A Cognitive-Functional Approach to Nominalization in English [Reprint 2015 ed.] 9783110903706, 9783110178098

Citation preview

A Cognitive-Functional Approach to Nominalization in English

W G DE

Cognitive Linguistics Research 26

Editors

René Dirven Ronald W. Langacker John R. Taylor

Mouton de Gruyter Berlin · New York

A Cognitive-Functional Approach to Nominalization in English

by Liesbet Heyvaert

Mouton de Gruyter Berlin · New York 2003

M o u t o n de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter G m b H & Co. K G , Berlin

© Printed on acid-free paper which falls within the guidelines of the A N S I to ensure permanence and durability.

ISBN 3 11 017809 5 Bibliographic information published by Die Deutsche

Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at < h t t p : / / d n b . d d b . d e > .

© Copyright 2003 by Walter de Gruyter G m b H & Co. K G , D-10785 Berlin All rights reserved, including those of translation into foreign languages. N o part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Printed in Germany

To Wouter, Judith and Oscar

Acknowledgements

The present study reflects a research project to which many people, each in their own way, have contributed. I would like to thank those people here. First and foremost I would like to thank Kristin Davidse. As my Ph.D. supervisor, she was involved in my research on nominalization patterns from its earliest stages onwards. I am deeply grateful for the continual encouragement she has given me, for the openness with which she has shared her own learning with me, and for the precision with which she has read my texts, time and again supplying extensive commentary. This book, in which any remaining shortcomings are my own, in the first place represents the dialogue I have been able to conduct with her. I would also like to thank Hubert Cuyckens, Ad Foolen, Ronald Langacker and Willy Van Langendonck for their rigorous comments on an earlier version of this book. In addition, I am grateful to René Dirven and Ronald Langacker who, as editors of the Cognitive Linguistics Research Series, guided me through the proposal and writing process and whose careful comments on various chapters of the book I have greatly appreciated. I thank the managing editor of the series, Birgit Sievert, for her patience and for her professional assistance throughout the publication process. I thank Jürgen Benteyn for his swift and always good-humoured help with the formatting of the manuscript, and Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte for their general support. Thanks also to Lieven Vandelanotte for his help with the drawings and to An Laffut and Lieven Vandelanotte for proof-reading parts of the manuscript. My thanks also go to a number of people who have in the past years commented on my work and helped me with their suggestions and criticisms. In addition to Hubert Cuyckens, Kristin Davidse, Ad Foolen, Ronald Langacker and Willy Van Langendonck, these are: Eirian Davies, Renaat Declerck, Dieter Kastovsky, An Laffut, Odo Leys, Lachlan Mackenzie, Bill McGregor, Hans-Jörg Schmid, Lieven Vandelanotte, Frederike van der Leek, Jean-Christophe Verstraete and Emma Vorlat. The University of Leuven has contributed to the publication of this book by allowing me to work on it as a postdoctoral fellow of the Research Council (postdoctoral grant PDM/02/034). I am most grateful to my parents and my parents-in-law for their unfailing support. Finally, I would like to thank Wouter, without whose unconditional support, practical help and love I would never have been able to bring this book to a good end.

Contents

Part I Towards a theoretical-descriptive approach to nominalization

Chapter 1: Introduction

3

Chapter 2: Theoretical assumptions

9

1. 1.1. 1.2. 1.3. 2. 2.1. 2.2. 3. 3.1. 3.2. 3.3. 3.3.1. 3.3.2. 3.4. 4. 5. 5.1. 5.2. 6.

Language: The relation between system and usage A usage-based approach to language Cognitive Grammar: The interface between system and usage The language system as network The natural symbolic relation between lexicogrammar and semantics A natural description of language The natural or non-arbitrary relation between grammar and meaning Constructions: Composition, classification and function A functional approach to composition Levels of functional analysis A functional approach to classification An internal functional analysis of constructions Classes as complex categories: Schema and prototype Composition and classification: Conclusion Functional layers of organization Constructions and the relations between them Agnation and enation: A definition Agnation and enation: Theoretical significance and heuristic value Theoretical assumptions: Conclusion

9 9 11 13 15 15 17 19 20 21 23 24 26 28 29 33 35 37 40

Chapter 3: Nominalization

41

1. 1.1. 1.2.

42 43 44

A usage-based account of nominalization Schematization and extension Entrenchment

χ Contents 2. 3. 3.1. 3.2. 3.2.1. 3.2.2. 4. 5. 5.1. 5.2. 5.3. 6.

Nominalizations as non-arbitrary symbolic units Nominalizations as composition and reclassification Constructions as assemblies of symbolic units Nominalization as functional reclassification Rank shift and reclassification Nominalization as functional reclassification A multifunctional account of nominalization The role of agnation in the analysis of nominalization Networks of agnation Agnation as a descriptive heuristic The schematic nature of agnation Towards a theoretical-descriptive approach to nominalization: Conclusion

46 47 48 49 49 51 52 52 54 55 56 58

Chapter 4: The functional organization of nominal and clause

61

1. 2. 2.1. 2.1.1. 2.1.2. 2.2. 2.3. 2.3.1. 2.3.2. 2.3.3. 2.4. 2.5.

From noun type specification to nominal From process type specification to clause The clausal head Finite vs. atemporal clausal heads The clausal head, Objects and Complements Instantiation and quantification of the clausal head Clausal grounding The finite element Person deixis Clausal grounding: Conclusion The speech-functional role of the Subject The functional organization of nominal and clause: Conelusion

61 70 72 73 76 78 81 82 85 86 87 92

3.

Nominalization as functional reclassification

94

Part Π Deverbal -er nominalization Chapter 5: Deverbal -er suffixation: Towards a descriptive position 0. 1. 1.1.

Introduction -er nominalizations: Agent names or Subject names? Agentivity and salience

99

99 101 102

Contents xi 1.1.1. 1.1.2. 1.1.3. 1.2. 1.2.1. 1.2.2. 1.2.3. 1.3. 2. 3.

-er nominalizations as natural symbolic units -er suffixation as a complex conceptual category The verbal base of deverbal-er nomináis Non-agentives as external arguments -er nomináis profiling oblique participants -er nomináis based on unaccusatives Middle-based-er nomináis -er nominalizations as Subject names Lexicalization versus ad hoc nominalization Towards a coherent account of deverbal -er nominalizations

104 106 108 110 115 116 117 118 120 123

Chapter 6: The middle construction

125

1. 2. 3. 4. 4.1. 4.2. 5.

Middle formation: A state of the art The representational versatility of the middle formation Towards a semantic typology of middle constructions Middle constructions and modality The modality of 'letting' The middle as 'letting' construction Conclusion

126 128 132 137 140 143 147

Chapter 7: A multifunctional approach to -er nominalization

149

0. 1. 1.1. 1.2. 2. 3. 3.1. 3.1.1. 3.1.2. 3.1.3. 3.2. 3.2.1. 3.2.2. 3.2.3. 3.3. 4. 5.

149 151 152 153 154 158 159 159 161 163 165 166 168 172 173 174 176

Introduction Representational semantics Agentive-er nominalization Non-agentive-er nominalization A semantic typology of non-agentive -er nominalizations The constructional properties of -er nominalization Lexicalized -er nominalizations Non-agentive -er nominalizations Agentive -er nominalizations A note on instrumental -er nominalization Ad hoc nominalizations Phorie ad hoc nominalizations Non-phoric ad hoc nominalizations Ad hoc-er nominalization: Conclusion The constructional properties of -er nomináis: Summary Deverbal -er suffixation as Subject-profiling Conclusion

xii

Contents

Part ΙΠ Factive nominalization

Chapter 8: Factive nominalization: Towards a descriptive position

181

0. 1. 2. 2.1. 2.2. 2.3. 2.4. 3. 3.1. 3.2. 4. 4.1. 4.2.

181 183 185 186 187 188 194 195 195 198 200 201 203

Introduction Factivity as truth presupposition Factivity as embedded projection Embedding vs. taxis Projection vs. expansion Facts as embedded projections Conclusion Delineation of the fact category Ambiguity and vagueness Gerundive nominalizations: Acts or facts? Towards an internal, nominal analysis Halliday's dependency analysis Type vs. instance nominalization

Chapter 9: A functional analysis of factive nominalizations as nominal constructions

207

1. 1.1. 1.2. 2. 2.1.

207 208 209 211 213

2.2. 2.3. 3. 3.1. 3.2. 3.2.1. 3.2.2.

ί/ζα/-facti ves: From finite clause to proper name strategy Downranking as functional reclassification The nominal paradigm the fact íAútf-constructions: A case of apposition The structural characteristics of the fact thatconstructions The functional properties of the fact íAaí-constructions the fact /to-structures: Conclusion Gerundive facts: From atemporal clausal head to common noun or proper name strategy From atemporal clausal head... .. .to common noun or proper name strategy Gerundive nominalizations: from transformation to nominal constituent Towards a systematic account of the nominal properties of gerundive nominalizations

217 221 221 222 228 229

231

Contents

3.2.3. 3.3. 3.4. 4.

The nominal functioning of gerundive nominalization: Conclusion Factive gerundive nominalization: Integrating the clausal and nominal paradigms Factive gerundive nominalization: Conclusion The internal organization of factive nominalizations: Conclusion

xiii

239 241 244 244

Chapter 10: Conclusion

249

Appendix

255

Notes

257

References Index

267 281

Parti Towards a theoretical-descriptive approach to nominalization

Chapter 1 Introduction

The focus of the present study is on deverbal nominalization in English. The analysis of nominalization that is presented in it has mainly been prompted by two observations. First, the existing descriptive analyses of nominalization fail to embed the peculiarities of specific nominalization types in a coherent and systematic theoretical account of the fundamental constructional mechanisms that underlie nominalization in general. Secondly, the description of nominalization systems tends to be biased towards the clausal categories which they realize, to the neglect of their nominal characteristics. Following Lees's (1960) influential analysis of nominalization in English, nominalizations are still basically regarded as clause-like constructions used in nominal units in an unfathomable way. The absence of a coherent theoretical perspective on the process of nominalization has led to descriptive fragmentation. Morphological processes of nominalization such as -er or -ee suffixation, for instance, seem hardly relatable to nominalization types which incorporate clause-like structures, as in [His answering the phone] surprised me. Also, the one-sided, clause-based nature of most descriptions has elucidated only part of the lexicogrammatical and semantic properties of nominalizations. These problems reflect deficiencies of theory as well as of description. A central theoretical problem is, for instance: how can we, given the prevailing emphasis on structural constituency, model nominalization, which seems to elude and transcend ordinary constituency relations? A central descriptive problem is: which features of clause and nominal enable the reclassification from the one into the other that is generally held to be involved in nominalization? Most attention so far seems to have gone to the reclassification of verb into noun, and the 'reification' of a situational referent into an entitized one that seems to go with it. However, as a reclassification mechanism, this is too thin. There must be other factors, mainly analogies between clause and nominal, which motivate the possibility of reclassifying a clausal configuration into a nominal one. To tackle the theoretical as well as the descriptive issues, this study is composed of a general theoretical-descriptive part (Part I) and two case studies (Parts Π and ΠΙ). Part I presents the theoretical and descriptive constructs required to model nominalization. Parts Π and ΙΠ present two case studies and determine in how far the proposed theoretical-descriptive ap-

4 Introduction proach can account for the lexicogrammatical properties and meanings of specific nominalization systems. Part Π zooms in on the system of deverbal -er nominalization (e.g., remind -» reminder) and in Part ΙΠ, three types of factive nominalization are considered (viz. [That he killed her] surprises me, [The fact that he killed her] surprises me and [His killing her] surprises me). The main theoretical motif in this study is that nominalization strongly calls for a functional rather than purely structural approach. Crucially, the reclassification from verbal/clausal starting point into nominal class is functional in nature: nominalization involves the functional reclassification of a particular level of assembly in the organization of the clause into a nominal structure. A nominalization always adopts external nominal functions, but, as will become clear, it may also acquire internal nominal functional organization. The reason why the identification of nominalizations as constructions of the nominal class requires a functional view on composition is that nominalizations do not always have discernible structural components for each function which they realize: as in ordinary nomináis, the internal functions which nominalizations realize are not necessarily reflected iconically in their structure. It is therefore only a functional analysis which can shed light on what nominalized structures have in common with the iconically and non-iconically coded members of the nominal class in general. Moreover, it is only by adopting a functional viewpoint that one can relate nominalizations which contain part of a clause or a full clause (e.g., her signing the contract, that she signed the contract) to nominalizations at word level (e.g., baker). Like nominalizations based on a simple verb stem (e.g., bake —• baker), nominalizations containing a clausal unit reclassify this clausal unit into a unit with nominal functional behaviour. Because the analysis of nominalization requires a radically functional approach, I will in Chapter 2 discuss two linguistic frameworks which have shown a particular interest in the functional nature of language, i.e. Cognitive Grammar (as developed in Langacker 1987a and 1991) and SystemicFunctional Grammar (Halliday 1994). I will go more deeply into some of their basic tenets, such as, firstly, the interaction between language usage and the language system (Section 1); secondly, the natural symbolic relationship between lexicogrammar and semantics (Section 2); thirdly, the 'intrinsic' functional nature of linguistic structure and the need for a functional approach to the notions of 'rank shift' and 'class' (Section 3); fourthly, Halliday's multifunctional approach to language (Section 4); and, finally, the need for an accurate analysis of not only the syntagmatic layer of linguistic organization, but also of the paradigmatic, systemic aspect of language (as revealed by the schematic networks or system networks of

Introduction 5 related structures and options in which both Cognitive Grammar and Systemic-Functional Grammar situate a construction) (Section 5). In Chapter 3,1 will show how these basic theoretical assumptions enable us to identify some of the lacunae and weak points in the existing approaches to nominalization. Importantly, these principles also point out a possible course of action in the description of nominalized constructions. Going through the main theoretical divisions of Chapter 2 again, I will argue, firstly, that nominalization patterns have to be categorized both by schematization and extension. It also has to be taken into account that nominalizations may display varying degrees of entrenchment or automatization: they may be lexicalized and function as fixed expressions in language, or they may only be established as grammatical patterns or schematic units, without themselves being included as conventional units in the language system (Section 1). Secondly, I will stress the importance of viewing nominalizations, like all constructions, as natural symbolic units of which the meaning is encoded in the lexicogrammar (Section 2). Thirdly, I will argue that nominalizations have to be viewed as functional configurations which are situated on a particular level or rank of functional organization (Section 3). As such, they find themselves in between a 'higher' rank in which the nominalization itself serves a function and a 'lower' level that constitutes the nominalization's own internal organization. A full description of a nominalized structure requires an analysis of its external, synthetic functioning, of its internal, analytic functional outlook, and of the way in which the external and internal functions are combined in the construction itself. Put differently, some nominalized constructions may represent an at first sight inexplicable mixture of external, nominal behaviour and internal, clausal categories, but the nominalized structure itself succeeds in combining these seemingly divergent functions into one construction and therefore has to be accounted for. Moreover, for a number of nominalizations, we need a concept of downranking, which allows the possibility of functional configurations occurring at a level of organization which is not the one on which they normally function: clauselike nominalizations such as íAaí-clauses can then be identified as having been downranked from clausal to nominal level while preserving part of their internal clausal outlook. By the same token, downranking entails its own form of reclassification and involves external as well as internal reclassification: by being reclassified, the clausal structure takes on the grammatical and semantic properties inherent to the nominal unit. Failure to recognize the reclassification involved in the rankshifted type of nominalization has been one of the main gaps in the description of nominalization so far.

6

Introduction

Fourthly, nominalizations are best described from a multifunctional perspective. They encode representational semantics, they are related to the ground and they serve textual functions. Their import cannot therefore be reduced to their representational semantics (Section 4). Finally, I will also argue that, if it is used systematically, the paradigmatic relationship of agnation forms an important descriptive heuristic in the analysis of nominalization patterns (Section 5). It helps to shed light on the clausal categories that are present in nominalization, as well as on the similarities and differences between distinct nominalization types such as factive thatnominalizations, the fact ^/-constructions and gerundive factives. The theoretical principles discussed in Chapters 2 and 3 reveal that an insightful description of the functional properties of nominal and clausal organization is needed before one can tackle those of nominalized constructions. In Chapter 4, therefore, I will complete the theoretical-descriptive framework by elaborating on the internal functional organization of nomináis and clauses. I will argue that nominalization or the functional reclassification from processual starting-point to nominal construction is possible due to the fundamental correspondences that exist between the functional organization of the nominal and that of the clause: clausal categories can be mapped onto what is basically a nominal structure because the functional organization of clauses and nomináis is to a large extent equivalent (Langacker 1991). Chapter 4 will be organized around the functional categories which Langacker (1991) identifies for the nominal and the clause, i.e. those of type specification, instantiation, quantification and grounding. I will first demonstrate the relevance of these functional categories to the description of clauses and nomináis (Sections 1 and 2). In a third section, then, I will formulate a tentative analysis of the basic constructional mechanisms that underlie nominalization. I will argue that the integration of clausal and nominal properties in nominalized constructions develops along the lines of the functional categories of type specification, instantiation, quantification and grounding. Parts Π and ΠΙ of this volume will then present two descriptive case studies. The nominalizations which I will analyze in them broadly cover the spectrum of clausal levels of assembly to which nominal reclassification can apply: they include nominalizations derived from a simple verb stem, i.e. -er nominalizations (e.g., baker, bestseller) (Part II); nominalizations based on what I will call an atemporal clausal head (e.g., signing the contract) and nominalizations which are centred on finite clauses (e.g., that she signed the contract, the fact that she signed the contract) (Part ΙΠ). The latter three structural types of nominalizations will be treated together because they have part of their semantics in common and can all be 'factive'.

Introduction 7 Each case study will start with a chapter in which I look at the main insights thus far offered in the literature and establish my own descriptive position. In Chapter 5 an overview will be given of the literature on deverbal -er nominalization; Chapter 8 will discuss the most interesting analyses that have thus far been proposed for the system of factive nominalization. In my own descriptive analysis of -er derivation and factive nominalization (developed in Chapter 7 and Chapter 9 respectively), I will attempt to show that it is possible to move beyond existing descriptions and provide an answer to some of the long-standing moot points in the analysis of specific nominalization types if the functionally-oriented approach which I set out in Part I is systematically applied. Importantly, my analyses will focus not on the external functioning of the nominalized constructions, but on their internal functional organization, i.e. on the clausal and nominal categories which they integrate. Apart from the general functional perspective which I will take and the central role which I will assign to the functional categories of the nominal and the clause, two main motifs can be said to run through both my descriptive case studies. First, my analysis will be based on the assumption that the semantics of a construction is encoded in its lexicogrammatical properties and can therefore be revealed through a careful analysis of those properties. I will, in other words, not only present a detailed analysis of the lexicogrammar of -er nominalizations and factive nominalizations; I will also try to shed light on the basic meanings which these nominalization types encode. Secondly, in each of these studies, the description of the syntagmatic integration of clausal and nominal properties in the internal organization of the nominalized constructions is supplemented with observations about the paradigmatic relations which the nominalized structures hold. In other words, the syntagmatic properties of nominalizations will be systematically considered against the background of the larger system network which they form part of and which they select from. Not surprisingly, the paradigmatic relations that will turn out to be most valuable to the elucidation of nominalizations are those with clausal structures and with nomináis. First, due to the clausal nature of their starting point and the clausal categories which they embody, nominalizations tend to be related to clausal structures. The identification of related or 'agnate' clause types will be shown to play a crucial role in the analysis of nominalized constructions. In my description of -er nominalizations (Chapter 7), I will thus elaborate on the systematic relationship that exists between -er suffixation (e.g., bestseller) and the clausal middle construction (e.g., That book sells well) and I will show how an accurate analysis of the lexicogrammatical properties of middle clauses

8 Introduction (presented in Chapter 6) sheds a new light on the system of -er nominalization. Apart from being related to clausal structures, however, nominalizations also connect with nominal constructions: by being made to function within the nominal paradigm, nominalizations enter into relation with other nominal structures, nominalized as well as non-nominalized. The gerundive type of factive nominalization, that-f&ctives and the fact ¿Ααί-constructions can thus be considered as nominalized structures which are systematically related to each other and are linked up within the system of language. In my analysis of factive nominalization, I will show that it is only by taking into account the properties of each of them that one can come to a more accurate description of the overall system of factive nominalization and its factive reading (Chapter 9). Importantly, the various types of factive nominalizations do not only relate to each other: as nominal constructions, they also enter into relation with specific types of non-nominalized nominal structures. Once again, the elucidation of the precise paradigmatic relations which each nominalization type has with 'ordinary' nominal constructions will be argued to be vital to the elucidation of its properties. In the Conclusion (Chapter 10), finally, I will attempt to answer the question as to what constitutes a nominalization.

Chapter 2 Theoretical assumptions

Nominalization represents a theoretical challenge and has to be situated in an overall theory of the language system. It is therefore essential to first clarify the basic theoretical assumptions on which my analysis of nominalization is based. This is even more important as my approach is eclectic and some of the terminology which I use requires clarification. In this chapter I will therefore zero in on the linguistic system and discuss the various 'levels' that can be distinguished in it: in a first section, I elaborate on the language system as such and its relation to language usage·, the second section deals with the various strata that can be distinguished in language and focuses on the natural relation between lexicogrammar and semantics; in Section 3,1 deal with the ranks or the levels of organization that can be distinguished within the grammar and I introduce two of the notions that are crucial for the analysis of nominalization, viz. reclassification and rank shift. Section 4 focuses on the multifunctional approach which I advocate for the description of nominalization patterns; and, finally, Section 5 points out the heuristic value of considering the relations that exist between different constructions.

1. Language: The relation between system and usage 1.1. A usage-based approach to language The analysis of nominalization set out in this study is rooted in a view of language which is at once functional and cognitive. It is functional in that it builds on the assumption that the nature of the language system is "closely related to the social and personal needs that language is required to serve" (Halliday 1970b: 142): the language system is viewed as functional with respect to the particular needs of the language user. More specifically, the language system is claimed to be shaped by the functions which it serves, i.e. they are built into it. An analysis that is functional in design can thus "relate the internal patterns of language - its underlying options, and their realization in structure - to the demands that are made on language in the actual situations in which it is used" (Halliday 1970b: 165) (in Section 4 the

10

Theoretical

assumptions

functional level of organization in language will be discussed in more detail). Apart from viewing language as functional, I also adhere to the claim that the language system - and the constructions in it - is determined by certain cognitive abilities of the language user: the language system is viewed as an integral part of human cognition (Langacker 1987a). Language is held to refer to "those aspects of cognitive organization in which resides a speaker's grasp of established linguistic convention", and the linguistic abilities of the language user are believed to "not necessarily constitute an autonomous or well-delimited psychological entity" (Langacker 1987a: 57): there are a number of basic psychological phenomena involved in language use which are essential to the language system, without being limited to it (Langacker 1999). Because the cognitive processes which they embody equally contribute to their make-up, an analysis of linguistic structures should articulate what we know of cognitive processing, i.e. it must have "psychological reality" (Langacker 1987a: 42). The language system is, however, not only shaped by the functions which it serves and by the cognitive abilities of the ones who use it: it also itself sanctions language use. "Usage events" (Langacker 1987a: 66) or actual "instantiations" of language (Halliday 1992: 20; Langacker 1999: 99) are sanctioned by the restrictions and options of the system of language underlying them: they are 'systemically motivated' (Langacker 1987a: 426). As Hjelmslev (1961: 39) puts it, an actual instance of language (which he calls a 'process' or 'text') only "comes into existence by virtue of a system's being present behind it, a system which governs and determines it in its possible development". This is not to say that language is a generative or constructive system which generates well-formed, novel structures as its output: it is not the language system which is responsible for assembling novel expressions. Construing language is viewed as "something that speakers do" (Langacker 1987a: 65), which ties in well with Hjelmslev's 'processual' interpretation of language use (Hjelmslev 1961: 39). Actual language use is "a problem-solving activity that demands a constructive effort and occurs when linguistic convention is put to use in specific circumstances" (Langacker 1987a: 65). In short, the language system may provide the necessary "symbolic potential", but it is left for the speaker to recognize this potential, i.e. to "exploit it in a fashion that responds to all the varied constraints inherent in the situation" (Langacker 1987a: 66). It is the dynamic interaction between system and usage, or the confrontation of the conventionalized language units in human cognition with the constructive effort of the language user which forms the "crucible of lin-

Language: The relation between system and usage 11 guistic structure" and the "source of language change" (Langacker 1987a: 70). If we are to elucidate particular constructions in the language system, therefore, we must consider them from a usage-based perspective, i.e. we must attach substantial importance to their instantiations in actual language use. Hjelmslev (1961: 11) speaks of the "empirical principle", which states that "a theory must be capable of yielding, in all its applications, results that agree with so-called (actual or presumed) empirical data". A grammar should be at once a grammar of the language system and a grammar of the uses which it is put to (Halliday 1994: xxii); it must work bottom-up rather than top-down (Langacker 1988: 132). Langacker's cognitive model of grammar is aimed at modelling the language system in a usage-based and psychologically plausible way (Langacker 1987a, 1988, 1990, 1991, 1999). It describes the interface between system and usage, and the insight into the organization of the language system gained from it in detail and with reference to the various psychological processes that play a role in it. Because Cognitive Grammar posits a fundamental equivalence between, on the one hand, the interface between system and usage, and, on the other hand, the organization of the language system itself, - both of which are argued to be modelled on the same, essentially cognitive principles - I feel that it goes a long way towards the formation of a usage-based theory of language. I will therefore start with a brief discussion of Langacker's analysis of the interaction between language use and the language system, and of the model which he proposes to account for the organization of the language system. A more detailed analysis of particular aspects of the language system will follow in later sections, together with a discussion of some basic, functional theoretical constructs which I consider to be compatible with Langacker's model and prerequisite for a truly usage-based account of language.

1.2. Cognitive Grammar: The interface between system and usage Langacker considers the following psychological abilities of the language user to be crucial for language use: symbolization, composition, comparison/ categorization and schematization. Symbolization builds on the general ability of association, whereby "one kind of experience is able to evoke another" (Langacker 1999: 94). Symbolization in language concerns the "relation between a structure in semantic space and one in phonological space" (Langacker 1987a: 77). When an established symbolic relationship becomes automatized and acquires 'unit' status, it forms a symbolic unit (e.g., [[OPENER]/[opener]],

12

Theoretical

assumptions

which associates the semantic unit [OPENER] with the phonological unit [opener]). Composition refers to the ability to integrate two or more (component) structures to form a composite structure (Langacker 1999: 94). Composition gives rise to relationships on the syntagmatic plane of language (Langacker 1987a: 75). An example of a composite structure is can opener, which combines the symbolic units [[CAN]/[can]] and [[OPENER]/[opener]], which in turn consists of [[OPEN]/[open]] and [[-ER]/[-er]]. Apart from symbolization and composition, the abilities of comparison and schematization are also relevant to language use. According to Langacker (1987a), the language user solves the problem of coding a detailed, context-dependent conceptualization in terms of a linguistic expression by making comparisons or categorizing judgements (Langacker 1987a: 6573): he/she assesses whether a specific expression can be categorized as a member of the category defined by a conventional linguistic unit or a unit that is widely shared by the language users of a speech community. A usage event is therefore always systemically motivated or "structured and evaluated with reference to the conventional units of a grammar" (Langacker 1987a: 426). Because structures which at first sight appear very different, may be "quite comparable in a coarse-grained view" (Langacker 1999: 93), categorization necessarily involves the ability of schematization: to determine whether a usage event satisfies the specifications of specific conventional units, one may have to abstract away from certain points of difference and portray the distinct structures with lesser precision and specificity (Langacker 1999: 93). A usage event can then either be identified as elaborating/instantiating a category, or it can be said to extend it. When a linguistic category is elaborated or instantiated by a usage event, the latter conforms to the specifications of the category but is, schematically speaking, "characterized in finer-grained detail" (Langacker 1999: 93). An elaborative or instantiating novel usage is conventional or - in more traditional terminology - grammatical·, it is conforming to the linguistic convention embodied by the linguistic unit, i.e. it is fully sanctioned by it. When, on the other hand, a discrepancy exists between the category that is defined by the linguistic unit and the usage event which is related to it, the category is extended and the usage event is either ill-formed or non-conventional. In general, a novel usage, irrespective of whether it merely elaborates or extends a conventional unit, can become conventionalized or acquire unit status via entrenchment (Langacker 1987a: 59). When the unit [[MOUSE]/ [mouse]], for instance, was first used with reference to a piece of computer

Language: The relation between system and usage 13 equipment, its meaning was extended. This extension has by now achieved unit status because of its frequent use and the entrenchment resulting from it (Langacker 1999: 108). Entrenchment or the automatization of certain structures is a matter of degree and the boundary between units and nonunits fluctuates continually because "every use of a structure reinforces it and entrenches it more deeply, whereas non-use has the opposite effect" (Langacker 1999: 100).

1.3. The language system as network Crucially, the relationships of symbolization and composition, comparison and schematization, which the language user establishes when construing language, are also claimed to determine the internal organization of the language system itself. The terminology used to describe language use consequently applies to the language system as well. The language system consists of three types of units: firstly, semantic (e.g., [PRINT]), phonological (e.g., [p]) and symbolic units (the latter associate a semantic with a phonological unit, e.g., [[PRINT]/[print]]); secondly, schematic symbolic units or grammatical patterns (e.g., [[V/...][[-ER]/[-er]]), and, thirdly, the categorizing relationships of elaboration/instantiation, extension and mutual similarity (Langacker 1988: 134, 147). These units have acquired unit status because they are entrenched: the language user can employ them "in largely automatic fashion" because they constitute 'pre-packaged assemblies' (Langacker 1987a: 57). They are conventional in that they are shared and recognized as being shared by a substantial number of language users (Langacker 1987a: 62). The simplest kind of symbolic unit is the morpheme, "in which a semantic and a phonological structure participate as unanalyzable wholes in a symbolic relationship" (Langacker 1987a: 58; see also Halliday 1961). The combination or (syntagmatic) integration of basic symbolic units leads to the formation of a composite symbolic structure or a grammatical construction. The term 'grammatical construction' applies to the component structures which the composite structure consists of, to the relationship of integration that exists between them and to the resulting composite structure (Langacker 1987a: 277). In Cognitive Grammar, a construction is situated on the syntagmatic plane of composition, as well as on a schematic plane, which is orthogonal to that of the syntagm and distinguishes between schemata and specific instantiations of schemata. The patterns for combining simpler symbolic units into larger ones are in this perspective analyzed as constructional

14 Theoretical assumptions schemata, which capture generalizations and serve as 'templates' for the assembly of novel expressions (Langacker 1988: 148): the schema [[V/...][[-ER]/[-er]] is thus instantiated by [[OPEN/open]-[-ER/-er]]. Because it is believed that a speaker's knowledge of a construction is not restricted to a single rule or schema standing in isolation, but also embraces the speaker's "knowledge of how the rule is 'implemented' with respect to more specific structures" (Langacker 1988: 151), the more specific instances of constructional schemata are argued to be part of the system of language as well. As Langacker puts it (1987a: 29), it is "gratuitous to assume that mastery of a rule like Ν + -s, and masteiy of forms like beads that accord with this rule, are mutually exclusive facets of a speaker's knowledge of his language; it is perfectly plausible that the two might sometimes coexist". In addition to being composite, a grammatical construction is therefore also complex: it does not reside in a single structure, but rather in "a family of structures connected by categorizing relationships" (Langacker 1988: 149). The 'family of structures' that makes up a complex category is best viewed as a schematic network or a network in which "linguistic structures of any kind and any size are linked in pairwise fashion by categorizing relationships" (Langacker 1999: 103): the superordinate nodes in the network are then schematic and the subordinate nodes are either subschemata or more specific structures which elaborate/instantiate or extend the schema. An example of a schematic network is given in Figure 1 (based on Langacker 1988: 131): SCHEMA THING X

DOG dog

PL -s

PL -s

TREE tree

PL -s

Figure 1. A schematic network In including not only the categorizing relationship of extension, which is focused on in prototype theory (e.g., Rosch 1975, 1978, cited in Langacker 1987a; Lakoff 1987), but also categorization based on schematization (i.e. elaboration or instantiation), the network model distinguishes itself from prototype theory. As Langacker (1988: 139) points out, it is necessary to in-

The relation between lexicogrammar and semantics 15 elude schematization among the categorizing relationships because a category's extension from the prototype tends to be "accompanied by a certain amount of upward growth, as schémas are extracted to generalize over a more diverse array of category members". The vision of the language system that emerges through Langacker's model "is one of massive networks in which structures with varying degrees of entrenchment, and representing different levels of abstraction, are linked together in relationships of categorization, composition, and symbolization" (Langacker 1999: 95). The description of a construction involves the elucidation of the relationships which the construction embodies as well as the relations which it contracts with other constructions in the network. In the following sections, I will zoom in on specific aspects of these relationships. I will start with the relationship of symbolization and briefly discuss the natural or non-arbitrary nature of the symbolic relation between lexicogrammar and semantics (Section 2). In Section 3, I will consider the syntagmatic relations that exist between the components of a construction and lay out the main properties of a radically functional view of composition and classification. Section 4 will go more deeply into the various types of functional relationships that are at work in linguistic patterning and it will discuss Halliday's tristratal functional interpretation of linguistic structure. Finally, in Section 5, I will look at the various types of paradigmatic relationships that obtain between specific constructions in the language system and I will point to their value as a descriptive 'heuristic'. I will argue that, apart from the categorizing relationships of elaboration/instantiation, extension and mutual similarity which Langacker discusses, the relation of agnation, introduced in Gleason (1965), should be included in the description of linguistic patterning as well.

2. The natural symbolic relation between lexicogrammar and semantics 2.1. A natural description of language The idea that a description of language, apart from being usage-based, should also be natural is prominently present both in Langacker's cognitive model of grammar and in Halliday's systemic-functional approach. In general, the notion of 'naturalness' refers to the concern to develop a description of language which "deals with data in their own terms, with full regard for the richness, subtlety, and complexity characteristic of linguistic phe-

16 Theoretical assumptions nomena" (Langacker 1987a: 13). A natural model of language is a model which respects the nature of language as a complex system. In Langacker (1987a), a natural analysis of language is, for instance, argued to accommodate not only discrete, but also non-discrete aspects of language structure; it is said to carry the simplifications and idealizations which are sometimes needed to begin with the analysis of language not too far, so as not to lose sight of the actual complexity of the subject matter; and, finally, a natural description is claimed to operate with substantive constructs, or constructs which are sufficiently clear to be elucidated "with a certain amount of precision and detail" and which are real in that they correspond to "some actual feature of linguistic organization" (Langacker 1987a: 22). One of the unnatural distinctions which circulate in linguistics is that between lexis and grammar, and, related to it, the assumption that linguistic theory needs the constructs of lexicon and syntax to describe it. Syntax is thought of as the "domain of generality and regularity, of productive rules with fully predictable outputs", and the lexicon is viewed as "the domain of irregularity, idiosyncrasy, and lists" (Langacker 1987a: 26). The distinction between lexicon and syntax has, for instance, been argued to underlie the differences between nominalizations such as John's refusing the offer and John's refusal of the offer (Chomsky 1970: 187): the latter, so-called 'derived' type of nominal has been argued to form part of the lexicon, mainly because the semantic relation which it has with the verb from which it is derived tends to be idiosyncratic. Nomináis of the former, i.e. gerundive, type, by contrast, are claimed to be part of the syntactic component of language, being derived from clauses by means of 'syntactic transformations' without fundamental changes to the clausal meaning (Lees 1960; Chomsky 1970): their meaning is argued to be related to that of the underlying clause in a regular way. Gerundive nomináis of the 'mixed' (Chomsky 1970: 215) or 'action' type (Lees 1960), such as John's refusing of the offer, however, make clear that the constructs of 'lexicon' and 'syntax' lack substance and cannot be strictly distinguished between. Like gerundives of the type John's refusing the offer, the meaning of action nominalizations seems regular, which suggests that they are 'syntactically' derived; and yet, action gerundives are not fully productive (e.g., *the being of a woman·, *his having of a car) and they have the internal structure of a nominal, two features which are alleged to classify them with lexically-based nomináis. The problematic use of the constructs of 'lexicon' and 'syntax' in the analysis of nominalizations thus illustrates that there is no reason to assume that the distinction between lexical and grammatical issues is a sharp one (as also argued in Langacker 1987a: 18, 26). In Halliday's work, lexis is for that reason referred to as

The relation between lexicogrammar and semantics 17 "most delicate grammar" (Halliday 1961: 267), its relation with grammar is argued to take the form of a cline or a continuum (an idea which is also found in Langacker's work) and the term lexicogrammar is used instead of grammar. Apart from the artificial distinction between lexis and grammar, the relationship that has most often been misrepresented is the fundamental relation between lexicogrammar and semantics: grammar and semantics have often been conceived as autonomous entities or separate 'components' of language (Langacker 1987a: 12; see also Halliday 1988). A description of language which is truly natural, however, can only be arrived at when the relation between grammar and semantics is recognized to be natural or non-arbitrary.

2.2. The natural or non-arbitrary relation between grammar and meaning Language is essentially symbolic in nature in that it consists of an openended set of linguistic symbols or signs, "each of which associates a semantic representation of some kind with a phonological representation" (Langacker 1987a: 11). In the case of most of the lexical items which are not further analyzable into morphemes, the symbolic association between meaning and form is arbitrary, i.e. there is no natural relation between the meaning of the lexical item and the particular phonological form it is mapped onto (Saussure 1972). Well-known exceptions to the arbitrariness of unanalyzable lexical items are instances of onomatopoeia, such as hiss and buzz. One of the basic tenets of both cognitive and functional models of language is that not only lexical items but also grammatical patterns are signs or symbolic combinations of meaning and form. As Langacker (1987a: 12) puts it, "morphological and syntactic structures themselves are inherently symbolic" and this "above and beyond the symbolic relations embodied in the lexical items they employ". A grammatical construction is thus symbolic not only in that it contains symbolic elements: the patterns along which these symbolic components are integrated is held to be symbolic as well. Grammatical patterning itself is believed to associate a particular meaning with a particular form.1 Most importantly, and unlike in most simple lexical items, the symbolic relationship established in grammatical patterns is held to be non-arbitrary or natural, with the grammar 'encoding' or 'realizing' the semantics (Halliday 1994: xvii). However, without further qualification, the Saussurean, bipolar interpretaton of the linguistic sign, which posits a link between meaning and pho-

18

Theoretical

assumptions

nological form, seems unfit to capture the non-arbitrary nature of grammatical structure: "A semiology in which meaning is said to be directly associated with phonology cannot be sensibly linked to a theory of grammar" (Davidse 1991: 1). As already hinted at in Saussure's work, it is necessary to posit a third or intermediate stratum, in between the semantic and phonological ones, to account for the non-arbitrary nature of grammatical structure. Halliday (1987, 1992) has used Hjelmslev's model of the linguistic sign in an attempt to clarify the relation between lexicogrammar and semantics.2 Hjelmslev (1961) takes over the Saussurean concepts of content and expression to talk about the linguistic sign and proposes that the content plane of the linguistic sign has to be further analyzed into content-form and content-substance. Content-form, which Hjelmslev describes as languagespecific encoding (Hjelmslev 1961: 52), represents according to Halliday (1987) the lexicogrammatical stratum of language. It is by passing through this lexicogrammatical stratum that general communicative 'purport' (Hjelmslev 1961: 51) or meaning which "can be 'translated' between languages" (Davidse 2000a: 33) is formed into content-substance, i.e. into sign-specific meanings, which can then be mapped onto a particular phonological form (Hjelmslev 1961: 52). The linguistic sign is, in short, not an expression that points to a content outside the sign. Rather, it is "an entity generated by the connexion between an expression and a content" (Hjelmslev 1961: 47), with the contentlevel further divided into a lexicogrammatical and a semantic plane. Crucially, it is only by virtue of content-form or lexicogrammar that contentsubstance or semantics exists: content-substance appears by "the form's being projected on to the purport, just as an open net casts its shadow down on an undivided surface" (Hjelmslev 1961: 57). Or, as Langacker (1987a: 98) puts it, semantic structure is conceptual structure that is "shaped for symbolic purposes according to the dictates of linguistic convention". The relationship between content-substance and content-form is therefore natural or non-arbitrary·, the meaning of a linguistic sign is encoded or realized via the lexicogrammar of the particular language it belongs to. Any attempt to construe semantics and lexicogrammar as discrete components of the language system is, consequently, at odds with the very nature of the linguistic sign. To establish the meaning of 'symbolically complex' units (Langacker 1987a: 82) - i.e. units which integrate two or more symbolic units into a grammatical construction - , then, we have to analyze their lexicogrammar in detail. Such an analysis mainly involves an elucidation of various 'contexts' (Langacker 1987a: 401). In what follows, I will concentrate on the

Composition, classification andfunction 19 syntagmatic context of the linguistic sign and the relations that are established through composition. In Section 5,1 will discuss the systemic context of linguistic signs or their position "within the schematic networks that collectively constitute the grammar of a language" (Langacker 1987a: 401).

3. Constructions: Composition, classification and function3 ... the components of a complex expression should not be thought of as providing the material used to construct it - theirfunction is rather to categorize and motivate facets of the composite structure .... (Langacker 1991: 508) Owing to the structural diversity of nomináis, a universally valid schematic characterization must be couched in terms of meaning and semantic function rather than formal properties such as constituency. (Langacker 1991: 54) In this section, a radically functional approach to the analysis of constructions is described, which views them as functional configurations, the components of which are essentially component functions and not necessarily component structures. A radically functional analysis of composition thus posits that it is functions rather than, for instance, constituents, which form the overarching motivating principle in composition. The main properties of a radically functional approach to composition will be sketched out as follows: in a first part (Section 3.1), I will focus on the role of functions in the analysis of composition and posit that the components of a construction are best viewed as component functions. In Section 3.2,1 argue in favour of integrating different levels of functional organization in the description of a construction, levels which together constitute a hierarchy or 'rank scale' (Halliday 1961) of functional configurations, ranging from 'lower' to 'higher' levels of functional organization. In Section 3.3, classes of symbolic units are argued to be best established on the basis of their specific functional behaviour, not only in their external functioning, but also - and this is more controversial - with respect to their own internal organization. Langacker's (1991) description of the four main functions of nomináis and clauses will be argued to have paved the way for such a 'radically functional' view of class.

20

Theoretical assumptions

3.1. A functional approach to composition In Langacker's (1987a) model of language, composition is said to involve components or component structures which are turned into a composite structure or a grammatical construction. The notion of 'component' captures the compositional relation by which components are integrated into a composite whole. Like the terms 'construction' and 'assembly', it hints at the active nature of the constructive efforts of the language user. 'Component' is preferred to the notion of 'constituent', which, as pointed out by McGregor (1997), is too much linked to one specific type of syntagmatic relationship, viz. that of constituency (i.e. "the syntagmatic relationship between a part and the whole to which it belongs", McGregor 1997: 21). The term 'constituent' moreover suggests that the components of a construction are building blocks out of which the composite structure is formed (Langacker 1987a, 1999). A 'building-block' view on composition, however, fails to account for those constructions in which "components are only partially discernible (or even indiscernible) within the composite whole" (Langacker 1999: 152). Even though composite structures like these are non-prototypical or 'marked', they form an essential part of the nature of language, and should therefore be included in an account of composition that aims to be natural. It is the description of constructions in which not all components are discernible that illustrates best the assets of a functional approach to composition. Components which are only partially discernible or indiscernible cannot be referred to as component structures. As Langacker's discussion of non-canonical nomináis makes clear, such components are best described in functional terms, i.e. as component functions which are not realized by specific component structures. A functional analysis of such noncanonically coded constructions not only sheds light on the constructions themselves, but it also succeeds in revealing what they have in common with other, canonically coded members of their class. Let me illustrate this briefly (a more elaborate account of the functional interpretation of nomináis given in Langacker 1991 follows later in Chapter 4). In a canonically coded nominal construction, Langacker argues, each semantic function that is characteristic of nomináis is associated with a distinct element of structure (Langacker 1991: 54). An instance of a canonical nominal is, for instance, the nominal those three black cats, in which grounding (or the link with the speech event) is effected by the demonstrative pronoun those, quantification by the numeral three and the type specification is provided by black cats.4 Non-canonically coded is a nominal like she. Even though it contains no visible structural assembly, it can be argued

Composition, classification and function 21 to realize functions that are similar to those encoded in those three black cats: it profiles a single instance of the schematically specified type 'human female', is inherently definite, and its designatimi is identified as being distinct from the speech-act participants (Langacker 1991: 148). It is thus the functional analysis of she which establishes it as a member of the class 'nomináis' and links it up with other nominal constructions. (From a strictly structural perspective, she seems to have hardly anything in common with nomináis like those three black cats.) The analysis of a construction thus essentially involves elucidating the functions which it expresses (e.g., Haas 1954; Halliday 1994; McGregor 1997) and constructions can be interpreted as 'configurations of functions' (Halliday 1994). In the following section, I will zoom in on the unique role which functions play in the description and classification of constructions.

3.2. Levels of functional analysis In Haas (1954), it is claimed that a functional analysis of composition is necessarily two-directional: functional relations are said to characterize a construction both analytically and synthetically. An analytic definition of a construction involves the analysis of the functions served by its components, whereas a synthetic definition refers to the functions which the composite whole itself plays in a larger configuration (Haas 1954: 61). A construction can, in this perspective, be described in functional terms externally (synthetically) as well as internally (analytically). Apart from the external and internal functions of a construction, the construction itself should be considered as well, as the level at which the various external and internal functions are integrated.5 In this way, the construction itself constitutes a third level of linguistic analysis. It is not "algorithmically deducible" from its components, but must be treated in a nonreductive way, "as a separate entity in its own right" (Langacker 1987a: 87). The levels of analysis that are needed for the description of constructions reflect a kind of hierarchy among the symbolic units of a language: symbolic units seem to be situated among layers of 'lower' and 'higher' levels of organization, whereby the lower levels define the construction analytically or internally, and the higher levels situate the construction externally or synthetically. To account for this hierarchy, Halliday (1961) introduced the construct of the rank scale. In his original interpretation of the notion of 'rank' - which is very much based on a constituency-oriented model of language - a rank grammar is said to be "one which specifies and

22

Theoretical assumptions

labels a fixed number of layers in the hierarchy of constituents, such that any constituent... can be assigned to one or other of the specified layers, or ranks" (Halliday 1966b: 111). The various levels of organization that can thus be distinguished correspond to specific, determinate stretches of structure, characterized by a distinction in 'size', i.e. the sentence, clause, phrase, word and morpheme (Halliday 1961). These levels are claimed to be related in such a way that, "going from top (largest) to bottom (smallest), each 'consists of one, or more than one, of the unit next below" (Halliday 1961: 251).6 The analysis of a construction, then, requires that it be accounted for at all ranks. This is called the 'requirement of total accountability' in Halliday (1966b: 113). A clause, for instance, is analyzed into phrases, which themselves consist of words, which, in turn, are made up of morphemes. The rank scale allows for downward 'rank shift', whereby a symbolic unit is used at a rank which is either equal to or lower than its own rank. The use of a clause as nominal, as in [That Caesar was dead] was obvious to all (Halliday 1994: 264), for instance, is considered to be a case of downward rank shift. The limitations of this "strong version" of the rank hypothesis (McGregor 1991: 121) and the need for refining and restricting it have been repeatedly pointed out (e.g., Matthews 1966; Huddleston 1988; McGregor 1991). Firstly, it has to be acknowledged that certain items, such as conjunctions (e.g., and, but), interjections (e.g., hey, eh) and adjuncts (e.g., surely, probably, yet) cannot be included in the rank scale (Matthews 1966; McGregor 1991). Secondly, the requirement of total accountability at all ranks is arguably more illuminating in some cases than in others. As Matthews (1966: 103) points out, an utterance like Yes! consists of a morpheme assigned to sentence level, and analyzing it as 'one sentence, which is one clause which is one phrase which is one word which is one morpheme' does not seem to add anything to its analysis. Many criticisms are, moreover, provoked by the 'structuralist' formulation of the rank scale in Halliday (1961), where the rank scale was presented as being a rank scale of (canonically coded) structures. Nonetheless, the idea that constructions occupy a particular position on a rank scale of symbolic units is valuable to the description of constructions, provided it is interpreted functionally and it is not detached from Halliday's overall functional approach to linguistic organization. Halliday himself has stressed that ranks are basically a unit's "most generalized functional environment" (Halliday 1966a: 66; [italics mine]) and that the identification of the rank of a unit "is a possible first step in the specification of what Haas (1966: 125) calls 'functional relations'" (Halliday 1966a: 65). Moreover, Halliday (1994: 22) links the constituency analysis implied in

Composition, classification andfunction 23 the rank scale to so-called minimal or functional bracketing, that is "bracketing together only those sequences that have some function relative to a larger unit". (Functional bracketing differs from 'immediate' constituent analysis, which brackets together units that may not have any function at all and which reveals "the order in which all the pieces are put together, pair by pair", Halliday 1994: 23.) From a functional perspective, the rank scale can thus be viewed as a hierarchy of basically functional configurations and the analysis of a construction can be said to involve a description of its functioning in higherlevel configurations, a description of the lower-level functions which it embodies and the integration of these two into a coherent analysis of the rank that lies in between, i.e. that of the construction itself.

3.3. A functional approach to classification Symbolic units are typically assigned to particular classes because of their (external) behaviour in larger configurations and, in some cases, because of basic semantic features which they have. In functionally-oriented approaches to language, it is usually pointed out that classes of symbolic units are best derived on the basis of the larger structural configurations in which they can occur, i.e. on the basis of how they function in them (e.g., Halliday 1961; Quirk et al. 1985; McGregor 1997). The class of the noun phrase or nominal, for instance, is thus characterized as a class of symbolic units which "typically functions as subject, object, and complement of clauses and as complement of prepositional phrases" (Quirk et al. 1985: 245). In Croft (2000: 16), syntactic categories, including classes, are argued to be "construction-specific" or derivative only "from the constructions that define them". Apart from classification on the basis of external functioning, one also finds - be it much less frequently - notional criteria to describe a number of basic classes such as the noun and the verb, and certain of their subclasses, such as count and uncount or mass nouns. Probably the most elaborate notional characterization of nouns and verbs is the conceptual or 'intrinsic' characterization given in Langacker (1987a, 1987b, 1991). It is necessarily abstract or schematic because of the extremely heterogeneous nature of nouns and verbs (Langacker 1987b: 58; 1991: 15). According to Langacker's intrinsic characterization, a noun profiles a thing or "a region in some domain, where a region is defined abstractly as a set of interconnected entities" (Langacker 1991: 15). Verbs, on the other hand, are basically relational and profile 'interconnections' (Langacker 1991: 19). Within

24

Theoretical assumptions

the noun class, then, the distinction between the subclasses of count and mass nouns mainly depends on "whether the profiled region is construed as being bounded within the scope of predication" (Langacker 1991: 18): count nouns (e.g., lake) include the boundaries of what they designate in their scope of predication, whereas mass nouns (such as water) do not. The external, grammatical functioning of a symbolic unit and the highly schematic semantic characterizations of some basic classes have weighed heavily on discussions about grammatical classes. Interestingly, internal properties have not: internally, the various instantiations of a particular class are typically considered to be too diverse to be generalized across. Nomináis, which vary from having a common noun as head, to being a proper name or a pronoun, seem to be no exception to this. The internal structural diversity that is characteristic of many classes has, in fact, explicitly been said to render classification on the basis of internal properties impossible. Halliday (1961: 261) thus states that a class is "not a grouping of members of a given unit which are alike in their own structure": he (1961: 261) holds that, in terms of the rank scale, classes are derived 'from above' and not 'from below'. In what follows, an alternative approach to grammatical classes will be presented which does build in the perspective from below. In a first part, I will argue that a radically functional approach to classification includes the internal functional organization of symbolic units, irrespective of whether the internal functions are symbolized separately or not. A functional approach to grammatical class which includes the internal functions in its analysis manages to identify symbolic units which are structurally widely divergent as members of one and the same class. In a second part, I will argue that classes which are functionally derived allow for a finer-grained description of what can be said to constitute a prototypical instantiation of a class.

3.3.1. An internal functional analysis of constructions A radically functional approach to classification identifies classes not only in terms of their behaviour in larger functional configurations, but also by the presence of a particular set of component functions. More particularly, it views classes as combinations of certain external and internal functions. The inclusion of internal functional properties in the classification of a symbolic unit is in accordance with Haas's claim that "many constructions can be adequately defined only by a combination of analytic and synthetic criteria ..., and a classification of them will require a similar combination"

Composition, classification andfunction

25

(Haas 1954: 70). Until recently, it was difficult to see how such an analysis could be envisaged of the nominal, in which, in view of the topic of this study, we are particularly interested. While generalizations that deal with the nominal's external functioning are fairly readily arrived at (as, for instance, in Quirk et al. 1985), internal functional generalizations are not. This is not only because of the apparent disparity of the units that can function in the nominal slots of a construction, but also, and more fundamentally, because the question of the internal functioning of the nominal has tended to be approached from an exclusively structure-based perspective. It is only Langacker's (1991) analysis of the internal functional organization of the nominal that makes clear that the various instantiations of the class can be generalized across also internally. Langacker's analysis is pathfinding, both descriptively and theoretically, and adopts a view of the nominal which is very much 'radically functional' in that it recognizes that "semantic function (rather than constituency) is the critical factor for understanding their internal organization" (Langacker 1991: 51). The semantic functions which, according to Langacker, are characteristically realized by nomináis are those of type specification, instantiation, quantification and grounding (Langacker 1991). The "universally valid schematic characterization" which Langacker proposes for the nominal is that it profiles "a thing construed as an instance of some type and further incorporates some specification of quantity and grounding" (Langacker 1991: 54). Simple nouns provide nothing more than a type specification: they specify "the basis for identifying various entities as being representatives of the same class", but are "not tied to any particular instance of that class" (Langacker 1991: 53). Full nomináis such as the site, an excellent site and two convention sites in the Midwest, on the other hand, presuppose instantiation of the type in question and designate one or more instances. In full nomináis, "information is furnished concerning both the number of instances and their status vis-à-vis the speech-act participants" (Langacker 1991: 53). These semantic functions, Langacker (1991: 54) argues, are realized in all nomináis, be it in many different ways. If they are coded separately in the nominal's structure, then the head noun, together with its adjectives and other modifiers that render it more precise, provides the type specification (e.g., excellent convention site in the Midwest); the quantifier is added as a separate layer (e.g., three black cats), and a grounding element is appended at the outermost layer (e.g., those three black cats). Many nomináis, however, fail to display this type of assembly, in which case "the semantic functions in question are not uniquely associated with distinct levels of constituency" (Langacker 1991: 54). Apart from pronominal nomináis

26 Theoretical assumptions (such as she, which was discussed earlier), proper names probably represent one of the most extreme cases of departure from the iconic type of coding: in proper names, Langacker (1991: 59) proposes, "type, instantiation, quantity, and grounding are conflated in a single expression". In the proper name Stan, for instance, a type specification is incorporated (that of 'male human'), and because it characterizes a specific person, it presupposes instantiation and quantification. Grounding is subsumed as well, "for the nominal is definite and portrays the profiled individual as being uniquely apparent to the speaker and hearer on the basis of this name alone" (Langacker 1991: 59). Because these functions are fulfilled, the proper name qualifies as a nominal, and this in spite of its non-canonical coding (Langacker 1991: 59). As Langacker (1991: 53) concludes, the functions of type specification, instantiation, quantification and grounding thus allow us to "profitably examine the organization of nomináis from the standpoint of semantic function, abstracting away from any details of structural implementation". Classification by internal functional properties should be distinguished from intrinsic or notional classification. Notional classification makes reference to abstract cognitive events and, to make sure that all members of the class (including the less prototypical ones) instantiate the class schema, it is necessarily highly schematic (Langacker 1987b: 54).7 As Langacker (1991: 52) himself points out, the characterization 'thing or region in some domain' applies to nominal predications "as a broadly defined class": it is too abstract to distinguish simple nouns like site from full nomináis like an excellent convention site, "each of which profiles a thing and qualifies as a noun in this inclusive sense of the term" (Langacker 1991: 52). The intrinsic classification of nomináis is especially meant to distinguish them from relational predications. The functional description in terms of type specification, instantiation, quantification and grounding, in contrast, applies to the internal organization of nomináis and helps to identify the differences between nominal constructions and simple nouns (Langacker 1991: 53). Classification in general, then, is intrinsic as well as functional, and the functional criteria involve both the external functions served by the symbolic unit and the internal functions of its components.

3.3.2. Classes as complex categories: Schema and prototype Langacker's functional description of nomináis shows that it is feasible to include the internal functional organization of a construction into its classification: the functional characterization of nomináis as 'grounded instances

Composition, classification andfunction 27 of some type', for instance, is applicable to all the members of the class 'nominal'. Put differently, all nomináis instantiate the same schema, which "captures the pertinent generalization" (Langacker 1988: 130). All nomináis are fully compatible with its specifications, but characterize them in finer detail or "elaborate the schema in different ways along various parameters, to yield more precisely articulated notions" (Langacker 1987a: 68). By combining a schema with various instantiating structures, classes form complex categories, or categories residing "in a family of structures connected by categorizing relationships" (Langacker 1988: 149). Among the categorizing relationships in a complex category, however, Langacker distinguishes not only categorization by schema, but also categorization by prototype or extension. He (1987a: 371) views the two modes of categorization as "inherently related and describable as aspects of a unified phenomenon", and his network model is a 'synthesis' of categorization by prototypes and by schémas. Categorization by prototype offers a perspective on classification which is not so much alternative as complementary to that of categorization by schema. Both categorization by schema and categorization by prototype can be said to involve comparing a usage event or target structure with a conventionalized unit or standard. Instantiation implies that all the specifications of the standard are satisfied by the usage event. The standard is then viewed as a schema which only differs from its instantiations in degree of schematicity or 'granularity'. This is true of, for instance, the schematic characterization of nomináis as 'grounded instances of some type'. Extension of the standard, on the other hand, implies that the target structure is only partially sanctioned by the standard or deviates from it. The standard is then conceived as the prototype of the category. It follows that a 'natural' approach to classification should not only look for schematic generalizations, but should also take account of the occurrence of prototypical and less prototypical instances of a class (Langacker 1987a: 371). Instead of considering all members of a class as being fully compatible with a schema, categorization by prototype judges class membership through a "perception of similarity" (Langacker 1987a: 69), which is based on what is conceived to be a prototypical instance of the class. Prototypicality thus results in "degrees of membership based on degrees of similarity" (Langacker 1987a: 371; [italics mine]). The criteria used to settle on the prototype of a particular class - and with them, the prototypes themselves - are rather divergent. Semantic/intrinsic arguments can be distinguished from arguments that concern the structural implementation of specific internal functions.

28 Theoretical assumptions With regard to the general classes of language, Langacker argues that they are "centered on a prototype that reflects a basic aspect of our everyday experience" (Langacker 1991: 521). The archetypal conception of 'physical object' is thus generally held to constitute the semantic prototype for nouns. In his functional analysis of nomináis, then, Langacker takes the structural way of coding as starting-point: prototypical are held to be nomináis "whose structure most directly mirrors semantic function" (Langacker 1991: 143), i.e. nomináis which reflect their component functions through distinct and overt levels of constituency.8 It follows that, in Langacker's view (as opposed to, for instance, that of Van Langendonck 1999), prototypical nomináis incorporate both a head noun and a determiner (Langacker 1991: 143). Examples of prototypical nomináis are this cup, some milk, a big house, those three black cats. Non-prototypical are nomináis such as three from Toledo, the poor, for Harold to resign·, they occur less frequently or "only in special grammatical circumstances" (Langacker 1991: 143). Interestingly, by linking prototypicality to iconicity, Langacker's analysis of the internal organization of nomináis also allows one to distinguish degrees of non-prototypicality: some nomináis, for instance, deviate from the prototype in that the semantic functions which they realize are associated with the nominal as a whole (e.g., personal pronouns, proper names, ίΑαί-nominalizations; Langacker 1991: 148), others consist of two nomináis, related by means of apposition (e.g., the fact that whales are mammals·, Langacker 1991: 149); still others lack a head noun because the grounding predication itself functions as nominal (e.g., that, these, some, any).

3.4. Composition and classification: Conclusion In short, a radically functional approach to composition gives priority to functions rather than to constituents and posits that an accurate description and classification of symbolic units has to be based on a functional analysis which includes the external and the internal functional organization of the unit in question. These functional levels are situated on a rank scale of functional configurations. Yet another characteristic of a functional analysis of linguistic patterning is that it considers the components of a composite whole to serve several functions at the same time. This kind of 'multifunctional' view on language will constitute the focus of the next section.

Functional layers of organization

29

4. Functional layers of organization The key to a functional interpretation of grammatical structure is the principle that, in general, linguistic items are multifunctional. (Halliday 1994: 30) Many of the functional theories of language are 'multifunctional' and posit that constructions are characterized by different layers of functional organization (e.g., the Prague School; the functional schools of Dik and of Givón; the systemic-functional school of Halliday). The multifunctional approach to linguistic organization which I present here is that of Halliday's systemic-functional model (with McGregor 1997 as most interesting recent development in it), and this because of its emphasis on the semiotic principle or the correlation between symbolization and semantic value for each layer of organization. The systemic-functional interpretation of the multifunctional hypothesis can be illustrated by means of the categories of Subject, Theme and Agent in English (based on Halliday 1970b, 1988, 1994). If we take an English clause like (1)

The duke gave my aunt this teapot.

the element considered as Subject is the duke. Its role in the clause can be characterized in various ways: the duke can be analyzed as the Agent of the process with which it is construed (i.e. gave); it can be said to be "the concern of the message" (Halliday 1994: 30), and it is that of which something is predicated. Not all English Subjects, however, realize these three features. Consider the following clause: (2)

This teapot my aunt was given by the duke.

The Subject here is my aunt, but the Agent-role is realized in the fry-phrase (i.e. by the duke) and the message concerns this teapot. The latter two features - the role which a unit plays in the process designated by the main verb and being the concern of the message - appear to be features which can, but need not be, associated with the Subject: they can also be realized by other components of the clause.9 What at first sight appear to be characteristic properties of a single unit thus turn out to be basically functions on the clausal level. Rather than being properties or aspects of a specific unit (in this case, the Subject-unit), they are distinct functions which are either mapped onto each other in a

30

Theoretical

assumptions

single component (as in 1) or are realized by distinct components (as in 2). In the case of a unit such as the duke in (1), Halliday has labelled its various functions as Theme, Subject and Actor: as a Theme, the duke is the "point of departure" of the message conveyed by the clause (Halliday 1994: 34); the duke is also Actor or the "doer of the action" (Halliday 1994: 30); and, finally, in its function of Subject, the duke is the one "on which rests the truth of the argument" (Halliday 1994: 30). A unit such as the duke in (2) is thus multifunctional, i.e. it serves several functions in the clause at a time. To fully appreciate Halliday's multifunctional view on linguistic units, three aspects of it have to be further looked into, viz. the so-called metafunctional origin of the functions which it involves, its conflgurational nature, and, thirdly, the significance that is attributed to the paradigmatic or systemic choices that lie behind each metafunctional configuration. The paradigmatic or systemic aspect of constructions will constitute the focus of Section 5. The metafunctional and conflgurational nature of Halliday's multifunctional approach to linguistic organization will be elaborated now. Let us start with the notion of 'metafunction': what characterizes a metafunction and how does this relate to the «íw/rifunctional view on linguistic units? Language users are believed to select from among a large number of interrelated options which together constitute as it were the 'meaning potential' of language (Halliday 1970b: 142). It is one of the most fundamental claims of the systemic-functional approach to language that the options which are embodied in this meaning potential and which are reflected in the grammatical options of a language "combine into a very few relatively independent 'networks'" (Halliday 1970b: 142). Within the massive network of options that constitutes the grammar of a language, one can thus distinguish certain clusters or sub-networks in which the "selections made by the speaker at one point tend to determine, and be determined by, the selections he makes at another" (Halliday 1979: 61). These clusters of lexicogrammatical options have been interpreted by Halliday as very general functions of language and they have therefore been called 'metafunctions' (Halliday 1994). Halliday distinguishes the ideational, the interpersonal and the textual metafunction. The ideational or representational metafunction covers all those options in language which enable language users to express their "experience of the real world, including the inner world of ... [their] own consciousness" (Halliday 1970b: 143). These options either represent experience directly in terms of processes, participants and circumstances (i.e. experientially), or they represent experience indirectly, in terms of the fundamental logical relations in language (i.e. logically) (Halliday 1979: 59). hi the clause, for instance, the grammatical system of transitivity, which construes the world

Functional layers of organization 31 into a set of process types and participants, is basically experiential in function. The function of Actor served by the duke in example (1) is an experiential function. Halliday argues that the logical metafunction determines the relations that exist between the various clauses within a clause complex (e.g., between the clauses in I would, if I could, but I can't; Halliday 1994: 218); but it is also at work in the nominal, where it imposes a head-modifier structure (e.g., those two splendid old electric trains, with trains as head; Halliday 1994: 191). The interpersonal metafunction concerns the communicative function of language or language as exchange: it incorporates the grammatical options which a language user has to adopt a particular speech role and assign one to others (e.g., making assertions, asking questions, giving orders). The interpersonal metafunction also covers the linguistic options which the language user has for relating what he says to the moment of encoding/decoding, i.e. for grounding it in the speech event. The principal grammatical system for the interpersonal metafunction of language is that of mood. The function of Subject is identified by Halliday as an interpersonal function (Halliday 1970b, 1994; for a more elaborate discussion of the Subject, see Chapter 4). The ideational and the interpersonal metafiinctions together constitute the two main kinds of meaning in language. The third, textual metafunction "breathes relevance into the other two" (Halliday 1994: xiii): it covers the options found in language "for making links with itself' (Halliday 1970b: 143), i.e. for creating texts. The clause, for instance, is organized as a message "by having a special status assigned to one part of it", called the Theme. The latter then combines with the rest of the clause into a message (Halliday 1994: 37). In the metafunctional hypothesis lies the intrinsic character of Halliday's systemic-functional approach to language: the metafunctions link up the two general purposes of language in life (the ideational and the 'active' or interpersonal; Halliday 1994: xiii) with particular options that can be chosen by the language user. Importantly, each of the metafunctions links up with particular lexicogrammatical realizations: they represent formmeaning, or better, form-function units (McGregor 1997) and cannot therefore be characterized as either 'semantic' or 'syntactic' functions (as are, for instance, the functions in Dik 1980).10 A second main characteristic of Halliday's multifunctional hypothesis is its configurational focus, which follows from the fact that the meanings of particular functions necessarily derive from their relationship to the other functions with which they are structurally associated: the function of Actor, for instance, is interpretable only in its relation to other functions of the same kind, i.e. other 'representational' functions such as Process and Goal

32

Theoretical assumptions

(Halliday 1994: 35). It is "the structure as a whole, the total configuration of functions, that construes, or realizes, the meaning" (Halliday 1994: 34). Each of the three metafunctions is therefore held to be reflected in a structure or configuration of functions (Halliday 1994: 30). The strands of meaning which the three metafunctions form thus constitute three different structural layers. Importantly, because not every element in a construction discharges a function in every metafunction or layer of structure, constructions cannot always be analyzed exhaustively in terms of all of the metafunctions (Halliday 1994: 35). Some elements thus have a purely textual function (e.g., conjunctions), while others serve a primarily interpersonal function (e.g., modal Adjuncts such as probably, certainly, actually). It is at the level of the clause that these metafunctional layers of organization are most clearly in evidence (Halliday 1979: 77). An example of a three-layered metafunctional analysis of a clause is given in Table 1: the first strand of analysis (consisting of the functions of Theme and Rheme) is that of the textual metafunction; the second layer is the interpersonal one (Complement-Subject-Finite-Predicator-Adjunct), and the third layer represents the experiential functions realized by the clause (Goal-RecipientProcess-Actor). Table 1. A multifunctional analysis of the clause This teapot Theme Complement Goal

my aunt Rheme Subject Recipient

was

given

Finite Predicator Process: material

by the duke Adjunct Actor

It has repeatedly been pointed out by Halliday that this type of 'segmental' analysis of the clause does not "tell the whole story" (Halliday 1994: 36). For one thing, it captures the experiential metafunction more accurately than it does the interpersonal and the textual metafunctions. In his radically semiotic model of language, McGregor (1997) therefore argues that syntagmatic relationships other than that of constituency underlie the interpersonal and the textual metafunction. He proposes to view the interpersonal metafunction, for instance, as a grammatical sign of which the formal aspect is formed by the syntagmatic relationship of 'conjugation'.11 Conjugational relations are basically whole-whole relationships, which either involve what McGregor (1997: 210) calls 'scoping' ("in which a unit applies over a certain domain, leaving its mark on the entirety of this domain") or 'framing' ("in which a unit delineates the domain over which it

Constructions and the relations between them 33 applies, marking it off from everything else"). Μ the clause, the declarative and interrogative moods are thus claimed to have 'scope' over the full clause. An example of framing is formed by inter-clausal relationships whereby one of the clauses expresses 'represented speech': like "a frame around a picture, the representing clause delineates the represented clause from its context... indicating that it is to be viewed or evaluated in a different way" (McGregor 1997: 253).

5. Constructions and the relations between them [The] Language system involves regularly recurring patterns in sentence structures. ... Such recurrent patterns within sentences, however, do not exhaust the system of a language. There is an additional set of relationships that grammar must describe, and this is at least as large and diverse. These exist between pairs of sentences.... (Gleason 1965: 195) Ultimately, the grammarian takes account in his statement of those relations which seem to fit into an interconnecting system ramifying far and wide through the language. (Gleason 1965: 202) It is a basic tenet of systemic-functional linguistics that functions cannot only be defined by reference to the (syntagmatic) configuration of functions which they belong to, but that they are primarily options or choices from a paradigmatic system: a function in the system is held to be defined by the total set of features from which it has been selected (e.g., 'past' is defined by reference to 'present' and 'future'), just like a structural function is defined by the total structural configuration it occurs in (e.g., 'modifier' by reference to 'head') (Halliday 1970b: 61). The metafunctional layers of organization in the syntagm are, in other words, based on paradigmatic systems which, like them, are metafunctionally motivated. A systemic or paradigmatically oriented description - i.e. a description of the systems that lie behind particular functions in a syntagmatic configuration - is therefore considered to necessarily complement a syntagmatic or structural description. Hence the term systemic-functional. In systemic-functional grammar, a 'paradigmatic' description takes the form of a system network, which maps out the various options of the language user in hierarchically ordered systems. For each metafunctional layer of organization in a construction, system networks of interrelated choices can be set up. Any option in the system can then be linked to particular wordings or instances of language use. An example of a system network

34

Theoretical assumptions

which accounts for part of the interpersonal choices made in the English clause would be the one in Figure 2 (based on Halliday 1976: 93): it gives an overview of the basic mood system of the clause and the various options in it, each of which stands for a list of constructions. The option 'interrogative, yes-no question', for instance, represents clauses such as Will you close the door? and Did she climb that mountain?. yes/no

• interrogative

[ C

WH-

• indicative

assertion

• declarative

exclamation

clause

• imperative

C

jussive optative

Figure 2. An example of a system network (Halliday 1976: 93) In what follows, I will elaborate on the systemic or paradigmatic side of constructions and, following Gleason (1965), I will point to the theoretical and methodological significance of including into the analysis of particular construction types the relations that exist between the constructions of the language system. In terms of Halliday's system networks: the relations that exist among the various options (i.e. constructions) of a system network are highly relevant to the analysis of any construction that belongs to it. Or, in terms of Langacker's network model of language: not only the relation between a construction and the schema which it instantiates/elaborates or that between a construction and the prototype which it extends should be analyzed, but also the relation between the various instantiating constructions themselves, as well as that between constructions which (seemingly) instantiate distinct schemata. I will first zoom in on Gleason's notions of enation and agnation and discuss the inter-constructional relationships which they represent (Section 5.1). In a second part, the theoretical significance of agnation and enation, and their heuristic value will be pointed out and illustrated (Section 5.2).

Constructions and the relations between them

35

5.1. Agnation and enation: A definition What do the notions of 'agnation' and 'enation' stand for? Gleason (1965) proposes the terms 'agnation' and 'enation' to describe two basic kinds of relationships which he distinguishes between constructions. The terms are derived from Latin enatus 'related on the mother's side' and agnatus 'related on the father's side' and they are deemed appropriate to designate "two contrasting types of relations, neither of which can exist without the other" (Gleason 1965: 199fn2). The relationship of enation links up constructions with "identical structures", that is, "the elements ... at equivalent places ... are of the same classes, and ... the constructions in which they occur are the same" (Gleason 1965: 199). The relationship between the clauses in (3a) and (3b) is one of enation: (3)

a. The dog bit the man. b. The cat ate the canary.

Agnation is a relationship between two grammatical structures which have the same major lexical items, but are different in structure, as in (4a) and (4b). They are agnate because their "relation in structure is regular and systematic", i.e. "it can be stated in terms of general rules" (Gleason 1965: 202). (4)

a. The dog bit the man. b. The man was bitten by the dog.

Importantly, for a relation to be identifiable as agnation, it must involve large numbers of constructions. Rather than being a "unique relation between isolated pairs" of constructions (Gleason 1965: 202), agnation requires that groups of constructions are found which are agnate or systematically related to each other in a similar way. Agnation "is based on the pervading patterns of the language", and it is, therefore, "always a recurrent thing, involving large numbers of sentences" (Gleason 1965: 202). As the examples in (5) and (6) make clear, the identification of groups of agnates (linked up by means of ) always presupposes the identification of enate constructions (indicated by I I I ) (based on Gleason 1965: 202):

36

Theoretical

(5)

He saw it.

It was seen by him.

He heard it.

It was heard by him.

He felt it.

«

It was felt by him.

(6)

assumptions

He galloped the horse. Ill He marched the soldiers.

He walked the dog.

Ό

The horse galloped. Ill The soldiers marched. Ill The dog walked.

It is only because the clauses He heard it and He felt it, which are enates of He saw it display a relation of agnation with It was heard by him and It was felt by him, which are enates of It was seen by him, that the relationship between He saw it and It was seen by him in (5) can be analyzed as being one of agnation. Likewise, relations cannot be identified as enate if they do not have identical sets of agnates. Constructions such as The man saw a stranger and The man seemed a stranger, for instance, are not enate because they belong to different agnation networks: A stranger was seen by the man is possible, while A stranger was seemed by the man is not. The man seemed to be a stranger, on the other hand, is acceptable, while The man saw to be a stranger is not (Gleason 1965: 203). The identification of agnate structures is, in other words, an ideal heuristic tool for disambiguating constructions which are seemingly identical (Davidse 1998a: 283). Because it is possible that constructions share some agnates, but behave differently with respect to others, Gleason also recognizes partial enation or enation between structures which are only partially structurally identical (which is elaborately illustrated in Laffut 2000). As pointed out by Davidse (1998a: 283), it seems consistent with Gleason's way of thinking to speak of a relationship of non-enation when constructions are structurally non-identical. The relationship between The man saw a stranger and The man seemed a stranger can then be labelled as being one of 'non-enation'. In short, relationships of agnation and enation cannot be stated clearly without one another: they are mutually defining notions that make up a two-dimensional set to analyze constructions (Gleason 1965: 201). Most importantly, they show that "formal evidence is not restricted to 'observable characteristics' of the syntagm in question, but ... also includes the syntagm's systematic relation to its agnates" (Davidse 1998a: 284). In more

Constructions and the relations between them mainstream terminology, one would say that formal evidence includes the syntagm's behaviour with regard to syntactic tests (Davidse 1998a: 284).

5.2. Agnation and enation: Theoretical significance and heuristic value Gleason's claim that agnate and enate relationships form an integral part of the analysis of a construction is theoretically significant in that it assigns a central role to the paradigmatic or systemic aspect of language use. As pointed out by Davidse (1998a), it has had a great impact on Halliday's thinking, and even though Halliday has never theorized much about agnation (Davidse 1998a: 287), his identification of system networks and, linked to it, of the various metafimctions, is based on inter-constructional relations. For one thing, the categories in his system networks essentially represent the proportionality of relations between constructions. As Halliday puts it, a category's label is "no more than the name of a proportional relation" between constructions (Halliday 1994: xxxii). Proportional relations and, consequently, linguistic categories are defined by agnation and enation (Davidse 1998a: 287): more precisely, they are established as well as semantically interpreted by means of the twodimensional set of agnation/enation. To give an example, the categories in the system network in Figure 2 represent constructions of the following types: (7)

a. You opened the door, (declarative: assertion) b. Who opened the door? (interrogative: WH-question) c. Did you open the door? (interrogative: yes-no question) d. Open the door! (imperative: jussive)

Between constructions such as those in (7a) and (7c), situated on the horizontal level of (8), there exists a proportional relation of agnation, which is confirmed by the enate constructions on the vertical axis: (8)

You opened the door. She climbed the mountain. He washed the car.

::

Did you open the door? Did she climb the mountain? Did he wash the car?

The proportionality which these constructions show (indicated by '::') centres on their Subject-Finite unit: more particularly, in the paradigm in the left-hand column, the Subject precedes the finite verb, while in the para-

37

38

Theoretical

assumptions

digm on the right-hand side, it is the Finite that precedes the Subject. It is this proportionality which is captured by the categories labelled as 'declarative' and 'interrogative'. Importantly, these categories also derive their 'meaning' from it: linked to the declarative order of the mood element is the interpersonal speech function of 'giving information', while the interrogative involves the demand of information (Halliday 1994). On a inore general level, the Subject-Finite unit can thus be said to serve a central role in what is called the interpersonal function of language. Particularly interesting, then, is that the inter-constructional relations that underlie the categories in Halliday's system networks differ depending on the metafunctional layer of language which they represent (Davidse 1998a): while interpersonal networks such as the one in Figure 2 map out agnate structures (as shown in 7) and consist of categories "dealing with clausal variants that are, by and large, applicable to any one clausal syntagm" (Davidse 1998a: 290), agnation has a different use in system networks of the ideational domain. There, agnation is mainly used to distinguish non-enate structures: the categories of 'intransitive', 'transitive' and 'ditransitive' clauses, for instance, do not label proportionalities of agnate structures, but of non-enate constructions (Davidse 1998a). Consider the following ideational network (based on Halliday 1968): s - intransitive (Actor - process) e.g., John is running clause goal-directed (Actor - process - Goal) e.g., John hit the ball transitive descriptive (Initiator - process - Actor) e.g., John marched the prisoners Figure 3.

An example of an ideational system network

The non-enate status of the constructions John hit the ball and John marched the prisoners, and, consequently, of the categories 'goal-directed' and 'descriptive', is revealed through the different sets of agnates which they can be related to (Davidse 1998a: 293). The construction John marched the prisoners is part of the paradigm illustrated in (9): (9)

a. John marched the prisoners. b. John made the prisoners march. c. The prisoners marched.

Constructions and the relations between them

39

d. What the prisoners did was march. e. What John made the prisoners do was march. As shown in (10), the construction John hit the ball does not belong to that same paradigm: (10) a. John hit the ball. b. *John made the ball hit. c. *The ball hit. d. *What the ball did was hit. e. *What John did to the ball was hit. The categories in the ideational network in Figure 3, in short, represent non-enate constructions, identified on the basis of the distinct agnation paradigms 'behind' them. Importantly, these agnation paradigms also serve "as a heuristic to interpret the semantic difference" between John hit the ball and John marched the prisoners (Davidse 1998a: 293): the agnates of John hit the ball show that the ball is the 'undergoer' or Goal of the process, whereas those of John marched the prisoners reveal that the prisoners is Actor. Crucially, the relations of agnation and non-enation, which seem to underlie Halliday's metafunctional distinction between interpersonal and ideational categories, can be said to tie in with the overall import of the metafunctions. The interpersonal layer of organization in language is mainly concerned with fitting representational content into the communicative event of the language exchange. It represents different ways of construing one and the same experiential meaning. It seems natural, then, that the various construals which the interpersonal function of language can give rise to, are linked up by means of relationships of agnation, which relate constructions of which the lexical items are identical, but the structure is different. The representational metafunction of language, in contrast, describes and categorizes the language user's experience of the world and seems to do that primarily by ordering or classifying the processes and participants in it into distinct categories. Not surprisingly, these categories are not agnates of each other (representational functions do not operate with identical lexical items organized in different patterns). Rather, they are nonenate, i.e. they are structurally, nor lexically identical. In short, the relations of agnation, enation and non-enation form an important undercurrent of Halliday's Systemic-Functional Grammar: they motivate the categories or options that are distinguished in his system networks and thus help to draw up the paradigmatic side of a construction,

40

Theoretical assumptions

and, ultimately, the system of language. When used as a heuristic for the description of particular constructions, they seem to serve one of two purposes: either they disambiguate structures which seem identical, but are non-enate and they help to interpret the semantic differences between them (e.g., John hit the ball vs. John marched the prisoners)·, or they reveal the various construal types which a construction can occur in and indicate the semantic impact of each construal type (e.g., He opened the door vs. Did he open the door?). In any case, the relations which constructions hold with other constructions seem to form an essential part of the 'family of structures' that constitutes a grammatical construction (Langacker 1988: 149).

6. Theoretical assumptions: Conclusion In this chapter, I have set out the overall cognitive-functional view of the language system in which my analysis of nominalization has to be situated. Among the most prominent features of this cognitive-functional perspective, I have argued, are: its usage-based character (Section 1.1); the attention that is devoted in it to both schematization and extension or prototypicality (Sections 1.2 and 1.3); its view of grammatical constructions as natural symbolic units which encode meaning (Section 2.2); its multifunctional analysis of constructions (Section 4); and, finally, the interest which it shows in the systemic or paradigmatic side of constructions and the importance which it attaches to inter-constructional relationships (Section 5). I have also argued in favour of a radically functional approach to composition and to the construct of 'class' (Sections 3.1 and 3.3) and I have pointed to the existence of a functional hierarchy or rank scale on which constructions are situated and which identifies them as combinations of specific external and internal functions (Section 3.2). In the next chapter I will confront these theoretical insights with analyses that have been suggested for nominalization. I will show how the cognitive-functional framework set out in Chapter 2 not only enables us to identify the lacunae and weak points in the literature but also gives an indication as to how to move beyond them and set up a more coherent and systematic account of the constructional mechanisms that underlie nominalization.

Chapter 3 Nominalization Nominalization is not only pervasive but theoretically significant .... (Langacker 1991: 22) The focus of this study is on deverbal nominalizations, i.e. on nominal constructions which, at some layer of their assembly, incorporate a reclassified verbal component. The nominalizations included in the discussion differ as to the level of assembly at which the nominalization process takes place: firstly, they may be structures that have been transcategorized from verb to noun class, i.e. nominalizations at the level of the word and with a simple nominal head (examples 1 and 2); secondly, they may nominalize structures that are intermediate between a verb and a full clause and thus form complex nominal heads (illustrated in 3); thirdly, they may consist of clauses functioning as full nomináis (as in 4) (unless otherwise mentioned, the source of the examples is the COBUILD corpus, henceforth CB): (1) (2)

computer, teacher, bestseller ... removal of references to the monarch is an act of sedition and a breaking of the oath of office taken by politicians. (CB) (3) Since old Nanapush had saved her from death, since he was her only friend on the reservation, her visiting the man stood to reason. (CB) (4) I am not a father who regrets that he didn 't see them enough because of work. (CB) My interest in nominalization has been prompted by the observation that, in spite of the fact that nominalizations have been extensively discussed in the literature, no systematic and coherent theoretical-descriptive approach of nominalization has been provided yet. Too often, nominalizations are regarded as 'empoverished' clauses, rather than as constructions in their own right. Moreover, descriptive accounts of their lexicogrammatical properties tend to be biased towards representational categories (such as the process-participant relations which they express), to the neglect of other functional categories. And finally, few attempts have been made to connect the lexicogrammar of the various nominalization types to the meanings which they realize. General problems such as these are not simply descriptive but can be related to theoretical problems such as the absence of the notions of rank and rank shift, the lack of attention that is paid to form-

42 Nominalizaîion meaning correlations and the absence of a multifunctional perspective in the description of nominalization patterns. The theoretical-descriptive model which I will attempt to develop presupposes a coherent theoretical view of the language system. It is in the dialectic between theory and description that the answers to many of the moot points in the analysis of nominalization lie. The analysis of nominalization itself then turns out to be "theoretically significant" (Langacker 1991: 22) and contributes insights to issues such as form-function relations, the multifunctional analysis of linguistic organization, and rank and class. In what follows, I will present a first confrontation between theory and the phenomenon of nominalization. I will follow the order in which I set out the theoretical model in Chapter 2.1 will start by pointing out some of the properties of nominalizations that reflect the usage-based character of the system that lies behind them (Section 1). I will then reflect on the status of nominalized constructions as symbolic units characterized by a natural relation between semantics and lexicogrammar (Section 2). In Section 3, issues related to the composite nature of nominalizations will be discussed, together with the systemic-functional notions of 'rank shift' (Halliday 1961) and 'reclassification' (McGregor 1997). The advantages of a multifunctional interpretation of nominalization will be treated in Section 4, followed by an analysis of the role of agnate relationships in the analysis of nominalization patterns (Section 5).

1. A usage-based account of nominalization Nominalization constitutes a domain that is characterized by the frequent occurrence of novel formations and extensions and by varying degrees of productivity. Not surprisingly, therefore, the analysis of nominalization can benefit greatly from a usage-based approach or an approach which views nominalizations as 'systemically motivated' constructions which are structured with reference to the conventional units of language, and linked to them in terms of schematization and extension (Section 1.1); an approach, also, which recognizes that, as a result of the constant interaction between language use and the language system, nominalizations are characterized by varying degrees of entrenchment, which not only explain (diachronic) changes in productivity but also account for the distinction between 'lexicalized' and 'ad hoc' nominalizations (Section 1.2).

A usage-based account of nominalization 43 1.1. Schematization and extension First, the analysis of a specific nominalized construction involves a reconstruction of the various categorizing judgements involved in its construal. It entails an attempt to reconstrue the constructional schemata which the nominalization instantiates or extends and tries to map out the family of structures which the nominalization, as a complex category, is related to. Two claims of Cognitive Grammar relating to the nature of complex categories are particularly relevant to the analysis of nominalization (Langacker 1987a, 1988, 1991, 1999): firstly, categorization by prototype and categorization by schema are claimed to represent complementary modes of categorization; and, secondly, categorization by schema is said to involve both highly abstract and low-level schemata, the latter being probably more important for the computation of novel expressions than the former (Langacker 1988: 133). Langacker has repeatedly pointed out that "extension tends to be accompanied by schematization", i.e. "that the 'outward' growth of a network by extensions from a prototype tends to induce its 'upward' growth via the extraction of higher-level schémas" (Langacker 1999: 102). When, in other words, a construction extends a schema, the language user still perceives some similarity between the extension and the schema which it extends, and this similarity - representing the commonality between the basic schema and the extension - constitutes "a schema having the two for instantiations" (Langacker 1988: 139). Extension is thus inextricably linked to schematization and leads to the extraction of a schema that generalizes over a more diverse group of category members (Langacker 1988: 140). Langacker also stresses the importance of recognizing both high-level schemata which generalize across all instances of a construction type and low-level schemata which hold for only a limited set of data and explain the variation that exists among them. Low-level schemata cannot be considered in isolation from the high-level schema which they instantiate because they primarily specify which of the options that are sanctioned by the highlevel schema "are in fact conventionally exploited" (Langacker 1991: 47). How does this view on categorization tie in with the analysis of nominalizations? When a number of nominalized constructions are parallel in formation and a pattern is discernible among them, the language user extracts a constructional schema that represents their commonality (Langacker 1991: 47). The various instantiations of that schema are then categorized as members of one particular type of nominalization (e.g., deverbal -er nominalization; gerundive nominalization), and their structure is considered to be "motivated rather than arbitrary" (Langacker 1991: 47). In

44

Nominalization

addition to this highly abstract schema, various low-level schemata can then be distinguished which account for variations within the system and which, along with the high-level schema, constitute the backbone of the schematic network that characterizes the nominalization type. Research into nominalizations, however, has tended to be primarily oriented towards lowlevel schemata, and it often refrains from trying to identify the high-level, generalizing schema that lies behind them. A case in point is the analysis of deverbal -er nominalization. Descriptions of deverbal -er nomináis mostly consist of elaborate enumerations of the wide range of entities which the -er suffix can profile (e.g., human Agents as in teacher and hunter, inanimate Agents as in eye-opener and reminder, Instruments as in eraser and poker, Locations, as in diner and sleeper, etc.). Because -er nominalizations are so diverse in profile, few attempts are made to link up the lowlevel schemata which they instantiate with the high-level schema that generalizes across them. Yet, schematic networks consist of high-level and low-level schemata. If it aims at being accurate, an account of a particular nominalization type therefore necessarily involves a description of both.

1.2. Entrenchment Schematic networks are not only characterized by different degrees of specificity (ranging from concrete expressions to low- and high-level schemata); the units that belong to them also display varying degrees of entrenchment or automatization. Constructions fall "along a continuous scale of entrenchment in cognitive organization" (Langacker 1987a: 59) and differ in the degree to which they form "well-rehearsed routines" (Langacker 1999: 93). Two phenomena related to the degree of entrenchment of a unit are particularly relevant to the analysis of nominalization patterns, viz. productivity and lexicalization. First, the likelihood that a structure is activated and leads to the construal of novel units - i.e. its productivity seems to correlate positively with both entrenchment and specificity (Langacker 1991: 48). Secondly, along with specificity, the degree of entrenchment of a unit also turns out to play an important role in the lexicalization of it. As far as lexicalization is concerned, the parameters of specificity and entrenchment are equally determining: structures come to form part of the "set of fixed expressions in a language" (Langacker 1991: 45) if, firstly, they are high in specificity and represent individual expressions rather than schemata, and, secondly, if they are highly entrenched, i.e. have achieved the status of conventional units (Langacker 1991: 45). Note that symbolic

A usage-based account of nominalization 45 complexity is not relevant to lexicalization: lexical items can take the form of morphemes, words, phrases and even longer expressions (Langacker 1991: 45; see also Halliday 1961: 274). Lexicalized constructions are to be distinguished from grammatical patterns which are sufficiently entrenched to be established as schematic units, but not as specific expressions. Grammatical patterns are either situated at word level (traditionally called 'morphology') or at a level higher than the word (i.e. 'syntax').12 Lexicalizations are thus not always simple words, and non-lexicalized items are not necessarily larger than the word. Both for that reason, and because the degree of entrenchment which determines the lexicalization of a unit fluctuates continually, the boundary between lexicon and grammar is not a sharp one, but rather forms a continuum. The distinction between lexicalized constructions and constructions which are schematic rather than specific is highly relevant to the discussion of nominalization patterns such as those in (5) and (6): (5) Alex Gough, the Welsh destroyer of Del Harris's hopes on Tuesday ....(CB) (6) ... a passenger steamer, which crossed the Channel escorted by a destroyer. (CB) The nominalizations in (5) and (6) show that, in a schematic network that elaborates one and the same abstract schema (in this case, that of [[V/.. .][[-ER]/[-er]]), some instantiations are lexicalized, while others are not. Whereas destroyer in (6) forms a lexicalized nominal and has itself achieved the status of a conventional unit, destroyer in (5) is what I will call an 'ad hoc' nominalization: it has been formed according to the same high-level schematic symbolic unit [[V/...]-[[-ER]/[-er]] as has the lexicalization destroyer, but it is not included in the language system as a particular, fixed expression.13 As to the productivity of particular schematic units - i.e. the likelihood of their being activated for the construal of novel units - Langacker (1991: 48) remarks that low-level schemata are more likely to be invoked than more abstract schemata. He (1991: 48) argues that evidence in favour of this claim comes from the fact that instantiations of a schema are normally concentrated in certain regions which correspond to low-level schemata, rather than being "distributed randomly through the space of possibilities defined by the highest-level schema". Langacker's claim that it is low-level rather than high-level schemata which determine the production of novel expressions is comparable to what is traditionally referred to as analogy or the "direct modelling of novel expressions on the basis of familiar ones"

46 Nominalization (Langacker 1999: 144). It differs from analogy in its emphasis on the role of schematization: "even the learning of specific expressions (required as the basis for analogy) involves abstraction and schematization from actual usage events" (Langacker 1999: 144). The low-level nature of productivity which Langacker describes can be clearly observed among the agentive and the non-agentive instances of -er nominalization. Both are characterized by certain 'regions' of productivity or low-level schemata which give rise to instantiations that are sufficiently entrenched to be established as units. Agentive regions are, for instance, those of agent-like instrumental -er nomináis (e.g., computer, printer, parser), 'second-order' Agents (as inpageturner 'a book that almost forces you to turn the pages' and yawner 'an event that makes you yawn') and 'professional' Agents (such as programmer and fire-fighter). Non-agentive regions of productivity are especially those that designate non-agentive Instruments (e.g., front-loader, stroller, stepper, walker) and items of food (e.g., broiler 'a chicken suitable for broiling', cooker 'an apple for cooking', sipper 'a drink that you have to sip', dipper 'vegetable/fruit/other snack that has to be dipped before being eaten').

2. Nominalizations as non-arbitrary symbolic units In spite of the fact that nominalizations are symbolic units which encode their meaning, research on nominalization patterns often seems to consider the analysis of lexicogrammatical properties as an end in itself and only rarely attempts to link it up with the semantics of the construction. Certain lexicogrammatical features moreover tend to be - erroneously - considered as 'meaningless' and are therefore frequently left unanalyzed. The tendency to leave things at a detailed account of a construction's lexicogrammatical properties only is perhaps best illustrated by the following nominalizations: (7) (8) (9)

And it's the third anniversary of the opening of East Germany's borders. (CB) Opening a book is like opening the door into a wonderful new world. (CB) I still regret opening the door so suddenly, (personal example)

In Lees' (1960) account of nominalization, the nominalization in (7) is labelled 'action nominal', the nominalization in (8) is described as a 'gerundive nominal referring to actions', and that in (9) is said to be a 'gerun-

Nominalization as composition and reclassification

47

dive nominal referring to facts'. In spite of the fact that he gives an excellent description of the lexicogrammatical characteristics of each of these types of nominalizations, Lees does not succeed in tracing the origin of the nominalizations' semantic differences in them. Since Lees (1960), numerous analyses have been devoted to these nominalization types, some of which significantly elucidate certain of their properties and the meanings associated with them (e.g., Kiparsky and Kiparsky 1971; Schachter 1976; Halliday 1985, 1994; Langacker 1991; Davidse 1991). Yet, it remains unclear why, for instance, the nominalization opening the door into a wonderful new world in (8) designates what Lees calls an 'action', while opening the door so suddenly in (9) is factive in meaning. Likewise - and in spite of the fact that their lexicogrammatical differences have been extensively described - the differences in meaning between the 'action' nominal in (7) and the 'gerundive nominal' in (8) remain to be clarified. One of the reasons why the semantic distinction between nominalizations like the ones in (7) and (8) has not yet been fully accounted for, is that the semantic import of the grammatical marker of in (7) has tended to be ignored (Langacker 1991: 35): along with the 's of Zelda 's in a nominalization like Zelda's signing of the contract, o/has often been assumed to be 'empty' and inserted for purely grammatical purposes (Langacker 1991: 35). And even if grammatical markers like's and of are not explicitly argued to be meaningless, they are not taken into account in the analysis of the nominalization in which they function. The analysis of nominalization offered in Halliday (1994) and Halliday and Matthiessen (1999), for instance, ignores the presence of periphrastic markers by relating nominalizations such as the writing of business programs and writing business programs to the same clausal agnate, i.e. to (people) write business programs. However, grammatical markers such as the and of form part of the lexicogrammatical properties of nominalizations and motivate a facet of their composite structure. As Langacker (1991: 35) points out, they "invariably have some kind of conceptual import, which may be quite abstract but is nonetheless essential to their function". A natural account of nominalizations has to take account of all lexicogrammatical properties, including those the import of which at first sight seems negligible.

3. Nominalization as composition and reclassification In addition to relationships of categorization and symbolization, a construction also establishes compositional relationships by integrating two or more component symbolic units into a complex unit or a symbolic assembly.

48 Nominalization Two aspects pertaining to the assembly of symbolic structures are particularly important for nominalization: first, the analysis of a construction should not only deal with the construction as product, but should also consider the order of assembly of its component parts, i.e. the construction as process (this will be dealt with in Section 3.1). Secondly, both within and above assemblies of symbolic structures, different levels of organization can be distinguished which are functionally defined and which together form a hierarchy or rank scale. One aspect of the analysis of a construction is to situate it relative to this rank scale. I will show that, in the case of nominalizations, a distinction can be made between, on the one hand, nominalizations which reclassify a verbal unit of the rank of the word into a nominal unit of the same rank, and, on the other hand, reclassification which involves a clausal unit being 'downranked' to nominal rank (Section 3.2).

3.1. Constructions as assemblies of symbolic units The importance of looking at the order of assembly of the component parts of nominalized constructions is illustrated by the analysis which Langacker (1991: 32) gives of the following nominalizations: (10) Sam's washing of the windows was meticulous. (11) Sam s washing the windows was a shock to everybody. Langacker shows that these nominalizations essentially differ as to the level of organization at which the nominalization process takes place: in the action nominalization in (10), the verb stem wash has first been nominalized and turned into a noun, i.e. washing, which then functions like any other noun as nominal head. In the gerundive nominal in (11), on the other hand, the nominalization process applies to a "higher-level structure" or "a processual expression that has all the ingredients of a finite clause except an explicit subject and a grounding predication" (Langacker 1991: 32). When this higher-level structure is nominalized, it comes to function as a whole as nominal head, which can then be premodified by Sam's. The order in which the component parts of the nominalizations in (10) and (11) have been assembled thus determines their structural outlook.14

Nominalization as composition and reclassification 49 3.2. Nominalization as functional reclassification Nominalization types differ according to the level of organization at which the nominalization process takes place (see also Langacker 1991). As pointed out in the introduction to this chapter, three types of nominalizations can be distinguished: nominalizations at the level of the word (e.g., teacher, Sam's washing of the windows), nominalizations which nominalize a structure that lies in between a verb and a full clause (e.g., Sam's washing the windows) and, finally, nominalizations consisting of full clauses (e.g., that Sam washed the windows). The latter two types deviate from the 'normal' rank scale of units in that they represent nomináis or phrases which consist of clausal or clause-like structures. They have therefore often been regarded as problematic, and it has even been claimed that /AaZ-structures are not nominalizations (e.g., Dik 1997; McGregor 1997). In what follows, I argue that structures with a clause-like component (washing the windows as in Sam's washing the windows) and even constructions which have all the characteristics of a full clause (that Sam washed the windows) can be said to be nominalized and 'reclassify' a clausal unit into a nominal one, and this both externally and internally. A notion that is crucial for the analysis of this kind of nominalization, however, is that of 'rank shift', which was introduced by Halliday (1961) as a constituency-bound mechanism, and which I have reinterpreted in a radically functional way in Chapter 2 (Section 3.2). McGregor (1997) has added a further capital clarification by pointing out that rank shift entails reclassification. I will discuss this point and its relevance to nominalization in Section 3.2.1. In a second part (Section 3.2.2), I will briefly introduce the idea that all nominalizations - i.e. those which imply rank shift as well as those which do not - involve a functional reclassification which is both external and internal.

3.2.1. Rank shift and reclassification Halliday speaks of downward rank shift, when "an item normally having the function of (entering into the paradigmatic and syntagmatic relations associated with) rank χ characteristically 'loses' these functions on taking over those of rank y" (Halliday 1966b: 114). A clause which is rankshifted to phrase level (Halliday 1961: 253 speaks of 'group' level) thus "cannot enter into direct syntagmatic relations with clauses outside the structure of that group" (Halliday 1966b: 114). As Halliday defines it, rank shift basically involves a change in the external functions of a symbolic unit: a

50 Nominalization clause which is rankshifted to phrase level, for instance, acquires the external functioning potential of phrases (as is shown, in, for instance, [That he committed the crime] surprised us, where the that-clause has come to function as Subject of the verb surprised). Halliday (1961: 261) concludes that "by reference to the rank scale, classes are derived 'from above' (or 'downwards') and not 'from below' (or 'upwards')". Because rank shift involves "a change in part-of-speech membership" (McGregor 1997: 128), McGregor defines downward rank shift as reclassification. Most importantly, and against Halliday, he holds that rank shift involves external as well as internal reclassification. He (1997: 127) defines rank shift as "the process whereby a unit of a given rank is as it were demoted in size, and reclassified as a unit of lower rank, as a result of which it takes on the grammatical and semantic properties inherent to the lower ranking unit". He (1997: 128) further points out that, semantically speaking, the reclassification of, for instance, a clause into a nominal unit implies that the clausal meaning is 're-presented' "as a thing - a secondorder entity - rather than an ongoing phenomenon located at some point in space and time". In addition, McGregor emphasizes that there must be some formal registration of the rankshifted status of a unit and he argues that rankshifted clauses are characterized by a reduced number of grammatical options that are available. Among the options that are no longer available are "those which permit the anchoring of the situation to a specific occurrence in the referent world" (McGregor 1997: 130). From this premise, McGregor (1997: 131) draws the conclusion - wrongly in my opinion - that finite clauses cannot be rankshifted or nominalized, whereas non-finite clauses can. (A similar view is expressed in Dik 1997.) Finally, McGregor (1997: 130) states that, apart from having a reduced structural potential, the embedded clause may also acquire morpho-syntactic properties that are normally associated with nomináis. In my opinion, McGregor rightly points to the importance of including the internal properties of rankshifted units into their description and classification. Yet, he fails to clarify their significance to the analysis of rankshifted units, because he only considers overtly realized structural properties, instead of taking in a functional perspective. By focusing on the clausal structural properties of the downranked clause only, McGregor fails to observe and account for its nominal functioning. This results in inexplicable discrepancies between the external (nominal-like) functioning of particular rankshifted units (e.g., as Subject of a clause as in [That he was the one that killed her] surprised us most) and the internal status of these units, which McGregor describes as non-rankshifted. McGregor's analysis also fails to clarify the correspondences in meaning and in external functioning

Nominalization as composition and reclassification

51

of finite and non-finite nominalizations (such as [That he killed her] surprised us and [His killing her] surprised us). In McGregor's view, gerundive (non-finite) nominalizations are rankshifted and discharge a function in the clause; that-structures, by contrast, are not rankshifted and serve no such function. As Langacker (1991: 35, 449) points out, however, when a finite clause comes to function as a constituent in another clause, the finite grounding which it establishes as a clause may retain its processual character, but it is viewed through an additional viewing frame which construes it holistically, as a unitary entity. I will moreover show in my discussion of factive nominalizations in Chapter 9 that full clauses and intermediate clausal expressions which are downranked are fitted into the nominal paradigm and connect with other, non-nominalized nomináis.

3.2.2. Nominalization as functional reclassification I will show that all nominalizations - irrespective of the level of assembly at which the nominalization process has taken place - are essentially nominal constructions and involve the functional reclassification of a nonnominal into a nominal unit. Each nominalized construction thereby adopts external but also internal functions of the target class. In those cases where a simple verbal root is nominalized, typically by means of a nominalizing suffix, the nominal status of the reclassified unit is not problematic. However, if nominalizations contain clear units of clausal patterning, then the nominal character of the construction has eluded linguists more - or it has not been pursued much in linguistic analysis. In fact, a 'constructional' analysis of these nominalizations (which integrates their clausal and nominal characteristics) has not been achieved. Even though the theoretical concepts of reclassification and rank shift are available, a descriptive breakthrough with respect to these clause-like nominalizations is needed. This descriptive breakthrough, I will show, can be realized by means of the functional categories which Langacker (1991) has proposed for the clause and the nominal and which I will zoom in on in Chapter 4. In particular, all nominalizations will be argued to realize the highly abstract schema which Langacker (1991) identifies for the nominal and according to which a nominal realizes a 'grounded instance of some type'. I will show, however, that the implementation of these functions ranges from being discernible (e.g., in the case of action nomináis such as the opening of the door) to being non-prototypical (e.g., in gerundive nominalizations such as [Opening the door] was not a good idea).

52 Nominalization 4. A multifunctional account of nominalization The account of nominalization which I will attempt to develop is also built on the assumption that constructions should be given a multifunctional analysis. Nominalizations encode ideational, interpersonal and textual meaning. Their semantic import cannot be reduced to, for instance, their representational semantics and certainly not to the state of affairs that is represented by the verbal component. What is encoded of their relation to the speech event (i.e. their interpersonal meaning) and to the immediate textual context (their textual functioning) has to be brought into the picture as well. In Chapter 4,1 will argue that Langacker's description of the nominal in terms of type specification, instantiation, quantification and grounding allows me to integrate these functional strands. Research on nominalization patterns has thus far focused primarily on ideational or representational functions: nominalizations have mainly been analyzed in terms of the kind of experience which they designate. In Matthiessen (1995) and Halliday and Matthiessen (1999), for instance, nominalizations are primarily seen as a resource for reconstruing the speaker's experiences in a nominal way. However, by describing an action nominalization such as Tom's cleaning of the kitchen and a gerundive nominal like Tom's cleaning the kitchen only in terms of the kind of process that underlies them and the participants that accompany it, one cannot shed light on the important constructional differences that exist between the two nominalizations. Another example of a nominalization type that has mainly been analyzed in ideational terms is -er nominalization: suffixation with -er has been described as a system that is basically agentive in profile, but allows for non-agentive formations as well (e.g., among others, Marchand 1969; Levin and Rappaport 1988; Ryder 1991, 1999b). The processes that are being reclassified by it have been argued to be typically transitive (Lemmens 1998). However important it is for the analysis of -er nominalization, a purely experiential description of -er suffixation, however, is unable to identify a high-level constructional schema that generalizes across all instances of -er nominalization, irrespective of which type of process they derive from and of whether they are agentive or non-agentive.

5. The role of agnation in the analysis of nominalization The analysis of nominalization can also benefit from the identification of agnate constructions, or structures which nominalizations are systematically related to. Gleason (1965: 211) himself has pointed out that agnation

Agnation and nominalization 53 can relate "two constructions which are not on the same grammatical level", as is the case with the clause the boy runs and the nominalization the boy's running (Gleason 1965: 212). In fact, relationships of agnation, enation and non-enation have always figured in descriptive accounts of nominalization patterns (and, indeed, of structural patterns in general: the central idea underlying the construct of 'transformation' in Transformational Grammar was that of capturing systematic relationships between different structures). In Lees's (1960) Grammar of English Nominalizations, for instance, it is thus relationships of agnation which underlie the transformational relation between nominalized structures and 'deep structure' clauses (Lees [1960] 1968: 32, 65). Nominalizations and the clauses they are said to be transformationally derived from are agnates in that they have the same major lexical items, but differ in structure: (12) his rapid drawing of the picture (13) his bringing up the box (14) John is the signer of the check.

» o

He drew the picture rapidly. He brought up the box. John signs the check.

Notice that the 'tests' which Lees uses to identify the differences in structural behaviour of action nomináis (e.g., his drawing of the picture) and gerundive nomináis (e.g., his drawing the picture) are likewise based on the assumption that constructions can only be analyzed in more detail if their systemic or paradigmatic relations are considered: in the examples in (15) and (16), for instance, the paradigmatic relations of Aw and of drawing as they function in his drawing of the picture and his drawing the picture are considered (Lees 1968: 65-66): (15) a. his drawing of the picture b. his drawing the picture (16) a. his drawing of the picture b. his drawing the picture

«

o

the drawing of the picture *the drawing the picture *his having drawn of the picture his having drawn the picture

Agnation may have been used before in the analysis of nominalization, but it has never been fully exploited. Often relationships of agnation fail to be interpreted and linked up with the semantics of the nominalized construction in question (e.g., Lees 1960). Or they are used in a way which deviates from Gleason's (1965) understanding of them in that, rather than being viewed as a network or grouping of structures, agnation is restricted to one-to-one correspondences between nominalizations and clausal struc-

54 Nominalization tures (as happens in Lees's transformational account of nominalization, but also in Halliday 1994 and Halliday and Matthiessen 1999). In my analysis of nominalization, I will attempt to use agnates more systematically. Three aspects of agnation seem to me important to the description of nominalization patterns. First, relationships of agnation are always part of elaborate networks which characterize a specific construction. Secondly, agnation can be used as a heuristic not only for the disambiguation of structures (as discussed in Chapter 2, Section 5), but also for the identification (and description) of the components of a construction. And, finally, relationships of agnation are best captured in schematic terms. They enable one to derive high-level schemata capturing the general meaning of a construction type.

5.1. Networks of agnation As I pointed out in Chapter 2 (Section 5.1), Gleason does not view agnation as "a unique relation between isolated pairs of sentences", but argues that it is based on the 'pervading patterns' of language and that it is "therefore always a recurrent thing, involving large numbers of sentences" (Gleason 1965: 202). Gleason has also pointed out that one agnate structure highlights a particular grammatico-semantic feature of a construction, but does not uniquely identify the structure which it agnates with. A detailed picture of the grammatico-semantic choices that are realized by a particular structure can thus only be provided by agnation networks, in which each agnate structure sheds light on one or more of the structure's properties. Gleason's understanding of agnation uncovers one of the central weaknesses of the transformational-generative and the systemic-functional approaches to nominalization, viz. their being focused on one - necessarily clausal - agnate.15 Clausal agnates are very important to the analysis of nominalized structures, but other, non-clausal agnates may contribute to the identification of the lexicogrammatical properties of a nominalization as well. The focus on one agnate structure which is clausal in nature tends to lead to incomplete or even inaccurate analyses. For instance, in Halliday and Matthiessen's (1999) account of nominalization, both a structure such as Tom's cleaning the kitchen and Tom's cleaning of the kitchen would be related to the clause Tom cleans the kitchen. This, however, fails to identify the relevant features of the nominalizations it is related to and needs to be accompanied by a set of other agnates to distinguish the gerundive nominal Tom's cleaning the kitchen from the action nominalization Tom's cleaning of the kitchen}6

Agnation and nominalization 55 The central role which is assigned to clausal agnates moreover tends to reduce nominalizations to clause-like constructions, rather than viewing them as unique - nominal - constructions in their own right. This is especially the case in the transformational-generative approach, which views nominalizations as constituting the 'surface level' of language, while their 'source sentences' (Lees 1968: 65) are conceived of as the 'deep' or more fundamental level of linguistic organization. The distinction between deep and surface structures is rejected in Cognitive Grammar, which claims "that grammatical structure is almost entirely overt: things really are what they appear to be, provided we know how to interpret them properly" (Langacker 1987a: 27). Systemic-functional linguistics likewise rejects the distinction between deep and surface levels of organization, but by focusing solely on clausal agnates nonetheless fails to analyze nominalizations as constructions in their own right.

5.2. Agnation as a descriptive heuristic As noted in Chapter 2 (section 5.1.), Gleason (1965: 203) shows that agnates can crucially be used to disambiguate apparently identical structures. This also applies to the disambiguation of nominalizations. Consider, for instance, the nominalized structures in (17) and (18): (17) a. I still regret not going to school. (18) a. Not going to school is not a very sensible thing to do. At first sight, the structure not going to school in (17a) is identical to that in (18a). Yet, if one considers the set of agnates that is related to not going to school in (17a) and in (18a) (illustrated in 17d, 17e, 18d and 18e), as well as the paradigmatic options of their component functions (illustrated in 17b, 17c, 18b and 18c), it becomes clear that they are different structures: (17) b. I still regret not having gone to school. c. I still regret their not going to school. d. I still regret that I did not go to school. e. I still regret the fact that I did not go to school. (18) b. *Not having gone to school is not a very sensible thing to do. c. *Their not going to school is not a very sensible thing to do. d. *That you do not go to school is not a very sensible tiling to do. e. *The fact that you do not go to school is not a very sensible thing to do.

56 Nominalization In my opinion, yet another function of agnation lies in the identification of the component parts of a construction. A good example here is the analysis of the nominalization in the following clause: (19) Sam's washing the windows was a shock to everybody. (Langacker 1991: 32) I have pointed out in Section 3.1 that Langacker's analysis of the order of assembly of this nominalization is essential to distinguish it from the action nominalization Sam's washing of the windows. Langacker (1991: 32) has pointed out that the gerundive nominalization in (19) nominalizes a "higher-level structure" or "a processual expression that has all the ingredients of a finite clause except an explicit subject and a grounding predication". The fact that this 'higher-level' structure combines washing and the windows into one component unit of the assembly Sam's washing the windows, which is then premodified by Sam's, is confirmed by agnation. Washing the windows agnates, for instance, with wash the windows as it is used in the constructions in (20), (21) and (22): (20) You will wash the windows, won't you? (21) What will you do? Wash the windows. (22) I will wash the windows. Will you? Yes, I will. The clauses in (20), (21) and (22) illustrate that, also within clauses, a verb (minus the grounding elements of tense and modality) and its Object(s) combine into a separate component unit (Halliday 1994). The clause-level agnates thus provide evidence in favour of the claim that the order of assembly described in Langacker (1991) is the correct one and that washing the windows constitutes a component unit in the assembly of Sam's washing the windows.

5.3. The schematic nature of agnation In this final section, I put forward my view that agnation is best thought of as involving degrees of schematicity·. relationships of agnation between specific examples of constructions always involve relations between the schemata which they instantiate. I also argue that agnate relationships between nominalizations and clauses reveal the importance of clausal categories to the analysis of nominalized constructions.

Agnation and nominalization 57 Both claims can be illustrated by a critique of the transformationalgenerative approach to nominalizations which was formulated by Dik (1967). One of the points which Dik argues for is that nominalizations and the clauses they are related to should be viewed as two distinct constructions: often "the 'derived' complex word contains a semantic aspect not present in the proposed underlying structure", and "the proposed underlying structure contains semantic aspects not present in the 'derived' complex word" (Dik 1967: 378). Dik gives the example of John is the signer of the check, which in Lees (1968: 32) is identified as a transformation of the clause John signs the check. While the clause contains an explicit indication of tense, the nominalization is said to be "neutral in relation to time, but [it] can be referentially used in different ways" (Dik 1967: 378): it can, for instance, be interpreted as designating a past event, as in John signed the check. Clauses, Dik concludes, "necessarily contain an aspect of tense, which is not present in the final derived structure" (Dik 1967: 378). Dik's analysis of the relationship between John is the signer of the check and John signs/signed the check illustrates two characteristics of agnation which seem to me essential for the description of nominalization. First, it shows that relationships of agnation are basically schematic in nature and that, in some cases, a relationship of agnation is captured more accurately in schematic terms than as a relation between examples. Secondly, Dik's description of the possible temporal meanings of John is the signer of the check and of their (schematic) link with the clausal tense system provides a first indication of the precise nature of the schematic relationship between nominalized constructions and their clausal agnates. I now develop these two points. I have thus far described agnation as a relationship between specific examples. Each of these examples, however, instantiates a constructional schema. Agnation can therefore be said to essentially involve a relationship between schemata. In fact, in some cases the agnate relationship is perhaps better captured in schematic terms than by means of relations between specific examples. A good example is Dik's analysis of what I describe as an 'ad hoc' nominalization, i.e. signer of the check. Dik - in my view, correctly - argues that it is 'neutral' in time and adopts a particular tense only in the context in which it is used. Various specific clausal constructions can, in other words, be said to agnate with signer of the check: e.g., he signs the check, he signed the check or perhaps even future he will sign the check. One way of generalizing across the different temporal interpretations is by viewing them schematically. Ad hoc -er nominalizations can then be said to relate to the overall temporal mode of grounding in clauses. The relationship of agnation between ad hoc -er nominalizations and temporally groun-

58

Nominalizatìon

ded clauses can, in turn, be captured by means of high-level schémas of which both are instantiations. This abstract schema could, for instance, be argued to be that of 'temporal grounding', or, on an even more abstract level, that of clausal grounding in general. This brings me to a second aspect of the schematic nature of agnation which is relevant to the analysis of nominalizatìon: the schematic relationship of agnation that exists between nominalizations and clausal constructions alerts us to the role of schematic clausal categories in nominalized constructions. The clausal category of grounding, for instance, somehow appears to be relevant to the analysis of the system of -er nominalizatìon. Agnation thus does not only help to disambiguate seemingly identical nominalizations and identify the component units of a nominalized structure; the clausal agnates of a nominalized construction also provide information concerning the internal functional behaviour of nominalizations and the clausal categories that play a role in it. I will argue that this information is crucial for the description of specific nominalizatìon types and for the identification of the basic mechanisms that underlie deverbal nominalizations.

6. Towards a theoretical-descriptive approach to nominalizatìon: Conclusion I have argued in this chapter against the tendency to view nominalizations as 'empoverished clauses' whose main purpose is to periphrastically realize processes and their participants and of which certain lexicogrammatical properties (such as the inclusion of, for instance, of and '5) can be ignored. Nominalizations, I have suggested, should be viewed as constructions in their own right: they are reclassifications of non-nominal into nominal units or units which have adopted nominal functions both in their external and in their internal functioning. An accurate description of these nominal units involves a description of, firstly, their symbolic status (as units which encode a particular meaning); secondly, of the compositional relationships which they realize (in particular, of their component functions)·, and, thirdly, of the (paradigmatic) relationships which they enter into as complex categories (i.e. with the high- and low-level schemata which they instantiate or extend and with agnate constructions). At the same time, however, nominalizations are reclassifications of various levels of assembly of the clause and it is clauses which they most typically agnate with. A natural account of the component functions of nominalizations and of the meanings which they encode can therefore only

Conclusion 59 be arrived at when nominalizations are analyzed not only in terms of the nominal but also in terms of the clausal functional categories that they realize. More specifically, analyses of nominalization patterns should be focused on the way in which nominalizations integrate nominal and clausal categories. In the next chapter (Chapter 4), I will therefore zoom in on the functional categories which, in my opinion, are best qualified to elucidate the integration of clausal and nominal functions in nominalizations, viz. the functions of type specification, instantiation, quantification and grounding (Langacker 1991). Crucially, Langacker has argued that these functions figure in nominal structures as well as in clauses and that they thus reveal a fundamental parallellism between the functional organization of clauses and ordinary nomináis. After I have dealt with the functional organization of nomináis and clauses (Sections 1 and 2), I will set out the main principles of what a functionally-oriented theoretical-descriptive approach to nominalization may look like (Section 3).

Chapter 4 The functional organization of nominal and clause

The key to a descriptive analysis of nominalizations which does not lose sight of the various theoretical concerns raised in Chapter 3 lies in an accurate account of their functional organization. This, in turn, presupposes an insightful account of the functional resources of both the nominal and the clause: only when the parallelisms between the functional organization of nomináis and clauses are clear can the process of nominalization itself be elucidated and the mechanisms that lie behind specific nominalization types be identified. I indicated earlier that Langacker's description of nomináis and clauses, combined with elements of Halliday's multifunctional approach to linguistic organization, offers a very productive perspective on the functional make-up of the nominal and the clause. This chapter will therefore first zoom in on the functional categories which Langacker (1991) and Halliday (1994) distinguish. Langacker's analysis of nomináis and finite clauses as 'grounded, quantified instances of some type' will form the leitmotif of the chapter and will be elaborated in detail; Halliday's metafunctional hypothesis and, in particular, his interpersonal analysis of linguistic organization will be shown to throw a complementary light on the assembly of clauses, from their innermost layer of organization (i.e. the process type specification) to the outermost layer or grounding elements. I will first discuss the assembly of the nominal (Section 1) and then elaborate in detail on the parallels and differences that exist between the organization of the nominal and that of the clause (Section 2). In a final section, I will turn to nominalized constructions and sketch the main principles that underlie the functional reclassification which they realize (Section 3).

1. From noun type specification to nominal The overall function of a nominal is to refer to a thing and make it "a momentary focus of attention" within the speech event (Langacker 1991: 53). To single out the thing one wants to talk about - to the exclusion of other things - one can basically employ two strategies. Either one provides the thing with its own, unique label and uses a proper name·, or one follows a strategy based on type specifications which one then specifies in terms of

62

The functional organization of nominal and clause

quantity, as well as in terms of their relation to the speech event and its participants (the ground) (Langacker 1991: 53). It is by being grounded in the speech event that the instance of the type specification that one wants to talk about becomes uniquely apparent to the speech participants. However different the latter, common noun strategy may seem from that based on proper names, both strategies give rise to nomináis that conform to the schematic characterization of the nominal as a 'grounded, quantified instance of some type'. They only differ in the "structural implementation of these semantic functions" (Langacker 1991: 148): unlike proper names, nomináis with a common noun as head tend to reflect the semantic functions which they realize iconically, with the type specification as the innermost layer of organization, the grounding predication as the outermost layer and intermediate between the type specification (or simple noun) and the grounded type (or nominal), the functions of instantiation and quantification (Langacker 1991: 54). Which is the contribution of each of the semantic functions which Langacker (1991) distinguishes to the ultimate grounded status of the thing designated by the nominal? And how are these functions encoded in the structure of the nominal? Both formally and semantically, the type specification constitutes the innermost functional layer. Semantically speaking, it makes "an initial delimitation among the potential objects of thought" and "specifies the basis for identifying various entities as being representatives of the same class" (Langacker 1991: 53). In the nominal my office, for instance, the noun office specifies the type or class of entities that is being talked about and thus forms the representational core of the nominal (Halliday 1994: 189). Structurally speaking, it is the head noun which provides the type specification, along with any adjectives or other modifiers which render the type specification more precise and add certain refinements: in the nominal an excellent convention site, the head noun convention site is thus rendered more specific by the inclusion of excellent. Postmodifying structures can also add to the type specification, as illustrated by in the Midwest in the nominal an excellent convention site in the Midwest (Langacker 1991: 52). Together with the head noun, these modifiers create a 'higherorder type specification' which forms a kind of nucleus within the nominal and which typically "supplies a vast amount of detailed conceptual content" (Langacker 1991: 143). A nominal type specification is instantiated when it is thought of as "having a particular location in the domain of instantiation" (Langacker 1991: 57). For nomináis, the typical domain of instantiation is space. The instantiation of a type specification implies that one conceives the domain of instantiation as being able to support the simultaneous manifestation of

From noun type specification to nominal 63 multiple entities. It involves anchoring the type specification at a particular spot which is different from other possible locations (Langacker 1991: 57). Only very rarely, a noun type specification is not instantiated (and grounded) and functions as a type specification in the clause: Langacker (1991: 69) gives the French examples of Alain est professeur 'Alain is a professor' and Pierre est médecin 'Pierre is a doctor'. An example of an English non-instantiated noun type may be Mary became Queen (Davidse p.c.). What counts as an instance of a type specification depends on the kind of noun type that is instantiated. Common nouns can, in this respect, be divided into three categories: there is, firstly, the category of the singular count noun which designates a discrete entity (e.g., pebble). An instance of the singular count noun pebble is then a single small stone. There is the category of plural nouns (e.g., pebbles), which Langacker classifies among the mass nouns and which designates a 'replicate mass' in that "considerable prominence is accorded to the discrete entities out of which the mass is constituted" (Langacker 1991: 78).17 An instance of the plural noun pebbles is necessarily a set consisting of two or more stones. A third category of common nouns, then, is formed by non-plural mass nouns such as gravel, which designate a non-replicate mass and focus on the continuous nature of the profiled mass. An instance of the mass noun gravel must be perceived as internally uniform rather than particulate (Langacker 1991: 78). Instantiation of a type specification is an important prerequisite to both quantification and grounding, but, unlike the latter, it cannot receive separate structural symbolization: prototypically, instantiation is indicated by determiners which also serve the function of grounding the instantiated type (Langacker 1991:58). If a nominal is to serve its function of singling out a particular instance of a specified type, it must also include some indication of the size or quantity of the profiled instance (Langacker 1991: 53). Importantly, quantification of an instance does not specify the number of instances which the nominal profiles: every nominal profiles a single instance of a type, and this irrespective of whether the type specification is singular or plural (Langacker 1991: 81). The nomináis the pebble, the pebbles and the gravel thus each designate a single instance of the type specifications pebble, pebbles and gravel. Even when a quantifier is added, as in the seven pebbles, the nominal profiles only one instance. Rather than specifying the number of instances, quantification merely gives more information concerning the size of the designated instance (Langacker 1991: 81).18 Nomináis select from a variety of quantifiers to give a finer-grained indication of the size of the instance which they designate. Langacker distin-

64

The functional organization of nominal and clause

guishes between absolute or true quantifiers, on the one hand, and relative quantifiers, on the other. Absolute quantifiers "offer a direct description of magnitude" (Langacker 1991: 83). Examples are seven, many, few, little. Relative quantifiers specify a quantity in relation to a 'reference mass', which in the default case consists of the "maximal instantiation of the pertinent category" (Langacker 1991: 82). A relative quantifier such as all in all dogs, for instance, profiles a quantity that is identical to the reference mass. In most dogs, in contrast, the profiled mass only constitutes a subpart of the reference mass, be it one "that comes reasonably close to exhausting it" (Langacker 1991: 82). Because they situate the instances which they quantify relative to a definite reference mass, relative quantifiers also serve a grounding function. This explains why they cannot be used with other grounding predications (e.g., *the all dogs, *those most dogs). Absolute quantifiers do not imply a reference mass: the nomináis many dogs, several dogs and seven dogs give a (vague or precise) indication as to the quantity of dogs involved, but they do not situate this quantity relative to a reference mass. Absolute quantifiers do not therefore serve a grounding function and can be preceded by a grounding predication (e.g., those three dogs, the many cats I have owned) (Langacker 1991: 82-83). Langacker argues that every nominal contains some indication of quantity, either in absolute terms (e.g., three cats) or proportionally (e.g., most cats). He does not, however, elaborate on the quantificational information provided by those nomináis which leave the size of the profiled instance implicit or characterize it only 'schematically' (Langacker 1991: 89), such as proper names and nomináis with articles, zero-determiners, demonstratives and possessives. It is interesting in this respect to consider the analysis which Milsark (1976, 1977) gives of determiners according to the type of quantification that they can express (Davidse 1999: 216-217). Milsark's analysis is to be situated in the context of a systematic distributional analysis which he makes of the use of determiners in the existent NP of existential constructions, more precisely in the unmarked 'cardinal' type of existential such as There are five weak spots in the human body (for an extensive discussion of existentials, see Davidse 1999). Milsark distinguishes, firstly, 'cardinality words' which "express the size of the set of entities denoted by the nominal with which they are construed" (Milsark 1977: 23). Cardinal determiners can be used in the existent NP of cardinal existentials and they include, apart from the quantifiers which Langacker (1991) classifies as absolute, also the indefinite article a, the zero-determiner with plural and mass nouns in non-generic reading and, finally, cardinal numbers. If determiners do not give a cardinal value themselves, Milsark argues, they "must always be understood with reference to a set of some cardinality"

From noun type specification to nominal 65 (Milsark 1977: 23), i.e. they are what have been called 'relative' quantifiers. The relative group of determiners comprises the definite article, bare plural and mass nouns with generic reading, pronouns, proper names, possessive and demonstrative determiners, genitives and various relative quantifiers, such as all, every, both, most (Davidse 1999: 217). The latter group of determiners is excluded from the cardinal existential construction. Davidse (1999) argues that Milsark's analysis of the quantifying properties of determiners can be integrated with Langacker's analysis of the nominal as 'quantified instance of some type'. The distinction between absolute (or cardinal) and relative or proportional quantification can, in other words, be extended to all nomináis, irrespective of whether they contain an explicit quantifier. Some of the analyses which are proposed in Davidse (1999) are the following. With regard to non-generically used bare mass nouns and plural count nouns (as in 1 and 2), Davidse argues that they realize absolute quantification (Davidse 1999: 222), because they alternate systematically with nomináis introduced by absolute quantifiers such as some, much and many (illustrated in 3 and 4): (1) (2) (3) (4)

There's milk on the table. There are mice in the shed. There is some/ much milk on the table. There are some/ many/ three mice in the shed.

Generic bare plurals and mass nouns (as in 5 and 6), in contrast, are claimed to receive a relative quantificational reading (Davidse 1999). They resemble proper names in that they designate unique instances. Or, as Carlson (1978) puts it, they are proper names of kinds conceived of as individuals. Unlike proper names, however, the instances which they designate are instances in type space and not in physical space: the nomináis koalas and laurel do not refer to specific, spatially-instantiated animals or plants. Rather, they have to be understood with reference to a type universe, in which they constitute unique individuals (for a more detailed account, see Davidse 1999: 209-215). They realize relative quantification because they "quantify universally, as it were, over the one instance contained by the reference mass in type space" (Davidse 1999: 225). (5) (6)

Koalas carry their young in pouches. Laurel has red berries.

Finally, nomináis which include a possessive determiner (see 7) or a genitive (as in 8) typically realize relative quantification: they invoke the

66

The functional organization of nominal and clause

concept of a reference mass (all the instances of a type specification in the current discourse space), and present the instances of the type specification which they profile as coinciding with either all or part of this reference mass (Davidse 1999: 223-224): (7) My students protested loudly. (8) the teacher's books With the size of the profiled instance specified, it remains for the speech-act participants to establish contact with the instance: the objective of the speaker in using a nominal is to refer to an instance of a type which the hearer can identify. The speaker has to make the profiled instance into something that can be exchanged in communication, something that can be 'talked about'. As Langacker (1991: 91) puts it, "the speaker (S) and hearer (H), who jointly form the ground (G), face the task of coordinating their mental reference to some instance t¡ of type Τ drawn from the reference mass RT". Within the system of identification, the basic options are definite vs. indefinite identification. Indefinite identification presents instances which are not presumed known to the hearer; definite identification marks an instance as 'presumed known'. Definite identification occurs when "full coordination of reference" between speaker and hearer is achieved (Langacker 1991: 91). As pointed out in Davidse (2001: 8), the functions of identification and quantification have not always been properly distinguished: it is thus unclear how the definite-indefinite distinction applies to the various determiners that are found in the nominal, and how this distinction relates to that between relative and absolute quantifiers. Davidse (2001) suggests viewing all definite reference as implying relative quantification. She (2001: 9) argues that "speaker and hearer can be thought of as coordinating their mental reference to P, the actually designated mass, which, through the operation of definite reference, is delineated as an identifiable set or mass". Unlike indefinite reference, definite reference delineates or delimits the portion of the reference mass which the hearer should make mental contact with. Note that this definition of definite reference is compatible with that by Langacker (1991) in terms of 'full coordination of reference to some instance t, of type Τ drawn from the reference mass Rf '; it only differs from it in that the mechanism of relative quantification is brought into the definition and is identified as a prerequisite for definite identification. Davidse (1991: 9) thus takes over the schematic representation of definite reference in Langacker (1991: 92) and only changes the profiled mass t¡ into a Ρ resulting from a presupposed operation of relative quantification:

From noun type specification to nominal

67

Figure 4. Definite identification (see Langacker 1991: 92)

Domain of instantiation Figure 5. Indefinite identification (see Langacker 1991: 92) In Davidse's (2001: 10) approach, indefinite reference "does not delimit a mass or a set of instances" but only involves "the hearer recognizing instances as a t¡ of type T". As Langacker (1991: 104) puts it, indefinite refer-

68

The functional organization of nominal and clause

enee puts an instantiation on the scene but is "not sufficient to put the hearer in mental contact with a uniquely determined instance of the category". In the example Hand me a wrench (Langacker 1991: 104), for instance, the speaker tells the hearer 'what sort of thing' he wants, but a wrench does not delimit a set of one specific wrench. In that the hearer has to be able to recognize an instance corresponding to the type specification used by the speaker, indefinite reference presupposes 'type-identifiability' (Gundel et al. 1993: 275, as cited in Davidse 2001: 13). Besides representing the instantiation relation between t¡ and T, Figure 5 also indicates that speaker and hearer both have to make mental contact with the type T. It is thus not only determiners that contribute to the reference act: the type specification makes an essential contribution to it as well: as Davidse (2001: 12) points out, "successful reference is dependent as much on a sufficiently informative and contextually well-positioned 'categorization' as on the appropriate choice of determiner". In fact, the type specification turns out to play a different role with definite and indefinite identification. In nomináis which are introduced by a definite determiner, Davidse (2000b, 2001) argues, the type description defines a 'reference mass' of 'all the instances corresponding to that type in the discourse context'. The following examples are given to illustrate this point: (9)

When God asked why they were wearing clothes, it was the man who spoke first. (10) Here are the exams that still need correcting. (11) If you take care of your exams that still need correcting, I'll take care of mine. In (10) and (11), the contextually relevant reference mass is 'all exams that still need correcting'. While the definite article in (10) refers to all of those, the possessive your in (11) picks out only some instances. In (9), the contextually relevant reference mass (in the context of the Garden of Eden) consists of only one instance of the type 'man'. The use of the definite article to refer to 'man' is therefore correct. Indefinite nomináis, then, merely designate "'some' instantiation of a type, without indicating whether there are - or aren't - other instances of the type in the discourse context" (Davidse 2000b: 1113). The type specification has a purely "classificational" function (Davidse 2000b: 1113). The designated instance may be presented as indefinite, but specific (as in 12): the instance is then being introduced into the discourse for the hearer's benefit. Or no specific instance may correspond to the indefinite nominal for either speaker or hearer (as in 13) (Davidse 2000b: 1113):

From noun type specification to nominal 69 (12) I had only just arrived when a man suddenly entered the restaurant and started shooting. (13) A tie is always an appropriate present for a man who is rather conservative. I have discussed the semantic functions of type specification, instantiation, quantification and grounding and I have shown how they contribute to the functional organization of the nominal. These functions can be realized by separate words or phrases, ordered in a layered kind of organization, as in those three black cats,19 or they may not be reflected iconically. There are, for instance, nomináis in which the semantic functions are associated with the nominal as a whole. Langacker (1991: 148) gives the examples of proper names, personal pronouns and nominalized finite clauses. Proper names (e.g., John, Belgium) carry with them a type specification (e.g., 'human male', 'nation'). Van Langendonck (1999: 109) refers to it as the 'categorical presupposition' of the proper name and argues that it forms its basic level meaning. Proper names are moreover semantically singular, i.e. they always designate a single instance of the type specification (also when the latter is plural, as in the Alps, the Philippines, Van Langendonck 1999: 109-110). Finally, they are grounded in a definite way. As pointed out in Van Langendonck (1999: 111), the definite nature of proper names is either expressed by a fixed definite article (as in the Nile) or it is incorporated in the structure of the proper name itself (as in John). Like proper names, personal pronouns such as I and she profile a single, definitely grounded instance of a schematically specified type (e.g., 'human' in the case o f / ; 'human female' in the case of she). The distinguishing property of pronouns is that their profile is entirely determined by the speech event and the participants in it: they are either first- and secondperson pronouns referring to the speaker and hearer respectively, or, if an entity is profiled that is distinct from the speech participants, they are in the third person (Langacker 1991: 148). The nominalization of a finite structure, then, also constitutes a nominal, "as witnessed by its ability to occur alone as subject or direct object" (Langacker 1991: 148): (14) That prices will continue to rise is obvious. The type specification, Langacker (1991: 149) argues, is in this case provided by the proposition conveyed by the nominalized clause. Because the clause is finite and uniquely related to the time and the participants of the

70

The functional organization of nominal and clause

speech event, the type specification which it implies has only a single instance which is definitely grounded. Other nomináis consist of two structures which are themselves nomináis and which are in apposition: the two components have "equal claim to the status of profile determinant" and the resulting, composite structure is a "doubly-characterized, doubly-grounded thing" (Langacker 1991: 149). Examples of such appositional structures are the fact that whales are mammals·, all those kittens and my daughter the lawyer (Langacker 1991: 149). Still other nomináis consist of a grounding predication only (e.g., that, these, some, any, all). They can be said to incorporate a type specification, in spite of the fact that they lack a head noun. A nominal such as these thus schematically characterizes the entity which it grounds as being a replicate mass or plural; each, on the other hand, incorporates the type specification of a discrete entity (Langacker 1991: 151). In short, the semantic functions of type specification, instantiation, quantification and grounding can be encoded in many different ways and they do not always receive individual symbolization. Yet, they are realized in every nominal and together, they serve to turn a 'thing' into something which is related to the speech event and its participants, something which one can zoom in on and talk about. Crucial for the analysis of nominalizations is Langacker's claim that the "relation between a simple noun and a nominal is directly analogous to that between a simple verb and a finite clause" (Langacker 1991: 191). In the next section, I will go more deeply into the functional organization of the clause. Again, the analysis that is presented is based on Langacker (1991). More than in the description of the nominal, however, a number of complementary or alternative analyses that have been proposed for certain aspects of clausal assembly are included as well. The latter are especially based on Halliday (1994), Davidse (1992b, 1997), Heyvaert (2000) and Verstraete (2002).

2. From process type specification to clause The pivotal element in the clause is the verb. It may stand on its own (as in 15a), or it may form part of a more elaborate kind of verb group with a varying number of auxiliaries (as in 15b and 15c): (15) a. She cleaned the room. b. She may clean the room. c. She must have been cleaning the room.

From process type specification to clause

71

Langacker (1991) argues that, on the basis of semantic function and grammatical behaviour, the verb group falls apart into two parts, viz. into a grounding predication and a type specification. Importantly, this distinction does not coincide with that between the main verb and its auxiliaries. Rather, Langacker suggests "that the specification of tense and modality be analyzed as the grounding predication, with the remainder of the verb group (other auxiliaries and the main verb) regarded as a complex clausal head analogous to a head noun" (Langacker 1991: 195). For the verb groups in (15), this leads to the following analysis (with the grounding predications in italics): (16) a. She cleaned the room. b. She may clean the room. c. She must have been cleaning the room. Semantically speaking, tense and modality seem to stand apart from the rest of the verb group in that they are "the only auxiliary elements that specifically invoke the ground as a reference point" (Langacker 1991: 195): tense relates the process to the time of the speech event, while the presence of a modal explicitly situates the profiled relationship with respect to the judgement of the speaker. A structural argument in favour of the proposed division comes from the fact that an infinitive (marked by to) can be formed on the main verb together with any combination of auxiliaries, minus the markers of tense and modality. It is thus possible to have to clean, to be cleaned, to have cleaned, to have been cleaning. It is not possible to do the same with a verb group which is marked for tense or modality, e.g., *to should wash, *to had been washing (Langacker 1991: 195). There is, in other words, a distinction that cuts right through the auxiliaries of the verb group and distinguishes those auxiliaries that serve to ground the process type from all the others. Together with the main verb, the latter, nongrounding auxiliaries make up the clausal head. In what follows, I will successively discuss the properties of the clausal head (Section 2.1), the way in which it is instantiated and quantified (Section 2.2), and the means that exist of grounding it (Section 2.3). Both in the instantiation and in the grounding of a process type specification, the Subject will be argued to play an important role. A third function of the clausal Subject, viz. that of 'speech functionally responsible' element in the clause (Halliday 1994), will be described in Section 2.4.

72

The functional organization of nominal and clause

2.1. The clausal head Langacker argues that the head-part of the verb group, which provides the process type specification of the clause, may be quite complex internally: like the head in nomináis, the essential semantic content provided by the main verb (e.g., clean) may be elaborated with any desired degree of precision by the non-grounding auxiliaries with which it is combined (Langacker 1991: 196). Each auxiliary element is said to impose its profile on the structure it combines with, "thereby deriving a distinct higher-order type specification" (Langacker 1991: 197). It is, however, only "the leftmost verb in the sequence" (Langacker 1991: 196) which determines the profile of the entire verb group. If a verb group is to function as clausal head and provide a clausal type specification, Langacker (1991: 199) argues, its leftmost verb must profile a 'process', or a series of component states which are distributed continuously through time and scanned in a sequential way (Langacker 1991: 21, 246). The function of the auxiliaries have (present perfect), be (progressive aspect) and be (passive voice) is precisely to impose a processual profile on the atemporal relation that is formed by the affixal/inflectional pair members -ed (perfect aspect), -ing (progressive aspect) and -ed (passive voice) (Langacker 1991: 199). More particularly, the "members of a given pair must co-occur if the resulting expression is to be processual and thus able to serve as clausal head" (Langacker 1991: 200). Examples of clausal heads which illustrate the use of specific auxiliary pairs are the following: (17) a. They must have known it all the time, (perfect aspect) b. She will be singing in several operas this year, (progressive aspect) c. You will be notified in due course, (passive voice) The auxiliary pairs can also be combined: (18) a. They have argued that the Inkatha leader has been deliberately orchestrating the recent violence. (CB) (perfect and progressive) b. Samples from 90 horses are being tested at the Australian Animal Health Laboratory, in Geelong. (CB) (progressive and passive) c. A computer program which diagnoses schizophrenia has been developed by a University of Queensland research team. (CB) (perfect and passive) d. And will the Queensland Government agree? Well, the papers have been being put into the Cabinet in-basket and taken out of the

From process type specification to clause 73 Cabinet in-basket since some time in March. (CB) (perfect, progressive and passive) Two aspects of Langacker's notion of 'clausal head' deserve further attention. First, there is the fact that clausal heads are presented as being oriented towards being grounded, in finite clauses. Secondly, it remains to be seen whether the verb group minus the grounding predication exhausts the type specification of the clause or whether other clausal elements form part of the clausal head as well. I will start with a discussion of the processual nature of the clausal head and formulate an analysis of the clausal head in English which basically distinguishes two options, i.e. a finite and an atemporal one (Section 2.1.1). In a second part, I will present arguments in favour of viewing a clause's Objects and Complements as part of the clausal head or type specification (based on Halliday 1994 and Davidse 1997) (Section 2.1.2). 2.1.1. Finite vs. atemporal clausal heads Langacker's discussion of the clausal head focuses on finite clauses, i.e. on clausal heads which are grounded in the speech event in terms of tense or modality. I would like to broaden this picture by proposing that the grammatical system of clausal heads offers not one, but two options: verb groups with a processual profile cannot only be grounded; they can also be atemporalized (see also Heyvaert 2000). Atemporal clausal heads constitute an option in the functional organization of the clause which should be accounted for. Which evidence can be adduced in favour of distinguishing a systematic atemporal alternative to grounding the clausal head? I mentioned earlier that one of the main structural arguments which Langacker gives in favour of distinguishing the grounding predications of tense and modality from the rest of the verb group is that it is possible to add a to-infinitive to any verb group minus the markers of tense and modality (e.g., to clean, to have been cleaning, but *to had been cleaning, *to must clean) (see Langacker 1991: 195). It seems to me, however, that this does not only illustrate that it is the non-grounding auxiliaries (like have and be), rather than the grounding predications, which form a strong unit with the main verb. These structures also show that the combination of main verb and non-grounding auxiliaries which Langacker calls 'clausal head' cannot only be turned into a finite expression, but also into a /o-infinitive. The infinitival form can, in fact, be

74

The functional organization of nominal and clause

attached to an equally wide range of auxiliaries and auxiliary combinations as the finite forms. The to-infinitive can, moreover, be replaced by -mg:20 (19) washes/washed has washed has been washing

washing/ to wash having washed/ to have washed having been washing/ to have been washing

Similar to grounding predications, the -ing suffix and the to-infinitive can be used with both so-called 'perfective' and 'imperfective' verbs. Imperfective verbs profile processes with identical component states (e.g., know —> knowing/ to know). Perfective verbs profile some kind of change due to the differences that exist among the component states (e.g., hit -» hitting/ to hit) (Langacker 1987a: 254-267). I propose that the to-infinitive and the -ing forms in (19) represent clausal heads that have been atemporalized. They constitute an alternative to grounding the clausal head by tense or modality. Significantly, grounding a clausal head and atemporalizing it are mutually exclusive choices. Atemporal markers cannot therefore be added to a grounded structure, nor can an atemporalized structure be grounded: (20) a. *to was sitting on that bench (Langacker 1991: 421) *shoulding have sat on that bench (Langacker 1991: 421) b. *should being killed *will to play in the garden However, if the atemporal -ing suffix is elaborated into a structure with be, thus forming the auxiliary pair be V-ing, a new processual structure or clausal head is formed and grounding becomes possible again: when, for instance, being dropped is turned into be being dropped, it can be grounded in the speech event, e.g., is being dropped. Can atemporalized clausal heads effectively figure as head of the clause, and which is their semantic import? Like clausal heads that have been grounded in the speech event, atemporalized clausal heads can serve the function of head of the clause. As with finite clausal heads, a varying number of other clausal components (e.g., Subject, Object, Complement) tends to accompany them. In the non-finite clause in (21), for instance, the apples functions as Subject, and the ¿y-phrase realizes the Agent of the process. In (22), then, the atemporal clausal head is accompanied by an adverbial {conscientiously) and an Object (the woods he loved).

From process type specification to clause 75 (21) Those who ate the apples ran off wild and mad into the forest - the apples having been poisoned by a woman who had loved Merlin .... (CB) (22) There his father ... lived the life of a recluse, conscientiously tending the woods he loved. (CB) Observe that the clauses in which atemporal clausal heads function are necessarily subordinate, i.e. they cannot stand alone as independent clauses. I will come back to this briefly in Section 2.3.3. As to their semantic import, I hold that atemporal clausal heads basically represent a different way of looking at the process type specification. They do not ground it or instantiate it, but preserve its status as type specification. The change of processual structures such as play or have been playing into the atemporal structures playing or having been playing, for instance, does not instantiate them.21 It does, however, change the way in which the component states of the verb type are scanned: rather than the step-by-step, serial way of scanning which is typical of processual structures (such as kill, have killed, be killing) and finite clauses, atemporalized clausal heads offer a holistic view and activate the component states of a verb cumulatively, through summary scanning (as defined in Langacker 1991: 554). They thus present the component states of a verb as "a single gestalt" (Langacker 1991: 21). This is true for both atemporal clausal heads in -ing and for those that take a io-infinitive. It is beyond the scope of the present study to examine the precise semantic differences that exist between -ing and to-infinitives as they function in clausal heads. It is important, though, to point out that the -ing of an atemporal clausal head should not be confused with progressive -ing. Progressive -ing has been characterized as "restricting its profile to a series of component states that does not include the initial and final states" and as rendering a verb "imperfective" in that its component states "are construed at a level of schematicity that neutralizes their differences" (Langacker 1991: 209). Verbs that are clearly imperfective cannot therefore take progressive -ing. As pointed out before, however, the -ing of atemporal clausal heads can be used with perfective as well as imperfective verbs (e.g., knowing the truth, having been there) and the clausal head which it forms can itself contain a progressive (e.g., having been washing; having been playing in the garden). As Langacker has moreover pointed out to me, an atemporal clausal head such as eating that whole chocolate cake in [Eating that whole chocolate cake] was a big mistake does not take an internal perspective on the atemporalized process (i.e. one that excludes its endpoints). Rather, the atemporal clausal head designates the entire event of eating the cake. In

76

The functional organization of nominal and clause

that respect, the -ing of atemporal clausal heads also differs from constructions in -ing that follow verbs of perception (e.g., We saw [the ship sinking], We heard [him yelling]) in which -ing can be attributed the same value that it has in the progressive (i.e. it excludes the endpoints of the event and offers an 'inside perspective') (Langacker 1991: 443). To conclude, then, the process type specification which forms the core of the English clause is either grounded in the speech event, or it is excluded from grounding by being explicitly atemporalized. In the latter case, the process type specification remains ungrounded. Atemporal or ungrounded clausal heads need not, however, remain uninstantiated: in what follows, I will argue that atemporal clausal heads, like all clausal heads, are instantiated by being combined with a Subject. Crucially, the role of the clausal Subject should be distinguished from that of the Object, which is primarily type-specifying: the clausal Object forms part of the clausal head, which the Subject serves to instantiate. I will first elaborate on the typespecifying role of the clausal Object (Section 2.1.2), and will only in Section 2.2. turn to the instantiating properties of the Subject.

2.1.2. The clausal head, Objects and Complements In Langacker's functional analysis of the clause, it is the verb group minus the markers of tense and modality which is presented as clausal type specification or head. The inclusion of the participants that are involved in the process and realized as Subject or Object of the clause is argued to turn the process type specification into an instance (Langacker 1991: 34). Davidse (1997: 420) has proposed an alternative view according to which the type specification of the clause also includes "any Objects, Complements, and circumstantial Adjuncts that may be present in the clause". Instead of serving the role of instantiator, the Objects, Complement and Adjuncts merely 'subcategorize' the process type expressed by the verb stem "into more specific process types" (Davidse 1997: 422). This claim builds on Halliday's (1994) observation that a fundamental structural split seems to exist between the Subject and the finite verb (and polarity), on the one hand, and the rest of the clause, on the other. Halliday views this split as being essentially interpersonally motivated, i.e. it divides the clause into two main parts according to the function which they serve in turning the representational content of the clause into a message that can be exchanged in communication. In Halliday (1994)'s interpersonal analysis of the clause, the Subject-Finite complex is referred to as the 'Mood element', and the rest of the clause is labelled the 'Residue'. Semantically

From process type specification to clause 77 speaking, the Mood is argued to relate what is expressed in the Residue to the context of the here-and-now of the speech event (Halliday 1994).22 Davidse (1997, p.c.) brings together Halliday's construct of 'Residue' with Langacker's functional analysis of the clause and views the Residue as realizing the clausal process type specification: the Objects, Complement and circumstantial Adjuncts that belong to the Residue, she argues, serve to 'subcategorize' the general process type expressed by the verb group minus the grounding elements. Or, as Davies (1979: 64) puts it, a clause minus its grounding devices is a "complex name" or a description, to which the grounding devices can be attached. In his study of the Subject-Predicate relation, Strawson (1974) observes that the Subject serves to 'individuate' the non-particulars that are expressed by the Predicate: the Subject gives "an individually distinguishing specification of a particular situation of a certain general type" (Strawson 1974: 95). In Davidse's view, the Residue (or clausal head) thus contains not only the main verb and its nongrounding auxiliaries (as argued in Langacker 1991), but also the Objects, Complements and Adjuncts that function in the clause. In the examples in (23) and (24), the clausal head is in italics: (23) She cleaned the room. (24) She must have gone to school. The fundamental structural and semantic distinction which appears to be at work within the structure of the clause between the Subject and the grounding predications of tense and modality, on the one hand, and the nongrounding auxiliaries, the main verb and the Objects and Complements that follow it, on the other hand can be represented as follows (the analysis that is given is based on the clauses I am writing you a letter and You 're not being a pest, are you?)·.

Subject

Mood element Finite Polarity

You

1 P·, pres.tense 2 p., pres.tense

Residue 10 DO Complement

TYPE SPECIFICATION , ... you a be writing ^

st

/

Predicator

not

, , . be being

Figure 6. The Residue as clause type specification

a pest, , (are you?)

78

The functional organization of nominal and clause

We have thus far identified the semantic function of the Residue as being that of clausal head, providing the type specification within the clause. In the next section, I will go more deeply into the question of how instantiation and quantification of the clausal head are realized.

2.2. Instantiation and quantification of the clausal head Broadly speaking, the instantiation of a process type is in many ways similar to instantiation in the nominal, as described in Langacker (1991). Like the instantiation of a noun type specification, instantiation of a process type implies that one conceives the domain of instantiation (in this case time) as having sufficient expanse to support the manifestation of multiple instances and that the profiled processual instance is situated at a specific location which is distinct from other possible locations. As in the nominal, instantiation of a process type always results in one instance, the size of which can be further specified through quantification, while its relation to the speech event can be established through grounding. As pointed out by Langacker (1991), the basic functional schema which clauses realize thus does not differ fundamentally from that of the nominal. With Davidse (1992a, 1997), I hold that it is by being tied to a particular Subject that the type specification expressed in the clausal head or Residue is turned into the conception of an instance.23 Because typically, instances are also grounded in the speech event, however, it is difficult to find evidence in favour of the claim that it is actually the Subject which fulfils the function of instantiator of the clause. As pointed out in Davidse (1991: 364), the best place to observe the precise function of the Subject in English is in non-finite, nominalized clauses, where the grounding effect of the Finite is cancelled. Consider the following examples: (25) Writing a letter is something that can be done in half an hour. (26) Your having written us a letter made all the difference. The non-finite structure in (25) merely represents a type specification, which is not tied to a particular instance. The structure in (26), in contrast, has been linked up with a Subject (which is periphrastically realized by your) and represents an instance of the type specification. In other words, by adding a Subject to an otherwise ungrounded clausal type specification, the latter is turned into the conception of an instance. Notice that the instantiated nature of the clausal type specification in (26) is not only made clear by the presence of a Subject: it is also reflected in the 'secondary' tense

From process type specification to clause 79 having written which is marked on the clausal type itself. This type of secondary tense in the non-finite verb group is not deictic, i.e. it does not itself anchor the clausal type in the here-and-now of the speech event (Halliday 1994: 204). It does, however, imply grounding (or primary tense), which, in turn, presupposes that the clausal type specification is viewed as being instantiated. It can be observed here that the claim that the Subject has special status with respect to the instantiation of the clause fits in with the Cognitive Grammar analysis of the Subject as 'primary focal participant' (Langacker 1991: 321) and that of 'initial reference point accessed in arriving at the processual conception' (Langacker p.c.). As pointed out to me by Langacker, it also meshes nicely with other ideas that have emerged from research in Cognitive Grammar. It reveals, for instance, an interesting parallel between the instantiation of the clausal process type and that of a nominal type with possessor: in both of them instantiation is effected by the element that is characterized as the primary reference point (i.e. the Subject and the possessor respectively) (see also Langacker 1993). In my opinion, the Subject in that respect differs fundamentally from Objects or 'secondary' focal participants, which, I argued before (Section 2.1.2), cannot themselves instantiate a type specification (though they can serve to specify in more detail the instance that has been established by the clausal type specification and the Subject). How, then, can the size of the single clausal instance which the Subject establishes through instantiation be further specified, i.e. how are clauses quantified!? Langacker (1991: 197, 421) proposes that it is the nongrounding auxiliaries of aspect (perfect and progressive) that serve to quantify the verb's processual profile: they "indicate, with respect to some temporal reference point, whether the process designated by the content verb has been completed or is still in progress (hence only partially accomplished)" (Langacker 1991: 197). At the same time, however, Langacker argues that these non-grounding auxiliaries serve to derive a higher-order type specification, and he compares their function to that of number in the nominal, where the plural of a noun like pebbles creates a type specification that is distinct from that of pebble·, while pebbles designates a replicate mass type (which belongs to the category of mass nouns), pebble designates a discrete entity and represents a count noun (Langacker 1991: 78). The claim that the non-grounding auxiliaries of aspect derive a higher-order type specification, however, seems to conflict with the idea that they express quantification, which presupposes instantiation (Davidse 1997: 423). On this point, I will thus follow Davidse (1997), who argues that it is not the non-grounding auxiliaries which quantify over a clause's processual

80

The functional organization of nominal and clause

instance, but the system of polarity. Davidse (1997) refers to Davies (1979), who argues that polarity expresses 'occurrence value' ('Ov') and that "One or other Ov (positive/negative) is built into every description which involves a predication" (Davies 1979: 67). When the speaker uses negative polarity, as in My son didn 't sleep yesterday, he or she states that the occurrence value of the instance is 'nil' (Davidse 1997: 423). By using positive polarity, in contrast, the occurrence value of the instance is claimed to be 'at least one', as in My mother called yesterday. As pointed out in Halliday (1994: 88), polarity in between positive and negative is expressed by modality. Epistemic modality of probability realizes the polarity of'"either yes or no', i.e. maybe yes, maybe no, with different degrees of likelihood attached" (Halliday 1994: 89). Epistemic modality of usuality is argued to be equivalent to 'both yes and no', i.e. "sometimes yes, sometimes no, with different degrees of oftenness attached" (Halliday 1994: 89). Deontic modality of obligation, then, realizes 'do' and 'don't' and intermediate degrees; that of willingness expresses 'will' and 'won't' and intermediate degrees. Notice that the polarity that is linked to occurrence value is, like nominal quantification, essentially associated with instantiation and grounding. It should be distinguished from what has traditionally been called 'internal' negation or negation that follows a modal auxiliary and negates the type specification rather than the modality itself (Davidse 1997: 424), as in: (27) You guys must / not have closed the door all the way. (CB) (28) Your friend must/ not have very good eyes. (CB) (29) You may/ not have read in the Constitution about Election Day for the Supreme Court.... (CB) (30) The researchers say the findings suggest women may/ not have equivalent access to these procedures (CB) Importantly, Davidse's proposal to assign a quantifying role to polarity is in line with the claim that instantiation is situated in the Subject and grounding is realized by both Subject and Finite (see Section 2.3). Quantification presupposes instantiation (only the size of an instantiated process type can be further specified) and it is tied up with grounding (only quantified instances can be grounded). Not surprisingly, therefore, quantification is situated within the Subject-Finite or Mood-element of the clause. The relation between the functional analysis of the clause as a 'grounded quantified instance of some type' and the systemic-functional interpersonal analysis of clause structure (as presented earlier in Figure 6) can now be refined as follows:

From process type specification to clause 81 Residue

Mood element Subject

Finite

INSTANTIATION /

You

Predi-

Polarity

cator

QUANTIFICATION ist 1 p., present tense 2 p., present tense

Complement

TYPE SPECIFICATION be writing

not

IO DO

be being

you a letter. a pest, (are you?)

Figure 7. Clausal instantiation and quantification In this view, the clause closely resembles nomináis with common noun head, realizing the functions which it encompasses iconically, with the process type specification as the innermost layer of organization and the functions of instantiation and quantification added to it. The clause can leave it at that and profile a quantified instance of a type which is not grounded in the speech event, witness the occurrence of non-finite clauses with atemporal clausal heads. In Section 2.3,1 will look into quantified instances of a process type that are grounded and I will give an overview of the types of grounding which they incorporate.

2.3. Clausal grounding In many functional descriptions (Hengeveld 1989; Langacker 1991; Halliday 1994), it is held that the clause is grounded or related to the here-andnow of the speech event by the finite element of the verb group. Typically, the finite element is said to ground the clause either through tense, by reference to the 'temporal zero-point' or the time of encoding and/or decoding of the utterance (Declerck 1991a), or by means of modality, i.e. by reference to the judgement of the speaker as to how 'probable' or how 'obligatory' occurrence of the process instance is. In what follows, I will discuss two analyses which refine this basic model of clausal grounding. In Section 2.3.1,1 elaborate on the idea that all tensed finite clausal heads ground the clause in what is basically a modal, epistemic way of grounding (Verstraete 2002). In Section 2.3.2, I discuss a proposal by Davidse (1997) to analyze the clause as being grounded not only in terms of tense/modality, but also in terms of grammatical person or person deixis. The Subject is argued to

82

The functional organization of nominal and clause

contribute to this type of clausal grounding by serving as the objective or explicit realization of grammatical person. It is thus identified in its second main function, in addition to that of instantiator of the clausal process type.

2.3.1. The finite element In his discussion of the grounding predications in the English verb group, Langacker (1991) distinguishes between a schematic characterization of their properties and one in terms of more specific values which the grounding predications may have. At the schematic level, he (1991: 244) argues, "time per se is not a factor" and "the system is purely one of modality" (Langacker 1991: 244). Clauses which have a tensed verbal group but lack a modal auxiliary are, in this perspective, viewed as realizing a schematic type of modality. Langacker (1991: 242fnl) classifies the schematic modality of finite full verbs as being 'epistemic' in nature or related "to what a particular person knows (or thinks he knows)". A similar claim is made by Davies (1979, 2001) and Verstraete (2002). Both argue that the indicative mood (or the verbal group that contains a finite full verb rather than a modal auxiliary) realizes epistemic modality, or encodes - as do the epistemic modal auxiliaries - the speaker's position with respect to the plausibility of the propositional content of the clause. Unlike epistemic modals, however, the indicative does not express probability or possibility. Rather, it expresses the speaker's "commitment to the truth of the proposition" (Verstraete 2002: 58), i.e. the proposition is 'known to be true' (Davies 2001: 230). In an attempt to integrate this schematic epistemic modality of the indicative into the overall distinction between epistemic and deontic modality, Verstraete (2002), basing himself on Halliday (1970a) and Davies (1979), redraws the traditional distinction between epistemic and deontic modals and posits that a fundamental distinction can be made between, on the one hand, various types of epistemic grounding (including the indicative, 'objective' deontic modality and 'dynamic' modality), and, on the other hand, the 'subjective' type of deontic modality, together with the imperative. What precisely does the notion of subjective deontic modality stand for and what distinguishes the subjective type of deontic modality from objective modality? The notions of subjective and objective modality go back to a distinction that was first made in Halliday (1970a). Halliday argues that two main subtypes can be distinguished in the overall system of modality. A first type of modality represents a "form of participation by the speaker

From process type specification to clause 83 in the speech event" in that the speaker "associates with the thesis an indication of its status and validity in his own judgment" (Halliday 1970a: 335). This type of modality is called 'interpersonal' modality or 'modalization'. It comprises the epistemic modal assessments of probability/ possibility/ certainty, as well as those of usuality/ typicality. A second type of modality is called 'modulation' and is argued to be "part of the thesis", i.e. "part of the ideational meaning of the clause" (Halliday 1970a: 336). Under modulation, Halliday includes the modals of volition and ability (grouped together as 'dynamic' modals by Palmer 1990: 2, as cited in Verstraete 2002: 20), as well as instances of deontic modality to do with obligation and permission. In Lyons (1977), the terms 'subjective' and 'objective' were introduced to refer to the interpersonal and ideational types of modality respectively: subjective or interpersonal modality thus encodes the speaker's position with regard to the propositional content of the clause. Objective or ideational modality, in contrast, is internal to the proposition (Verstraete 2001: 1507).24 Halliday (1970a) and Verstraete (2001, 2002) suggest that a subjective type of modality exists in addition to that of epistemic modality: they argue in favour of recognizing a subjective type of deontic modality as well. Examples of subjective deontic modality are given in (31) and (32). The modals in (33) and (34) illustrate objective deontic modality (based on Verstraete 2002: 38): (31) Almost daily, I was told No, Clarkson, you are a fool and you may not wear training shoes for school. (CB) (32) For taking my sister's life, and all the other lives, too, she must be locked up forever. (CB) (33) But Ramadan means more than just physical deprivation. It has spiritual and moral obligations, too. Followers must refrain from bad thoughts, words and actions .... (CB) (34) As you know, Mr. Karmal, a king must always return a gift with something of similar value. (CB) In the clauses in (31) and (32), the modal serves to encode a particular position of the speaker with regard to the propositional content. The only difference between this type of deontic modality and epistemic modality concerns "the relevant domain for the position", i.e. "deontic modality encodes positions concerning the desirability of actions, whereas epistemic modality encodes positions about the plausibility of propositions (Verstraete 2002: 40). The examples in (33) and (34), in contrast, do not encode a particular position of the speaker, but they "report on the existence of a particular ob-

84

The functional organization of nominal and clause

ligation without necessarily committing the speaker to it" (Verstraete 2002: 38). The speaker merely describes the obligation, but is not necessarily the source of it. As such, the modals in these examples are objective or ideational in nature and closely related to the dynamic modalities of willingness and ability. With the existence of a subjective type of deontic modality established, the system of modality in English can be schematized as follows: Table 2. The system of modality in English subjective or interpersonal modality/modalization -epistemic modality -probability, possibility, certainty -usuality, typicality -deontic modality -obligation, permission25 objective or ideational modality/ modulation -dynamic modality -willingness, ability -deontic modality -obligation, permission26

When the various types of subjective and objective modals are analyzed in terms of whether or not they establish epistemic grounding, another picture emerges and the main distinction turns out to be that between subjective deontic modals (and imperatives), on the one hand, and all the other types of modality and the indicative, on the other: Table 3. Epistemic vs. non-epistemic grounding Epistemic grounding:

Non-epistemic grounding:

-indicative mood -epistemic modal auxiliaries -'objective' deontic modality -'dynamic' modality -imperative -'subjective' deontic modality

Epistemic grounding turns out to be linked to tensed verbal groups, to all those modality types that can realize tense (i.e. those of objective deontic modality and dynamic modality), and, of course, to epistemic modal auxiliaries. Non-epistemically grounded are the imperative and the subjective type of deontic modality, both of which are tenseless. It can be observed that Verstraete (2002) thus analyzes clauses with objective modality (of the deontic or the dynamic type) as establishing epistemic grounding.27 An important indication of the presence of epistemic modality in clauses with objective modality is the presence of tense marking in them. Unlike subjec-

From process type specification to clause 85 tive deontic modals, objective deontic modality does not encode the speaker's commitment to the obligation, but merely describes the existence of such an obligation. And this can be located in past, present or future, just like any other proposition (Verstraete 2002: 63): (35) In Bangalore, computer companies have to/ had to/ will have to do privately what they cannot depend on the government to do. The same counts for clauses with dynamic modality, which are also tensed (Verstraete 2002: 65): (36) He could write, he could play any instrument.... (CB) Significantly, the presence of tense in clauses such as those in (35) and (36) distinguishes them from clauses with subjective deontic modality: subjective deontic modals apply to tenseless or virtual domains. The state of affairs that combines with subjective deontic modals is desired and, consequently, purely virtual. To conclude, the finite element does not simply ground the clause either by tense or by means of modality: at a schematic level, all clauses that realize tense can be analyzed as being modal in nature. They establish epistemic modality and express the speaker's position with regard to the plausibility of the prepositional content of the clause. As Davies (2001) argues, the semantics of indicative mood and of epistemic modality can thus be treated as an integrated continuum. Interestingly, the objective type of deontic modality and the dynamic modalities of ability and volition also seem to establish epistemic modal grounding and thus form part of the epistemic continuum as well. It is only the subjective type of deontic modality and the imperative which do not imply any epistemic position on the part of the speaker about the plausibility of the proposition. Instead, they express the speaker's desire or wish and therefore deal with states of affairs which are purely virtual and tenseless.

2.3.2. Person deixis Davidse (1997) proposes that, in addition to being grounded by the Finite, the clause is also grounded by 'person deixis': the process instance is not only located with respect to the ground by being related to the speech event in terms of modality and tense, but also by being tied to 'grammatical person', which may be either the speaker/encoder, the hearer/decoder, or some

86

The functional organization of nominal and clause

third person (Davidse 1997: 425). It is obvious that first and second person refer to the ground: they refer to the speech participants who are primary constituents of the ground. With regard to third person, note the unique referential status of the Subject which is pointed out by Keenan as a 'truly universal' property of Subjects: the Subject's reference "must be determinable by the addressee at the moment of utterance. It cannot be made to depend on the reference of other NPs which follow it" (Keenan 1976: 313). The structure * He-self admires John cannot therefore replace John admires himself. A third-person Subject is thus the non-speaker/hearer referent that is most immediately accessible for identification by speaker and hearer. How is the system of person deixis encoded in the clause? Davidse argues that it is coded subjectively or implicitly (Langacker 1991: 93) by the Finite, i.e. through person and number marking, and objectively or explicitly by the Subject.28 The Subject thus makes explicit the grammatical person which is marked subjectively on the Finite.29 Notice that there is no person and number marking on non-finite or atemporal clausal heads: it seems logical that atemporal clausal heads, which do not ground the process type temporally or modally, also fail to realize the person deixis that is inextricably linked to it. As to the Subject function in non-finite clauses, then, I suggest that it is reduced to its instantiating role. It is only when a Subject functions in a finite clause that it co-operates in the grounding of the clause by establishing objective person deixis. Whereas the nominal may be grounded either in terms of spatial proximity (by means of demonstratives) or in terms of person deixis (via possessives), the clause is thus grounded both in terms of temporal/modal proximity and person deixis (Davidse 1997: 424).

2.3.3. Clausal grounding: Conclusion Summarizing, in my description of the various systems that operate in language to relate quantified instances of a process type to the ground, I have highlighted two issues: first, every finite clause encodes, in one way or another, the speaker's position towards the proposition expressed by the instantiated process type. Every grounded clause thus realizes a schematic modal feature, and this may or may not be combined with an indication of tense. Secondly, grounding is more than the expression of temporal/modal proximity: finite clauses also identify the instance which they ground with respect to the speaker and hearer of the speech event, and they encode this by means of person deixis or grammatical person, marked on the finite verb and the Subject. The overview of the functional organization of the clause

From process type specification to clause 87 can now be completed, with grounding located both in the Subject and in the finite element of the clause: Mood element Subject

Finite

INSTANTIATION

/ You

Polarity QUANTIFICATION

GROUNDING 1 p., pres.tense 2 p., pres.tense

Predicator

TYPE SPECIFICATION

be writing not

Residue IO CompleDO ment

you a letter.

be being

a pest, (are you?)

Figure 8. The role of the Subject and the Finite in clausal grounding The Subject-Finite or Mood-element of the clause does not only serve to ground the clausal type specification. According to Halliday (1994), the Mood-element also defines the speech functional value of the clause. It expresses the "illocutionary force of an utterance" or "its function as a speech act" (McGregor 1997: 213). This illocutionary force is closely tied up with the grounding function of the Mood-element: non-finite clauses cannot therefore occur as independent speech acts in the discourse. Verstraete (2002) points out that this is due to the fact that non-finites do not have grounding that is expressive of the position of the speaker: "Speaker/interlocutor-positioning is a functional prerequisite for a propositional content to be put forward as a speech act in discourse: what functions in discourse is not bare descriptions of SoAs, but speakers' and interlocutors' positions with respect to these SoAs" (Verstraete 2002: 79). The speech functional value of finite clauses identifies a third function of the Subject, viz. that of 'speech functionally responsible' element (Halliday 1994). In the following and last section, I will briefly elaborate on the speech functional role of the Subject.

2.4. The speech-functional role of the Subject I have thus far shown that a clause is always grounded by person deixis and in terms of modality: when the clause realizes epistemic modality, it en-

88

Thefiinctionalorganization of nominal and clause

codes the speaker's position with regard to the plausibility of the proposition; subjective deontic modality, in contrast, expresses the speaker's position with respect to the desirability of the proposition. For a clause to function as a fixll speech act, however, it has to provide an indication as to who takes responsibility for the modal position that is expressed in it. This indication is provided by the basic clause types declarative, interrogative and imperative: together with the modal positions that can be taken in the clause, these clause types serve to define the speech function of the clause (Verstraete 2002: 83). In what follows, I will first briefly discuss these speech functions and show how they indicate whether it is the speaker or the hearer that is responsible for the modal position that is implied in the clause. My discussion of the speech functions is based on Davies (1979) and, in particular, on Verstraete (2002: 83-106). In a second part, I will zoom in on the speech functional role of the Subject. My analysis of the role of the Subject in the clause as speech act builds on Halliday (1994) and Davidse (1997). Declarative and interrogative structures serve to encode alternative assignments of responsibility for the modal position which they imply: while the declarative assigns responsibility to the speaker in his or her own turn, the interrogative transfers responsibility to the hearer in the next turn. As Davies (1979: 51) puts it, in an interrogative, the speaker assigns to the addressee "the role of subsequent teller in so far as the speaker's use of this construction incorporates a demand for reply". In the declarative examples in (37) and (39), the speaker thus takes the responsibility in his or her own turn and is committed to the assessment of possibility realized by might and the assessment of certainty implied in the indicative in (39). The interrogatives in (38) and (40), on the other hand, signal that the speaker transfers the responsibility to the hearer in the next turn. Verstraete (2002: 85) observes that this does not imply that the hearer will take this position: "rather, transfer of responsibility implies that it is up to the interlocutor to accept, reject or modify the modal position which the speaker has chosen as the starting point for the exchange" (the examples are based on Verstraete 2002: 85): (37) Gene therapy might very well become a major new revolution in medicine. (CB) (38) Might it be of some value? (CB) (39) Then he got us involved in Iran-Contra. (CB) (40) Were the police involved in that? (CB)

From process type specification to clause 89 Note that in imperatives, as in declaratives, it is the speaker that takes responsibility for the modal position: the speaker expresses the desirability of a certain action and takes responsibility for it: (41) Leave this room at once! The same difference between declarative and interrogative can be observed when the position that is encoded in the clause is one of subjective deontic modality. The declarative in (42) thus signals that the speaker takes responsibility for the position of permission that is encoded by may, whereas the interrogative in (43) signals transfer of the responsibility to the hearer. Note that, again, this transfer does not imply that the hearer will take the position. It only implies that the hearer will decide about the issue of permission raised by the speaker: (42) You may return to your duties, Colonel Pople. (CB) (43) May we sit down? (CB) What about the role of the Subject in declaratives, interrogatives and imperatives? Halliday (1994: 76) views the Subject as the 'modally responsible element' or the element that is "so to speak, being held responsible responsible for the functioning of the clause as an interactive event". Now that we have distinguished between two basic modal positions (i.e. that related to the plausibility or truth of the proposition and that related to the desirability of the proposition), Halliday's analysis of the Subject as 'modally responsible' can be understood as follows. The Subject is not responsible in the way that the speaker or hearer is. It is not responsible for the modal position of plausibility/truth/desirability that is implied in the speech act: it is the speaker who assumes a particular position and the hearer who is assigned a position. Instead, the Subject is the entity which the speaker holds responsible for what the modal position involves, i.e. the Subject is presented as being 'responsible' for the truth or the plausibility of an epistemically grounded clause, and for the success of the obligation/desire in a subjective deontic clause and an imperative. The precise nature of the speech functional role of the Subject thus varies according to the type of schematic modality that is realized: in clauses that realize epistemic modality, the Subject is the entity in which the plausibility or truth of the proposition is vested: (44) a. He gave the book to Mary, b. Mary was given the book.

90

The functional organization of nominal and clause

c. The book was given to Mary. (45) a. Did he give the book to Maiy? b. Was Mary given the book? c. Was the book given to Mary? In the declaratives in (44), it is the speaker who assumes responsibility for the epistemic modal position of truth, but it is the Subject in which the truth of the proposition is vested: whether or not the proposition is actually true is presented as being dependent on the Ae-person in (44a), on Mary in (44b) and on the book in (44c). If one wants to find out whether the propositions in these examples are true, the Subjects indicate who or what to turn to: one can ask the he in (44a), one can turn to Maty and check with her in (44b), and one can try to locate the position of the book in (44c). The interrogatives in (45) are in this respect similar to the declaratives: they also present their Subjects as the entities on which the truth of the proposition depends. Different from the declaratives in (44), however, they transfer the responsibility for the modal position of truth to the hearer, and thereby also ask the hearer to decide on whether the Subjects support the truth of the proposition. In clauses with objective deontic modality (which imply epistemic modality and belong to the epistemic continuum), the Subject is responsible for the existence of the obligation. Consider, for instance, the following examples (based on Davidse 1997): (46) a. He has to tell mother the truth. b. The truth has to be told to mother. c. Mother has to be told the truth. In (46a), the responsibility for the existence of the obligation is attributed to the Ae-Subject: the Subject is not only the Agent who will have to do the 'telling', the Subject is also held responsible for the fact that an obligation exists in the first place. The Ae-Subject may, for instance, be identified as the one who has not told the truth to mother yet, as a result of which he is now responsible for the fact that the obligation of telling her exists. In (46b) and (46c), by contrast, the obligation is made to depend on characteristics of the truth and of mother. There may, for instance, be "some moral imperative associated with 'the truth'", or the speaker's respect for mother may explain the existence of the obligation in (46c) (Davidse 1997: 427). In clauses with subjective deontic modality, then, as well as in imperatives, the modal position which is assumed by the speaker relates to the desirability of the proposition. Because no epistemic modality is involved, the

From process type specification to clause

91

Subject is not so much held responsible for the existence of the obligation, but for carrying out the desired process: it depends on the Subject whether the command will actually be carried out and will thus be successful. In active clauses, the Subject is held responsible for carrying out the desired process (as shown in 47a and 47b): (47) a. (You) be quiet! b. You must leave at once! Halliday's notion of responsibility can, however, also be recognized in passive imperatives and with passive subjective deontic modality. In (48a), the Subject has been dissociated from the Agent of the process (i.e. the one who invites), but is still responsible for the success of the command (Halliday 1994: 76). In (48b), the Subject is responsible for the success of the obligation as well, though she may not willingly co-operate in it. It is, nonetheless, the Subject that is assigned a central role in the obligation: it does not matter who does the locking up, as long as it leads to her being locked up: the success of the desired process is made to depend on the Subject ( Ί will not rest until she has been effectively locked up')· (48) a. (you) Get yourself invited to their meeting, will you? b. She must be locked up forever for what she did to me. Observe the difference between (48b) and its objective deontic alternative She has to be locked up forever·, in the latter, the sAe-Subject is held responsible for the existence of the obligation: she has committed a crime, and therefore the obligation of locking her up exists. Like the analysis of the Subject as clausal 'instantiator', its characterization as 'modally responsible' element in the clause is reconcilable with the Cognitive Grammar analysis of Subjects as 'primary focal participants' (Langacker p.c.). Modal responsibility is assigned to the Subject-entity by the speaker and therefore resides in the speaker's conceptualization of the scene (Langacker 1990b). Like the notion of 'primary focal participant', that of 'modally responsible element' is thus essentially a subjective category. It manages to specify the precise nature of the speaker's conceptualization in more detail by linking it to the modal position (of plausibility, truth or desirability) which the speaker takes in the clause. Importantly, the subjective characterization of Subjects in terms of 'responsibility' is schematic and generalizes across the range of Subjects that can be found in the clause. Subjects can be said to differ, however, as to whether they are also 'objectively responsible', i.e. as to whether there is an objective basis for

92

The functional organization of nominal and clause

selecting them as modally responsible elements (Langacker p.c.): while in the examples in (44) to (48), some objective motivation for conferring focal prominence or modal responsibility to the Subject-entity was identified, this is not possible for an example such as It is obvious that Belgium is a great place for beer drinkers·, it cannot be said to objectively motivate the speaker's choice of construing it as Subject. I therefore agree with Langacker (p.c.) when he argues that the objective situation described may offer varying degrees of motivation for conferring modal responsibility to an element, but that it is ultimately the speaker who chooses which element will function as Subject. In short, the Subject-function serves to instantiate the process type expressed by the clausal head; it co-operates in grounding instantiated types by relating them to the system of person; and, finally, the Subject is speech functionally responsible: it is the entity which the speaker picks out as being responsible for the plausibility/truth of the proposition or for the success of the obligation.

2.5. The functional organization of nominal and clause: Conclusion The description of the internal functional organization of nomináis and clauses in terms of Langacker's functions of type specification, instantiation, quantification and grounding reveals that, at a highly schematic level, nomináis and finite clauses realize the same functional schema: both represent grounded instances of a type (as argued in Langacker 1991). Importantly, these functions enable us to generalize over all instances of the class of the nominal, as well as over all finite clauses, irrespective of whether the functions are discernible (explicitly coded) or not. They thus show that the classes of the nominal and the clause can be defined not only in terms of their external functions and their intrinsic properties, but on the basis of their internal organization as well. Interestingly, when Halliday's metafunctions are brought into the picture, some of the differences between the nominal and the clause show up. Firstly, in nomináis the ideational or representational content is located in the nominal head, which "supplies a vast amount of conceptual content" (Langacker 1991: 143). Whereas the noun type specification and its modifiers serve to characterize the designated entity, the grounding predications in nomináis are "wholly extrinsic" and do not pertain to "properties inherent in the entity itself' (Langacker 1991: 143). Clauses, by contrast, do not show compartmentalization of the representational content which they express, but rather have their representational content as a strand of meaning

From process type specification to clause 93 that is realized throughout the clause (Halliday 1994: 179). The representational functions in the clause are thus not located in the clausal head only (even though the clausal head does provide the basic categorization of experience in the clause): representational functions are mapped onto the Subject as well. Halliday (1994: 35) argues that the textual metafunction in the clause, which relates the clause to the larger discourse context, is expressed by means of "peaks of prominence located at beginnings and endings": the textual component in the clause expresses the textual status of parts of the discourse by assigning them to the boundaries of the clause. There is thus a special significance attached to 'coming first' and 'coming last' (Halliday 1979: 69). In the structure of the nominal (which is less flexible in terms of word order), by contrast, textual functions are primarily located in the determiner: determiners can thus relate the nominal to the discourse context or to an immediate referent "within the text" (Halliday and Hasan 1976: 33). Determiners, in other words, establish a direct link with the speech event and its participants or they have an immediate referent in the discourse or text. Finally, as to the interpersonal metafunction, it was shown that Halliday's (interpersonal) Mood-Residue analysis of the clause matches Langacker's functional description of the clause quite closely. While the Residue expresses the clausal type specification, the Mood-element turned out to realize instantiation, quantification and grounding. Not surprisingly, therefore, the interpersonal differences between the clause and the nominal - which also realizes a 'grounded instance of a type' - are smaller than are those in terms of the representational and the textual metafunctions. Interpersonally speaking, both the nominal and the clause consist of a component that is ungrounded and combines with grounding elements that relate it to the speech event, thus turning it into something that can be exchanged in communication. In the clause, it is the Mood-element consisting of the Subject-Finite link that is most essential to grounding and realizes instantiation and quantification; in the nominal, it is determiners (in the broad sense) that assume the functions of instantiation, quantification and grounding. Now that we have gained insight into the functional parallels and differences that exist between nomináis and clauses, we can turn to nominalization again and try to establish the impact of the functional organization of the nominal and the clause on the process of nominalization.

94

The functional organization of nominal and clause

3. Nominalization as functional reclassification I argued in Chapter 3 (Section 3.2) that one of the main problems related to the traditional analyses of nominalization is that they fail to see nominalizations as constructions in their own right, i.e. as constructions which combine clausal and nominal categories in a unique way but which are essentially nominal in status and function as such in larger configurations. The nominal character of nominalizations derived from a verb stem is not controversial. That of nominalizations containing clause-like units, as in [Being able to play the piano] has always been a dream of mine has eluded linguists more and has not been accurately described yet. The nominal features of the latter type of nominalization have often failed to be recognized because they tend to be encoded non-prototypically, i.e. as non-discernible functional components. I argued in Chapter 3 that for the internal analysis of nominalizations, a functional account of the nominal class is needed which is sufficiently schematic to cover all nominal constructions, prototypical as well as non-prototypical. As I have shown in Section 1 of this chapter, the functional analysis of the nominal in terms of type specification, instantiation, quantification and grounding can provide such an account and thus makes an analysis of the nominal status of the various types of nominalized constructions - including those that contain clause-like units - possible. In addition to the nominal characteristics of nominalized constructions, the clausal categories which accompany the clausal level of assembly that forms the starting point of nominalizations have to be accounted for as well. The schematic characterization of the clause which I have set out in this chapter (Section 2) allows us to define the distinct levels of assembly which nominalizations can start from more accurately (as first suggested in Langacker 1991): nominalization either applies to a simple process type specification or verb stem, to what I have defined as an atemporal clausal head or to a fully grounded (or finite) clause. I also pointed out in Chapter 3 (Section 5) that nominalizations should not be reduced to being clause-like constructions, but should be analyzed as being systematically related to clausal constructions through agnation. More specifically, I argued that relationships of agnation between nominalizations and clausal constructions can shed light on the clausal categories that play a role in nominalization. The functional model of the clause and, in particular, the description of the types of grounding that are found in it and of the various functions served by the Subject, I will show, form an excellent tool for the description of the clausal agnates of nominalized structures.

Nominalization as functional reclassification 95 Finally, I posited in Chapter 3 (Section 2) that the literature on nominalization tends to lose sight of the natural symbolic nature of nominalized constructions: nominalizations encode their meaning in their lexicogrammatical properties, and only a careful analysis of these lexicogrammatical features can shed light on the semantics of the nominalized constructions. I have thus far argued that nominalizations are combinations of nominal and clausal functional categories. Crucially, the lexicogrammar of nominalizations is not made up of a random combination of clausal and nominal categories. Rather, nominalizations reclassify a clausal level of assembly into a nominal construction and thereby integrate the functional values attached to the clausal starting point with the functional organization of the nominal structure in which it comes to figure. Importantly, this integration cannot be described from a representational perspective only: it has to be accounted for in terms of a multifunctional analysis which also considers, for instance, its interpersonal properties or the role which each component plays in linking the nominalization to the ground (as argued in Chapter 3, Section 4). Once again, the functions of type specification, instantiation, quantification and grounding prove to be pivotal to such an analysis: one of the most attractive properties of these functions for the description of nominalizations is precisely that they schematically characterize both the clausal and the nominal mode of organization and thus make it possible to come to an integrated analysis of the nominal and clausal contributions to nominalized constructions, and this in terms of one and the same functional framework. I will show that there exist interesting correlations between the clausal level of assembly which a nominalization derives from, the type of reclassification that applies to it (i.e. involving rank shift or not) and the nominal strategy that is adopted (i.e. proper name or common noun strategy). In short, the model of functional organization of the nominal and the clause that has been presented in this chapter provides the descriptive breakthrough that was needed to move beyond the existing analyses of nominalization. It helps to identify the main principles that lie behind the process of nominalization and, together with the theoretical concerns which I raised in Chapters 2 and 3, it provides a descriptive-theoretical framework for the analysis of specific nominalization systems. To shed light on the functional reclassification mechanism which underlies nominalizations and to find out more about the precise nature of the integration which nominalizations establish between nominal and clausal categories, I will now embark on the description of specific nominalization types. I will discuss nominalizations that are based on a simple verb or process type specification (as in teacher and the signing of the contract), nominalizations that derive from an atemporalized clausal head (as in I

96

The functional organization of nominal and clause

don 7 regret [having waited for him]), and nominalizations that take a full finite clause as starting point (as in [That he left her] surprised me). In spite of the fact that they derive from different levels in the functional assembly of the clause, all of these structures are nominalizations, i.e. they have been reclassified and have come to function as nomináis. My analysis is aimed at elucidating the reclassification process that leads from processual starting point to nominalization. In Part Π, I will look at deverbal -er nominalizations. In Part ΙΠ, I will turn to the system of factive nominalization, which includes three lexicogrammatical subtypes: gerundive nomináis (e.g., his being fired), thatnominalizations (e.g., that he was fired) and the fact /Aai-nominalizations (e.g., the fact that he was fired). When discussing gerundive nominalizations, I will also mention gerundive nomináis with an 'action' meaning (e.g., [Firing someone] is a terrible thing to do) and action nominalizations (e.g., the writing of that book).

Part II Deverbal -er nominalization

Chapter 5 Deverbal -er suffixation: Towards a descriptive position

0. Introduction In this chapter, I will discuss the literature on deverbal -er nominalization and single out the two main descriptive themes of my own analysis (presented in Chapter 7). First, as regards the semantic profile of -er nomináis, the question will be raised whether the system can (still) be characterized as agentive. Secondly, the distinction between lexicalized -er nomináis (e.g., destroyer 'warship') and 'ad hoc' nominalizations (e.g., the destroyer of our hopes) will be introduced, together with the question of how to integrate the differences between these types of -er nominalizations in the overall analysis of -er suffixation. As to the profile of deverbal -er nominalizations, it is widely agreed that, diachronically speaking, the -er suffix (including its orthographic variants in -or and -or) is agentive·, the Old English suffix -er(e) primarily served to derive Agents, and then especially personal or human Agents (e.g., bœcere 'baker' from bacan 'to bake', bëâtere 'beater, boxer' from bëâtan 'to beat', cwellere 'killer' from cwellan 'to kill', for more examples, see Marchand 1969; Kastovsky 1971, 1985). The few non-agentive coinings that exist in Old English, Kastovsky (1971: 295) points out, "are more or less restricted to translations from Latin, and thus probably did not belong to the general vocabulary": examples of Latin-based, patientive -er nomináis are scëawere 'buffoon, actor' (from scëawian 'to look at, see', recorded as a gloss for Latin scurra) and scëawere 'mirror' (from scëawian 'to look at, see', a translation of Latin speculea). Designating the Location rather than the Patient of the process of looking is scëawere in the meaning of 'watch-tower' (Kastovsky 1971: 295). The instrumental type of -er nominal, which is so productive in Modern English (e.g., eraser, toaster, screwdriver, nutcracker), is hardly found among Old English -er nomináis: the only example of an Old English instrumental -er nominalization found by Kastovsky is pünere 'pestle' from pünian 'to pound' (Kastovsky 1971: 295). Nowadays, suffixation of a verb with -er still prototypically results in a nominalization of which the referent is the Agent of the base process (hence the label nomina agentis; see, among others, Jespersen 1914-1929,

100

Deverbal -er suffixation: Towards a descriptive position

6; Lees 1960; Marchand 1969; Quirk et al. 1985). Apart from human Agents, -er nomináis also designate animate Agents or animals (e.g., retriever, wood-pecker) and immaterial Agents (e.g., pointer, reminder). Instrumental nomináis like scraper, scratcher, toaster, transmitter have been argued to come close to being material Agents (see, for instance, the -er entry in the Oxford English Dictionary and Marchand 1969). However, Modern English has also seen the emergence of many different types of non-agentive deverbal -er nominalizations. The non-agentives that are most often identified in the literature are those based on verbs of cooking (as in 1), those denoting items of clothing (as in 2), locative nouns (as in 3), and, finally, nomináis designating a variety of other non-agentive entities (as in 4):30 broiler roaster steamer (2) jumper sneakers slippers (3) sleeper

'a young chicken suitable for broiling' 'a kind of meat suitable for roasting' 'an edible clam' 'a type of pullover' 'casual shoes with rubber soles' 'a kind of shoe that can be slipped into' 'a train you can sleep in'; 'a bed or sofa in a train, in which you can sleep' diner 'a place where you can dine' (4) bestseller 'a book or other item that sells well' dipper 'a vegetable/ fruit/ other snack that has to be dipped before being eaten' scratcher 'a lottery ticket that has to be scratched to reveal the potentially winning patterns' (1)

The issue that needs to be tackled here is whether the Modern English -er suffix can still justifiably be analyzed as agentive, or whether another characterization has to be looked for which describes the present distributional properties of the suffix more accurately. Put differently, is agentivity still the highest-level schema which deverbal -er nomináis instantiate, or do we have to look for another schema? And if so, what does this alternative high-level schema look like? Not only the occurrence of non-agentive formations has to be accounted for. Deverbal -er nominalizations also differ according to the degree to which they are entrenched·. I referred earlier to the occurrence of 'ad hoc' nominalizations (as in 5) and lexicalized -er nomináis (as in 6):31

-er nominalizations: Agent names or Subject names ? 101 (5) Alex Gough, the Welsh destroyer of Del Harris's hopes on Tuesday ....(CB) (6) destroyer 'a small, heavily armed warship' Lexicalizations are included in the language system as 'fixed expressions' (Langacker 1991: 45) and name what are basically types of things or people. Ad hoc nominalizations, in contrast, represent -er formations which are derived on the basis of the constructional schema of -er nominalization to serve a specific discourse purpose. These two types are characterized by differences in degree of entrenchment and specificity, as well as by a number of semantic and structural differences. It has, for instance, been observed that ad hoc nominalizations tend to be postmodified by an o/-phrase (i.e. of Del Harris's hopes) realizing what would clausally have been a participant in the process. Lexicalizations typically do not take this kind of o/-phrase. In what follows, I will go through the analyses of -er nominalization that have been proposed thus far. I will first look at descriptions of the semantic profile of -er nomináis (Section 1) and then at analyses of the distinction between lexicalizations and ad hoc nominalizations (Section 2). Characteristic of the analyses of the profile of -er nominalization is that they view all -er nomináis as designating what is basically a semantic role (Ryder 1991, 1999a, 1999b), an argument structure configuration (Levin and Rappaport 1988; Mackenzie 1990; Rappaport Hovav and Levin 1992) or a complex conceptual category (Panther and Thornburg 2002). The label of 'Subject name', I will show, has thus far been brought into the discussion of -er nominalization only occasionally and without reflection on the actual value and use of the Subject in clausal contexts (e.g., Marchand 1969; Kastovsky 1971; Bauer 1983; Booij 1986, 1992). As for the differences between lexicalized and ad hoc -er nominalizations, I will point out that they have been repeatedly discussed in the literature (e.g., Jespersen 1914-1929, 6; Strang 1968, 1969; Levin and Rappaport 1988; Mackenzie 1990; Rappaport Hovav and Levin 1992), but they have not yet been brought together into a coherent analysis, nor has the distinction between lexicalized and ad hoc nominalizations been integrated in the analysis of deverbal -er suffixation in general.

1. -er nominalizations: Agent names or Subject names? Research into the semantic profile of deverbal -er nominalization has always struggled with the system's tolerance of non-agentive -er nomináis.

102 Deverbal -er suffixation: Towards a descriptive position Some of the early accounts of the suffix simply ignore the existence of nonagentive -er nomináis and refer to -er as an agentive suffix (e.g., Oxford English Dictionary; Lees 1960; Quirk et al. 1985). Others acknowledge that -er can also derive non-agentive nomináis, but do not attempt to integrate the non-agentive cases in the system of -er suffixation in general: their account of -er nominalization is restricted to a mere list of possible formations (e.g., Jespersen 1914-1929, 6; Quirk et al. 1972; Kastovsky 1985). Still others point to the disparity between agentive and non-agentive -er and conclude that the non-agentive type is not part of the system of -er nominalization, that it is idiosyncratic or non-systematic in nature (e.g., Keyser andRoeper 1984; for Dutch: Booij 1986; Taeldeman 1990). Linguists who have tried to come to terms with non-agentive -er nominalization are, firstly, Ryder (1991, 1999a, 1999b), who offers an analysis of the system of -er nominalization which focuses on the prototypically agentive status of the designated entity and its salience within the event in which it figures; secondly, Panther and Thornburg (2002), who suggest that -er nominalization forms a complex conceptual category which is centred on an agentive prototype to which a large number of other meanings are linked through metaphor and metonymy; and, thirdly, Levin and Rappaport (1988) and Rappaport Hovav and Levin (1992), who analyze -er nomináis from the perspective of argument structure configurations. In the following sections, I will have a closer look at the descriptive analyses which they propose.

1.1. Agentivity and salience The analysis which Ryder (1991, 1999a, 1999b) proposes to account for the variety of profiles in deverbal -er nominalizations is "semantic/pragmatic" in focus (Ryder 1999b: 291). It analyzes -er nominalization "in terms of semantic case, event structure and prototype reanalysis" (Ryder 1991: 299) and builds on the assumption that semantic roles "are better viewed as prototypes rather than as absolute categories" (Ryder 1991: 300). While some participants will thus "be considered agents by everyone, other, less prototypical ones will be considered agents to the extent that they can be construed as matching or approximating the prototype" (Ryder 1991: 300). Ryder argues that the non-agentive cases of -er can all be regarded as extensions or reanalyses of the agentive -er prototype, reanalysis being possible when a non-agentive entity is conceived as being salient or particularly noticeable within the designated event (Ryder 1991: 309; 1999b: 288-289). Salience is what relates deverbal -er nominalizations to denominal ones

-er nominalizations: Agent names or Subject names ? 103 (e.g., prisoner, weekender) in which "the referent of the -er noun ... should be higher in salience than that of the nominal base" (Ryder 1999b: 290). The schema which is instantiated by all -er nominalizations - deverbal and non-deverbal - can thus be summed up as "a person, animal or thing saliently connected with X" (Ryder 1999a: 307). How did in Ryder's opinion the distinct extensions of the agentive prototype in deverbal -er nominalizations come about? She (1999b: 287) argues that two factors are important here, viz. an entity's salience and identifiability in the event, and, secondly, its location on the causal chain of the event in which it figures (Ryder 1999b: 287). The first and most obvious extensions of Agent -er nomináis are those nomináis that designate Instruments. Instruments can easily be foregrounded as salient, agent-like entities: "because they are independent of the Agent in many ways, the event can be construed as having them as the head of the part of the causal chain that is in focus, and so, like Agents, they are highly salient" (Ryder 1999b: 288). Especially the great proliferation of implements and machines from the 16th century onwards explains the extension of the almost exclusively agentive -er suffix to instrumentais in the early Modern English period (as attested in Dalton-Puffer 1994). Once -er nomináis came to designate Instruments, "clothing intended to be worn while performing central actions in an episode" was construed as similar to an Instrument used in that episode (Ryder 1991: 305). It is the 'identifiability' of clothing items, or the fact that they are "specifically designed for certain activities, and so can be readily identified by those activities" (Ryder 1999b: 290) which explains why various -er nomináis denoting items of clothing have since been coined: e.g., loafers, sneakers, waders, loungers, jumper. Patientive -er nomináis referring to food, then, profile structures in which "the agent is outside the boundaries of the episode, leaving the original patient as the most agent-like element remaining" (Ryder 1991: 309): the nominal steamer 'edible clam', for instance, profiles a clam that steams. Nomináis like broiler, baker and frier are analyzed similarly. Ryder (1991: 309) says they all designate "agent-like active patients". Comparable to the instrumental and clothing types of -er nomináis, "the food is sufficiently salient and ... sufficiently identifiable by the verb" to become the referent of an -er nominalization (Ryder 1999b: 289). The salience of these items of food is attributed to two features in particular: firstly, the processes which they are related to refer to a "specialized type of preparation" (Ryder 1999b: 289), which the animals, vegetables and fruit are bred or selected for; and, secondly, "the food that is cooking, at least in modern kitchens, is often the most prominent participant actually present in the event" (Ryder 1999b: 289). Ryder refers here to the fact that the "per-

104

Deverbal -er suffixation: Towards a descriptive position

son in charge of the meal performs few if any overt actions during the cooking process, and neither does the device in which or by which the food is cooked" (Ryder 1999b: 289). Although originally the referents of the food-related nomináis may have been interpreted as active Patients, Ryder (1991: 309) claims, they were reanalyzed as truly patientive (e.g., steamer as a clam that steams was reanalyzed as a clam that is steamed) and thus paved the way for -er formations "that have few if any agent-like qualities", e.g., scratcher 'a lottery ticket that is scratched to reveal the potentially winning patterns' (Ryder 1991: 310) and dipper 'something that is dipped before being eaten' (Ryder 1991: 309). That the latter, in spite of their lack of agent-like properties, can be profiled by an -er nominalization, Ryder (1999b: 289) argues, is only due to their being sufficiently "identifiable in their event". Ryder's analysis of -er nominalization has the merit of trying to offer an integrated analysis of agentive and non-agentive -er nominalization on the basis of a wealth of data. It includes interesting observations, for instance on the relation between instrumental -er nomináis and the agentive prototype of -er nominalization. Ryder's approach, however, in my opinion, also has its weak points and in the next section, I will point to some of them. In Section 1.1.2 I will discuss Panther and Thornburg's (2002) alternative cognitive proposal for the analysis of -er nominalization and show that it significantly improves on the analysis proposed by Ryder and contains several valuable descriptive observations concerning the semantics of -er formation. In Section 1.1.3, then, I will argue that, as shown in Lemmens (1998), a detailed analysis of the nature of the base verbs of deverbal -er nominalizations is necessary to arrive at an accurate description of their profile.

1.1.1. -er nominalizations as natural symbolic units Ryder argues that non-agentive -er nominalizations are extensions of the agentive prototype which have in common with agentives that they designate entities which are conceived of as being salient and identifiable within the designated event. I consider Ryder's analysis in terms of salience and identifiability problematic in two respects. First, Ryder accounts for nonagentive -er nominalizations in terms of a semantic property that is typical of agentives and she thus suggests that the semantics of non-agentives can be reduced to the vague concept of 'agent-like salience'. Secondly, Ryder tries to shed light on the semantics of -er nominalizations but she ignores their constructional properties.

-er nominalizations: Agent names or Subject names? 105 First, Ryder creates the - in my opinion false - impression that agentive and non-agentive deverbal -er nomináis have blended into one schematic concept of 'agent-like salience'. She repeatedly stresses that salience is very much an agentive property: because the most salient participant in the causal chain is typically the Agent of the process, non-agentive entities that become "the most salient part of their event's causal chain" are said to be agent-like or "behave like Agents" (Ryder 1999b: 288). Ryder argues that this is true of "many Instruments and a few Patients" (Ryder 1999b: 288), such as the cooking-related nomináis. She admits, however, that nomináis such as scratcher, dipper and sipper designate "fairly prototypical patients that have few if any agent-like qualities" (Ryder 1991: 309). The claim that non-agentive and agentive -er nominalizations realize one and the same meaning of '(agent-like) salience' conflicts with the language user's interpretation of non-agentive -er nominalizations: nomináis such as kneeler 'a chair which you have to kneel on' and squeezer 'a container that you have to squeeze to extract something from it' are not interpreted as having an agent-like meaning of salience. Rather, they designate entities that are done to or affected by the process. That non-agentives have to be interpreted as non-agentive rather than as salient, agent-like entities is also revealed by -er nominalizations that constitute clear cases of lexical ambiguity (Taylor 1995). Nominalizations such as reader and baker, for instance, have an agentive and a patientive reading {reader either profiles 'someone who reads' or 'a compilation of literature'; baker designates 'someone who bakes' or 'a food - meat, fruit or vegetable - that is suitable for baking'). Rather than realizing one 'merged' meaning of salience, the agentive and non-agentive readings of these -er nominalizations constitute irreducible senses in their own right. The linguistic issue that should be addressed, therefore, is why one and the same suffix can designate salient entities as divergent as Agents and Patients, while leaving intact their respective agentive and non-agentive readings. The notions of salience and identifiability cannot shed light on this issue, because they fail to address the constructional properties of deverbal -er suffixation. Ryder's analysis is purely semantic in design and does not consider the unique constructional link which -er establishes between an entity and a process. I view -er nomináis as natural symbolic units or units of which the meaning is encoded in the lexicogrammar (see Chapter 3, Section 2). An account of -er nominalizations, in my opinion, therefore has to begin with a careful description of their constructional properties and it will have to distinguish deverbal -er nomináis from -er nomináis with nonverbal bases. Nominalizations with a verbal root and nominalizations with a non-verbal root have a compositional value and a meaning that are neces-

106

Deverbal -er suffixation: Towards a descriptive position

sarily different. Denominal -er nominalizations, moreover, do not integrate verbal and nominal categories into one structure, since they do not involve reclassification (as defined in Chapter 3, Section 3). In terms of Langacker's network model of language, a highly abstract schema which generalizes across all instances of -er nominalization (deverbal as well as others) may well be possible; but deverbal -er nominalizations instantiate a lower-level, less abstract constructional schema, which, among other things, has to do justice to the variety of process types that can be nominalized by -er and the range of participants in these processes that can be profiled. 1.1.2. -er suffixation as a complex conceptual category Like Ryder, Panther and Thornburg (2002) adhere to a purely semantic approach to -er nominalization and reject the possibility of a constructionally-based description of it. Unlike Ryder, however, they do not posit a general abstract meaning for all -er nomináis. Instead they argue that the system of -er suffixation has a central sense to which most other senses of -er nomináis are "more or less directly linked" (Panther and Thornburg 2002: 193). The central sense is that of a "human Agent who performs an action or engages in an activity to the degree that doing so defines a primary occupation" (Panther and Thornburg 2002: 193). Nomináis that designate non-human entities which are perceived to be human in some respect (e.g., retriever, warbler, grasshopper, skyscraper) are metaphorically related to this central sense through personification. Instrumental -er nomináis, Panther and Thornburg suggest, can either be viewed as strong substitutional metonyms for human Agents or (more plausibly) as historical extensions of the -er prototype which are "metonymically motivated by conceptual contiguity" (Panther and Thornburg 2002: 177). -Er nomináis with event referents (e.g., thriller, eye-opener, pageturner), finally, are described as instances of reiflcation. Apart from the fact that it presents what is probably the most elaborate account of -er nominalization thus far, Panther and Thornburg's analysis in my opinion significantly adds to the existing descriptions of -er nominalization in two ways: firstly, it opens up interesting new perspectives for the analysis of non-deverbal -er formation and its relation to deverbal -er\ secondly, it contains some valuable observations with respect to the semantics of instrumental and truly non-agentive -er formations. First, in their extensive overview of deverbal and non-deverbal -er nomináis Panther and Thornburg manage to reveal systematic conceptual

-er nominalizations: Agent names or Subject names? 107 resemblances between denominai and deverbal -er nomináis. They argue that both verb-based and non-verb-based -er nominalizations evoke action scenarios which include activities, participants and a specific setting. Verbbased -er formations evoke such action scenarios directly, i.e. through their verbal base; non-verb-based nomináis evoke them metonymically, by naming a component (e.g., Patient, Instrument, Location, Time) of the action scenario, which comes to function as reference point (Panther and Thornburg 2002: 158). The action must then be metonymically accessed from that reference point. It is especially among prototypically agentive and instrumental -er nomináis that interesting parallels turn out to exist between the nomináis with a verbal base and those with a non-verbal base: nomináis such as tinner, slater, hatter, philosopher, for instance, can be said to resemble deverbal -er formations like baker, programmer and teacher in that they also designate a human who occupationally performs an action, be it an action that is metonymically linked to the nouns on which they are based (i.e. tin, slate, hat, philosophy). Similarly, nomináis such as threewheeler and doubledecker are said to metonymically realize Instruments because their base "names a 'design feature' that evokes a scenario in which the entity with such a design feature is used' (Panther and Thornburg 2002: 170). In short, Panther and Thornburg (2002) show that denominal -er structures can realize metonymically many of the meanings which deverbal -er nominalizations evoke directly. Interestingly, Panther and Thornburg (2002: 173) also point to a number of semantic similarities between instrumental and truly non-agentive deverbal -er nomináis such as broiler, cooker, reader, poster. The latter, they suggest, are "conceptually fairly close to Instruments" because like Instruments they are "purpose-designed entities" (Panther and Thornburg 2002: 173). They are either "designed for a special purpose" or they have "inherent properties that make them suitable for certain purposes" (Panther and Thornburg 2002: 173): a broiler is a chicken that is suitable for broiling and a squeezer is a bottle designed to be squeezed. Panther and Thornburg's observations are, I think, basically correct (the 'purpose-designed' status of most patientive -er nomináis has also been hinted at in Levin and Rappaport 1988 and Lemmens 1998). However, they cannot explain why a prototypically agentive suffix such as -er has come to designate instrumental as well as truly non-agentive entities that are designed or suitable for a specific purpose. The so-called "conceptual closeness" of instrumental and non-agentive -er nomináis does not explain why they make use of one and the same, prototypically agentive suffix to designate two fundamentally different participants, i.e. agent-like and patientive ones. The basic extension from agentive to non-agentive nomináis that

108

Deverbal -er suffixation: Towards a descriptive position

seems to have taken place within the system of deverbal -er suffixation is, in other words, left unaccounted for. hi Chapter 7 I will show that it is only when the compositional value of instrumental and patientive -er nomináis is analyzed more closely that the reason behind their semantic resemblances becomes clear. Summarizing, Panther and Thornburg show that denominal -er nomináis can be profitably examined from the perspective of metonymy and that their semantics resembles that of deverbal -er nomináis to a certain degree. They also make interesting observations concerning the semantics of instrumental and patientive -er nomináis. In my own analysis of -er nominalization, I will restrict myself to deverbal -er nomináis because their semantics is encoded in the process-participant configuration which they designate through the constructional link which they establish between a verb and the -er suffix. I will, more particularly, argue that it is only the analysis of the constructional properties of the various types of -er suffixation that can shed a new light on their semantics, and, ultimately, on the high-level schema which all deverbal -er nomináis instantiate.

1.1.3. The verbal base of deverbal -er nomináis Among the lexicogrammatical features that must be looked at in more detail to uncover the meaning of deverbal -er nominalization and shed light on the occurrence of agentive and non-agentive -er nomináis are those of the base verb. In Lemmens (1998), it is shown that the distinction between transitive and ergative processes, as described in Davidse (1991, 1992b, 1998b), is also operative on the morphological level of deverbal -er derivation. What distinguishes a transitive from an ergative process? Characteristic of transitive processes (e.g., chase, hit, kill) is that they are Actorcentred: their "most central participant" is the Actor, and the "ActorProcess complex is grammatically more nuclear and relatively more independent" (Davidse 1992b: 110). The basic Actor-Process complex can be extended only to include a Goal, as in The lion is chasing the tourist. Ergative processes such as break, open and roll, in contrast, are "Mediumcentred", with the Medium as "most nuclear participant" (Davidse 1992b: 110) (e.g., The glass broke). The basic Medium-Process constellation can only be opened up to include an Instigator, as in The cat broke the glass. While the transitive Goal is a "totally 'inert' Affected", the ergative Medium "co-participates in the process" (Davidse 1992b: 118). In ergative one-participant constructions such as The glass broke, this active co-

-er nominalizations: Agent names or Subject names? 109 participation of the Medium in the process is foregrounded and the Medium is presented as "semi-" or "quasi-autonomous" (Davidse 1998b).32 The differences that characterize the transitive and ergative paradigms in the clause, Lemmens (1998) argues, are also at work in deverbal -er formation. Lemmens (1998: 127) observes, firstly, that corpus material shows -er nomináis to be primarily based on transitive processes, which is intuitively plausible given the prototypically agentive profile of -er nominalization and the Agent-centredness of the transitive paradigm. Nomináis that profile the Goal of a transitive process are far less common, but nonetheless exist (e.g., dipper, sipper, scratcher). Importantly, they cannot be attributed agent-like properties, but designate inert, affected entities (Lemmens 1998: 137).33 Lemmens further holds that the majority of -er nomináis derived from ergative processes designates the Instigator of the process (e.g., opener, freezer). The claim that -er nomináis are typically Agent-profiling then schematizes over nomináis that designate the Actor of a transitive process and nomináis that profile the Instigator of an ergative one (Lemmens 1998: 131). Against Keyser and Roeper (1984: 395), who regard Mediumprofiling -er nomináis as "idiosyncratic", Lemmens moreover argues that -er formations designating the Medium of an ergative process can be accounted for as extensions of the agentive prototype. Medium-profiling -er nomináis are motivated by the Medium's potential for independence, which is tied up with its co-participating role in the process. Lemmens (1998: 136) states that the "higher the independence, the more likely it becomes that an -er formation may profile this participant rather than the Instigator". A nominal like cracker, for instance, designates the Medium of the ergative process crack and foregrounds the fact that the particular type of wafer which it refers to makes a cracking noise when you eat it (Lemmens 1998: 135). Lemmens manages to refine Ryder's account of -er nominalization in terms of extension of the agentive prototype by applying the transitive/ergative distinction to it. Especially the non-agentive type of -er nominalization can now be described more accurately, viz. as involving a Goalprofiling and a Medium-profiling type, with the former constituting the most radical departure from the agentive prototype. Medium-profiling -er nomináis can be said to profile a co-participating participant in the process: while an Instigator may still be part of the semantic base, they present the Medium as a semi-autonomous or independent entity. Illustrative of Medium-profiling -er nomináis are the food-related -er nomináis (e.g., steamer, broiler, baker) which Ryder identifies as "agent-like active patients" (Ryder 1991: 309). The fact that they are more agent-like than, for

110 Deverbal -er suffixation: Towards a descriptive position instance, dipper, sipper and scratcher can now be argued to be due to the ergative nature of the process which they nominalize and, related to it, to the Medium which they profile.34 More in general, Lemmens's account of how the transitive and ergative paradigms operate in -er derivation points to the importance of considering the contribution of the verb from which -er nomináis are derived: the compositional value of a deverbal -er nominalization largely depends on its verbal base. To conclude, then, both Ryder's analysis and Lemmens's refinements of it are indicative of the rich representational or ideational semantics of deverbal -er nominalization. The analysis of the representational functions which are realized in a nominalization system forms an essential part of a multifunctional approach to nominalization (as discussed in Chapter 3, Section 4). Representational categories are important for the description of deverbal -er nominalizations in that they reveal the variety that exists among them and allow us to identify the category's prototype, i.e. the Agent-profiling type of -er nominal. However, representational categories seem unable to generalize across all instances of deverbal -er suffixation. The category of Agent, for instance, fails to account for the non-agentive instances of -er nominalization. In representational terms, moreover, the non-agentive type of -er nominalization is necessarily viewed as a peripheral, non-prototypical instance of -er nominalization. A schematic view on -er nominalization which, I argued in Chapter 3 (Section 1.1), necessarily complements an analysis in terms of prototype cannot therefore be based on the representational categories of agency and process type. In the following section, I will examine an attempt at schematization suggested in Levin and Rappaport (1988) and Rappaport Hovav and Levin (1992).

1.2. Non-agenti ves as external arguments Levin and Rappaport (1988) and Rappaport Hovav and Levin (1992) claim that the derivation of -er nomináis "refers not to particular semantic-role labels but rather to argument-structure configurations" (Levin and Rappaport 1988: 1067). -Er nomináis may bear a wide range of semantic relations to the verbs from which they are derived, but "they most systematically correspond to the external argument of the base verbs" (Rappaport Hovav and Levin 1992: 145) and therefore only derive from verbs which have external arguments (Levin and Rappaport 1988: 1070). What does the term 'external argument' stand for? The notion of 'external argument' was introduced in Williams (1980, 1981) to refer to a particular, "distinguished argument, not a syntactic posi-

-er nominalizations: Agent names or Subject names ? I l l tion, a case, or something else" (Williams 1981: 83). The external argument is the argument "that corresponds to the NP in a sentence of which a [verb] phrase with that item as its head is predicated" (Williams 1981: 84). The example which Williams gives to illustrate the notion is that of the verb hit. Hit has two arguments, viz. one which serves the semantic role of Actor and another one which is Theme. The Actor is the external argument because in a clause it "must be specified in a position external to the verb phrase of which hit is the head" (Williams 1981: 84). The Theme, on the other hand, "must be specified within the verb phrase" (Williams 1981: 84). This seems to correspond to what in Levin and Rappaport (1988: 1074) is understood by 'external argument': the external argument of a verb is defined there as its "underlying (D-structure) subject", while an internal argument is an "underlying (D-structure) direct object". Passives, middles and ergative one-participant structures illustrate that external arguments do not necessarily correspond to the so-called "surface structure" Subject. In passive clauses, for instance, the arguments of the verb are reorganized and "the internal argument of the verb occupies the canonical subject position" (Haegeman and Guéron 1999: 123). In the passive clause Louise was invited (by Thelma), the internal argument Louise thus occupies the Subject position, while the external argument Thelma "may appear optionally in a PP adjunct" (Haegeman and Guéron 1999: 123). Ergative, one-participant structures such as The ball rolls (called 'unaccusative' in Perlmutter 1978) have only one argument, which functions as Subject and is not an external, but an internal argument (Levin and Rappaport 1988: 1074). In the corresponding two-participant structures, the internal argument functions as Object (e.g., I roll the ball). Middle clauses such as This book reads well, finally, constitute yet another type of clause in which the Subject is the internal rather than the external argument of the verb (Williams 1981; Keyser and Roeper 1984). In their analysis of -er nominalizations, Levin and Rappaport (1992: 145) discuss four groups of nomináis which, they claim, corroborate their "external argument generalization", viz. instrumental -er nomináis, nomináis derived from the spray/load-type of verbs, nomináis based on intransitive verbs, and, finally, nomináis which are related to the middle form of the verb. As far as instrumental -er nomináis are concerned, Levin and Rappaport claim that they must designate "intermediary" instruments or instruments which are able "to perform the action in some sense autonomously" (Levin and Rappaport 1988: 1071). The verb open, for instance, can take -er suffixation to refer to an intermediary instrument involved in the opening of something, as in

112

Deverbal -er suffixation: Towards a descriptive position

(7)

a. Doug opened the can with the new gadget, b. The new gadget opened the can. (opener)

Verbs that take what have been referred to as "facilitating" or "enabling" instruments (Marantz 1984 and Wojcik 1976, as cited in Levin and Rappaport 1988: 1071), in contrast, do not allow instrumental -er nomináis: (8)

a. Bill ate the meat with a fork. b. *The fork ate the meat. (*eater)

Interestingly, it is also intermediary instruments which in the clause can function as Subject, as illustrated in The new gadget opened the can and The key unlocked the door (Levin and Rappaport 1988: 1071). Instrumental Subjects, Levin and Rappaport claim, "are only found with verbs that have external arguments independently" (Levin and Rappaport 1988: 1071), i.e. with verbs that always take an external argument: the instrumentais that figure as Subjects of these verbs can be said to "satisfy the lexical requirements of certain predicators for external arguments" (Levin and Rappaport 1988: 1071). Because instrumental -er nominalizations only profile intermediary instruments or instruments which in the clause can appear as Subjects and designate external arguments, Levin and Rappaport conclude that they can be analyzed as designating external arguments (Levin and Rappaport 1988: 1072). A second group of nomináis which is claimed to provide support for the external argument generalization consists of nomináis derived from spray/load verbs (as illustrated in (9) and (10), based on Levin and Rappaport 1988: 1072): (9)

a. Bill loaded cartons onto the truck, b. Bill loaded the truck with cartons. (10) a. Jack squirted water on the plants, b. Jack squirted the plants with water. Especially the behaviour of the Locatum-argument of spray/load verbs (i.e. the argument which is either realized as a bare nominal such as cartons and water, or as a vw/A-phrase, as in with cartons and with water), is interesting according to Levin and Rappaport: being an internal argument, the Locatum can never appear as Subject of the clause or be profiled by an -er nominalization (Levin and Rappaport 1988: 1073): (11) ""Cartons loaded the truck, ("loader)

-er nominalizations: Agent names or Subject names ? 113 Nomináis such as loader, sprayer, spreader and sprinkler do not designate the Locatum-argument of load, spray, spread or sprinkle, but rather profile an (intermediary) instrument, which in the clause is realized either as Subject or in the form of a vw'fA-phrase (Levin and Rappaport 1988: 1073): (12) a. The crane loaded the truck. b. Bill loaded the truck with cartons with a crane. With respect to -er nomináis derived from intransitive verbs, Levin and Rappaport claim that they "should be based on a unergative rather than an unaccusative verb, since only the single argument of a unergative verb is an external argument" (Levin and Rappaport 1988: 1074). The distinction between unergative and unaccusative intransitives goes back to Perlmutter (1978) and Burzio (1986, as cited in Levin and Rappaport 1988: 1074). Unergatives largely coincide with what in Davidse (1991, 1992b) are called 'intransitives' (e.g., run, climb, work and sin). The unaccusative category includes all the ergative verbs of Davidse's model in their 'non-caused' sense, as in the door opened, the ball rolled, the water froze, as well as processes such as disappear, happen and die. The single argument of unaccusatives "is not an external argument but rather a direct internal argument, that is, an underlying (D-structure) direct object" (Levin and Rappaport 1988: 1074). It cannot therefore undergo -er nominalization: (13) *disappearer, *happener, *dier, *exister -Er nominalizations derived from verbs with both an unaccusative and a transitive use can only profile the external argument of the transitive use of the verb (Levin and Rappaport 1988: 1075), i.e. they can profile the Instigator but not the Medium of the ergative process: (14) opener, dryer, freezer, heater Yet another group of -er nominalizations which is claimed to corroborate the external argument generalization mainly consists of nomináis derived from cooking verbs. They refer to the "theme argument of the base verbs (the entity that changes state)" (Levin and Rappaport 1988: 1077). Examples are baker, broiler, fryer, roaster and steamer (Levin and Rappaport 1988: 1077). -Er nomináis which are not based on verbs of cooking but seem to behave comparably are bestseller, loaner and sticker (Rappaport Hovav and Levin 1992: 149). They "present a problem for an account in terms of the notion of external argument" (Levin and Rappaport 1988:

114 Deverbal -er suffixation: Towards a descriptive position 1077), because they are generally assumed to profile an internal rather than an external argument. Levin and Rappaport (1988: 1078) point out, however, that these nomináis "do not receive the interpretation that would be expected if they were derived directly from the transitive or unaccusative uses of the related verb". Rather, their meaning seems closer to the interpretation that the related verb receives in the middle structure (Levin and Rappaport 1988: 1078). A nominal like broiler, for instance, designates a chicken which is bred or intended to be broiled, which comes closer to the interpretation of a middle construction such as This chicken broils well than to This chicken broils. Apart from semantic correspondences, middles and nominalizations in -er also show the same constraints with respect to the verbs which they can be based on: both of them, it is argued, can only be derived from verbs with 'affected objects'. A structure like *this planet sees well and the corresponding -er nominal *seer are, for that reason, impossible (Levin and Rappaport 1988: 1079). Does the relationship with middle formation imply that cooking -er nomináis, like all -er nominalizations, profile the external and not the internal argument of the base verb? Levin and Rappaport (1988: 1078) admit that most accounts of the middle construction consider the single argument of middles as internal (e.g., Keyser and Roeper 1984; Roberts 1987), but they argue that the occurrence of -er nomináis like broiler, baker and steamer might be taken as support for an "alternative analysis of the middle construction" (Levin and Rappaport 1988: 1078). The analysis which they propose involves "the externalization of the direct internal argument of the verb in the lexicon" (Rappaport Hovav and Levin 1992: 149). An -er nominalization which is related to middle formation therefore profiles an external rather than an internal argument. Levin and Rappaport's analysis is meant to show that -er nominalizations cannot be generalized across in terms of semantic roles but only in terms of the notion of 'external argument'. The latter, however, proves to be an ill-defined notion: what it seems to stand for is 'being the most likely candidate for serving as Subject', and this, by and large, corresponds to being the Agent of the process or being an entity that can be conceived as agent-like. They thus fail to distinguish the representational function of Agent from the 'interpersonal' or instantiating/grounding function of Subject (see Chapter 4, Sections 2.2 and 2.3). Levin and Rappaport also bring in the Subject of middle constructions, but without investigating the true extent of this category in English. This leaves a number of -er types unaccounted for. In what follows, I will briefly run through those types of -er nominalization which Levin and Rappaport fail to elucidate. The main ones are: -er nomináis designating what I will call 'oblique participants' (Laffut

-er nominalizations: Agent names or Subject names? 115 2000), -er nomináis designating the single participant of an ergative unaccusative verb, and, finally, the middle-related type of -er nominal (as a systematic category). (Note that many of these are not accounted for by Ryder 1991, 1999b, Lemmens 1998 and Panther and Thornburg 2002 either.)

1.2.1. -er nomináis profiling oblique participants Following Laffut (2000: 139), I will use the term "oblique participant" to refer to participants which are clausally realized in the form of a prepositional phrase (as in 15). Laffut (2000: 137-156) lists a number of semantic and formal arguments which show that oblique participants differ from circumstantials as in (16) in that they are basically "complements" and elaborate a salient e-site in the verb with which they are used (Langacker 1987: 305) (examples based on Laffut 2000): (15) a. The pope was killed by a sniper. b. Spray Static Guard on your comb or brush. (16) a. The pope was killed on a Sunday. b. Floyd cut the salami in the kitchen. The existence of -er nomináis that profile oblique participants counters part of Levin and Rappaport's analysis of instrumental -er nomináis, of nomináis based on spray/load, verbs and of nomináis derived from unaccusative verbs. Nomináis like stroller, walker and viewer fail to profile the agentlike, external argument of the processes of strolling, walking and viewing (this is also pointed out by Ryder 1999b: 272-273): instrumentais such as stroller and viewer cannot be analyzed as doing the strolling and viewing themselves, i.e. a stroller does not stroll and a viewer does not itself view. Instead, they seem to profile oblique participants. Apart from oblique participants which indicate the means whereby a process can be carried out, -er nomináis can also designate the place of a process, as in: (17) I work on the stepper, I do squats, bench presses, abdominals, concentrated curls - it's all short, sharp shocks and it's very effective. (CB) (18) Various crashing noises erupted from our pew as our new friend searched for his lost gold earring beneath the wooden kneelers. (CB)

116 Deverbal -er suffixation: Towards a descriptive position (19) Our welcoming ensemble features leather wherever your body touches. Includes a sleeper with innerspring mattress and TV-tilt headrest. (CB) These nomináis illustrate that -er nomináis derived from intransitive verbs (in Levin and Rappaport's terminology, unergative verbs) do not necessarily designate the external argument of the base verb: rather than profiling the entity which does the stepping/kneeling/sleeping, they designate the surface that you step on, the chair that you kneel on and the bed that you sleep in. Finally, Levin and Rappaport only focus on the Locatum argument of spray/load verbs (e.g., Bill loaded the truck with cartons, Bill loaded cartons onto the truck) and they ignore the Location of the process, as if that category is not relevant to -er suffixation (e.g., Bill loaded the truck with cartons, Bill loaded cartons onto the truck). Yet, I have found several -er nomináis which designate the Location of the verb load: (20) [about a boat:] She came up the Thames to Chelsea Harbour, and was craned out on to a low-loader. (CB) (21) The most popular type of washing machine is the front-loader but there are still alternatives. (CB) (22) Top-loaders are useful for a kitchen short on space as they can be placed at the end of a run of units. (CB) The Location cannot be analyzed as the external argument of load. The fact that it can be profiled by an -er nominalization thus forms yet another counterargument to the external argument generalization which Levin and Rappaport propose for -er nominalization.35

1.2.2. -er nomináis based on unaccusatives Levin and Rappaport are right in pointing out that (unaccusative) verbs of existence and disappearance do not normally take -er suffixation (e.g., *exister, *happener, *disappearer). Yet, in some cases they do, witness the following (attested) examples of -er nomináis with a one-participant unaccusative as basis: (23) When I get busy, I tend to forget to water my plants, so it's a good thing I've got a plant that's a wilter. The minute it gets a little low on

-er nominalizations: Agent names or Subject names ? 117 water, it wilts, so it's a reminder it's time to water everything again. (Ryder 1999b: 274) (24) A few thought they had noticed someone resembling the man in the picture. I wasted two days tracking one of the supposed resemblers and found no resemblance at all. (Kastovsky 1986b: 410) An ad hoc formation with disappear and die likewise cannot be excluded: e.g., He is a good disappearer, You have peaceful diers and troubled diers. What's more, -er nomináis which nominalize the one-participant or 'unaccusative' use of ergative verbs are also found: there is not only the example of cracker (Lemmens 1998: 135); also the food-related nomináis broiler, baker, steamer and cooker designate the Medium of the ergative verb which they are derived from. They cannot therefore be analyzed as external arguments. 1.2.3. Middle-based -er nomináis Finally, Levin and Rappaport's attempt to analyze the so-called 'middlebased' type of -er nominal as profiling the external argument of the base verb is circular. As pointed out by Ryder (1999b: 275), Levin and Rappaport "hypothesize that -er forms must always appear as external arguments of their root verbs and then ... decide that the subjects of middle verbs are external arguments primarily based on the possibility of turning the verbs into bases for -er nomináis". The relation which, they suggest, exists between certain non-agentive -er nomináis and middle constructions in itself, however, is intriguing. The middle construction captures the semantics of certain non-agentive -er nomináis better than any other clausal structure. A nominal like broiler, for instance, does seem to be more closely related to the middle construction This chicken broils well than to the non-middle one-participant structure This chicken broils (Levin and Rappaport 1988: 1078). The relation between the nominal bestseller and the middle construction This item sells best is perhaps even more obvious. Yet another nominal with a meaning that comes close to that of the corresponding middle construction is spreader 'butter which is particularly spreadable' (Ryder 1999b: 273): spreader designates a type of butter that spreads well. Even though, when looked at in terms of its relation to the middle construction, spreader seems to connect with other (non-agentive) -er nomináis, it cannot, however, be accounted for by Levin and Rappaport's external argument analysis: it profiles the Locatum-argument of the verb spread (you spread butter on your sandwich with a knife), which is an internal rather

118 Deverbal -er suffixation: Towards a descriptive position than an external argument (Ryder 1999b: 273). That it nonetheless exists forms yet another indication that the external argument generalization is untenable and that Levin and Rappaport's analysis of -er nominalization in terms of the argument structure of the base verb cannot generalize across the agentive and non-agentive cases of -er nominalization. As with Ryder's analysis in terms of semantic roles, it is especially with regard to the category of non-agentive -er nomináis that many questions remain unanswered. The external argument generalization may not hold, but the observation that some non-agentive -er nomináis are close in meaning to middle formation opens up a highly interesting, new perspective on non-agentive -er nominalization. As I pointed out in Chapter 3 (Section 5), systematic relations or relationships of agnation between nominalizations and clauses shed light on the clausal categories which are realized in nominalizations. Without doubt, the most characteristic property of middle constructions is that they are centred on the relation between a non-agentive Subject and an active finite verb (e.g., This shirt irons easily, That floor cleans well). Agentive participants are, in any case, prototypical Subjects. If it can be shown, then, that, together with agentive -er nominalizations, all the nonagentive cases of -er suffixation can be related to clausal Subjects, this would lead to a generalization that covers the distributional properties of -er nominalization much better than the generalization in terms of semantic roles which Ryder proposes, the generalization in terms of conceptual contiguity discussed in Panther and Thornburg (2002) or that in terms of argument structure configurations, suggested by Levin and Rappaport. In the following section, I will look at how the function of Subject has figured in descriptive accounts of -er nominalization up till now.

1.3. -er nominalizations as Subject names The label of 'Subject' has thus far been assigned to -er nominalizations only sporadically, and then particularly as an empty label representing a purely syntactic category. The term 'Subject' is in fact often used in one and the same breath with the semantic role that is typically mapped onto it, viz. that of Agent. In the same way, patientive -er nomináis are then classified as belonging to the 'Object' type of -er nomináis. Marchand's (1969) account o f - e r nominalization nicely illustrates this specific use of the concept of Subject in the description of -er suffixation. Marchand (1969) distinguishes between a Subject type of deverbal -er nominalization, an Adverbial Complement type and an Object type. In the Subject group, he situates all those nomináis that are agentive, ranging from

-er nominalizations: Agent names or Subject names ? 119

human Agents (e.g., preacher, hunter) over non-human, animate Agents (e.g., retriever, warbler) and material Agents or Instruments (e.g., blotter, lighter, toaster) to immaterial Agents (e.g., eye-opener, reminder). Nomináis profiling what he calls an 'Adverbial Complement' are slipper 'you slip into the shoe' and waders 'waterproof boots'. The Object type of -er nominal, finally, is made up by the patientive cases of -er nominalization, exemplified by a nominal such as tier 'pinafore' from you tie it (Marchand 1969: 275). A comparable analysis, but then of Dutch non-agentive -er nomináis is found in Van Langendonck (1988). Even though he recognizes the generalizing power of assuming a Subject profile for both agentive and nonagentive -er nomináis (Van Langendonck 1988: 326), Van Langendonck prefers to distinguish between what he calls "Subject", "Object" and "prepositional Object" nomináis. The Object type consists of those -er nomináis which are patientive in profile; the prepositional Object group contains circumstantials which are realized by means of a prepositional phrase (e.g., loper [runner], i.e. 'something that you walk on', based on Boon 1972: 94). The Subject group, finally, consists of the prototypical, agentive instances of -er nominalization, as well as of instrumental nomináis and other, inanimate but agent-like cases of -er suffixation. Interestingly, Van Langendonck (1988: 325-326) gives two reasons for not adhering to a generalization in terms of the Subject category: first, he does not see any evidence in favour of passive structures (with a patientive Subject) being able to underly nominalization processes. Secondly, the occurrence of prepositional Object profiling nomináis is considered problematic for a generalization in terms of Subject, since prepositional phrases cannot be construed as clausal Subjects. To include the latter type of -er nominalization in the analysis of the overall system, Van Langendonck thus proposes an analysis which distinguishes between Subjects/Agents, Objects/Patients and prepositional Objects. In short, descriptive accounts of deverbal -er suffixation in English and Dutch typically use the Subject-label as if it were identical with that of Agent (see Chapter 2, Section 4). The notion of 'Subject' is not considered to generalize across the agentive and the non-agentive instances of -er nominalization. In fact, arguments are provided against using it to cover agentive as well as non-agentive -er nominalizations. It is, for instance, pointed out that (Dutch) -er nomináis that designate oblique participants do not seem relatable to a clause in which they function as Subject. Patients, then again, can be realized as clausal Subjects, viz. of middle and of passive clauses, but not all patientive -er nomináis appear to nominalize processes that can figure in a clausal middle construction (as pointed out by

120

Deverbal -er suffixation: Towards a descriptive position

Booij 1986), and evidence that morphological processes can be passive in nature is argued to be sparse.36 So, if the concept of Subject is going to be used to generalize across all deverbal -er nominalizations, and if the middle construction is going to be linked up with non-agentive -er nominalization, it will have to be shown that these three arguments are unjustified: -er nomináis that profile oblique participants will have to be provided for; non-agentive -er nominalizations in general have to be shown to be more closely related to middle formation than to passive clauses, and, finally, non-agentive nomináis derived from verbs that do not normally occur in clausal middle constructions have to be integrated in the analysis. The Subject function, moreover, has to be attributed a semantic function in its own right, which can, but is not always, mapped onto the representational function of Agent.

2. Lexicalization versus ad hoc nominalization I have thus far concentrated on the occurrence of agentive as well as nonagentive -er nomináis. Cutting across the network of deverbal -er nominalization is, however, yet another distinction that has to be accounted for, viz. that between lexicalized and ad hoc -er nominalizations: (25) a. saver of lives, as in he is the saver of many lives b. a lifesaver 'lifeguard' (Levin and Rappaport 1988: 1070) These -er nominalizations are characterized by different degrees of entrenchment and serve distinct functions in discourse. Both these differences have been described in various ways. A lexicalized nominalization such as lifesaver has been argued to serve a primarily labelling or naming function (Kastovsky 1986a, 1986b). Ad hoc nominalizations have been termed "syntactic recategorizations" (Mötsch 1979, cited in Kastovsky 1986b: 411), "entirely predictable nominalizations" (Quirk et al. 1985: 1547) or "deictically used" structures (Ryder 1999b: 282). They are said to serve the textual function of condensing information and of "referring back to something contained in the previously occurring text" (Kastovsky 1986b: 411; see also Mackenzie 1990: 141-143). I prefer to use the term 'ad hoc nominalization' because it captures the fact that these nomináis are not sufficiently entrenched to be lexicalized and therefore have to be derived 'on the spot', on the basis of the constructional schema of -er derivation. Their being used for anaphoric reference or reference to something in the previous text does not distinguish them from lexicalized-er nominalizations,

Lexicalization versus ad hoc nominalization 121 which can likewise be used anaphorically once they come to function in the structure of the nominal. Still, it cannot be denied that ad hoc formations are frequently used to rephrase a previous clause by nominalizing it, as happens in the following examples: (26) One guy jumped right into the fight, but his friend immediately vanished. Hie police came and hauled off the fighter, after which the vanisher promptly reappeared, laughing, (cited in Ryder 1999b: 283) (27) "Not him," says Mrs. Cook. "This one, though. A bit of a poseur, but he moves very nicely." I will not describe the nice mover, to spare his blushes, but let us say that he had dressed for the part. (CB) Interestingly, lexicalized -er nomináis and their ad hoc counterparts are characterized by significant lexicogrammatical differences. As Mackenzie (1990: 133) points out, lexicalizations by and large function as ordinary nouns "in that their syntactic behaviour is indistinguishable from comparable words that cannot be related to a verb (author, butcher, etc.)". Ad hoc -er nominalizations, in contrast, are only "partially nominalized" (Mackenzie 1990: 136-137). Strang (1968, 1969) calls them "minimal" nominalizations. Unlike lexicalized -er nomináis, they cannot shift the characteristic accent position of the verb (as illustrated in 28a). They take adjectives that resemble adverbiale in that they "do not ascribe a quality to the referent but rather indicate the manner, frequency or extent of the activity associated with the referent" (Mackenzie 1990: 137), as in (29): (28) a. He is a PHOtographer of snow-scenes. b. He is a professional phoTOgrapher. (Mackenzie 1990: 136) (29) a. He is a hard worker. b. He works hard. (Mackenzie 1990: 137) Ad hoc nominalizations are said to "inherit argument structure" (Levin and Rappaport 1988: 1069): they can be postmodified by an o/-phrase realizing what would clausally have been a participant of the process (e.g., saver of lives; destroyer of Del Harris's hopes). And, finally, they seem to be necessarily agentive in profile (Rappaport Hovav and Levin 1992: 132). Non-agentive ad hoc -er nominalizations do not appear to exist. Ad hoc nominalizations do not even allow for an instrumental profile: a nominal like, wiper, for instance, may receive an instrumental or a human Agent reading, but "a wiper of windshields may receive only an agentive reading" (Levin and Rappaport 1988: 1069).

122 Deverbal -er suffixation: Towards a descriptive position The semantic differences that are linked to these lexicogrammatical properties have been hinted at in Levin and Rappaport (1988) and Rappaport Hovav and Levin (1992). Ad hoc nominalizations are described as event nomináis because, unlike their lexicalized counterparts, they seem to designate processes that have been carried out. As Levin and Rappaport put it, "Someone may be called a lifesaver even if he or she has never saved anyone, but someone may not be called a saver of lives unless he or she has actually been involved in saving lives" (Levin and Rappaport 1988: 1070). While ad hoc nominalizations designate events that have occurred, lexicalizations are non-event nomináis or nomináis that do not presuppose the actual occurrence of the process. However, as the nominalizations in (30) and (31) (i.e. breather and sleeper) make clear, Levin and Rappaport's characterization of ad hoc nominalizations as designating actual events is not accurate: not all of them designate events that have actually occurred: (30) Recent research has established that merely breathing smoke-laden air - passive or 'second-hand' smoking - puts the breather at risk. (CB) (31) Trouble sleeping is the most common complaint, including an inability to get to sleep, terrible, frightening dreams that wake the sleeper in a state of anxiety, and waking early feeling unrefreshed but unable to get back to sleep. (CB) The nomináis breather and sleeper refer to the one who breathes and the one who is sleeping respectively, but they do not designate specific, actual occurrences of the processes of breathing or sleeping. As to the "ability to inherit complement structure" (Rappaport Hovav and Levin 1992: 132), then, this does not seem to be possible in ad hoc nominalizations only: the nominalizations in (32) and (33) are both lexicalizations and they are followed by o^phrases which periphrastically realize an Object which specifies the range or scope of the process (in 32) or the Beneficiary of the teaching (in 33): (32) Mrs. Vogt was a teacher of German and French in a secondary school.... (CB) (33) Mrs. Emerson, the second-grade teacher of Tony, the youngest boy ....(CB) Unlike what is suggested in Levin and Rappaport (1988) and Mackenzie (1990), the inclusion of an o/-phrase is thus not a distinctive feature of ad hoc nominalizations.

Towards a coherent account of deverbal -er nominalization 123 Levin and Rappaport's distinction between event and nonevent nominalizations may be somewhat inaccurately formulated, but it does reveal that -er nominalizations not only differ in terms of the entity which they designate, but, that, depending on whether they are lexicalized or not, they differ in lexicogrammatical properties and realize different meanings. An account of the system of deverbal -er suffixation has to offer a satisfactory explanation for these differences and integrate them in the overall schematic network of -er nominalization.

3. Towards a coherent account of deverbal -er nominalizations The system of deverbal -er nominalization may have started out as an agentive system, it can no longer be characterized as such: non-agentive -er nominalizations are numerous and too productive to be described as nonsystematic. In the next two chapters, I will defend the claim that deverbal -er suffixation has evolved from profiling Agents to profiling Subjects. I will argue that an analysis of -er nominalization in terms of the Subject function provides a more accurate generalization for the present distributional properties of -er nominalization than one in terms of agency. In particular, I will show that the various types of non-agentive -er nominalizations are systematically equivalent to the Subject of middle constructions and of so-called "Setting-Subject" constructions (Langacker 1991: 346; e.g., This tank does not squirt water.). I will moreover argue that deverbal -er suffixation construes equivalents to the basic grounding or deictic function of the clausal finite verb. The correspondences between the clausal Subject-Finite unit and -er nominalization elucidate the semantics of the -er suffix significantly. The evolution in the system of deverbal -er nominalization may run parallel to certain changes of predominance in the various functional layers of organization in the clause. Within the Prague School (which has always integrated the comparative and the diachronic in its functional approach), it has been hypothesized that the English clause has developed from an Agent-oriented to a predominantly Subject-oriented form of organization. Or, as Vachek (1966: 92) puts it, "the main function of the grammatical subject in an English sentence is no longer to express an agent". Unprototypical or non-agentive Subjects as in middle constructions are on the rise. Significantly, it is this clause type, which combines a non-agentive Subject with an active finite verb, which turns out to be particularly important for the analysis of non-agentive -er nominalization. It will therefore constitute the focus of the next chapter. In Chapter 7, then, I will present my own

124 Deverbal -er suffixation: Towards a descriptive position analysis of -er nominalization, along the lines of the theoretical-descriptive framework set out in Part I.

Chapter 6 The middle construction

This chapter37 focuses on the clausal middle construction, as exemplified in the following sentences:38 (1) (2) (3) (4)

More than 50 years later, the book still sells well. (CB) The new Holden Berlina handles like a junior sports sedan.... (CB) And it's latex paint, so it cleans up quickly and easily with soap and water. (CB) If they wanted to do Eliot, why didn't they do The Magi? It's narrative, it reads well.... (CB)

Middles typically combine an active verb form with a Subject that is patientive in nature and thus reminiscent of passive constructions.39 Semantically speaking, the middle is often associated with a facilitative meaning (Fawcett 1980; Kemmer 1993): it is said to express how easy or difficult it is to realize a particular process with respect to the entity construed as Subject. Prototypical middle constructions are construed with the adverb easily, as in (5) (6)

When no longer required, the discs remove easily. (CB) Compact lipstick size design, slips easily into your handbag or pocket. (CB)

Rather than account for the unique constructional link which middles realize between a non-agentive Subject and an active verb form, most analyses of middle formation describe middle constructions by reference to other constructional paradigms and, as a result, leave many questions unanswered. In this chapter, I will start with a brief sketch of the two main existing approaches to middle formation (Section 1). I will show that they tend to reduce the representational potential of middle formation and focus on only part of the middle system. Among the middle constructions that have been systematically ignored in the literature are, for instance, those based on intransitive processes, with an oblique participant as Subject. They will be the main topic of Section 2.1 will then zoom in on two aspects of middle formation which, in my opinion, are crucial to identify the high-level schema which middle constructions instantiate: I will first discuss the vari-

126

The middle

construction

ous semantic profiles of middle formation, and argue that some middles, rather than profile the facility with which a process can be carried out, focus on the destiny of the Subject or on the result of carrying out a particular process on it (Section 3). In a final section, then, I will elaborate on the essentially modal nature of the relationship which middles establish between a Subject and a Finite (Section 4).

1. Middle formation: A state of the art In the literature on middle formation, some have stressed the active, oneparticipant nature of the construction and have claimed agentive status for its Subject, while others have ascribed passive value to the construction, emphasizing that its Subject is affected and that an Agent is implied. I will refer to the former as the ergative and the latter as the passive approach to middles. Those who situate middle constructions close to ergative oneparticipant structures are, among others, Hale and Keyser (1987), Sinclair et al. (1990), Francis et al. (1996) and Sohn (1998). They argue that middles resemble ergative one-participant structures in that, firstly, they allow for the addition of an "event of causation" (Hale and Keyser 1987: 6), which is generally assumed to be characteristic of ergative one-participant structures such as the one in (7): (7) (8)

The boat sank. This pan cleans easily.

o

A torpedo sank the boat. You clean this pan.

Secondly, like ergative one-participant construals, middles are said to focus on the participant that is central to the event depicted by the clause and in which a change of some sort manifests itself (Hale and Keyser 1987: 6). By being realized in a one-participant structure, this change is presented as being effected "autonomously" or "without the participation of an external agent" (Hale and Keyser 1987: 7). The proponents of the ergative approach attribute varying degrees of agency to the middle Subject, which is described as "agent-like" in that it "corresponds to a participant which is, in an intuitively clear sense, responsible for the process depicted in the predicate" (Hale and Keyser 1987: 9; see also Lakoff 1977). The notion oí responsibility is discussed at length in Van Oosten (1977, 1986), who even characterizes the middle Subject as "energy source of the action" (Van Oosten 1986: 85) and as being able to "bring about the action of the predicate independently" (Van Oosten 1986: 93). The Subject's responsibility is said to follow from its properties,

Middle formation: A state of the art

127

which, Van Oosten claims, represent the linguistic raison d'être of middle constructions: "the purpose of the patient-subject construction is precisely to enable the speaker to assert that ... properties of the patient bear the responsibility for the occurrence of the action of the verb" (Van Oosten 1986: 93). In Lakoffs view as well, "the point of using the patient-subject construction is to say that properties of the patient are more responsible for what happens than the agent is" (Lakoff 1977: 248). An alternative analysis of middle formation is found in, among others, Sweet (1891), Jespersen (1914-1929, 3), Halliday (1967), Smith (1978), Keyser and Roeper (1984), Fellbaum (1986) and Fellbaum and Zribi-Hertz (1989). All of them start from the assumption that the middle realizes one participant overtly, but contains two. The two-participant approach to middle formation identifies the Subject of the middle as patientive. The Agent of the process is argued to be necessarily implied. Evidence in favour of the patientive nature of the middle Subject is said to come from the fact that middles cannot be used in the imperative (Fellbaum and Zribi-Hertz 1989: 19) (illustrated in 9), and from the fact that the middle Subject, unlike the Subject of an ergative one-participant construction, cannot be presented as doing things by itself (see 11) (Keyser and Roeper 1984: 405):40 (9) *Read easily, book! (10) a. The door closed slowly. b. The door closed slowly all by itself. (11) a. The book read slowly. b. *The book read slowly all by itself. As Keyser and Roeper (1984: 405) point out, a contradiction exists between the notion all by itself and the Agent which is implied in middles (see also Smith 1978): the middle Subject cannot be perceived as agent-like because the Agent of the process is implied.41 The Subject has moreover been claimed to be necessarily affected by the process (Levin 1993: 26): (12) *This poem understands easily. (13) *The Eiffel Tower sees easily from my window. Finally, following Sweet (1891), the middle has frequently been compared to and sometimes even been located within the passive paradigm (Halliday 1967; Keyser and Roeper 1984; Declerck 1991b). At the same time, however, it is pointed out that middles also differ from ordinary passives. Most importantly, middle constructions do not allow the implied Agent to be realized overtly, while passives do (Keyser and Roeper 1984: 406):

128

The middle

construction

(14) John was hit. John was hit by Bill. (15) Bureaucrats bribe easily. « *Bureaucrats bribe easily by managers. 2. The representational versatility of middle formation42 The deadlock which characterizes the literature on middle formation can only be broken by a detailed analysis of the different verb types that can function in the middle and the participant roles that can be mapped onto the Subject. First, the construction type of middle formation cannot be reduced to one or other verb type: it is widely acknowledged - even by those who situate middles close to the ergative one-participant paradigm - that ergative as well as transitive verbs can function in the middle. Examples (16) and (17) illustrate the middle use of the transitive verbs remove and wipe; the middles in (18) and (19) are based on ergatives, i.e. flatten and disperse: (16) Broiler rack removes easily. (CB) (17) In Britain, however, the traditional image of schools drama as being all chalk and blackboards does not wipe easily from people's minds. (CB) (18) Nylon carpets are prone to static ... and the pile flattens easily. (CB) (19) It's heavier, like an emulsion with the texture of chocolate mousse and it doesn't disperse easily. (CB) Apart from transitive and ergative verbs, however, also intransitive verbs can figure in middle constructions, a possibility which, while recognized in descriptions of the Dutch middle construction, has thus far been almost unanimously denied in the literature about English.43 The Dutch middles in (20), (21) and (22) qualify as intransitive. The example in (20) is based on Haeseryn et al. (1997: 52); that in (21) was extracted from the DDL corpus44; the middle in (22) is taken from De Vries (1910: 132): (20) Asfalt fietst prettiger dan grind. asphalt cycles more pleasantly than gravel 'it is more pleasant to cycle on asphalt than it is on gravel' (21) Sommige bomen zitten nu eenmaal lekkerder.... some trees just sit more comfortably 'some trees are just more comfortable to sit in'

The representational versatility of middle formation 129 (22) Nieuwe schoenen lopen lastig. new shoes walk with difficulty 'new shoes are difficult to walk with' The entities functioning as Subject are normally felt to be outside the nucleus of the clausal process and the participants that are directly involved in the process: they are oblique participants of the main verb, providing either spatial information (e.g., on asphalt, in trees) or information on the means whereby a process takes place (e.g., with/in new shoes).45 Only Van Oosten (1986), Smith (1978) and Rosta (1995) have hinted at the existence of English middle constructions of which the Subject is not patientive. They do not, however, offer a systematic analysis of their properties, let alone integrate them in the analysis of middle formation in general. After mentioning the structure This music dances better than the other one ('to this music you can dance better than to the other one', 1986: 84), Van Oosten, for instance, decides that although "the subject in this construction does not, then, need to be a patient but can be an instrument or a locative or hold some other relation to the verb, I will for simplicity continue to call this the patient-subject construction" (Van Oosten 1986: 85). The fact that intransitive middles are rare in English has doubtless contributed to the absence of a coherent description of their properties: even large-scale corpora such as COBUILD hardly contain intransitive middles. The following list therefore contains not only middles drawn from the COBUILD corpus, but also middles that were suggested by native speakers of English (labelled p.c.), and middles that were found in the literature on middle formation: (23) This music dances better than the other one. (Van Oosten 1986: 8) (24) The Strathayr track is racing consistently nowadays. (Barry Blake p.c.) (25) The pitch is playing truly/well. (Barry Blake p.c.) (26) [about a tennis court:] It is slightly coarser, so it plays a bit slower. (CB) (27) "They rolled the green just before the match and it ran three seconds faster," said Curtis.... (CB) (28) You wouldn't give my tip a chance of landing Aintree's Martell Red Rum Chase ... but with the ground riding slower, he should improve dramatically. (CB)

130

The middle construction

The Subjects of these middle constructions designate either the means that is used to carry out the process (music) or they specify its location (examples 24 to 28). They are normally realized as prepositional phrases: you dance to music, run on a track, bowl on a pitch, play on a court, run on a green and ride on a ground. Yet another type of intransitive middle has an Instrument for Subject (Smith 1978: 103): (29) The brush paints well. (30) The knife cuts well. These middles can be paraphrased as 'with this brush/knife you can paint/cut well'; they are ambiguous between a middle reading and an ordinary active reading that profiles the agent-like involvement of the Instrument in the process. Middle formation thus turns out to be remarkably versatile as to the verb classes which it allows for - transitive, ergative and intransitive - and the roles which can be mapped onto the Subject. The distinct representational relations in middle constructions form an essential part of the semantics of middle constructions: the relationship between a Patient and a transitive process (This meat cuts easily), for instance, is fundamentally different from that between a Complement and an intransitive process (This court plays slower), or that between a Medium and an ergative process (This emulsion doesn't disperse easily). An account of the representational functional strand in the middle thus forms an essential part of the multifunctional approach which in general is vital to come to an accurate description of structural patterns (see Chapter 2, Section 4). It is, moreover, only when considering the full range of representational options which middle formation allows for that it becomes clear that many of the properties which have in the literature been ascribed to middle formation are not applicable to all middle constructions. For one thing, the middle cannot be said to systematically alternate with a two-participant structure in which the Subject is turned into Object (a claim made by proponents of the ergative approach), because intransitive middles contain only one participant (e.g., The green ran faster but *You ran the green faster). Neither can it be analyzed as realizing a kind of passive twoparticipant structure (e.g., The green ran faster but *The green was run faster). Intransitive middle constructions also form evidence against an agentive interpretation of the Subject in middles: like Patients, the Subject of intransitive middles cannot be interpreted as energy source of the process. In general, it has to be observed that the Subject of middle constructions is not necessarily affected (as argued by the proponents of the 'pas-

The representational versatility of middle formation

131

sive' view on middle formation) and it does not always undergo a change (as claimed by the adherents of the ergative view on middle formation), witness the following examples: (31) Girls don't make it just because they are skinny. It's about how they photograph and move. (CB) (32) His earlier short stories don't read so well. (CB) The girls in (31) are not affected or changed by being photographed. Likewise, the short stories in (32) are not affected or changed by being read. The middle Subject may not be agentive, but is the Agent of the process then obligatorily implied? As pointed out by Langacker, a middle like This ice cream scoops out quite easily cannot but be interpreted as involving an Agent which does the scooping out: we do not picture the ice cream scooping itself out (Langacker 1991: 334). It is, in fact, the semantics of the transitive verb scooping out which invokes the Agent: if for This ice cream scoops out quite easily we do not imagine the ice cream scooping itself out, it is because we follow the semantic instructions given by the active transitive verb scoop out, which necessarily implies an Agent. What is presented as most prominent by the middle construction, however, is the relationship between the process and the non-agentive entity construed as its 'focal participant' or Subject (Langacker 1991: 335). Intransitive middles can be given an analogous semantic analysis: they most prominently present an intransitive process in its relation to a Location or Means, but they imply an Agent which actually executes that intransitive process. In, for instance, The court plays a bit slower, an Agent is invoked which does the playing. Likewise, in This music dances better than the other one, an Agent is implied which dances to the music. Van Oosten and Lakoff s emphasis on the agent-like responsibility of middle Subjects should therefore be toned down: it is widely agreed that the middle attributes the fact that a particular process can be carried out to the inherent properties of the entity construed as Subject (apart from Van Oosten 1977, 1986 and Lakoff 1977, also Fellbaum 1985, 1986; Langacker 1991; Massam 1992; Rosta 1995; Lemmens 1998), but the fact that the Subject has properties that make a certain process possible should not be confused with agentivity. Or, as Rosta (1995: 128) puts it: In The book sold well... the properties of the book are a necessary condition for the book to sell well, but need not be (and typically are not) a sufficient condition; if it were not for the properties of the book, it would not sell well, but even though the book has the properties it has it will not sell

132 The middle construction well unless some other participant (an agent) gets involved in getting the book to sell well. Rather than assign agent-like value to its Subject, the middle, in other words, foregrounds the fact that the Subject-entity has properties which influence the occurrence of a particular process. It has been pointed out that when non-human entities function as middle Subjects, these inherent properties are often part of their design (Levin and Rappaport 1988; Lemmens 1998). As Lemmens (1998: 80) puts it, "the properties emphasized in the middle construction are often also those for which the entity has been designed in the first place". In the case of the latex paint in (33) and the lipstick in (34), for instance, it is obvious that cleaning up easily with soap and water and slipping easily into your handbag or pocket is part of what they are designed for: (33) And it's latex paint, so it cleans up quickly and easily with soap and water. (CB) (34) Compact lipstick size design, slips easily into your handbag or pocket. (CB) Animals, then again, are often bred for a particular process: a chicken that broils well has probably been bred to be broiled (Levin and Rappaport 1988: 1078).

3. Towards a semantic typology of middle constructions Importantly, middle constructions do not only vary according to the process type which they are centred on and the non-agentive entity which they profile: middles also differ depending on which facet of the interaction between the non-agentive Subject and the process they highlight (note that this makes them particularly suitable for being used in advertising, which, as pointed out by Yoshimura (forthcoming), frequently resorts to middle constructions). In the literature on middle formation, one has tended to focus on facility-oriented middles, or middles that specify how easy or difficult it is to carry out a particular process on the Subject-entity. Facilityoriented middles are typically construed with the adverb easily. However, as pointed out in Heyvaert (1997), corpus data reveal the existence of middle constructions that do not so much specify how the process can be carried out, as indicate that it can be carried out (i.e. its feasibility) or comment on the result of carrying it out. Still other middle structures highlight

Towards a semantic typology of middle constructions 133 the destiny of the Subject-entity by pointing out where it can be placed. Interestingly, as with the facility-oriented type of middle, the semantic properties of the result- and destiny-oriented types of middle formation can be systematically related to the adverbials or the oblique participants which they are used with. Feasibility-oriented middles are characterized by the absence of a specific type of adverb. The facility-oriented reading highlights the fact that the properties of the Subject entity are such that the process can be carried out either easily or with difficulty: (35) The window opened only with great difficulty. (Langacker 1991: 334) (36) [about conditioning milk:] ... rinses easily away and really works! (CB) Facility-oriented middles thus zoom in on the carrying out of the process and specify whether that is hindered or facilitated by the properties of the Subject. Structurally speaking, facility-oriented middles stand out in that they typically take the adverb easily. Other adverbials, such as with great difficulty in (35), are also found, though much less frequently. A type of middle which is closely related to facility-oriented middles in that it also focuses on the process being carried out is that of middles which provide a more general indication of the way in which it can be carried out. This type includes middles which incorporate an adverbial expressing a quality judgement as in (37) and (38), and middles which contain a comparison of quality as in (39) and (40) (Kemmer 1993: 147): (37) That is easily done because the car handles superbly. (CB) (38) If they wanted to do Eliot, why didn't they do The Magi? It's narrative, it reads well.... (CB) (39) The new Holden Berlina handles like a junior sports sedan.... (CB) (40) Her life story reads like the Hollywood movies she's starred in. (CB) Still other middle constructions comment on the properties of a particular entity by pointing out how much time it takes to carry out a certain process on it: (41) "They rolled the green just before the match and it ran three seconds faster," said Curtis .... (CB) (42) [about a cosy car seat protector:] Quickly attaches/removes with elastic straps and velcro tabs. (CB) (43) This title usually ships within 2-3 days, (www.amazon.com)

134

The middle construction

The middle constructions discussed so far all zero in on the effect of the properties of the Subject-entity on how the process can be carried out. However, the nucleus of non-agentive Subject and active verb phrase in the middle can also focus on various other stages. I distinguish between three other possible foci of middle constructions: middles can simply highlight the feasibility of the process (i.e. whether or not the process can be carried out); they can zero in on what the process is destined for; or they can focus on what the result is like when the process is carried out. First, middles of the feasibility type (Davidse 1991: 44) merely focus on whether the properties of the entity construed as Subject make a process possible. They do not specify how this can be done or whether or not it can be done easily. Examples are: (44) [about a cook book holder:] Folds up and packs away for convenient storage. (CB) (45) This umbrella folds up. (Fellbaum 1986: 9) (46) I thought we were out of gas, but the car DRIVES! (Fellbaum 1986: 9) (47) This dress buttons. (Fagan 1992: 57) Examples like these illustrate that the most prototypical cases of middle formation may be those with an adverb (such as easily, welt), but that the presence of an adverb is not an absolute prerequisite: middles which focus on whether it is feasible to carry out a certain process on a particular entity do not seem to be characteristically accompanied by specific adverbs. Some middle constructions contain a locative oblique participant (Fagan 1992: 80). Interestingly, the occurrence of such Locatives appears to account for the destiny-oriented focus which characterizes these middles (Davidse p.c.): (48) [about a children's coat:] Outer flap wraps around little hands and secures with Velcro.... (CB) (49) The purpose of the device is to alert deer of your approach, not the other way around. It is a little whistle that attaches to your car with self-adhesive tape. (CB) (50) The ultimate travel pillow. Resteaz fixes to the headrest providing comfortable head and neck support. (CB) (51) Playset folds up into a storage case with handle for easy carrying. (CB)

Towards a semantic typology of middle constructions 135 Destiny-oriented or 'telic' (Dirven p.c.) middle constructions either focus on where an entity has to be placed to make it function or they merely specify where it can be stored when it is not being used. In either case, the "deliberately designed properties" (Lemmens 1998: 80) that characterize the Subject of many middle constructions are prominently present. Examples of destiny-oriented middles which specify where an entity has to be placed in order to function are those in (48), (49) and (50): the middle in (48) tells the parent who buys the children's coat to wrap the outer flap around the child's hands·, in (49), it is pointed out to the person who buys the whistle that, in order to use it, it has to be attached to the car. Likewise, in (50), it is specified that the travel pillow has to be fixed to the headrest. The middle in (51), then, does not tell you to use the playset by folding it up into a storage case; instead, it indicates that the playset has been designed in such a way that when you do not use it, you can fold it up into a storage case. Notice that destiny-oriented middle constructions often indicate not only where an entity can or should be placed (Location), but also mention how this can be realized (Means), as in (48) and (49) {with Velcro and with selfadhesive tape). Finally, there exist middle constructions which comment on the result of carrying out a certain process on the entity construed as Subject. In accordance with Heyvaert (1997) I will call them result-oriented middles. An example of a result-oriented middle is (52), in which the woman functioning as Subject is characterized as not photographing well. This does not mean that it is difficult to photograph her (facility-oriented meaning), but that, if you photograph her, the pictures resulting from it are hardly ever beautiful. As also pointed out by Fellbaum (1986: 10), the middle in (53) emphasizes that the result of washing the item of clothing was good: there was little shrinkage and no puckering. (52) She does not photograph well.... (CB) (53) [about a piece of clothing:] It washed well with little shrinkage and no puckering. (CB) Notice the difference in meaning between the adverb well as it is used in (52) and (53) and the well of middle constructions such as the following (Fellbaum 1986: 10): (54) This book sells well. (55) This car drives well.

136 The middle construction These middles do not mean that the result of selling the book or driving the car is good. Rather, the meaning of well in (54) comes close to fast or in large quantities, whereas in (55), well can be paraphrased as easily, smoothly. The adverb well, in other words, has various different contextual entailments, which in the middle either trigger off a facility- or qualityoriented reading (as in 54 and 55) or a result-oriented one (as in 52 and 53). Other result-oriented middles are: (56) [about a bag:] It measures 37 χ 60 cm ... and folds up neatly. (CB) (57) She does not photograph well, and the portraits of her pinched features snatched outside the High Court only added to the picture of a shrew. (CB) The middle in (56) emphasizes that the result offolding up the bag is likely to be neat. As in middle constructions with an inanimate Subject in general, it is implied that the Subject-entity has been designed in such a way that this result is possible. The middle in (57) emphasizes that photographing the Subject-entity will probably not lead to nice pictures. The facility-, quality-, feasibility-, destiny- and result-oriented middle structures which I have thus far discussed all 'target' (Langacker p.c.) one particular facet of the complex event that is profiled by the middle construction: feasibility-oriented middles highlight the beginning of the event, facility- and quality-oriented ones its middle stage, while destiny- and result-oriented ones zoom in on the end stage of the event. Corpus data reveal that middle constructions often combine two or more of these foci, and this by integrating adverbs and oblique participants which are characteristic of different semantic types. Consider the following middle constructions: (58) Cards and shapes store neatly in the desk. (CB) (59) Wood valance attaches in seconds, magnetically, to a standard curtain rod. (CB) (60) A full hood, which neatly stows behind the after sets, would make the cockpit a part of the accommodation and keep off the rain or sun. (CB) (61) [about a kitchen timer clock:] Magnetically attaches to any metal surface .... (CB) (62) [about an emergency ladder:] Compact so that it can be stored under the bed or in a wardrobe, the ladder fixes quickly and easily under the window ledge. (CB) (63) Compact lipstick size design, slips easily into your handbag or pocket. (CB)

Middle constructions and modality 137 The middle in (58), for instance, highlights the destiny of the cards and shapes (i.e. in the desk) and it specifies that the result of storing them in it is likely to be neat. The middle construction in (59), then, focuses on the time span in which the process can be carried out (in seconds), the way in which it can be carried out (magnetically) and the destiny of the wood valance (to a standard curtain rod). To conclude, middle constructions may be prototypically patientive and facility-oriented, but they allow for other processes and Subject-entities as well and they can highlight a wide range of facets of the interaction between the non-agentive entity construed as Subject and the process. What seems to characterize all middle constructions, irrespective of the process which they are centred on, the entity which they construe as Subject and the facet(s) which they select for highlighting, is that the properties of the Subject-entity are presented as making (or not making) it conducive to a particular process. This conduciveness, I will argue in the following section, is basically a modal meaning which is realized through the unique constructional link which middles establish between a non-agentive Subject and an active finite verb: it is the specific way in which middle constructions code the relation between a process and a non-agentive entity - be it Patient, Location, Means or Instrument - which typifies the construction and provides the right generalization.

4. Middle constructions and modality It has been argued before that middles imply a modal meaning (Fellbaum 1985; Fagan 1992; Massam 1992; Iwata 1999). In Massam (1992: 122), it is posited that all middle constructions contain what is called a null element "which is in essence equivalent to the modal can". Like the auxiliary can, middles are argued to realize the modality of possibility/ ability (Massam 1992: 124; Fagan 1992: 54). Fagan (1992) thus suggests the following paraphrases for the middle constructions in (64a): (64) a. [about a kind of siding:] It nails easily. It cuts easily. b. It can be nailed easily. It can be cut easily. (Fagan 1992: 54) However, no evidence has thus far been adduced in favour of a modal interpretation of the middle: the claim that middles realize the modality of possibility/ability is only based on the observation that middles seem to be paraphrasable by clauses with can followed by a passive verb. In this section, I will try to provide arguments for the modal nature of middle con-

138 The middle construction structions by analyzing their semantic and lexicogrammatical properties in a more systematic way. First, middle constructions do not imply epistemic possibility. This becomes clear when we add modal forms like possibly or it is possible that to them: (65) Possibly, this pan cleans easily. The meaning of this middle deviates from that of This pan cleans easily, in which no adverbial modal is present: rather than create "double modality" (Halliday 1970a: 330fnl2) and reinforce the meaning of the middle This pan cleans easily, the epistemic modality of possibility that is expressed by the adverbial possibly introduces a new scope and forms a middle which comes closer to the structure (66) This pan may clean easily. in which the epistemic modality of possibility is explicitly realized by means of the modal auxiliary may. Unlike in (65), adding the adverbial modal possibly to (66) results in a double modal structure in which the modal may is reinforced: (67) Possibly, this pan may clean easily. The epistemic modality of possibility, in other words, only appears to be present in those middles which contain an overt indication of it, in the form of a modal auxiliary (e.g., may), a modal adverbial (e.g., possibly), or both. Other attested examples of explicit epistemic modality in middle constructions are the following: (68) Plunge under cold water: the thin elastic membrane and any pieces of fat should peel off easily. (CB) (69) This book could sell. (Roberts 1987: 233) The modal should in (68) realizes the modality of virtual certainty, and, as Roberts points out, the modal could in (69) "can only be epistemic here", i.e. it can only mean it is possible for the book to sell and not the book was able to sell (Roberts 1987: 233). Including the speaker's comment on the possibility or probability of the process is, in short, an option in the system of middle formation that requires the explicit realization of a modal auxil-

Middle constructions and modality 139 iary or adverbial. It does not shed light on the relation between the Subject and the finite verb in middle formation in general. Do middle constructions then implicitly realize the deontic or root modality of ability? Research on middle formation has in general never got beyond the observation that middles can be paraphrased by passive clauses with the modal auxiliary can. As also hinted at in Roberts (1987), the middle itself is unable to contain the modal can in its ability meaning: (70) *This book can sell. (71) *This pan can clean easily. In my view, passive clauses with the modal auxiliary of ability can do not agnate with middle constructions because they cannot be systematically related to them: as I argued earlier, not all middle constructions can be related to a passive construction: (72) *The green was run faster. (73) *That music was danced better. Passive caw-clauses also fail to shed light on the constructional link which middles establish between a non-agentive Subject and an active Finite. The modality of ability, moreover, especially foregrounds the Agent of the process: as a rule, the modality of ability "relates to and is intrinsic to the actor" (Halliday 1970a: 339). In clauses with active voice, the Subject is thus "actor with respect to the modality as well as with respect to the process" (Halliday 1970a: 339): in the clause Jones can drive, Jones is identified as the one who has the ability to drive. In passive clauses the relation of ability is between the implied or fry-Agent and the process, rather than between Subject and process: in, for instance, any problem can be solved, it is not the problem that is characterized as having the ability to solve, but the implied Agent.46 Middle clauses, however, focus not so much on the implied Agent, as on the non-agentive Subject and its properties. In the next section, I will formulate an alternative modal analysis of middle formation in which the middle is viewed as realizing a forcedynamically interpreted type of objective modality. In a first part (Section 4.1), I will briefly elaborate on the specific view of modality that lies behind the analysis which I propose, viz. Langacker's (1991) and Talmy's (2000) force-dynamic approach to modality as well as the view of modality which posits that all clauses realize a schematic form of modality, as described in Chapter 4 (Section 2.3.1; based on Verstraete 2002). In a second part, then, I will give arguments in favour of my interpretation of the mid-

140 The middle construction die, referring to the general semantics of the construction type and to a number of constructions that seem to be systematically related to the middle (Section 4.2).

4.1. The modality of 'letting' The modality that is characteristic of middle constructions can be described as objective, with the Subject as the locus of potency directed towards the realization of the process, not by bringing it about itself, but by letting it be carried out. Where exactly can we situate this form of modality in the overall system of modality? First, Langacker (1991: 269) argues that one of the properties of the historical antecedents of the modern English modals is that their Subject is "the locus of some kind of potency directed at the landmark process, i.e. a physical or mental force that, when unleashed, tends to bring about an occurrence ofthat process" (Langacker 1991: 270). Which are these historical antecedents? Diachronically speaking, the English modals started out as main verbs: the modal adverbial will, for instance, derives from a verb with the meaning of want. Similarly, the modal can originally meant that its Subject had the knowledge or mental ability to do something (Langacker 1991: 269). The original expressions are comparable to such modern English locutions as 'want/know how' to swim. In the case of ' want/know how' to swim, then, it is the Subject which has the potency to unleash a force towards the realization of the process of swimming. The historical evolution from main verb, through root or deontic modality, to epistemic modality can be described "as a matter of the locus of this potency becoming progressively less salient and well-defined" (Langacker 1991: 272). While at the main verb-stage, the locus of potency is the Subject-entity, i.e. "an objectively construed participant accorded maximal salience" (Langacker 1991: 272), the notion of potency has in later stages undergone subjectiflcation, i.e. it has become increasingly implicit. It seems to me, however, that in the 'dynamic' types of modality (i.e. volition or willingness and ability), the conception of potency is still located in the entity construed as Subject. The modality of willingness, for example, as realized by the modal auxiliary will, is always Subject-oriented and expresses the will or willingness of the entity designated by the Subject. In the clause He will do anything to please her, the //¿-person is the one that is willing to please, i.e. that is directed towards the realization of the process of pleasing. Likewise, in the passive clause He won't be helped, it is the Subject-entity that is not willing to be helped (Declerck 1991b: 363). In

Middle constructions and modality 141 clauses which express the modality of ability, it is clear that the potency is located in the Subject as well (e.g., He can play the piano). I propose that the modality of middle constructions comes close to the kind of modality realized in the Modern English dynamic modals. In the middle construction as well, it is the Subject which forms the locus of potency directed towards the realization of the process. However, the potency of the middle Subject is not the one which is typically associated with causation, i.e. it does not itself bring about the process. Rather, the Subject has properties that make it conducive to a particular process: by being construed as Subject of a middle construction, a non-agentive entity is presented as having the necessary properties or potential to let a process be carried out. Interestingly, 'letting' has in Talmy's force-dynamic framework (Talmy 2000: 443) been characterized as being part of the "greater modal system" of language. The letting-foTce which is unleashed by the middle Subject is, unlike the locus of potency that unleashes it, typically realized maximally subjective, i.e. it is not explicitly realized in the clause. I will show that one of the arguments in favour of the presence of an objective, dynamic type of modality in middle constructions comes from the occurrence of a subtype of middle construction in which it is explicitly realized: structures such as The door would not open express "negation of willingness" (Quirk et al. 1985: 229) or the absence of inclination on the part of the Subject. They thus explicitly deny the letting modality that is implied in the middle construction. Like the dynamic modals of willingness and ability, middle constructions realize a kind of modality that is objective·, the modality of letting is internal to the proposition and specifies the representational meaning of the relationship that is established in it. Like dynamic modals, middle constructions therefore also have schematic epistemic modality (as described in Chapter 4, Section 2.3.1, based on Verstraete 2002). This may either be realized as indicative mood, or it takes the form of an explicit epistemic modal, expressing the modalities of possibility, probability or certainty and situating "the process at varying distances from the speaker's position at immediate known reality" (Langacker 1991: 246), as illustrated in (74) and (75): (74) Plunge under cold water: the thin elastic membrane and any pieces of fat should peel off easily. (CB) (75) This book may sell, (personal example) Because it can take explicit epistemic modals, the modality that is realized in the middle comes closest to the periphrastic type of dynamic modals

142

The middle

construction

such as be able to and be willing to, which can also be used with explicit epistemic modals (e.g., He may be able to come, He may not be willing to do it). When middles are used in the indicative (which expresses the speaker's commitment to the truth of the proposition, see Verstraete 2002; Davies, 2001), they can be part of immediate or 'proximal' reality, which typically coincides with the present tense; or they can be situated in non-immediate reality, i.e. in the past (Langacker 1991: 245).47 Examples of past middles are: (76) The lap portion of the car seat belt slotted into place easily, but the belt clip at the back was rather difficult to thread. (CB) (77) "They rolled the green just before the match and it ran three seconds faster," said Curtis .... (CB) Prototypically, however, middle constructions have the present tense and designate a characteristic or state of the Subject-entity, which "is in principle instanced by a number of dynamic situations, but the simple present does not actually refer to any of these" (Declerck 1991b: 182). Because they do not receive a "single event reading" (Declerck 1986: 172), middles like these can be described as generic. Either they designate a characteristic of an individual, as in (78), or they characterize a type or class of entities, as in (79): (78) She does not photograph well.... (CB) (79) The new Holden Berlina handles like a junior sports sedan .... (CB) In short, the modality of 'letting' which is realized in the middle construction can be situated close to the dynamic modalities of willingness and ability: like them, it situates the locus of potency that is directed towards the realization of the process in the Subject; it forms part of the representational meaning of the proposition; it also has schematic epistemic modality and it can be situated in the past. Like the periphrastic modals be willing to and be able to, the modality of letting can, moreover, take explicit epistemic modality. I therefore suggest describing the letting modality of middle formation as representing a dynamic type of modality which designates being conducive to: when an entity is construed as the Subject of a middle construction, it is said to be conducive to a particular process. In Section 4.2, I will go more deeply into some of the characteristics of middle constructions which support their analysis as 'letting' constructions.

Middle constructions and modality

143

4.2. The middle as 'letting' construction Which arguments corroborate the analysis of middle constructions as realizing the meaning of 'letting', viewed as an instance of dynamic modality? First, the letting analysis appears to offer an explanation for the centrality of the Subject-function in middle formation, i.e. the fact that the nonagentive Subject-entity is felt to be responsible for the process and the emphasis that is placed on the role of the properties of the Subject in the carrying out of the process. The middle Subject forms a clear example of what I described in Chapter 4 (Section 2.4) as being the speech functional role of the Subject. The Subject in middle constructions is held responsible for the epistemic modal position that is taken by the speaker: it is pointed out as being responsible for the epistemic position of truth that characterizes middles without explicit modal auxiliaries or adverbials and for the epistemic possibility/probability that is expressed by middles which contain an explicit modal. The (objective) properties of the Subject-entity, however, clearly motivate its being construed as Subject by the speaker. Importantly, this speech functional type of responsibility of the middle Subject should not be confused with causal or agentive responsibility, as has been done in Hale and Keyser (1987) and Van Oosten (1986). The letting analysis entails that the Subject in middle constructions is identified as the locus of the potency or force that is directed towards the process. The middle Subject has that force due to its properties, which are typically physical in nature: they seem to prototypically involve one's ability to manipulate things (e.g., That rack removes easily, That lipstick slips easily into your handbag, This conditioner rinses easily away), but they can also be visual (as in That girl photographs well), auditory (e.g., That music dances better) or they can involve one's sense of taste (e.g., That wine drinks like water, This chicken broils well). An example in which the middle Subject's potency can, in one reading, be interpreted as primarily mental in nature is That novel reads well. All middle constructions, including the less prototypical ones, are thus centred on the potency of the Subject. The Subject's potency and the fact that that potency is directed towards the realization of the process is what was characterized earlier as the conduciveness of the middle Subject to the process. Importantly, the relationship of conduciveness that characterizes middle constructions is not to be interpreted as one of agentivity, but as one of 'letting'. The middle Subject is not itself the energy source of the process (it is not agentive), but it is or is not - conducive to the implied Agent carrying out the process. This approach remedies the uneasy vacillation between the straight agentive and patientive interpretations that are found in the literature on middle forma-

144

The middle

construction

tion (discussed in Section 1 of this chapter as the ergative and the passive interpretations of middle formation). The non-agentive entity construed as Subject of a middle structure is a letting-entity or a conducive entity, i.e. one whose properties either let through the energy flow instigated by the Agent, or blocks it. Conduciveness, and consequently also the notion of 'letting', can be argued to constitute general properties of middle constructions in that they schematize over the various process types and Subject-entities which the middle can take, as well as over the wide range of facets of the interaction between them which middles can select for highlighting. The schematic characterization in terms of letting and conduciveness counterbalances the traditional focus on the prototypical middle construction with a facilityoriented profile and a patientive Subject. It constitutes a schema which is instantiated by all middle constructions, prototypical as well as nonprototypical. In, for instance, the green ran faster in (80), the properties of the green are claimed to have let one run faster on it; the middle in (81) conveys that the properties of the ladder are such that it lets one fix it quickly and easily under the window ledge. Likewise, the middle in (82) expresses that the properties of the Subject-entity are such that it lets the process of washing be carried out with a good result. The umbrella in (83), finally, is characterized as letting one fold it up. (80) "They rolled the green just before the match and it ran three seconds faster," said Curtis .... (CB) (81) ... the ladder fixes quickly and easily under the window ledge. (CB) (82) It washed well and didn't shrink much .... (CB) (83) This umbrella folds up. (Fellbaum 1986: 9) Additional evidence in favour of the letting analysis of middle formation comes from two related constructions which pick up on the meaning of letting. A first related construction consists of what is basically a middle (it has a non-agentive Subject and takes an active verb form), but then with the negative modals won't and wouldn't. It is illustrated in the following examples: (84) ... Mr. Fabius chose to exit by a false door. ... the scenery began to shake in response to the ex-prime minister's desperate efforts, but still the door would not open. (CB) (85) ... Mrs. Dambar was having some difficulty with the new dress she had bought .... For some reason, it wouldn't zip all the way up in the back. (CB)

Middle constructions and modality

145

(86) I tried to open the door, but the key wouldn 't turn. (Quirk et al. 1985: 229) (87) I am at α sentence that will not write. (Jespersen 1914-1929, 3: 349) It has been pointed out in the literature that the negation of will/would in these clauses seems to "have something of the personificatory force of 'refusal'" (Quirk et al. 1985: 229; see also Ehrman 1966; Huddleston 1969): mrs. Dambar's new dress in (87), for instance, as it were refuses to be zipped. The modal auxiliary won 't/wouldn't in this use denies the typical, letting properties of the Subject and shifts the focus to the fact that the Subject instead hinders the implied Agent in carrying out the process. When the modal will in middle constructions is followed by internal negation, its meaning seems to move more towards the epistemic, even though one can also argue that there is a notion of preventing or resisting something that is not desired, as in (88) and (89):48 (88) Just press and personalize clothes in minutes with clear, legible imprints that will not wash out. (CB) (89) Place gifts in cardboard boxes that won't crush easily. (CB) Interestingly, the notion of 'hindrance' which is, in some cases, implied in the use of the modals won't and wouldn't can also be realized by the lexical verb refuse: (90) ... although the key turns, the door will refuse to lock (or to unlock). (CB) (91) ... when the matches refused to strike. (Jespersen 1914-1929, 3: 348) These constructions may be considered as variants of middle clauses with won't and wouldn't (as in 92 and 93), but then with the notion of hindering which the latter imply realized lexically: (92) The door won't lock. (93) The matches wouldn't strike. Another construction which helps to shed light on the middle construction is the laat zieh 'let itself - construction in Dutch. Dutch has a construction type that makes use of the auxiliary laten 'let', followed by the reflexive pronoun zieh 'itself and an active infinitive (the examples were extracted from the IDL corpus):

146 The middle construction (94) GSM-antwoordapparaat laatzich gemakkelijk kraken. mobile phone-answering machine lets itself easily crack 'mobile phone-answering machine cracks easily' (95) Haar veelbewogen Hollywood-leven laat zieh lezen als een spannend meisjesboek. her turbulent Hollywood-life lets itself read like an exciting girl's book 'her turbulent Hollywood-life reads like an exciting girl's book' Significantly, this construction type has many features in common with middle formation: the participant functioning as Subject is non-agentive; the verb form is active and it is usually accompanied by some modification of how the process can be carried out with respect to the Subject-entity, but it may just as well express mere feasibility, as in the following examples (from the LDL corpus): (96) De uitkomsten daarvan laten zieh als volgi samenvatten .... the results of that let themselves as follows summarize 'the results can be summarized as follows ....' (97) Een beetje bitter, maar het laat zieh drinken. somewhat bitter, but it lets itself drink 'even though it is somewhat bitter, it drinks' In Fagan (1992: 211), the German variant of this type of 'let itself construction is elaborately discussed and classified as a type of middle construction: (98) Der Wagen läßt sich angenehm fahren, that car lets itself pleasantly drive 'that car is pleasant to drive' The Dutch 'let itself - construction is mentioned and likewise classified as a type of middle structure. I prefer to treat it as a close agnate of middle formation. Interesting about it is that it lexicalizes the notion of 'letting' which is claimed here to be implied in all middle constructions. It appears as if in Dutch, the 'letting' aspect of middle formation can either be realized highly subjectively (implicitly), or objectively, i.e. in the form of a lexical verb.

Conclusion 147

5. Conclusion I have suggested that the existing analyses of middle formation can be refined in three ways. First, the construction's representational potential can be mapped out more accurately and turns out to include also intransitive processes. Secondly, middle constructions vary according to which facet of the designated event that they highlight: in addition to facility-oriented middles, also feasibility-, destiny- and result-oriented middles can be distinguished and linked up with the occurrence of specific adverbs or oblique participants. And, finally, the link between a non-agentive Subject and an active Finite, rather than restriction to a specific verb class, typifies the middle construction. The essentially modal value of the relationship which middles establish, and, more particularly, the letting force-dynamic status of it, allows us to siring all the characteristics of middle formation together. The middle's emphasis on the properties of the Subject-entity ties in with the Subject being the locus of the potency of the letting modality; the fact that the middle profiles a letting relationship explains the meaning of conduciveness which all middles have in common, as well as the non-agentive nature of the Subject and the need for an implied Agent to actually carry out the process. By being construed as Subject of a middle construction, a non-agentive entity thus acquires speech functional responsibility and its properties and the fact that they let one carry out certain processes (in a particular way) are highlighted. In Chapter 7,1 will show how these properties of the clausal middle construction shed light on the category of nonagentive -er nominalization and, ultimately, on the category of deverbal -er derivation itself.

Chapter 7 A multifunctional approach to -er nominalization Within a schematic network ..., certain nodes and relationships are far more prominent and important than others, both cognitively and linguistically. In particular, special significance attaches to the nodes that function as the category prototype and as the highest-level schema. The prototype is significant because of its developmental priority and notable cognitive salience. As the primary basis for extension, it defines the center of gravity for the category. The highest-order schema is significant because it embodies the maximal generalization that can be extracted as a characterization of the category membership. (Langacker 1987a: 380-381)

0. Introduction The analysis of -er nominalization which is presented here is aimed at identifying the high-level schema which deverbal -er nominalizations instantiate. It essentially posits that deverbal -er suffixation implies a relationship between an entity and a process which is strikingly similar to the SubjectFinite relation in the clause. Like the clausal Subject, the entity which is profiled through -er nominalization is either agentive or non-agentive; like the Finite in clauses, deverbal -er suffixation can ground the process type which it nominalizes in a variety of ways. As to the actual profile of -er derivations, I will argue that it is equivalent to that of the clausal Subject. It is the function of Subject that will be identified as the high-level schema which all deverbal -er nominalizations realize. Nominalizations which at first sight seem to resist a Subject interpretation are those designating non-agentive entities. Among the arguments that have been raised against viewing them as Subject-profiling are, firstly, that there are no grounds for assuming that a morphological process such as -er nominalization can have a patientive entity as Subject; that oblique participants such as Locatives and certain Instruments are not easily construed as Subjects; and, finally, that not all non-agentive -er nomináis can be related to a middle construction.49 The hypothesis which I put forward is nonetheless that deverbal -er nominalization originated as an agentive system, but that the productivity of non-agentive nomináis suggests that it has now turned into a Subject-oriented system. The rearrangement of options which seems to have taken place within the network of -er formations may well

150 A multifunctional approach to -er nominalization have run parallel with the increased productivity of clausal middle formation, and even with the Subject-centred nature of contemporary English as such. Possibly, the rise in -er nomináis based on nouns and other, more complex expressions (such as hardliner, down-and-outer, lefthander) will constitute another factor of instability within the system of -er nominalization in general, which will again refashion the system and may have repercussions for deverbal -er nominalization as well. But for now, the Subject generalization seems to be the most accurate generalization for deverbal -er nominalizations. I claim that non-agentive -er nominalizations can be accounted for in terms of subjecthood, provided that the link between non-agentive -er nominalization and middle formation is recognized for what it is, viz. as a relationship of agnation which is based on a fundamental correspondence between the speech event-related choices made within the Subject-Finite unit of the middle and those made in non-agentive -er nominalization. Middle constructions are thus much more fundamental to the analysis of non-agentive -er derivation than has been suggested thus far (in Levin and Rappaport 1988; Rappaport Hovav and Levin 1992; Lemmens 1998). To elucidate their role, however, it is essential that their own Subject-Finite structure is interpreted correctly, i.e. as being centred on an active finite verb and allowing for a wide range of verb types, including intransitives. If non-agentive -er nomináis are related to middle constructions, then, this would solve two things. Firstly, non-agentive -er nomináis construing a Patient are not linked to a verb in the passive voice. Secondly, oblique participants such as Locatives also fall within the generalization and can be said to relate to the Subject function of the middle. The relationship of agnation between middle formation and non-agentive -er nominalization should, moreover, be seen as a basically schematic relation between systems. Non-agentive -er nomináis instantiate schemata which are partly identical to those instantiated by middle formation. Each system, however, also has its own schemata, which account for the unique character of its instantiations. In short, if analyzed precisely, the relationship of agnation between the middle construction and non-agentive -er nominalization can refute all arguments raised against a Subject generalization for deverbal -er nominalization. To map out the overall schematic network of deverbal -er suffixation and unveil the schematic choices which it has in common with the Subject and the Finite in clausal constructions, a multifunctional analysis is required: the representational semantics of -er nominalization has to be integrated with the constructional (or, in Halliday's terms, the interpersonal or speech-event related) status of the relationship which -er nomináis imply

Representational semantics 151 and with their textual functioning (see Chapter 4, Section 3). It is only through a multifunctional analysis that we can string together the form and meaning aspects of -er nominalization and come to an account of its properties which is truly natural. The outline of my analysis is as follows: first (in Section 1), I will give a brief overview of the representational options which -er nominalization offers and list some of the subschemata which can be said to instantiate the schematic categories of agentive and non-agentive deverbal -er suffixation. In a second part, I propose a semantic typology of non-agentive -er nominalization, modelled after that of middle clauses (Section 2). In the third and most substantial - section, I present an elaborate account of the constructional relationship which -er nomináis establish between an agentive or non-agentive entity and a process type. I go more deeply into the distinction between lexicalizations and ad hoc nominalizations and point to the different kind of interpersonal relation which they imply between an entity and a process and the differences in textual functioning that seem to correlate with it. In a fourth and final section, then, I elaborate on the Subjectgeneralization which I propose for the system of deverbal -er nominalization.

1. Representational semantics Depending on the process type that is nominalized, -er nomináis profile different participant roles. At the extreme, "every verb defines a distinct set of participant roles that reflect its own unique semantic properties" (Langacker 1991: 284). If we distinguish between agentive and non-agentive -er nominalizations, then, this is necessarily very much a schematic distinction, which covers a wide range of profiles. In Systemic-Functional Grammar, it has, for instance, always been emphasized that not all two-participant clauses can be analyzed in terms of a 'causation' model, and that, strictly speaking, only clauses which realize an action model and take 'material' verbs can be said to construe an Actor-Goal type of semantics. Material processes or 'processes of the external world' prototypically designate 'outer experience', i.e. actions and events. They are to be distinguished from mental processes or processes of consciousness, related to "inner experience" (Halliday 1994: 107), and relational processes, which relate two items of experience to one another, either by classifying ('this is a kind of the other') or by identifying ('this is the same as that') (Halliday 1994: 107). The participants that are involved in relational and mental process clauses are 'Agents' and 'Patients' only in a very general, schematic sense

152

A multifunctional approach to -er nominalization

and more refined characterizations can be provided for them: in a mental clause such as I admire you, for instance, the /-participant can be thought of as the Senser and the Fou-participant the Phenomenon (Halliday 1994: 117). In a relational process clause such as John is the leader, John can be identified as the Token and the leader as the Value which is supplied for it. In short, if in my analysis of -er nominalizations, I use the notions of 'agentive' and 'non-agentive', it is implied that they cover a whole range of entities which differ depending on the type of verb which the nominalization derives from. In the next section, I will sketch the main options within each schematic category. My analysis elaborates on Lemmens' (1998) account of the transitive and the ergative paradigms and their impact on -er nominalization. The criteria for discriminating between the various subtypes which I discuss are based on Hopper and Thompson (1980), Davidse (1991) and Halliday (1994).

1.1. Agentive -er nominalization An -er nominalization can be high in transitivity and designate the volitional, human Agent of a transitive process with an affected Object (e.g., carver, cutter). -Er suffixation can also foreground the Agent of a process that is lower in transitivity, but still two-participant (e.g., writer, dancer). In the latter case, the processes underlying the nominalization are 'pseudoeffective' (Davidse 1991), i.e. they can contain an Object, but only one which specifies the "range or scope of the process" (Halliday 1994: 146), as in write a book and dance the Charleston. The book and the Charleston are not 'done to' (*What he did to the book was write it, *What they did to the Charleston was dance it). When an agentive -er nominalization derives from an ergative process, it profiles the Instigator of the process and another entity is understood to be co-participating in the process (e.g., burner, opener, freezer). Finally, an -er nominalization can also profile the Agent of a truly one-participant or intransitive process (e.g., sinner, laborer). Interestingly, some -er nomináis designate an Agent which does not itself carry out the process, but causes someone else to carry it out, such as yawner in Normally, certification of the votes by the Secretary of State is a yawner (CNN-news, 27 November 2000) and pageturner 'a well-written, exciting book' (for more examples of this type, see Ryder 1999b: 276). They designate what is actually a second-order Agent (Halliday 1994: 172): they focus not so much on the Agent which carries out the process, but on the entity which causes it to act. A pageturner, for instance, is a book that is so good that it almost forces the reader (i.e. the first-order

Representational semantics

153

Agent) to turn the pages. Similarly, a weeper is a movie that makes you weep and may, in fact, have been designed to make you weep (Panther and Thomburg 2002: 181). These constitute the main subtypes of agentive -er nominalization. Cross-cutting this classification and further refining it are various other distinctions. One can, for instance, distinguish between human Agents (e.g., baker, preacher), animals (e.g., wood-pecker, retriever), inanimate agentlike entities or Instruments which designate material processes (e.g., toaster, stapler, scraper) and inanimate Agents based on immaterial processes (e.g., pointer 'a piece of advice or information which helps you to understand something or find a way of making progress', reminder, thriller) (based on Marchand 1969; see also Panther and Thornburg 2002). Somewhat more peripheral in terms of agentivity, but nonetheless closer to the agentive than to the non-agentive group of -er nomináis are those nomináis that derive from mental processes or processes of sensing (e.g., admirer, lover, believer) (Halliday 1994). They designate the Senser of the mental process which they are based on. Nomináis like container and holder are derived from relational processes expressing possession and part-whole relations and belong to the agentive rather than to the nonagentive schematic category because they do not designate the things that are in the container/holder, but the entity which contains/holds them. The schematic category of agentive -er nominalization, in short, spans a wide range of participant roles.

1.2. Non-agentive -er nominalization Non-agentive -er nomináis fall apart into two main categories: there are those that designate a patientive participant, and there are -er nomináis that profile oblique participants such as Locatives and Instruments (for the notion of oblique participant, see Chapter 5, Section 1.2.1). Within the first group, a further distinction can be made between nomináis that profile prototypical Patients or inert, affected entities (e.g., scratcher, squeezer), nomináis that designate entities which merely specify the Range of the process (Halliday 1994) (e.g., reader, leaner), and nomináis which profile the Medium or central participant of an ergative process, which is not purely patientive but coparticipates in the process (e.g., broiler, cooker, steamer).50 The other group of -er nomináis can be further divided into, firstly, nomináis profiling oblique locative participants (e.g., kneeler 'a kind of chair which you have to kneel on', stepper 'a raised surface on which you

154 A multifunctional approach to -er nominalization have to step as a way of working out' Jotter 'something which you can jot things on or in'); and, secondly, nomináis designating instruments which lack agent-like properties, such as walker, stroller and viewer. Without doubt, more and finer-grained representational categories can be distinguished among agentive and non-agentive -er nominalizations. This is, however, as far as I will take the representational description: the main aim of my analysis is to integrate the overall representational semantics of deverbal -er nominalization with the system's constructional properties. In the following section, I will elaborate on the semantics which all non-agentive -er nominalizations have in common, and I will propose a semantic typology of non-agentive -er derivation much along the lines of that which I set out for middle constructions in Chapter 6. The semantic correspondences between non-agentive -er nominalization and middle formation will then in Section 3 be shown to derive from a more fundamental, lexicogrammatical equivalence between the Subject-Finite unit in middle constructions and the relationship which deverbal -er suffixation establishes in non-agentive nomináis.

2. A semantic typology of non-agentive -er nominalizations By turning a non-agentive entity into the Subject of an active Finite, middles were shown to foreground the conduciveness of the Subject-entity towards the Agent's carrying out of the action. I claim that the link which non-agentive -er nominalizations establish is based on a largely equivalent mechanism: the constructional link which the -er suffix imposes on the representational relation between entity and process also leads to a general meaning of conduciveness which characterizes non-agentive -er nomináis. In this section, I will discuss the various semantic profiles which nonagentive -er nominalizations can have. In Section 3, I will show that the semantic correspondences between middle formation and non-agentive -er nominalization (as hinted at in Levin and Rappaport 1988 and Lemmens 1998) are more systematic than assumed till now and shed a new light on various non-agentive -er nominalizations that were thus far considered problematic. In general, non-agentive -er nominalizations designate entities of which the properties are conducive to a specific process. As pointed out by Lemmens (1998: 139), in the case of non-agentive -er nomináis these characteristics often "almost predetermine" the profiled entities for a particular process: as with the middle construction, the properties of the profiled entity have typically been designed or intended for a particular process, but much

A semantic typology of non-agentive -er nominalizations 155 more than middle clauses, -er nomináis focus on this process, which forms the ultimate "destination" of the profiled entity (Lemmens 1998: 138). Think, for instance, of Levin and Rappaport's (1988: 1078) analysis of broiler as 'a chicken bred to be broiled'. Another example is killers, designating 'cattle ready for slaughter' (Oxford English Dictionary on CD-rom, as cited in Lemmens 1998: 138): (1)

As killing cattle or killers (cattle ready for killing) they are inferior to corn-fed stock, (attested in 1937, Supplement to the Oxford English Dictionary 1972)

As pointed out by Lemmens (1998: 138), killers "is not to be interpreted as indicating that these animals possess qualities that facilitate the actual killing". Rather, killers refers to animals that have been fattened in a particular way, so that they are suitable for being killed for their meat. The orientation towards the process which they have been designed/bred for seems characteristic of the large majority of non-agentive -er nominalizations and goes hand in hand with the nomináis' overall emphasis on the properties of the profiled entity (see also Panther and Thornburg 2002). I claim that the semantics of non-agentive -er nominalizations can be further differentiated according to the facet(s) of the interaction between the nonagentive entity and the process which they profile. This means that middle formation and non-agentive -er nominalization show interesting parallels in terms of the lower-level schemata which they instantiate (for the distinction between high-level and low-level schemata and its relevance to nominalizations, see Chapter 3, Section 1.1). Unlike middle constructions, in which adverbials and prepositional phrases can provide indications as to which facet of the designated event is highlighted, however, non-agentive -er nominalizations are typically characterized by a certain indeterminacy with respect to the facet that they focus on. Most nominalizations do not contain any overt indication as to whether they highlight the beginning (feasibility), the middle (facility or quality) or the end (destiny or result) of the event. As a result, they are open to more than one interpretation or they combine various facets of the designated event in their semantics. Only two groups of non-agentive -er nominalizations are more explicit as to which facet they zero in on: one of them consists of nomináis such as bestseller, easy-rider and easy-walker which specify how (well) the properties of the profiled entity let the process be carried out. The other group includes the nomináis front-loader, top-loader and low-loader and zooms in on where the properties of the profiled entity allow the process of loading to be carried out. First, when a book or other item is called a bestseller,

156

A multifunctional approach to -er nominalization

what is being referred to is that it sells well/ easily: the -er nominalization picks up on the facility- or quality-oriented meaning which is prototypically associated with middle formation and specifies how easy it is to carry out the process on the entity construed as Subject or how fast it is sold. As pointed out by Lemmens (1998: 139), even a nominal like seller, in which the modal adverbial has been left out, is oriented towards the facility/speed with which the action of selling is carried out: if you describe a particular thing as a seller, it is implied that it sells well. Two other examples of nonagentive -er formations that highlight the way in which a process can be carried out are easy-rider and easy-walker. An easy-walker is the latest type of stroller, which has properties that facilitate walking with it because it has only three wheels which are highly manoeuvrable. An easy-rider is a kind of motorbike which rides easily or smoothly, as in (2)

The 1.41 rides smoothly while the 1.6 is much firmer. (CB)

Nomináis such as top-loader, front-loader and low-loader, then, do not focus on the ease or difficulty with which the process of loading is carried out, but rather specify where a particular entity has to be loaded to make it serve the function it has been designed for: front-loader and top-loader refer to a type of washing machine or clothes drier which has to be loaded by inserting the clothing at the front or from the top. A low-loader designates a kind of vehicle of which the carrying platform is low, so that you can load things onto it from a low position. Of course, the fact that what is loaded onto the low-loader does not have to be lifted to a high position greatly facilitates the loading. That is not, however, what the -er nominal highlights: like front-loader and top-loader, low-loader primarily indicates how the profiled entity can be put to use, by specifying where the loading has to take place. As such, these nomináis resemble the destiny-oriented type of middle construction, which was argued to be typically accompanied by a Locative, as in (3): (3)

It is a little whistle that attaches to your car with self-adhesive tape. (CB)

Most non-agentive -er nominalizations do not specify which facet they foreground. Some of them seem to combine feasibility and facility/quality·. a viewer, for instance, enables you to view photographic transparencies (more easily); a walker is a frame which helps babies or disabled or ill people to walk (more easily); a stroller allows you to take your baby out and stroll (comfortably) with it; a sleeper in a train is a bed or sofa in which you

A semantic typology of non-agentive -er nominalizations 157 can sleep (well); a train that is referred to as a sleeper is a train which has all the facilities to let travellers sleep.51 The various non-agentive -er nomináis that designate items of clothing also seem to profile especially if {feasibility) and how (facility) they allow a certain process to be carried out: waders are boots with which you can wade; slippers are heelless shoes which one can slip into easily.52 Other non-agentive -er nominalizations also express feasibility, but especially foreground the fact that undergoing the process on which the nominalization is based is what the profiled entity has been designed for (Lemmens 1998). It seems to me that nomináis such as kneeler, squeezer, jotter and stepper especially foreground the processes which the entities that they profile have been designed for. A kneeler thus designates a type of chair which is designed in such a way that you can kneel on it; a squeezer is a kind of bottle that one has to squeeze to extract something from it; a jotter is meant to jot things down in; a stepper is a raised surface that one has to step on as a way of working out. A final group of non-agentives, then, can be argued to resemble the result-oriented type of middle formation. I am thinking here of the -er nominalizations that profile items of food, such as cooker, steamer, broiler and fryer. They designate animals and vegetables that have, in many cases, been bred or raised to be prepared in a certain way.53 But they also imply that, if a particular process is carried out on the item of food in question, the result will be good. Middle structures such as those in (4), which Levin and Rappaport (1988: 1078) associate with non-agentive -er nomináis referring to food, primarily hint at how to prepare the food to achieve the most delicious or the best possible result: the adverbial well is used here in its resultoriented meaning. A boiler, for instance, designates a type of chicken that is best boiled and used for soup; a cooker is an apple which is best cooked and used for, for instance, applesauce. (4)

this type of chicken broils well: broiler this type of apple cooks well: cooker this type of clam steams well: steamer this type of chicken/ rabbit fries well: fryer

In short, the semantic distinctions that were introduced for the analysis of middle formation seem pertinent to the description of non-agentive -er nominalization as well. Importantly, they elucidate what all non-agentive -er nominalizations have in common, viz. the conduciveness of the profiled entity towards a specific process and the fact that the properties of the profiled entity are responsible for it. It is the properties of the easy-walker that make it easy to walk with it; it is because it has an opening at the front, that

158 A multifunctional approach to -er nominalization a front-loader can be loaded via the front; a viewer has properties that allow you to view photographic transparencies with it; a squeezer is squeezable, i.e. it has properties which enable you to squeeze it when you have to extract something from it; a broiler has certain properties which turn broiling into a good way of preparing it. Like middle clauses, non-agentive -er nominalizations may even foreground specific facets of the interaction between the process and the non-agentive entity which they profile, ranging from the feasibility of the process, the facility or speed with which the process can be carried out, and the place it is oriented towards, to the result of carrying it out. However, since few non-agentive -er nominalizations explicitly indicate the facet which they want to highlight, vagueness with respect to the distinction between facility-, destiny- and result-orientation seems to be part of the system of non-agentive -er nominalizations.

3. The constructional properties of -er nominalization In this section, I will show that, if the lexicogrammatical properties of agentive and non-agentive -er nominalizations are considered more closely, evidence can be provided for the claim that it is the Subject profile which constitutes the basic commonality between agentive and non-agentive -er nominalization. Importantly, cutting across the constructional properties of deverbal -er nominalizations is the distinction between lexicalizations and ad hoc nominalizations: lexicalizations can be agentive or non-agentive in profile, while ad hoc nominalizations are always agentive (Levin and Rappaport 1988). Before we try to offer a comprehensive account of the system of deverbal -er nominalization in English, therefore, the lexicogrammatical commonalities between non-agentive -er nominalizations and agentive lexicalizations have to be analyzed and the properties of ad hoc nominalizations have to be accounted for. I will first look into non-agentive -er nominalizations and consider their relation with agentive lexicalizations. In Section 3.2, the category of ad hoc nominalization will be described. It is not until Section 4, then, that I consider the system of deverbal -er suffixation in its entirety and elaborate on the generalization which I propose for it.

The constructional properties of -er nominalization 159 3.1. Lexicalized -er nominalizations 3.1.1. Non-agentive -er nominalization In this section, I want to put forward the claim that the semantic correspondences between non-agentive -er nominalization and middle formation which I described in Section 2 reflect a more fundamental level of correspondence between both construction types. I argue that non-agentive -er nominalizations establish a relation between a non-agentive entity and a process which is equivalent to that of the Subject-Finite unit in middle clauses: like middle clauses, non-agentive -er nomináis designate a relationship that is characterized in terms of objective, dynamic modality, and, like middle constructions, non-agentive -er nominalizations have a schematic, epistemic modal claim mapped onto this relationship: (5) (6) (7)

broiler bestseller easy-rider

'a chicken that broils well' 'a book that sells well' 'a motorbike that rides easily'

First, I propose that the semantics of conduciveness which characterizes non-agentive -er nominalizations derives from the same type of modality as that expressed by the middle. I argued in Chapter 6 (Section 4) that the relationship which middle constructions establish between a non-agentive entity and a process is one of objective modality (Halliday 1970a): rather than construing the speaker's position with regard to obligation or likelihood, the modality realized in middle constructions was argued to form "part of the content of the clause, expressing conditions on the process referred to" (Halliday 1970a: 338). It was moreover classified as an instance of dynamic objective modality, because, like the Subject of clauses with the dynamic modalities of volition and ability, the middle Subject forms the (objectively realized) locus of potency directed towards the realization of the process. Unique about the middle construction, then, was that the middle Subject does not bring about the process itself, but lets an implied Agent bring it about. Thus, the meaning of conduciveness that typifies middle constructions was argued to derive from the objective, dynamic letting modality expressed by their Subject-Finite unit. Non-agentive -er nominalization can in my opinion be argued to instantiate a type of objective, dynamic modality which is similar to that of middle formation. As in middle constructions, this modality is objective or part of the ideational content structure of the non-agentive -er nominalization, where it significantly adds to the semantics of the process-entity constella-

160 A multifunctional approach to -er nominalization tions that are designated and serves to derive a more elaborate meaning: a cooker is therefore more than an apple that cooks and a broiler is more than a chicken that broils. In particular, the objective type of modality that is realized in non-agentive -er nomináis comes close to the dynamic modalities of volition and ability, because it relates a process to an entity which is conceived as the locus of potency of the modality: the profiled entity in non-agentive -er nominalizations may be non-agentive, but it is the locus of potency of the process because it lets a process be carried out, i.e. it is conducive to a specific process. A kneeler is thus conducive to kneeling, i.e. it enables you to kneel on it; a sleeper is a train that has beds that let you sleep; a walker is a frame that makes it possible for you to walk, it lets you walk. Importantly, these correspondences with the middle's Subject-Finite complex suggest that non-agentive -er nominalizations can be analyzed as profiling what is clausally construed as the locus of potency of a dynamic modal relationship, i.e. they profile a subject-like entity. Like middle constructions, non-agentive -er nomináis moreover imply a schematic, epistemic modal claim or position in addition to the objective modality of their propositional content (see Chapter 4, Section 2.3.1). As in middle structures, the epistemic modal position of the speaker in nonagentive -er nominalizations relates to the 'truth' of the proposition: the relationship which is designated by non-agentive -er nomináis is presented as 'known to be true' and it is the profiled entity that is held responsible for the truth of it (Chapter 4, Section 2.4). In particular, non-agentives imply that the entity which they focus on has properties that make it 'modally' or 'speech functionally' responsible for the truth of the proposition: the fact that the profiled entity is conducive to a particular process and lets it be carried out is, as in middle constructions, attributed to the properties of the profiled (Subject) entity. Kneelers, walkers and sleepers thus have properties that make them conducive to the processes of kneeling, walking and sleeping. The fact that non-agentives imply that the modal responsibility of the relationship which they designate is assigned to the entity which they profile is reminiscent of the modal responsibility of the Subject in the clause. It therefore provides additional support for the claim that nonagentives relate to the clausal Subject profile. It is the modal value of the relationship which non-agentive -er nominalizations imply which explains why they are so-called 'non-event' nominalizations, i.e. why they do not presuppose that the process which they nominalize has actually taken place (Levin and Rappaport 1988; Rappaport Hovav and Levin 1992). Something can be called a squeezer before anyone has actually squeezed in it precisely because the -er nominal makes a modal claim about the entity which it profiles: the entity is presented as having

The constructional properties of -er nominalization

161

properties that make it conducive to a particular process, but, crucially, this does not entail that this process has already been carried out. In section 3.1.2, I zoom in on agentive lexicalizations. Agentive -er nominalizations do not present a problem for the Subject-generalization which I propose: Agents are prototypically mapped onto the clausal Subject (see, for instance, Langacker 1991; Halliday 1994). However, it is interesting to see that the properties of agentive lexicalizations support the analysis which I have suggested for non-agentive -er nominalizations: like nonagentive -er nominalizations, agentive lexicalizations imply a relationship between a (Subject-) entity and a process which is basically modal in nature.

3.1.2. Agentive -er nominalizations Agentive -er nominalization is typically associated with occupational terms such as teacher, baker, preacher, life-saver, fire-fighter, trainer. Levin and Rappaport (1988) classify these occupational terms as 'non-event' nominalizations. They argue that many of them can be used to refer to people "before they have engaged in the activity denoted by the related verb" (Levin and Rappaport 1988: 1070): one can, for instance, be a life-saver and not have saved any lives yet. What seems to lie behind Levin and Rappaport's characterization of agentive lexicalizations as non-event nominalizations is, once again, the modal nature of the relationship which agentive -er nominalizations realize. It appears to me that, like non-agentive -er nominalizations, agentive lexicalizations tend to establish a relationship between an entity and a process which is basically modal in nature. The human agentive prototype of -er nominalization (which diachronically speaking is the oldest type of -er nominal) seems to instantiate types of modality which come close in meaning to what Langacker (1991) identifies as the historical antecedents of the modern English modals. Especially the antecedent of can, which indicated "that its subject had the knowledge or mental ability to do something" and that of may, which specified that the Subject "had the necessary strength or physical ability" (Langacker 1991: 269) are of interest to the analysis of human agentive lexicalizations such as baker, fisher, engraver, wood-cutter and painter, all of which are found in Old English (Kastovsky 1971). Importantly, many Modern English agentive -er nominalizations still assign some degree of ability to the Agent which they profile, irrespective of whether this is a human being, an animal or an inanimate entity. Like clauses that realize dynamic modality, agentive -er nomináis moreover add

162 A multifunctional approach to -er nominalization a schematic or epistemic claim to the objective modality of the relationship which they designate, presenting it as 'known to be true to the speaker'. As in clausal structures, the profiled entity is held responsible for the truth of the proposition: it is the properties of the profiled entity which are responsible for the fact that the relationship of ability is 'true'. A life-saver thus refers to a person who knows how to save lives·, a fire-fighter designates someone who knows how to fight a fire. A retriever is able to retrieve birds and other animals which hunters have shot; a warbler is a kind of bird which can warble or sing pleasantly. Likewise, a transmitter can transmit television and radio programmes and a toaster can toast bread. It is the modal nature of the relationship which agentive nomináis like these establish which accounts for their so-called 'non-event' reading: a life-saver is essentially a type of person who has the ability or knows how to save lives. He or she need not have saved any life yet. Notice that it is implied in many agentive lexicalizations that the profiled entity brings about the process regularly, a baker is someone who knows how to bake bread, but will not be called a baker unless he or she bakes bread regularly (professionally). When referring to animate entities, agentive -er nominalizations in general tend to designate characteristic or habitual behaviour: apart from the dynamic modality of ability, these nomináis seem to make the epistemic claim of 'permanent habit or characteristic' which can be realized by the clausal indicative as well (e.g., He bakes bread can be understood as expressing a characteristic of the Subject, Declerck 1991b: 183). Nominalizations such as drinker, gambler and woodpecker, then again, characterize the profiled Agent in terms of a persistent habit: a drinker is someone who will drink; a gambler is someone who will gamble. They thus realize the dynamic modality of "strong volition" (Declerck 1991b: 361). Importantly, the type of modality that seems to be characteristic of agentive -er nominalizations in general is dynamic and situates the locus of potency which it implies in the profiled (Subject-) entity: in agentive -er nominalizations, the profiled entity is presented as the locus of potency directed towards the realization of the process. Unlike in non-agentive -er nominalizations, however, the potency associated with the profiled entity of agentive nominalizations does not let someone else carry out the process, but it brings the process about itself. In Talmy's terms, rather than implying the force of letting, agentive -er nominalizations thus tend to realize that of causing which is linked to the modalities of volition and, especially, ability. As I pointed out earlier, the causality of agentive -er nominalizations can be either first-order or second-order, as in laugher 'a movie or book that makes people laugh' and pageturner 'a book that makes you turn the

The constructional properties of -er nominalization 163 pages'. Rather than being a direct causer, the profile of an agentive -er nominalization may also be an 'instigative setting' (Vandenberghe 2001): nominalizations such as gusher 'a flowing oilwell' and squirter 'a container that ejects liquid in a jet from a narrow opening' relate to what have been termed "Setting-Subject" clauses by Langacker (1991: 346), as in (8) (9)

A supertanker continues to gush oil off the coast of Spain. (CB) Tank is non-functional and does not squirt water. (CB)

It is argued in Vandenberghe (2001) that Setting-Subject constructions like those in (8) and (9) are based on ergative verbs (Davidse 1992b) and allow for an apparent setting to be used in Subject-position. By being construed as Subject, however, the (agent-like) instigative force that characterizes the setting in its function of Emitter of the flow of the substance, is foregrounded. As an Emitter, the setting then "either instigates the motion as part of its functioning, or due to its malfunctioning as a container, for instance a crack or hole ..." (Vandenberghe 2001: 10). In conclusion, agentive lexicalizations have in common with nonagentive -er nominalizations that the relationship between Subject and verb which they imply tends to be a modal one, with the profiled entity or Subject as the locus of potency. Characteristic of agentive -er lexicalizations, then, is that rather than profiling a subject-like entity which lets an Agent carry out the process, they profile an agentive Subject or a subject-like entity which brings about the process itself.

3.1.3. A note on instrumental -er nominalization Interestingly, the differences and correspondences between agentive and non-agentive -er lexicalizations also shed a new light on the unique position of the large group of instrumental -er nominalizations within the system of deverbal -er nominalization. Instrument-denoting nomináis have often been analyzed as inanimate extensions of the human agentive prototype of -er nominalization (e.g., Marchand 1969; Ryder 1991, 1999b). Many of them can indeed be related to clausal agent-like participants, as they are mapped onto the Subject function: (10) a computer computes data a printer prints data a transmitter transmits data a carrier carries people and things

164

A multifunctional approach to -er nominalization

However, I pointed out before that a significant number of instrumental -er nomináis cannot be attributed agent-like properties. Nomináis like knocker and poker, for instance, seem to be most acceptable in an instrumental with meaning (Ithe knocker knocks on the door; Ithe poker pokes the wood in the fire, but: you knock on the door with the knocker, you poke the wood with the poker). In the case of stroller, viewer and walker, an agentive reading is totally excluded: a stroller does not stroll·, a walker does not walk and a viewer does not itself view. Now that the agentive and the non-agentive type of lexicalized -er nominalizations have been described and their respective properties have been accounted for, it becomes clear that instrumental -er nomináis find themselves in a unique position in the system of deverbal -er suffixation: instruments by definition hover between being able to carry out a process themselves (as agent-like participants) and letting others carry it out (as non-agentive entities). Those instrumental -er nomináis that are nonagentive thus resemble middle constructions in that they profile an entity that has properties that let an implied Agent perform a particular action (e.g., stroller, viewer, walker). Agent-like instrumentais, on the other hand, foreground the agent-like ability of the tool which they refer to (as in transmitter, toaster). A large group of instrumental -er nomináis lies in between the agentive and the non-agentive type of -er nomináis: depending on which perspective is chosen, they can be conceived as agentive or as non-agentive. Even Instruments which strongly depend on the intervention of a human Agent in order to fulfill their function can, in certain contexts, be presented as agentlike, witness the following attestations ofpeeler and grater: (11) All Purpose Peeler - Peels fruit and vegetables quickly and safely. Simply the best. (CB) (12) ... a clever stainless steel double grater that grates ingredients finely or coarsely.... (CB) Instrumental -er nominalizations, in short, embody one of the fundamental choices that is offered by -er suffixation, i.e. that between an agentive and a non-agentive, letting profile. The fact that their position in the system of deverbal -er nominalization is significantly elucidated by the relationship of agnation which I propose for non-agentives and middle constructions and by the modal nature of agentive lexicalizations, once again provides support in favour of a middle-related analysis of non-agentives and a Subject generalization for deverbal -er nominalizations in general.

The constructional properties of -er nominalization 165 In the next section, I will show that the analysis of ad hoc nominalizations helps to complete the picture of the Subject-Finite-like relationship which deverbal -er nominalizations encode. While lexicalizations were shown to prefer modalized Subject-Finite relationships, it will be argued that ad hoc nominalizations exploit the other option which the SubjectFinite unit in clauses offers and designate primarily temporally grounded instances of a process type.

3.2. Ad hoc nominalizations I pointed out earlier that the distinction between lexicalized -er nominalizations and ad hoc formations is centred on differences in the degree to which -er nominalizations are 'entrenched': while lexicalizations are sufficiently automatized to be themselves included as units or fixed expressions in the language system, ad hoc formations are derived 'ad hoc', on the basis of a schematic unit or structural pattern (Chapter 3, Section 1.2). Especially interesting for my analysis of deverbal -er suffixation is that lexicalizations and ad hoc formations also differ in their semantic and formal properties. As to the meaning of ad hoc nominalizations, I referred earlier to Levin and Rappaport's (1988) 'event' interpretation, according to which ad hoc nominalizations designate events or processes that have actually occurred, in contrast with the so-called 'non-event' value of lexicalizations. Regarding the formal properties of ad hoc nominalizations, I have in Chapter 5 (Section 2) pointed to the frequent occurrence of adverbial-like adjectives (e.g., He is a hard worker, Mackenzie 1990: 137), the impossibility to change the accent position of the verb (e.g., He is a PHOtographer of snow-scenes, Mackenzie 1990: 137), and the frequent use of postmodifying structures (e.g., the Welsh destroyer of Del Harris's hopes, CB). I will argue that, in spite of their semantic and structural differences, lexicalized and ad hoc -er nominalizations instantiate the same high-level constructional schema of deverbal -er nominalization. Like lexicalized -er nominalizations, ad hoc ones can be shown to establish a relationship which is similar to that which in the clause is realized between the Subject and the Finite. Of the options that are available within the Subject-Finite constellation, however, ad hoc nominalizations tend to choose options which are partly different from those realized in lexicalizations. Ad hoc nominalizations, in other words, realize a low-level schema which is different from that of lexicalizations. More precisely, unlike lexicalizations, ad hoc nominalizations tend to establish a relationship between an entity and a process which is primarily temporal (see also Strang 1968, 1969).

166 A multifunctional approach to -er nominalization To map out the properties of ad hoc nominalizations and establish their precise position within the system of deverbal -er nominalization, it is important to distinguish between two types of ad hoc nominalizations. Ad hoc nominalizations seem to follow either one of two strategies: (13) One guy jumped right into the fight, but his friend immediately vanished. The police came and hauled off the fighter, after which the vanisher promptly reappeared, laughing. (Ryder 1999b: 283) (14) I'm not a frequent resigner. (CB) In (13), the -er nominalizations vanisher and fighter have definite, specific reference and refer back to participants which were introduced in the preceding discourse context. I will call them 'phoric ad hoc nominalizations'. Phorie ad hoc nominalizations differ from ad hoc nominalizations such as resigner in (14), which is indefinite and part of a non-referential nominal: rather than establish the presence of a referent in the universe of the discourse, it ascribes a quality to the entity construed as Subject of the copular clause (Kuno 1970). In what follows, I will consider the properties of both types of ad hoc nominalizations more closely.

3.2.1. Phoric ad hoc nominalizations What characterizes phoric ad hoc nominalizations like the ones in (13) is that they realize specific reference to a referent that has been established in the discourse or in the situational context. The identity of the referent of ad hoc instances can typically be retrieved exophorically (i.e. in the context of situation) or anaphorically (in the preceding text) (Halliday and Hasan 1976). An example of a nominalization which refers to an instance that can only be identified in the context of situation is when someone who is listening to an opera asks Who is the singer? (Jespersen 1914-1929, 6: 231). The nominalization singer then refers to a specific instance in the situational context, i.e. the one who is singing now.54 The nomináis in (15) illustrate anaphoric reference or reference to an entity that was introduced in the preceding discourse (see also 13): (15) One of them was faking. ... Could the faker keep up free association ...? The faker, whichever he was, had practised or had natural talents. (Kastovsky 1986b: 410)

The constructional properties of -er nominalization 167 Phorie ad hoc nominalizations are characterized by the fact that they designate an actual relationship which is situated within the temporal framework of the discourse. Rather than being entrenched common nouns with a conventional, modal meaning, they establish an ad hoc relationship which depends on the context for its interpretation and precise temporal grounding. Unlike non-phoric ad hoc nominalizations (which I will discuss in Section 3.2.2), they realize a relationship which is always temporally grounded. The nominal singer in Who's the singer?, for instance, refers to the one who is singing (now). The nomináis fighter and vanisher in (16) designate the one who was fighting and the one who had vanished respectively. (16) One guy jumped right into the fight, but his friend immediately vanished. The police came and hauled off the fighter, after which the vanisher promptly reappeared, laughing. (Ryder 1999b: 283) It is the temporally grounded status of phoric ad hoc formations like these which lies, I think, behind Levin and Rappaport's (1988) notion of 'event' nomináis. The nominal strategy which phoric ad hoc nominalizations use to establish specific reference is that of definitely grounded common nouns with relative quantification. Phoric ad hoc nominalizations cannot opt for the nominal strategy of indefiniteness. The context always makes clear what the relevant reference mass is. In the case of the nomináis in (15), for instance, the ad hoc formation faker quantifies universally over the reference mass (which consists of only one instance, as indicated by one of them) by profiling a quantity that is identical to it. Other examples which illustrate the definite status of phoric ad hoc nominalizations are: (17) "Not him," says Mrs. Cook. "This one, though. A bit of a poseur, but he moves very nicely." I will not describe the nice mover, to spare his blushes, but let us say that he had dressed for the part. (CB) (18) Recent research has established that merely breathing smoke-laden air - passive or 'second-hand' smoking - puts the breather at risk. (CB) Lexicalized -er nominalizations, in contrast, can be used with the various grounding strategies that are available for ordinary common nouns as well and they are either definite or indefinite. They can be definite and have relative quantification, as illustrated by the second occurrence of teacher in (19):

168 A multifunctional approach to -er nominalization (19) I was a teacher at an approved school and I saw a staff member beat up a boy. I was asked to turn a blind eye but I couldn't and because of that I was shunned by colleagues who expected me to support the teacher. (CB) The nominalization the teacher has definite reference and refers back to the staff member mentioned earlier. The contextually defined reference mass of teacher thus consists of only one instance, over which the teacher quantifies universally. Lexicalizations can also have indefinite reference. As pointed out in Chapter 4, when it is used indefinitely, the nominal type specification does not define a reference mass, but serves a purely "classificational" function (based on Davidse 2000b: 1113). Indefinite lexicalized nomináis may present the designated entity as indefinite, but specific (and therefore paraprasable as 'a certain X', Dirven p.c.), as in: (20) But it was when a teacher realized the lead singer of the group, Freddie Mercury, was a bisexual who had died of AIDS last fall, that the controversy began. (CB) Often, however, no specific instance is referred to, as in (21): (21) At my children's school, they're looking for a teacher. Phorie ad hoc formations, which designate specific, contextually identifiable instances, differ most radically from lexicalized -er nominalizations. The difference between lexicalizations and ad hoc formations that are used non-referentially is more subtle: the non-phoric ad hoc nominalizations which I will discuss in Section 3.2.2 in some respects come closer to lexicalizations than to phoric ad hoc nominalizations. They are not only indefinite, but the type of relationship which they designate is not always temporal and can imply modal grounding as well. Indefinite ad hoc formations thus seem to be non-entrenched derivations which are modelled closely after lexicalizations.

3.2.2. Non-phoric ad hoc nominalizations In this section, I will show that nomináis such as resigner in I'm not a frequent resigner are situated in between phoric ad hoc -er nominalizations and true lexicalizations. Unlike the former, they have indefinite grounding,

The constructional properties of -er nominalization 169 but in contrast with the latter, they tend to make their type specification more informative, by elaborating it with pre- and postmodifiers. As argued in Davidse (2000b, 2001), the type specification in indefinite constructions has a purely classificational function and it "has to enable the hearer to conceptualize instances corresponding to the categorization used by the speaker" (Davidse 2001: 13). For this reason, the speaker will take care to use a type specification which is sufficiently identifiable or clear. In the case of ad hoc formations, I would suggest, this tends to lead to more elaborate type specifications: because these -er formations are not themselves entrenched or lexicalized units, they tend to be turned into more elaborate, higher-order type specifications. The higher-order type specification in ad hoc formations is typically realized by means of explicit markers: as in ordinary nomináis, a higher-order type specification is derived when pre- and/or postmodifying structures are added to the head noun, i.e. the -er derivation as such: (22) ... he could swing the ball and was a precocious taker of wickets. (CB) (23) She was, in that relatively brief time, a girlfriend, a pregnant woman, ... the carrier of an unwanted second child,.... (CB) (24) I was too self-absorbed to be a giver of pleasure. (CB) (25) To get herself back out, she befriends a myriad of oddballs including ... a professional loser of boxing matches. (CB) A more elaborate, higher-order type specification may also be derived contextually, without the inclusion of pre- or postmodifiers. In that case, the borderline between ad hoc formations and lexicalizations is reached. Let's start with the use ofpremodifìers in ad hoc formations. Apart from ordinary epithets, indicating some quality of the type of thing that is being profiled (as in 26 and 27), indefinite ad hoc -er nominalizations frequently take classifying adjectives, which indicate the subclass of the entity being designated (Halliday 1994: 184). The fact that these -er derivations are deverbal noun type specifications may explain the frequent use of what Mackenzie (1990: 137) calls "adverbial adjectives" or classifying adjectives that "do not ascribe a quality ... but rather indicate the manner, frequency or extent of the activity", as in (28), (29) and (30): (26) Minnie is no sentimental praiser of the past. (Jespersen 1914-1929, 6: 229) (27) ... he could swing the ball and was a precocious taker of wickets. (CB)

170 A multifunctional approach to -er nominalization (28) Legal secretary about her boss: I swear, the moment I need to talk to Max, he's suddenly gone. I'm beginning to think he's a professional vanisher. (Ryder 1999b: 282) (29) Ask a frequent flyer like Margie Crase .... (CB) (30) I'm not & frequent resigner. (CB) The use of adjectives such as frequent, regular, professional is noteworthy in that they can be argued to realize lexically what many 'entrenched' -er nominalizations incorporate in the grounding that is implied in them, viz. the fact that a specific process is carried out habitually, regularly or professionally. In ad hoc -er nominalizations, the relationship that is established is temporal, but notions of habituality and professionality can be expressed by adjectives. As noted above, these adjectives serve to derive higher-order type specifications: in the nominal a frequent flyer like Margie Crase, for instance, frequent derives the higher-order type specification frequent flyer and Margie Crase is identified as an instance of the type of people that can be classified as frequent flyers. Likewise, in (28), the heperson is identified as an instance of the class of professional vanishers. Type specifications such as frequent flyer and professional vanisher may not be entrenched or conventional, but semantically they often function in a way which is very similar to lexicalized -er nominalizations. What is special about the postmodifying elements in ad hoc formations is that they tend to be of the restrictive type, i.e. they restrict "the head noun's type specification" (Langacker 1991: 432): they provide information which is essential for the -er derivation to develop into the higherorder or more elaborate type specification which is necessary for it to perform its classifying function, associated often not just with indefinite grounding in the nominal, but also with the clausal function of predicative Attribute as in (31) to (33): (31) You don't have to be a passionate lover of opera to adore Verona's opera season. (CB) (32) Miss Garden was no giver of confidences. (CB) (33) John Kamm is head of Occidental Chemical in Hong Kong and a frequent traveller to China. (CB) Unlike what has been suggested in, for instance, Levin and Rappaport (1988) and Mackenzie (1990), postmodification is also possible in nomináis with lexicalized -er nouns. However, because the head of such nomináis is a lexicalized -er noun, the postmodifier may in that case serve various functions: it may be restrictive and derive a higher-order type specifica-

The constructional properties of -er nominalization

171

tion which is still more elaborate than the type specification of the lexicalization (as in 34). Or it may be non-restrictive, as in the nominalization in (35): (34) ... Allen has told his lover of 12 years that when the custody hearing begins next week there would be nothing of her left standing. (CB) (35) The teacher, of West London, was led away to the cells .... (CB) Notice that some of the more elaborate type specifications, consisting of an ad hoc -er derivation followed by a postmodifier, seem to have become entrenched or 'lexicalized' themselves, and are therefore explicitly included in the dictionary (that it is not necessarily words that can become entrenched or lexicalized was already pointed out in Chapter 3, Section 1.2). The pattern 'be a lover o f , as in As a young man, he was a lover of highspeed sports (CB), is thus included in the Collins Cobuild Dictionary, in addition to the lexicalization lover 'someone with whom one has a sexual relationship'. Another example is reminder 'letter sent to tell you that you have not done something, such as pay a bill' and 'be a reminder of as in The British are about to be given a sharp reminder of what fighting abroad really means (Collins Cobuild Dictionary, 1995). As argued above, the frequent inclusion of postmodifying structures and of premodifiers in ad hoc nominalizations seems to be motivated by the fact that for non-entrenched, ad hoc -er derivations to function as noun type specifications, it is necessary to derive a higher-order, more elaborate type specification first. When ad hoc nominalizations do not take any (explicit) pre- or postmodifiers, they are difficult to distinguish from true lexicalizations and only differ from them in that they are not entrenched. Typically, they exploit the habitual meaning which is implied in some agentive lexicalizations. Consider, for instance, the clause Sinatra told the judge his old friend was 'a giver, not a taker'. The nominalizations giver and taker do not belong to the entrenched units of the language system and, as illustrated in (36) and (37), they typically need elaboration if they are to function as noun type specifications: (36) Miss Garden was no giver of confidences. (CB) (37) ... he could swing the ball and was a precocious taker of wickets. (CB) In his oldfriend was 'a giver, not a taker ', however, they are used as if they are lexicalizations and designate the type specification 'someone who has the habit of giving/taking'.

172

A multifunctional approach to -er nominalization

The fuzziness of the categories which non-phoric ad hoc formations realize suggests that they constitute the transition area between phoric ad hoc formations and entrenched nominalizations: they are modelled after lexicalizations, but they are not sufficiently automatized yet to function as lexicalized units. They thus illustrate that lexicalizations and ad hoc nominalizations "mark the end points of a functional scale and are completely dissociated only in certain extreme cases" (Kastovsky 1986b: 412). Like phoric ad hoc derivations, the relationship which they imply is often temporal in nature, as is shown by Levin and Rappaport's (1988: 1070) examples of lifesaver and saver of lives : Levin and Rappaport in my opinion correctly point out that a life-saver need not have saved any lives yet, while a saver of lives as in He is a saver of many lives implies that he has saved many lives. Some non-phoric ad hoc nominalizations, however, come closer to being lexicalizations in that they establish a relationship which has modal characteristics. Being agentive in profile, ad hoc nominalizations then mostly imply the modality of ability, as illustrated by the nominalizations in (38) and (39): (38) ... he could swing the ball and was a precocious taker of wickets. (CB) (39) I was too self-absorbed to be a giver of pleasure. (CB)

3.2.3. Ad hoc -er nominalization: Conclusion To conclude, ad hoc -er nominalizations represent derivations which are not entrenched, fixed expressions of the language system (and hence not evoked unconsciously, Dirven p.c.), but which are consciously derived ad hoc, to serve a specific discourse need. They turn out to imply temporal grounding and they follow either one of two nominal grounding strategies. Ad hoc nominalizations are either definitely grounded, phoric nomináis and imply a relationship between a particular, context-identifiable entity and a process which receives specific temporal grounding in the discourse context. Or they are used indefinitely (typically non-phorically) and may also imply modal grounding. Indefinite ad hoc nominalizations designate instances of a type which, in order to discharge its 'classifying' function, tends to be elaborated by pre- and/or postmodifiers.

The constructional properties of -er nominalization 173 3.3. The constructional properties of -er nomináis: Summary To summarize, I have successively discussed the lexicogrammatical properties of non-agentive -er nominalizations, agentive lexicalizations, and phoric versus non-phoric ad hoc nominalizations. Two of their properties stand out: first, all of them imply a Subject-Finite relationship, but the exact nature of it differs for each of them; and, secondly, the types of nominal grounding which the nomináis containing the -er derivation receive also add to our understanding of these different types. As for non-agentive -er nominalizations, the various semantic correspondences between the system of middle formation and non-agentive -er nominalization were shown to be based on a fundamental lexicogrammatical equivalence: like middle constructions, non-agentive -er nouns were argued to establish the objective, dynamic modality of letting. The relationship of agnation between middle constructions and non-agentive -er nominalizations thus turned out to elucidate the properties of the latter significantly and it helped to identify the clausal categories which non-agentives imply (see Chapter 3, Section 5.3). Agentive lexicalizations were also found to have objective, dynamic modality, viz. that of ability or volition. They were moreover shown to typically realize the force-dynamic model of causality. Ad hoc nominalizations all seem to imply temporal grounding. If they are used phorically, they are necessarily temporal. When used in a non-phoric way, they were shown to come close to being lexicalizations and, in some cases, imply modality. Lexicalized -er nominalizations can be used with specific, definite reference, as in (40a) and (40b) and with specific (41a) or non-specific (41b) indefinite reference; or they are non-referential, as in (42a) and (42b): (40) a. Mrs. Emerson, the second-grade teacher of Tony .... (CB) b. Various crashing noises erupted from our pew as our new friend searched for his lost gold earring beneath the wooden kneelers. (CB) (41) a. But it was when a teacher realized the lead singer of the group, Freddie Mercury, was a bisexual who had died of AIDS last fall, that the controversy began. (CB) b. In that town, there's a diner where the portions are gigantic. (42) a. I was a teacher at an approved school .... (CB) b. Is it an eater or a cooker and how long does it keep? (CB) Phoric ad hoc nominalizations choose the specific, definite type of the common noun strategy:

174 A multifunctional approach to -er nominalization (43) One of them was faking. ... Could the faker keep up free association ...? The faker, whichever he was, had practised or had natural talents. (Kastovsky 1986b:410) Finally, non-phoric ad hoc nominalizations seem to be associated especially with indefinite, non-referential nomináis: (44) I was too self-absorbed to be a giver of pleasure. (CB) Thus, the nominal strategies of grounding of-er derivations seem to correlate to a certain extent with the types distinguished on the basis of the kind of Subject-Finite relationship which the -er noun implies: in lexicalized -er nouns the clausal layer is modal and the nominal layer is typically indefinite; in phoric ad hoc formations, on the other hand, the clausal layer is temporal and the nominal layer turns out to be more typically definite. In the following and last section of this chapter, I will attempt to put more semantic flesh on the Subject-generalization which I have proposed for deverbal -er nominalizations.

4. Deverbal -er suffixation as Subject-profiling This section is aimed at identifying what all deverbal -er nominalizations have in common. In previous sections, I have foregrounded the various possible Subject-Finite relationships which the noun type specification formed through deverbal -er suffixation can imply and the range of representational roles and processes that can be mapped onto these relationships. Different subtypes of -er nominalization were identified on the basis of correspondences with distinct Subject-Finite relations such as modal and temporal grounding in agnate clausal constructions. Here I want to describe the impact of -er suffixation in general by zooming in on what all -er nominalizations have in common. To come to a schematic characterization of -er nominalization, it is essential that a distinction be made between the various Subject-Finite relations which -er nouns have in their base, on the one hand, and the actual profile of -er formations, on the other: different types of Subject-Finite relationships may form the base of an -er nominalization, but -er suffixation always foregrounds the Subject-entity of these relationships. Unlike clauses, -er nominalizations do not profile the processual relationship between the Subject, Finite and clausal head, but they zoom in on the entity that is involved in it (which itself comes to function as a noun type specification that can be instantiated and grounded in the

Deverbal -er suffixation as Subject-profiling 175 speech event). It is therefore in the Subject profile that the semantic commonality of all -er nominalizations lies. I pointed out in Chapter 5 that the notion of Subject has figured in the analysis of -er nominalization before, but then as a more or less empty notion, the meaning of which is primarily linked to the representational role that is prototypically mapped onto it, viz. that of Agent (e.g., Marchand 1969; Bauer 1983; Keyser and Roeper 1984; Levin and Rappaport 1988; Mackenzie 1990; Rappaport Hovav and Levin 1992). Halliday's multifunctional analysis of the clause, however, makes clear that the functions of Subject and Agent form part of two distinct layers of organization (see Chapter 2, Section 4). The Subject function clusters together with the various options which are available to the language user to ground the proposition and assume a modal position with regard to it. As such, the Subject serves an 'interpersonal' function. The Agent, on the other hand, expresses how the speaker experiences the world which he describes and forms part of the 'representational' metafunction of language. If we are to come to an accurate account of the Subject function, and, related to it, of the profile of -er nominalizations, therefore, the Subject function has to be described in terms of the interpersonal functions which it serves. I argued in Chapter 4 that for the description of the function of Subject, a syntagmatic or configurational analysis is required which considers the different steps in the organization of the clause as a construction related to the speech event, viz. the functions of process type specification, quantification and grounding. I thus identified three distinct functions of the clausal Subject: it was argued to function as instantiator of the clausal process type specification; it was shown to co-operate with the finite element of grounded clauses by establishing objective or explicit person deixis; and, finally, it was identified as the modally responsible element of the clause (Halliday 1994). I suggest that what deverbal -er nominalizations profile is comparable to the function of the Subject in clauses. Firstly, -er nominalizations characterize people and things in terms of a particular process type which they instantiate. -Er suffixation applies to a process type specification and essentially represents a mechanism whereby a process type (given by the base verb) is instantiated by tying it to an entity (marked by -er). As in the clause the instantation is not necessarily causal in nature but may also involve other types of force, such as as that of 'letting'. Secondly, -er nominalizations imply categories of grounding and thus also profile the 'grammatical person' to which the process is tied. More specifically, the person deixis which they establish is that of the 'third person' (Davidse p.c.). Finally, like the clausal Subject, the entity which comes to be profiled

176 A multifunctional approach to -er nominalization through -er suffixation is held 'responsible' for the proposition that is implied in -er nominalization: in it is vested the truth of the proposition. The entity either has a certain potential for carrying out a process or for letting others carry it out (in lexicalizations); or it is an entity that has done/ is doing/ will do something (in phoric ad hoc nominalizations). Lexicalized -er nominalizations in addition typically profile entities in which, as in the Subject of the dynamic modals of ability, volition and letting, a certain potency is located: the profiled entity has the necessary skills/ knowledge/ physical appearance/ design/ inherent properties to carry out or let others carry out a process. In phoric ad hoc nominalizations that establish a relationship of which the precise temporal status has to be retrieved from the context no such potency is implied.

5. Conclusion In conclusion, by analyzing the correspondences between the clausal Subject-Finite unit and -er nominalization more closely, a new perspective on -er nominalization has been offered. While so far, its semantics had been approached as representational semantics only, I have shown that the most fundamental generalization behind -er nominalization is constructional: -er nomináis profile an entity that is related to a process much like a clausal Subject is related to the finite verb. In agentive -er nomináis, the profiled entity is typically also the 'doer' of the process. In non-agentive -er nomináis, by contrast, we can observe the added meaning that is associated with the Subject of the clausal middle construction, viz. that of conduciveness. By positing a relationship of equivalence between non-agentive -er nominalizations and the Subject-Finite unit in middle formation, a new light was shed on many longstanding controversies: non-agentive cases of -er nominalization (patientives, as well as oblique participants) were accounted for and identified as full members of the system of deverbal -er nominalization, rather than as idiosyncratic or peripheral formations. The large group of instrumental -er nomináis was described more accurately as hovering between being able to carry out a process itself (like agentive -er nomináis) and enabling others to carry it out (like non-agentives). And, finally, the analysis of the Subject-Finite complex of middle constructions also revealed the commonalities between agentive and non-agentive lexicalizations, viz. their dynamic modal status and the identification of the Subject as the locus of potency of the modality. The force-dynamic characteristics of the modal relationships which agentive and non-agentive -er nominalizations establish, then again, revealed their basic difference (i.e. letting vs.

Conclusion 177 causing). Finally, the modal value of lexicalized -er nomináis was shown to contrast with the predominantly temporal focus of ad hoc nominalizations. To conclude, a multifunctional approach to deverbal -er nominalization, or an approach which is not restricted to the representational properties of the system, but includes the constructional categories which it realizes and the textual functioning of the nouns that result from it, manages to offer a new perspective on the mechanisms of -er derivation. It unveils the interaction between the nominal and clausal interpersonal categories of type specification, instantiation and grounding and it shows that the high-level semantics of -er nominalization is identifiable only in terms of these categories.

Part III Factîve nominalization

Chapter 8 Factive nominalization: Towards a descriptive position

0. Introduction The gerundive, that- and the fact i/zaZ-constructions that I want to focus on in Chapter 9 are of the following types: (1) To this day he regrets that one of the unanticipated side effects of the Welfare State was the disappearance of granny and grandpa from the family hearth into the old people's homes. (CB) (2) I don't like his having to do so much .... (CB) (3) One characteristic regarding farm holidays is the fact that the different offers are so varied. (CB) Structures like the ones in (1) and (2) have traditionally been described as complement clauses, or as clauses functioning as Subject or Object in another clause (e.g., Kiparsky and Kiparsky 1971; Menzel 1975; Noonan 1985; Wierzbicka 1988).55 All three construction types are typically characterized as being 'factive', a feature generally thought of as the "presupposition by the speaker that the complement of the sentence expresses a true proposition" (Kiparsky and Kiparsky 1971: 345). Strikingly, the system of factive nominalization tends to be primarily associated with ¿Aaí-structures, which are conceived as prototypical instances of factive nominalization. As to the position of gerundive constructions and the fact /Aai-structures, opinions diverge. Lees (1968: 59-60), for instance, includes only ίΑαί-structures in his category of 'factive nomináis': (4)

That he came was obvious.

Yet, in his analysis of 'gerundive nomináis' he acknowledges that some of them are factive in meaning (Lees 1968: 71-72): (5) His eating vegetables is surprising. The Kiparskys (1971), on the other hand, seem to view gerundives (as in 6b) primarily as 'tests' for identifying factive predicates and their that-

182 F active nominalizatioti: Towards a descriptive position complements: "Only factive predicates allow the full range of gerundial constructions ... to stand in place of the that-clause" (Kiparsky and Kiparsky 1971: 346). The fact ^/-constructions (as in 6c) are likewise primarily viewed as periphrastic structures testing for the factivity of /Aa/-structures (Lees 1968: 65; Kiparsky and Kiparsky 1971: 346-347): (6)

a. that the dog barked during the night b. the dog's barking during the night c. the fact that the dog barked during the night

In the transformational literature on factives, it has even been suggested that the head noun fact is present in the so-called 'deep structure' of all factive nominalizations, also in those that lack a factive head noun in their 'surface structure' (Kiparsky and Kiparsky 1971: 356; Stockwell et al. 1973: 584; Menzel 1975: 89). Halliday (1994: 266, 269) characterizes the fact /^/-constructions straightforwardly as constituting a subtype of factive nominalization. Gerundives he analyzes as borderline cases between factive and non-factive complementation: he says that there exists only a minimal distinction and perhaps even blending between structures like (7) (8)

She liked the snow falling, [factive] She watched the snow falling, [non-factive]

Halliday consequently focuses exclusively on the fact that- and thatstructures in his discussion of factive constructions. Davidse (1991, 1994), finally, takes Halliday's analysis one step further and lists a number of arguments in favour of excluding gerundives from the category of factive complementation all together, basically arguing that factive nominalizations are necessarily derived from grounded clauses (Davidse 1991: 355). Why is it that /Aa/-nominalizations are generally regarded as factive prototypes, while the linguistic status of other nominalized constructions with a factive meaning is contentious? The existing analyses of factive nominalization have all tended to concentrate on one construction type. I will consider factive gerundives, /¿art-structures and the fact /Aa/-constructions together and show that it is only by taking into account the structural properties of each of these agnate constructions that one can chart the options that exist within the system of factive nominalization and gain more insight into the 'factive presupposition'. In this chapter, I will sketch the most important analyses that have thus far been given of both the external and the internal functioning of the various types of factive nominalizations. In

Factivity as truth presupposition

183

Chapter 9 I will then propose a new, radically functional description of the internal properties of factive nominalizations.

1. Factivity as truth presupposition Without doubt the best-known attempt to describe factivity is that by the Kiparskys (1971). They argue that factivity arises when the "speaker presupposes that the embedded clause expresses a true proposition, and makes some assertion about that proposition" (Kiparsky and Kiparsky 1971: 348). They show that the notion of factivity distinguishes verbs such as regret, resent, and ignore, which are factive and combine with factive complements, from non-factive predicates such as claim, assert, suppose (Kiparsky and Kiparsky 1971: 345-347). They observe a similar distinction among copular clauses: while clauses with be significant, odd, exciting are factive in meaning, adjectives such as likely, possible, true and false give rise to a non-factive reading: (9)

a. I regret that it is raining, [factive] b. I suppose that it is raining, [non-factive] (10) a. It is significant that he has been found guilty, [factive] b. It is likely that he has been found guilty, [non-factive] Two aspects of the analysis of factivity presented in Kiparsky and Kiparsky (1971) are especially noteworthy: first, the Kiparskys' analysis reveals a correlation between the factive category and specific formal features and it points to a number of formal differences between factive and non-factive complementation. Secondly, it suggests that factivity results from a combination of features which relate to the factive nominalization itself, the predicate of the clause in which it is embedded and the role of the speaker in the clause. First, Kiparsky and Kiparsky list a number of agnates of factive complements which help to distinguish factive complements from non-factive ones. As I pointed out earlier, //¿«/-constructions are claimed to be systematically replaceable by the fact íAírt-structures and gerundive nominalizations (see 6) (Kiparsky and Kiparsky 1971: 346). The Kiparskys also argue that, unlike factive nominalizations, most non-factive nominalizations can have their Subject turned into the Subject of the main clause, as illustrated in (1 lb) (Kiparsky and Kiparsky 1971: 346): (11) a. It is likely that he will accomplish even more, [non-factive]

184

Factive nominalization: Towards a descriptive position

b. He is likely to accomplish even more. But c. It is relevant that he has accomplished even more, [factive] d. *He is relevant to have accomplished even more. Factive nominalizations cannot be turned into an 'accusative and infinitive construction' either, as is made clear in (12d) (Kiparsky and Kiparsky 1971: 358): (12) a. I understand that Bacon is the real author, [non-factive] b. I understand Bacon to be the real author. But c. I regret that Bacon is the real author, [factive] d. *I regret Bacon to be the real author. Moreover, extraposition is shown to be optional in clauses with factive complements, but obligatory in those with non-factive complements (Kiparsky and Kiparsky 1971: 346): (13) a. That there are porcupines in our basement makes sense to me. [factive] b. It makes sense to me that there are porcupines in our basement. But c. *That there are porcupines in our basement seems to me. [nonfactive] d. It seems to me that there are porcupines in our basement. Finally, Kiparsky and Kiparsky (1971: 362) show that both factive and non-factive clauses can be replaced by it, which is the "pro-form of noun phrases" (see 14a and 14b), while only non-factive clauses can take so, or the "pro-form of sentences", as illustrated in (15): (14) a. John supposed that Bill had done it, and Mary supposed it, [non-factive] b. John regretted that Bill had done it, and Mary regretted it, [factive] (15) a. John supposed that Bill had done it, and Mary supposed so, [non-factive] b. *John regretted that Bill had done it, and Mary regretted so, [factive]

too. too. too. too.

What is interesting about the Kiparskys' approach is not only that they situate factivity within a specific formal paradigm. Important is also that they do not restrict the notion of factivity to one structural element only: in

Factivity as embedded projection

185

their discussion of the factive constructions in (16) and (17), for instance, the Kiparskys first state that the "factive sentence ... carries with it the presupposition 'it is raining'" (Kiparsky and Kiparsky 1971: 348). Then they continue by saying that the "speaker presupposes that the embedded clause expresses a true proposition" (Kiparsky and Kiparsky 1971: 348). Their conclusion, finally, is that "predicates which behave syntactically as factives have this semantic property" (Kiparsky and Kiparsky 1971: 348). (16) It is odd that it is raining. (17) I regret that it is raining. As Delacruz (1976) points out, the Kiparskys thus use the notion of the factive presupposition in three different senses: to refer to the sentence as a whole, to the main verb that functions in it, and, thirdly, to the speaker uttering the clause (Delacruz 1976: 180). And, of course, the complement or nominalization itself is also called 'factive' (Kiparsky and Kiparsky 1971: 356). There is no reason why the four interpretations of factivity should be incompatible.56 In fact, they illustrate that, rather than being traceable to one structural unit only, factivity results from afusion of several factors. The Kiparskys may point to a number of interesting lexicogrammatical differences between clauses which are factive in meaning and clauses which are not, but the precise relationship between the formal characteristics of the factive category and the notion of factivity is never made clear. As Davidse (1994: 260) puts it, "tracing systematically different paradigmatic affiliations to establish distinct constructions is α heuristic technique, not α grammatical explanation". The notion of factivity should be linked up with the category's grammatical behaviour, and because they refrain from doing so, the Kiparskys ultimately fail to shed light on it. The Kiparskys' characterization of facts as propositions that are 'presupposed true' thus remains vague and requires further elaboration.

2. Factivity as embedded projection The analysis of factive constructions given in Halliday (1968, 1994) and Halliday and Hasan (1976) is centred on two basic constructs, viz. that of embedding (vs. hypotaxis), on the one hand, and that of projection (vs. expansion), on the other. The Kiparskys' factive paradigm is in Halliday's approach linked to the system of embedded projection. What do the notions of embedding and hypotaxis, projection and expansion stand for?

186 F active nominalization: Towards a descriptive position 2.1. Embedding vs. taxis First, the term 'embedding' is not new: it is also found in mainstream linguistics, where it is used in a meaning which is closely associated with that of 'complementation': embedded clauses are defined as complements or subordinated clauses that function as arguments of another clause (Noonan 1985: 42). Traditionally, the /^-clauses in (18) and (19), for instance, are both analyzed as embedded clauses or complements (examples taken from Noonan 1985: 90): ( 18) I think that Floyd didn 't hit Roscoe. (19) I regret that Floyd didn't hit Roscoe. Halliday, however, posits a fundamental difference between clauses such as those in (18) and (19) and he uses the notion of embedding for the thatclause in (19) only. He claims that the undifferentiated concept of 'subordination' which is used in mainstream linguistics actually covers two distinct types of clausal relationships, viz. embedding and hypotaxis. In Halliday's view, embedded clauses are clauses that have been rankshifted in that they are made to function at a rank which is lower than that of the clause (see Chapter 3, Section 3.2.1). Halliday argues that it is only embedded or rankshifted clauses which can function as constituents within another clause and serve a function in it: the clause [that Floyd didn't hit Roscoe] in (19) thus forms an embedded constituent within the clause I regret that Floyd didn't hit Roscoe and serves the function of Object in it. Embedded or rankshifted clauses should be distinguished from hypotactic clauses (as in 18). Instead of being embedded or structurally integrated as a constituent in another clause, hypotactic clauses depend on the clause with which they are used (Halliday and Hasan 1976: 136, 221). While a sentence such as I regret that Floyd didn't hit Roscoe thus consists of only one clause which has another one embedded in it, a clause such as I think/said that Floyd didn't hit Roscoe consists of two clauses, of which the second is dependent on or hypotactically related to the first. Hypotaxis is, in other words, a relation between clauses that form a 'clause complex', whereas embedding involves a simple clause in which another clause is downranked. Notice that, in addition to being hypotactic, the relation between clauses in a clause complex can also be paratactic (traditionally referred to as 'coordination'). Unlike hypotactic clauses, paratactic clauses are equal in status: there is no dependent and dominant clause, but only an 'initiating' and a 'continuing' one (Halliday 1994: 218). An example of

Factivity as embedded projection

187

paratactic clause combining, Halliday (1994: 250) argues, is formed by quotes: (20) Brutus said: 'Caesar was ambitious'. The claim that the traditional notion of subordination actually covers two distinct types of clausal relations has been made by Halliday as early as 1968. Various other studies have since suggested that a distinction should indeed be made between, on the one hand, part-whole relationships or embedding, and, on the other hand, part-part or tactic relationships (i.e. parataxis and hypotaxis) (McGregor 1997). With regard to quotes, for instance, it has been argued that "the quoted sentence is not syntactically or semantically a part of the sentence that contains it" (Partee 1973: 418; see also Munro 1982). As to the construct of subordination, Haiman and Thompson (1984: 510) argue that it is "at best a negative term which lumps together all deviations from some 'main clause' norm, and thus treats as unified a set of facts which we think is not a single phenomenon". They point out that the systemic-functional framework is 'exceptional' in that it has been careful to distinguish between hypotaxis and embedding (Haiman and Thompson 1984: 521; see also Matthiessen and Thompson 1988; Hopper and Traugott 1993). Before I consider the arguments that can be given in favour of Halliday's distinction between embedding and taxis and before I turn to the analysis of factive nominalizations as embedded constituents, another distinction has to be introduced, viz. that between projection and expansion.

2.2. Projection vs. expansion Cross-cutting the categories of embedding and taxis are those of projection and expansion. Halliday (1994: 219) argues that the relationship between clauses within a clause complex or, in the case of embedding, within a simple clause cannot only be described in terms of dependency and embedding, but also in terms of more general semantic concepts. He posits the existence of two fundamental types of semantic relationships between clauses, viz. those of expansion and projection. When clauses are combined and each clause represents "the phenomena of experience themselves" (Halliday and Hasan 1976: 131), they are said to be related through expansion. Expansion thus implies that each of the clauses involved in the relationship forms a "direct representation of (nonlinguistic) experience" (Halliday 1994: 250). In the clause complex When-

188 F active nominalization: Towards a descriptive position ever the horse stopped, he fell o f f , for instance, both clauses are direct representations of experience. When, on the other hand, one of the clauses in the relationship represents experience not directly but instead designates "linguistically processed phenomena" (Halliday 1968: 195), Halliday prefers to speak of projection. Projections are metaphenomena (Halliday 1994: 249): they are representations "of a (linguistic) representation" (Halliday 1994: 250). They represent "phenomena already encoded in language" which then participate in other linguistic structures (Halliday and Hasan 1976: 131). Among the most important subtypes of projection which Halliday distinguishes are the following (Halliday 1994): (21) 'Was Caesar ambitious?' asked Mark Antony, [quote] (22) Mary said/thought that she wanted to go home, [report] These clause complexes differ in terms of the process type that functions in their main clause and in terms of the relationship that exists between the two clauses which they consist of: quotes are paratactically related to clauses which typically contain a verbal process (i.e. asked in 21). Reports are hypotactically related to the 'dominant' clause: in (22), that she wanted to go home thus depends on the main clause Mary said/thought. Reports can follow mental processes (e.g., thought), as well as verbal processes (e.g., said). When they are construed with a mental process, they are traditionally referred to as 'indirect speech' representations.

2.3. Facts as embedded projections Halliday (1968, 1994) analyzes facts as projections which are not tactically related to the clause with which they are used, but are embedded in it as a constituent. Facts are said to typically combine with clauses that contain mental and relational processes, and, less frequently, verbal processes. An example of a factive nominalization that functions in a relational clause or a clause that sets up a relation between two entities (Halliday 1994: 119) is given in (23): (23) Mr. Archer said it was significant that Queensland, a growing sector in the Australian property industry, had been granted this important recognition. (CB) Mental processes taking factives typically express a feeling of affection (Halliday 1994: 118):

Factivity as embedded projection 189 (24) I regret the role of mother isn't generally more highly regarded by everyone. (CB) Finally, in some cases, also verbal processes can be followed by a fact: (25) Its owners cheerfully admitted that the bird was mad, and seemed genuinely offended when I suggested it needed to be put down. (CB) Characteristic of factive nominalizations is that they are not projected by the process of the clause in which they function, but come "as it were ready packaged in projected form" (Halliday 1994: 264). A clause with a factive nominalization in it thus represents some form of interaction with a prepackaged proposition: a fact "is an entity interacted with, or mentally manipulated", it is "the source of a specific affect ... or something which is cognitively or linguistically reacted to", as in Don't forget [that you are a Christian] and He didn 't want to admit [that he was a Christian] (Davidse 1994: 276). Non-factive projections, in contrast, are projected by the process of the clause which they combine with: rather than being factive, the projected clause gives the 'content' of the locution or idea that is implied in it. In sentences such as those in (26) and (27), the ¿Ααί-structure is thus hypotactically projected by the verbs think and point out and it specifies the thoughts and locutions of the main process's Subject; in the factive constructions in (28) and (29), on the other hand, the ¿Ααί-constructions are not projected by but responded to or interacted with by the Subject of the main verb: (26) David thinks that all championships would be better run on Grand Prix lines. (CB) (27) King pointed out that Meredith's admission was merely a crumb thrown black Americans in lieu of real equality for all. (CB) (28) Probably you resent that peace demonstrations include kids who can wait out the war in college. (CB) (29) I regret that it all happened. (CB) Where do embedded projections come from, then? Whose ideas do they represent? In Kiparsky and Kiparsky (1971) it was argued that facts are presupposed true by the speaker, i.e. that it is always the speaker who is committed to the truth of the factive proposition. As pointed out by Davidse (2003), however, 'speaker factivity' is only one possible interpretation of factivity and two other types of factivity need to be recognized, i.e. 'processer factivity' and 'speaker-processer factivity'. Where do these notions

190

F active nominalization: Towards a descriptive

position

come from and what do they stand for? In his (1976) article, Delacruz observes that the sentence Bill regrets that John resigned can receive a reading in which both the speaker and Bill regard the proposition John resigned as a fact, as well as a reading in which John resigned is a fact to Bill (i.e. the processer of the mental process regret) only. Building on Delacruz's observation and on the basis of a corpus study of factive constructions, Davidse (2003) suggests that there are three basic interpretations of the notion of factivity and that these interpretations can be linked to a number of general semantic schemata. Firstly, facts may be facts for the speaker only when they function in discourse contexts that imply no other consciousness than that of the speaker (e.g., in relations of 'addition' or 'causality', Davidse 2003; e.g., Add to that the fact that over half the menus the NRA analyzed include sandwiches, and you 've got a trend worth noting, CB). In environments which either represent or imply a 'second consciousness' distinct from the speaker's, by contrast, facts may be facts either to both the speaker and the processer that is interacting mentally with the proposition or to the processer only. In The council have woke [sic] up to the fact that Glasgow and a lot of areas are slums (CB), for instance, the use of the verb wake up conveys that the speaker was subscribing to the factivity of the proposition that Glasgow and a lot of areas are slums before the council came to accept it (Davidse 2003). A pure 'processer fact', on the other hand, can be found in What the 37 year-old finds most soul destroying about Haiti is the fact that things have not changed since his childhood (CB): the processer the 37year-old is depicted here in his or her emotional reaction to a proposition which s/he perceives as a fact but which is not necessarily a fact to the speaker (Davidse 2003). Note that even facts that are facts to the processer only represent propositions which are not projected by the clause with which they combine - as are reports and quotes but which are 'pre-existent to' (Davidse 2003) and 'manipulated by' the relation in which they participate. By characterizing facts as projections, Halliday manages to elucidate their commonality with clause complexes that have typically figured in the discussion of factives, i.e. those involving direct and indirect speech and thought. It is the status of factive nominalizations as embedded constituents which turns out to be crucial to distinguish them from the other types of projection. In what follows, I will go more deeply into the arguments that have been given in the literature to support the analysis of factives as embedded in rather than hypotactically related to the clause with which they combine. The discussion is based on Halliday (1968, 1994), Halliday and Hasan (1976) and Davidse (1991, 1994).

F activity as embedded projection

191

Halliday's claim that factive clauses are embedded entails that they are viewed as downranked constituents which serve a function in another clause. Which arguments can be given in favour of the downranked status of factive structures? Evidence that ί/ζαί-factives fill a nominal slot in the clause comes from the observation that, unlike their hypotactic counterparts in (31), the nominalized ίΑαί-constructions in (30) can be coded as Subject of the passive (as in 30a), they can be given thematic prominence by being fronted (e.g., 30b), and they can be extraposed as Objects, while the Object slot is filled by it (see 30c): (30) a. That they had to wait in line was resented by people. b. That they had to wait in line people resented. c. Most people resented it that they had to wait in line. (31) a. *That they had to wait in line was insisted by Drew. b. *That they had to wait in line Drew insisted. c. *Drew insisted it that they had to wait in line. Factive that-structures also seem to be semantically entitized or reified: as Davidse (1994: 271) points out, "there is a strong general tendency for facts to be replaceable by nominal groups": while in (32) and (33), for instance, the /Aai-factives can be said to systematically relate to ordinary nominal groups, the hypotactic ί/ζαί-structures in (34) and (35) do not allow for such an alternation (examples based on Davidse 1994: 271): (32) a. Most people resented that they had to wait in line. b. Most people resented the long waiting period. (33) a. Tom regrets that they 're getting divorced. b. Tom regrets their divorce. (34) a. Drew insisted that they had to wait in line. b. *Drew insisted the long waiting period. (35) a. Sue thinks that they 're getting divorced. b. *Sue thinks their divorce. The conceptual reification of the clause in ίΑαί-factives is, moreover, accompanied by internal structural behaviour which is different from that of ordinary or non-rankshifted clauses. First, downranked ίΑαί-structures are limited in terms of the thematic freedom which they allow their constituent parts. Unlike hypotactically projected that-clauses (as in 36), for instance, that-factives (as in 37) cannot have their main verb thematized (Hooper 1975: 99):

192 Factive nominalization: Towards a descriptive position (36) a. Drew insists that wait they shall/will. b. Sue thinks that divorce they will. (37) a. *People resent that wait they will. b. *Tom regrets that divorce they will. Another structural characteristic which distinguishes embedded from hypotactic /^/-structures is that the former readily accommodate non-sequent tenses or deictically absolute tenses relating directly to the speaker-now rather than being coded relative to the projecting clause (Kiparsky and Kiparsky 1971: 359). An example is Ijust remembered you are a socialist, in which the factive /Aa/-clause is not grounded relative to the main process remembered, but has direct deictic ties to the speaker-now. It should be observed that the fact that //zai-factives can be fitted in with the nominal paradigm is not to say that they do not also differ from nonnominalized or ordinary nomináis. 77aa/-factives are nominalized to a large degree, but being embedded clauses, they will always be non-prototypical nomináis. Unlike with ordinary nomináis, for instance, the use of factive iAa/-structures as Subject-nominal in interrogatives is restricted (e.g., Kuno 1973: 363; Noonan 1985: 82; Huddleston 1988: 145fn; Davidse 1994: 270): (38) ? Is that she did it now known? (Davidse 1994: 270) (39) ? Is that Floyd left town significant? (Noonan 1985: 82) As far as extraposition is concerned, by contrast, that-factives are more flexible than ordinary nomináis in that they can be extraposed: (40) It's significant that Floyd is drunk. (Noonan 1985: 83) (41) It makes sense to me that there are porcupines in our basement. (Kiparsky and Kiparsky 1971: 346) Moreover, as is pointed out by Davidse (1994: 270), a factive /Aa/-structure like that in (42) does not allow for the passive: (42) That you work so hard pleases us. *We are pleased by that you work so hard. In spite of the fact that they do not take over all of the functions which an ordinary nominal can fulfill in the clause, however, factive /Λα/-structures clearly function as nomináis and their embedded or nominal constituent status shows up in many of their lexicogrammatical properties.

F activity as embedded projection

193

Because it fails to distinguish embedding from hypotaxis, some of the early literature on factive complementation is riddled with examples of what are actually non-factive clauses or clauses that are hypotactically related to the main clause. Examples of hypotactic structures that have thus wrongly been situated in the factive paradigm are, for instance, the following constructions cited in Lees (1968: 59-60): (43) I know that he came. (44) I told her that he saw us. It also becomes clear now that the 'tests' which the Kiparskys use to distinguish factive from non-factive clauses primarily pick up on the distinction between hypotaxis (non-factive) and embedding (factive). That factive clauses cannot be substituted by so, for instance, has to do with their embedded, nominalized status: as shown by Halliday and Hasan (1976: 131-134), substitution can only be used with the same function and class as the wording which it substitutes for. In the case of hypotactic clauses, so is possible because the clause which it replaces serves the function of clause. By contrast, embedded or rankshifted clauses have lost their clausal status and function first and foremost as nomináis in another clause (Davidse 1994: 274). Reference, in contrast, does not depend on whether the reference item has the same function or class as the clause which it refers to (Halliday and Hasan 1976; Davidse 1994). It is therefore possible to have substitution in (45), but not in (46). In (46) only reference is acceptable: (45) Do we really have to wait in line? It seems so. Drew says so. (46) I regret that you had to wait in line. I regret it/*so too. Further, it is the nominalized status of factive clauses that makes fronting possible, either as Subject of the passive, or as a marked Theme of the clause (Davidse 1991: 337). Hypotactic structures do not normally have the same flexibility, because they do not represent constituents of the clause, but depend on it (see 48): (47) That he had been wrong was admitted by him. (48) IThat he had been wrong was said by him.57 Finally, the fact that factive nominalizations cannot have their Subject turned into the Subject of the main clause (as shown in 49a) or figure in 'accusative and infinitive constructions' (as illustrated in 49b) is a direct result of their embedded status as well (Davidse 1991: 354). Because fac-

194

F active nominalization: Towards a descriptive position

tive clauses are rankshifted into nominal slots as constituents of the clause, they cannot move beyond the boundaries of the nominal element of structure. (49) a. *He is relevant to have accomplished even more, [factive] b. *I regret them to have accepted, [factive]

2.4. Conclusion Halliday's interpretation of facts as embedded projections is innovative and elucidates the relationship between the meaning of factivity and the lexicogrammatical properties of factive constructions. By identifying facts as projections, Halliday manages to shed light on the resemblances which factive constructions show with clause complexes in which speech and though act verbs figure (e.g., factive He resents that she did it vs. non-factive He says/thinks that she did it). Halliday's approach thus situates fact clauses within their natural habitat, i.e. that of metaphenomena, ideas or linguistically processed phenomena, as opposed to direct representations of the way in which the language user experiences reality. Importantly, by breaking up the undifferentiated mainstream concept of 'subordination' into embedding and hypotaxis, Halliday also succeeds in singling out the main difference between facts and projecting clause complexes with reports: evidence is provided in favour of the embedded, constituent status of facts and the hypotactic, dependent status of reports. Factive fAai-clauses turn out to be downranked nominalized constituents which display much of the flexibility of ordinary nomináis with regard to assuming Subject or Object/Complement position in the clause and they are to a certain degree also semantically entitized. Rather than being 'presupposed true by the speaker', it was argued, facts are propositions which are 'pre-existent' to the relation in which they participate (Davidse 2003). Their being pre-existent rather than projected by the main clause is reflected in their embedded status. The concept of embedded projection thus significantly elaborates and refines the Kiparskys' somewhat vague characterization of facts as being 'propositions that are presupposed true by the speaker'. However, Halliday's account of factive nominalization also leaves many questions unanswered. I see two main lacunae, both of which are basically concerned with the delineation of the category of factive nominalizations: first, the distinction between hypotaxis and embedding is not always clearcut and it is sometimes hard to draw the line between truly factive (embedded) projections and non-factive (hypotactic) projections. Secondly, Halli-

Delineation of the fact category 195 day focuses exclusively on finite factives (i.e. on ίΑαί-constructions and the fact /Aa/-nominals) and leaves the category of gerundive nominalizations undiscussed. Yet, the analysis of gerundive nominalizations is highly pertinent to the description of factivity and the borderline between factive gerundives or gerundives which are embedded projections, on the one hand, and non-factive gerundives, on the other, is at times hard to draw. In what follows, I will briefly elaborate on some of the remaining problems with respect to the delineation of the category of factive nominalization defined as embedded projection.

3. Delineation of the fact category The delineation of the category of factive nominalization is problematic in several respects. First, there exist constructions which combine properties of both the factive and the non-factive category of predicates. Constructions belonging to this type have been called 'semi-factives' (e.g., Karttunen 1971; Hooper 1975). There also exist a large number of ambiguous predicates or predicates that can be interpreted either factively or non-factively. Finally, the analysis of gerundive nominalizations poses problems of delineation as well.

3.1. Ambiguity and vagueness Kiparsky and Kiparsky (1971: 360-361) point out that there exist numerous predicates that are ambiguous and have both a factive and a non-factive use. A sentence like that in (50a), for instance, is said to have a factive and a non-factive reading, evidence for which is provided by the minimal pairs in (50b) and (50c) (Kiparsky and Kiparsky 1971: 360): (50) a. They reported that the enemy had suffered a decisive defeat. b. They reported the enemy to have suffered a decisive defeat, [nonfactive] c. They reported the enemy's having suffered a decisive defeat, [factive] The sentence in (50c) implies "that the report was true in the speaker's opinion" (Kiparsky and Kiparsky 1971: 360). It is factive and consists of a systematic agnate of the factive type of ¿/¿«/-structure, viz. a gerundive nominalization. The to-infinitive in (50b), on the other hand, "leaves open

196 F active nominalization: Towards a descriptive position the possibility that the report was false" (Kiparsky and Kiparsky 1971: 360). As I argued before, the accusative and infinitive construction is only possible with non-factive clause complexes, because it requires a direct relationship between two clauses. The Kiparskys' claim that the projected clause is not presupposed true and may be 'false' picks up on the hypotactic nature of the clause complex: the clause the enemy to have suffered a decisive defeat is hypotactically projected by the process of the main clause and represents the projection of the main clause Subject, rather than that of the speaker. The hypotactic clause in (50b) thus implies that the fact that the enemy had suffered a decisive defeat is only what they reported; it does not necessarily coincide with the speaker's opinion. An example of a verb which can be used in different verb senses and which, depending on the meaning in which it is used, is either factive or non-factive is explain (Kiparsky and Kiparsky 1971: 361): (51) I explained Adam's refusing to come to the phone, [factive] (52) I explained that he was watching his favorite TV show, [non-factive] In (51), explain is used in the meaning of 'give reasons for', while in (52) it means 'say that S to explain X' (Kiparsky and Kiparsky 1971: 361). In Halliday's model, (51) is considered factive because the structure Adam's refusing to come to the phone functions as embedded projection: rather than being projected or directly asserted by the clause I explained, the I-person expresses some assertion about the (pre-existent) proposition. The thatclause in (52), on the other hand, is a non-factive, hypotactic projection of the main predicate: it is directly projected by the Subject of the main process and represents the locution that is implied in explain. The Kiparskys rightly observe that the list of verbs with both a factive and a non-factive reading is quite extensive: many constructions are ambiguous, even those that contain what seem to be prototypically factive verbs like regret·, the sentences in (53) and (54), for instance, differ in meaning from the one in (55): regret is factive in (55) and expresses a kind of affect, with the that-clause an embedded projection. In (53) and (54), on the other hand, the ίΑαί-clause is hypotactically projected by regret, which is used in the meaning of 'regret to say'. The /Aa/-clauses in it consequently realize a locution rather than a fact. (53) Irma regrets that she cannot answer your letters personally. (CB) (54) The Times also regrets that enclosures accompanying letters cannot be returned. (CB)

Delineation of the fact category 197 (55) He trained during the Gulf War and even regrets that he was never called up to take part in any action. (CB) Apart from ambiguous predicates, there exist also predicates which are characterized by vagueness and which have features in common with both the factive and the non-factive category. Among the processes that have in the literature been argued to belong to this type are verbs describing "processes of knowing or coming to know" (Hooper 1975: 117), such as know, discover, find out, learn, and predicates such as deny, doubt, it is possible, it is true, it is false (see, among others, Kiparsky and Kiparsky 1971; Karttunen 1971; Hooper 1975; Wierzbicka 1988). (56) Anyone who has seen Internet knows that it uses amazing technology. (CB) (57) ... she also found out that her father has been in touch with Derek all this time. (CB) (58) ... he discovered that Sharon's parents, both known drug takers, had been repeatedly attacking their daughter. (CB) (59) The agent of a sacked Coronation Street star denies that she is on the verge of a nervous breakdown. (CB) (60) It is true that in the Seventies there was no shortage of 'stupid gymnasties', as the Bhagwan once called the act of lovemaking. (CB) (61) ...it is possible that this year we may suffer again. (CB) It has repeatedly been pointed out in the literature that predicates like the ones in (56) to (61) are located somewhere in between 'true' factives and non-factives. The process know has thus been described as "semantically factive", but "syntactically non-factive" (Kiparsky and Kiparsky 1971: 348).58 It is, for instance, impossible to say *I know the fact that John is there or *I know John's being here, which seems to suggest that know is non-factive. Also, know takes the accusative and infinitive construction, which is normally only possible with hypotactic or non-factive structures (e.g., I know him to be there, based on Kiparsky and Kiparsky 1971: 348). Similar observations have been made with regard to it is possible and it is true·, like factive predicates, they cannot have their Subject raised to the Subject-position of the main clause (e.g., It is possible that he will accomplish even more cannot be turned into *He is possible to have accomplished even more). Like factives, moreover, they may or may not take extraposition (e.g., It is possible that we may suffer again or That we may suffer again is possible). Unlike factives, however, clauses construed with predicates such as it is possible and it is true do not alternate with the fact that-

198

F active nominalization: Towards a descriptive position

constructions and gerundives (e.g., *The fact that we may suffer again is possible/true·, *John's being ill is possible/true) (Kiparsky and Kiparsky 1971). For one thing, structures like these show that the distinction between hypotactic projection and embedded projection is not always clear-cut and that analyses of particular configurations are likely to confront us with the existence of categories that lie in between hypotactic and embedded projection. Constructions like the ones in (56) to (61) thus reveal the need for further clarification of the way in which factive /^-structures are related to the clause with which they combine and ultimately, their semantic and structural properties will have to be integrated in a description of factive nominalization that is aimed at being exhaustive. My own analysis of factive nominalization, however, will focus on the internal functional properties of the factive category, rather than on the delineation of the external contexts in which it occurs.

3.2. Gerundive nominalizations: Acts or facts? The problems related to the delineation of the factive category within the group of gerundive nominalizations were already pointed out by Lees (1960) in his groundbreaking study on nominalization in English. In his analysis of what he calls 'gerundive nomináis', Lees (1968: 58) shows that a distinction should be made between the following two subtypes: (62) a. Eating vegetables is healthful. b. * Eating vegetables was a great surprise. (63) a. His having eaten vegetables was a great surprise, b. *His having eaten vegetables is healthful. The nominalization in (62) is said to illustrate a type of gerundive nominal that is 'action-referrent'. It is said to be characterized by the fact that it "cannot have any expressed subject, nor any auxiliary" (Lees 1968: 72). The structure his having eaten vegetables in (63), on the other hand, is classified with the 'fact-referrent' type of gerundive nominalizations. It is said to occur "with or without the genitive subject ... or the auxiliaries" (Lees 1968: 72). Lees repeatedly stresses that the contrast between both types of gerundive nomináis is "as yet imperfectly understood" (Lees 1968: 58) and that "little beyond these bare facts is as yet known about this further distinction among these nomináis" (Lees 1968: 72).

Delineation of the fact category

199

Lees's description of gerundive nomináis makes clear that problems of delineation can be expected in the case of nomináis that lack an explicitly realized Subject: such nomináis can be either action-referrent or factive. When subjectless gerundive nomináis also lack auxiliaries, it becomes even more difficult to identify their precise import, as is shown in the examples in (64) to (67): (64) I think my only regret, as I say, was not being able to train for the job I wanted to do. (CS) (65) Threatening people will get you nowhere. (Halliday 1994: 248) (66) Tearing up my new paper dolls was mean. (Thompson 1973: 378) (67) Gathering quietly around the campfire each evening was nice. (Thompson 1973: 377) The situation becomes even more complicated when it turns out that, unlike what Lees claims, the action type of gerundive nomináis (i.e. the type of gerundive nominal that is not factive but 'action-referrent') can in some cases take a possessive or periphrastic Subject, as in: (68) Techniques for doing this include the caregiver's exposing the infant to various contingent and discrepant experiences. (CB) (69) My husband speaks very well, but his job involves my answering the phone on his behalf quite a bit of the time. (CB) Lees's analysis of gerundive nominalizations is interesting in many respects. First, it reveals that within the category of gerundive nominalizations, a further distinction can be made between different subtypes, and that the identification of these subtypes is problematic. The existence of an action-referrent and a factive type of gerundive nominalization and the problems that are associated with their delineation show that gerundive nominalizations are highly interesting to the analysis of factive nominalization and explore the borderline between factivity and non-factivity. Secondly, Lees's criteria for distinguishing between factive and nonfactive gerundives are intriguing but clearly require further elaboration. Importantly, the notion of 'embedded projection' as defined by Halliday (1994) does not suffice to further clarify the distinction either. Especially the notion of 'projection' turns out to be insufficiently understood to make a precise demarcation possible between embedded projections and embedded expansions: while the difference between projection and expansion in hypotactic clause complexes is fairly straightforward, in the case of embedded constructions, it is not always obvious whether a structure represents a

200 Factive nominalization: Towards a descriptive position direct representation of experience or a metaphenomenon. Do the embedded structures threatening people and gathering quietly around the campfire each evening in (65) and (67), for instance, represent direct representations of experience or are they facts, i.e. ideas or linguistically processed phenomena? And which lexicogrammatical properties motivate their interpretation? Like the occurrence of ambiguous and vague predicates, the problems of delineation which characterize the category of gerundive nominalizations thus reveal the need to further refine the analysis of factive nominalization as embedded projections. Finally, the analysis which Lees gives of action and factive nominalizations may not suffice to draw the line between factivity and non-factivity, but it does make clear that an analysis of the internal structural properties of factive nominalizations might be able to contribute to their analysis as embedded projections or facts. By pointing to the tendency for action gerundives to occur without Subject and auxiliaries, while factive gerundives can take auxiliaries and frequently include a Subject, Lees suggests that there exists a correlation between a nominalization's meaning and its internal structural behaviour. My approach to factive nominalizations builds on Lees's insights and will largely consist in analyzing the internal, nominal properties of the various types of factive nominalizations. In the following section, I will therefore consider the proposals that have thus far been made with respect to the internal properties of factive constructions.

4. Towards an internal, nominal analysis The interest in the external structural behaviour of factive nominalizations and the attempts to map out the various contexts in which factive clauses can be embedded, pioneered by Kiparsky and Kiparsky (1971), have elucidated many aspects of the system of factive nominalization. One of the drawbacks of the focus on the external functioning of factive clauses has been the lack of attention devoted to their internal structural properties. The internal features of factive constructions have mostly been left unanalyzed, and have not been integrated with the way in which the nominalization functions externally as a nominal constituent. They have, moreover, tended to be treated as irrelevant to the semantic interpretation of factive nominalizations. If the internal structural properties of factive nominalizations are considered, it is the presence of clausal features which is typically highlighted: factives are conceived of as "imperfect nomináis" or as nomináis "in which the verb is still alive as a verb" (Vendler 1967: 131). 7%a/-structures in par-

Towards an internal, nominal analysis

201

ticular are analyzed as being fully clausal internally. Their nominalized status is attributed solely to the fact that they function as nomináis externally and that they can serve as Subject and Object in the clause: as Kruisinga and Erades (1953: 108) put it, embedded /Aa/-structures show "no connection between function and form". The various labels that have been used in the literature for embedded iAa/-structures are all meant to capture this apparent contradiction between their internal, clausal properties and their external, nominal-like behaviour: /Aaf-structures have been referred to as 'noun clauses' (Sweet 1891: 170), 'Subject/Object clauses' (Poutsma 1929; Kruisinga 1932; Kruisinga and Erades 1953), 'clausal nominalizations' (Comrie 1976; Cortine and Thompson 1985) and 'nominal clauses' (Quirk et al. 1985: 1047). Unlike ίΑαί-structures, gerundive nomináis have typically been argued to have an internal structure that is partly nominal.59 The discussion of the internal structure of factive gerundives has, however, generally been restricted to listing its clausal and nominal aspects (an elaborate overview of the nominal and verbal features of gerundive nomináis can be found in, among others, Lees 1960; Chomsky 1970; Pullum 1991). With the exception of Lees (1960), linguists generally seem to have tacitly accepted that the internal structure of factive gerundives, like that of ί/ζαί-factives, does not contribute to the factive meaning with which it is associated. Analyses that consider the internal properties of the various types of factive nominalizations and try to link them up with the factive meaning of the category are sparse. In the next section, I will successively discuss the analyses offered by Halliday (1994) and by Langacker (1991) and show that both of them stand out in that they try to reconcile the nominal-like external behaviour of factive nominalizations with the properties of their internal organization.

4.1. Halliday's dependency analysis As I have shown earlier, Halliday's analysis of factive nominalizations as embedded constituents is primarily elaborated in terms of the relation that exists between an embedded structure, on the one hand, and the larger configuration which it forms part of, on the other. Also included in his descriptive analysis of factives, be it somewhat less prominent, is an analysis of the internal structure of embedded ¿Λαί-structures. Behind it is the idea that an embedded clause functions in the structure of a nominal, i.e. as a "constituent WITHIN" it (Halliday 1994: 242). More specifically, Halliday pos-

202 Factìve nominalization: Towards a descriptive position its that a factive /¿¿¡/-structure either functions as head of the nominal, or as postmodifier to a head noun (Halliday 1994: 242, 266): (70) That you 're sorry isn't enough. (71) The fact that you 're sorry isn't enough. Huddleston (1988: 145fn) criticizes Halliday's internal analysis of thatfactives and argues that, rather than functioning as head of a nominal, thatstructures are immediately Subjects or Objects in the clause. The main argument which Huddleston gives against Halliday's analysis is that embedded finite structures "cannot enter into construction with any of the Pre- or Postmodifiers that are found in genuine nominal groups" (Huddleston 1988: 145fh). McGregor (1997) voices a comparable criticism. As I pointed out in Chapter 3 (Section 3.2.1), McGregor argues that downward rankshift necessarily involves internal reclassification, whereby the downranked unit "takes on the grammatical and semantic properties inherent to the lower ranking unit" (McGregor 1997: 127). Unlike non-finite clauses, McGregor argues, finite clauses do not show any sign of reclassification: semantically speaking, a finite clause designates not a thing, but an ongoing phenomenon located at some point in space and time; structurally speaking, it is not characterized by any reduction in the number of grammatical options that are available for anchoring the situation to a specific occurrence. McGregor concludes that /¿«/-structures are not internally reclassified and cannot therefore be analyzed as rankshifted. Even though Huddleston has a point when he argues that /¿«/-structures cannot take pre- or postmodifying structures and should not therefore be analyzed as heads, the alternative analysis which he proposes basically goes back to the early definitions of embedded /¿¿//-structures as Subject or Object clauses that show no correlation between function and form (e.g., Poutsma 1929; Kruisinga 1932; Kruisinga and Erades 1953). McGregor's analysis likewise results in inexplicable discrepancies between the external functioning of finite embedded clauses (e.g., as Subject in [That he did it] came as no surprise) and their internal structural features, which are claimed to be those of a non-rankshifted clause. By analyzing finite clauses as non-rankshifted and non-finite clauses as rankshifted or reclassified, McGregor moreover fails to clarify the semantic correspondences between them. In the next section, we will see how Langacker's functional analysis of the nominal group and his 'type-instance' approach to nominalizations does allow one to grasp both the external and internal reclassification of factive /¿«/-clauses as nomináis.

Towards an internal, nominal analysis 203 4.2. Type vs. instance nominalization Langacker (1991: 23) suggests that nominalizations of the types illustrated in (72) can be profitably examined from the standpoint of "the internal organization of the relational predication, in particular whether it represents the conception of a process type or a specific instance of that type" (examples based on Langacker 1991: 422): (72) a. Zelda's reluctant signing of the contract surprised the entire crew. b. Zelda's reluctantly signing the contract surprised the entire crew. c. That Zelda reluctantly signed the contract surprised the entire crew. The nominalization in (72a) is called an 'action nominal' in Lees (1968: 64) and is to be distinguished from the 'action' type of gerundive nominal which I discussed earlier (e.g., his job involves [my answering the phone on his behalf quite a bit of the time] and [Eating vegetables] is healthful). The structure in (72b) represents a factive gerundive, and that in (72c) is a factive /Aa/-structure. Langacker (1991) zooms in on the internal functional organization of these nominalizations and argues that the structural differences that exist between them follow from the fact that they derive from three different "levels of organization in the assembly of a finite clause" (Langacker 1991: 33), i.e. a verb stem, an 'intermediate processual structure' and a finite clause. These structural levels, he says, correspond to a type, an ungrounded and a grounded instance respectively. The action nominalization Zelda's reluctant signing of the contract is analyzed as a 'type' nominalization because the nominalization process involves the reification of a verb stem (sign) into a noun (signing), of which the "semantic function ... is limited to specifying a type" (Langacker 1991: 33). Of course, if that noun is used as head of a nominal, it becomes an instance and the nominal means of instantiation and grounding apply. The gerundive nominalization in (72b), on the other hand, is claimed to derive not from a verb stem realizing a process type, but from an intermediate processual expression or ungrounded instance (sign the contract), which Langacker (1991: 34) defines as "a structure that is like a finite clause except for the absence of an explicit subject and a predication of tense or modality". The structure's instantiated status is said to follow from the fact that it "incorporates a fully specified direct object" (Langacker 1991: 34). The result is a 'complex noun' which functions as head noun, the Subject of which is expressed by means of possessive periphrasis (Zelda's signing the contract). Finally, unlike gerundive nomináis, /AaMiominals are said to derive from a

204

Factive nominalizatioti: Towards a descriptive position

grounded instance or a finite clause, whose internal structure is retained in the nominalization. Interestingly, Langacker also attempts to link up the structural properties of the action and gerundive types of nominalizations with the meanings which they express. In the case of action nominalizations, which start from a simple process type specification and are turned into simple nouns, he points out that they behave like nouns also semantically and can therefore profile various facets of the reified event, such as the manner in which it is carried out (73a), its duration (73b), its propriety (73c) and even the factuality of the event (73d) (Langacker 1991: 32): (73) a. Harvey's b. Harvey's c. Harvey's d. Harvey's

taunting of the bear was merciless. taunting of the bear lasted three hours. taunting of the bear was ill-advised. taunting of the bear came as a big surprise.

Compare this with the meanings in which the simple noun party can be used (Langacker 1991: 33): (74) a. The party was boisterous. b. The party lasted three hours. c. The party was ill-advised. d. The party came as a big surprise. As to the factive meaning of certain gerundive nominalizations, Langacker remarks that the "construal of a reified process as representing a particular instance renders it compatible with contexts that presuppose its factuality" (Langacker 1991: 34). He argues that a construction such as Harvey's taunting the bear can be interpreted factively because it is derived from the intermediate processual expression taunting the bear, which "represents an instance and not just a type" (Langacker 1991: 34). Because it is not finite and therefore not grounded, however, a gerundive nominal can also occur in non-factive contexts. Compare the factive use of Harvey's cruelly taunting the bear in (75) with the non-factive use in (76) (Langacker 1991: 34): (75) Harvey's cruelly taunting the bear was a severe blow to his campaign for presidency. (76) Harvey's cruelly taunting the bear is something that could simply never happen.

Towards an internal, nominal analysis

205

In short, Langacker's analysis of action nominalizations, gerundive nominalizations and that-structures suggests that an analysis of the internal organization of nominalized constructions can benefit from taking into account the functional level of assembly of their processual starting-point. It also shows that the functional constructs of type specification, instantiation and grounding are relevant to the description of nominalizations and shed light not only on many of their structural properties, but also on the meanings that are linked to, for instance, action and factive nominalizations. Some of the details of Langacker's type/instance analysis of nominalizations, however, are in need of further clarification: especially the analysis of gerundive nominalizations can be refined. Langacker's discussion of gerundive nomináis moreover only deals with gerundives that have a periphrastic Subject. As could be deduced from Lees's analysis of gerundive nominalization, however, it is especially in the group of gerundives that lack an explicit Subject that problems of delineation between factive and non-factive gerundives occur. The divergent internal properties of all the different types of gerundive nominalizations should be included in the analysis of factivity. In the next chapter, I will try to give a precise and systematic descriptive analysis of the internal functional properties of factive gerundives, thatfactives and the fact fAa/-constructions. As well, I will show that a functional analysis of all three subtypes of factive nominalization is a precondition to come to a precise analysis of the category of factive nominalization and to link up the notion of factivity with the lexicogrammatical properties of the nominalized structures that realize it.

Chapter 9 A functional analysis of factive nominalizations as nominal constructions

This chapter will present an analysis of the internal organization of factive nominalizations as nominal constructions. It aims at counterbalancing the prevailing focus on the external nominal contexts in which factive nominalizations figure as clause constituents. Assuming those as one aspect of the rank shift involved in factive nominalizations, it will complete the picture by tracing the internal reclassification involved in the various types of factive nominalizations, by virtue of which they fulfil nominal functions such as type specification and grounding. The factive nominalizations that I will consider are either based on full finite clauses (i.e. íAaí-factives and the fact /Aa/-constructions) or on an 'intermediate processual expression' (in the case of gerundive factives). I will try to account for their internal functional organization in terms of the functional model of the nominal and the clause set out in Chapter 4. In a first part (Sections 1, 2 and 3), I will successively zero in on thatfactives, the fact /Aai-constructions and factive gerundive nominalizations and describe how each of them makes the transition from processual starting-point to nominal construction. In Section 4, the main results of these descriptions and the implications for the analysis of the system of factive nominalization in general will be pointed out.

1. ίΑαί-factives: From finite clause to proper name strategy The clausal level of assembly which ίΑαί-nominalizations start from is that of a full finite clause or a clause which profiles a grounded instance of a process type (Langacker 1991: 34). It was shown in Chapter 4 that a finite clause grounds its process type specification in two ways: it establishes a temporal/modal link with the speech event, and it relates the process type to a grammatical person. Person deixis, it was argued, is expressed subjectively on the Finite (by means of Subject-Finite concord) and is objectively realized by the Subject (Davidse 1997). While the point that /Aa/-structures start from finite clauses is obvious and not controversial, the claim that the resulting nominalized //^/-constructions show parallels with specific nominal paradigms is more contentious.

208 Factive nominalizations as nominal constructions I pointed out before that it is widely agreed that factive /Aai-structures serve essentially nominal functions in the clause, viz. functions such as those of Subject and Object (e.g., among others, Sweet 1891; Poutsma 1929; Kruisinga 1932; Lees 1960; Kiparsky and Kiparsky 1971; Comrie 1976; Quirk et al. 1985). That ίΑαί-structures also fit into the nominal paradigm internally seems less obvious. The following two points are crucial to come to terms with the internal properties of íAaí-structures. First, it is important to recognize the downranked character of the reclassification which nominalized ί/ζαί-structures realize (Halliday 1994). And, secondly, it has to be acknowledged that the nominal paradigm allows for structures which do not realize the functions of type specification, quantification and grounding iconically.

1.1. Downranking as functional reclassification Let us start with the downranked status of iAcz/-constructions. As argued by Halliday (1994), factive ί/ζαί-structures are clauses that have been made to function at a rank which is lower than that of the clause, i.e. that of the nominal. It is because /Aa/-factives are downranked clauses that they have retained so many of their clausal properties. Deverbal -er nomináis and action nominalizations of the type the opening of the door simply contain reclassifications of a verbal unit into a noun, i.e. a unit which belongs to the same rank but realizes a different class. By contrast, fAai-structures are reclassified through rank shift and the finite clause which they start from is nominalized as a whole: a clause that is downranked is downranked with its internal clausal properties untouched. A downranked finite clause thus retains its finite status. The only effect which downranking has on the internal clausal properties of the downranked clause is that they are as it were viewed through an additional, nominalized viewing frame which construes the process in it "holistically" rather than scanning it sequentially or stepby-step (Langacker 1991: 440, 449) (for Langacker's notions of 'sequential' vs. 'holistic' scanning, see also Chapter 4, Section 2.1.1). Even though a downranked clause retains most of its clausal properties, its being downranked implies that it has to discharge the functions of the accommodating class. Downranking implies functional reclassification 'from below'. For factive ί/ζαί-clauses this means that they, in one way or other, discharge the nominal functions of type specification, instantiation, quantification and grounding. In the next section, I will show how, following Langacker (1991), embedded /Aa/-structures can be fitted into the nominal paradigm and linked up with nominal strategies.

that-/actives: From finite clause to proper name strategy 209 1.2. The nominal paradigm In Chapter 8,1 described how Halliday analyzes //^/-structures as basically realizing the common noun strategy. In his view, /Aa/-structures realize the head of the nominal (Halliday 1994: 266). He also claims that /Aa/-factives alternate with nomináis in which they do not function as heads themselves, but postmodify a nominal head, as, for instance, in the fact that he left her. As correctly pointed out by Huddleston (1988: 145fn), Halliday's analysis of /Aa/-structures as head of the nominal is questionable since, unlike ordinary nominal heads, /Aa/-structures cannot take any pre- or postmodification. While I agree with Huddleston's criticism and reject Halliday's analysis of /Aa/-structures as nominal heads, I do not follow Huddleston in the conclusion which he draws, viz. that /Aa/-structures cannot be analyzed as nomináis. I believe that they can be fitted into the nominal paradigm, if only one does not restrict the nominal strategies which a nominalized clause can adopt to that of common nouns only. I thus come to the second precondition for identifying the internal nominal functioning of //¡«/-structures: it has to be acknowledged that there exist nomináis which do not realize the functional schema of 'grounded instance of a type' iconically or by means of a nominal head and determiners. In fact, it is by no means exceptional to have nomináis which do not need determiners to instantiate and ground the type specification (Langacker 1991: 148, 150). Pronouns and proper names, for instance, are inherently definite. As to postmodifiers, they characterize a type of Thing in terms of some process within which it is directly or indirectly a participant (Halliday 1994: 188). Inherently definite units such as pronouns and proper names do not take this kind of postmodification. Huddleston's rejection of the nominal status of //ia/-structures on the basis of the absence of pre- and postmodifiers can now be argued not to hold: even among ordinary nomináis, examples are found of nomináis without any pre- or postmodifiers. Nouns such as pronouns and proper names are uniquely identifiable and fully characterized. They function as fully grounded nomináis in their own right (Langacker 1991). As pointed out by Langacker (1991: 149), /Aa/-factives are part of this nominal paradigm: embedded /AaZ-structures do not function as heads of the nominal, but they are fully grounded nomináis themselves. Factive that-structures have in common with pronouns and proper names that they are fully characterized and uniquely identified in relation to the speech context. Langacker (1991: 149) in my opinion rightly points out that in the case of nominalized thatstructures, "the unique circumstances of the grounding relationship are incorporated", as a result of which the "specified type ... has only a single

210 F active nominalizations as nominal constructions instance, with the consequence that the derived noun is inherently definite". Like pronouns and proper names, therefore, /Aai-clauses do not require "separate grounding" to function as nomináis (Langacker 1991: 148). I will refer to the nominal strategy which embedded /Aa/-structures employ as the 'proper name strategy'. The unique combination of clausal and nominal properties which thatfactives realize can now be mapped out in more detail. With ¿Aaí-structures, the nominalization process applies to a finite clause. By being downranked, this finite clause is reified and construed as an abstract entity, with its process scanned holistically rather than sequentially. Rather than effecting fundamental changes to the clausal internal properties of the finite clause, the downranking mechanism reclassifies the clause into a nominal, capable of discharging nominal functions such as grounding according to the proper name strategy. Like proper names, factive ίΑαί-clauses establish an entity which is definite in itself and can therefore function as a fully grounded nominal. Thus, by following Langacker's analysis of factive iAa/-constructions, one can analyze them as nominal-like not only externally, but also internally. This counters both Huddleston's and McGregor's analysis of embedded ίΑαί-structures. McGregor (1997: 127) argues that finite clauses cannot be rankshifted because rankshift or reclassification always implies that the downranked unit "takes on the grammatical and semantic properties inherent to the lower ranking unit". Finite clauses "show none of the formal properties ... which are characteristically associated with nominalizations" (McGregor 1997: 131). It is especially the finite or fully grounded status of /Aai-clauses which, according to McGregor, prevents them from being downranked as nomináis: he argues that finite clauses "always represent situations as situations, not as entities" (McGregor 1997: 131). Against McGregor (1997), it can now be argued that thatnominalizations do function as nomináis internally. First, as pointed out by Langacker (1991: 448), it is essential to acknowledge that with embedded /Aa/-structures "there are multiple levels of organization to consider, and multiple paths through which the contents of a /Aa/-clause are accessible from the ground". Reclassification need not be marked by means of a reduction of the grounding options that are available within the downranked clausal unit. Nor does it always require that the reclassified unit adopts particular morphological and syntactic properties that are characteristic of the accommodating class. Instead, reclassification is junctional in nature. In the following section, I will zoom in on the fact ¿/¡^/-constructions and show how they provide additional evidence for the 'proper name' analysis of í/zaí-structures.

the fact ihat-constructions: A case of apposition 211 2. the fact /Art-constructions: A case of apposition As pointed out in Chapter 8 (Section 4.1), Halliday analyzes the fact thatconstructions like those in (1) and (2) as that-factives functioning as postmodifiers of the head noun: (1) He was, I think, troubled by the fact that the Democratic Party couldn't put together a majority on the national level.... (CB) (2) And I don't like the fact that it's been carnivalized. (CB) Quirk et al. (1972: 647, 1985: 1260) likewise treat the that-clauses in the fact ¿Aírf-constructions as postmodifiers of the fact-noun. To distinguish them from the restrictive type of relative postmodifying clauses (as in 3), they describe /Aai-structures in the fact iAaf-constructions as 'appositive'. (3)

The news that appeared in the papers this morning was well received. (Quirk et al. 1985: 1244)

As the most obvious difference between relative and appositive postmodification they point out that appositive that functions as a conjunction, rather than as a relative pronoun. Halliday also discusses the different status of the embedded that-clause in restrictive relative clauses and in the fact thatclauses. In restrictive relative clauses, the clause is said to form a direct representation of reality and of how the language user experiences it (Halliday 1994: 242). The relationship between the restrictive relative that-clause and the antecedent is consequently one of expansion. The that-c\mse in the fact f/^/-constructions, in contrast, itself forms a projection or metaphenomenon (see Chapter 8, Section 2.2). Rather than representing reality in a direct way, it represents a phenomenon that has already been linguistically processed (Halliday 1994: 264). As indicated in Chapter 8, in Halliday's view, facts come as 'ready-packaged' projections, i.e. they are not projected by the clause in which they function. Yet another type of that-clause which is considered to be postmodifying is exemplified in (4) and (5): (4) the thought that she might one day be a queen (Halliday 1994: 264) (5) the news that the team had won (Quirk et al. 1985: 1244) Quirk et al. (1985) group these //¡ai-structures with the appositive type of postmodifying clauses. Halliday (1994: 263) calls them "embedded locutions and ideas", functioning as postmodifiers within a nominal group. He

212

F active nominalizations as nominal constructions

argues that, like ί/ζαί-clauses postmodifying the noun fact, they represent projections or metaphenomena, but, unlike in the fact iAaf-structures, "the projecting element is the noun that is functioning as Thing" (Halliday 1994: 263). In the nominal in (4), for instance, it is the noun thought which projects the that- clause that follows it. The analyses which Quirk et al. (1972, 1985) and Halliday (1994) suggest for the relationships between embedded /Aa/-clauses and the noun classes that precede them can now be summarized as follows: Table 4. The analysis of 'postmodifying' ίΑαί-clauses as presented in Quirk et al. (1972, 1985) and Halliday (1994) the fact that apposition head-postmodifier / the ίΑαί-clause is an embedded projection or metaphenomenon

restrictive relative clauses relative head-postmodifier expansion the that-clause is an embedded direct representation of experience

the thought/claim that apposition head-postmodifier projection the /Aa/-clause is an embedded projection or metaphenomenon

Notice that all embedded fÄa/-clauses are analyzed as postmodifiers, in spite of the fact that embedded projections realize an appositive relation and restrictive relative clauses a relative relation. Secondly, it can be observed that Halliday analyzes the semantic relationship of the that-clause and the noun that precedes it only for restrictive relative clauses and embedded locutions/ ideas: the relation between the fact and the that- clause in the fact ¿/zai-constructions is not further specified. In the next section, I will focus on the relationship of apposition which has been argued to exist between the the fact ίΑαί-nominal and the thatclause. In a first part (Section 2.1), I will discuss the structural characteristics of restrictive apposition and compare them with those realized by restrictive relative clauses and in the thought/claim /Aa/-constructions. I will suggest that Quirk et al.'s (1985) and Halliday's (1994) analysis of the fact ^¿-constructions in terms of modification is incompatible with their restrictive appositive status. In Section 2.2, then, I will zoom in on the functional characteristics of the appositive units and compare the features of the fact iAa/-constructions with those of non-nominalized cases of apposition.

the facttiaat-constructions:A case of apposition 213 2.1. The structural characteristics of the fact /Aaf-constructions Quirk et al. (1972: 647, 1985: 1260-1261) analyze the fact ίΑαΖ-clauses as consisting of two units which are related to each other in a restrictive, appositive way (see also Langacker 1991: 149). But what characterizes restrictive apposition? Apposition is said to be a relationship between "constituents of the same level" which are "identical in reference" (Quirk et al. 1972: 620). In, for instance, Paul Jones, the distinguished art critic, both Paul Jones and the distinguished art critic refer to the same person: the relation between them can therefore be described as appositive. Appositive units are restrictively linked when they form a single information unit, i.e. when they are uttered on a single tone unit and are not separated by commas in writing: the apposition between Paul Jones and the distinguished art critic in Paul Jones, the distinguished art critic can thus be analyzed as non-restrictive. Examples of restrictive apposition are the famous critic Paul Jones, the singer Robeson and the River Thames. The characterization of the fact ίΑαί-structures as appositive nominal groups thus implies that they establish a relation between two units which are of equal status. At the same time, however, Quirk et al. (1985: 1261) classify the /¿¿¡/-clauses in the fact /Aû/-constructions as postmodifiers. I argue that the appositive nature of the fact /Aaf-constructions is not reconcilable with their being described in terms of modification (see also Biber et al. 1999: 645) and that, consequently, the use of the notion of postmodification in the description of the fact /Aai-constructions is unjustified. Let us have another look at Halliday's account of the structural relations that may exist between members of a complex. Halliday (1994: 221) argues that the notions of hypotaxis/parataxis and expansion/projection (described in Chapter 8, Sections 2.1 and 2.2) "define complexes at any rank" and can also be used to describe complex relations at the level of the nominal. In general, parataxis is the "linking of elements of equal status", while hypotaxis is the "binding of elements of unequal status" (Halliday 1994: 221). The initiating and continuing element in a paratactic relationship are 'free' "in the sense that each could stand as functioning whole", while in a hypotactic relationship only the dominant element is free (Halliday 1994: 221). Parataxis thus forms an alternative to a relationship of modification, which always involves one element modifying another (Halliday 1994: 218). As far as I can see, the the fact ίΑαί-nominal and the ίΛαί-clause are paratactically and not hypotactically linked to each other: they are not related through modification, but form elements on the same level, with each of them having the potential to stand on its own. The construction the fact that Caesar was dead can, for instance, be replaced by the fact or by that

214 Factive nominalizations as nominal constructions Caesar was dead without affecting the acceptability of the clause in which it is embedded:60 (6)

a. The fact was obvious to all. b. That Caesar was dead was obvious to all.

Rather than functioning as a postmodifier, therefore, the /Aa/-clause in the fact fAai-constructions has a status which is equal to that of the the fact that-nominal. The that-clause cannot be analyzed as postmodifier, and neither can it be classified as Qualifier to the fact-noun in terms of its representational semantics: the /AaZ-clause does not characterize the type of Thing that precedes it in terms of a process in which it is a participant (as argued in Halliday 1994: 188). Instead, it represents a second Thing. To fully grasp the nature of the paratactic relationship that exists between the the fact that-nomm&X and the /Aa/-clause, its semantic status should be analyzed as well (Halliday 1994: 243). The main distinction to be made here is that of expansion versus projection, expansion being a relationship between direct representations of experience and projection involving metaphenomena or linguistically processed ideas (Halliday 1968; 1994). Halliday distinguishes three types of expansion relationships. Elaboration implies that one element restates another element in other words, specifies it in greater detail or exemplifies it. A clausal example of elaboration is given in (7a). When two elements are related by means of extension the second element adds some new element to the first one, gives an exception to it or offers an alternative (e.g., 7b). Enhancement, finally, involves a second clause which qualifies the first one with some circumstantial feature of time, place, cause or condition (e.g., 7c) (Halliday 1994: 220). (7)

a. John didn't wait; he ran away, [elaboration] b. John ran away, and Fred stayed behind, [extension] c. John was scared, so he ran away, [enhancement]

Each of these expansion types can be combined either with parataxis (as in all the examples in 7) or with hypotaxis. In fact, in considering the various possible mappings, Halliday defines 'apposition' as a combination of elaboration and parataxis (Halliday 1994: 225): when two items are in apposition, the second item elaborates the first and specifies it in more detail. The type of semantic relationship which the fact ίΑαί-constructions establish can now be identified. Halliday (1994) only gives a semantic characterization of embedded /AaZ-clauses: they are projections without explicit projecting clause. However, when a that-factive comes to function as the

the fact that-constructions: A case of apposition 215 continuing element in a paratactic, appositive relationship with a nominal containing the fact noun, it also enters into a semantic relationship with that NP. It seems to me that the paratactic relationship between the thatstructure and the fact NP is one of elaboration : the ίΑαί-clause restates and specifies the 'fact' referred to by the fact NP; it clarifies what exactly the fact is about. We can now determine the precise position of the fact /¿ai-structures with respect to the other construction types described as 'postmodifying' by Quirk et al. (1985) and Halliday (1994). First, restrictive relative clauses can be analyzed as postmodifiers. They have the function of qualifying the type of Thing designated by the head: the general type news in the news that appeared in the papers this morning is thus further subcategorized as (news) having appeared in the papers. Importantly, as pointed out by Langacker (1991: 432), a restrictive relative clause "restricts the head noun's type specification", i.e. it constitutes an element of the type specification. Nominal grounding therefore applies to the combination of head noun and restrictive relative clause (see also Davidse 2000b). Because they function as postmodifiers, restrictive relative clauses cannot function on their own: (8)

a. The news was well received. b.*That appeared in the papers this morning was well received.

Finally, restrictive relative clauses establish a semantic relationship which can be characterized in terms of elaboration (with the relative wA-element in the embedded clause restating the nominal antecedent, Halliday 1994: 243) or as one of extension (whereby the relative clause is viewed as adding a new element to the nominal type, thus turning it into the higher-order type specification that is grounded by the nominal): (9)

the man who came to dinner

As to embedded locutions and ideas or íAa¿-clauses following nouns such as assertion, claim, rumour, thought, belief, view, news, Halliday (1994: 263) suggests that they be analyzed as containing "projecting nouns": the relation between the two appositive units which they establish is in that case one of projection. Embedded locutions and ideas are said to differ from the fact /^-constructions in that they can be systematically related to clause complexes in which the that-structures are ranking rather than embedded projections: (10) the assertion that/ They asserted that Caesar was ambitious.

216 F active nominalizations as nominal constructions (11) the thought that/ They thought that Caesar was ambitious. (12) the claim that/ They claimed that Caesar was ambitious. However, I feel that there is an important distinction in status between the /^¿-structures in the projecting clause complexes in (10) to (12) and the /Aa/-clauses that are embedded in the corresponding nomináis. While in the projecting clause complexes, the relationship between the two clauses is hypotactic, embedded locutions and ideas are appositive constructions and paratactically combine two units which are of equal status. This can be deduced from the fact that the two components can function on their own, as illustrated in (13) and (14) (the existence of /Àaf-clauses which represent embedded locutions and ideas was first pointed out in Davidse 1994): (13) a. The rumour that he killed her is not true, b. That he killed her is not true. (14) a. Just the thought that he may have killed her is unbearable, b. That he may have killed her is unbearable. In view of the fact that there is no reason to assume that a noun can serve a projecting function, I propose to analyze constructions of this type also as establishing a semantic relationship of elaboration. Like the fact thatstructures, the /Aa/-clauses in the claim/thought /Aai-constructions restate and specify the nominal that precedes it. Like the fact /^/-constructions, the claim/ thought /^/-constructions thus fit into the paradigm of apposition and realize a combination of parataxis and elaboration. It follows that the appositive relationship established in a nominal such as the thought that Caesar was ambitious does not differ from that in the fact //^/-constructions: both are based on embedded projections, both paratactically relate the two units which they consist of, and in both the thatclause restates and specifies the nominal that precedes it. In general, the relationships that are established in the fact //¡^/-constructions, restrictive relative clauses and constructions of the type the claim that can now be summarized as follows, with the shaded areas indicating that the proposed analysis deviates from the analysis suggested in Quirk et al. (1972, 1985) and Halliday (1994) (see Table 4):

the fact Ubat-constructìons: A case of apposition 217 Table 5: An alternative analysis of embedded ίΑαί-clauses

the fact that

restrictive relative clauses

the thought/claim that

apposition

relative

apposition

parataxis

head-postmodifier, i.e. hypotaxis

parataxis

expansion: elaboration the ίΑαί-clause itself is an embedded projection or metaphenomenon

expansion: extension/ expansion: elaboration elaboration the that-ciause is an the f/zaf-clause itself is an embedded direct repre- embedded projection or metaphenomenon sentation of experience

The overview of the various structural and semantic relationships that can be established between nominal and that-clause makes it possible to demarcate the properties of the category of the fact iAaf-constructions more accurately. What characterizes the fact /Aa/-structures is that they establish a structural relationship between two elements which are considered to be of equal status, viz. the fact and the iAa/-clause. The semantic relation between them is one of elaboration: the ίΑαί-clause restates and further specifies the the fact-nommàì. It is especially their paratactic status which turns out to distinguish appositive structures such as the fact /Aa/-constructions from restrictive relative clauses.61 Interesting to observe is also that, unlike what is claimed in Halliday (1994), there is no fundamental structural difference between the fact /Aa/-structures and the thought/claim thatconstructions. As I will show in the next section, the precise difference between these constructions only becomes clear when their functional properties are considered more closely. To come to a comprehensive account of the properties of the fact ¿ΑαΖ-structures, therefore, the functional organization of the fact íAaí-nominalizations has to be looked into as well.

2.2. The functional properties of the fact í/zaí-constructions As far as their functional organization is concerned, appositive structures stand out for several reasons. Most strikingly, appositive nomináis contain two units "each of which is itself a nominal". As Langacker (1991: 149) puts it, a composite structure such as the singer Robeson is "a doubly-

218

Factive nominalizations as nominal constructions

characterized, doubly-grounded thing". For nomináis to be able to function in the most prototypical type of restrictive apposition (e.g., the famous critic Paul Jones, the singer Robeson, the number three) they should, moreover, "display minimal definite determination" (Van Langendonck 1999: 116): in other words, the nomináis that figure in restrictive apposition are necessarily definite. Either they have a definite article (e.g., in the famous critic, the singer) or they do not take a determiner because they are fully grounded themselves (as in Paul Jones, Robeson). Because they "designate the same conceived entity but describe it in different ways" (Langacker 1991: 432), there are also interesting differences between nomináis in apposition. Restrictive apposition of the type the singer Robeson is characterized by the fact that the first unit "is preceded by a definite determiner ... and is more general than the second appositive" (Quirk et al. 1972: 638). Unlike the second nominal, the first appositive unit can take modifiers freely, e.g., famous in the famous critic Paul Jones (Lee 1952; Quirk et al. 1972, 1985). The second appositive seems to necessarily lack an explicit determiner and it cannot take modifiers (Haugen 1953; Van Langendonck 1999). It is, for instance, impossible to have *the poet (the) great Burns instead of the great poet Burns (Van Langendonck 1999: 117). These structural differences have been argued to reflect important differences in function: while the first element in a construction like the poet Burns is classificational, the second is specific (see, for instance, Lee 1952; Haugen 1953; Quirk et al. 1972, 1985; Langacker 1991; Schmid 2000). Importantly, Van Langendonck (1999: 113) claims that restrictive apposition, or 'close apposition', represents "a crucial criterion for proper names". In close apposition, he argues, "the unit with the potentially most specific reference is a proper name" (Van Langendonck 1999: 116), while the other one functions as a common noun and helps to "specify the subclass to which the proper name belongs" (Van Langendonck 1999: 117). Behind this claim lies the view that proper names are essentially identified by the way in which they function in a construction. Van Langendonck (1999) argues that a distinction should be made between proprial lemmata, on the one hand, and the grammatical category of proper names, on the other. Proprial lemmata represent 'dictionary entries' and may be used in various ways. In the case of the name John, for instance, it is possible to use it with unique reference, i.e. to refer to a uniquely identifiable person with that name (as in 15), or it can be used as a common noun, as illustrated in (16) (Van Langendonck 1999: 100): (15) John attended a meeting today.

the fact Oast-constructions: A case of apposition 219 (16) You are talking about a different John. What constitutes a proper name, then, is derived from the way in which an item functions in a construction and units which are not proprial lemmas may also be used as proper names (Van Langendonck 1999: 96). The close apposition test thus identifies not only proprial lemmas (e.g., Robeson in the singer Robeson), but also structures of which the primary function is not that of proper names. Van Langendonck (1999: 118) gives the following examples: (17) a. The song 'The Seven Drunken Knights' b. The question 'What does it mean to live in modern society?' As Van Langendonck (1999: 118) points out, "semantically and syntactically" structures like these "behave as proper names in these appositional structures in the same way as the units in the standard appositional syntagms": they show individual reference and they can function on their own syntactically. The restrictive type of apposition realized by the fact /Actf-structures significantly resembles that found in constructions like the singer Robeson and the poet Burns. As in ordinary nominal appositive structures, the first nominal (the fact) and the second nominal (the ίΑαί-clause) differ in terms of how they describe the fact which they designate, but they are identical in reference. Notice that the that-clause in the fact ίΛαί-constructions thus does not specify the nominal type fact in greater detail (as do restrictive relative ίΑαί-clauses), but relates to the nominal the fact as a whole. Like restrictive appositive structures such as the poet Burns, moreover, a the fact ί/ιαί-structure starts with a common noun which is preceded by a definite determiner and which can be premodified, as in: (18) The ugly fact that he was holding a gun indicated his guilt. (Quirk et al. 1985: 1262) The primary function of the first nominal in the fact ί/ζαί-structures is classificational: it indicates which 'class' or type of things the second nominal belongs to. As Schmid (2000: 14) points out, the first nominal (which he calls the "shell noun") serves the "semantic function of characterizing and perspectivizing complex chunks of information which are expressed in clauses or even longer stretches of text". Van Langendonck (1999: 120) describes the function of the common noun in restrictive apposition as designating the 'basic level' meaning or the category of the second

220 Factive nominalizations as nominal constructions nominal. Like the second nominal in the appositive structure the poet Burns, then, the /AaZ-clause in the fact //¡aZ-structures is more specific and specifies which fact is being referred to. Because the finite //ia/-clause is uniquely grounded, it functions as a full nominal (see Section 1). In terms of Van Langendonck's analysis of restrictive appositive constructions, the //¡¿¡/-clause can be said to serve the function of proper name in appositive the fact fAa/-constructions. The position of the fact that-structures in the overall system of factive nominalization can now be clarified: they classify the (factive) /Aa/-clauses which they contain explicitly in terms of the nominal class of fact; they identify the class which the embedded /Aa/-constituent belongs to as being that of facts. It is here that the main difference with embedded locutions and ideas lies: in the thought/claim /Aa/-structures, the //¡«/-clause is identified as realizing the category of locutions and ideas, rather than that of facts. In contexts that allow for a factive as well as a non-factive reading, a /Aa/-clause which stands on its own as nominal may therefore be interpreted either as an embedded locution or as a fact: (19) a. That she refused the job makes sense to me. b. The fact that she refused the job makes sense to me. c. The news/rumour that she refused the job makes sense to me. The inclusion of the fact-noun, moreover, helps to "activate one single idea, because nouns have a higher potential for reification than nominal thatclauses" (Schmid 2000: 367). Because they include an explicit and nominal classification of the //¡«/-clause, it can be argued that the fact //¡«/-structures come closer to non-nominalized or ordinary nomináis, which might explain why they can be used in clausal contexts that do not allow ordinary thatfactives (see, for instance, Davidse 2003), as in (20) and (21): (20) a. We are pleased by the fact that you work so hard. (Davidse 1994: 270) b. *We are pleased by that you work so hard. (21) a. ... it's going to take some adjusting to the fact that Anchorage is a one-paper town. (CB) b. *... it's going to take some adjusting to that Anchorage is a onepaper town. Finally, the fact //¡«/-constructions also confirm the analysis of thatfactives as full nomináis that use the proper name-strategy. In the fact thatstructures, the embedded /Aa/-clauses function as second nominal in the ap-

Gerundive facts 221 positive relation to the fact. As such, they enter into a relation with the nominal paradigm that normally functions in that particular position. Because in non-nominalized cases of appositives (like the poet Burns), the type of nominal that figures as second element in a restrictive appositive relation has the properties of a proper name (e.g., Burns), the /Aa/-clause in the fact ί/ιαί-constructions can be said to connect with proper names. The analysis of the nature of the relationship which the fact /¿«/-constructions realize thus confirms what I suggested in my analysis of ίΑαί-factives, viz. that /Aa/-factives are paradigmatically related to nomináis realizing the proper name strategy.

2.3. the fact /Aaí-struc tures: Conclusion The transition from clausal starting point to nominal structure realized by the fact ¿Ααί-structures can be summarized as follows: like /Aai-factives, the fact ίΛαί-constructions start from a finite clause which is downranked and functionally reclassified as a nominal, more specifically, one following the proper name strategy. Because it functions as a proper name, it can also enter into a relationship of apposition with other nomináis. The appositive construction which the that-clause comes to function in then categorizes or classifies the that-clause as being a fact.

3. Gerundive facts: From atemporal clausal head to common noun or proper name strategy To reveal the differences and the similarities between that-factives (including those that function in appositive the fact ίΑαί-constructions) and gerundive factives such as those in (22) and (23), I propose that it is necessary to consider three things: firstly, the clausal level of assembly which gerundive nomináis start from; secondly, the nominal strategies which the embedded clausal unit in gerundive nominalizations adopts and, thirdly, the way in which the nominal and clausal properties interact in the resulting nominalized structure. In this section, I will elaborate each of these aspects in more detail. (22) Since old Nanapush had saved her from death ..., her visiting the man stood to reason. (CB) (23) She appreciated the clerk's taking the time to be so friendly ... (CB)

222 Factive nominalizations as nominal constructions 3.1. From atemporal clausal head... It is fairly obvious that the main difference between iAai-factives and gerundive factives is that the former are based on a finite clause, while the latter are not. Which clausal level of assembly it is that gerundive nominalizations derive from, however, is unclear. In his analysis of nominalizations, Langacker (1991: 33) argues that gerundives start from an "intermediate level which profiles an instance that is left ungrounded". A gerundive nominalization such as Zelda's signing the contract is thus claimed to derive from the processual expression sign the contract. I will propose an alternative analysis (see also Heyvaert 2000) which views gerundive nominalizations as being based on an 'atemporal clausal head' (as discussed in Chapter 4, Section 2.1). A nominalization such as Zelda's signing the contract will be analyzed as deriving from the atemporal clausal head signing the contract. Rather than being an ungrounded instance, the processual starting point of gerundive nominalizations is thus an atemporalized process type specification. This is relevant for the analysis of gerundives as nominal constructions: I will argue in Section 3.2. that it is the iype-status of the clausal head which determines the nominal strategies that gerundive nominalizations can adopt. We will start from a number of well-known observations concerning the internal status of gerundive nominalizations. As Lees (1968: 65-66) points out, gerundive nominalizations can take auxiliaries (as in 24a), they can be used with an adverbial (illustrated in 24b) and they are construed with participants which are not periphrastically realized but as clausal constituents (see 24c) (see also, among others, Fraser 1970; Chomsky 1970; Pullum 1991; Hye Suk Yoon 1996; Malouf2000): (24) a. his having brought up the box b. his drawing the picture rapidly c. his drawing the picture Gerundive nominalizations differ in these respects from so-called 'action nomináis' like the signing of the contract (Lees 1960). Action nominalizations are generally considered to be more 'nominal' in nature: they cannot take any auxiliaries (as shown in 25a), they are used with adjectives rather than with adverbials (see 25b), and they realize their participants periphrastically (illustrated in 25c): (25) a. *his having brought up of the box b. his rapid drawing of the picture

Gerundive facts

223

c. his drawing of the picture It has been deduced from these lexicogrammatical properties that in action nomináis, the suffix -ing itself has nominalizing power and creates a noun. Declerck (1991b: 496) refers to this type of nominalization as 'nominalized gerunds'. The -ing structure found in gerundive nominalizations, in contrast, seems to be processual in nature (e.g., among others, Quirk et al. 1985; Declerck 1991b; Pullum 1991; Hye Suk Yoon 1996; Taylor 1996). If we compare the processual properties of gerundive nominalizations with those of atemporal clausal heads (see Chapter 4, Section 2.1), some striking similarities can be observed. First, as in atemporal clausal heads, the -ing suffix in gerundive nominalizations can attach to both perfective and imperfective verbs. Lees (1968: 66) already pointed out that gerundive nominalizations can be formed on all types of verbs, including 'stative' ones, while action nomináis such as the signing of the contract require what he calls 'action' verbs: (26) a. his having a hat, but *his having of a hat b. his resembling his mother, but *his resembling of his mother c. his believing her, but *his believing of her Moreover, the -ing structure in gerundive nomináis can be systematically related to a fo-infinitive, which constitutes the atemporal alternative to -ing in atemporal clausal heads used in clauses as well: (27) a. his having a hat -» for him to have a hat b. his resembling his mother -» for him to resemble his mother These to-infinitive constructions can be used in many factive contexts in which gerundive nomináis can also be used: (28) For him to have a hat/His having a hat surprised me. (29) For him to resemble his mother/ His resembling his mother came as a surprise. Significantly, the processual constructions found in gerundive nominalizations include clause-like realizations of Objects: it was argued in Chapter 4 that in the clause, Objects are part of the processual type specification and, thus, of the atemporal clausal head (Davidse 1997). As we will see in Section 3.3, the processual expression in gerundive nominalizations can also take a (periphrastically realized) Subject, as illustrated in John's sign-

224 Factive nominalizations as nominal constructions ing the contract. In non-finite clauses as well, atemporal clausal heads combine with Subjects, as in: (30) Those who ate the apples ran off wild and mad into the forest - the apples having been poisoned by a woman who had loved Merlin .... (CB) Finally, like atemporal clausal heads, the clausal structure in gerundive nominalizations can include a range of auxiliaries: (31) Their having to attend it sometimes puts upon them. (CB) (32) ... her having been able to follow it into full, ecstatic awareness of God was sheer grace. (CB) (33) ... its originality and sheer brilliance vindicate completely her having been declared Doctor of the Church. (CB) Importantly, the modal auxiliaries have to in (31) and be able to in (32) and the perfect auxiliary in having been declared in (33) do not themselves ground the clausal head: they are atemporalized by the -ing suffix and therefore essentially belong to an atemporalized process type specification in which they further modify the simple verb stem. In short, the lexicogrammatical properties of gerundive nominalizations suggest that the processual level of assembly which they start from is that of an atemporal clausal head. This analysis ties in with well-known observations about the clause-like status of gerundive nominalizations. As to the precise effect of the -ing suffix on the verb it is added to, then, I would argue that, as in atemporal clausal heads, -ing imposes an atemporal profile on the process and forms an alternative to grounding it. Rather than relating the process to the speech event by means of modality/tense, the atemporal profile of the -ing suffix confirms the ungrounded status of the process type specification and serves to derive a higher-order type specification. -Ing only changes the way in which the process is looked at: rather than scanning the component states of the process in the step-by-step or sequential way that is characteristic of finite clauses, -ing scans the component states of the process type holistically. As also argued in Chapter 4 (Section 2.1.1), it activates the component states of the verb cumulatively, through summary scanning. Because they realize summary scanning rather than sequential scanning, atemporal clausal heads are likely candidates for nominalization: as pointed out before in the analysis of that-factives (Section 1.1), the nominalizing of clause-like units (such as finite clauses) turns them into unitary entities of

Gerundive facts 225 which the component states are scanned simultaneously or holistically and which can be manipulated and used in other clauses (Langacker 1991: 440). By using atemporal clausal heads in nominal units, an additional viewing frame is, in other words, imposed on them. Because internally, atemporal clausal heads already offer a holistic perspective on the component states of the process type, they come closer to the holistic end-profile of the nominalized construction in which they figure than do embedded //^/-structures (where the holistic, nominalized perspective applies to finite clauses or clauses which have sequential scanning internally). It is probably their internal atemporal status which explains why gerundive factives, like the fact //^/-constructions, can figure in nominal contexts which are not open to //^/-structures. The example in (34a), for instance, alternates with a the fact /Aa/-construction, but not with a /Aa/-structure: (34) a. Last year, Mgr. Jacques Gaillot ... was ousted by the Vatican for openly promoting the use of condoms to prevent the spread of Aids. (CB) b. ... was ousted for the fact that he openly promotes the use of condoms to prevent the spread of Aids. c. *... was ousted for that he openly promotes the use of condoms to prevent the spread of Aids. Importantly, describing gerundive nominalizations as containing a specific functional level of assembly of the clause provides an alternative to the traditional analyses of gerundive nominalizations which describe them vaguely as 'clause-like' or 'verbal' (e.g., among others, Lees 1960; Quirk et al. 1985; Pullum 1991; Hye Suk Yoon 1996; Malouf 2000). The analysis proposed here makes clear that gerundive nominalizations are 'clause-like' because they derive from the functional level of the atemporal clausal head, which includes Objects, adverbials and non-grounding auxiliaries of the clause. The processual levels of assembly which deverbal nominalizations can start from (Langacker 1991) can now be identified as being either a verb stem or simple process type specification, a higher-order, atemporalized process type specification, or a full finite clause. In my analysis of gerundive nominalizations, I will show that there is a fourth level of assembly which nominalizations can be derived from, viz. that of ungrounded instances or atemporal clausal heads with a Subject in the objective case (Section 3.3). It has to be noted that the analysis of gerundive nominalizations as being based on a clausal head that realizes a clausal type specification deviates from the analysis which Langacker suggests for them. As I pointed out be-

226 Factive nominalizations as nominal constructions fore, Langacker (1991: 33) analyzes the gerundive nominal Zelda's signing the contract as being derived from the 'ungrounded instance' sign the contract. It is the inclusion of the Object which is held responsible for the instantiated nature of the structure sign the contract·. "The fact that the nominalized structure incorporates a fully specified direct object supports the claim that it represents an instance and not just a type" (Langacker 1991: 34). Langacker adds that even though "-ing attaches to the verb from which this structure inherits its processual profile, it is the structure as a whole with its clause-like internal organization already established - that is nominalized" (Langacker 1991: 32). I see two main differences between Langacker's analysis and the one which I propose. First, my analysis considers the Object as part of the clausal head or type specification and assigns an instantiating role to the Subject only. Arguments in favour of these claims were given in Chapter 4, Section 2.2 (see also Davidse 1997). Secondly, in my description of gerundive nominalizations, the functional level of assembly at which the nominalization process takes place is that of signing the contract rather than sign the contract. Of course, the atemporal clausal head signing the contract itself is based on the simple process type specification sign the contract. Crucially, however, sign the contract is not the structure which the nominalization applies to. Langacker's analysis remains vague about the precise nature of the -ing of gerundive nominalizations and the difference with the nominalizing -ing in action nomináis such as the signing of the contract. By describing gerundive nominalizations as being based on an atemporal structure which includes the atemporal marker -ing, I explicitly identify the nature of the -ing suffix in them as being verbal rather than nominalizing. It can be pointed out here that the verbal status of the -ing suffix in gerundive nominalizations such as Zelda's signing the contract is not recognized by everyone. While in many of the descriptions of gerundive nominalizations it is posited that the suffix -ing is verbal (e.g., among others, Horn 1975; Schachter 1976; Quirk et al. 1985; Pullum 1991; Hye Suk Yoon 1996), some have claimed that it has nominalizing value (Baker 1985; Milsark 1988). As Milsark (1988: 622), for instance, puts it: "The structures that have classically been called gerundives - that is, those that occur in subject position and in object position of prepositions and of verbs such as regret and enjoy - are ... those whose verbs have assumed nominal categorization as a result of the affixation of -ing". At the same time, Milsark considers the -ing form in gerundive nominalizations to be functioning as head of a gerundive clause, and he argues that "the nominal features of the derived verb form percolate up its projection path, ultimately relabeling the entire gerundive clause as NP" (Milsark 1988: 613). In my opinion,

Gerundive facts 227 Milsark's analysis confuses the atemporal effect of the -ing suffix with nominalization. As I pointed out before, the holistic perspective which -ing offers on the component states of the process comes close to the holistic viewpoint which nominalization in general has on clause-like structures (including finite structures) (see Langacker 1991). The two should not, however, be confused: the atemporal status of -ing does not itself effect nominalization, witness the occurrence of atemporal -ing in subordinate (or non-downranked) non-finite clauses. The atemporal status of -ing does, however, seem to be conducive to nominalization and an atemporal structure can, when nominalized, function in contexts that are more typically 'nominal' in nature than can nominalized expressions which are finite. Milsark's analysis may confuse atemporalization with nominalization, but it does make clear that it remains to be explained how an atemporal clausal head can be nominalized and fulfil nominal functions in the clause. What is it that turns an atemporal clausal head into a gerundive nominalization with nominal-like behaviour? In Hye Suk Yoon (1996), it is suggested that the atemporal clausal component which gerundive nominalizations derive from is nominalized by a process of 'zero derivation', which leaves intact the internal processual status of the component but enables it to function as a nominal in larger configurations. I do not see any reason for analyzing the nominal behaviour of atemporal clausal heads in gerundive nominalizations as resulting from what is basically a morphological process, i.e. that of zero-derivation. Rather, I propose that, like the finite clause in rAa/-factives and the fact ί/ιαί-constructions, the atemporal clausal head in gerundives is downranked, with its internal processual status left intact. More specifically, it is only when Halliday's construct of downranking is interpreted as functional reclassification that the lexicogrammatical properties of factive gerundives can be accounted for and the differences and similarities with 'action' nominalizations can be explained. Both turn out to involve the reclassification of a processual structure, but they realize two different types of reclassification. The -ing of an action nominalization such as the signing of the contract realizes the reclassification of a verb stem (sign) into a noun (signing): it nominalizes the verb and changes the class it belongs to, but it does not alter the rank which the verb functions on: both the base verb and the derived noun belong to the rank of the word. Action nominalization thus realizes a type of reclassification which does not involve downranking. In the gerundive nominalization signing the contract, in contrast, the nominalization process is not realized by -ing: the -ing suffix is merely an atemporal verbal marker and does not shift the class of the verb it attaches to from verb to noun. Rather, the entire atemporal clausal head is downranked within the structure of the nominal.

228

Factive nominalizations as nominal constructions

Crucially, the downranked status of gerundive nominalizations is, like that of ^«/-structures, primarily functional in nature and the downranked clausal head adopts the functional properties of a nominal head. In the following section, I will go more deeply into the precise nature of the functional reclassification of atemporal clausal heads in gerundive nominalizations. As in my analysis of that-factives and the fact //¿ai-constructions, I will zoom in on the paradigmatic relations which the nominalized processual structure enters into and try to identify the precise nature of the nominal strategy which it connects with. To fully grasp the properties that characterize factive gerundives, I will situate factive gerundives in the broader system of gerundive nominalization and consider non-factive types of downranked gerundive constructions as well.

3.2. ... to common noun or proper name strategy In Pullum (1991), the rather provocative claim is defended that gerundive nominalizations are nomináis with a verb phrase head: the gerund in gerundive nominalizations is claimed to be "a verb buried inside the nonfinite verb-phrase head of a noun phrase" (Pullum 1991: 786). Pullum bases his claim on the observation that the head of gerundive nominalizations does not behave like ordinary nominal heads in a number of respects (Pullum 1991: 768-770): the head of gerundive nominalizations can, for instance, take clausal negation not, which cannot be used with ordinary nominal heads (illustrated in 35); gerundive nominalizations cannot take attributive adjectival modification (see 36), nor do they permit restrictive relative clause formation (as illustrated in 37). Finally, unlike in ordinary nomináis, when the head of a gerundive nominalization follows a genitive, it cannot be ellipsed and interpreted anaphorically (see 38): (35) a. my not having left, but *my not books b. John's not destroying the city, but ""John's not destruction of the city (36) his kind offer, but *his kind walking me home (37) *his leaving her that you predicted (38) a. I was amazed at Stacy's eagerness and at Morgan's too. b. *I was amazed at Stacy's being so eager and at Morgan's too. Against Pullum (1991), I claim that gerundive nominalizations do not result from some kind of 'mismatch' of a verbal head and a nominal constituent. I will argue that the downranked atemporal clausal heads that fig-

Gerundive facts 229 ure in gerundive nominalizations are functionally reclassified and come to function either as nominal heads or as full nomináis. In fact, as in embedded ίΑαί-clauses, there exists a correlation between the internal level of grounding of the embedded processual unit and the nominal (grounding) strategy which the downranked unit adopts. In my analysis of that-factives, I have argued that the proper name strategy which a downranked finite clause adopts corresponds to its profile of uniquely identified, grounded clause (see also Langacker 1991). In the previous section I have shown that gerundive nominalizations derive from a level of assembly which profiles what is basically a type specification. The nominal behaviour of gerundive nominalizations will now be revealed to be in accordance with the specific status of its head. I will start with a brief recapitulation of Lees's (1960) description of gerundive nominalizations and the distinction which he makes between 'action' and 'factive' gerundive nomináis. I will then go more deeply into the 'non-transformational' analysis which Schachter (1976) offers of the internal nominal functioning of gerundive nominalizations (Section 3.2.1). In section 3.2.2, then, I will set out my own view of gerundive nominalization.

3.2.1. Gerundive nominalizations: from transformation to nominal constituent I explained in Chapter 8 that Lees (1960) identifies two semantic subtypes of gerundive nominalizations, i.e. factives and action-referrent gerundives (the latter category is to be distinguished from Lees's category of 'action nomináis' such as the signing of the contract). Some of the examples which he gives of the factive type are reproduced in (39). In (40), gerundive nominalizations of the action type are given: (39) a. His eating vegetables is surprising, b. His dressing himself is funny. (40) a. Eating vegetables is healthy, b. Dressing oneself is fun. Gerundive nominalizations which are action-referrent, Lees (1968: 72) argues, "cannot have any expressed subject, nor any auxiliary". Factive gerundives, in contrast, can occur "with or without the genitive subject" and they can take auxiliaries (Lees 1968: 72). An example of a factive nominalization without Subject is the following:

230 Factìve nominalizations as nominal constructions (41) Having gone pleased us. Lees specifies that, if in factive gerundives the Subject is not realized overtly, the Subject of the gerundive nominalization will be interpreted as being similar to that of the matrix clause, or to another participant realized in it: in (41), for instance, the Subject of having gone is interpreted as being the we that is implied in the us of the matrix clause. Schachter (1976) tries to formulate an alternative to the transformational approach given by Lees and argues in favour of "an analysis in which gerundive nomináis are assigned an underlying structure more like that of ordinary noun-headed noun phrases" (Schachter 1976: 206). Two aspects of his account of gerundive nominalizations are particularly interesting: first, he suggests that the interpretation of subjectless gerunds "has to do with the more general matter of the interpretation of nomináis with and without overt determiners" (Schachter 1976: 210). Secondly, he compares the nominal behaviour of gerundives with that of uncount nouns and plural count nouns. He argues that subjectless gerunds "are actually determinerless gerundive nomináis" (Schachter 1976: 211) which are either related to the generic use of uncount nouns and plural count nouns, as in (42) or designate specific instances, as in (43) (Schachter 1976: 214-215): (42) a. Going to the beach is enjoyable. b. Milk does something for every body. (43) a. I enjoyed going to the beach yesterday. b. There's milk all over the kitchen floor. The gerundive nominal in (42a) is said to refer to a class of activities, rather than to a specific instance. It is comparable to the generic use of uncount nouns. The gerundive nominal in (43a), in contrast, designates a specific instance and is comparable to the non-generic use of the uncount noun in (43b). Schachter (1976: 215) concludes that the "determinerless NP itself is thus merely a class name, and this name may be used to refer either to the whole of the class or to some specific instance of it, according to the context". Schachter's analysis of gerundive nominalizations breaks new ground in that it is the first to point out systematic resemblances between gerundive nominalizations and ordinary, non-nominalized nominal structures·, it compares the downranked clausal head in gerundive nominalizations with heads of ordinary nomináis and points out correspondences that exist between the determination of gerundives and that of uncount nouns and plural count nouns. By bringing into the picture the nominal properties of gerundive

Gerundive facts

231

nominalizations, Schachter's analysis breaks quite radically with the linguistic tradition of approaching nominalizations as 'transformed clauses' and as consisting of clausal categories only. His description of the nominal status of gerundives, however, also leaves room for improvement: Schachter's analysis can be broadened out and other types of gerundive nominalizations should be included in the description. Certain aspects of the analysis which he proposes for determinerless gerundives can moreover be refined. In the following section, I will show how a new light can be shed on gerundive nominalizations by tracing how the downranked clausal heads function in them as reclassified full nomináis or nominal heads. I will argue that a downranked clausal head can either adopt the common noun strategy (using determiners or quantifiers) or the proper name strategy (with the reclassified clausal head functioning as full nominal).

3.2.2. Towards a systematic account of the nominal properties of gerundive nominalizations The various subtypes of gerundive nominalizations that will be analyzed are illustrated in the following examples: (44) a. Only that could account for their having a suitcase with them. (CB) [possessive pronoun, factive] b. Techniques for doing this include the caregiver's exposing the infant to various contingent and discrepant experiences. (CB) [genitive, non-factive] (45) a. ... showing one's feelings too clearly (or being highly sensitive) is regarded as a 'feminine' trait. (CB) [no determiner, nonfactive] b. Having attended Harvard can be helpful. (Schachter 1976: 215) [no determiner, factive or non-factive] (46) a. Of course I regret getting pregnant. (CB) [no determiner, factive] b. Going there was fun. (Postal 1970: 478) [no determiner, nonfactive] (47) a. This burning the midnight oil of yours has got to stop. (Schachter 1976: 218) [definite determiner, non-factive] b. Any talking loudly on your part will be punished. (Schachter 1976: 218) [indefinite determiner, non-factive]

232

Factive nominalizations as nominal constructions

By considering the nominal functioning of the downranked atemporal clausal heads in these nominalizations, I will show that each of them constitutes a different subtype of gerundive nominalization. I will start with what are considered to be the prototypical instances of factive gerundive nominalizations, i.e. gerundive nomináis containing a possessive determiner or a genitive preceding the downranked atemporal clausal head, as illustrated in (44a), (48) and (49): (48) I appreciate your taking the time to talk to us. (CB) (49) I think the country's making a statement is a lot bigger than partisan politics (CB). It is this type that is most often referred to when the co-occurrence of nominal and clausal properties in gerundive nominalizations is pointed out. It is also this type of gerundive nominalization which most clearly shows that the atemporal clausal head has been reclassified, in this case as a nominal head. As such, it adopts the common noun strategy and can function like any other nominal head: it can be preceded by either a possessive pronoun (as in 48) or a genitive (see 49), which serve to identify a definite and specific subset of it. The instance which gerundive nominalizations like these refer to is part of the concrete, physical world of instances and is identified in terms of the system of person, defined from the standpoint of the speaker (Halliday 1994: 181). As regards the use of the genitive in factive gerundive nominalizations, it should be pointed out that factive gerundive nominalizations do not seem to occur with the indefinite type of genitive illustrated in (50): (50) Is there a/some diligent student's notebook here? As argued by Woisetschlaeger (1983; cited in Davidse 1999: 224), the identity of a/some diligent student's notebook is not retrievable and nonspecific. The fact that I have not come across factive gerundives with this type of genitive suggests that the genitive that functions in gerundive nominalizations is necessarily referential. Importantly, the definite, specific type of reference which is realized by the factive nominalizations in (48) and (49) is not unique to factive nominalizations: there exist non-factive or action-referrent gerundive nominalizations with a possessive or a genitive preceding the atemporal clausal head as well. This 'action' type of gerundive nominalizations - even though occasionally mentioned in the literature (e.g., Quirk et al. 1985: 1064; De-

Gerundive facts 233 clerck 1991b: 497) - has always been overshadowed by the factive gerundive prototype. Consider the following examples: (51) Techniques for doing this include the caregiver's exposing the infant to various contingent and discrepant experiences. (CB) (52) My husband speaks very well, but his job involves my answering the phone on his behalf quite a bit of the time. (CB) (53) And then all of the sudden, it fractures again and a third time, and the third time, there's an enormous ripping sound, and that is her having the stroke. (CB) (54) Some researchers have felt that having a mental illness would lead to the individual's occupying a position in a lower social class .... (CB) None of the nominalizations illustrated here can be replaced by a factive the fact that or the fact «^construction. Rather than being facts, they represent a type of gerundive nominalization which differs from the category of action gerundives that is described in Lees (1960) and Schachter (1976) in that it does not profile a class or type, but links the nominal head to a specific instance the identity of which is defined by the possessive or genitive which precedes it. Let us turn to another type of gerundive nominalization. Which nominal strategy can action gerundive nominalizations of the type illustrated in (45a), (55), (56) and (57) be said to follow? (55) The bedrock of character is self-discipline. A related keystone is being able to motivate and guide oneself, whether in doing home-work, finishing a job or getting up in the morning. (CB) (56) Self-awareness also involves recognising one's strengths and weaknesses and seeing oneself in a positive but realistic light.... (CB) (57) But others, like this junior, admitted that knowing about AIDS and protecting oneself from it are two very different things. (CB) Schachter (1976: 214) compares the nominal behaviour of gerundive nominalizations like these to the generic use of uncount nouns (take, for instance, the use of milk in Milk does something for every body)·, like the related generic type of mass nouns, these nomináis are said to name a kind or class of activity or condition, rather than a specific instance of it. In Davidse (1991: 364), nominalizations such as those in (55) to (57) are characterized as "merely specifying a 'type' of a process". I will call them the generic category of action gerundives.

234 Factive nominalizations as nominal constructions Which kind of nominal grounding strategy do generic action gerundives follow? I refer here to the analysis of generically used mass nouns and plural count nouns in Davidse (1999). Following Carlson (1978), Davidse posits that mass nouns and plural count nouns which are generic in profile behave like proper names with respect to quantification and grounding: their generic reading excludes in principle explicit quantification, because they designate just one instance, viz. the kind or type which they refer to as such. Compare, for instance, the generic mass noun in (58) with its nongeneric counterpart in (59): (58) a. Milk helps to make your bones stronger. b. *Some milk helps to make your bones stronger. (59) a. There's milk on the table. b. There's some/much milk on the table. Similar to proper names, moreover, the use of the bare nominal is sufficient to identify which kind is being referred to: definite grounding is thus part of the internal structure of a generically used mass noun (Davidse 1999: 211). Rather than referring to an indefinite instance in physical space (as does milk in 59) or to a uniquely identified instance in physical space (as do ordinary proper names), however, generically used mass nouns designate specific instances in type space. They are basically names for kinds that are presented as instances. Notice that the definite status of generic action gerundives is supported by the fact that they can occur as second nomináis in close apposition: (60) Politics is the act of reallocating power. (CB) (61) Thirdly, the act of getting married is expensive. (CB) It can be concluded that the category of action nominalizations that refers to kinds is not only, as Schachter correctly argues, generic in profile, it is also essentially definite in nature and functions in the clause as a definitely grounded nominal. Interestingly, Schachter (1976) also points to the existence of generically used gerundives which contain the auxiliary of perfect aspect have. Remember that in Lees's analysis, the use of such an auxiliary is explicitly excluded from the action-referrent type of gerundive nomináis. Schachter (1976: 215) gives the following example: (62) Having attended Harvard can be helpful.

Gerundive facts 235 Schachter (1976: 215) argues that this nominal "seems to name a type of condition resulting from some completed activity". I believe Schachter's claim to be essentially correct: some action gerundives contain the auxiliary have. In my opinion, however, the clause in (62) can receive not only a generic reading, but also a reading in which a specific instance is being referred to. Rather than being necessarily generic, in other words, the nominal is ambiguous. On the generic interpretation, no particular instance is envisaged. The atemporal clausal head having attended Harvard - which designates a type specification - then comes to function as a definitely grounded nominal with generic reference. Notice that, as I argued in Chapter 4 (Section 2.2), the presence of the auxiliary of perfect aspect have does not itself instantiate or ground the process type specification. Of course, the secondary tense which have realizes is indirectly situated with respect to the ground. Yet, it does not itself instantiate or ground the process type. The nominalization having attended Harvard in (62) can, in my opinion, also be analyzed as designating a specific instance, the identity of which is retrievable from the context of speech. If, for instance, I am telling someone that I have attended Harvard, and that person answers that Having attended Harvard can be helpful, the gerundive nominalization is factive in meaning and designates a specific instance. As such, it agnates with the nominal the fact of your having attended Harvard. Factive gerundives of the type having attended Harvard will be discussed in more detail later on in this section. An example of a gerundive nominalization which resembles that in (62) in that it is ambiguous between a generic, action reading and a factive interpretation is that in (63): (63) Do you consider not having been to art school a disadvantage? Firstly the Subject of this nominal may be coreferential with the Subject of the matrix clause; the speaker then asks the yow-person how he or she feels about the fact that he/she has not been to art school. While this reading might be the most obvious one here because there is a nominal in the matrix clause, viz. you, which can readily be interpreted as identifying the Subject of the gerundive nominalization (which is not the case in 62), a generic reading cannot be excluded. One can, for instance, imagine the question being directed to the head of an art school, who has been to art school himself: the nominalization in that case does not refer to a specific instance, but it designates a kind of condition. The nominal not having been to art school thus comes to be interpreted as generic, on a par with having attended Harvard in its non-specific meaning. A generic reading, finally,

236

Factive nominalizations as nominal constructions

seems to be the most likely reading for the gerundive nominalization in (64): (64) Gandhi ... also reminded Americans of color not to be ashamed of their heritage of slavery; there was dishonor only in having been slaveholders. (CB) Unlike in (63), the Subject of the gerundive nominalization cannot be retrieved from the matrix clause: it is not the 'Americans of color' which should be ashamed of having been slaveholders, but rather the white inhabitants of America. The nominalization having been slaveholders may therefore be analyzed as profiling a kind of condition resulting from some completed activity, rather than a particular instance. As such, it has definite generic reference. Let us now turn to the type of gerundive nominalization that is exemplified in (65) and (66): (65) I resent being tagged a problem boy. (CB) (66) Being dropped by Scotland before the Five Nations was a disappointment .... (CB) In Schachter's (1976) view, these nominalizations designate specific instances and are comparable to the non-generic use of the mass noun milk in There's milk on the kitchen table. It was pointed out in Chapter 4 (Section 1) that mass nouns such as milk in There's milk on the kitchen table are used indefinitely and take cardinal quantification (Davidse 1999: 222), as can be inferred from the fact that they alternate systematically with absolute or cardinal quantifiers such as some and much: (67) a. There's milk on the table. b. There's some/ much milk on the table. Schachter's analysis of these nomináis thus implies that they are indefinite and take cardinal quantification. However, it is impossible to add quantifiers such as some and much to the gerundives in (65) and (66). (As I will show later on in this section, however, action-referrent gerundives can in some very marked cases take indefinite quantifiers.) Conceiving them as indefinite moreover seems to conflict with the fact that they designate specific instances of a process type. Against Schachter, I therefore argue that nominalizations like the ones in (65) and (66) function as proper names, much like the factive ίΑαί-structures that were discussed in Section 1 of this

Gerundive facts 237 chapter: they designate an individual instance of the reclassified process which is instantiated in physical space. The proper name analysis which I propose for the factive gerundive nomináis in (65) and (66) is supported by their semantic and formal behaviour. First, semantically speaking, the downranked atemporal clausal head is conceived as being instantiated and the instantiator of the process is retrievable from the context: it is, for instance, clear that being tagged a problem boy in (65) refers to the /-Subject of resent·, similarly, in (66), it is implied that it is the speaker who was dropped by Scotland before the Five Nations. The nominal into which the instantiated process type is reclassified can therefore be argued to function as a nominal with definite reference: as is the case with gerundives taking a genitive or a possessive, it is clear which instance is being referred to. Unlike in gerundive nominalizations with a genitive/possessive, however, there is no explicit determiner indicating the instantiator of the process and signalling the definite status of the nominalization. Rather than following the common noun strategy, nominalizations like those in (65) and (66) follow the proper name strategy: they designate an inherently definite instance of a process type. Because they function as definite proper names, they can occur as second nominal in a restrictive appositive structure: (68) ... had decided to boycott me for the mere fact of having written a book on Franco that was not a denunciation. (CB) (69) But, as he was now recognizing, the fact of having registered as a conscientious objector effectively closed the door .... (CB) It should be pointed out that, like gerundive nomináis with a possessive/genitive, this type of gerundive is not necessarily factive: there also exists an action-referrent type of gerundive that designates a specific instance of a process type but lacks a possessive or genitive, as in: (70) Getting together quickly for coffee would be fun. (Thompson 1973: 377) (71) Adjourning immediately at 4:00 was impossible. (Thompson 1973: 377) As with the factive gerundives in (65) and (66), the specific instance of the process type that is being referred to can be retrieved from the context. Action gerundives of this type are also definite, and they can therefore function in appositive constructions:

238

F active nominalizations as nominal constructions

(72) Prue drew a momentary blank, still flushed with guilt over being caught in the act of searching Luke's shack. (CB) (73) Ayatollah Khomeini had said the act of severing ties with the United States was a correct move. (CB) The final category of gerundive nominalizations which I will discuss assumes a rather marginal position within the overall system of gerundive nominalizations. It can be illustrated by the following example given by Schachter (1976: 218): (74) Any talläng loudly on your part will be punished. We can also think of examples such as (75) More/some/0 reporting from you would be welcome, (personal example) These nominalizations follow the common noun strategy in which an indefinite quantity of the nominal head is designated by the use of determiners such as any, some, 0 and more. As Schachter (1976: 218) points out, this constructional schema can also be used with definite determination, as in (76): (76) That burning the midnight oil of yours has got to stop. Remarkably, the head can take postmodification (from you, on your part and of yours), which specifies the instantiator of the downranked clausal process type. The gerundive nominalizations in (74) to (76) therefore designate, as observed by Schachter, instances of a process type, rather than being generic in focus. However, gerundives of this type which do not specify the instantiator of the process periphrastically seem to viate more towards generic reference, as in (77) More/some/0adjourning example)

at 4.00 p.m. would be welcome, (personal

It should also be noted that this - rather marginal - type of gerundive seems possible only with action gerundives, not with factives.

Gerundive facts 239 3.2.3. The nominalfunctioning of gerundive nominalization: Conclusion Gerundive nominalizations are interesting in that, on the one hand, they apply the proper name strategy (found also with factive thatnominalizations) which reclassifies the clausal structure as a fully grounded nominal. On the other hand, they also display the common noun strategy, by downranking the atemporal clausal head into nominal head position and integrating it with nominal determiners. Used as proper names, gerundives can refer either to classes or individual instances of process types (depending on whether the process itself is contextually instantiated or not). As for the gerundives that follow the determiner plus common noun constructional schema, the common type with possessive/genitive always refers to individual instances of processes, precisely because the possessive/genitive designates the instantiator of the process. The marked type with indefinite (e.g., some, any, 0 ) or definite determiner allows periphrastic expression of the instantiator of the process by postmodification (from you, of yours, on your part). If this is not the case, this type seems inclined to generic reference. All the options listed above are represented by the horizontal rows in Table 6. Factive gerundives turn out to display more restrictions with regard to these options than action ones. Action gerundives have all possibilities, but factive gerundive nominalizations never designate classes of a process they always refer to definite instances of a process, either by means of the proper name or the common noun strategy. The options available for factive and action gerundives are charted in the columns of Table 6.

240

Factive nominalizations as nominal constructions

Table 6. An overview of the nominal strategies adopted by gerundive nominalizations

generic reference; proper name strategy

reference to an individual situation; proper name strategy

reference to an individual situation; common noun strategy

definite, specific

definite or indefinite, specific

ACTION gerundives

FACTIVE gerundives

Eating vegetables is healthy.

/

Adjourning at 4:00 was impossible.

I resent being tagged a problem boy.

... his job involves my answering the phone on his behalf.

... the country's making a statement is a lot bigger than partisan politics.

This burning the midnight oil of yours has got to stop.

/

Some reporting from you would be welcome.

The analysis of the internal nominal behaviour of gerundive nominalizations thus helps to distinguish the various subtypes of gerundive nominalization more accurately, and it provides valuable information about the nature of factive gerundives, as opposed to gerundives which designate actions. In this respect, some headway has been made with the aim of complementing the functional analysis of gerundives 'from above', e.g., as complements of the clause, with one 'from below', looking at the functions

Gerundive facts 241 of grounding, quantification, instantiation and type specification of the nominal constructions which gerundives form. In the next section, I will concentrate on how precisely the functional properties of the clausal level of assembly which factive gerundive nomináis derive from are integrated with the functional organization of the nominal.

3.3. Factive gerundive nominalization: Integrating the clausal and nominal paradigms I have thus far argued that factive gerundive nominalizations are based on atemporal clausal heads which are functionally reclassified and downranked either as the common noun head of a nominal (with a possessive or a genitive preceding it), or as proper names (with no determiner). Importantly, even when gerundives come to function as common nouns, they do not behave as prototypical members of the common noun category. Pullum (1991) rightly observes that gerundives fail to display certain properties that characterize non-nominalized common noun heads (e.g., adjectival modification and restrictive relative clauses are impossible; see Section 3.2). I claim that it is only by analyzing how the processual functional organization of the downranked clausal head is integrated with the nominal structure of factive gerundives that we can understand such restrictions and the phenomenon of factive gerundives as such. Let us recapitulate the nominal forms of determination which factive gerundives can establish. We have seen in the previous section that factive gerundives always function as definitely grounded nomináis: they function either as common nouns that are grounded by a possessive or a genitive in front of the atemporal clausal head (e.g., Only that could account for [their having a suitcase with them], CB); or they function as proper names and the instantiator of the specific instance which they designate is retrievable from the matrix clause or the context (e.g., [Being dropped by Scotland before the Five Nations] was a disappointment...., CB). Strikingly, factive gerundive nominalizations cannot take any of the other ordinary nominal determiners: structures like [Caroline's/her opening the door] surprised me danài regret [opening the door] do not relate to (78) a. *The opening the door surprised me. b. *I regret that opening the door.

242 Factive nominalizations as nominal constructions It should also be noted that the genitive of gerundive nominalizations alternates with nouns which are in the objective or common case, as shown in examples (79) and (80) (for a detailed overview of the contexts in which the common case is used rather than the genitive, see Poutsma 1929 and Kruisinga and Erades 1953).62 This type of gerundive structure constitutes a fourth clausal level of assembly from which deverbal nominalizations can be derived, i.e. that of the non-finite clause. (79) As matters stand, Tory MPs who resent John Major outmanoeuvring them in last year's leadership coup will always find journalists more than willing to indulge their conspirational ambitions. (CB) (80) I don't approve of the crew's only remaining man going there. (Lees 1968:72) How can the non-occurrence of articles and demonstrative pronouns and the possibility of using the objective case besides the genitive be explained? In the literature on gerundive facts, the possessive and the genitive have mostly been referred to as the Subjects of the nominalized process (e.g., among others, Kruisinga and Erades 1953; Jespersen 1914—1929, 3; Lees 1960; Wasow and Roeper 1972; Thompson 1973; Schachter 1976). Like Subjects, the possessive and genitive can indeed designate various experiential roles: they can realize an agent-like participant (e.g., the clerk's taking the time to be friendly·, her visiting the man) or a Patient (e.g., his being dropped from the ticket). Depending on which participant is profiled, the atemporal clausal head which they are used with can be active as in (81) or passive as in (82): (81) I think the country's making a statement is a lot bigger than partisan politics. (CB) (82) Its volume alone would be impressive, but its originality and sheer brilliance vindicate completely her having been declared Doctor of the Church. (CB) Like the clausal Subject, moreover, the Subject in gerundive nominalizations can be in the objective case, as was shown above. Significantly, those gerundive nomináis that are factive in meaning but lack an explicit determiner all turn out to necessarily imply a Subject. So, while there is no explicit Subject in the nominalization having introduced us in he regrets having introduced us, the Subject is understood to be the Ae-person of the matrix clause.

Gerundive facts

243

The importance of the Subject to factive gerundive nominalizations is particularly clear in ambiguous nominalizations such as Do you consider [not having been to art school] a disadvantage? and [Having attended Harvard] can be helpful. I pointed out earlier that these nominalizations are either generic action gerundives or they are interpreted as factive nominalizations. Crucially, their interpretation as factives depends on the identification of a Subject: if, for instance, not having been to art school is addressed to someone who has been to art school himself and the you of the matrix clause cannot therefore be interpreted as instantiating the process type, and no Subject can be identified in the context, then the nominalization is generic rather than factive. If having attended Harvard is used after someone has told the speaker that he/she attended Harvard, the Subject of the nominalization is known and the nominalization is factive in meaning. Factive nominalizations, in other words, seem to necessarily have a Subject, which is either explicitly identified or it is implied but retrievable. It is obvious that the central role which the Subject plays in factive nominalizations primarily relates to the clausal level of organization in the nominalization: the Subject function is an essentially clausal function. In my discussion of atemporal clausal heads in Chapter 4 1 have identified the role of the Subject as being that of instantiator: by being tied to a Subject, an atemporal clausal head was argued to convert into the conception of an instance. The obligatory explicit or implicit inclusion of a Subject in factive gerundive nominalizations can thus be argued to instantiate the atemporal clausal head that forms their starting point. Factive gerundives, then, can choose between the nominal common noun or proper name strategy, but on the clausal level, they necessarily establish instantiation of the atemporalized process type which they contain. The integration of the clausal instantiation which factive gerundives realize and the nominal strategies which they use to ground the reclassified clausal head can be schematically represented as follows: Table 7. An overview of the integration of clausal instantiation and nominal grounding in factive gerundive nominalizations

NOMINAL instantiation, quantification, grounding definite common noun proper name

possessive/genitive /

CLAUSAL instantiation Subject realized Subject implied

244 Factive nominalizations as nominal constructions 3.4. Factive gerundive nominalization: Conclusion To conclude, in my analysis of factive gerundives, I have first identified the clausal level of assembly which they start from as being that of an atemporal clausal head. The fact that gerundive nomináis include clausally realized Objects, auxiliaries and adverbials rather than adjectives was thus given a natural explanation. It was then argued that, like the finite clause in thatfactives, the atemporal clausal head of gerundives is downranked and functionally reclassified into a unit of the nominal class. I focused on the nominal paradigms which the downranked atemporal clausal head in gerundive nominalizations comes to function in and described in detail which nominal grounding strategies the various types of gerundive nominalizations follow. I have also set up a typology of gerundive nominalizations which clarifies the position of factive gerundives in the overall system of gerundive nominalization. In a final part, then, I have shown that the restrictions that exist on the internal nominal behaviour of factive gerundives follow from the influence of their clausal layer of organization. I have pointed out that gerundive factives require a Subject that instantiates the reclassified clausal type specification, and this Subject is either explicitly realized in the form of a possessive or genitive, or it is implied and can be retrieved from the context. Factive gerundives, it can be concluded, successfully manage to integrate their clausal layer of instantiation (of an atemporalized process type) with the nominal grounding that is required for a noun to function in the speech event.

4. The internal organization of factive nominalizations: Conclusion The analysis of the internal functional properties of /Aai-factives, the fact ¿Ααί-constructions and gerundive factives provides clear evidence in favour of my claim that nominalization essentially involves the functional reclassification of a non-nominal into a nominal unit, whereby the nominalized unit not only takes over the external functions of nominal constituents in the clause but also adopts the latter's internal nominal behaviour: all three types of factive nominalizations start from a clausal level of assembly but realize nominal strategies of grounding. The clausal structure which factive nominalizations reclassify was shown to come to function either as nominal head or as a full nominal itself. Crucially, the precise nominal strategies followed by the various types of factive nominalizations were shown to be in accordance with the status of the clausal unit which they reclassify. That-

The internal organization offactive nomináis: Conclusion 245 factives incorporate uniquely grounded processual instances. As nomináis, they refer to these uniquely grounded instances of process types, by using the proper name strategy. 7%a/-factives are thus fully grounded nomináis. The appositive status of the fact //^/-constructions was argued to support the analysis of /Aa/-factives as proper names, while the the /àc/-nominal was shown to indicate the category or 'basic level' meaning of the thatstructure. Finally, gerundive factives are all derived from an atemporalized clausal head. As a clausal structure this clausal head is not grounded, but in gerundive factives it is always instantiated by a Subject. The Subject of gerundive factives may be either expressed explicitly by a possessive determiner, or by the genitive or objective case, or it may be implied and retrievable from the matrix clause or context. In the case of factive gerundives with possessive determiner or genitive, clausal and nominal instantiation were shown to be integrated seamlessly with each other. The description of the three types of factive nominalizations also illustrated the importance of taking a functional perspective to the internal organization of nominalized constructions. The structural analysis of thatfactives given in Halliday (1994), according to which embedded thatstructures are the heads of the nominal turned out to be inconsistent with the radically functional analysis of ^/-structures as proper names (see also Langacker 1991). 77ta/-factives thus also illustrate the need for a functional understanding of the concept of rank shift, rather than one based on structural constituency (see Chapter 2, Section 3.2). In the analysis of the fact /Aa/-structures as well, Halliday's head-postmodifier analysis was shown to be unsatisfactory and to overlook the peculiarities of the appositive relationship realized in them. Functionally, the paratactic relation which the fact /Aitf-structures establish is one of double grounding, with both the fact and the /Aa/-structure realizing a fully grounded and definite nominal (Langacker 1991). In terms of their representational semantics, the thatstructures in the fact //¡^-constructions were shown to function not as Qualifiers elaborating the type designated by the head noun, but as Things themselves. In my analysis of gerundive nominalizations, then, I pointed out that they follow either the common noun strategy, with the atemporal clausal head functioning as head of the nominal; or they realize the proper name strategy, with the atemporal clausal head functioning as fully grounded nominal. Factive gerundives were shown to be always grounded definitely. It was also shown that for the nominalization to have the potential of being factive in meaning, the atemporal clausal head has to be linked to the category of instantiation, i.e. to the Subject. In terms of their representational semantics, gerundive nominalizations turned out to be derivable

246 Factive nominalìzatìons as nominal constructions from both perfective and imperfective verbs and, if realized, their Subject could be mapped onto both agentive and non-agentive participant roles. Importantly, the analyses that were presented of the internal functional organization of factive nominalizations point to the need to consider not only prototypical or canonically coded nomináis but also more peripheral realizations of the nominal paradigm. I mean two things by this. First, when analyzing the nominal behaviour of nominalizations, it is essential that one does not only look for canonical types of nominal coding. The discussion of the internal nominal behaviour of nominalized constructions has often foundered because nominalizations were expected to behave as it were more 'typically nominally' than do ordinary, non-nominalized nomináis themselves. Gerundive factives preceded by a possessive or a genitive, for instance, have thus generally been considered to display some degree of nominal internal behaviour, and this because they realize the canonical type of nominal coding whereby the functions of type specification and grounding are realized iconically, by separate components. Bare gerundive factives and í/zaí-factives, in contrast, do not realize grounding iconically and have therefore hardly been recognized as being nominal-like internally as well. It is only by taking into account the existence of non-canonical nominal codings such as non-separate symbolization of a function that one can move beyond the existing descriptions of nominalizations and come to a coherent analysis of their external and internal functional properties. However, it is not only vital to consider non-canonical nominal codings in the analysis of nominalizations; it should also be acknowledged that nominalizations which realize the reclassification of a clausal structure will always remain peripheral members of the nominal paradigm, i.e. members which do not realize all of the properties that characterize prototypical nomináis. Nominalized constructions cannot be expected to behave exactly like non-nominalized nominal structures: their processual starting point necessarily influences much of their internal organization. To conclude from this that nominalizations like the ones which I discussed in this chapter are not nominal-like internally, however, is too drastic. Nominalizations can be said to realize the same schematic construction or configuration of functions as do ordinary nomináis. The functions invoked throughout the analysis of that-factives and factive gerundives are those of type specification, instantiation, quantification and grounding. What about the notion of factivity, then? What do the various types of factive nominalizations have in common semantically, and how can this be related to their lexicogrammatical properties? I believe that the obligatory presence - explicit or implicit - of a Subject in gerundive factives sheds a new light on the notion of factivity. Langacker (1991: 34) already noted

The internal organization of factive nomináis: Conclusion

247

that the "construal of the reified process as representing a particular instance renders it compatible with contexts that presuppose its factuality". While in Langacker's viewpoint, however, instantiation is not exclusively linked to the Subject function, the analysis of the internal, functional organization of gerundive nominalizations shows that clausal instantiation depends on the presence of a Subject. Factivity does not require finiteness, as is suggested by the tendency to focus on factive fAa/-structures only (e.g., Lees 1960; Kiparsky and Kiparsky 1971; Halliday 1994; Davidse 1994). Rather, it is about being conceived as an instance. My analysis of the internal functional properties of ίΑαί-factives, the fact íAútf-constructions and gerundive factives reveals that what they all have in common and thus seems to constitute the minimal condition for factive nominalization is the instantiation of the process type which they reclassify. While gerundive factives profile the instantiated nature of the nominalized process type (and consequently foreground the role of the Subject), /Aai-structures realize process types which are also grounded in the speech event. Yet, the internal organization of factive nominalizations cannot itself exhaustively explain how the meaning of factivity arises. Think, for instance, of the gerundive nominalizations which I described as 'instantiated actions'. Table 6 shows them to realize exactly the same options as do factive gerundives: they can follow the 'instantiation-of-a-type' model that is characteristic of common nouns (e.g., ... and that is [her having a stroke]), or they can follow the proper name strategy and function without determiners (e.g., Going there was fun). In spite of the fact that their internal organization does not seem to differ from that of factive gerundives, these nominalizations are not factive: rather than representing metaphenomena, they designate direct representations of experience. That these two subtypes of gerundive nominalization realize different systems of nominalization only becomes clear when we consider the constructions which they can be said to agnate with: factive gerundives agnate systematically with the fact thatconstructions and /Aai-factives; instantiated action nomináis belong to a paradigm which also contains generic action nominalizations designating a class. The lack of internal functional differences between factive gerundives and instantiated action gerundives illustrates that the analysis of factive nominalization cannot suffice with descriptions of the internal functional properties of its various subtypes. As already suggested in Kiparsky and Kiparsky (1971), factivity results from a combination of features and does not depend on the nominalized construction only: the context in which it is embedded is crucial as well. I hope to have shown, however, that the almost exclusive focus on the external functioning of factive constructions in the literature should be counterbalanced by careful descriptive analyses

248

Factive nominalizations as nominal constructions

of the nominalized constructions' internal functional organization. The description of the internal functioning of factive nominalizations does not only establish the unique position of each subtype of factive nominalization within the overall system, it also contributes to the identification of the schematic category of factivity and of the various lexicogrammatical realizations of it.

Chapter 10 Conclusion

In this study I have tried to show that it is in the dialectic between theory and description that lies the key to an analysis of nominalization which moves beyond existing descriptions and is able to shed a new light on specific nominalization types and on the process of nominalization in general. My study has been based on the assumption that an accurate description of specific nominalization patterns presupposes a coherent theoretical view on the language system and an insightful description of nominal and clausal categories. A significant part of it has therefore been devoted to the discussion of a number of basic theoretical notions and to the introduction of nominal and clausal functional categories. However, this study also makes clear that it is descriptive analyses of nominalized constructions which identify the theoretical constructs that are pertinent to the analysis of nominalization. It is also only through the description of nominalization patterns that we can come to a better understanding of the constructional mechanisms that lie behind the process of nominalization in general. The largest part of this book has therefore been devoted not to theory, but to the description of two specific systems of nominalization. The analysis of deverbal -er nominalization and of the various types of factive nominalizations has provided new insight into the general process of deverbal nominalization. Deverbal nominalization can now be defined as the functional reclassification of a specific functional level of assembly in the clause into a construction which does not only serve nominal functions in larger configurations but internally makes use of nominal means of instantiation, quantification and grounding. The analysis of nominalization requires the following major steps: firstly, the level of assembly which the nominalization derives from has to be identified; secondly, it has to be determined whether the reclassification that applies to it involves rank shift or not; thirdly, we have to identify the exact nominal strategy which the nominalized unit adopts; and, finally, the precise lexicogrammatical integration of the clausal starting point with the nominal end-status of the nominalized construction has to be considered: by bringing together their internal lexicogrammatical features and their external functional behaviour, we can describe the unique semantic import of specific nominalization systems. First, the various clausal levels of assembly which nominalization can apply to turn out to coincide with the main functional levels in the organi-

250 Conclusion zation of the clause (as identified in Chapter 4): nominalization can apply to a simple process type specification or verb stem (e.g., bake -» baker, sign -» the signing of the contract)·, it can start from a complex process type specification or atemporal clausal head (e.g., playing the piano -» [Her playing the piano] surprised us); nominalization can be based on a non-finite clause (e.g., John playing the piano —» [John playing the piano] surprised us), or it can apply to a finite clause (e.g., John plays the piano —> [That John plays the piano] surprises us). The clausal categories which deverbal nominalizations, unlike non-nominalized nomináis, imply (think, for instance, of the categories of modal and temporal grounding in deverbal -er nominalizations) are closely related to this clausal starting point of the nominalization process. Through nominalization, then, the various clausal levels of assembly are turned into nominal constructions, i.e. they are reclassified into the nominal class. While the nominal status of nominalizations derived from verb stems is unproblematic, the internal nominal character of nominalizations based on more elaborate units of clausal patterning is contentious. As my analysis of gerundive nominalizations and /Aa/-factives reveals, the internally reclassified nature of clause-like nominalizations shows up only when, firstly, their downranked status is acknowledged and, secondly, when non-iconic types of nominal coding are considered. The notion of downranking and, behind it, the idea that language consists of levels of functional organization which together make up a hierarchy or rank scale, is vital to the analysis of nominalizations which contain units of clausal patterning: such nominalizations establish reclassification by shifting a clausal configuration of functions - with its internal clause-like structure intact - from clausal to nominal rank. Through rank shift, a set of functions which normally figures in the clause thus comes to function at nominal rank and is fitted into the functional organization of the nominal. Generalizing, we can say that deverbal nominalization always involves reclassification: in some cases reclassification relates two units of the same rank (i.e. reclassification from verb to noun). In other cases, reclassification is realized through rank shift. To recognize the precise nominal paradigm in which rankshifted clausal units come to function, a view of nominal organization is needed which is radically functional and describes nominal composition in terms of component functions which are not necessarily realized by discernible component structures (see Chapter 2, Section 3.1). As the description of gerundive and iAaZ-nominalizations makes clear, nominalizations, just like 'ordinary' nomináis, do not necessarily realize the prototypical, iconic form of coding that is characteristic of common nouns. An analysis of nominalizations which tries to force them into what is felt to be the nominal prototype (i.e.

Conclusion 251 the common noun strategy) is therefore bound to fail and will not be able to reveal the internal nominal status of the majority of nominalizations that involve downranking. The descriptive breakthrough needed to account for both iconically and non-iconically coded nomináis and, consequently, to identify the internal nominal properties of downranked clausal structures, was shown to be provided by Langacker's functional analysis of the nominal in terms of type specification, instantiation, quantification and grounding. I have then shown that, by systematically analyzing the internal organization of nominalizations from this functional perspective, their classification as nominal can be based not only on their external functioning, but also on their internal functional behaviour. The precise nominal paradigm which reclassified units fit into varies according to the clausal level of assembly which is being nominalized. If a verb stem or simple process type is nominalized, it is necessarily turned into a common noun or nominal type specification (see Langacker 1991), and this irrespective of the categories of instantiation and grounding that may be implied in the relationship that is established in the resulting nominalization (think, for instance, of the analogy between the semantics of -er nouns and the Subject-Finite unit). An atemporal clausal head is either turned into a common noun (which is then instantiated and grounded by means of determination) or, when instantiation is implied, it functions as a full nominal or proper name. If the atemporalized process type specification is construed as a unique instance, but then in type space, it follows the proper name strategy of generic mass nouns. Likewise, a downranked finite clause is reclassified and comes to function as a proper name, as also pointed out by Langacker (1991). The same can, in my opinion, be argued for nominalizations which apply to the fourth level of assembly which I have identified, viz. that of ungrounded instances or non-finite clauses with their Subject in the objective case (e.g., John playing the piano —» [John playing the piano] surprised us). Importantly, the functional characteristics of the clausal level of assembly are thus integrated with those of the nominal construction in which the reclassified unit comes to function. Downranked clausal units adopt in some cases the common noun strategy (i.e. in gerundive nomináis of the type his opening the door and that burning the midnight oil of yours), but far more often, they function as full nomináis or proper names. When nominalization involves rank shift, in other words, the default nominal strategy seems to be that of proper names, which provide a unique name for an instantiated process or for a unique process type (in generic gerundives). In gerundive nomináis, it is only when the Subject-entity instantiating the

252 Conclusion downranked process is foregrounded, that the common noun strategy applies and the Subject is explicitly realized as determiner (as in his opening the door) or as postmodifier (as in any talking loudly of yours). In short, rather than being inexplicable combinations of nominal-like external behaviour and clause-like internal categories, nominalizations are basically nominal constructions, externally as well as internally. A full description of a nominalized construction, therefore, requires an analysis of its external, nominal functioning, of the clause-like categories which it still contains, as well as of the way in which the nominalized construction fits these clause-like categories into the nominal paradigm. It is only when the latter, third level of analysis is included in the description of nominalizations that their internal nominal status can be revealed and the lexicogrammatical integration of clausal and nominal categories can be identified. It is moreover in the integration of clausal and nominal categories which nominalizations realize that important clues to their semantics lie. In my analysis of -er nominalization, I have, for instance, argued that the semantics of the noun type specification resulting from -er derivation is basically analogous to that of the clausal Subject. The integration which -er nominalizations establish between a process type and an entity was shown to be strikingly similar to the relationship which in the clause is realized by the combination of Subject and Finite (and the categories of temporal and modal grounding realized in it). In my analysis of gerundive nominalizations, then again, I showed that for gerundives to be able to receive a /active reading in the clause, they have to integrate the clausal conception of an instantiated process with a definitely grounded nominal structure. I argued that gerundives can realize this integration in two ways: either they follow the common noun strategy and use a determiner which does not only ground the nominal but also instantiates the downranked process type specification; or they construe the atemporal process type as a full (definite) nominal whose instantiation of the downranked process type by a Subject is contextually realized. Like factive ^/-clauses, factive gerundives were shown to necessarily designate instances rather than types. Like them, they were shown to nominalize these instances as definitely grounded nomináis. The factive reading, I concluded, is not necessarily linked to finite structures, but to instantiated processes which function as definite nomináis in factive clausal contexts. To summarize, the approach to nominalization which has been presented in this study breaks away from tradition in several respects. It shows that reclassification is both external and internal: nominalized structures that function as nomináis in larger configurations are reclassified also internally and take on the nominal behaviour of particular nominal strategies.

Conclusion 253 My approach thus complements the prevailing focus on the clausal categories that figure in nominalizations and considers nominalizations first and foremost as nominal constructions which integrate the clausal and nominal categories which they contain into one nominalized structure. Importantly, my analysis of nominalization also moves away from the descriptive fragmentation that characterizes the literature on nominalization and offers a unified account of 'morphological' types of nominalization (such as -er derivation) and nominalizations which involve more elaborate, clause-like structures. Finally, the analysis of deverbal -er nominalizations and of factive nominalization shows it is possible to link up the lexicogrammatical properties of nominalizations with the meanings which they realize. The study of nominalization patterns that has been presented here and the overall framework which I have proposed for the analysis of nominalized constructions is based on nominalization systems that exemplify the various clausal levels of assembly which nominalization can start from. Many nominalization types have, however, been left undiscussed. Undoubtedly, a detailed analysis of their lexicogrammatical and semantic properties will modify and refine the present model. Among the nominalization types that might be particularly interesting are nominalizations derived by -ee suffixation and derived nomináis of the type the destruction of the city. Action nominalizations such as the signing of the contract should also be considered more closely. Unlike gerundive nominalizations, action nominalizations and derived nomináis necessarily realize their participants by means of nominal elements (i.e. nominal pre- and postmodifiers) and determination in them is not restricted to the possessive type, but it includes the article as well. As to -ee derivation, the increased productivity of agentive -ee nomináis (e.g., escapee, attendee) has to be accounted for and the resemblances and differences with -er derivation have to be identified. Also, by elaborating especially on the internal structure of nominalizations, I have chosen to focus on only part of the story: the analysis which I propose for the internal organization of nominalizations has to be supplemented with a systematic overview of the external contexts in which nominalizations can occur and of the broader, textual motives that inform their occurrence.

Appendix Non-agentive -er nominalizations

baker bestseller boiler broiler cooker diner dipper dunker front-loader fryer gusher jotter keeper killers kneeler knocker

loaner low-loader poker reader roaster scratcher sipper sleeper sleeper slipper(s) squeezer squirter steamer stepper

'a food (meat, fruit or vegetable) that is suitable for baking' 'a book or other item that sells well' 'a chicken suitable for boiling' 'a young chicken suitable for broiling' 'an apple for cooking* 'aplace where you can dine' 'vegetable/fruit/other snack that has to be dipped before being eaten' 'a doughnut that is preferably dunked before being eaten' 'an appliance, such as a washing machine or a clothes drier which allows for material to be inserted at the front' 'a young chicken or rabbit suitable for frying' 'a flowing oilwell' 'a small notebook' 'something that should be kept' 'cattle suitable for killing' 'a kind of chair, which you have to kneel on' 'a piece of metal on the front door of a house with which you can knock on the door, to attract the attention of the people inside* 'something that is loaned to someone' 'a vehicle in which the carrying platform is kept low for ease in loading' 'a metal bar to move coal or wood in a fire' 'a (course)book consisting of a compilation of literature ' 'a kind of meat suitable for roasting' 'a lottery ticket that one has to scratch to reveal the winning patterns' 'a drink that one has to sip' 'a train you can sleep in' 'a bed or sofa in a train, in which you can sleep' 'a kind of shoes that can be slipped into' 'a kind of container that you have to squeeze in order to extract something from it' 'a container that ejects liquid in a jet from a narrow opening' 'an edible clam' 'a raised surface on which you have to put your foot, as a way of working out'

256

Appendix

stroller top-loader viewer walker

'a small chair on wheels, in which a baby or small child can sit and be wheeled around' 'a washing machine or clothes drier which can be loaded from the top' 'a device for viewing photographic transparencies' 'a special kind of frame which is designed to help babies or disabled or ill people to walk'

Notes

1.

2.

3. 4.

5.

6.

7.

8.

It can be pointed out here that, as McGregor (1997: 2) remarks, in claiming that grammar is a semiotic or symbolic system, "there is no implication that every identifiable pattern in the organization of a language is necessarily semiotic". It is, for instance, possible that the fact that a particular morpheme always occurs in final position does not have any semiotic significance. The claim that language is semiotic should rather be taken to mean that "a certain non-trivial subset of the organizing patterns are semiotically significant" (McGregor 1997: 2). Note that Halliday's interpretation of Hjelmslev (1961) is to a large extent based on his own theoretical views on grammar: Hjelmslev's Prolegomena to a Theory of Language itself hardly refers to grammar. For this section, I have benefitted from discussion with Kristin Davidse. Notice that, strictly speaking, even a nominal like those three black cats is not fully canonical because the expression of quantity which it contains extends beyond the (well-defined) constituent three to the demonstrative pronoun (those instead of, for instance, this) and the head noun (cats instead of cat) (Langacker 1991: 54). In Halliday (1979: 58), it is argued that the organization of a linguistic unit may be considered from three different perspectives: at its own level ("its relation to other elements identified at the same level as itself'), from above, i.e. in terms of "its relation to elements at the next (or some) higher level" and from below, or in "its relation to elements at the next (or some) lower level". Halliday thus considers a third level of analysis of linguistic units to consist of the syntagmatic relationships which a unit holds to other units at the same (rank) level. I include syntagmatic relations under the 'external' functional relations of a unit and define the third level differently, i.e. as the level of the unit itself, where all external and internal relations come together. The metaphor which Halliday uses to describe the rank scale is, in fact, reminiscent of the 'building-block' metaphor (Langacker 1987a, 1999): the various levels of the rank scale are claimed to appear as it were "one inside the other, in a sort of one-dimensional Chinese box arrangement" (Halliday 1961: 250). Note that in Langacker's model, grammatical classes are, like constructions, conceived as symbolic units or as bipolar linguistic symbols: the class of nouns, for instance, "is defined by a schema that we can represent as [[THING]/[...]], where [THING] is a schematic semantic unit (...) and [...] a schematic phonological unit" (Langacker 1987a: 81). Nomináis with a common noun as head are thus considered to be more prototypical than proper names. Van Ltingendonek (1999) is of the opinion that it is not common nouns which are prototypical, but proper names, and this for

258

Notes

several reasons: because of their primary referring or identifying function, because the meanings which proper names have are of a presuppositional nature and therefore more easily processed, and, finally, because proper names display the unmarked nominal feature of definiteness and are prototypically singular, countable, non-generic and concrete. Van Langendonck argues that it is proper names rather than common nouns which receive a 'truly iconic motivation', namely as nouns which mostly occur without an article (Van Langendonck p.c.). (Note that the characterization of prototypical nouns as serving the pragmatic function of reference is due to Croft (2000), who argues that cross-linguistically, certain unmarked or prototypical combinations of pragmatic function and class can be observed: a noun, for instance, has the primary function of 'reference to an object', verbs have the function of 'predication of an action', and adjectives that of 'modification by a property'. It seems to me, however, that Croft's pragmatic/ communicative functions are not meant to point to prototypical instances within a class, but rather characterize classes in general, as opposed to other classes. As such, they form a highly schematic type of characterization and are valid for all members of the class. Like Langacker's description of nomináis as 'things or regions in some domain', the characterization of nomináis as 'referring to an object' therefore does not distinguish between nouns and nomináis. The schematic function of 'reference' as used by Croft (2000) should be distinguished from that of 'grounding', which applies to the link which nomináis - but also clauses - establish with the speech event. It is with respect to this link, or more specifically, with respect to the structural implementation of it, that common nouns and proper names can be said to differ: unlike common nouns, proper names do not need individually symbolized grounding predications tagged onto them in the nominal to be grounded in the speech event; they are inherently grounded.) 9. I will in Chapter 4 present a more elaborate analysis of the Subject. Note that the analysis of the Subject that is given here is comparable to that presented in Langacker (1991: 305-321), where prototypical Subjects are argued to rank highly with respect to the so-called "topicality factors" (based on Givon 1976, 1978, as cited in Langacker 1991: 306): they are agentive, human, definite and they constitute "the figure within the profiled relationship" (Langacker 1991: 308). However, Langacker also argues that a general characterization of the Subject cannot be based on the notion of topic, but "must be subjective in nature (i.e. a matter of construal rather than conceptual content" (Langacker 1991: 312). The definition which he proposes is that of 'relational figure'. 10. In Dik (1980), a distinction is made between 'semantic', 'syntactic' and 'pragmatic' functions. Semantic functions are functions such as Agent, Goal, Recipient; the Subject and Object are regarded as syntactic, and, finally, functions such as Theme and Topic are considered to be pragmatic. As Dik himself points out, however, the terms are somewhat unfortunate and chosen "mainly for traditional reasons" (Dik 1980: 13). The so-called 'syntactic'

Notes

11. 12.

13.

14.

259

functions of Subject and Object, for instance, rather than being treated as semantically 'empty', are given a semantic interpretation. Like Halliday, moreover, McGregor (1997) relates the logical metafiinction to the syntagmatic relation of dependency. Importantly, in Cognitive Grammar, as in Systemic-Functional Grammar, it is posited that there is "no fundamental distinction between morphological and syntactic constructions, which are fully parallel in all immediately relevant aspects" (Langacker 1987a: 82; see also Halliday 1961). Both morphological and syntactic structures realize meaning in a natural or non-arbitrary way. A similar distinction can be observed in the system of -ing nominalization: the nominal opening in (1) is lexicalized, whereas opening in (2) forms an ad hoc formation based on a schematic unit: (1) He squeezed through a narrow opening in the fence. (CB) (2) It entails a kind of patient regard, an opening of oneself toward the subject.... (CB) However elucidating the analysis of a construction's order of assembly may be, it does not (always) tell the whole story: as pointed out by Langacker (1987a, 1991, 1998, 1999), composite structures tend to exhibit partial rather than full compositionality and, secondly, the contribution of their component parts is not always recognized, i.e. they are not always fully analyzable. First, as far as compositionality is concerned, the fact that components can be recognized within an assembly of symbolic units "does not entail that these components exhaust its characterization", i.e. the composite whole "may have properties above and beyond those of its components, which may in turn be manifested in it only imperfectly" (Langacker 1987a: 87). The composite structure is then either more specific than what can be computed from its components, or it forms an extension of the expected compositional value (Langacker 1999: 153). Nominalizations - particularly nominalizations at the level of the word - are constant reminders of the partial nature of the compositionality of constructions. A structure like ruler, for instance, is more frequently understood as a device for measuring things than as a device used for ruling lines and its meaning thus significantly extends the expected compositional value; similarly, a computer is much more than just 'something that computes' (Langacker 1999: 128). In short, any assembly of symbolic structures must be treated as a separate entity in its own right and "with many possible sources of extracompositionality" (Langacker 1998: 4), such as discourse and context, particular domains of knowledge, mental spaces, metaphor and metonymy. The discrepancy between an expression's expected compositional value and the actual meaning that becomes its conventional value facilitates a decline in the analyzability of the construction (Langacker 1999: 128): the language user becomes less cognizant of the components of the assembly or no longer co-activates the composite structure and its components (Langacker 1987a: 462). Analyzability is a matter of degree: while in a nominalization like teacher, the components may still be perceived, this is less true for, for instance, nomináis like propeller and drawer. Many of the -er nomi-

260

15.

16.

17.

18.

19.

20.

Notes nalizations that designate items of clothing likewise seem to have lost much of their analyzability, to the extent even that nomináis like jumper, sneakers and loafers no longer seem to be recognized as being derived from the processes ofjumping, sneaking and loafing. The systemic-functional view of nominalizations is closely tied up with the concept of 'grammatical metaphor': nominalization is presented as a major resource for the creation of metaphorical rather than 'typical' or congruent lexicogrammatical realizations of semantic categories. A nominalization such as the cast's brilliant acting, for instance, is thus viewed as the metaphorical counterpart of the clause the cast acted brilliantly (Halliday and Matthiessen 1999: 229). To describe the relationship between the congruent and the metaphorical realizations, Gleason's concept of agnation is used and it is claimed that every metaphorical structure has "an agnate form corresponding to its congruent realization" (Ravelli 1988: 141). Agnates that help to distinguish gerundive nominalizations from action nominalizations are, for instance, non-finite modifiers (e.g., Any boy [cleaning the kitchen] will be rewardedfor it) and non-finite clauses (e.g., With [Tom cleaning the kitchen], we can be sure that it will be very clean). Both structures agnate systematically with gerundive nominalizations but not with action nomináis: like gerundive nomináis, for instance, they allow for imperfective processes (conpare, for instance, any boy [being prepared to help] and with [Tom being prepared to help] with the gerundive nominal [Tom's being prepared to help] surprised me and with the action nominal * Tom's being of prepared; *Tom's being of a good husband. Plurals and non-plural mass nouns are grouped together by Langacker (1991) because they behave alike in numerous respects. They can, for instance, both occur as full nomináis without quantifier or overt grounding predication (e.g., They're looking for diamonds/ gold); they do not tolerate the indefinite article (e.g., *a diamonds/ gold); and both can be used with a number of quantifiers not permitted with count nouns (e.g., most diamonds/ gold/ *diamond; a lot of diamonds/ gold/ *diamond) (see Langacker 1991: 77). Note that for a plural, "the size of the instance will be given by the number of its component entities", but collectively, these entities "constitute just one instance of the plural-noun type" (Langacker 1991: 81). As pointed out by Langacker (1991: 54), even in those three black cats, the different structural levels are not realized entirely separate from each other: number is indicated not only by the quantifier three, but also by the grounding predication those and by the head noun. Only when the clausal head is progressive in profile, is -ing impossible: the auxiliary pair be playing can thus not be turned into an -ing form (*being playing). Langacker (1991: 232) attributes this to the fact that "simpler alternatives are available that do essentially the same job, i.e. they provide effectively equivalent information without creating potentially awkward ambiguities". The -ing form playing is a simpler alternative for being playing and therefore makes the latter redundant.

Notes

261

21. Important evidence in favour of the type status of atemporal clausal heads comes from gerundive nominalizations and the way in which they function in the structure of the nominal. I will not, however, go into their properties here and refer to Chapter 9 for a discussion of the status of gerundive nominalizations as atemporalized process type specifications. 22. While both Halliday (1994) and Langacker (1991) agree on the grounding function of the finite element in the clause, they differ as to where exactly they draw the boundary between the grounding and the non-grounding part of the verb group: Langacker considers only the actual markers of tense and modality as grounding; in Halliday's analysis, auxiliaries like did/does/was/ had/have are considered to be part of the finite or grounding element, and this primarily because they are picked up in tags, such as didn't you?, has he?, wasn't she?. In Langacker's model, the verb group has responded can thus be divided into -s (which grounds the verb group) and have responded (which constitutes the non-grounding, head-part of the verb group). In Halliday (1994), that same verb group is analyzed as has (grounding) and responded (non-grounding). When no auxiliaries are present, Halliday's analysis of the verb group coincides with that in Langacker (1991): a verb group such as responded is divided into -ed and respond. 23. In Davidse ( 1997) and Heyvaert (2000), the view is defended that the clause is instantiated (as well as grounded) by the Subject-Finite complex, rather than by the Subject only. Especially on the basis of my analysis of gerundive nominalizations (presented in Chapter 9), I have come to an analysis which is slightly different in that it considers only the Subject to fulfill an instantiating role. I view the function of the finite or grounding element of the verb group as being that of grounding the instance in the speech event (which, of course, presupposes instantiation, but crucially does not foreground it). As argued by Davidse (1997), the Subject then co-operates in the grounding by establishing objective or explicit person deixis (see Section 2.3.2). The Subject thus serves two distinct, though related functions: it instantiates the process type specification of the clause and, if the clausal head is finite, it co-operates in the grounding. 24. The notions of 'subjective' and 'objective' as they are used here in the context of modality should be distinguished from the use which Langacker (1985, 1990b) makes of them: in Langacker's work they do not refer to whether or not a linguistic element is related to the speaker, but they relate to how explicitly the speaker is present in an utterance. The more explicit the speaker figures in the utterance, the more 'objective' its status is said to be. As Verstraete (2001: 1513) points out, the notions of subjectivity and objectivity used by Langacker are not entirely unrelated to the notions used in the classification of modality: "they can be regarded as an additional dimension for the subjective pole, depending on how explicitly the speaker-relatedness figures in the utterance". 25. The subjective deontic modality of obligation/permission is expressed by the prototypical uses of deontic auxiliaries such as must, should, shall, may.

262

Notes

26. The objective deontic modality of obligation and permission is typically realized by non-Speaker-Hearer-oriented deontic modal auxiliaries, such as have to, ought to, can. 27. Verstraete (2002: 65) observes that, when the objective modal auxiliary is a periphrastic modal such as have to or be able to, the schematic, epistemic modality that accompanies objective modals can surface in the form of an epistemic modal auxiliary, as in He may have to leave his house; She might be able to come over after all. 28. Note that the notions of 'subjective' and 'objective' are used here in a meaning that is related to the way in which they are used by Langacker (1985, 1990b): the Subject is said to establish person deixis objectively because it puts the person deixis 'on stage' as a maximally salient entity; the person deixis that is realized in the finite element, on the other hand, is subjective, or minimally salient: it remains implicit (based on Langacker 1991: 93). 29. In Langacker (1991: 247), a similar claim is made with respect to Subjectverb agreement: "what is traditionally known as 'subject-verb agreement' is analyzed as being part of the grounding predication". 30. An extensive list of English non-agentive -er nominalizations has been included in the appendix. 31. The distinction between lexicalized and 'schematic' or 'ad hoc' nominalizations was introduced in Part I, Chapter 3, Section 1.2. 32. This view on English transitivity, in which a fundamental split is posited between transitive and ergative clause construals is proposed in Davidse (1992b, 1998b) and Davidse and Geyskens (1998). Davidse's view is rooted in the British structural-functional tradition of Halliday (1967, 1968), Lyons (1969) and Cruse (1972, 1973, as mentioned in Davidse and Geyskens 1998). It should be distinguished from Perlmutter's (1978) distinction between 'unergative' and 'unaccusative' verbs: broadly speaking, one can say that Perlmutter's 'unaccusative' category includes all the ergative verbs of the structuralfunctional tradition, as well as intransitive verbs such as fall, die and stumble (for a detailed comparison of the two traditions, see Davidse and Geyskens 1998). 33. Following Levin and Rappaport (1988), Lemmens claims that Goal-profiling -er nomináis are related to the middle use of the transitive base verb: "as with the middle construction the focus is on the properties of the Affected" (Lemmens 1998: 137). Lemmens's account of the relation between Goal-profiling -er nomináis and middle formation will be discussed in Chapter 7, Section 2. 34. Interestingly, a category which is not included in Lemmens's account (it was not accounted for in Ryder's approach either) is that of -er nomináis that profile so-called 'oblique participants' such as jotter, stepper, kneeler, viewer, stroller (I will explain the notion of 'oblique participant' (based on Laffut 2000) in Section 1.2). Together with the Goal- and Medium-profiling group of -er nomináis, they make up the non-agentive type of -er nominalization. I will go more deeply into this type of -er nominalization in Section 1.2.1.

Notes

263

35. In fact, it has been pointed out by Laffut (2000: 145) that the Locatum is the Patient or primary affected of the process, but "because of the involvement in the locative process, the Location can be analyzed as a second affected", both when it is realized prepositionally and when it functions as a bare noun phrase. 36. Deverbal nominalization types which have been analyzed in the literature as offering an active and a passive voice option are those of -ion and -ment suffixation, as in: (1) a. the critics' hostile reception of the play (active, 'they received the play') b. the play's hostile reception by the critics (passive, 'the play was received') (Quirk et al. 1985: 1289) (2) a. the committee's appointment of John (active, 'they appointed him') b. John's appointment by the committee (passive, 'he was appointed') (Lees 1968: 68) 37. Many aspects of the description of the middle proposed in this chapter are also discussed in Heyvaert (1997), Heyvaert (1998) and Davidse and Heyvaert (in press). 38. Structures like these have been variously referred to as activo-passives (Jespersen 1914-1929, 3: 347), medio-passives (e.g., Declerck 1991b: 203), process-oriented passives (Halliday 1967), facility-oriented passives (Fawcett 1980: 148), patient-subject constructions (Van Oosten 1977: 459) and pseudointransitives (Lyons 1969: 366; Smith 1978: 103). I will use the term 'middle construction' (as used by, among others, Keyser and Roeper 1984; Fellbaum 1986; Hale and Keyser 1987; Fagan 1992; Kemmer 1993). 39. Hence labels like activo-passive, medio-passive and patient-subject construction. 40. Even if The door closed slowly is analyzed as a middle construction, it is turned into an ergative one-participant construction as soon as the phrase all by itself is added to it: The door closed slowly all by itself czanot but be interpreted as an ergative one-participant construal. 41. Note that the necessarily unexpressed nature of the Agent in middle constructions has inspired some authors to call them pseudo-intransitives (Lyons 1969; Smith 1978). 42. For the notion of 'representational' or 'ideational' fonctions, see Chapter 2, Section 4. 43. The intransitive type of middle is, in fact, well established in Dutch and has received a fair amount of attention in the literature on Dutch middle constructions (e.g., De Vries 1910; Holierhoek 1980; Van den Toorn 1982; Hoekstra and Roberts 1993; Ackema and Schoorlemmer 1994; Peeters 1999). 44. The IDL corpus is the 38 million word corpus of the Instituut voor Nederlandse Lexicografie 'Institute for Dutch Lexicography' in Leiden. 45. Intransitive middles in Dutch agnate with impersonal constructions in which the oblique participant is realized as a prepositional phrase:

264

46.

47.

48.

49.

50.

51.

52.

Notes (1) Het fietst prettiger op asfalt dan op grind. it cycles more pleasantly on asphalt than on gravel 'it is more pleasant to cycle on asphalt than gravel' (2) Het zit lekkerder in sommige bomen. it sits more comfortably in some trees 'it is more comfortable to sit in certain trees' (3) Het loopt lastig in nieuwe schoenen. it walks difficult in new shoes 'it is difficult to walk in new shoes' Ehrman (1966: 74) points out that passive clauses with can often highlight that "there are certain positive qualities of the subject such that the way is cleared for the predication", as in These engines can be removed from a boat with relative ease (Ehrman 1966: 14). Passives like these resemble middles in that they focus on the properties of the non-agentive Subject entity. The fact that they make use of a passive verb, however, constitutes an important difference with middle constructions. Middles have often been claimed to be necessarily generic (e.g., Keyser and Roeper 1984; Roberts 1986; Fagan 1992; Masam 1992; Sohn 1998). As also pointed out in Rosta (1995) and Iwata (1999), this is not true: middle constructions can just as well designate specific events. An example of a nongeneric middle construction is The lap portion of the car seat belt slotted into place easily (CB). Observe that will can also be used in middles to make "a prediction or a generalization about someone's behaviour" (Declerck 1991b: 413-414), as in Her face is her best asset because she has very good bone structure which will photograph well (CB). These objections were raised in discussions of Dutch -er nominalization in Van Langendonck 1988 and Booij 1986 (discussed in Chapter 5). They can, however, be argued to apply to English -er suffixation as well. Notice that, unlike Ryder (1991, 1999b) and Lemmens (1998), I categorize nomináis related to processes of cooking (broiler, boiler, baker, cooker) with the non-agentive group of -er nominalizations. I have two reasons for that. Firstly, even though these nomináis profile participants which co-participate in the process, they clearly also have a (human) Agent in their base which does the broiling and cooking. Secondly, as I will show in Section 2 of this chapter, the semantics of -er nomináis of cooking comes closer to that of nonagentive, middle-related -er nomináis than to that of agentive nomináis (as also pointed out in Levin and Rappaport 1988). A comparable reading is plausible for the nominal diner, which may have been construed analogously to sleeper, in the Oxford English Dictionary, diner is defined as 'a railway dining car' and the attested examples which are given of it are all drawn from the railway context, e.g., A new dining car which ...is the first diner... built by that company. As I pointed out in Chapter 3, many of the -er nominalizations designating items of clothing are characterized by a certain loss of analyzability (Lan-

Notes

53.

54. 55.

56.

57.

58.

265

gacker 1987a, 1999) and are no longer recognized as -er derivations: are nomináis such as jumper, loafers and sneakers still analyzable as deriving from the processes ofjumping, loafing and sneaking? Or have they lost their verbal meaning and evolved into simple nouns? The idea that food is raised to be processed in a particular way is clearly behind the following corpus example: ... can you supply me with any information about this apple? Is it an eater or cooker and how long does it keep? (CB). The nominal singer can of course also have its lexicalized meaning here and designate a professional singer. I follow Noonan's (1985) definition of (clausal) complementation here, which is in line with the literature on factive complementation in general (e.g., Rosenbaum 1967; Kiparsky and Kiparsky 1971; Stockwell, Schachter and Partee 1973; Menzel 1975): "By complementation we mean the syntactic situation that arises when a notional sentence or predication is an argument of a predicate. For our purposes, a predication can be viewed as an argument of a predicate if it functions as the subject or object of that predicate" (Noonan 1985: 42). Clausal complements differ in this respect from non-clausal or nominal complements, which are prototypically conceived of as referring to Objects, rather than to both Objects and Subjects (e.g., Radford 1997: 498, 519; Haegeman and Guéron 1999: 21, 124). As Delacruz (1976: 180) puts it, "Possibly whenever the use of a predicate in a sentence involves a presupposition one could say that the sentence as a whole has or carries a presupposition. And if a sentence carries a presupposition, one might say that whoever asserts it presupposes whatever is presupposed by the sentence". As pointed out in Davidse (1994: 280-281), there is a special sense associated with fronted locutions. The sentence That he had been wrong was definitely said by him, for instance, has a semantic value which differs from that of He said that he had been wrong: "On the one hand, ... it specifies the content of the Sayer's speech act. On the other hand, the embedded status of the locution entails speaker commitment to the 'fact' that this locution was produced by the Sayer" (Davidse 1994: 281). Also structurally speaking, fronted locutions come close to facts: they can function as Subjects of the passive; they cannot be substituted, but only be referred to (That was definitely said by him). They can function as second nominal in appositive constructions (The words: 'Iwas wrong' were definitely said by him, embedded locution: quote). Davidse therefore proposes to consider the fronted type of locutions as embedded constituents, rather than as tactically related clauses. As correctly pointed out by Wierzbicka (1988: 49), this is "a contradiction in terms", which "undermines the whole idea that semantic factors (such as factivity) 'account for the complement system of English, and other languages as well'".

266

Notes

59. This has inspired Noonan (1985: 60) to coin the term 'nominalized complements' for them, to be distinguished from fAai-structures and infinitivals which are termed 'complements'. 60. As I will point out in Section 2.2, not all clausal contexts allow for replacement of a the fact ίΑαί-clause by a that-clause. This has primarily to do with the higher degree of 'nominality' of the fact /Aai-clauses: they can be fitted into more clausal contexts than can /AaZ-clauses. In Subject position, however, the fact that- structures are typically replaceable by í/taí-structures. 61. Structures can be ambiguous between a restrictive relative reading and an appositive, factive one, witness the nominal the fact that she forgot in The fact that she forgot is unimportant (Langacker 1991: 432). On the relative clause interpretation, the fact is interpreted as a participant in the relative clause: she forgot the fact. In its factive reading, in contrast, that she forgot is what constitutes the fact. 62. The possessive pronouns in gerundive nominalizations can also be turned into objective case (see, among others, Sweet 1898; Poutsma 1929; Lees 1960; Declerck 1991b): e.g., We have always been a close couple and shared everything, so I never resented [him leaning on me when he was depressed]. (CB); I do not like [him coming here so ofien], (Sweet 1898: 120).

References

Databases: COBUILD corpus, Birmingham INL corpus, Leiden Ackema, Peter, and Maaike Schoorlemmer 1994 The middle construction and the syntax-semantics interface. Lingua 93: 59-90. Bauer, Laurie 1983 English Word-Formation. Cambridge: Cambridge University Press. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 Longman Grammar of Spoken and Written English. London: Longman. Booij, Geert 1986 Form and meaning in morphology: the case of Dutch 'agent nouns'. Linguistics 24: 503-517. 1992 Morphology, semantics and argument structure. In Thematic Structure. Its Role in Grammar, I. M. Roca (ed.), 47-67. Berlin: Foris Publications. Boon, René 1972 -£r-afleiding in modern Nederlands. Synchronische analyse en inleidend onderzoek [-Er derivation in modern Dutch. A synchronic analysis and introductory investigation]. M.A. thesis, Department of Linguistics, University of Leuven. Carlson, Greg 1978 Reference to Kinds in English. Bloomington: Indiana University Linguistics Club. Chomsky, Noam 1970 Remarks on Nominalization. In Readings in English Transformational Grammar, R. A. Jacobs, and P. S. Rosenbaum (eds.), 184— 221. Waltham, Massachusetts: Ginn. Comrie, Bernard 1976 The syntax of action nomináis: a cross-language study. Lingua 40: 177-201. Comrie, Bernard, and Sandra A. Thompson 1985 Lexical nominalization. In Language Typology and Syntactic Description III, T. Shopen (ed.), 349-407. Cambridge: Cambridge University Press.

268

References

Croft, William 2000

Parts of speech as language universale and as language-particular categories. In Approaches to the Typology of Word Classes, P. M. Vogel, and B. Comrie (eds.), 65-102. Berlin/New York: Mouton de Gruyter. Dalton-Puffer, Christiane 1994 Are Shakespeare's agent nouns different from Chaucer's? - On the dynamics of a derivational sub-system. In Studies in Early Modern English, D. Kastovsky (ed.), 45-58. Berlin/New York: Mouton de Gruyter. Davidse, Kristin 1991 Categories of experiential grammar. Ph. D. diss., Department of Linguistics, University of Leuven. 1992a On the real function of Subject. Paper delivered at the 19th International Systemic-Functional Congress, Sydney, 13-18 July 1992. 1992b Transitivity/ergativity: the Janus-headed grammar of actions and events. In Advances in Systemic Linguistics: Recent Theory and Practice, M. Davies, and L. Ravelli (eds.), 105-135. London: Pinter Publishers. 1994 Fact projection. In Perspectives on English. Studies in Honour of Professor Emma Vorlat, K. P. Carlon, K. Davidse, and B. Rudzka-Ostyn (eds.), 259-286. Leuven: Peeters. 1996 Turning Grammar on itself: Identifying Clauses in Linguistic Discourse. In Meaning and Form, Systemic Functional Interpretations, Meaning and Choice in Language: Studies for Michael Halliday, M. Berry, C. Butler, R. Fawcett, and G. Huang (eds.), 367-393. Norwood, New Jersey: Ablex Publishing Corporation. 1997 The Subject-Object versus the Agent-Patient asymmetry. Leuven Contributions in Linguistics and Philology. 86 (4): 413-431. 1998a Agnates, verb classes and the meaning of construals. The case of ditransitivity in English. Leuven Contributions in Linguistics and Philology. 87 ( 3 ^ ) : 281-313. 1998b On transitivity and ergativity in English, or on the need for dialogue between schools. In English as a Human Language: to Honour Louis Goossens, J. Van der Auwera, F. Durieux, and L. Lejeune (eds.), 95-108. Muenchen: Lincom. 1999 The semantics of cardinal versus enumerative existential constructions. Cognitive Linguistics 10 (3): 203-250. 2000a Semiotic and possessive models in relational clauses: thinking with grammar about grammar. Revista Canaria de Estudios Ingleses 40: 13-35. 2000b A constructional approach to clefts. Linguistics 38 (6): 1101— 1131.

References 2001

269

Postdeterminers: their secondary identifying and quantifying functions. Preprint nr. 177, Department of Linguistics, University of Leuven. 2003 A corpus check of the factive presupposition. In Configurations of Culture. Essays in Honour of Michael Windross, A. Remael, and K. Pelsmaekers (eds.), 115-126. Antwerpen: Garant. Davidse, Kristin, and Sara Geyskens 1998 Have you walked the dog yet? The ergative causitivization of intransitives. WORD 49(2): 155-180. Davidse, Kristin, and Liesbet Heyvaert in press On the so-called 'middle' construction in English and Dutch. In Empirical Approaches to Contrastive Linguistics and Translation Studies, S. Granger, J. Lerot, and S. Petch-Tyson (eds.). Amsterdam: Rodopi. Davies, Eirian 1979 On the Semantics of Syntax. Mood and Condition in English. London: Croom Helm. 2001 Propositional attitudes: a semantic model of epistemic mood and modality in English. Functions of Language 8 (2): 217-250. Declerck, Renaat 1986 The manifold interpretations of generic sentences. Lingua 68: 149-188. 1991a Tense in English. Its Structure and Use in Discourse. London: Routledge. 1991b A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha. Delacruz, Enrique B. 1976 Factives and proposition level constructions in Montague Grammar. In Montague Grammar, B. H. Partee (ed.), 177-199. New York: Academic Press. De Vries, Wobbe 1910 Opmerkingen over Nederlandsche syntaxis. [Comments on Dutch syntax]. Tijdschrift voor Nederlandsche Taal- en Letterkunde [Journal of Dutch Linguistics and Literature] 29: 122-165. Dik, Simon C. 1967 Some critical remarks on the treatment of morphological structure in transformational generative grammar. Lingua 18: 352383. 1980 Studies in Functional Grammar. London: Academic Press. 1997 The Theory of Functional Grammar. Part II. Complex and Derived Constructions. K. Hengeveld (ed.). Berlin/New York: Mouton de Gruyter. Ehrman, Madeline E. 1966 The Meanings of the Modals in Present-day American English. The Hague: Mouton.

270

References

Fagan, Sarah 1992

The Syntax and Semantics of Middle Constructions. Cambridge: Cambridge University Press.

Fawcett, Robin P. 1980 Cognitive Linguistics and Social Interaction. Towards an Integrated Model of a Systemic Functional Grammar and the Other Components of a Communicating Mind. Exeter/ Heidelberg: Julius Groos Verlag and Exeter University Press. Fellbaum, Christiane 1985 Adverbs in agentless actives and passives. Papers from the Parasession on Causatives and Agentivity at the Twenty-first Regional Meeting of the Chicago Linguistic Society. 21-31. 1986 On the Middle Construction in English. Indiana: Indiana University Linguistics Club. Fellbaum, Christiane, and Anne Zribi-Hertz 1989 The Middle Construction in French and English: A Comparative Study of its Syntax and Semantics. Indiana: Indiana University Linguistics Club Publications. Francis, Gill, Susan Hunston, and Elizabeth Manning 1996 Collins COBUILD Grammar Patterns 1: Verbs. London: HarperCollins. Fraser, Bruce 1970 Some Remarks on the Action Nominalization in English. In Readings in English Transformational Grammar. R. A. Jacobs, and P. S. Rosenbaum (eds.), 83-98. Waltham, Massachusetts. Gleason, Henry A. 1965 Linguistics and English Grammar. New York: Holt, Rinehart and Winston. Haas, William 1954 On defining linguistic units. Transactions of the Philological Society. 54-84. Haegeman, Liliane, and Jacqueline Guéron 1999 English Grammar. A Generative Perspective. Oxford: Blackwell Publishers. Haeseryn, Walter, K. Romijn, Guido Geerts, J. de Rooij, and M. C. van den Toorn (eds. ) 1997 Algemeen Nederlandse Spraakkunst [General Dutch Grammar]. Groningen: Martinus Nijhoff, Deurne: Wolters Plantyn. Haiman, John, and Sandra A. Thompson 1984 'Subordination' in Universal Grammar. Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society. 510-524. Hale, Kenneth, and Jay Keyser 1987 A View from the Middle. Lexicon Project Working Papers 10. Cambridge: MIT.

References

271

Halliday, Michael A. K. 1961 Categories of the Theory of Grammar. Word 17 (3): 241-292. 1966a Some notes on 'deep' grammar. Journal of Linguistics 2: 57-67. 1966b The concept of rank: A reply. Journal of Linguistics 2: 110-118. 1967 Notes on transitivity and theme in English. Part I. Journal of Linguistics 3 (1): 37-81. 1968 Notes on transitivity and theme in English. Part III. Journal of Linguistics 4 (2): 179-215. 1970a Functional diversity in language as seen from a consideration of modality and mood in English. Foundations of Language 6 (3): 322-361. 1970b Language structure and language function. In New Horizons in Linguistics, J. Lyons (ed.), 140-165. Harmondsworth: Penguin Books. 1976 System and Function in Language. G. R. Kress (ed.). London: Oxford University Press. 1979 Modes of meaning and modes of expression: types of grammatical structure, and their determination by different semantic functions. In Function and Context in Linguistic Analysis: Essays Offered to William Haas, D. J. Allerton, E. Carney, and D. Holdcroft (eds.), 57-79. Cambridge: Cambridge University Press. 1985 An Introduction to Functional Grammar. London: Edward Arnold. 1988 On the ineffability of grammatical categories. In Linguistics in a Systemic Perspective. J. D. Benson, M. J. Cummings, and W. S. Greaves (eds.), 27-51. Amsterdam/Philadelphia: John Benjamins. 1992 How do you mean? In Advances in Systemic Linguistics. Recent Theory and Practice, M. Davies and L. Ravelli (eds.), 20-35. London: Pinter. 1994 An Introduction to Functional Grammar (2nd ed.). London: Edward Arnold. Halliday, Michael A. K., and Ruqaiya Hasan 1976 Cohesion in English. London: Longman. Halliday, Michael A. K., and Christian M. I. M. Matthiessen 1999 Construing Experience Through Meaning. A Language-based Approach to Cognition. London: Cassell. Haugen, Einar 1953 On resolving the close apposition. American Speech 28: 165170. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics 25: 127-157.

272

References

Heyvaert, Liesbet 1997 Patientive -er nomináis. Leuven Contributions in Linguistics and Philology 86 (4): 433-456. 1998 Nonagentive deverbal -er nominalization in English and Dutch: a contrastive analysis. Languages in Contrast 1 (2): 211-243. 2000 Gerundive nominalization: from type specification to grounded instance. In Constructions in Cognitive Linguistics. Selected Papers from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997, A. Foolen, and F. van der Leek (eds.), 103-121. Amsterdam/Philadelphia: John Benjamins. 2001 Nominalization as an interpersonally-driven system. Functions of Language 8 (2): 287-329. in press Nominalization as grammatical metaphor: on the need for a radically systemic and metafìinctional approach. In Metaphor in Systemic-Functional Perspectives, L. Ravelli, A.-M. SimonVandenbergen, and M. Taverniere (eds.). Amsterdam/ Philadelphia: John Benjamins. Hjelmslev, Louis 1961 Prolegomena to a Theory of Language. Translated by F. J. Whitfield. Madison: University of Wisconsin Press. Hoekstra, Teun, and Ian Roberts 1993 Middle constructions in Dutch and English. In Knowledge and Language. Volume 2. Lexical and Conceptual Structure, E. Reuland, and W. Abraham (eds.), 183-220. The Netherlands: Kluwer Academic Publishers. Holierhoek, C. M. G. 1980 Werkwoorden van Waarneming. Aspekten van hun Systematiek en Ontwikkeling [Verbs of perception. Aspects of their functioning and development]. Den Haag: Universitaire Pers Leiden. Hooper, Joan B. 1975 On assertive predicates. In Syntax and Semantics. Volume 4, J.P. Kimball (ed.), 91-124. New York: Academic Press. Hopper, Paul J., and Sandra A. Thompson 1980 Transitivity in grammar and discourse. Language 56: 251-299. Hopper, Paul J., and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Horn, George M. 1975 On the Nonsentential Nature of the POSS-ING Construction. Linguistic Analysis 1 (4): 333-387. Huddleston, Rodney 1969 Review article on Madeline Ehrman, 'The meanings of the modals in present-day American English'. Lingua 23: 165-176. 1988 Constituency, multifunctionality and grammaticalization in Halliday's functional grammar. Journal of Linguistics 2: 137-174.

References

273

Hye Suk Yoon, James 1996 Nominal gerund phrases in English as phrasal zero derivation. Linguistics 34: 329-356. Iwata, Seizi 1999 On the status of an implicit arguments in middles. Journal of Linguistics 35: 527-553. Jespersen, Otto 1914-1929 A modern English grammar on historical principles. 7 volumes. London: George Allen and Unwin. Karttunen, Lauri 1971 Some observations on factivity. Papers in Linguistics 4: 55-69. Kastovsky, Dieter 1971 The old English suffix -er(e). Anglia. Zeitschrift für Englische Philologie. 89: 285-325. 1985 Deverbal nouns in Old and Modern English: from stemformation to word-formation. In Historical Semantics: Historical Word-Formation, J. Fisiak (ed.), 221-261. Berlin/New York: Mouton de Gruyter. 1986a The problem of productivity in word formation. Linguistics 24: 585-600. 1986b Diachronic word-formation in a functional perspective. In Linguistics across Historical and Geographical Boundaries. In Honour of Jacek Fisiak on the Occasion of His Fiftieth Birthday. Volume 1: Linguistic Theory and Historical Linguistics, D. Kastovsky, and A. Szwedek (eds.), 409-421. Berlin/New York: Mouton de Gruyter. Keenan, Edward L. 1971 Two Kinds of Presupposition in Natural Language. In Studies in Linguistic Semantics, C. Fillmore, and D. Langendoen (eds.), 45-52. New York: Holt, Rinehart and Winston. 1976 Towards a universal definition of 'Subject'. In Subject and Topic, C. N. Li (ed.), 303-333. New York: Academic Press. Kemmer, Suzanne 1993 The Middle Voice. Amsterdam/Philadelphia: John Benjamins. Keyser, Samuel J., and Thomas Roeper 1984 On the middle and ergative constructions in English. Linguistic Inquiry 15: 381—416. Kiparsky, Paul, and Carol Kiparsky 1971 Fact. In Semantics: An Interdisciplinary Reader in Philosophy, Linguistics, and Psychology, D. D. Steinberg, and L. A. Jakobovits (eds.), 345-369. Cambridge: Cambridge University Press. Kruisinga, E. 1932 A handbook of present-day English. 5th edition. Groningen: Noordhoff.

274

References

Kruisinga, E., and P.A. Erades 1953 An English Grammar. 8th edition. Groningen: Noordhoff. Kuno, Susumu 1970 Some properties of non-referential noun phrases. In Studies in General and Oriental Linguistics. Presented to Shirô Hattori on the Occasion of His Sixtieth Birthday, R. Jakobson, and S. Kawamoto (eds.), 348-373. Tokyo: TEC Company. Constraints on Internal Clauses and Sentential Subjects. Linguis1973 tic Inquiry 4 (3): 363-385. Laffiit, An 2000 The locative, image impression and material/product constructions. A functional description of the impact of NG-PrepP variation. Ph. D. diss., Department of Linguistics, University of Leuven. Lakoff, George 1977 Linguistic Gestalts. Papers from the 13th Annual Regional Meeting of the Chicago Linguistics Society. 236-287. 1987 Women, Fire, and Dangerous Things: What Categories Reveal about the Mind. Chicago: University of Chicago Press. Langacker, Ronald W. 1985 Observations and Speculations on Subjectivity. In Iconicity in Syntax, J. Haiman (ed.), 109-150. Amsterdam/Philadelphia: John Benjamins. 1987a Foundations of Cognitive Grammar. Theoretical Prerequisites. Stanford: Stanford University Press. 1987b Nouns and verbs. Language 63: 53-94. 1988 A Usage-Based Model. In Topics in Cognitive Linguistics, B. Rudzka-Ostyn (ed.), 127-161. Amsterdam/Philadelphia: John Benjamins. 1990a Concept, Image, and Symbol. The Cognitive Basis of Grammar. Berlin/New York: Mouton de Gruyter. 1990b Subjectification. Cognitive Linguistics 1: 5-38. 1991 Foundations of Cognitive grammar. Descriptive Application. Stanford: Stanford University Press. 1993 Reference-point constructions. Cognitive Linguistics A: 1-38. 1998 Conceptualization, Symbolization and Grammar. In The New Psychology of Language. Cognitive and Functional Approaches to Language Structure, M. Tomasello (ed.), 1-39. Mahwah, New Jersey: Lawrence Erlbaum Associates. 1999 Grammar and Conceptualization. Berlin/New York: Mouton de Gruyter. Lee, Donald W. 1952 Close apposition: an unresolved pattern. American Speech 27: 268-275.

References

275

Lees, Robert Β. 1960/1968 The Grammar of English Nominalizations. Bloomington: Indiana University Research Center in Anthropology, Folklore and Linguistics, publication 12. Lemmens, Maarten 1998 Lexical Perspectives on Transitivity and Ergativity. Causative Constructions in English. Amsterdam/Philadelphia: John Benjamins. Levin, Beth 1993 English Verb Classes and Alternations. Chicago: University of Chicago Press. Levin, Beth, and Malka Rappaport 1988 Nonevent -er nomináis: a probe into argument structure. Linguistics 26: 1067-1083. Lyons, John 1969 Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. 1977 Semantics. Volume 2. Cambridge: Cambridge University Press. Mackenzie, Lachlan J. 1990 First Argument Nominalization in a Functional Grammar of English. Linguistica Antverpiensia 24: 119-147. Malouf, Robert 2000 Verbal gerunds as mixed categories in head-driven phrase structure grammar. In Syntax and Semantics. Volume 32. The Nature and Function of Syntactic Categories, R. D. Borsley (ed.), 133— 166. San Diego: Academic Press. Marchand, Hans 1969 The Categories and Types of Present-day English Wordformation. A Synchronic-Diachronic Approach. 2nd edition. München: C. H. Beck'she Verlagsbuchhandlung. Massam, Diane 1992 Null objects and non-thematic subjects. Journal of Linguistics 28: 115-137. Matthews, Peter H. 1966 The concept of rank in 'Neo-Firthian' grammar. Journal of Linguistics 2: 101-109. Matthiessen, Christian M. I. M. 1995 Lexicogrammatical Cartography: English Systems. Tokyo: International Language Sciences Publishers. Matthiessen, Christian M. I. M., and Sandra A. Thompson 1988 The structure of discourse and 'subordination'. In Clause Combining in Grammar and Discourse, J. Haiman, and S. A. Thompson (eds.), 275-329. Amsterdam/Philadelphia: John Benjamins.

276

References

McGregor, William B. 1991 The concept of rank in systemic linguistics. In Functional and Systemic Linguistics: Approaches and Uses, E. Ventola (ed.), 121-138. Berlin/New York: Mouton de Gruyter. 1997 Semiotic Grammar. Oxford: Clarendon. Menzel, Peter 1975 Semantics and Syntax in Complementation. The Hague: Mouton. Milsark, Gary L. 1976 Existential Sentences in English. Bloomington: Indiana University Linguistics Club. 1977 Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3: 1-29. 1988 Singl-ing. Linguistic Inquiry 19 (4): 611-634. Munro, Pamela 1982 On the transitivity of 'Say' verbs. In Syntax and Semantics. Volume 15. Studies in Transitivity, P. J. Hopper, and S. A. Thompson (eds.), 301-318. New York: Academic Press. Noonan, Michael 1985 Complementation. In Language Typology and Syntactic Description. Volume 2. Complex Constructions, T. Shopen (ed.), 42-140. Cambridge: Cambridge University Press. Partee, Barbara H. 1973 The syntax and semantics of quotation. In A festschrift for Morris Halle, P. Kiparsky, and S. R. Anderson, 410-418. New York: Holt, Rinehart and Winston. Peeters, Roger J. 1999 The Adjunct Middle Construction in Dutch. Leuven Contributions in Linguistics and Philology 88 (3—4): 355-401. Perlmutter, David M. 1978 Inpersonal passives and the unaccusative hypothesis. Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society. 157-189. Postal, Paul M. 1970 On Coreferential Complement Subject Deletion. Linguistic Inquiry 1 (4): 439-500. Poutsma, H. 1929 A Grammar of Late Modern English. 2nd edition. Groningen: Noordhoff. Pullum, Geoffrey K. 1991 English nominal gerund phrases as noun phrases with verbphrase heads. Linguistics 29: 763-799. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1972 A Grammar of Contemporary English. London: Longman. 1985 A Comprehensive Grammar of the English Language. London: Longman.

References

277

Radford, Andrew 1997 Syntactic Theory and the Structure of English. A Minimalist Approach. Cambridge: Cambridge University Press. Rappaport Hovav, Malka, and Beth Levin 1992 -ER nomináis: implications for the theory of argument structure. In Syntax and Semantics. Volume 26: Syntax and the Lexicon, T. Stowell, and E. Wehrli (eds.), 127-153. New York: Academic Press. Ravelli, Louise J. 1988 Grammatical metaphor: an initial analysis. In Pragmatics, Discourse and Text. Some Systemically-inspired Approaches, E. Steiner, and R. Veltman (eds.), 133-147. London: Pinter. Roberts, Ian G. 1987 The Representation of Implicit and Dethematized Subjects. Dordrecht: Foris. Rosenbaum, Peter S. 1967 The Grammar of English Predicate Complement Constructions. Cambridge, Massachusetts: MIT Press. Rosta, Andrew 1995 'How does this sentence interpret?' The semantics of English mediopassives. In The Verb in Contemporary English. Theory and Description, B. Aarts, and C. F. Meyer (eds.), 123-144. Cambridge: Cambridge University Press. Ryder, Mary E. 1991 Mixers, mufflers and mousers: the extending of the -er suffix as a case of prototype reanalysis. Proceedings of the Berkeley Linguistics Society 17: 299-311. 1999a Complex -er nomináis. Where grammaticalization and lexicalization meet? In Between Grammar and Lexicon, E. ContiniMorava, and Y. Tobin (eds.), 291-331. Amsterdam/Philadelphia: John Benjamins. 1999b Bankers and blue-chippers: an account of -er formations in present-day English. English Language and Linguistics 3 (2): 269297. Saussure de, Ferdinand 1972 Cours de linguistique générale. Edition critique préparée par Tullio de Mauro. Paris: Payot. Schachter, Paul 1976 A nontransformational account of gerundive nomináis in English. Linguistic Inquiry 7 (2): 205-241. Schmid, Hans-Jörg 2000 English Abstract Nouns as Conceptual Shells. From Corpus to Cognition. Berlin/New York: Mouton de Gruyter.

278

References

Sinclair, John, Gwyneth Fox, Stephen Bullón, Ramesh Krishnamurthy, Elizabeth Manning, and John Todd 1990 Collins COBUILD English Grammar. London: HarperCollins. Smith, Carlota 1978 Jespersen's 'move and change' class and causative verbs in English. In Linguistic and Literary Studies in Honor of Archibald A. Hill, M. Jazayery, E. Polomé, and W. Winter (eds.), 101-109. The Hague: Mouton. Sohn, Joong-Sun 1998 The English middle and inchoative. Kansas Working Papers in Linguistics 23 (1): 1-16. Stockwell, Robert P., Paul Schachter, and Barbara H. Partee 1973 The Major Syntactic Structures of English. New York: Holt, Rinehart and Winston. Strang, Barbara M. H. 1968 Swift's agent-noun formations in -er. In Wortbildung, Syntax und Morphologie: Festschrift zum 60. Geburtstag von Hans Marchand, H. Brekle, and L. Lipka (eds.), 217-229. The Hague: Mouton. 1969 Aspects of the history of the -er formative in English. Transactions of the Philological Society. 1-30. Strawson, Peter 1959 Individuals: An Essay in Descriptive Metaphysics. London: Routledge. Sweet, Henry 1891 A New English Grammar. Logical and Historical. Oxford: Clarendon. Taeldeman, Johan 1990 Afleidingen op -seh semantiek, produktiviteit en integratie in een globale verantwoording van deverbatieve nomina [Derivations in -sel: semantics, productivity and integration in an overall account of deverbal nominalizations]. In Betekenis en Produktiviteit. Gentse Bijdragen tot de Studie van de Nederlandse Woordvorming [Meaning and productivity. Gent contributions to the study of Dutch word-formation], J. De Caluwe (ed.), 77-115. Studia Germanica Gandensia. Talmy, Leonard 2000 Toward a Cognitive Semantics. Volume 1. Concept Structuring Systems. Cambridge, Massachusetts: MIT Press. Taylor, John R. 1995 Linguistic categorization: prototypes in linguistic theory. 2nd edition. Oxford: Clarendon. 1996 Possessives in English. An Exploration in Cognitive Grammar. Oxford: Clarendon.

References

279

Thompson, Sandra A. 1973 On subjectless gerunds in English. Foundations of Language 9: 374-383. Vachek, Josef 1966 The Linguistic School of Prague. Bloomington: Indiana University Press. Vandenberghe, Wim 2001 Instigative Setting-constructions: Force Dynamic Research on 'New' Types of Agency. Leuven Contributions in Linguistics and Philology 90: 365-390 Van den Toorn, M.C. 1982 Nederlandse Grammatica [Dutch grammar]. Groningen: Wolters Noordhoff. Van Langendonck, Willy 1988. -ER als akkusatief en -LING als ergatief suffix in een Zuidbrabants dialekt [-ER as an accusative and -LING as an ergative suffix in a South-Brabant dialect], Spektator 17 (4): 317-327. 1999 Neurolinguistic and syntactic evidence for basic level meaning in proper names. Functions of Language 6 (1): 95-138. Van Oosten, Jeanne 1977 Subjects and agenthood in English. Papers from the 13th Annual Regional Meeting of the Chicago Linguistics Society. 451-471. 1986 The Nature of Subjects, Topics and Agents: A Cognitive Explanation. Bloomington: Indiana University Linguistics Club. Vendler, Zeno 1967 Linguistics in Philosophy. Ithaca/New York: Cornell University Press. Verstraete, Jean-Christophe 2001 Subjective and objective modality: Interpersonal and ideational functions in the English modal auxiliary system. Journal of Pragmatics 33: 1505-1528. 2002 Interpersonal grammar and clause combining in English. Ph. D. diss., Department of Linguistics, University of Leuven. Wasow, Thomas, and Thomas Roeper 1972 On the subject of gerunds. Foundations of Language. International Journal of Language and Philosophy 8: 44—61. Wierzbicka, Anna 1988 The Semantics of Grammar. Amsterdam/Philadelphia: John Benjamins. Williams, Edwin 1980 Predication. Linguistic Inquiry 11 (1): 203-238. 1981 Argument Structure and Morphology. The Linguistic Review 1: 81-114.

280

References

Yoshimura, Kimihiro forthcoming The Middle Construction in English: A Cognitive Linguistic analysis. Berlin/New York: Mouton de Gruyter.

Index

action nominalization, 46-47,48, 51, 52, 203-204 ad hoc nominalization, 45, 57,100101,117, 120-123,158, 165172,259 event reading, 122,165, 167, 172 nominal grounding, 167-169, 172,173-174 non-phoric, 168-172 phoric, 120-121, 166-168, 172 postmodification, 121-122, 170-171 premodification, 121, 169-171 temporal grounding, 166-167,172, 173-174 Agent Actor vs. Instigator, 108-109, 152 diachronic origin of deverbal -er, 99, 123 modal value of agentive -er, 161-163 profile of deverbal -er, 99-110, 161-163 schematic category, 151-153 systemic-functional view of, 29-30 agnation construct of, 35-36 descriptive heuristic, 37-40, 55-56 networks of, 54-55 relevance to nominalization, 6, 7, 52-58,94 schematic nature of, 56-58, 150 Transformational Grammar, 53-55 analogy, 45-46 apposition, 213,217-218

double grounding, 70, 217-220 vs. postmodification, 211-217 argument external, 110-115 internal, 111-114 assembly construction, 20 level of, 6,41,49, 94-95, 222, 225-226, 242, 244-245, 249252 order of, 48,226 atemporal clausal head, 73-76, 81, 86, 94, 222-228,243 auxiliaries grounding vs. non-grounding, 71,224 categorization elaboration/ instantiation, 1215, 27, 34 extension, 5, 12, 27, 34,43-44 schématisation, 5, 12, 14,27, 43 class external functional properties, 19, 23-24, 26, 50, 92 functional approach to, 4,19, 23-28, 50-51, 92 internal functional properties, 19, 24-26, 92 notional definition of, 23, 26 clause correspondences with nominal, 6, 70, 71, 81,92-93 functional approach to, 6, 7092 grounding, 71, 81-87 instantiation, 78-79, 81 quantification, 79-81 type specification, 71-78

282

Index

clausal head finite vs. atemporalized, 73-76, 81, 86, 94 inclusion of Objects, 76-77 Cognitive Grammar, 4-5, 11, 13, 15, 17, 91, 258 conventional, 12, 13 common noun strategy nominalizations, 95, 201-202, 209, 232-233, 238,240, 241244, 246, 257-258 ordinary nouns, 61-62, 81 comparison, 11, 12 complexity conceptual category, 101, 106108 membership of a family of structures, 14, 15, 26-27,43 symbolic complexity, 18, 19, 44-45, 47 composition, 11-14,47-49, 259 analytic definition of, 21 building-block view of, 20, 257 component functions, 19, 20 as integration, 21, 95, 251-252 radically functional approach to, 19-28 synthetic definition of, 21 constituency, 3, 19, 20, 21-23, 25, 28 construction construct of, 13 integration, 21, 95, 251-252 radically functional approach to, 19-28 downranking or rank shift, 5, 22, 49-51, 95, 186, 208,227-228,239,249250,251 elaboration instantiation, 12, 14 embedding, 185-187, 188-194 enation

the construct of, 35-40 partial enation, 36 non-enation, 36, 39 entrenchment conventionalization, 12, 13, 172 degrees of, 5, 15,44-46, 100101, 165 ergativity, 108-110, 111, 262 Instigator and Medium, 108110 unaccusatives, 113-117 -er nominalization agentive profile, 52, 99-110 agnation with the middle, 7, 14, 117-118, 150,154-158, 159161,262

argument structure, 110-118 conduciveness, 154-158, 159 complex conceptual category, 106-108

denominal, 102-103, 106-107, 150 destiny-oriented, 156 facility-oriented, 155-157 feasibility-oriented, 156-157 high-level schema, 100, 110 low-level schemata, 44 modal value of agentives, 161163 modal value of non-agentives, 159-161 result-oriented, 157 salience, 102-106 Subject profile, 118-120, 123, 154,158,159, 174-176 verbal base of, 108-110 expansion, 185, 187-188, 211-217 elaboration, 214-217 extension of the prototype, 12-13, 14-15, 26-28,43-44 type of expansion, 214-217 factive embedded projection, 185-195

Index external functioning, 183-185, 185-187, 191-194,247 instantiation, 204-205, 241244, 245-248 internal functioning, 200-205, 207-248 nominalization, 181-205, 207248 presupposition, 181, 183-185, 189-190 finite element grounding, 81-85, 93 modality, 81-85 tense, 81, 84-85 fixed expression, 5,44, 101, 165, 172 generic bare plurals and mass nouns, 65, 234 middle, 142, 264 gerundive nomináis, 198, 229230,233-236, 238, 240 gerundive nominalization action, 16,46-47, 199, 232233, 237-238, 240, 247 atemporal clausal head, 222228,243 common noun strategy, 232233,238,240, 241-244 factive, 6, 8,46-47,48, 181182, 183, 195-196, 198-200, 201,203-205, 221-244,244248 generic action, 198, 229-230, 233-236,238, 240 -ing, 224-227 instantiation, 204-205, 241244, 245-248 internal functioning, 200, 203205, 222-244 possessive/genitive, 232, 242, 266 proper name strategy, 233-238, 239-240,241-244

283

Subject, 242-244, 244-248 typology of, 231-240 ground, 62, 71, 81 grounding clausal, 6, 57-58, 81-92, 260261 definite vs. indefinite, 66-68, 167-169,173-174 Finite, 81-85 interpersonal status, 31, 93, 95 nominal, 6, 25-26, 66-69, 167169,172, 173-174 quantification, 64-66 Subject, 85-86 holistic scanning, 51, 75, 208,210, 224-225 head clausal, 71-78 nominal, 62-63, 92 hypotaxis, 185-188, 190-194, 196198,213-217 iconic encoding vs. non-iconic encoding, 4, 2021,24, 25, 28,51,62, 63,6970, 81, 92, 94, 208, 209-210, 246,250-251 ideational metafunction, 6, 30-31, 38-39, 52, 95 modality, 83-84 -ing, 72, 74-76, 224-227, 260 instantiation clause, 6, 78-81, 86 elaboration, 12, 13, 14, 27, 34, 43 gerundive nomináis, 204-205, 241-244,245-248 nominal, 6,25-26, 62-62 usage event, 10, 27, 33 Instrumental -er, 100, 103, 107, 111-112, 115, 149, 163-164

284

Index

interpersonal metafunction, 6, 31, 34, 38-39, 52, 61, 76, 93, 95, 150,175 modality, 82-84 letting, see modality lexicalization, 5,44-45, 100-101, 120-123,158, 159-164 non-event reading, 122, 160, 172 postmodification, 122 lexicogrammar construct of, 16-17,45 relation with semantics, 17-19 metafunctions as clusters of lexicogrammatical options, 30, 33, 37-39 as configurations of functions, 31-33 interpersonal/constructional, 6, 31,34, 38-39, 52,61,76, 93, 95, 150, 175 representational/ideational, 6, 30-31, 38-39, 52, 92-93, 95, 110, 128-132, 150-154 textual, 6, 31, 52, 93, 151 metaphor, 106-108 metonymy, 106-108 middle construction, 111, 125-147, 263 agency, 131 agnation, 7, 114, 117-118, 150, 159-161 conduciveness, 137, 141, 143144, 147, 154 destiny-oriented, 134-135, 136 ergative approach, 126-127, 263 facility-oriented, 132-133, 136 feasibility, 132-134, 136 intransitive middles, 128-130, 263 modal value, 137-142, 159 passive approach, 127-128

quality-oriented, 133,136 result-oriented, 135-136 Subject profile, 131-132, 137, 140-147 modality agentive -er, 161-163 deontic, 82-85, 89-91, 139 dynamic, 83-85,140-141, 159, 161 epistemic, 82-85, 89-91, 138, 159 'letting', 141-147 middle constructions, 137-142 non-agentive -er, 159-161 objective vs. subjective, 83, 140-141, 159, 261 schematic, 82, 89-91, 141, 159-160 modalization, 83 modulation, 83 multifunctional approach, 4, 6,2933, 52, 95, 150, 177 natural description of language, 15-17, 20 non-arbitrary relation between lexicogrammar and semantics, 17-19, 105 network schematic, 14, 15,45 system, 4, 7, 13, 30, 33-34, 3739 nominal class, 25, 92 correspondences with the clause, 6, 70, 71, 81, 92-93 functional approach to the, 6, 20-21, 25, 61-70 vs. noun, 25-26 non-agentive -er, 100, 103-104,107, 115-118, 149, 153-158 agnation with the middle, 159161

Index puipose-designed status of, 107, 154-155 semantic typology, 154-158 noun nominal, 25-26, 62 non-plural mass noun, 63, 234, 260 plural noun, 63, 234, 260 proper noun, 26, 61-62, 69 singular count noun, 63 oblique participants definition, 115 -er nominalization, 115, 149150,153-154, 262 paradigmatic relations of components, 33, 53, 55, 228 of constructions (agnation, enation), 6, 7, 8, 33-40, 52-58 parataxis, 186-187,213-217 Patient Goal or Medium, 108-110 person deixis, 85-86 polarity, 79-80 productivity, 44-45, 149-150 projection, 185, 187-190, 194, 211217 pronouns, 21, 69 progressive -ing, 75-76, 260 proper name strategy, 61-62, 65, 69, 95, 209-210, 233-238, 239-240, 241-244,246, 258 prototype criteria for identification of the, 27-28, 257-258 extension of, 15, 27-28,43-44, 149 reanalysis of -er, 102-104 theory, 14 quantification clausal, 6, 79-81 nominal, 6, 25-26, 63

285

quantifier absolute, 64—66 relative, 64-66 rank scale downranking or rank shift, 5, 22,49-51, 95, 186, 208, 227228, 239, 249-250,251 radically functional approach to, 4, 5,19, 22-23, 28,49-51 structural, 21-23, 257 reclassification external/synthetic vs. internal/analytic 4, 5, 7,49-51,208210,249-252 functional approach to, 4, 5051, 94-96, 208,227-228, 249250 structural approach to, 49-50 reference mass, 64, 66-68 representational metafiinction, 6, 30-31, 38-39, 52, 95, 110, 128-132, 150-154 modality, 83-84 Residue, 76-77, 93 responsibility causal or agentive, 143 middle Subject, 126-127, 131— 132,143 objective, 91-92, 143 speech-functional or modal, 8792, 143 salience, 102-106 sanctioning, 10, 12, 27,43 scanning serial or sequential, 75, 208 summary or holistic, 75, 208, 224-225 schema constructional, 13,43, 106 -er nominalization, 100, 103, 106,110,174-176 high-level vs. low-level, 43-45, 100, 106,149

286

Index

schematic characterization of class, 19, 23-25 schematic modality, 81-85, 141 schematic nature of agnation, 56-58, 150 schematic network, 14, 15,45, 149 schematization the psychological ability of, 11, 12 in the language system, 13, 1415,43 Setting-Subject construction, 163 Subject gerundive nomináis, 241-244 instantiating function, 78-79, 86, 175, 204-205, 241-244, 245-248, 261 locus of potency, 140-144, 159-163, 176 middle, 126-127,131-132, 143-144 person deixis, 85-86,175, 262 primary focal participant, 9192, 258 profile of -er, 118-120, 149150, 159-164, 174-176 speech-functional role, 87-92, 175-176 systemic-functional view of, 29-32 Subject-Finite unit -er nominalization, 149, 159, 174 grounding, 76-87, 93 interpersonal nature, 37-38, 8193 mood options, 37-38, 87-92 speech-functional value, 87-92 syntagmatic relations, 7, 12, 13, 1923, 33 symbolic relation lexicogrammarsemantics, 4, 5, 7, 17-19,41, 46-47, 105,257

the psychological ability of symbolization, 11 unit, 11, 12, 13,17, 105, 257 system interaction with usage, 4, 9-15 systemic motivation, 10, 12, 42 system network, 4, 7, 13, 30, 33-34, 37-39 systemic (or paradigmatic) relations, 4, 19, 33-40, 52-58 Systemic-functional Grammar, 4-5, 15, 17, 29, 33, 258, 259 textual metafunction, 6, 31, 52, 93, 151 íAaí-nominalization, 49, 69-70 embedded or downranked projection, 185-195, 195-198, 208 factive, 6, 8, 181-205,207-210 factive prototype, 181-182 hypotactic, 186-187, 189, 191194, 195-198 internal functioning, 200-203, 209-210 locutions/ideas, 193, 265 proper name strategy, 209-210, 220-221 Theme Government-and-Binding, 111 Systemic-Functional grammar, 29-32 the fact //¡«/-nominalization, 6, 8, 181-182, 183, 190, 195, 197198, 202,211-221 apposition vs. postmodification, 211-217 doubly-grounded nominal, 217221

vs. embedded locutions/ideas, 211-212,215-217, 220 vs. relative clauses, 211-212, 215,217, 266 transcategorization, 41, 227 transitivity, 108-110, 262 Actor and Goal, 108-110

Index unergatives or intransitives, 113, 115-117 type specification clausal, 6, 72-77 definite vs. indefinite reference, 68, 168-169 nominal, 6, 25-26, 62-63, 6668 unaccusatives, see ergativity unergatives, see transitivity units conventional, 5, 10, 12,44

287

symbolic, 11, 13, 17, 19, 257 schematic, 13,45,165 semantic, 13, 17 phonological, 13, 17 usage usage-based approach, 11,42-46 event, 27 verb group internal complexity, 70-72, 261 processual profile, 72, 74 atemporal, 73-76

Cognitive Linguistics Research Edited by René Dirven, Ronald W. Langacker and John R. Taylor Mouton de Gruyter · Berlin · New York

1 2 3 4 5 6 7 8 9 10 11 12

13 14 15 16

Ronald W. Langacker, Concept, Image, and Symbol. The Cognitive Basis of Grammar. 1990. Paul D. Deane, Grammar in Mind and Brain. Explorations in Cognitive Syntax. 1992. Conceptualizations and Mental Processing in Language. Edited by Richard A. Geiger and Brygida Rudzka-Ostyn. 1993. Laura A. Janda, A Geography of Case Semantics. The Czech Dative and the Russian Instrumental. 1993. Dirk Geeraerts, Stefan Grondelaers and Peter Bakema, The Structure of Lexical Variation. Meaning, Naming, and Context. 1994. Cognitive Linguistics in the Redwoods. The Expansion of a New Paradigm in Linguistics. Edited by Eugene H. Casad. 1996. John Newman, Give. A Cognitive Linguistic Study. 1996. The Construal of Space in Language and Thought. Edited by Martin Pütz and René Dirven. 1996. Ewa D^browska, Cognitive Semantics and the Polish Dative. 1997. Speaking of Emotions: Conceptualisation and Expression. Edited by Angeliki Athanasiadou and Elzbieta Tabakowska. 1998. Michel Achard, Representation of Cognitive Structures. 1998. Issues in Cognitive Linguistics. 1993 Proceedings of the International Cognitive Linguistics Conference. Edited by Leon de Stadler and Christoph Eyrich. 1999. Historical Semantics and Cognition. Edited by Andreas Blank and Peter Koch. 1999. Ronald W. Langacker, Grammar and Conceptualization. 1999. Cognitive Linguistics: Foundations, Scope, and Methodology. Edited by Theo Janssen and Gisela Redeker. 1999. A Cognitive Approach to the Verb. Morphological and Constructional Perspectives. Edited by Hanne Gram Simonsen and Rolf Theil Endresen. 2001.

17

Emotions in Crosslinguistic Perspective. Edited by Jean Harkins and Anna Wierzbicka. 2001. 18 Cognitive Linguistics and Non-Indo-European Languages. Edited by Eugene Casad and Gary B. Palmer. 2003. 19.1 Applied Cognitive Linguistics I: Theory and Language Acquisition. Edited by Martin Pütz, Susanne Niemeier and René Dirven. 2001. 19.2 Applied Cognitive Linguistics II: Language Pedagogy. Edited by Martin Pütz, Susanne Niemeier and René Dirven. 2001. 20 Metaphor and Metonymy in Comparison and Contrast. Edited by René Dirven and Ralf Pörings. 2002. 21 Grounding. The Epistemic Footing of Deixis and Reference. Edited by Frank Brisard. 2002. 22 Cristiano Broccias, The English Change Network. 2003. 23 Cognitive Approaches to Lexical Semantics. Edited by Hubert Cuyckens, René Dirven and John Taylor. 2003. 24 Cognitive Models in Language and Thought. Ideology, Metaphors, and Meanings. Edited by René Dirven, Roslyn Frank and Martin Pütz. 2003. 25 Rong Chen, English Inversion. A Ground-before-Figure Construction. 2003. 26 Liesbet Heyvaert, A Cognitive-Functional Approach to Nominalization in English. 2003.