Ag), in which case they are Class β clauses, or as other adjuncts (Ad), in which case they are Class γ clauses. Such clauses will be classified as β/γ.

'The purpose of the Government Bill to abolish resale price maintenance is to ensure that the full force of competition is brought to bear and that profits and dividends are not kept at an artificially high level by maintained prices', he said.

(El.23) These sex differences, which develop shortly before spawning, can also be initiated by injection of anterior pituitary extracts ... (J1.97) Let us say that anything which is introduced, or can be introduced, into a remark by an expression is a term ['term' in italics] (K.146)

134 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

6.43 Class γ

Class γ ('agentless passives') is the central and most frequent passive clause class. It corresponds, roughly, to Groups III/IV, and has no expressed fry-agent, but may have 'direct' agent extension (which is usually animate) with subsequent systemic potential active transformation. 'Direct' denotes that agent extension and active transformation are possible within the same tense (cf. Class ε below).

No subdivision will be made here corresponding to Groups III and IV. The main reason for this deviation from the output results is that these two groups are more closely related than any other two (see Section 6.32, pp. 126ff.). This is all the more remarkable in view of their large sizes. Furthermore, the one feature which distin-guishes all Group IV members from Group III, and from any other group, is absence of adjuncts, and we would feel reluctant to make absence of what is usually an optional clause element a key feature in the final classification (see Section 3.1, pp. 25 f.). On the other hand, further predominant features, such as bound external clause relation, indicate that we may be justified in regarding Group IV as a functionally important subclass of Class γ.

Ί could have got a job in the war. I was offered a decent one, in the B.B.C....' (M1.21)

The person convicted was fined only a pound.' (E2.45) I life is lived. in the present # and /one . Iworks . for the füture# (B2.26) The invitation hadn't been very felicitously phrased, but it was entirely cordial.

(M6.13) Many varieties of laterals are heard in English, ... (1-41) Order has been restored without bloodshed and without concessions, ... (G.2) The word or its synonyms mean quite different activities according to the context and manner in which it is used. (H.13) At last we have high-fashion knitwear that can be washed by hand and in the washing-machine... (C.334)

6.44 Class δ

Class δ occupies an intermediate position on the 'passive scale' in that its members have both verbal and adjectival properties. Their verbal character is manifested in their potential transformation into extensive active clauses; their adjectival character in their potential transformation into intensive active clauses (see Criterion CL 13, pp. 83 f.), and in their ability to take coordination with adjectives, qualification, and lexically marked auxiliaries {feel, seem, etc.). Members of this class have, typically, animate subjects. Class δ is divided into two subclasses.

Class δ2 corresponds closely to Group V (see Section 6.32, p. 128). These 'emotive passives' usually have, or can have, quasi-agents (see Section 5.81, p. 104), particularly with the lexically unmarked auxiliary be (see p. 93).

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 135

Members of Class δ1 (we may call them 'attitudinal passives') do not take quasi-agents but are instead remotely related to Class β, and hence indirectly to γ, since they may occasionally have fry-agents which are typically inanimate. This is reflected in their position above δ2 in the classification.

Examples of Class δ1 ('attitudinal passives'; cf. Bolinger 1961b, 379ff.):

We are encouraged, therefore, to use the radar data to obtain drop-size distributions. (J5.272)

... when Watson felt compelled to call up Pratt, the crisis appeared to have passed. (F.23)

I've always been far too inclined to treat important objects as part of my own petty existence. (M5.29)

Examples of Class δ2 ('emotive passives' with quasi-agents have also been illustrated in Section 5.81):

Gerald was suddenly very annoyed. (M5.36) I'm entranced with the whole idea. (LI. 14) Mrs. Cressett ... was uncertain but interested. (M5.78) For a moment Rood looked offended. (M6.40) Indeed, there soon emerged a group of psychologists who were interested in problem-solving and who were critical of Thoradike's methods and theory. (H.35) ... it struck him that the girl was now rather frightened. (M6.102)

6.45 Class ε

Class ε ('nonagentive passives') is the most multifarious class of all. This class repre-sents the end of the 'passive scale' where the relation to active is most tenuous. When asked if members of this class may have agent extension and active transformation, native informants often vary considerably in their responses. Such reactions must be taken as evidence that, towards this end of the passive scale, it is no longer realistic to consider these clauses in terms of voice transformation potential. In Class ε clauses, 'external agents' are either unlikely or impossible, because no 'agent' (used in its non-technical sense, i.e. 'actor', 'doer', 'logical subject', etc.) is conceived of. Many such clauses may be said to have their agents 'internalized' (cf. Bolinger 1961 b, 378), which is manifested by such actual or potential features as coordination of their past participles with adjectives and other predicatives in equative clauses (W 36-7, p. 96), lexically marked auxiliaries (W 26-7, p. 94), and qualification (W 34, pp. 95f.). Moreover, ε-clauses may frequently have transmutation (CL 20, p. 90).

Class ε includes 'statal passives', which are related to Class γ in that they may take agent extension and active transformation, but only 'indirectly', i.e. a change of tense must take place. (See Criterion CL 17 and the subsequent discussion, pp. 85f.)

136 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

In the experiment, members of this subclass did not form a clear-cut group of their own but were found among the least typical members of Group III (see, for example, OTUs 107, 53, 105).

There he stood, in this difficult atmosphere, gleaming in his white coat, a tray in either hand on which was tastefully arranged a bowl of turtle soup, a dish of olives and celery and sliced tomatoes and a glittering array of tools to tackle further delights to come. (M4.58) Cavendish inquired with grave courtesy,'Is the thesis finished, Doctor?' (M4.57) the /general# who /said my \flanks are :türned# my /centre is bröken# /1 shall attäck# (B4.16) 'Sandra - listen' 'Not another word. The die is cast.' (LI. 17)

In the following example, the precise passive value of was posed is indicated by the aspectual contrast of the active had posed:

She was posed - for it was much as if her companion had deliberately posed her for his own pleasure and Appleby's - in a small shifting dapple of sunlight and shadow. (M6.97)

The subclassification of Class ε will not be further pursued here, apart from making a distinction between clauses with mutative and nonmutative values. The proper context of a detailed description of Class ε (as well as of δ) would include equative and factitive clauses (see Section 3.8 and Chapter 8).

Class ε1 has mutative value, and the auxiliary may be unmarked {be) or marked (become, get, etc., see Section 5.51, pp. 93f.).

This should be compared with Fig. 2, which shows the diffraction pattern when a stearic acid monolayer has been retracted on the silver, with the crystal in the same azimuth. (J5.301) Not until the sun was moving down over the blue hills of Donegal was the destiny of these two matches decided, and of all the long day the last hour must come first.

(F.28) As the modern world becomes more highly industrialized and mechanized,... (1.33) 'Think of the episode of supreme savage comedy when Hamlet jumps into her grave and gets jammed in it.' (M6.16) As a consequence the Order Paper gets silted up, questions get stale, and the intervals between a particular Minister's appearance at the Dispatch Box become longer and longer. (G.9) The significance of the mystery, however, was lost on Clarissa. (M5.21)

Lexically marked auxiliaries (notice gets silted up, get stale, become longer in three sequentially related clauses in the last but one example) emphasize the decreased voice relationship, since the agentive α-, β-, γ-classes normally take only be as passive auxiliary.

PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS 137

Many of the lexical verbs in these clauses are mutatively marked by their suffixes, as in specialize, departmentalize, etc.

Class ε2 has nonmutative value, and the auxiliary may be unmarked (be) or marked {look, remain, etc.).

It was not their sense of patriotism that was involved. It was their sense of smell. (LI.20)

... Nordic, Mediterranean, and Alpine, once discernible races, are now intermingled everywhere to a greater or less degree. (1.21) 'As far as poetry and that sort of thing is concerned, Lewis's taste simply didn't exist.' (M6.65) Often an older firm is disproportionately strong in foreign business, while a younger firm may be concentrated rather in the home industrial financing ... (G.10) They are most contented as they are. (LI.57) '... suppose that she was simply mistaken about the ink.' (M6.81) She was not normally given to calling people 'dear' or to leering at them, ...

(M5.18)

6.46 Class ζ

Class ζ comprises compounds generally, where the isolation from the lexical verb is morphologically marked (see Section 2.73, pp. 21 f.). Although compounds are morphologically isolated, they are not necessarily syntactically isolated with regard to voice, in so far as they may occasionally have agents (cf. example in Section 2.74 and below). One compound in the Taxonomic Passive Corpus (OTU 95) was in fact grouped with Group III. But most compounds in the material are nonagentive, and have, frequently, lexically marked auxiliaries and coordination with equative predica-tives. The lexically marked auxiliary and the compound participle are often in close collocation (pass unnoticed, come unstuck, etc.).

But Cavill was unimpressed by this sally. (M6.54) The significance of these photomechanical changes is unknown but they demon-strate that the pineal cells are sensitive to light. (J 1.103) 'Not all the parts of a thought can be complete; at least one must be "unsaturated" or predicative·, otherwise they would not hold together.' (K.152) In relying upon the grammatical phrases, 'substantival expression' and 'expres-sion containing a verb in the indicative mood', the distinction seems both parochial and unexplained ... (K. 148) Exodus flights are, no doubt, made also by individuals from the smallest possible populations but pass unnoticed. (J5.349) But her question remained unspoken. (M5.45) As long as clause four stands unaltered and as long as the variably influential section of the party which subscribes to a full-blooded interpretation of it goes undefeated, so long will the contradiction remain. (G.l 1)

138 PASSIVE CLAUSE CLASSIFICATION: COMPUTATION AND RESULTS

... the South Africans' final anxieties had been self-inflicted. (F.25) Old Professor Wali-Anu will be heart-broken. (LI.8) Pig breeders have not been as panic stricken as some of their spokesmen, ... (G.5) It was hoped ... that recent agitation on those lines was ill-founded and artificial.

(G.5)

6.5 THE PASSIVE SCALE A N D VOICE RELATION

The classes we have set up are not clearly delimited, and they differ in size and degree of interrelation: α and β have much in common; so have δ and ε, but the latter two classes also accommodate clearly distinct subclasses; δ, for example, has a cline within itself, one end of which is ß-like, the other ε-like, and so on. A representation in a single dimension can only inadequately depict the complexity of their relation-ships.

The key in Figure 6:4 (p. 117)is a diagrammatic representation of the classification. It is designed to be used in applying the key criteria, but it is not intended to be anything more than a coarse guide to the major classes. Progression may take place only in the direction of the arrows, along an itinerary which is determined by a series of binary choices at the nodal points. The diagram should be self-explanatory, except for the double entries leading to Class δ: entry may be made either via 'potential extensive active transform' or via 'no potential extensive active transform'. The reason for this is that the criterion of potential extensive active transformation applies less readily to some members of Class δ than the small number of OTUs in Group V led us to believe. It seems therefore realistic to admit this fact by providing two paths leading to the same goal.

This weakening of our prime criterion - potential extensive active transformation -is only to be expected, as we proceed down the scale, away from the pole with close transformational voice relation. Passive clauses (in the wide sense that we have used the term 'passive') may be seen as related to active clauses but in different ways and to varying degrees. Agentive clauses, at the top end of the passive scale, are trans-formationally related to active clauses (transitive, ditransitive, factitive, see Section 3.8), whereas non agentive clauses, at the bottom end of the passive scale, have syntagmatic affinity with active equative clauses. In between the two extremes of the scale there are intermediate classes with mixed properties. Generally speaking, we may say that, as the degree of transformational voice relationship decreases, the degree of syntagmatic relationship increases.

7

T H E U S E O F T H E P A S S I V E

This chapter will be devoted to a study of the use of the passive voice in a fairly extensive selection of contemporary English, the Major Passive Corpus (see Section 1.5, pp. 8 f.). Passives that occurred in this material will be classified according to the system set up in Chapter 6, and discussed with regard to clause types, verbal group types, and texts.

7.1 FIVE PASSIVE CLAUSE TYPES

In Chapter 3 we discussed some major clause types, particularly in the light of their voice relations. Having narrowed the scope to the passive, we shall now introduce a more detailed clause typology for this term in the voice system. Although the clause names PASSIVE TRANSITIVE (ext-0), PASSIVE DITRANSITIVE (ext-C), and PASSIVE FACTITIVE (int-C) are unsatisfactory in that they are modelled on the names of their corresponding active clauses and hence imply, misleadingly, that there is always a direct active-passive relationship, it seems nevertheless useful to have an indication in easily convertible terms of potential clausal voice relation, provided the transforma-tional constraints are also borne in mind.

We can achieve an increase in the delicacy of our analysis by subdividing passive transitives into clauses which have and clauses which do not have a complement-like prepositional adjunct in close collocation with the verb. The clauses which do not have such collocation retain the original clause label, whereas the others are further distinguished according to their internal extensive and intensive clause relations, as these terms have been defined in Section 3.7 (pp. 33 ff.). The three types may be illustrated by the following examples.

PASSIVE TRANSITIVE (ext-0):

Many famous American and English philologists were trained in Germany in the nineteenth century, ... (1.16)

PASSIVE SEMIDITRANSITIVE (ext-A):

'How are we going to jockey Middleton into accepting the editorship that will undoubtedly be offered to him?' (M5.27)

140 THE USE OF THE PASSIVE

PASSIVE SEMIFACTITIVE (int-A): ... the London School of Economics and Political Science has come to be regarded as a thoroughly English academic institution. (G.6)

The semifactitive type of clause differs from the sem'ditransitive in having an equative relation between its subject and the head of the prepositional phrase.

-* the LSE is a thoroughly English academic institution

Consequently it has, for example, the characteristic number restriction, which does not apply to extensive clauses.

*the LSE has come to be regarded as thoroughly English academic institutions -*• the editorship ... will undoubtedly be offered to them

The close linking of verb and preposition is perhaps particularly striking in those verbs which have the alternative possibilities of retaining or omitting the preposition, as for example:

(ext-C) The letter was given her. (ext-C") She was given the letter. (ext-A) The letter was given to her. (int-C) This was considered a duty. (int-A) This was considered as a duty.

The first two transforms represent the primary and secondary passive ditransitive clause types; the third is like the primary except for the preposition to. Despite obvious parallelism between the two, such a phrase will be called A and not C, i.e. it will not be considered an indirect object (see Section 3.9, p. 37).

The fact should not be overlooked that the ίο-phrase also has affinity with other prepositional phrases which are not similarly related to complement-structures. Compare, for example,

The letter was addressed to her.

which has no transforms

""The letter was addressed her. *She was addressed the letter.

On the other hand, these prepositional phrases should be distinguished from adjuncts with clear adverbial value and less cohesion between verb and preposition, as in

The letter was addressed in the post-office.

There is no ready way of establishing the diverse values of prepositional phrases, and for the present purpose we shall have to be content with a simple dichotomy into what may be called DOLOSE (as in addressed to her) and ΛΌΡΕΝ (as in addressed in the post-office). The criterion used for this separation will be the test with wA-questions

THE USE OF THE PASSIVE 141

(see Criteria A 66 and A 69 in Section 5.8, p. 102). Class Λ CLOSE has question-forms with wA-pronoun (whojm/ or what) + preposition, whereas Class Λόρεν had question-forms with wA-adverb (where, when, how, etc.) ± preposition. The question-form should have unmarked tonicity and falling tone, as in

/who(m) was the letter addressed -to# /where was the letter addressed#

In case of overlap, i.e. when either question-form seems equally natural, the guiding principle has been to consider the pronominal criterion as superordinate to the adverbial criterion. This applies, for example, to such phrases as

She was considered as a friend.

which may have either of the following question-forms (cf. Section 3.73, pp. 34f.):

What was she considered as! How was she considered?

Λ c l o s e is intermediate between C and Λ ό ρ ε ν . In clause designation, this will be indicated by '-A' attached to the extensive (ext-A) and intensive (int-A) clause labels; in the clause name, by the prefix 'semi' added to the corresponding major clause type ('semiditransitive' and 'semifactitive'). Λόρεν (which may occur in any clause type) will not be specially registered here, and hence no distinction is made between clauses with or without such adjuncts.

7.2 THREE MAJOR PASSIVE CLASSES

Excluding compounds from the present study, we may conveniently deal with the following three major passive classes separately: agentive, quasi-agentive, and non-agentive passives.

7.21 Agentive passives

Agentive passives (Classes α, β, β/γ, γ; see Sections 6.41-3, pp. 132if.) probably correspond fairly closely to the general concept 'passives' in English. The centrality of agentive passives to the passive construction as a whole is evidenced by their high frequency in our material: 2696 occurrences, i.e. almost three quarters of all collected passive clauses. Of our agentive passives, 80 per cent are agentless (Class γ), and the remaining 20 per cent of agentful clauses are equally divided between the animate (Class a) and inanimate (Class β) by-agent classes. This proportion of γ-passives is only slightly higher than for the Minor Passive Corpus, which had over 76 per cent (497 out of 650 occurrences; see Section 4.11, p. 40). It conforms with the figures given by Jespersen: 'between 70 and 94 per cent' (1924, 168), and 'over 70 per cent of passive sentences found in English literature contain no mention of the active subject' (1933, 121). All the agents in our a- and ß-passives are fey-phrases. The figure for ß-passives also includes the ambivalent class of Janus-agents (β/γ, see Section 6.42,

142 THE USE OF THE PASSIVE 2000

Fig. 7:1. Frequencies of the five passive clause types in the agentive classes.

p.133), which accounts for over one fifth of the 265 occurrences with inanimate iy-agents.

In respect of clause type, 73 per cent (1970 occurrences) of the agentive passives are transitive, 9 per cent (249) semiditransitive, 3 per cent (74) ditransitive, 5 per cent

THE USE OF THE PASSIVE 143

TABLE 7:1 The use of clause types in agentive passives

Agentive Class α ^ v class Class α Class β Class β/γ Class γ Total

Clause Class β Class β/γ Class γ

type \

Passive 216 195 55 1504 1970 transitive (8%) (7%) (2%) (56%) (73%)

(ext-0)

Passive 16 7 0 226 249 semiditransitive (1%) (0%) ( · ) (8%) (9%) (ext-A)

Passive 8 0 0 66 74 ditransitive (0%) (•) (·) (2%) (3%) (ext-C)

Passive 9 1 0 125 135 semifactitive (0%) (o%) (.) (5%) (5%) (int-A)

Passive 17 7 0 244 268 factitive (1%) (0%) (.) (9%) (10%) (int-C)

Total 266 210 55 2165 2696 (10%) (8%) (2%) (80%) (100%)

(135) semifactitive, and 10 per cent (268) factitive (see Figure 7:1). We shall now illustrate some of the passive clause types in relation to the passive classes. The frequencies are given in Table 7:1; the verbs are listed in the Appendices (pp. 167ff.).

The predominance of the PASSIVE TRANSITIVE clause type (ext-0) is particularly notable in agentful constructions, where as many as 88 per cent (466 occurrences) are of this type (see Appendix la, pp. 167ff.).

(a) The children - who have not been told the seriousness of their father's illness - were met by their mother, ... (E2.18)

(a) So far I have been concerned solely to set out certain associated distinctions, or aspects of one distinction, which have historically been made or recognized by philosophers. (K.142)

(β) The obscurity of the words baffled the audience; and they were not much helped by Mrs Middleton's demonstration of a bird in flight. (M5.87)

(β) The sucker is bounded at the edges by a series of lips, ... (J1.88) (β/γ) Loss of inhibitory power by avenacin on enzymic hydrolysis was confirmed

by growth experiments ... (J5.325) (γ) 'You wish that Packford had been murderedΤ (M6.60) (γ) Four decisive ideas in the Nathan Report have been introduced. (G-8) (γ) 'Her parents were killed in the Cafi de Paris raid,' said John. (M5.79)

144 THE USE OF THE PASSIVE

(γ) In situations in which the functional characteristics are grasped, the solution can be transferred and applied in a variety of situations. (H.50)

(γ) ... if the academical world insisted on its narrow limits, then other means of disseminating the truth must be found. (M5.17)

Passive transitive clauses include not only clauses with referential subjects but also those with proleptic it anticipating a finite or nonfinite clause, which operates as complement in active clause structure. Hence, proleptic it as passive clause subject disappears in transformation to active:

-*• numerous workers have demonstrated that ...

(a) It has been demonstrated by numerous workers that the shape of the pW curve of myosin adenosine triphosphatase at 25° C. is not altered from that shown in Fig. 1 at higher ionic strengths, ... (J5.296)

(γ) Yesterday it was learnt that a rocket had, for the first time, been destroyed in mid-flight by another rocket. (G.13)

(γ) Biologically it may be found that there are few differences between an African negro and a white Scandinavian. (1-23)

We also include proleptic //-clauses which are followed by direct speech, and those which are 'parenthetic' (see Appendix lb, pp. 183ff.).

(γ) It might be asked at this point: 'Why doesn't an examination of logic help us to discover the methods we use when we think out problems?' (H.47)

(γ) The first [point] was how the greatly increased number of Africans which it was envisaged would be admitted to the new legislature should be elected.

(G.12)

Clauses with it anticipating a nonfinite verb clause are rare (5 γ) as compared with those anticipating finite verb clauses (4 α and 116 γ). See Appendix Ic, p. 185.

(γ) 'it may be decided not to rely exclusively on fixed-site missiles.' (G.17)

PASSIVE SEMIDITRANSITIVE clauses (ext-A) have prepositional phrases as adjuncts where the prepositions collocate closely with the verb. The collocations are listed in Appendix Ha, pp. 185 ff. There are 16 α-, 7 β- and 226 γ-passives in the corpus.

(a) '... Packford ... was then robbed of it by someone who killed him in the process?' (M6.43)

(a) It [this word] was also applied by Shakespeare to human beings ... (1.24) (β) Adults, however, were found mostly on the upper surfaces of stones, shells

and glass, to which they are confined by their need for a smooth surface. (J5.328)

(γ) Bureaucratic clerks in all their hideous, inhuman behaviour were charged with the deed·, (M5.12)

THE USE OF THE PASSIVE 145

(γ) It is surprising that more has not been heard of him, although shortly after the war he was almost in the Walker Cup side. (F.28)

In this clause type, there are only two occurrences of proleptic subject (it) anticipating a finite verb clause (see Appendix lib, p. 188).

(γ) ... it hadn't in fact been mentioned to him that she was a domestic employee of the genteel variety. (M6.86)

PASSIVE DITRANSITIVE clauses (ext-C) constitute the least frequent type in the material (8 a, 66 γ). There are three exponent classes for C in the corpus: nominal, nominal + preposition, and finite clause. Nominal as complement (see Appendix ilia, pp. 188 f.):

(α) He was given this puppy by a farmer in the Welsh hills. (Ml.18) (γ) We aren't allowed a tragedy nowadays without a banana-skin to slip on and

make it funny. (L2.42) (γ) By using removable blocks different problems can be set the animal. (H.32)

Nominal + preposition as complement (see Appendix IHb, p. 189):

(a) But it [the house] had been made a mess of, comparatively late in its history, by some owner with a taste for the Gothic. (M6.69)

(γ) ... she might judge she was being made fun of. (M6.61)

Finite verb clause as complement, including the parenthetic type (see Appendix IIIc, p. 189).

(α) Ί am assured by Packford's physician that our friend had no rational occasion to fear for his health.' (M6.35)

(γ) Ί was told you wanted to see me,' Gerald said. (M5.32) (γ) 'Suicide is a crime, I have been told.' (M6.85)

In common with semiditransitive clauses, PASSIVE SEMIFACTITIVE clauses (int-A) have close collocation of verb and preposition, but whereas the former display a variety of prepositional exponents, semifactitive verbs collocate almost exclusively with as. There are 135 occurrences and three kinds of prepositional adjuncts: preposition + nominal, preposition + present participle, and preposition + adjective, or past participle.

preposition + nominal as close adjunct (see Appendix IVa, pp. 189f.):

(α) Before the learned journals could lumber into reasoned appraisal, the whole thing had been accepted as gospel by the common reader and become estab-lished as a plain fact of literary history. (M6.17)

(γ) 'You'// be classed as a second-rater.' (M5.38) (γ) ... he could see no reason why Radford's time over the longer distance

should not be accepted as a world's record. (F.30)

146 THE USE OF THE PASSIVE

Preposition + adjective or past participle as close adjunct (see Appendix IVb, p. 191):

(a) they [those who were killed in the First World War] are < remembered# . /with grief . as young# . and /gäy# and /beautiful# . by /those who survived -them#> (B2.16; <narrow and rhythmic»

(γ) It is, for example, possible to argue that in these regions Communist policy should no longer be treated as united or consistent. (G.10)

Preposition + present participle as close adjunct (see Appendix IVc, p. 191):

(a) Haemopoietic tissue occurs in the intestinal wall of the larva and this has been regarded by some as representing the spleen. (J 1.93)

(β) ... and Barber was deceived by one of Goddard's faster balls into attempting a too ambitious cut. (F-25)

(γ) [the three people] /have been ar!rested# . and /chärged# with /[trying to Nsmüggle] a : [Czech Nnätional] out of the ci>untry# (D.9)

(γ) Every few months the gap at Geneva is reported as being narrowed. (G.10)

PASSIVE FACTITIVE clauses (int-C) have C exponence similar to A in Type int-A apart from absence of preposition: nominal, present participle, adjective, past participle, etc. An additional, and highly frequent, exponent class is the fo-infinitive. In all, there are 17 a, 7 β, and 244 γ.

Nominal as complement (see Appendix Va, p. 191):

(γ) He used to be called'Crossbones'. (L1.34) (γ) The 'return' flight has been considered a necessity without which a species

would lose its breeding sites. (J5.349)

Adjective or past participle as complement (see Appendix Vb, p. 192f.):

(α) ... only a limited amount of money will be made available by the British Transport Commission (or the Government) for pay increases, ... (G.8)

(β/γ) It [this cheese] is kept fresh ... by the special wrapping (C.96) (γ) The animal is placed hungry at R with the direct and previously learned

pathway R-X blocked at Β. (H.42)

To-infinitive as complement (see Appendix Vc, pp. 192f.):

(α) ... and the primary unimolecular dissociation which it represents was shown by a number of workers, by the use of free radical 'traps' to inhibit any subsequent reactions, to give rise to the formation of two benzoyloxy radi-cals ... (J4.36)

(γ) You're a famous author and all famous authors are expected to give lectures as a matter of course. (LI .5)

(γ) ... Latin grammar was thought to embody universally valid canons of logic. (1.12)

THE USE OF THE PASSIVE 147

Present participle as complement (see Appendix Vd, p. 193):

(γ) There was still a mass of stuff to delight any authentic student of bibliopegy who should be set browsing in the place; (M6.83)

(γ) ... the low-temperature product of rhesus rhodopsin can be seen regenerating following irradiation with λ500 ηιμ light. (J5.293-4)

7.22 Quasi-agentive passives

Quasi-agentive passives (Class δ) are intermediate between agentive and nonagentive passives and have mixed verbal and adjectival properties. Potential active transforma-tion demonstrates their verbal nature, and among the adjectival features are actual or potential qualifiers and lexically marked auxiliaries (see Section 6.45, pp. 135ff.). Quasi-agentive passives, which are further characterised by animate subjects, have two subclasses: δ1 ('attitudinal passives') and δ2 ('emotive passives'). The first of these is the more closely related to agentive passives in often having the possibility of extension with 6y-agent. The second class may have quasi-agents, which differ from ordinary by-agents in that they are, typically, inanimate and embrace not only phrases with prepositions other than by but also ίΑαί-clauses and to-infinitives (see Section 5.81, p. 104).

Compare, for example, the following series:

(δ1) He was (felt) entitled to do this.

[his position] entitled him to do this •to do this entitled him *it entitled him to do this ) ^ ^ ^ *it made him /feel/ entitled to do thisj

(δ2) He was (felt) embarrassed to do this.

*[his position] embarrassed him to do this to do this embarrassed him it embarrassed him to do this it made him /feel/ embarrassed to do this

Attitudinal passives, totalling 56 occurrences in the material, occur in a limited set of clause types: in semiditransitive clauses with to + nominal, in factitive clauses with to + infinitive, and in semifactitive clauses with to + present participle.

The Africans were heavily committed to the principle of'one man, one vote'. (G.12) 'For one thing, you haven't been obliged to earn your living ...' (Ml.20) In plural societies where we can still mould the future we are deeply committed to promoting the experiment of partnership. (G.4)

148 THE USE OF THE PASSIVE

Emotive passives constitute a homogeneous class, and it is interesting that its defining syntactic characteristics are matched by semantic similarity (as between for example flummoxed, non-plussed, perplexed, staggered, and startled) or even 'macrosemantic' similarity (as between flummoxed, etc. and interested, delighted, dissatisfied, obsessed, overwhelmed, relieved, worried, and so on). The verbs of this class have the property of being 'connected with a specific human "reaction"' (Nowell-Smith 1954, 85; cf. Chomsky 1965a, 50f.). Many of its verbs may of course appear in other passive classes, but then with different syntactic and semantic attributes. Compare the following two sentences:

(γ) ... the sting was taken out of Wakefield by an injury to Fox after only thirteen minutes' play which left him hobbling on the left wing for the rest of the game. Rollin, too, was hurt, and these misfortunes certainly greatly reduced Wakefield's striking power. (F.23)

(δ2) Hali would have been bitterly hurt if I had suddenly abandoned him and gone home with Chris. (LI.87)

Unlike the second clause with hurt, the first will not, for example, admit of a lexically marked auxiliary and a qualifier: * Rollin felt very hurt ... (but Hali would have felt very hurt ...); nor of transformation to an intensive active clause: *It made Rollin hurt ... (but It would have made Hali bitterly hurt ...). Semantically, too, the two verb uses are different: in the first clause, hurt refers to physical injury; in the second, to emotional injury. Other verbs, which are similar in having both these meanings, and corresponding grammatical correlates, are depress, exhaust, upset, wound, etc.

There are 227 occurrences of emotive passives in the material. In addition to the discussion in Sections 5.81 and 6.45 (pp. 135ff.), the following examples will adequately illustrate this class.

... it was the bowler rather than the batsman who was surprised. (F.27)

... at the moment I am more than disconcerted·, I am profoundly worried. (LI. 19)

... only on occasions were they [the umpires] not entirely satisfied with the absolute fairness of a delivery. (F.30) She was delighted with the phrase - it would serve so well to interpret the academic world to her literary friends. (M5.34)

7.23 Nonagentive passives (Class ε)

With quite a large proportion (18 per cent, 666 occurrences) of the passives in our material, agent extension is unlikely or impossible (see Section 6.46, pp. 137 f.). As we might expect, the distinction between nonagentive and other passives is far from clear-cut, and a more refined analysis than ours would arrive at further subclasses at this end of the passive scale. Most of the marginal cases have be as auxiliary with mutative value; when the auxiliary is marked, the possibility of agent extension appears to be even more remote than when it is unmarked.

THE USE OF THE PASSIVE 149

The only serious contender to be as agentive passive auxiliary is get. But it is rare: in the entire corpus there are not even half a dozen agentive gei-passives, none of which is agentful. Although it is probably symptomatic that the occurrences below are either from speech or written dialogue, there is no indication in our material that the ge/-passive is common in colloquial English. However, as we shall see in Section 7.4, the passive as a whole is by no means infrequent in the spoken material.

'But Cintio's yarn may [orig. ital.] have got translated into English, you know, without the translation's having survived.' (M6.20) this is a free country and criticisms should be made and they get treated with the contempt that they deserve (B5.6.8) 'Is it that place that got started in the thirties - I used to see it - called The Socialist Bookshop or something like that?' (Ml. 19)

The agentiveness becomes more doubtful in:

People ... splash about with washleather and water, exhaust themselves and get thoroughly soaked. (C.542) 'Think of the episode of supreme savage comedy when Hamlet jumps into her grave and gets jammed in it.' (M6.16) We know very well how some thinking sticks to the point and moves steadily towards its conclusion while other thinking runs round in circles or drifts off into blind alleys or gets bogged down\ (H.ll-2)

Become usually has the specific aspectual function of indicating gradual change, which is often enhanced by modification with more and more, increasingly, etc. and suffixation with -ize (conventionalize, industrialize, mechanize, organize, etc.).

Taking the argument still farther, it is becoming increasingly widely recognized that African numerical superiority in Kenya will have ultimately to be reflected in the constitution. (G-l) ... people who now speak one and the same birth tongue may, in fact, become so highly specialized and departmentalized in all their words and ways that they will no longer be able to understand one another ... (1.9-10) In many simple problems these four phases may become telescoped in a single, swift coordinated phase of activity - but in complex situations all phases may be distinctly observed. (H.54)

The majority of the lexically marked auxiliaries in Class ε are mutative {get and become), and there is only a scatter of nonmutative auxiliaries (remain, lie, look, etc.).

The post lay scattered on the table where Yves had searched it eagerly for possible evidence of cheques. (M5.73) ... long before half-time Wakefield's forwards, still tired perhaps after their efforts against Hull, looked well beaten. (F.23)

150 THE USE OF THE PASSIVE

TABLE 7:2 The use of verbal group types in agentive passives

Verbal group type Class α Class β Class γ Total

d 38 115 588 741 (14%) (43%) (27%) (28%)

D 117 61 536 714 (45%) (23%) (25%) (26%)

ad/AD 38 58 700 796 (14%) (22%) (33%) (30%)

bd 56 22 225 303 (21%) (8%) (10%) (11%)

BD 9 5 72 86 (3%) (2%) (3%) (3%)

cd 4 2 17 23 (2%) (1%) (1%) (1%)

CD 1 0 8 9 (0%) (·) (0%) (0%)

abd/ABD 3 0 14 17 (1%) (.) (1%) (1%)

other 0 2 5 7 (subjunctive, imperative) (.) (1%) (0%) (0%)

Total 266 265 2165 2696 Total (100%) (100%) (100%) (100%)

Compare this last example with beat as a γ-passive, where be does not commute with look'.

With the exception of the men's and women's doubles all the top seeds were beaten in the finals of the St. Annes lawn tennis tournament on Saturday. (F.23)

Just as get and become have the functions of indicating incipient action and gradual change, so these nonmutative auxiliaries mark the aspect of state. Be can also be used in 'statal passives' (see Section 5.51, pp. 92f.), but since it is aspectually unmarked, auxiliaries like lie, look, and remain may serve to indicate the aspect unambiguously. Nonagentive passives will be further discussed in the next chapter.

7.3 VERBAL GROUP STRUCTURE

The verbal group structure of all the agentive clauses in the Major Passive Corpus was analysed in terms of the types set up in Section 2.2 (pp. 10 f.). Conflating present and past forms (denoted by lower and upper case letters respectively), we find that over half the passive verbal groups are Type djD, and that the next commonest are Types ad/AD and bd/BD; there is only a scatter of other types. The frequencies of the major types were as follows (for a detailed breakdown into agentive classes, see Table 7:2):

THE USE OF THE PASSIVE 151

TYPE EXAMPLE

d/D (is/was eaten) ad/AD (may/might be eaten) bd/BD (has/had been eaten) cd/CD (is/was being eaten) abdjABD (may/might have been eaten) acd/ACD (may/might be being eaten) bed) BCD (has I had been being eaten) abcdjABCD (may/might have been being eaten)

OCCURRENCE PERCENTAGE

1455 796 389

32 17 0 0 0

(54%) (30%) (14%) (>1%) « 1 % ) ( · ) ( · ) ( · )

This overall distribution is very similar to the one for the Minor Passive Corpus discussed in Section 4.2 (pp. 43 if.). Since even in our present much larger material, there are no instances of the last six types, it is evident that they have a very low probability of occurrence. Our observation of a negative correlation in the Voice Corpus between Type c/C and the passive is strengthened by the rarity of this type in the present material (just over 1 per cent). We also noted a positive correlation between Types b, a, A and the passive: again, the high frequency of bd, ad, and AD in the present results would seem to support this observation. For the last two, the associative tendency is particularly striking in agentless clauses, and for Type bd in α-clauses.

The a/A-type auxiliaries are predominantly 'closed-class' (here defined as not taking a preceding auxiliary; see Section 2.33, pp. 14ff.). The frequencies for individual auxiliaries in Type ad/AD are as follows:

can may could will should must would be to might

CLOSED CLASS

shall 'd

ought to

177 102 84 79 69 54 48 45 27 14 4 3 2 2 1

OPEN CLASS

have to appear to be (uri)likely to be going to have got to be bound to seem to need to be able to come to tend to

52 7 6 5 4 2 2 2 2 2 1

7/ used to need

85

711

152 THE USE OF THE PASSIVE

There are about a dozen combinatory a/A-types, where open class auxiliaries have substructured verbal group combinations (see Section 2.33), as for example:

(A'+A" +D) He could remember only that he was sixty-four, could wonder only whether his growing lust was a simple case of enlarged prostate that would 'have to be dealt with'. (M5.13)

(A'+B+A"+D) To have detected the larger drops recorded by the radar in their observed concentrations the sampling volume would have had to be increased by a factor of 102-104. (J5.272)

Summing up, we may emphasize that the use of the passive verbal group is very much restricted to a fairly small number of simple structures: Types d/D, ad/AD, and bd/BD account for 98 per cent of all agentive occurrences in the Major Passive Corpus.

7.4 THE USE OF THE PASSIVE VOICE IN THE TEXTS

The use of the passive in the texts (listed in Section 1.4, pp. 6ff.) varies considerably, in respect of classes as well as overall frequencies. Taking the number of passive clauses per thousand words as the unit, the frequency ranges from 32 in one scientific text (J4) to 3 in the sample from television advertising (Text C), with an average of 11.3 for the whole corpus. At the same time, the frequencies for most of the texts of a text group are remarkably similar (see Table 7:3), and we can therefore achieve greater generality and a broader basis for subsequent statements by making confla-tions, not only of the texts of one group (such as El and E2) but also of related text groups (such as I and J). Instead of 28 individual texts, this will give us, for example, the eight text sets in Table 7:4. They are arranged in order from most to least frequent in their use of the passive voice, measured in terms of occurrences per thousand words.

The samples taken from scientific exposition have a clear lead in frequent use of passive constructions, with 23.8 and 21.5 clauses per thousand words for the learned (Text J) and popular (Text H) varieties, respectively.

Next follow the mass media in their covering of topical events in press and radio news (Texts Ε and D) and in editorials (Text G) with 16.4,14.2, and 15.8, respectively.

The two samples from prose dealing with arts subjects show rather different uses: the more popular introduction to linguistics (Text I) has 13.8 as against 9.4 for the learned philosophical discussion (K). Since the latter is a much Shorter text, their joint average remains as high as 12.7 passives per thousand words.

All the samples discussed so far belong to the broad category 'informative prose'. Next in our list follows a rather more disparate collection of material: speech, sports reports, and novels. With one exception (Text B3), the spoken material is surprisingly homogeneous in its use of passives, and the surreptitious recording (A) does not differ from unscripted discussions (B) in this respect. Their combined average frequency is 9.2, which is slightly higher than that for press sports reports (Text F).

THE USE OF THE PASSIVE 153

TABLE 7:3 The use of agentive passives in 28 texts

TEXT SIZE IN WORDS

OCCURRENCE OF PASSIVES

FREQUENCY PER 1 0 0 0 WORDS

A 5 , 0 0 0 4 0 8 . 0

B 1 5 , 0 0 0 4 0 8 . 0 B 2 5 , 0 0 0 4 4 8 . 8 B 3 5 , 0 0 0 8 0 1 6 . 0 B 4 5 , 0 0 0 4 5 9 . 0 B 5 1 5 , 0 0 0 117 7 . 8

C 2 8 , 0 0 0 8 3 3 . 0

D 5 , 0 0 0 71 1 4 . 2

E l 5 , 0 0 0 9 3 18 .6 E 2 5 , 0 0 0 7 1 1 4 . 2

F 3 0 , 0 0 0 2 6 9 9 . 0

G 3 0 , 0 0 0 4 7 4 15.8

Η 1 5 , 0 0 0 3 2 2 2 1 . 5

I 1 5 , 0 0 0 2 0 7 13 .8

J1 5 , 0 0 0 1 1 6 2 3 . 2 J 2 5 , 0 0 0 1 4 0 2 8 . 0 J 3 5 , 0 0 0 7 9 15 .8 J 4 5 , 0 0 0 1 6 0 3 2 . 0 J 5 1 5 , 0 0 0 3 3 7 2 2 . 5

Κ 5 , 0 0 0 4 7 9 . 4

L I 1 5 , 0 0 0 8 2 5 . 5 L 2 1 5 , 0 0 0 7 6 5.1

M l 5 , 0 0 0 4 0 8 . 0 M 2 5 , 0 0 0 4 3 8 . 6 M 3 5 , 0 0 0 5 5 1 1 . 0 M 4 5 , 0 0 0 2 6 5 . 2 M 5 3 0 , 0 0 0 2 3 1 7 . 7 M 6 3 0 , 0 0 0 2 5 7 8 . 6

2 8 t e x t s 3 2 3 , 0 0 0 3 6 4 5 11 .3

In view of their size and number, the samples from novels (M) should provide the most reliable basis of all for a frequency study. The figures for most of the individual texts are close to the average of 8.2 passives per thousand words. Novels and plays, which together make up the category 'imaginative prose' in our corpus, have similar passive frequencies. It is interesting to note that 'imitated speech' as represented in two comedies (Texts El and E2) has far fewer passive occurrencies than our recorded

THE USE OF THE PASSIVE

Class α-passives

[ Class ß-passives

Class γ-passives

Fig. 7:2. Frequencies of the agentive passive classes in eight text sets.

THE USE OF THE PASSIVE 155

TABLE 7:4 The use of passives in eight text sets

SAMPLE AGENTIVE QUASI- NON-TEXT SET SIZE PASSIVES AGENTIVE AGENTIVE TOTAL

IN WORDS PASSIVES

PASSIVES PASSIVES

'Science' (Texts H, J) 50,000 967 (19.3) 9 (0.2) 178 (3.6) 1154 (23.1) 'News' (Texts D, E, G) 45,000 611 (13.6) 21 (0.5) 77(1.7) 709 (15.8) 'Arts' (Texts I, K) 20,000 200 (10.0) 6 (0.3) 48 (2.4) 254 (12.7) 'Speech' (Texts A, B) 40,000 261 (6.5) 25 (0.6) 80 (2.0) 366 (9.2) 'Sports' (Text F) 30,000 211 (7.0) 13 (0.4) 45 (1.5) 269 (9.0) 'Novels' (Texts M) 80,000 358 (4.5) 141 (1.8) 153 (1.9) 652 (8.2) 'Plays' (Texts L) 30,000 37 (1.2) 64 (2.1) 57 (1.9) 158 (5.3) 'Advertising' (Text C) 28,000 51 (1.8) 4 (0.1) 28 (1.0) 83 (3.0)

Eight text sets 323,000 2696 (8.3) 283 (0.9) 666 (2.1) 3645 (11.3)

Frequencies in terms of occurrences per thousand words are given in brackets.

'actual speech'. The smallest passive voice consumer is television advertising with only 3 passives per thousand words.

The frequencies in the eight text sets and the proportions within each set of the three major passive classes (agentives, quasi-agentives, and nonagentives) are shown in Table 7:4. It shows that the absolute occurrence of nonagentives is rather similar in all the sets, and hence proportionately greater for the text sets with low overall passive frequency. There is a notably larger proportion of quasi-agents in the samples from imaginative prose than in any other set.

Figure 7:2 shows the distribution in the eight text sets of agentive clauses only, and a slight consequential reordering of the sets: sports reports and speech on the one hand, and advertising and plays, on the other, have changed places. The graph suggests that the proportions of Classes α, β, and γ remain largely constant throughout the sets. There is, however, a notable preponderance of Class β over α in science, and of Class α over β in sports and speech.

Summing up, the results of this inquiry into the use of the passive voice emphasize the centrality of the agentive classes in the majority of the samples, and particularly of the agentless class. Furthermore, the results indicate, firstly, that there is consider-able variation in overall passive frequency in the individual texts, the use in the most frequent being over ten times that in the least frequent; secondly, that in their use of the passive there is notable consistency among the several texts of a particular cate-gory; thirdly, that the major stylistic determining factor in the frequency of its use seems to lie in a distinction such as that between informative and imaginative prose, rather than in a difference of subject matter or between the spoken and written lan-guage. However, within the category of informative prose, passives are most com-monly found in scientific exposition. Speech occupies an intermediate position be-tween the most extreme passive users.

8

CONCLUSIONS

In this study of the English verb we have discussed, from various standpoints, some problems connected with voice: in Chapters 2 and 3, the typology of the finite verbal group and clause; in Chapter 4, the use of the voice terms and their cooccurrence with other grammatical categories in the Voice Corpus and the Minor Passive Corpus; in Chapters 5 and 6, a more delicate analysis of a small but more representative number of passive clauses (the Taxonomic Passive Corpus) in terms of a large number of criteria and an attempt at a classification of passive clauses using numerical taxonomy; in Chapter 7, a quantitative study with the classification system applied to all the passive clauses in the Major Passive Corpus, consisting of 28 sample texts and totalling almost a third of a million words. We have now reached a point in our inquiry when it should be possible to draw some general conclusions.

8.1 THE CONCEPT OF 'THE PASSIVE SCALE'

The concept of'the passive scale' (see Section 6.4, pp. 132ff.) is particularly important for the direct bearing it has on the relation of the two voice terms. This scale consists of a number of passive clause classes that have different affinities with each other and with actives.

At one end of the scale are the agentful classes, α (with animate agents) and β (with inanimate agents). Clauses belonging to these two classes have close systemic trans-formational relation with the active voice, that is to say, possess all the necessary prop-erties in the clause elements for the application of a transformational rule to convert them into their corresponding active clauses. The word 'systemic' is, however, an important qualification here, since it does not subsume transformational constraints that are imposed by, for example, clause sequence (see Criterion CL 16, p. 85), adjunct placement, and in general the requirements of balance between clause ele-ments, case alteration in pronouns with oblique forms (as in he : him), and auxiliary change between shall and will determined by the person of the subject.

The question of balance is not simply a stylistic point but can be a most important consideration (see, for example, Criterion CL 18e, p. 87). Although 1:40 in this example is an extraordinary proportion for the number of words in the subject and

CONCLUSIONS 157

in the agent, it is indicative of many significant differences between these two nominal elements in passive clauses. In the material that was analysed for element length, agents had about twice as many words as subjects. We found that the average for the subject was 3.3 words in the Taxonomic Corpus and 5 words in the Minor Passive Corpus, but for the agent 8 and 8.5 words respectively (see Sections 4.36, pp. 52ff.; and 6.3). Unlike subjects, agents are frequently coordinate, but are rarely expounded by personal pronouns (see Sections 4.34 and 6.32, p. 129). Hence, we may con-clude that one of the motivating factors in selecting the passive in favour of the active is the preference for placing heavy nominal groups at the end of sentences.

Differences between the two agentful classes include, beside agent gender, possibility of deletion and frequency of coordination (which are greater for animate than inani-mate agents); frequency of modification (which is greater for inanimate agents); finitude and noun class (animate agents, more often than inanimate agents, are definite - as opposed to indefinite - and mass nouns - as opposed to countable nouns; see pp. 125 f.). The positioning of Classes α and β on the scale has been determined by the fact that there is often structural indeterminacy between β and γ, since preposi-tional phrases may be interpreted as either agents or other adjuncts (see Section 6.42, p. 133). Such phrases, which have been called 'Janus-agents', occur in Class β/γ, This class is quite large (see Section 7.21, pp. 141f.), and constitutes a bridge between the agentful and agentless classes.

There are also other indications that there is not an abrupt separation between the classes: phrases with prepositions other than by may function as agents, or near-agents, as in

And now all tendency to laughter was submerged in parental admiration as the little girls advanced singing "Es ist ein' Ros' entsprungen" [ital]. (M5.85)

Furthermore, similar functions may be carried not only by prepositional phrases but also by adverbs, as in

The immigrant problem is widely admitted to be the most important election issue. (El.10)

Adverbs have, of course, no nominal head which can operate as active subject, but their agent-like nature is made plain by the tautologous effect of at least certain fry-phrases in addition to the adverb, as in

The immigrant problem is [*widely] admitted by everyone to be ...

and by the inadmissibility of the adverb in the active transform:

-+· everyone admits [""widely] that the immigrant problem ...

In some cases, whole clauses may have agent-like function (this is a regular feature of emotive passives):

158 CONCLUSIONS

Where there was a second sale at a higher price, and the agency decided the increase was caused because planning permission had been granted, they would collect the 75 per cent tax. (El.16)

The 'agent', which is here rather the cause (cf. Section 5.81, pp. 103f.), is felt to be the following clause. It is perhaps worth noting the similarity between the because-clause and an inanimate factive by- agent:

... the granting of planning permission ] . , , ° , , · . . . . . . ι caused the increase ... the fact that planning permission had been granted j

In our descent of the scale, we are now already in Class γ, whose key feature is absence of, but possible extension with, 6j-agent. Judging by the frequency of occurrence in all text samples, this class is central to the passive construction. In the corpus, it covered some 80 per cent of all agentive passives (i.e. Classes α, β, γ). The proportions of the agentful classes varied somewhat in the texts, but Classes α and β (including β/γ) had an average of about 10 per cent each (see Section 7.21, p. 141).

There are also features cooccurring with 'absent agent' that further isolate Class γ from α and β. Most interesting is perhaps the distribution for external clause-relation commented on in Section 4.7 (pp. 65 fF.), where we found a remarkable difference between Class γ, on the one hand, and Classes α and β as well as active clauses, on the other, in that the ratio of syntactically bound to free clauses was very much higher in Class γ than the rest (see Figure 4:11, p. 68). In the output of the Taxonomic Experiment, Group IV (which together with Group III roughly corresponds to Class γ) was at variance with all the other groups in that its clauses had no adjuncts and were almost invariably bound (pp. 127f.). These two sets of results, which were arrived at via different techniques and different corpora, further testify to the differ-ences between the agentive classes on the passive scale; whereas agentful and active clauses are similar in usually being free (as well as in having potential direct trans-formation relation), proportionately many more agentless agentive clauses are bound in their external clause-relation. Since other characteristics of Group IV include high frequency of pronoun subjects and a large proportion of coreference between the passive subject and the object or predicative of a preceding contiguous clause, we may assume that an important function of many Class γ-passives is that of postnominal modifier, clausal adjunct, etc. (cf. p. 128):

He tempted Cartwright with two half volleys which were royally punished... (F. 28) ... Lewis Packford himself had contrived a situation which could be exhibited as entirely ludicrous. (M6.84)

Furthermore, we noted for the Voice Corpus that the use of tense forms in the verbal group was not parallel in the active and passive (see Section 4.2, pp. 43 ff.): the passive Types ad, AD, and bd (see Section 2.2, pp lOf.) were more common than their active counterparts (a, A, b), whereas other tense forms were more frequent in the active; this

CONCLUSIONS 159

was particularly true for Type c/C as against cdjCD and for the more complex verbal group structures (ac/AC and bcjBC as against acd/ACD and bedjBCD). However, the high ratio of Type ad/AD in the passive applied to Class γ but not to Classes α and β; Type bd, on the other hand, was particularly frequent in the agentful classes.

Class δ consists of two subclasses, δ1 ('attitudinal passives') and δ2 ('emotive pas-sives'), and a chief characteristic of the latter (its range of possible agents) has given the whole class the name of 'quasi-agentive passives'. The two subclasses demonstrate various points of similarity and dissimilarity (see Sections 7.22, pp. 147f.; and 8.2 below). Emotive passives are unique in our classification in having a striking macro-semantic correlate to their defining syntactic attributes.

Class ε is rather more heterogeneous than δ, and is interesting because of its low place on the passive scale and its complex voice relations. These last two classes should be dealt with more fully using a different set of criteria and together with the related intransitive and equative clause types. Nevertheless, it will be of interest to indicate here their place in the overall description of voice relations in English.

8.2 TRANSFORMATIONAL A N D SERIAL VOICE RELATIONS

As we proceed down the passive scale, it becomes increasingly difficult to consider the passive in terms of a transformational voice relation, since this is being gradually replaced by a different relation which we call 'serial'. The theory of serial relationship states that sentences can be produced paradigmatically: "There is surely every reason to suppose that the production of sentences proceeds by a complex interplay involving 'transformation' and what Miller calls 'a sentence-frame manufacturing device'."1

The concept of serial relation will help us to account for many voice features which cannot easily be explained otherwise. Bolinger, whose idea of 'syntactic blending' is similar to serial relation, draws attention to the 'stepwise relationship to the passive' of adjectives like prone and loth in a passive series such as the following:2

I am forced (led, induced, etc.) to agree with you. (Active transformations normal.)

I am inclined (disposed, disinclined) to agree with you. (Active transformations less usual.)

I am minded to agree with you. (No active transformation.)

The illustrated three types correspond to our γ-, δ-, and ε-passives. We may add that for Class δ, there is a greater likelihood of inanimate than animate agent and a possi-bility of transformation to active intensive clause type:

This makes me inclined to agree with you. *This makes me forced to agree with you.

1 Quirk 1965, 213. 8 Bolinger 1961b, 378.

160 CONCLUSIONS

Since verbs in the quasi-agentive class can also occur in other classes, passive construc-tions with such verbs may be potentially ambiguous in minimal structures (although in the corpus they were hardly ever found to be so, and this is unlikely to be the case in actual context). For example, the sentence

He was interested in linguistics.

may be interpreted either as a γ-passive (serially related to Class a) or as a 52-passive (serially related to active equative clauses):

(a) He was (very much) interested in linguistics by his professor. (b) He was (very much) interested in linguistics. (c) He was (very) interested in linguistics. (d) He was (very) keen on linguistics.

The corresponding active transforms are: (a) His professor interested him in linguistics. (b/c) Linguistics interested him.

Although informants find the second of these the right transform of the 62-passive, it is worth noting that, unlike the passive, the active cannot take verb qualification with very.

Structures of the favoured passive factitive type W to F(see p. 145), have a range of active transform structures; in some cases there are no active analogues at all, and in many more cases, transforms are dubious. The following series may illustrate a postulated serial relationship for the passive voice in the left-hand column:

The ambivalence of Class δ2 can be satisfactorily accounted for by the concept of serial relation since verbs of this class have transformational as well as serial voice relation:

His sisters behaviour annoyed him very much. His sisters behaviour made him feel very annoyed and angry. He was (felt) very annoyed (and angry) at his sister s behaviour.

There is an aspectual serial relation operating when a change of tense produces a passive which permits transformation to active. The ε-passive

The house is already sold.

is indirectly related to the active

He was

allowed seen said calculated scheduled reputed

to do it.

-* They allowed him to do it. They saw him do it.

-» They said that he did it. They calculated that he would do it.

-* (?) They had scheduled him to do it. -* *They reputed him to do it.

CONCLUSIONS 161

The estate agent has already sold the house.

via the γ- and α-passives:

The house has already been sold. The house has already been sold by the estate agent.

Other examples of 'statal passives' (see pp. 92 f.) with indirect transformational relation are the following:

Cavendish inquired with grave courtesy, 'Is the thesis finished, Doctor? 'Not quite, Captain.' Ί heard you were coming with us to give a lecture on your findings to the New York Canford Institute.' (M4.57) 'We must leave it [the rat]', repeated Madeleine. 'We can't. It's all mauled - don't you see? We cannot [ital.] It's got to be killed.'

(Ml.25)

In principle, statal passives do not have expressed agents, but serial relation will help to explain occasional examples of statal passives with agents, as in

This memorial is placed here by his friends and neighbours in testimony of respect, affection, and gratitude*

which has no active transform

*His friends and neighbours place this memorial here ...

but is serially related to

His friends and neighbours have placed this memorial here ...

and its passive transform

This memorial has been placed here by his friends and neighbours ...

as well as to

This memorial is placed here in testimony of...

Lexical marking in the auxiliaries of nonagentive passives may be taken to denote one further remove from the agentive class, as in the following series leading to an active transformation:

The door remains bolted and barred. The door is bolted and barred. The door has been bolted and barred. The door has been bolted and barred by the shopkeeper. The shopkeeper has bolted and barred the door.

» Erades 1958/9.

162 CONCLUSIONS

This view of statal passives enables us not only to dispense with a rigid, unrealistic, and arbitrary dichotomy into 'passive' and 'nonpassive', but also to account satisfactorily for those extremes of serial relation that may be called 'serial isolates', as in

The rat reared up full stretch in a supreme convulsion, as if about to spring. But it was done for. It rolled over, twitching. Immediately its open eyes began to film. (Ml.28)

Here, only the last part of the series remains, since in accepted usage we now have no corresponding γ-, and α-passives or, indeed, active correlates:

*The rat had been done for {by the dog). *The dog had done for the rat.

OED has an entry do for (s. v. do 38), both active and passive, which is labelled 'colloq.', but the last quotation for the active form is dated 1876; interestingly, Web-ster (1961) acknowledges the serial isolate by making no reference to do for under do but having a separate entry done for, which is labelled 'adj . A similar serial isolate is the colloquial done in ('exhausted').

Since we no longer have to account for all passives in terms of transformation, serial relationship also bridges the gap, in a natural way, between noncompounds and compounds (Class ζ) and between verbal and nonverbal bases (see Sections 2.73-4, pp. 21 ff.; and 6.46, pp. 137f.). Agentful compound-passives are then related serially to agentful noncompound-passives instead of transformationally to actives, as in:

Everyone rejects me. I am rejected by everyone. Ί am unloved and rejected by everyone'... (E2.27) But Cavill was unimpressed by this sally. (M6.54)

Generally speaking, transformational voice relation operates most strongly at the top end of the scale, and serial relation most strongly at the bottom end, with varying degrees of both relations in between these two extremes, as may be illustrated by the set of sentences in Figure 8:1. Sentence (A) is the direct transform of α-, and the in-direct transform of the γ-passives; (B) is the direct transform of β. The δ-passives have two active analogues each, one in transformational and one in serial relationship. The ε-passive is serially related, on the one hand, to γ-passives, and, on the other hand, to both the active intransitive (G) and the equative (H) sentences. The particular ε — intransitive relation, which involves type-change in the verbal group, has been referred to as 'transmutation' (see Section 5.45, pp. 88 ff.).

8.3 THE DESCRIPTION OF THE ENGLISH PASSIVE

We must now conclude by attempting to answer some of the questions raised in Section 1.1, in particular whether the passive in English should be derived from actives

CONCLUSIONS 163

This law has been applied to all murderers by the president.

PASSIVB active The president has applied ^ A ^ this law to all murderers

(by/in an edict).

This law has been applied to all murderers by an edict.

An edict has applied this law to all murderers.

This law has been applied to all murderers in an edict.

This law has been applied to all murderers.

Everyone is obliged to obey this law.

[The president/An edict] obliges everyone to obey this law.

Everyone has to obey this law.

Everone is pleased to have this law.

It pleases everyone to have this law.

Everyone is glad to have this law.

This law is now (already) applied to all murderers.

This law now (already) applies to all murderers.

This law is now applicable to all murderers.

Fig. 8:1. Diagrammatic representation of some voice relationships. <><><> = transformational relation

= serial relation

164 CONCLUSIONS

by transformation or whether it should be treated as a clause type in its own right. There exists, in certain respects, a clear connection between the active and passive

which can be stated by transformational rules. At the same time, when native speakers are faced with a choice, in actual sentences, they almost invariably prefer one of the voice terms to the other. The preferred one is usually the active. The following extreme result, obtained as a by-product in an experiment designed to measure lin-guistic acceptability, may serve as an illustration of both these points. Asked to estimate its deviance, 73 informants rejected, two queried, and one accepted the following sentence:

A nice little car is had by me.

On the other hand, 74 informants accepted, one queried, and one rejected an active analogue:

I have a black Bentley.

Before giving a direct judgment of the acceptability of the sentences, the informants were asked to perform a simple grammatical operation on each of them, in this case a change from positive to negative. Whereas a single informant failed in the case of the active sentence, only 36 managed to produce the successful operation on the passive sentence (which, as it happened, was one of the lowest results for any of the 50 sen-tences in the test battery). From our point of view, it is interesting to note that as many as 23 informants transformed the passive sentence into the active without being asked to do so, whereas no one transformed another passive sentence in the battery,

Clothing was needed by the poor.

Such a telling demonstration could surely be possible only if there was a feeling of great discomfort with the passive in this sentence, as well as a clear transformational voice relationship, by means of which the informants could achieve the necessary rectification of the deviant sentence.4

Once the existence of this basic voice relation has been stated, we must draw atten-tion to the considerable drawbacks of a procedure by which one voice term is derived from the other by means of transformation. The favoured direction has been from active to passive (see Section 1.1). The advantage of this particular unidirectional derivation (instead of one from passive to active) is largely supported by experience in the present work. We may mention the following points:

(a) The low proportion of agentful passives (in fact only some 20 per cent of agent-ive clauses and less than IS per cent of all passive occurrences in our material; see pp. 141 ff.).

(b) The apparently general preference for the active and lower overall frequencies for the passive. Even in a text (Jl) of the group 'scientific exposition' (which was found to be the group with the highest occurrence of passives), the proportion of pas-4 See Quirk & Svartvik 1966,74

CONCLUSIONS 165

sives to actives did not exceed 1:3, even discounting active equative fee-clauses (p. 46). (c) The differences in the use of verbal group types, where the restrictions are largely

on the passive (see pp. 46ff.). Compare, for example, the following pair:

These Conservatives have not been winning seats lately. (El. 14) Seats have not been being won by these Conservatives lately.

(d) The greater restrictions on lexical verbs in the passive than the active. Apart from intransitive and equative verbs, examples include some uses of the verbs have, lack; contain, hold; become, fit, suit; mean, resemble·, befall·, cost, last, take, walk, weigh.6

(e) The limited and as yet not clearly defined range of prepositional verbs in the passive. These are verbs and prepositions that have sufficient cohesion to be able to operate together as transitive verbs in the passive (see Sections 2.71-2, pp. 19fT.), as in

But in this country there are singularly few murders which are conceived of as deliberately incident to a robbery. (M6.38)

Prepositional verbs in the material also include account for, call for, care for, call upon, deal with, depend upon, dwell upon, fire on, go into, look at, preside over, provide for, talk of, think of", the phrasal-prepositional verbs include get away with, and make off with. Some of the collocations (like look at) are very close and can occur in the passive with few or no restrictions, but others (like go into) will do so only under certain conditions. Compare, for example, the following pair of sentences:

I think a lot of that [problem] can be gone into and gone into pretty . pretty thoroughly - (B5.6.14) *The room was gone into at once.

The active analogue of the latter is, however, fully acceptable:

He went into the room at once.

The difference in acceptability between the two passives must, it seems, be accounted for by some concept like concrete/abstract subject or literal/metaphorical verb (compare The town/conclusion was arrived at).

(J) The restrictions on the relations between clause elements, such as coreference between subject and object. This can be manifested by reciprocal and reflexive pronoun objects or by secondary concord (see Section 3.7, p. 33; and Olsson 1961, 111 ff.) manifested by possessive pronoun:

' Cf. Gleason 1965, 307f. and Twaddell 1963, 9. A much mote acceptable passive of have than A nice little car is had by me (discussed above) is the clichd A good time was had by all (cf. Hill 1958, 203). Although this sentence appears to contradict the general rule that have in the sense of 'own', 'enjoy' does not occur in the passive, it is, rather, the exception that proves the rule. In fact, it is precisely because of its deviance (and institutionalization) that this sentence achieves its special effect of facetiousness. Have ('obtain', 'get') has no similar restrictions in the passive: Taxis can be had anywhere outside the station.

166 CONCLUSIONS

There is a further bonus for which most M.P.s are rubbing their hands - though Blackpool hotels are upset. (E2.7) The doctor shook his head, and applied himself once more to his notebook.

(M4.56)

A general difficulty with using transformations of sentences in context is the imbalance that may result. It may be caused by discrepancies in the lengths of clause elements, such as personal pronouns and other short agent exponents occurring with long sub-ject exponents. We have found that agents are generally heavier than subjects. Pronouns are very frequent as subjects but extremely rare as agents. As many as one third of the clauses in Corpus III and one half in Corpus II had coordinate agents, whereas coordination was rare in the subject (see Sections 4.34-36, pp. 52 ff.; 6.32, pp. 129 ff.).

Stylistic imbalance can further be attributed to disruption of the natural develop-ment of the discourse, as this has been stated in the principle of functional sentence perspective (FSP), which 'causes the sentence to open with thematic and close with rhematic elements. Very roughly speaking, 'thematic elements are such as convey facts known from the verbal or situational context, whereas rhematic elements are such as convey new, unknown facts'.® The following sequence may serve as illustration of how the passive may be obligatory both to avoid imbalance and to accord with the principle of FSP:

'Thinking ... may be provisionally defined as what occurs in experience when an organism, human or animal, meets, recognizes, and solves a problem.' This provisional definition was also favoured by Dewey, who pointed out that so long as our activity glides along smoothly, or as long as we allow our imagination to entertain fancies, there is no reflection. (H.30)

There are, as we have seen, many characteristic features of the passive voice that cannot easily be accounted for in terms of unidirectional active — passive trans-formation. Rather, as we proceed down the passive scale, in the direction away from the agentful classes, it becomes increasingly realistic, and economical, to consider the production of passive sentences in terms of serial relationship with equative and intransitive active clause types.

• Firbas 1964,112.

APPENDICES

LEXICAL VERBS OF AGENTIVE CLAUSES IN THE MAJOR PASSIVE CORPUS (SEE CHAPTER 7)

APPENDIX la. VERBS IN PASSIVE TRANSITIVE CLAUSES WITH REFERENTIAL SUBJECT

(see pp. MSf.)1

accompany accept address adopt advise announce aid appreciate appreciate arrange beat attempt butcher await call call carry carry on catch catch convince commission corner confer defeat confirm dismiss confiscate entertain construct examine create finance cut follow decide force decipher govern define join demand knock demonstrate meet describe no-ball detect nobble determine pay develop

1 Verbs with animate subjects in the first column; verbs with inanimate subjects in the second column.

APPENDICES

penalize devise punish discuss receive disturb reject do ride dominate stump drive support emphasize surround establish undo examine wrap explain

extend favour feel form formulate frame give grant grasp handle imply impose inspect investigate kill lay down lead make make off with miss note notice observe obtain offer order own pay penetrate pick up preside over promise

APPENDICES

provoke publish punish purify put up receive recognize repeat report represent require resist resolve return say score see serve set share sort out speak state study summarize support take think of treat understand undertake use value watch win

Class β: affect attract catch convert corner cover

abstract accomodate accompany affect arylate assist

APPENDICES

COW attack defeat bias deter blaze embarass block fascinate blur help bound hold bring about impress carry incense cause interrupt circumscribe motivate compensate muddle complicate overwhelm confirm precondition connect reduce consolidate retain control reward cover save create shock define split demonstrate startle destroy surprise determine take in dictate trouble discount uproot disguise

disrupt do effect elaborate eliminate evoke exceed facilitate follow force form give govern hinder hold improve increase

APPENDICES

indicate influence inhibit innervate interrupt justify limit line lower mar mark match occupy overstrain parallel permit play power precede predict prescribe prevent produce protect pull punctuate regenerate reinforce relieve remove reorganize replace represent require ruin set set off solve stamp in stimulate stop substantiate

APPENDICES

suggest summarize supplement support surpass surround symbolize take take over take up threaten trap unmask

Class β/γ: accompany achieve break down change complete confirm convert correct damage deliver describe destroy determine effect elucidate express further give identify illustrate imply initiate interpret introduce keep maintain mention modify

APPENDICES 173

Class γ: accost amuse arrest assassinate beat believe bite blackmail bowl bring bring up bury buy out cage call up capture carry cast out catch caution censure chain cheer conquer create cut defeat deplore detain

obtain prove reduce remove replace select settle show solve spread summarize support

abandon abolish absorb accept accomodate accomplish account for achieve acquire add adduce adjourn administer adopt advance affect alert allow alter analyse answer applaud apply approach arrange articulate ascertain ask assess

APPENDICES

detect assume discredit attempt dismiss avoid dissect await distribute backdate drop balance educate bear elect bind employ block enter blunt exclude bring execute bring about expose bring in fear bring out find broadcast fire on build flog bungle foil bunker garland bury give out buy hang calculate have call for hear cancel help care for hire carry hit carry out hunt cause hurt cauterize ignore cement injure change install characterize invite chase keep check kill clamp down knock clarify know clear away leave clear up look after close massacre coin mention combine mislead commit mollify comprehend

APPENDICES

move compress murder compute name concede note condemn oppose conduct outplay confirm outvote connect ovariectomize consider overrule construct pay continue persecute control place cool post correct press count provide for cover punish create pursue criticize put curtail put away cut recall cut down represent cut off respect dangle restrain deal with sack debate satisfy declassify scrutinize deduce see defer select define send delay shoot demonstrate show up depict smuggle describe stump design summon destroy take detect take in determine talk of develop teach dig up test diminish thrash discard threaten discipline train discover

APPENDICES

transfer discuss underpay disentangle usher disperse warn display withdraw distribute worship divide

divulge do draw draw up drink drop dry duplicate dwell upon effect elicit eliminate emancipate embody emphasize employ endorse envisage establish esterify evaluate evaporate evoke evolve exaggerate examine excavate exclude exercise exhibit expect explain explore expose express extend

APPENDICES

face

fashion

favour

fence in

film

filter

finance

find

find out

fit

fix

flight

flout

fly

follow

force

forecast

foreshadow

forget

form

formulate

found

frame

furnish

gauge

generate

get away with

give

go into

grant

grasp

halve

have

hear

heat

help

hit

hold

hurt

identify

ignore

illustrate

APPENDICES

imagine

implant

implement

impose

imprint

improve

inaugurate

include

incorporate

increase

incubate

induce

infer

inherit

initiate

install

instance

intend

interchange

interpret

introduce

invent

investigate

isolate

issue

judge

keep

know

launch

lay

lay off

learn

leave

lessen

lift

list

live

lose

lower

maintain

make

mask

APPENDICES

measure meet mention misconstrue miss mistake modernize mount move need neglect note nurse observe obtain occupy offer open organize overcome override oversubscribe overthrow overturn paint pair parallel particularize pass patch pay perform permit perpetuate phrase pick up pin down place plan play plot postulate

APPENDICES

practise preach predict prejudge prepare present preserve press prevent print probe produce project prolong propose prove provide provoke prune publish pull punish pursue push put quote radiate raise reach read realize rebuild recall receive recite recognize reconstruct record recover reduce refine refit

APPENDICES

reflect refuse regard register regret reintroduce release relieve remodel remove renegotiate renew reorganize repair repay repeat replenish report represent require reserve resist resolve restate restore retain retard return reveal reverse review revise revive revivify revoke rotate rouse run run up satisfy save say

APPENDICES

score scrap scrawl see seek seize sell send set set off set up settle shake show sing smash smother solve sort out speak spread start state stop store strengthen stress study submerge submit subsidize substantiate suggest sum supersede supply suspend tackle take take up tell test

APPENDICES

thrust touch trace transact transform transmit treat try try out tune turn down underestimate undermine underrate understand use use up vary view vindicate want warm wash waste watch weight win wind work work out wrap write

APPENDIX lb. VERBS IN PASSIVE TRANSITIVE CLAUSES WITH PROLEPTIC it AS SUBJECT ANTICIPATING A

FINITE VERB CLAUSE OR DIRECT SPEECH (see p. 144)

a : demonstrate show suggest support

APPENDICES

Class γ: accept agree announce appreciate argue ask assume believe claim concede conclude consider deduce deny discover emphasize envisage expect fear feel find foresee forget hope imagine indicate infer intend know learn note notice object observe point out recommend record remark remember report rumour say

APPENDICES

see show state suggest suppose think wish

APPENDIX Ic. VERBS IN PASSIVE TRANSITIVE CLAUSES WITH PROLEPTIC it AS SUBJECT ANTICIPATING A

NONFINITE VERB CLAUSE (to-INFINITIVE) (see p. 144)

Class γ: agree decide feel hold

APPENDIX Ha. VERBS IN PASSIVE SEMIDITRANSITIVE CLAUSES WITH REFERENTIAL SUBJECT

(see pp. 144f.)

Class a : convict (of) drive (to) rob (of) take (to)

apply (to) associate (with) bequeath (to) do (on) draw up (for) give (to) mill (for) put across (on) subject (to) submit (to)

Class β:

Class γ:

bully (into) confine (to)

admit (to) ask (about) banish (from) bring (into)

develop (into) impose (upon) stamp out (into) transform (into) accept (for) adapt (to) add (to) address (to)

APPENDICES

celebrate (for) apply (for) charge (with) apply (from) convict (of) apply (to) convince (of) ascribe (to) crowd (out of) associate (with) deprive (of) attach (to) discharge (from) attribute (to) introduce (to) blame (on) leave (with) blend (for) let in (on) bring (into) let out (of) bring (to) mesmerize (into) build up (on) pair (with) choose (for) pay (for) combine (with) punish (for) commit (against) put (into) commit (to) recompense (for) commute (for) remind (of) compare (to) rouse (to) compare (with) sentence (to) compile (on) sew (into) complete (into) show (to) conduct (against) strip (of) confuse (with) supply (with) consign (to) suspect (of) convert (from) take (for) convey (to) tip off (about) deduce (from) treat (to) defer (to)

deny (to) direct (at) direct (to) direct (towards) discuss (with) do (about) do (on) do (to) draw (from) efface (from) entrust (to) equip (with) exempt (from) expose (to)

APPENDICES

extend (for) extend (to) extract (from) focus (on) give (of) give (to) gouge (out of) grant (to) hear (of) impose (on) induce (in) infer (from) introduce (into) know (about) know (from) know (of) leave (for) leave (to) liken (to) make (about) make (for) make (into) make (of) make (on) make (to) motivate (towards) notify (to) offer (to) pass (into) pay (to) permit (to) place (in) place (on) postulate (for) prefer (to) present (to) promote (to) propose (for) provide (for) put (into) put (on) put (to)

APPENDICES

raise (about) refer (to) relate (to) reserve (for) restrict (to) reveal (to) say (about) say (of) send (to) separate (from) set (against) shield (from) show (to) snatch (from) spend (in) spend (on) subject (to) submit (to) substitute (for) take (into) tell (about) throw (on) tie (to) transform (into) translate (into) turn (into) use (of) voice (about) vouchsafe (to) write (to)

APPENDIX lib. VERBS IN PASSIVE SEMIDITRANSITIVE CLAUSES WITH PROLEPTIC it AS SUBJECT ANTICIPATING A

FINITE VERB CLAUSE (see p. 145)

Classy: bear (upon) mention (to)

APPENDIX Ilia. VERBS IN PASSIVE DITRANSITIVE CLAUSES WITH A NOMINAL AS COMPLEMENT

(see p. 145)

Class a: give give tell

APPENDICES

Class γ: allow ask deny

give permit set

excuse fine give grant offer promise refuse save show teach tell

APPENDIX Illb. VERBS IN PASSIVE DITRANSITIVE CLAUSES WITH A NOMINAL + PREPOSITION AS COMPLEMENT

(see p. 145)

Class γ: make (fun of)

APPENDIX IIIc. VERBS IN PASSIVE DITRANSITIVE CLAUSES WITH A FINITE VERB CLAUSE AS COMPLEMENT

(see p. 145)

Class a: assure

Class γ: advise ask inform notify persuade remind tell

APPENDIX IVa. VERBS IN PASSIVE SEMIFACTITIVE CLAUSES WITH PREPOSITION + NOMINAL AS CLOSE ADJUNCT

Class α: make (a mess of)

(see p. 145)

Class a: accept (as) describe (as) regard (as)

190 APPENDICES

Classy: accept (as) acknowledge (as) assess (as) bring (as) class (as) describe (as) employ (as) know (as) select (as) try (as)

take (as) use (as)

accept (as) acclaim (as) base (as) bring (as) calculate (as) categorize (as) characterize (as) choose (as) conceive (as) conceive of (as) consider (as) define (as) designate (as) develop (as) display (as) do (as) express (as) give (as) hear (as) identify (as) illustrate (as) intend (as) interpret (as) know (as) look at (as) need (as) predict (as) read (as) recognize (as) regard (as) see (as) summarize (as) take (as) treat (as) use (as)

APPENDICES

APPENDIX IVb. VERBS IN PASSIVE SEMIFACTITIVE CLAUSES WITH PREPOSITION + ADJECTIVE OR PAST PARTICIPLE

AS CLOSE ADJUNCT (see p. 146)

Class a : remember (as) accept (as) Class γ: describe (as) accept (as)

define (as) describe (as) exhibit (as) regard (as) take (as) think of (as) treat (as)

APPENDIX IVc. VERBS IN PASSIVE SEMIFACTITIVE CLAUSES WITH PREPOSITION + PRESENT PARTICIPLE AS CLOSE ADJUNCT

(see p. 146)

Class a :

Class β:

Class γ:

deceive (into)

accuse (of) charge (with) describe (as) inhibit (from) preclude (from) prohibit (from) report (as) restrain (from)

regard (as)

prevent (from) recognize (as) regard (as) report (as) see (as) single out (as) take (as)

APPENDIX Va. VERBS IN PASSIVE FACTITIVE CLAUSES WITH NOMINAL AS COMPLEMENT

(see p. 146)

Class a :

Class γ:

call

appoint call make name

call consider make name term

APPENDICES

APPENDIX Vb. VERBS IN PASSIVE FACTITIVE CLAUSES WITH ADJECTIVE OR PAST PARTICIPLE AS COMPLEMENT

(see p. 146)

Class a: keep

Class β:

Class γ:

make

make keep make

consider account keep assume kill call place consider

dub find leave maintain make prove render show

APPENDIX Vc. VERBS IN PASSIVE FACTITIVE CLAUSES WITH /o-INFINinVE AS COMPLEMENT

(see p. 146)

Class a: ask report invite show press think

Class β: bind forbid

Class γ: advise admit allow allow appoint assume ask believe believe bring call upon calculate depend upon claim expect consider intend declare invite design know expect

APPENDICES 193

leave find make guarantee mean hear observe hold prove intend recommend know report make repute mean require observe say order see regard teach report think say use see

show suppose take think

APPENDIX Vd. VERBS IN PASSIVE FACTITIVE CLAUSES WITH PRESENT PARTICIPLE AS COMPLEMENT

(see p. 147)

Class γ: leave set

see

BIBLIOGRAPHY

Akerlund, A. (1914), "A Word on the Passive Definite Tenses", Englische Studien 47.321-37. Allen, W. S. (1959), Living English Structure. London. Bach, E. (1964), An Introduction to Transformational Grammars. New York/Chicago/San Francisco. Bloomfield, L. (1933), Language. New York. Bolinger, D. L. (1961a), Generality, Gradience, and the All-or-None. The Hague.

(1961b), "Syntactic Blends and Other Matters", Language 37.366-81. Brose, B. (1939), Die englischen Passivkonstruktionen vom Typus "I am told a story" und "I am sent

for". Würzburg. Buyssens, E. (1950), "La conception fonctionnelle des faits linguistiques", Grammaire et Psychologie

(= Journal de Psychologie 43.37-53). Carvell, Η. Τ. & Svartvik, J. (1966), Linguistic Classification and Numerical Taxonomy. TheHague. Chomsky, N. (1957), Syntactic Structures. The Hague.

(1962), "A Transformational Approach to Syntax", Third Texas Conference on Problems of Linguistic Analysis in English in 1958 (Studies in American English). Austin, Texas. (1964a), Current Issues in Linguistic Theory. The Hague. (1964b), "Topics in the Theory of Generative Grammar" (to appear in Current Trends in

Linguistics 3, ed. by T. A. Sebeok). Crystal D. & Quirk, R. (1964), Systems ofProsodic andParalinguistic Features in English. The Hague. Curme, G. O. (1913), "The Proper Subject of a Passiv Verb", Modern Language Notes 28.97-101.

(1931), Syntax (=Curme, G. O. & Kurath, H., A Grammar of the English Language 3). Boston. Erades, P. A. (1950), "His books don't sell", English Studies 31.155-7.

(1958/9), "Points of Modern English Syntax", English Studies 39.286, 40.63-4. Finney, D. J. et al. (1963), Tables for Testing Significance in a 2 x 2 Contingency Table. Cambridge. Firbas, J. (1959), "Thoughts on the Communicative Function of the Verb in English, German and

Czech", Brno Studies in English 1.39-63. (1964), "From Comparative Word-order Studies", Brno Studies in English 4.111-26.

Firth, J. R. (1957), "A Synopsis of Linguistic Theory, 1930-1955", Studies in Linguistic Analysis (Special Volume of the Philological Society). Oxford.

Francis, W. N. (1958), The Structure of American English. New York. Frary, L. G. (1929), Studies in the Syntax of the Old English Passive with Special Reference to the Use of

'wesart and 'weordan". Language dissertations published by the Linguistic Society of America, No. 5. Baltimore.

Fries, C. C. (1940), American English Grammar (English Monograph No. 10, National Council of Teachers of English). New York. (1959), The Structure of English. London.

Fröhlich, J. (1951), Der indefinite Agens im Altenglischen (= Swiss Studies in English 25). Bern. Gaaf, W. van der (1928), "The Predicative Passive Infinitive", English Studies 10.107-14.

(1929), "The Conversion of the Indirect Personal Object into the Subject of a Passive Con-struction", English Studies 11.1-11, 58-67. (1930), "The Passive of a Verb Accompanied by a Preposition", English Studies 12.1-24.

Garvin, P. L., ed. (1963), Natural Language and the Computer. New York. Gleason, H. A. Jr. (1965), Linguistics and English Grammar. New York.

BIBLIOGRAPHY 195

Godfrey, J. (1965), "The Survey of English Usage", English Language Teaching 19.98-103. Green, A. (1913), The Dative of Agency. New York.

(1914), "The Analytic Agent in Germanic", The Journal of English and Germanic Philology 13.514-52.

Halliday, Μ. A. K. (1961), "Categories of the Theory of Grammar", Word 17.241-92. (1964), "Syntax and the Consumer", Monograph Series on Languages and Linguistics 17.11-24.

Harris, Z. S. (1957), "Co-occurrence and Transformation in Linguistic Structure", Language 33. 283-340.

Hatcher, A. G. (1949), "To Get/Be Invited", Modern Language Notes 64.433-46. (1956), "Syntax and the Sentence", Word 12.234-50.

Hendriksen, H. (1948), "The Active and the Passive", Sprakvetenskapliga Sällskapets i Uppsala förhandlingar 1946-8. Uppsala Universitets Arsskrift 1948:13.61-76.

Herdan, G. (1964), Quantitative Linguistics. London. Hill, A. A. (1958), An Introduction to Linguistic Structures. New York. Hill, L. A. (1964), "The Passive Again", English Language Teaching 18.69-74. Hockett, C. F. (1958), A Course in Modern Linguistics. New York. Jacobson, S. (1964), Adverbial Positions in English. Stockholm. Jespersen, O. (1909-49), A Modern English Grammar on Historical Principles. Parts 1-7. London/

Copenhagen. (1924), The Philosophy of Grammar. London. (1933), Essentials of English Grammar. London. (1937), Analytic Syntax. Copenhagen.

Jones, D. (1963), Everyman's English Pronouncing Dictionary. London. Joos, M. (1964), The English Verb. Madison/Milwaukee. Jud-Schmid, E. (1956), Der indefinite Agens von Chaucer bis Shakespeare (= Swiss Studies in

English 39). Bern. Kennedy, A. G. (1920), The Modern English Verb-Adverb Combination (Stanford University

Publications, University Series, Language and Literature 1:1). Kirchner, G. (1936-7), "The Verbs with Direct and Indirect Object Re-Examined", English Studies

18.1-16, 206-22; 19.97-112. (1951), "Past Participles prefixed by 'un- Used in true Verbal Function to Form an Emphatic

Passive of Negation", English Studies 32.218-9. Klingebiel, J. (1937), Die Passivumschreibungen im Altenglischen. Berlin. Koumari, A. (1956), Certain Uses of the English Passive Voice and their Teaching to Foreign Students.

Thessaloniki. Kruisinga, E. (1927-31), A Handbook of Present-Day English. Part I in 4th ed.; Part 2 in 5th ed.

Groningen. (1927), "Retained Accusatives in Passive Sentences", English Studies 9.38-40.

Kuiera, H. (1962), "A Note on the Digital Computer in Linguistics", Language 38.279-82. Kurtz, G. (1931), Die Passivumschreibungen im Englischen. Berlin. Lamb, S. M. (1961), "The Digital Computer as an Aid in Linguistics", Language 37.382-412. Lees, R. B. (1957), Review of Chomsky, Syntactic Structures in Language 33.375-408.

(1963), The Grammar of English Nominalizations. Bloomington. (1964), "On Passives and Imperatives in English", Gengo Kenkyu 46.28-41.

Long, R. (1961), The Sentence and its Parts. Chicago. Longacre, R. E. (1964), Grammar Discovery Procedures. The Hague. Lyons, J. (1963), Structural Semantics (Publications of the Philological Society, Vol. 20). Oxford. Marchand, Η. (1960), The Categories and Types of Present-Day English Word-Formation. Wiesbaden. McKerrow, R. B. (1922), "English Grammar and Grammars", Essays and Studies 8.148-67. Oxford. Meier, Η. H. (1953), Der indefinite Agens im Mittelenglischen (1050-1350) (= Swiss Studies in

English 34). Bern. Meyer-Lübke, W. (1925), "Vom Passivum", Neusprachliche Studien (= Die Neueren Sprachen,

Beiheft 6.158-71). (1926), "Vom Wesen des Passivums", Die Neueren Sprachen 34.161-84.

Mihailovid, L. (1963), "Some Observations on the Use of the Passive Voice", English Language Teaching 17.77-81.

196 BIBLIOGRAPHY

Mitchell, Τ. F. (1958), "Syntagmatic Relations in Linguistic Analysis", Transactions of the Philological Society 1958.101-18. Oxford.

Moroney, M. J. (1956), Facts from Figures (Pelican Book A 236). Mustanoja, T. (1960), A Middle English Syntax. Part I: Parts of Speech ( = Mimoires de la Societi

Neophilologique de Helsinki 23). Helsinki. Nida, E. A. (1960), A Synopsis of English Syntax. Norman, Oklahoma. Nowell-Smith, P. (1954), Ethics (Pelican Book A 293"). Olsson, Y. (1961), On the Syntax of the English Verb with Special Reference to HAVE A LOOK

and Similar Complex Structures (= Gothenburg Studies in English 12). Gothenburg. Owen, Ε. T. (1914), "The Relations Expressed by the Passive Voice", Transactions of the Wisconsin

Academy of Sciences, Arts, and Letters 17:1.77-148. Madison, Wisconsin. Oxford English Dictionary. Oxford. Palmer, F. R. (1965), A Linguistic Study of the English Verb. London. Pike, K. L. (1954-60), Language in Relation to a Unified Theory of the Structure of Human Behavior.

Parts 1-3. Glendale, California. (1962), "Dimensions of Grammatical Constructions", Language 38.221-44. (1963), "A Syntactic Paradigm", Language 39.216-30.

Poutsma, H. (1926-9), A Grammar of Late Modern English, Parts 1 and 2:2. Groningen. Quirk, R. (1960), "Towards a Description of English Usage", Transactions of the Philological Society

1960. 40-61. Oxford. (1962), The Use of English. London. (1965), "Descriptive Statement and Serial Relationship", Language 41.205-17.

Quirk, R. & Svartvik, J. (1966), Investigating Linguistic Acceptability. The Hague. Reichmann, W. J. (1964), Use and Abuse of Statistics (Pelican Book A707). Roberts, P. (1955), "Fries's Group D", Language 31.20-4. Robins, R. H. (1964), General Linguistics. London. Scheurweghs, G. (1959), Present-Day English Syntax. London. Sledd, J., A Short Introduction to English Grammar. Chicago. Smith, C. S. (1964), "Determiners and Relative Clauses in a Generative Grammar of English",

Language 40.37-52. Söderlind, J. (1951, 1958), Verb Syntax in John Dryderis Prose, Parts 1 and 2. Uppsala. Sokal, R. R. & Sneath, P. H. A. (1963), Principles of Numerical Taxonomy. San Francisco/London. Sorensen, H. S. (1958), Word-Classes in Modern English with Special Reference to Proper Names with

an Introductory Theory of Grammar, Meaning and Reference. Copenhagen. Stockwell, R. P. & Schachter, P. (1962), "Rules for a Segment of English Syntax" (mimeographed). Strang, Β. Μ. H. (1962), Modern English Structure. London. Svartengren, H. (1948), " 'Become' as an Auxiliary for the Passive Voice", Moderna Sprdk 42.272-81. Sweet, H. (1955), A New English Grammar, Part 1. Oxford. Turner, G. W. (1962), "The Passive Construction in English Scientific Writing", AUMLA 18.181-97. Twaddell, W. F. (1963), The English Verb Auxiliaries. 2nd ed. Providence, R. I. Visser, F. Th. (1941-56), A Syntax of the English Language of St. Thomas More. Louvain. Vogt, Η. (1950), "Un aspect du problfeme actif-passif dans le verbe", Journal de Psychologie 43.130-8. Vossler, Κ. (1925), "Das Passivum, eine Form des Leidens oder des Zustandes?", Die Neueren

Sprachen 33.401-7. Webster's Third New International Dictionary of the English Language (1961). Springfield, Mass. Zandvoort, R. W. (1960), A Handbook of English Grammar. London.

(1961), "Grammatical Terminology", English Studies Today, Second Series, 283-94. Bern.

I N D E X

A, see adjunct Λ-item, /I-set 58 abstract (nominal group head) 99, 108, 112, 165 acceptability, linguistic 6, 164f. actional passive 86, 93 actional-durative passive 93 active, see transformation et passim adjective 27, 96,101,128,135,145,191f. adjunct (A) 25f., 70, 97, lOlff., 127, 158

, agentive (Ag) 102ff., see also agent , close MCLOSE) 140ff„ 189ff., , criteria relating to the adjunct 74,79, lOlff. , exponent classes 29ff., 41, 57ff., 71,128 , nonagentive (Ad) 57 ff., 108f. , open (ΛΌΡΕΝ) 140ff. , position 30, 58ff„ 110,112,132,156 , sentence adjuncts 29 , verb-adjuncts 29

adverb 29, 108, 157 agent (Ag)

, Ay-agent 102ff., 123,125,133,141ff. , definition 30 , exponent class 41, 49 ff., 52, 116, 129ff.,

157, 166 , extension 108,112,123,126ff„ 134f.,

147ff., 158 , external 135 , internalized 135 , Janus-agent 104ff., 133, 141, 157 , length 52ff., 112, 129ff., 156, 166 , quasi-agent 102ff., 134, 147 , zero 126

agentful/agentless 9, 40, 81, 129, 134, 141, 151, 156 ff.

agentive/nonagentive 9, 40, 137f., 140ff., 147ff., 164

'agentless passives', see Class γ Akerlund, A. 3 Allen, W. S. 2 animate/inanimate 27f., 34, 50, 108, 112, 123f.,

141 'animate agent passives', see Class α

aspectual constraint 85 f. 'attitudinal passives', see Class δ1

auxiliaries 11 ff., 9Iff. .elliptic42,67, 80 , lexically marked/unmarked 93, 134ff.,

147 f. , mutative/nonmutative 92ff., 136, 148 , Type a 10,12ff., 91,128,151f.; closed/open

class 14ff., 151, 168; fully tense-marked, non-tense-marked, semi-tense-marked 16f., 91

, Type b 10,17, 91,126,168 , Type c 10,17ff„ 91f., 168 , Type d 10,19,92ff., 168

Bach, Ε. 1, 6,34 balance 156, 166 base, verbal/nonverbal 22,95 become-passive 92ff., 135f., 149f., Bloomfield, L. 31 Bolinger, D.L. 26, 89,135, 159 Brose, B. 3 Buyssens, E. 5 Ay-agent, see agent

C, see complement; C-full/less 65 Carvell, Η. T. 21, 82, 111, 124 case alteration 132,156 Chomsky, N. If., 83,90,105,148 CL, see clause 'CLASP' (classification program) 113 Class α ('animate agent passives') 40, 123, 132f.,

136, 138, 141 ff., 156ff. Class β ('inanimate agent passives') 40,123,132f.,

136, 138, 141 ff., 154f. Class β/γ ('Janus-agent passives') 133,138,141 ff. Class γ ('agentless passives') 40, 123,134ff., 138,

141 ff., 159 Class δ1 ('attitudinal passives'), Class δ® ('emotive

passives') 134f., 138, 157ff. Class ε ('nonagentive passives') 135ff., 138, 159 Class ζ ('compound passives') 137f. classification of passive clauses 72-138 clause 13, 25-38

198 INDEX

clause, criteria relating to the clause (CL) 74,75, 78 ff.

, elements V, W, S, C, A 25f., 41 , finite/nonfinite verb clause 27,29, 55ff., 71,

101, 104, 110, 144f., 147,183ff., 188 , parenthetic 81,144 , relation: external 31 ff., 41, 65ff., 71, 78ff.,

126ff., 158; internal 31, 69, 70, 82 , sequence 156 , types 33,36f., 41,62ff., 71,139ff.; extensive

(EXT, ex/)/intensive(INT, int) 33ff., 83f., 123, 128, 134, 139

cline distribution 116, 132 colligation 102 collocation 15, 102, 138, 165 common 27 f. communicative dynamism (CD) 38 commutation 22f., 124f., 128 complement (C) 20, 25ff., 41, 64f., 188f.

, criteria relating to the complement 74, 78, 101

, exponent classes 26ff., 55ff., 70f., 145 compounds 21, 86, 95 'compound passives', see Class ζ conclusive/non-conclusive verbs 85 f. concord, covert/overt 25; valency classes 16ff. concrete 99,108,112,125,165 conjunctional clauses 32, 67, 80 constituent structure 36 contextualization 21 contingency tables 116ff. continuum 26 conversion 37 coordination 28, 52, 96, 100, 125, 128f., 134f.,

137, 157, 166 copula 34, 38, 93 f. coreference 33,37. 81 ff., 124,127f., 158,165 corpus 5ff.

, Corpus I, see Voice Corpus , Corpus II, see Minor Passive Corpus , Corpus III, see Taxonomic Passive Corpus , Corpus IV, see Major Passive Corpus

countable 99,112,125 ff., 129,157 criteria 73, 74,123 Crystal, D. 6, 20, 82 Curme, G.O. 3, 86, 89

data: preparation of input data 111 ff.; voice data processing 39-71

definite 28,99,125 f., 129,157 deletion 35,106,125,157 delicacy, analytical 26, 138 determiner 26 diagnostic frames 21 diagnostic key 132ff. 'dichotomies' 112

dispensability 26 ditransitive clause 37, 62ff., 83, 138ff., 188f. do-periphrasis 12, 25 double-tail test 119 durative passive 92

'emotive passives', see Class δ2

emotive verbs 84 equative clauses 37,62ff, 84,86,93,135ff., 159ff. equative relation 140 Erades, P. A. 3, 90,161 exponential coreference 33 extensive, see clause

factitive clause 37,62ff.,83,86,136, 138ff, 160, 191 ff.

feature 11 Iff., 123 ff., 132ff, 158 finite verb clause, see clause finitude 28, 99,125f., 157 Finney, D. J. etal. 116 Firbas, J. 25, 38,166 Firth, J. R. 5 Francis, W. N. 3,4,15,19, 27, 29, 86,93 Frary, L. G. 3,92 free clauses 31ff., 65 ff., 81,128,158 Fries, C. C. 3,4,5,27,37,86,92 Fröhlich, J. 3 functional sentence perspective (FSP) 166

Gaaf, W. van der 3 Garvin, P. L. 41 gender 27f„ 34,50f., 57,70,112,125,157 ^/-passive 92ff, 136f., 149f. Gleason, H. A. Jr 165 Godfrey, J. 6 Green, A. 3, 31,103 group, linguistic 5 et passim Groups I-VI (taxonomic) 115f., 119ff.

Halliday, Μ. A. K. 5,25,26, 33 Hatcher, A. G. 3,93 Hendriksen, H. 3, 31 Herdan, G. 116 Hill, A. A. 3,4,10,13,15,165 Hill, L. A. 3 Hockett, C. F. 3

imperfective passive 86 'inanimate agent passives', see Class β indefinite 28,99,125f„ 129,157 indeterminacy, structural 133 infinitive

, to-full (to V) 14f., 27, 30, 57, 101, 104, 146f., 185,192f.

, to-less (K) 12,14f.,27 input 39, 41 intensive, see clause

INDEX 199

intransitive clause 37, 62ff., 159 ff. introductory there 86 invariance for person and number 12

Jacobson, S. 30 Janus-agent, see agent 'Janus-agent passives', see Class β/γ Jespersen, Ο. 1, 3, 4, 22, 31, 37, 86, 88, 89, 92,

102, 141 Joos, M. 3 Jud-Schmid, E. 3

Kennedy, A. G. 20 kinetic passive 86 Kirchner, G. 3 Klingebiel, J. 3 Koumari, A. 3, 31 Kruisinga, E. 2, 3,4, 86 Ku£era, H. 41 Kurtz, G. 3

Lamb, S. M. 41 Lees, R. B. If., 105 levels (in computer program) 112; (in significance

test) 118 Long, R. 90 Longacre, R. E. 37 Lyons, J. 22

'macrosemantic' similarity 147,159 Major Passive Corpus (Corpus IV) 9,139ff., 151,

167ff. Marchand, H. 22 mass nouns 99,112,125ff., 129,157 material 6ff. McKerrow, R. B. 2 Meier, Η. H. 3 Meyer-Lübke, W. 3 Mihailovic, L. 3 Minor Passive Corpus (Corpus II) 9, 39ff., 129,

157 Mitchell, T. F. 20 modification 26, 28, 96,112,124ff., 127 ff„ 157 f. Moroney, M. J. 116 Mustanoja, T. F. 3,103 mutative, see auxiliaries

N, see nominal element name 26,28,98,112,129 negation, postverbal enclitic 12,15,18 Nida, E. A. 3 nominal group 26f., 98,157 nominal element (N) 36, 145ff., 187f., 191 f. nominalization 26 f. 'nonagentive passives', see Class ε noun 26,52,55ff., 98,101,112,129,157 Nowell-Smith, P. 148

Null Hypothesis 116f. number 26, 28, 34,100,140 numerical taxonomy 72f., I l l

object, (O) 37f„ 127f„ 158 , direct {Od), indirect (O.) 37,140

obligatory 25, see feature Olsson, Y. 10, 16, 90 operational taxonomic unit (OTU) 73, 111 optional 25, 58, see feature OTU, see operational taxonomic unit output 39,41 Owen, Ε. T. 3 Oxford English Dictionary 23, 103,108, 162

p, see probability P, see predicative Palmer, F. R. 3 participle: present (Ving), past (Ved) 27, 30, 97,

101, 145, 191ff. passive of 'being' and 'becoming' 86 'passive scale' 134ff., 138, 156ff. passive voice: definitions 3ff., use of 139-55 perfective passive 86, 92 permutation 88 ff., 125 person 28,100,112 phrase, prepositional 20f., 29,96,104,109f., 140,

144 Pike, K. L. 5,11 plural (pi) 34 'positives' 112 Poutsma, Η. 1, 37,86,89,92,102 predicate 37 predicative (P) 37f., 127f„ 137,158 probability (p) 116ff. program (computer) 39, 41f., 112 proleptic it 13, 81,86,112,144ff., 183ff„ 188 pronoun 26,28,52, 55ff., 98,112,127,128f„ 157

qualification 95,128,134f., 147 f. 'qualitatives' 112 'quantitatives' 112 quasi-agent, see agent 'quasi-agentive passives', see Class δ question-adverb 35,102

Quirk, R. 6,10,15 f., 20, 25, 74, 82,116,159, 164

rank 5, 25 referential pronoun 28,98,112 Reichmann, W. J. 116,118 relationship: internal/external 115 ff.; serial 159 ff. relative clauses 32,67,78,128 relator: fixed, mobile 32, 67, 80; sequential 29 rheme 38, 166 Roberts, P. 95 Robins, R. H. 15,102

200 INDEX

S, see subject scale 26, 33, 38 Schachter, P. 34 Scheurweghs, G. 4, 86 semiditransitive 139ff., 185 ff. semifactitive 139ff., 189ff. semitransitive 62 sequential constraints 85,132 sequentially related clauses 25, 31 f., 65 if., 80 serial isolate 162 serial voice relation 159ff. significance, statistical 116ff. similarity

, coefficient 113 , inter-group, internal, intra-group 115 ff. , matrix 113ff. , mean 114ff.

single-tail test 118 singular (sg) 34 Sledd, J. 37 Smith, C. S. 33 Sneath, P. H. A. 72 Söderlind, J. 3 Sokal, R. R. 72 Serensen, H. S. 27 speech, direct 27, 55 ff., 64,144,183ff. statal passive 86,93,135,150,161 f. static passive 86 statistical assessment 116ff. Stockwell, R. P. 34 Strang, Β. Μ. H. 3,10,20,25,27, 29,93 structure 5 et passim style 70, 87,132 subject (S) 25ff., 37,118ff., 158

.converted 1, 31 , criteria relating to the subject 74, 77, 97 ff. , elliptic 32, 67, 80 , exponent classes 26ff., 41, 49ff„ 70f., 116,

124ff., 134,147, 156ff., 166 , external relation 81,158 , grammatical 89 , inverted 31 , length 52 ff., 70,112,128 ff. , logical 89 , referential 144 ff., 167ff., 185 , suppletion 84

subscript 34 substitution classes 27 superscript 34 Survey of English Usage 6 Svartengren, H. 3, 93 Svartvik, J. 6, 15f., 21, 72, 74, 111, 116, 124,164 Sweet, H. 31 syntactically bound clauses 3If., 65ff., 78ff., 127,

158 syntagmatic affinity 138

system 5 et passim systemic coreference 33,36,37 systemic voice transformation potential 36, 132,

134, 156

tag question 32, 67 Taxonomic Passive Corpus (Corpus III) 9, 73 ff.,

157 taxonomy, see classification term 5 et passim theme 166 transformation

, constraints 84,139 , dual 128 , potential 24, 26, 34, 36, 83ff., 88ff., 112,

123, 127 f., 134ff., 147f., 158 , theory 1 , voice relation 159ff.

transformational structure (of major clause types) 36

transitive clauses 37, 62ff., 83, 86, 138, 167ff„ 183ff.

transmutation 88 ff., 128f., 135,162 Turner, G. W. 3,86 Twaddell, W. F. 10,165 Types a, b, c, d, see verbal group

un-, negative/reversative 22, 86

V, see verbal group valency classes 16ff. verbs

, auxiliaries; lexical verbs 11, 19ff., 41, 67, 95 ff., 112, 165, 167

, phrasal/prepositional 19ff., 95, 165 verbal group (K, W) 10-24, 25f„ 41, 43ff„ 70f.,

150, 158 , criteria relating to the passive verbal group

74, 76, 91 ff. , simple 10, complex 10-19

verbful construction 38 verbless construction 30, 38, 55 ff. Visser, F. Th. 3 Voice Corpus (Corpus 1) 9, 39ff, 158 voice

, data processing 39-71 , distribution 70 , relation 138 , transformation, see tranformation

Vogt, Η. 5 Vossler, Κ. 3

W, see verbal group

Webster's Third New International Dictionary 162

Xä-test 46,116ff.

Zandvoort, R. W. 4,5,86,92