Modality in English: Theory and Description (Topics in English Linguistics 58)

Modality in English

≥

Topics in English Linguistics 58

Editors

Elizabeth Closs Traugott Bernd Kortmann

Mouton de Gruyter Berlin · New York

Modality in English Theory and Description

Edited by

Raphael Salkie Pierre Busuttil Johan van der Auwera

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Modality in English : theory and description / edited by Raphael Salkie, Pierre Busuttil, Johan van der Auwera. p. cm. ⫺ (Topics in English linguistics ; 58) Includes bibliographical references and indexes. ISBN 978-3-11-019634-4 (alk. paper) 1. English language ⫺ Modality 2. Modality (Linguistics) I. Salkie, Raphael. II. Busuttil, Pierre. III. Auwera, Johan van der. PE1315.M6M634 2009 4251.6⫺dc22 2009017662

ISBN 978-3-11-019634-4 ISSN 1434-3452 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” Copyright 2009 by Walter de Gruyter GmbH & Co. KG, 10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Laufen. Printed in Germany.

Contents

Introduction Raphael Salkie, Pierre Busuttil and Johan van der Auwera

1

Towards a typology of modality in language Paul Larreya

9

‘Not-yet-factual at time t’: a neglected modal concept Renaat Declerck

31

Semantic ascent, deixis, intersubjectivity and modality Keith Mitchell

55

Degrees of modality Raphael Salkie

79

Another look at modals and subjectivity Jelena Timotijevic

105

For a topological representation of the modal system of English Alain Deschamps and Lionel Dufaye

123

Epistemic might in the interrogative Gilbert Ghio

145

MAY in concessive contexts Jean-Claude Souesme

159

When may means must: deontic modality in English statute construction Ross Charnock

177

Legal English and the ‘modal revolution’ Christopher Williams

199

vi Contents Posteriority in expressions with must and have to: a case of interplay between syntax, semantics and pragmatics An Verhulst

211

Using the adjectives surprised/surprising to express epistemic modality Christiane Rocq-Migette

223

Commitment and subjectivity in the discourse of a judicial inquiry Juana I. Marín-Arrese

237

Hearsay adverbs and modality Agnès Celle

269

When Yes means No, and other hidden modalities Claude Rivière

295

Modality and the history of English adhortatives Manfred Krug

315

On the “great modal shift” sustained by come to VP Philippe Bourdin

349

List of contributors

375

Author index

377

Subject index

380

Introduction Raphael Salkie Pierre Busuttil Johan van der Auwera

This book contains some of the papers presented at the Second International Conference on Modality in English (ModE2), held at the University of Pau in September 2004. This conference followed the Conference on Modality in Contemporary English, held at the University of Verona in 2001, so in the same way this book is the successor to Facchinetti, Krug and Palmer (eds.) (2003), and Facchinetti and Palmer (eds.) (2004). There are at least three general themes. One is the definition of the concept of modality, its relation to other concepts, and the general strategies to approach modal notions. The second theme is the study of the English modals. The third is the analysis of modal constructions other than auxiliaries. Of course, there is also an overlap, especially between the first and the second themes, for the general papers all illustrate the point with English modal auxiliaries.

What is modality? How should one study it? For Paul Larreya it is important to distinguish between modality and modalization. The former is a mental system based on the notions of possibility and necessity. He discusses various subtypes, some known from other work (root, epistemic, deontic), and he argues that epistemic modality comes in two subtypes, called ‘problematic’ and ‘implicative’. He also discusses the status of volition. As for modalization, Larreya argues that we should distinguish five types and he argues that this is the right place to handle the effect of the past tense (yielding might as a ‘qualification’ of may) or the counterfactual uses. Most of the argumentation deals with modern English, but he also refers to German and to Old English.

2 Salkie, Busuttil and van der Auwera Renaat Declerck argues that modality crucially involves a concept of “not-yet-factuality”. He defines this notion relative to three other notions, viz. factuality, hypotheticalness and counterfactuality. He also describes subtypes of this notion and illustrates their use in the analysis of a variety of temporal constructions such as until or before clauses and the future tense. But, most importantly for this volume, modality too is claimed to manifest a dimension of the ‘not yet factual’. Keith Mitchell relates subtypes of modality to Lyons’s (1977) distinctions between first, second and third order entities, paying particular attention to what is sometimes called “existential modality”, as in footballers can be sex maniacs. He examines to what extent it makes sense to treat both modality and sentence mood as a kind of deixis, a question to which his answer is positive. This further leads him to consider the auxiliary do as a modal auxiliary and to throw light on quasi-subjunctive should (as in it’s incredible that we should both have the same birthday). Raphael Salkie’s chapter argues for a prototype approach to modality, which is furthermore taken to be of typological value. He defines core modality in terms of four properties and applies the definition to the English auxiliaries. He also relates the concept of modality to the concepts of irrealis, mood, and evidentiality. The prototype approach, Salkie further argues, proves its utility in that it sheds new light on two puzzles in the analysis of English modals, involving the distinctions between may and can, on the one hand, and must and should, on the other. The solutions he proposes rely on applying a notion of degree of modality based on distance from the prototype, and he gives reasons to prefer his analysis to earlier work which takes a similar approach. Jelena Timotijevic focuses on the relation between modality and subjectivity. She provides a survey of the way the notion has been used in the domain of modality. In her own proposal, based on ideas in Recanati (2004), the use of the notion relates to the distinction between the semantics and the pragmatics of modality: the subjective (uses of the) modals involve more pragmatics than the objective uses. Her proposal is inspired by, though not identical, to the one in Papafragou (2000). From this angle she also turns to the monosemy vs. polysemy debate, with monosemy being more appropriate for objective modals and polysemy for subjective ones.

Introduction

3

English modal auxiliaries The book contains six papers which are set within the Theory of Enunciative Operations, developed by Antoine Culioli (cf. Culioli 1990, 1995), three in this section and three in the next. For the readers who are not familiar with the terminology used in this approach to discourse analysis, we offer a few notes which, we hope, are going to be helpful. The explanations we propose are, necessarily, simplified and partial, but we hope that they will enable readers to appreciate the principal arguments of these six papers. For a more substantial account in English, see Groussier (2000); see also Groussier and Rivière (1996) for definitions in French of the key terms and suggestions of English equivalents, and Bouscaren et al. (2008) for a glossary in English with extensive English translations of key passages from theoretical work. The theory employs technical terms, as well as familiar terms which are used in specialised ways. A basic distinction is between the speaker (French locuteur) and the enunciator (Fr. énonciateur). A speaker is a human being who produces speech sounds. The enunciator is an abstract function, not a real person: the source of a set of operations which result in an utterance in a context. Operations take place on several different levels. The most primitive level is that of notions, which can be thought of crudely as the things that words refer to: for example, the notion of TABLE. Speakers of a language will agree that certain entities in the real world are tables, some are clearly not tables, and others are unclear cases. The theory conceives of this in topological terms: each notion is associated with a notional domain, which has an interior I (tables), an exterior E (not tables) and a frontier F (unclear cases). The next level is when notions are combined to produce a predicative relation – crudely: a proposition, the thing that is constant in John loves Mary, Does John love Mary?, John may love Mary, and so on. One operation on this level is scanning (Fr. parcours), in which the enunciator reviews all the possibilities without choosing one. The word any in Pop by any time or I looked for soap but I couldn’t find any is a marker of a scanning operation. Predicative relations can themselves be treated additionally as com-

4 Salkie, Busuttil and van der Auwera plex notional domains: they can be true in the real world, or false, or unclear as to truth value and thus in the frontier. In order for a predicative relation to become a real utterance, a third level of enunciative operations applies to situate it in space and time and in relation to the knowledge, wishes, etc of the enunciator. Modality has traditionally been treated within the theory in terms of three planes (Fr. plans), thought of as spaces within which operations apply. The first is the plane of representation, sometimes called the pre-modal plane, where the enunciator does not make a choice between the interior I and the exterior E. The second is the plane of validation where the enunciator chooses whether the predicative relation is true or false. The third is the hypothetical plane, sometimes called the plane of fiction, where the enunciator can consider imaginary or counterfactual situations, free from the constraints of reference to the real world. By using a modal operator such as a modal verb, the enunciator can operate on (alternatively, “work on” or “play with”) more than one plane in a single utterance. Recent work in the theory has distinguished between two different types of operations at the third level of enunciative operations. Operations which evaluate the properties of a situation are called qualitative, while operations which relate to actual occurrences of a situation are termed quantitative. As far as the metalanguage is concerned, the following terms appear in the papers which are set within the enunciative framework: Otherness (Fr. altérité): A general term for opposition or contrast between two linguistic items. The contrast between a negative and a positive sentence is one type of otherness. Path (Fr. chemin): If the enunciator is conceived as being detached from a notional domain, the choice of I or E can be represented as a path from detachment to either I or E. Value (Fr. valeur): This word is sometimes used in the theory in a way that will be familiar to most readers – in expressions like “positive vs. negative value”, or in the sense of “value of a variable”. In the expression “epistemic value”, however, the word is used to mean roughly “meaning” or “use” or “meaning and use”.

Introduction

5

Determination: Roughly this term refers to what determiners do to nouns – they make nouns more specific or establish their reference. Within the theory, however, the term has a broader use: most operations are thought of as determining linguistic items, including predicative relations, more precisely. This often involves situating one item in relation to another item which is already more highly determined. The most comprehensive of the six papers is by Alain Deschamps and Lionel Dufaye. They present an account of the English modal auxiliaries may, will, must, can and shall. This formal analysis is illustrated in detail for will and special attention is also given to the combination of the modals with negation. Gilbert Ghio analyzes the interrogative uses of epistemic might, as in Might this be the beginning of the end?, in the enunciative framework. It is argued that such questions can only be rhetorical and that the modal is outside the scope of negation in examples like Might this not be the moment to give …. He furthermore explains why this use is impossible for may and could. Jean-Claude Souesme analyses the concessive uses of may, as in Literature may not be as noble as mathematics, but …. He claims that the concessive effect is primarily or exclusively triggered by markers such as although and but, and that the meaning of may is very close to its normal epistemic one. Souesme then characterizes the semantics of may in terms of the framework of Culioli and he provides an overview of the types of contexts that allow the concessive effect. Two papers deal with the special uses of modal auxiliaries in legal English. Ross Charnock studies a legal use of the auxiliary may in which it has a meaning similar to must. The point is not simply that this coercive sense is an effect of pragmatics; rather, under certain conditions, it is taken to be an aspect of the literal meaning of may. The author analyses the coercive sense of may in three authentic cases. He tries to explain why legislators may even prefer this use of may to shall; the use of the latter is argued to be performative and thus bring along a sense of inevitability. He also comments on the fact that even though judges are supposed to take words in their literal sense, they are implicitly embracing a contextualist approach. Christopher Williams discusses the present status of shall in legal English. Though it has been the most common modal in legal English for at

6 Salkie, Busuttil and van der Auwera least six hundred years, its use is currently in decline, especially in Australia, South Africa and New Zealand. The author identifies the structures that take up the place left behind by shall in the shall-free texts, viz. the simple present, must, may and be to. He also speculates as to what will or will not happen with shall in the legal language of the US and the UK. An Verhulst focuses on the expression of time in modal utterances, more particularly, on posteriority in expressions with must and have to. She pays attention to the aspectual nature of the infinitive following must and have to, to pragmatic information, and to the role of adverbs of time and frequency.

Modality beyond modal auxiliaries The chapter by Christiane Rocq-Migette is a detailed corpus study of the epistemic modality expressed by constructions of the type I would be surprised if ... and it would be surprising if… It is shown that the use of a predicate of surprise can combine with other factors and yield the epistemic interpretation. Special attention is also given to the use of negation and tenses. Juana I. Marín-Arrese explores how dimensions and degrees of speaker commitment and subjectivity are expressed in three registers of spoken British discourse, all connected to a court case, but different as they pertain to either an institutional domain (the Government), a social domain (the BBC) or a private one (Family). The author analyzes the semantic categories and the formal tools, the latter including modal auxiliaries, but not restricted to these, and then sets out on a detailed corpus analysis. The general perspective is a cognitive linguistic one. Agnès Celle does not study verbs or verbal mood but adverbs, more particularly, the hearsay adverbs reportedly, allegedly and supposedly, working in the enunciative framework. The adverbs are considered modal in the sense that they crucially concern the speaker’s commitment to the utterance, or, for these adverbs, his or her lack of commitment. Starting from Greenbaum (1969), she develops a detailed corpus-based syntactic and semantic analysis of these three adverbs and of the way they also differ from other modal adverbs such as obviously or apparently. Claude Rivière, working within Culioli’s framework, examines two types of “hidden” modalities through a consideration of the it be … since / before… constructions. First, a negative meaning can be expressed without

Introduction

7

a negation being apparent or included in the meaning of any of the terms used. Conversely, a positive reading can be extracted from the not VERB for… negative construction. Secondly, the conditions that allow a negative reading are not the same for the since and for the before constructions. This asymmetry turns out to go deeper than the prospective/retrospective asymmetry and reveals another hidden modality: expectation. Manfred Krug turns to “adhortative” constructions, such as Let’s go. Adhortatives, for Krug, are expressions of syntactic mood, and the latter is supposed to be part of a (wide) category of modality. Krug first studies the diachrony of English let’s from the point of view of grammaticalization theory. Then he deals with synchronic variation allowing e.g. let’s you and I or let’s everybody. The negated patterns are also studied, again both diachronically and synchronically, with some attention to regional preferences (British don’t let’s vs. American let’s don’t). Philippe Bourdin, who uses the enunciative framework, is probably the furthest removed from the core of modality. Having studied a special use of go in Facchinetti et al. (2003), as in It went unnoticed, in this paper he turns to a special use of come, as in How did you come to learn Navajo? He compares this special use to an ordinary movement use as in He came to Paris to see Sam, in which the to-infinitive expresses a purpose. The latter is called a “Control” interpretation, which involves a component of intentionality, and the former a “Raising ” interpretation. Control go is the older use, and when it developed the Raising use a semantic shift, more particularly, a “demodalization” occurred, in the sense that the Raising use dropped the intentionality component. Modality is one big intrigue. Questions erstwhile considered solved become open questions again. New observations and hypotheses come to light, not least also because the subject matter is changing. May this book play a role in the denouement of the modal intrigue.

References Bouscaren, Janine, Jean Chuquet, Eric Gilbert, Hélène Chuquet, and Ronald Flintham. 2008 English definitions of key terms in the Theory of Enunciative Operations. Available from http://www.sil.org/linguistics/glossary

8 Salkie, Busuttil and van der Auwera _glossary_fe/defs/TOEEn.asp Culioli, Antoine 1990 Pour une Linguistique de l’Énonciation: Opérations et Représentations, Tome 1. Paris: Ophrys. [Parts of this work can be read online in Google Books, including an entire chapter in English entitled “The concept of notional domain”, pp. 67–81] 1995 Cognition and Representation in Linguistic Theory. Michel Liddle (ed.). Amsterdam: John Benjamins. Facchinetti, Roberta, Manfred Krug, and Frank Palmer (eds.) 2003 Modality in Contemporary English. (Topics in English Linguistics [TiEL] 44). Berlin: Mouton de Gruyter. Facchinetti, Roberta, and Frank Palmer (eds.). 2004 English Modality in Perspective: Genre Analysis and Contrastive Studies. (English corpus linguistics 1). Bern: Peter Lang. Greenbaum, Sidney 1969 Studies in English Adverbial Usage. London: Longman. Groussier, Marie-Line 2000 On Antoine Culioli’s theory of enunciative operations. Lingua 110.3: 157–182. Groussier, Marie-Line, and Claude Rivière. 1996 Les mots de la Linguistique: Lexique de Linguistique Ènonciative. Paris: Ophrys. Lyons, John 1977 Semantics 2. Cambridge: Cambridge University Press. Papafragou, Anna 2000 Modality: Issues in the Semantics-Pragmatics Interface. Amsterdam: Elsevier. Recanati, François 2004 Literal Meaning. Cambridge: Cambridge University Press.

Towards a typology of modality in language1 Paul Larreya

The typology presented in this paper is based on a distinction between modality and modalization. The latter will be defined as the use speakers make of modality, depending on (i) the type of knowledge they have, or do not have, concerning the situation which is submitted to the modal judgment, and (ii) on the type of knowledge the hearer is assumed to have, or not to have, concerning that same situation – henceforth referred to as the modalized situation. (In this paper, the word situation will be used to designate the referent of a proposition – consisting of a subject and a predicate – irrespective of whether this referent is a state, a change of state or an action.) The English modal verb forms (i.e. the modal auxiliaries and modal phrases like HAVE TO, BE ABLE TO, etc.) will be the main source of illustration.

1. Types of modality I will define modality as a mental system – or sub-system – based on the mutually related concepts possibility and necessity. This is obviously a “narrow” definition, which excludes what is sometimes called sentence modality (notably assertion and interrogation), but not negation, which is part of the relation between possibility and necessity. It also excludes irrealis, which may be associated with modality, but always remains distinct from it (see Larreya 2003), and such categories as evidential boulomaic or optative modality, which could and perhaps should be included in a wider definition. As has been known since at least Aristotle’s de Interpretatione, possibility and necessity – whether they be logical, physical or moral – are related by a double negation (It is possible that P = It is not necessary that non-P, and It is necessary that P = It is not possible that non-P). They constitute the core of modality, but by no means its entirety: there are a range of modal meanings which are situated at the periphery of what may be called the modal square (necessity / non-necessity / impossibility /

10 Paul Larreya possibility). The modal square can explain, for instance, the semantic resemblance between You can’t NOT go! and You’ve got to go!, but it does not make it possible to account for the nuance of meaning which separates the two forms. Or again – and still more importantly – although the core modal values are sufficient to explain why She can swim and It’s not impossible for her to swim are semantically related (as are I can’t lift this suitcase and It’s impossible for me to lift this suitcase), they cannot account for the fact that these forms are far from synonymous. Figure 1 is a schematic representation of the typology of modality proper (as distinct from modalization) proposed in sections 1.1-1.3 of this paper. The terms that appear in the tree diagram will be defined in these sections. Concerning problematic modality and implicative modality, however, a brief commentary may be useful at this stage: both He must be tired (used to express certainty) and You have to be mad to do that have to do with the attribution of some kind of truth-value to a proposition through some type of inference, but the inference is “stronger” in the latter case than in the former. Modality

Root modality

Physical modality He had to stop – he was exhausted.

Deontic modality You must stop.

Figure 1. Types of modality

Epistemic modality

Problematic modality

Implicative modality

He must be tired.

You have to be mad to do that.

A typology of modality

11

1.1. Root modality and epistemic modality The “root” v. “epistemic” distinction is part of a division which extends far beyond modality, or even language. The two categories belong to two different domains of human mental activity: the domain of affect and/or action and the domain of knowledge. The affect/action v. knowledge distinction corresponds to an important division in the domain of psychology (even though psychologists consider cognition never to be independent of affect): as is well known, cognitive and perceptual psychology constitute a field of research which is generally distinct from clinical psychology, social psychology, etc. In language, the affect/action v. knowledge opposition can be observed in several areas of morphology and syntax. In English, for example, it can be observed in the syntax of constructions which follow certain speech-act verbs (see Traugott 1989: 43–47; Sweetser 1990: 69–73; Larreya 2001b: 115–121), and, in Spanish, in the opposition between the verbs pedir and preguntar: the former denotes a “root” speech-act (you use it to ask the hearer to do something), while the latter denotes an epistemic speech-act (you use it to ask for some information from the hearer). It can also be observed in the semantico-syntax of such English conjunctions as because or so (see Sweetser 1990: 76–86). The case of because deserves to be discussed here, as it bears a particular relation to modality. Deléchelle (1989: 412–430) has shown the existence of a gradient between the two main uses of because.2 At the two ends of the gradient are what Deléchelle calls “explicative because” (as in He’s ill because he ate too much last night) and “justificative because” (He’s ill, because he didn’t turn up for work this morning). Deléchelle claims that (i) while the former is simply based on the expression of a “p because q” relation, the latter implicitly contains a proposition approximately paraphrasable as I say/think/conclude that…[p, because q]; and (ii) in some cases, the implicit proposition is replaced (or made explicit) by an adverb like apparently or perhaps, or by an epistemic modal and/or the phrase I think, as in: (1)

I think I must’ve hit my head, because I was like out on my feet. (quoted by Deléchelle, 1989: 419)

Concerning the relation between root and epistemic meaning (both within and outside the domain of modality), I will argue in favour of the same type of analysis as that proposed by Deléchelle concerning the relation between

12 Paul Larreya the two uses of because, and try to show that the relation is metonymic (not metaphoric) in nature. As a preliminary, a brief discussion of metonymy will be necessary. What follows only concerns metonymy in language, although, as is well known, metonymy is not limited to language. (An example of a conventionalized non-linguistic metonymy is a horseshoe sign used to indicate a bridle path.) The use I will make of the concept will rely on the assumption that, in language, any metonymy is made up of two interrelated components: a purely mental component, which is based on some sort of relation (for example the relation between a part and the whole, as when we say the strings instead of the stringed instruments), and a linguistic component, based on a more or less complex elliptical process. (A simple example is England[’s football team] won the match.) Let us now revert to the relation between root and epistemic modality – or rather between root and problematic modality. (As we shall see in section 1.3, the relation between root and implicative modality is different, and more complex.) My claim is that, as a general rule, problematic modality is derived from root modality through a metonymic process whose elliptical component consists of a particular type of proposition roughly equivalent to “I think that…”. (In a few cases, however, this movement is reversed: it is root meaning that is metonymically derived from epistemic meaning. For a discussion of this type of semantic mutation in expect/be expected to, be supposed to and be sure to, see Larreya 2001: 119–121.) Thus, assuming that the root meaning of the ambiguous sentence He must eat a lot of bread is, roughly speaking, “something requires that he eat a lot of bread”, I suggest that the (derived) epistemic-problematic meaning is “something requires that I think that he eats a lot of bread”. (In paraphrases, the verb require provides a convenient substitute for the modal predicate which underlies root necessity or possibility, and whose essential component may be assumed to be CAUSATION +/–VOLITION. Thus, Eventually he had to stop could be analysed as “eventually something/somebody [wanted and] caused him to stop”. On this hypothesis, see Larreya 1984: 73–100, and below, 1.4.1.) Similarly, He can/may eat a lot of bread can be glossed as “nothing requires that [I think that] he (does) not eat a lot of bread”. In other words, what constitutes the difference between root and epistemic-problematic meaning is the presence or absence in the semantic structure of a proposition approximately paraphrasable as I (or X) think(s) that… (or, in some cases, as I believe /say that…), which I will call a speech/thought-act

A typology of modality

13

proposition. This proposition can be considered “floating”: its presence or absence depends on the context, and is unclear in the (relatively rare) cases of indeterminacy between root and epistemic modality (e.g. in He left ten minutes ago – he could be home by now).

1.2. Physical and deontic modality As a first approximation (a more precise definition will be given in 1.4.1), physical and deontic modality, the two types of root modality, can be defined as, respectively, physical constraint/possibility and moral constraint/possibility – or their negations. The frontier between the two categories is not always clear. The sentence He had to abandon his project, for example, may be a case of indeterminacy or merger between the expression of physical obligation (…because he was exhausted) and that of moral obligation (…because he was duty-bound to renounce it).

1.3. Problematic and implicative modality In what follows, I will be using two terms – truth-value and implication – which may be considered to belong primarily to formal logic, but to which I will give purely linguistic definitions. The concept of truth-value will not be the binary concept of formal logic: it will apply not only to TRUE and FALSE but also to what may be termed intermediate values such as PROBABLE and POSSIBLE, and will be used when referring not only to propositions but also to their referents (i.e. situations). As to implication, it will be defined as a relation which can be paraphrased as if … then …, and which is established between two propositions (or between the situations they refer to), so that it will have little in common with the relation of implication of formal logic and with the truth-table which serves to define it. For the sake of convenience, however, the sign ⇒ will be used to represent linguistic implication. Epistemic modality (whether problematic or implicative) consists in the attribution of a truth-value to a proposition (or to the situation that constitutes the referent of that proposition). In the case of implicative modality (e.g. in You have to be mad to do that), the truth-value is one of the two extreme values (TRUE / FALSE). In the case of problematic modality (of which You must be tired and It may/might rain tomorrow are

14 Paul Larreya examples),3 it is an intermediate (or “weak”) value, such as PROBABLE or POSSIBLE. Implicative modality consists in linking two propositions (whose referents are two situations or sets of situations) by a relation of implication. The relation can be represented by A ⇒ B (A implies B), A being the antecedent of the implication and B the consequent. Thus, in You have to be mad to do that the antecedent is X does that and the consequent is X is mad: the meaning can be glossed as “you do that” IMPLY “you are mad” (or “being mad is a necessary condition for doing that”). At this point, it should be emphasized that implication is a “strong” relation. Indeed, as we have just seen, implicative modality involves absolute truth-values (TRUE and FALSE), unlike problematic modality, which is concerned with “weak” truth-values. Thus, You have to be mad to do that can be paraphrased as “if you do that you are mad” – not as “if you do that you are probably mad”. This strong nature of the truth-values involved is the main characteristic of implicative modality, clearly differentiating it from problematic modality. There are two types of implicative modality, which I will call explicit and elliptical. Explicit implication (of which You have to be mad to do that is an example) has two characteristics. Firstly, it explicitly mentions the antecedent of the implication in a clause or phrase which is syntactically linked to the modal form. Secondly, it is, in most cases, abductive (i.e. reverses the cause-consequence relation):4 it infers the cause (X is mad) from the consequence (X does that), so that, somewhat paradoxically, it is the consequence (not the cause) which is the antecedent of the relation (X does that ⇒ X is mad).5 Elliptical implication, whose main markers in the English modal system are WILL and SHALL, is characterized by two facts: (a) the antecedent of the relation is not mentioned explicitly (often because it is not at issue for the speaker) and (b) the relation is rarely abductive (i.e. rarely reverses the cause-consequence relation). The first characteristic makes it possible to symbolize elliptical implication as (A) ⇒ B, where the parentheses represent the fact that the antecedent remains unexpressed. In I’m sure he’ll win the match or in The phone’s ringing – that’ll be John, the modal judgment consists in inferring a consequence from a set of “known” facts (which are not specified). In Accidents will happen or in He’ll sit there for hours looking at the walls (where the modal meaning is “characteristic behaviour”), the unexpressed antecedent is an intrinsic characteristic of the

A typology of modality

15

referent of the syntactic subject. (Accidents will happen, for instance, can be glossed as something like “the nature of accidents is such that they necessarily happen”.) When the modal has a volitional meaning, the antecedent is a proposition whose predicate is volitive: roughly speaking, John won’t answer my questions (meaning John refuses to answer my questions) can be glossed as “John’s not wanting to answer my questions has as its consequence that he does not answer my questions.” (This case will be examined more extensively in sections 1.4.1-1.4.2).6 Unlike problematic modality, which in most cases is clearly distinct from root modality, implicative modality often overlaps with deontic modality. An example is In Oklahoma you have to be 18 to marry without parental consent. We can now examine the relations between implicative modality and the other types. These relations are rather complex – much more so than the often-studied relations between root modality and what I have called problematic modality. Although conceptually akin to problematic modality, implicative modality is very different from it in several respects. Furthermore, explicit and elliptical implication bear different relations to the other types of modality.

1.3.1. Explicit implication As we have seen, explicit implication is, as a general rule, abductive (it reverses causal relations). In: (2)

You have to be mad to do that.

(which can be glossed as X does that ⇒ X is mad), the proposition X does that is the antecedent (not the consequent) of the relation. As is well known, the consequent of a relation of implication is a necessary condition for the antecedent, and the antecedent is a sufficient condition for the consequent. So, an implicative relation like that of (2) can be considered from two points of view. If the focus is set on the necessary condition and the consequent, the relation may be glossed as “being mad is a necessary condition for doing that”. But the relation can also be considered as having the antecedent (X does that) as its point of origin. This, in fact, is what the speaker normally does: the basis for his or her judgment is the fact “X does that”. Now, the sufficient condition that corresponds to X does that ⇒ X is mad may make sense in formal logic, but (at least apparently) it does not in

16 Paul Larreya language: “X does that” is a sufficient condition for “X is mad” does not mean anything. This gloss, however, does make sense if a speech/thoughtact proposition (see above, section 1.1), which in this case would be I believe/say that…, is inserted: “X doing that is a sufficient condition for me to believe that X is mad”. The presence in the semantic structure of an ellipted speech/thought-act proposition is therefore a common feature of problematic and implicativeelliptical modality. Another common feature is the fact that problematic modality also uses as its basis – as a sort of antecedent – a set of known facts. Problematic modality, however, differs from implicative-elliptical modality as regards the nature of the relation established between the basis (or antecedent) and the conclusion (or consequent). As noted before, the relation is a weak one in the case of problematic modality, so that the truthvalues involved are also weak (they are, for instance, “probable” or “possible”), whereas in the case of implicative modality (whether explicit or elliptical, in fact) the relation is strong (or absolute), and the truth-values involved are “true” or “false”.

1.3.2. Elliptical implication The conceptual links between implicative-elliptical and deontic modality will be examined in section 1.4.3, through a discussion of the grammaticalization of WILL and SHALL. In the present section, I will merely (and perhaps more superficially) investigate these links through some of the uses of WILL. As it does not explicitly mention the antecedent on which the implication is based, elliptical implication is more similar to problematic modality than is explicit implication. One form of elliptical implication, which may be called strong conjecture (as in The phone’s ringing – that’ll be John), is indeed very similar to the type of problematic modality expressed by must (which does not explicitly mention the facts on which the modal judgment is based). There are, however, important semantic differences between conjectural will and epistemic must (see for instance Palmer 1990: 57–58), which may have something to do with the fact that the conjectural use of will is related to both its “habitual” use (in other words to the expression of characteristic behaviour) and to its “futural” use. (On this double relation, see Coates 1983: 178.)

A typology of modality

17

The “habitual” use of will is, as we have seen, fundamentally implicative, but it is marginally related to material possibility and perhaps to weak volition (see below,1.4.2): if, speaking of a house, someone says It will accommodate five people, they mean that it can accommodate five people, and perhaps that it is, so to speak, “willing” to accommodate them. The “futural” uses of WILL fall into two categories, between which there is much overlap and indeterminacy: predictive (as in Leeds will win the match) and volitive (as in OK, I’ll do the dishes in contexts where an essential part of the meaning is “I am willing to / I agree to do the dishes”). Predictive meaning is purely epistemic. Volitive meaning is based on a strong (implicative) relation whose antecedent is the referent of the syntactic subject’s volition. It is, therefore, akin to deontic modality. (This point will be investigated more thoroughly in section 1.4.4.)

1.4. Volition and modality German provides an argument in favour of regarding volition as part of modality: its modal system (which is idiosyncratic, and therefore may be thought to be the formal counterpart of a semantic system) includes a verb (WOLLEN) which fundamentally expresses volition. The case of German, however, is far from being general. (In English, for instance, the modal WILL, although etymologically akin to WOLLEN, is only marginally a marker of volition.) If the question is to be addressed from a purely semantic point of view, it seems difficult to regard volition as one of the prime constituents of modality; or, at any rate, it seems difficult to place it at the same level as possibility and necessity. It nevertheless plays an important role in modality, on several counts.

1.4.1. Volition and root modality Volition is present at two levels of the meaning of a deontic utterance like (3)

John must leave.

First, the concept of obligation expressed by the modal includes some sort of volition: part of the meaning of (3) is the fact that some practical or moral principle, and perhaps somebody, require (or want) John to leave. (For a hypothesis on the role of volition, alongside causation, in deontic

18 Paul Larreya necessity, see above, 1.1.) Second, the use of must (and, relatedly, the fact that “leaving” is a voluntary act, unlike, e.g., “sneezing”) carries with it the implication that, if John obeys the injunction contained in (3) and actually leaves, he will have decided to leave (although perhaps reluctantly). This semantic element, in my opinion, is part of the meaning of the utterance – and not an implicature of it. However, Gordon and Lakoff (1975: 85–87) analyse it, or something very similar to it, as one of the three “hearer-based sincerity conditions” of requests, and therefore situate it at the same level as another semantic element which I think is an implicature of such an utterance as (3), and which has to do with volition: if left to himself/herself, the person who is the object of the directive speech-act would probably not carry out the requested action. So, there are two volitional elements implicitly present in the meaning of such a form as (3). I will call the first external volition, and the second internal volition. (The use of the adjectives “external” and “internal” justifies itself if one considers the place and role of the volitional elements in relation to “John”.) Now, consider: (4)

John had to stop: a huge snowdrift was blocking the road.

in which the modal meaning of had to is “physical necessity”. As regards internal volition, there is no fundamental difference with (3): John had to stop implies that, although John did not stop of his own free will, his act was in some way intentional. As to external volition, it does not seem to be part of the semantic make-up of (4), and, in this context, John had to stop may be glossed approximately as “X caused John to stop”. In the domain of possibility, there is a similar difference between John can go in – the boss gave him permission, which implicitly refers to the boss’s volition (or, more precisely, to his/her willingness), and John can go in – he’s got the key / he’s clever enough, in which can denotes an ability which is not directly linked to any external volition. So, if a general conclusion can be drawn from the preceding cases, it is the presence or absence of external volition which differentiates deontic from physical modality.

A typology of modality

19

1.4.2. Subjectal (or simple) volition An example of this is (5)

OK, I’ll tell you.

The main difference between subjectal volition and external volition – exemplified by (3) – can be described as follows: in the case of external volition, the “subject” and the “object” of the volition are different, while in the case of subjectal volition they are identical. Subjectal volition also differs from internal volition (defined in 1.4.1): it is more explicit, and more central in the meaning of the modal form – although, as we shall see, it is only part of that meaning. A further distinction can be made within the domain of volition. (i)

Strong volition (as in I WILL stay here) is associated with the implicature that some physical obstacle, or some external volition, might prevent the accomplishment of the “willed” situation. (In the case of I WILL stay here, this implicature may be made explicit by the speaker adding something like …whether you like it or not.)

(ii) Weak volition (or willingness) is also associated with an implicature: that of the existence of some external volition directed towards the accomplishment of the modalized situation. Thus, OK, I’ll do the dishes could be said as a response to a request made by the hearer. As is well known, there is between weak volition and strong volition the same relation (a double negation) as between possibility and necessity: I am willing to stay is logically equivalent to I have no desire not to stay. Or again, You may stay is roughly equivalent to I am willing for you to stay, and You must stay to I (or something) require(s) you to stay. (iii) Indefinite volition carries no implicature of the existence of any external volition, either in favour or against the accomplishment of the modalized situation. (However, it is notionally nearer to strong than to weak volition, as there is no reason to think that it contains a double negation.) An example is I think I’ll have one more small whisky [...]. (BNC, H8M 870)

20 Paul Larreya 1.4.3. Factual and non-factual volition Compare: (6a) (6b)

I want to go. I will go.

The volition expressed in (5a) can be said to be non-factual: the sentence does not mean that the act of going will be accomplished. (It is possible to add: …but I may not be able to.) In (5b), the volition is of a different nature. First, it is only part of the meaning of will, the form which expresses it. (This question will be discussed in 1.4.4.) Second, it can be said to be factual: the sentence presents as real (as true) the future accomplishment of the act of going. (This use of the words factual and non-factual is of course a terminological hypallage: the factuality or non-factuality concerns the situation that is “willed” – not the volition itself.) Unlike in the case of (5a), it is not possible to add …but I may not be able to. Several other facts argue in favour of the factuality of WILL/SHALL used as markers of futurity. Among these is their compatibility with such epistemic adverbs as maybe or possibly: compare The horse is in excellent form – (maybe) he’ll win the race and The horse is in excellent form – (*maybe) he must win the race. (Non-factual epistemic verb phrases cannot be “weakened” by epistemic adverbs: they are already “weak”.) What is to be taken into account to determine the factuality or non-factuality of a given form is of course linguistic reality (i.e. what the speaker says), and not psychological reality (what the speaker thinks) or physical reality (what happens/happened/will happen in real fact). Factuality, indeed, is a necessary component of futurity (whether “volitive” or “predictive”), as distinct from future-time reference; for instance, may is not usually considered a marker of futurity when used to refer to a future situation, as in It may rain tomorrow. Conversely, “pure” volition is not, per se, factual. Such verbs as want (mentioned above), wish and lexical will are obviously not factual. Neither is the German modal verb WOLLEN, in spite of its common origin with the English modal auxiliary WILL. A further remark needs to be made about factuality. In the modal system of English – and probably of any natural language – a distinction may be made between two types of factuality: on the one hand a priori factuality, as in (5b) or in If you disconnect the battery the car won’t start, and on the other hand a posteriori factuality, as in I’ll have to take the bus this

A typology of modality

21

morning as the car won’t start or in He’ll stay there for hours doing nothing.7 The former generally carries less credibility than the latter, and, for this reason, may sound “weaker”.

1.4.4. Volition and implicative modality The semantic and morphosyntactic mutation that led from the Old-English verb WILLAN to the Modern-English modal auxiliary WILL (in other words, the grammaticalization of WILL) provides a good illustration of the relation between subjectal volition and implicative modality. The most important change that WILLAN underwent in early Old English was its gradual shift towards factuality. This was particularly clear when the verb was used in the present indicative with a first-person subject. Tellier (1962: 64–69) shows for instance how, in Beowulf, the form ic wille/wylle can express promise and/or some sort of immediate future (the semantic shift being from “I want to…” to “I’m going to…”), as in Nu ic, Beowulf, þec, / secg betsta, me for sunu wylle / freogan on ferhþe (“Now, Beowulf, best of men, I say to you, I will cherish you in my heart like my own son” – Beowulf, 946–948). Tellier also points to another important development in the semantic evolution of WILLAN: while there are in Beowulf only a few instances of the verb being used to denote what may be termed characteristic behaviour (cf. what Palmer, 1990: 136-137, calls “power” and “typical activity”), Alfredian prose is rich in examples of this type of use, in which the WILLAN + infinitive construction is obviously factual. One of his examples is se hunde wile aspiwan þone mete þe hine hefegaþ on his breostum (“the dog vomits the food that is heavy on its breast”).8 In fact, almost all of the semantic values of present-day English WILL – some of which seem to have little in common with the original meaning of volition – were already present in its ancestor WILLAN before the end of the Old English period. The main question posed by the grammaticalization of WILLAN/WILL is then the following: how did the original meaning of volition which characterized WILLAN come to evolve into such a wide variety of meanings? Any volition is necessarily directed at some situation (or state of affairs), and implies the existence of three elements: the volition itself (symbolized by V in Figure 2), the “willed” situation (S in Figure 2), and the relation between the volition and the “willed” situation (represented in

22 Paul Larreya Figure 2 by various types of arrow). This relation may be either nonfactual, as in I want to go, or factual, as in I will go. (In Figure 2, a dotted arrow represents a non-factual relation, and other types of arrow represent factual relations.) The main stages in the grammaticalization of WILLAN/WILL can be schematically represented as in Figure 2 (comments on schemas below): (i) (ii) (iii) (iv)

V …..> V ––> V/X ––> (X) ⇒

S S S S

Figure 2. From WILLAN to WILL

The first two stages – (i) and (ii) – simply correspond to the shift from nonfactual to factual volition. At these stages, the main focus of the meaning (represented in Figure 2 by bold type) is still on the volition proper. This, however, changed in proportion as the meaning of WILLAN/WILL + verb extended itself: the main focus moved from the first element of the relation to the second and third elements (i.e. from the origin of the relation to the relation proper and to the resulting situation). The third stage – (iii) – consists in the metonymic extension of the first element of the relation. The relation proper remains basically the same (it is a “factual” relation of the cause-to-effect type), but its origin (the “cause”) is no longer necessarily the volition of the referent of the subject: it may be, for instance, a set of his/her/its inherent characteristics (in the case of WILLAN/WILL expressing characteristic behaviour), or a set of known facts and circumstances whose consequence will be the accomplishment of the modalized situation (in the case of the expression of a prediction). The last stage – (iv) in Figure 2 – corresponds to the basic meaning of present-day English WILL, and is not very different from stage three. It is characterized primarily by a lesser emphasis on the origin of the relation (which origin may remain rather vague, and, as in stage three, need not be “volitional”), and by a greater emphasis on the relation itself, which is more abstract, and simply consists of an implicative link whose antecedent is unexpressed (hence the parentheses around X in the last line of Figure 2). The counterpart of WILLAN/WILL in the field of extra-subjectal (i.e. external) volition is obviously *SCULAN/SHALL. As is well known, what

A typology of modality

23

*SCULAN fundamentally expressed was obligation (even though its original meaning, which extended itself metonymically, was approximately that of present-day English owe), in other words some form of external volition. Its semantic evolution was somewhat parallel to that of WILLAN: very early in the history of Old English, it specialized in the expression of factual obligation. This of course only concerned present indicative forms, and was particularly clear in uses with a second-person subject; among other examples, Tellier (1962: 72) quotes Nu þu, Andreas, scealt eþre geneþan / in gramra gripe (“Now, Andreas, you shall soon venture into the grasp of foes” – Andreas, 950–951). With a first-person subject, and if the verb that followed denoted a “voluntary action”, *SCULAN could express selfimposed obligation – or, in other words, could express a promise, as in Ic þe sceal mine gelæstan / freode, swa wit furðum spræcon (“I shall fulfill my friendship to you, as we have agreed”– Beowulf, 1706–1707).9 From this stage on, the process of grammaticalization undergone by *SCULAN was basically the same as that undergone by WILLAN (even though it gave rise to a narrower range of meanings), and can be represented in the same way as in Figure 2, if the symbol V (subjectal volition) is replaced by a symbol standing for obligation or extra-subjectal volition. The end result (stage four in the figure) was nearly the same as in the case of WILLAN/WILL: *SCULAN/SHALL came to fundamentally express necessary consequence. The difference which, beyond this common characteristic, separates WILL from SHALL in present-day English is obviously linked to their respective origins. It can be described in terms of markedness v. unmarkedness. While WILL unmarkedly expresses necessity (or, more precisely, necessary consequence), SHALL is able to specifically express what may be called subjectively oriented necessity. In other words, SHALL may express a form of necessity which more or less involves the speaker/hearer, and/or may be chosen in preference to WILL if the speaker wants to give more formality to his/her utterance. This description of the semantic difference between SHALL and WILL is based on the following hypothesis: the modals MAY, MUST, NEED and SHALL are subjectively oriented (in contrast to CAN, DARE, OUGHT and WILL, which are neutrally oriented); they are able to (but do not necessarily) express a modal judgment which presents itself as reflecting the will or opinion of the speaker (or of some authority on whose behalf s/he speaks, or, in interrogative utterances, of the hearer), and/or, in some contexts, they are perceived as more formal than their neutrally oriented counterparts, which

24 Paul Larreya are essentially CAN, HAVE TO, NEED TO, WILL and OUGHT. (The latter characteristic, which is socially motivated, is linked to the former.)

2. Types of modalization The way of using modality (be it root or epistemic), and in most cases the choice of modal forms, depend largely on whether what is submitted to the modal judgment is (i) a situation whose existence is a well-established fact, or (ii) a situation whose existence cannot be ascertained – either because it is situated in the future (as in You must see that film!) or because it is outside the scope of direct knowledge (as in He must be at home at the moment). This is the domain of modalization, which can be defined as the way in which modality is used in utterances, depending on (i) the state of knowledge of the speaker concerning the modalized situation and (ii) the assumed state of knowledge of the hearer concerning that same situation. Table 1 represents the two fundamental types of modalization (a priori and a posteriori) and their subdivisions. Table 1. Types of modalization A priori modalization A posteriori modalization Simple Qualified Constative Evaluative Counterfactual modalization modalization modalization modalization modalization You could ask He must He should have You can ask It’s a good John. John. leave thing he told them. cigarette should have some rest. You must ask You should butts John. ask John. everywhere! He must be He should be He’ll sit there It’s not He should be here there by now. there by now. for hours. surprising he by now [and he’s should have not]. left. (First row of examples: modalization of root modality. Second row: modalization of epistemic modality.)

The first division is between a priori modalization and a posteriori modalization. In the case of a priori modalization, the speaker does not

A typology of modality

25

know, or pretends not to know, the exact truth-value of the modalized situation. In the case of a posteriori modalization, the speaker knows, or pretends to know, that truth-value. (However, we shall see that in the case of counterfactual modalization the object of the modalization is not the “known” situation itself.) A priori modalization subdivides itself into simple modalization (as in It may rain tomorrow) and qualified modalization (as in It might rain tomorrow). In English, qualified modalization is characterized by the use of the “past” tense, which adds a presupposition of doubt to the modal judgment. (On the non-temporal uses of the “past” tense, see Larreya 2003.) Constative modalization (unlike evaluative and counterfactual modalizations) establishes a close link between the modality and the modalized situation. In most cases, the utterance serves to inform the hearer of the existence of the modalized situation (or, in other words, it asserts that existence). In some cases, however, it presupposes the existence of the situation (an example is Since you will insist on calling him “my dear friend”, I’m surprised you don’t invite him over more often – BBC, Radio 4, 21 August 2004), and the modal necessarily bears nuclear stress. Here, modalization is effected through a mental movement which, starting from the situation, goes back to its causes or circumstances, so as to somehow explain its existence – even though the main focus of interest may seem to lie more on the situation itself than on its origin. In contrast, evaluative modalization is totally exterior to the situation. It presupposes the existence of the situation, and the modal form simply expresses the speaker’s opinion of it. The judgment may be deontic in nature (the situation being judged either good or bad – with various nuances of “goodness” or “badness”) or epistemic (the situation being judged either logical or illogical – here again in various nuances, expressed by such words as normal, surprising, etc.), with cases of indeterminacy between the two types of modality, as in (e.g.) It’s normal he should have some rest. In fact, it is in the superordinate clause – e.g. in It’s a good thing/It’s surprising that… – that the modal judgment is primarily expressed. The modal should expresses it redundantly in the subordinate clause, as the result of a sort of metonymic transfer which is somewhat similar to the phenomenon traditionally called tense concord. Counterfactual modalization is a type of a posteriori modalization in which the “known” situation is not directly the object of the modal judgment. In You should have told him, for instance, the “known” (or

26 Paul Larreya “real”) situation is “You did not tell him.” (The existence of this situation is presupposed; it is of course known to the speaker, but not necessarily to the hearer. On cases in which the presupposition corresponds to “new” information, see Larreya and Watbled 1994: 71–74.) The object of the modal judgment is a “theoretical” situation (“You told him”) which is an inverted image of the “real” (presupposed) situation: the truth-value of the “theoretical” situation is the contrary of that of the “real” situation. (You should have told him can be glossed as “There was a moral obligation for you to tell him – but you did not tell him.”) There are many cases of indeterminacy between counterfactual (a posteriori) modalization and qualified (a priori) modalization. An example is You should get more exercise, which is counterfactual in so far as it concerns the present period of time (it presupposes “You are not getting enough exercise”) and tentative in so far as it concerns the future.

3. Concluding remarks I have attempted to show that it is appropriate to make a distinction between modality proper and what I call modalization. Modality, however, is what poses the most difficult problems. If possibility and necessity are the two basic concepts on which it rests, their definition – at least as far as language is concerned – necessarily involves volition. Modalization cuts across types of modality. Within each type of modality, it influences the choice of the modal form. (For instance, counterfactual obligation cannot, in present-day English, be expressed by MUST. Constative obligation is very often expressed by HAVE TO – and rarely by MUST.) Modalization also has an effect on the syntax of modals. Thus, it is not fortuitous that (with a few rare exceptions, which involve must and might) the only modals whose past tenses can be used freely to express a narrative past are the two “objectively oriented” modals CAN and WILL used in constative contexts; this concerns two types of use of CAN – “physical or moral possibility” and “occasional characteristic or behaviour”, exemplified in (7) and (8) below – and three uses of WILL – “isochronal volition”, “habitual characteristic or behaviour” and “insistence”, exemplified in (9), (10) and (11): (7) (8) (9)

She could swim at the age of two. He could be very sarcastic. He was mad at her because she wouldn’t answer his questions.

A typology of modality

27

(10) She’d take the bus every morning at 8.30. (11) He WOULD make that remark.

Notes 1.

2.

3.

4.

5.

6.

This paper has greatly benefited from the thoughtful comments made by Johan van der Auwera, Pierre Busuttil, Raphael Salkie and an anonymous reviewer on an earlier version. I wish to express my gratitude to them. Any remaining errors and shortcomings are of course my own. According to Sweetser (1990: 76–77), there is a third type of use of conjunctions such as therefore, although or because in which the utterance is “the instrument of a speech act”. (Her example for because is What are you doing tonight, because there’s a good movie on.) This, in my opinion, is simply a variant of the epistemic use of the conjunctions – the only difference being that the verb of the ellipted proposition is not a verb of mental attitude but a verb of saying (I ask you ... in Sweetser’s example). For a more detailed discussion of Sweetser’s analysis of because and of the relation between root and epistemic modality, see Larreya 2001b: 110–114. What I call problematic modality is what is traditionally considered the core, or perhaps the whole, of epistemic modality. As regards the “futurity” uses of will or shall, there is wide variation in the literature: for some scholars, they are epistemic, and for others futurity constitutes a category of its own. What I will try to show is that these uses are part of the domain of implicative modality. Logicians generally insist that the cause-consequence relation and implication are two different concepts, which must not be confused. This, however, should not deter linguists from using causality and implication as linguistic concepts. The implicative relation of language, in particular, is in several respects very different from the implication of formal logic, and the causal relation is one of the forms that it can take. The main markers of explicit implication are HAVE TO, the negation of NEED and the negation of CAN (as in You can’t have your cake and eat it – which merges implicative and root modality) Concerning the English modal WILL, the claim made here is that, in all of its uses, it expresses a relation of implication whose antecedent is unexpressed. Among the various arguments in favour of this claim (developed in Larreya 1984 and 2001a), I will mention two: (i) the WILL v. BE GOING TO opposition [as in Palmer’s (1974: 164) famous example Don’t sit on that rock. It’ll fall/It’s going to fall] clearly shows the implicative nature of WILL, which contrasts with the non-implicative nature of BE GOING TO (which cannot, for

28 Paul Larreya

7. 8.

9.

instance, express characteristic behaviour); (ii) it is significant that the use of WILL (or of SHALL) as a marker of futurity is impossible (or at least restricted) in any subordinate clause that is semantically linked to its superordinate clause by a relation of implication and constitutes the antecedent of that relation, as in I’ll do what I (*will) like, The more you (*will) drink the thirstier you’ll be, etc. The a priori v. a posteriori opposition will be discussed in section 2. Tellier 1962: 107; the quotation is from Alfred’s translation of Cura Pastoralis. An even clearer example is the following (dated c. 1225), quoted by Visser (1969: 1682): Hund will in at open dure [“A dog will enter (the house) if the door is open”]. Ic sceal, however, did not necessarily express a promise. It could also express, for instance, “constative” (and therefore factual) obligation, as in ælce dæg ic sceal erian fulne æcer oþþe mare (“each day I have to plough one full acre or more” – Ælfric’s Colloquy). For a discussion of “promissory I shall”, “promissory I will” and their relation to obligation and volition, see (for instance) Visser 1969: 1603–1605.

References Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Deléchelle, Gérard 1989 L'expression de la cause en anglais contemporain – étude de quelques connecteurs et opérations. Thèse de doctorat d’État (Paris III). Gordon, David and George Lakoff 1975 Conversational postulates. In Syntax and Semantics, Vol. 3, Speech Acts, Peter Cole and Jerry L. Morgan (eds.), 83–106. New York, San Francisco, London: Academic Press. Larreya, Paul 1984 Le Possible et le Nécessaire: Modalités et Auxiliaires Modaux en Anglais Britannique. Paris: Nathan. 2001a Modal verbs and the expression of futurity in English, French and Italian. Belgian Journal of Linguistics 14: 115–129. 2001b La grammaticalisation du non–dit: le cas de la modalité épistémique. In Mélanges en l’Honneur de Gérard Deléchelle, Fabienne Toupin (ed.), 107–127. Tours: GRAAT.

A typology of modality 2003

29

Irrealis, past time reference and modality. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank R. Palmer (eds.), 21–45. Berlin, New York: Mouton de Gruyter. Larreya, Paul, and Jean–Philippe Watbled 1994 Linguistique Générale et Langue Anglaise. Paris: Nathan. Palmer, Frank R. 1974 The English verb. London: Longman. 1990 Modality and the English Modals. 2nd ed. London: Longman. Original edition, 1979. Sweetser, Eve 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Tellier, André 1962 Les Verbes Perfecto–Présents et les Auxiliaires de Mode en Anglais Ancien (VIIIe–XVIe s.). Paris: Klincksiek. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. Visser, Fredericus Th. 1969 An Historical Syntax of the English Language (Part Three). Leiden: E.J. Brill.

‘Not-yet-factual at time t’: a neglected modal concept Renaat Declerck

1. Introduction1 Some hypothetical ‘possible worlds’2 can be characterized in terms of a modal concept which has been neglected in the literature on modality, viz. the idea that the world in question is envisaged by the speaker but not yet factual at the time t to which it is anchored. This kind of hypothetical world, which we will refer to as ‘not-yet-factual at t’ is evoked by any expression that has posteriority as part of its meaning. The clearest cases are those in which the reference is to a future world. Thus, John will take the exam tomorrow evokes a world which is subjective in the specific sense that it is not-yet-factual at S (= speech time). (It follows that the situation that is temporally located in that future world by the future tense is also not-yet-factual at S.)3 Apart from drawing attention to the existence of not-yet-factual worlds (which are hypothetical worlds and should therefore be treated in any comprehensive discussion of modality), this article aims to show that the value ‘not-yet-factual at t’ pertains not only to situations that are located in the future (and situations that are represented as posterior to a past reference time), but also to the situations referred to in sentences like the following: (1a) [It’s high time] we left. (The situation of our leaving is envisaged at the time of speech, but is not yet a fact at that time.) (1b) [I wish] Jim would stop lying. (The situation of Jim ceasing to lie is not yet a fact at speech time, but is envisaged at that time as being weakly possible in the future.) (1c) [I saw Sam] before she had seen me. (= ‘I saw Sam at a time when she had not yet seen me’, ‘When I saw Sam, it was not yet a fact that she had seen me’)

32 Renaat Declerck In sum, the main purpose of this article is to have a closer look at the factuality values that a situation can have in a possible world. We will argue that there are four: factual, counterfactual, hypothetical and not-yet-factual. As we will see, this has important consequences for the definition of modality: we will conclude that modality can apparently be identified with ‘nonfactuality’. The following are the main observations and claims on which the analysis rests: (a) We will use situation as a cover-term for anything that can be expressed by a verb phrase. According to Lyons (1977), a situation is either a state, an action, a process (= change, development) or an event (= a nonagentive dynamic situation, e.g. a fall). The term actualize will be used as a cover-term for all the verbs that are typically associated with a kind of situation. The sentence The situation is actualizing can thus be said of a state that is holding, an action that is being performed, a change that is taking place or an event that is happening. (b) A ‘possible world’ is always a t-world, i.e. a world which is anchored to a given time t. This means that it is possible for a tensed proposition to be true at one time but false at another, in other words, that it may be true of one t-world but false of another. Thus, J.F. Kennedy is the President of the U.S. is true of any world holding at (= anchored to) some time in the course of 1961 but is false of the ‘objective’ S-world, i.e. the actual world holding at speech time – see section (d) below. By contrast, omnitemporal situations (referred to by generic or universal sentences like A horse is an animal) are factual in every objective world holding at any possible time. The unmarked form of t-anchoring is S-anchoring (where S means ‘speech time’). If the world referred to is an S-world and the sentence referring to it is in the present tense, no anchor time needs to be specified in the sentence, nor in its context. This explains why The weather is nice is fully interpretable in isolation: The hearer assumes that the speaker presents the sentence as true at S. It also explains why omnitemporal situations are as a rule referred to in the present tense: This follows naturally from the fact that S is one of the times at which the situation is factual and is the unmarked anchor time.

Non-yet-factual at time t’

33

(c) There is no essential difference between saying that a tensed proposition is ‘true of’ (Lyons 1977: 687) a particular t-world and saying that the actualization of the situation represented by the tensed proposition is ‘factual in’ that world. Thus, if the tensed proposition John is walking home is true of (= true with reference to) the objective S-world (= the actual world holding at S), the actualization of the situation (i.e. the performance of the action) is factual in that world. If the untensed proposition ‘John be walking home’ is not true of the objective Sworld, as in [I wish] John was walking home, the actualization of the situation is not factual in the objective S-world, i.e. it does not belong to the actualizations making up the objective S-world. In that case John is walking home is true of a counterfactual S-world, which means that the actualization of the situation is factual in that counterfactual Sworld. (As will become clear from paragraphs (e) and (j), similar conclusions can be drawn in connection with sentences in the past tense or future tense.) (d) We need to distinguish between ‘objective’ and ‘subjective’ t-worlds. An objective t-world is the unique real world that holds at a given time and which is judged real by an (imaginary) ideal outside observer viewing the world as it is at that given time. A subjective (or intensional) t-world is an alternative world which is not judged real by such an ideal outside observer but which is conceived of as real by some consciousness at a certain time. Such a t-world consists of the tensed (= anchored in time by their verb form) propositions that are deemed true by the world-creating consciousness at the given time. (That time may or may not be S, i.e. speech time.) Thus, the situation referred to by Amsterdam lies in Belgium (which is counterfactual in the objective S-world) is factual in the counterfactual S-world existing in the mind of a speaker who is convinced that this assertion is true at S. (e) Another distinction we need to make is that between ‘narrow t-worlds’ and ‘extended t-worlds’. A narrow t-world is a t-world comprising all the situations that are actualizing at a given time t. Tensed propositions can only be true of such a world if the tense represents the (actualization of the) situation referred to as simultaneous with t. An extended t-world is a world comprising all the situations that are actualizing at t or have actualized before t. Various tenses can be used to represent the actualization of a situation as factual in an extended t-

34 Renaat Declerck world. Thus, Caesar was Emperor of Rome, There have been many battles between the French and the English and Paris is the capital of France all refer to actualizations that are factual in the speaker’s extended subjective S-world (as well as in the extended objective Sworld). Since an extended t1-world includes a narrow t1-world, any actualization that is factual in an objective narrow t-world is also factual in the corresponding objective extended t-world. Thus if John is in London now, the tensed proposition John is in London is true not only of the narrow S-world but also of the extended S-world. (f) Most t-worlds are either factual, hypothetical or counterfactual at t.4 (These three possibilities correspond to the truth values ‘true’, ‘possibly true’ and ‘false’.) A t-world of any of these three types comprises only actualizations that are factual (at the relevant time). Thus, the (actualization of the) situation referred to by John is ill, If only John was ill! and John is perhaps ill is each time represented as factual in a given S-world, but that S-world is, respectively, the objective S-world, a counterfactual S-world (= a subjective S-world which is conceived of by the speaker as being incompatible with the objective S-world) and a hypothetical S-world (which in this case is an open-possibility S-world conceived of by the speaker). If there exist several t1-worlds (= worlds anchored to the same time t1), the actualization of a given situation can only be factual in one of these t1-worlds. Thus, if it is indeed the case that John is ill at S, the sentence John is ill represents the (actualization of the) situation of John being ill as factual in the objective S-world. In If only John {was / were} ill!, the proposition ‘John be ill’ is represented as factual in the counterfactual S-world evoked by If only (= I wish…). And in John is perhaps ill, the proposition ‘John be ill’ is represented as factual in the hypothetical S-world evoked by the modal adverb perhaps (= It is possible that…). (g) It does not follow, however, that the only factuality value that an actualization can have in a particular t-world is the value [+ factual]. The actualization referred to in If only John was ill! is represented as having the value [+ counterfactual] in the objective S-world because it is conceived of as factual in a counterfactual S-world. So far we have distinguished three types of t-worlds (viz. factual, counterfactual and hypothetical), and a particular actualization can be factual in one type

Non-yet-factual at time t’

35

of t-world only. This means that (for the time being – see section 2) we distinguish between three mutually exclusive ‘factuality values’ which can apply to any t-world, viz. [+ factual], [+ counterfactual] and [+ hypothetical]. (The latter two share the feature [– factual].) Thus, the actualization referred to in If only John was ill! is counterfactual in the objective S-world, but factual in a counterfactual S-world. Similarly, John is perhaps ill refers to a hypothetical S-world (created by perhaps) in which the actualization of the situation of John being ill is factual. The factuality value of this actualization in the objective Sworld is [+ hypothetical]. (h) Whether a sentence is positive or negative is irrelevant to the question whether the actualization of the situation is factual or not in a given tworld, or to the question whether the sentence (= tensed proposition) is true or not of that t-world. Thus both London lies on the Thames and London does not lie on the Rhine represent the (actualization of the) situation referred to as factual in the objective S-world, and both sentences are true of the objective S-world. It follows that we must distinguish between ‘negative sentences’ and ‘counterfactual sentences’, as well as between ‘negative situations’ and ‘counterfactual situations’. London does not lie on the Rhine is a negative sentence, referring to what we can call the actualization of a negative situation (i.e. a situation referred to in a negative sentence). The situation referred to in this sentence is factual in the objective S-world. By contrast, If I were you… refers to (the actualization of) a situation which is factual in a counterfactual S-world. This clause is a counterfactual clause referring to a counterfactual situation (= a situation that does not actualize in the objective S-world because it is incompatible with this world). (i)

‘Hypothetical t-worlds’ – see (g) above – may be created by various means. As we have seen, one possibility is the use of an epistemic modal adverb like perhaps. Another possibility is the use of a nonfactive (opaque) verb like think, claim, etc., which creates an ‘intensional’ world. Another type is the (‘open’ or ‘tentative’) ‘theoretical world’ (Declerck & Reed 2001: 71–72) created by a conditional clause like If John {comes / should come / came / were to come} tomorrow… As will be argued below, a subjective t-world created by an expression of posteriority, such as the future tense form in I will do it tomorrow,

36 Renaat Declerck is also a hypothetical t-world. And still other possibilities will be discussed in section 3. (j)

Some hypothetical t-worlds can be characterized in terms of a modal concept which has been neglected in the literature on modality, viz. the idea that the t-world in question is envisaged by the speaker but not yet factual at the time t to which it is anchored. This kind of hypothetical world, which we will refer to as ‘not-yet-factual at t’ is evoked by any expression that has posteriority as part of its meaning. The clearest cases are those in which the reference is exclusively to a future world. Thus, John will trim the hedge evokes a t-world which is subjective in the specific sense that it is not-yet-factual at S but is predicted (at S) to become factual at some future time t. The proposition ‘John trim the hedge’ is true of that not-yet-factual t-world (in which actualization is envisaged by the speaker), but the not-yet-factual tworld itself is neither factual nor counterfactual at S: It is a hypothetical S-world.5 So, what we are arguing is that the situation referred to by John will trim the hedge does not form part of the narrow objective S-world (or of a narrow subjective S-world), because that situation is not (thought of as) actualizing at S, and that it does not belong to the corresponding extended S-world either, because it has not actualized before S – see (e) above. Since the situation is predicted to actualize in the future, it does not belong to a counterfactual S-world either (at least not to a ‘purely counterfactual’ S-world – see section 2 below). The S-world to which it does belong is a world which is conceived in the speaker’s mind at S. This world in question is a ‘not-yet-factual’ S-world, i.e. an S-world in which the actualization of the situation of John trimming the hedge is envisaged but is not factual yet.

Apart from drawing attention to the existence of not-yet-factual t-worlds (which are hypothetical worlds and should therefore be treated in any comprehensive discussion of modality), this article aims to show that the value ‘not-yet-factual at t’ pertains not only to future situations (and situations that are represented as posterior to a past reference time), but also to the situations referred to in sentences like (1a–c) (repeated here): (1a) [It’s high time] we left. (The situation of our leaving is envisaged at the time of speech, but is not yet a fact at that time.)

Non-yet-factual at time t’

37

(1b) [I wish] Jim would stop lying. (The situation of Jim ceasing to lie is not yet a fact at speech time, but is envisaged at that time as being weakly possible in the future.) (1c) [I saw Sam] before she had seen me. (= ‘I saw Sam at a time when she had not yet seen me’, ‘When I saw Sam, it was not yet a fact that she had seen me’) We will investigate what exactly ‘not-yet-factual at t’ means, how this interpretation comes about, and how the concept can be accommodated in an analysis of epistemic modality in English.

2. More on the definition of ‘not-yet-factual at t’ In this section we will argue that sentences representing the actualization of a situation as not-yet-factual in a t-world combine two ideas, viz. ‘counterfactual at t’ and ‘predicted or otherwise envisaged at t’. The latter idea implies ‘envisaged as possibly actualizing at a time posterior to t’, i.e. ‘envisaged as factual in a hypothetical world that might exist at a time posterior to t’. (Envisage thus refers to something weaker than prediction. It just means that the speaker reckons with the possibility that a situation will actualize.) We thus claim that the factuality value ‘not-yet-factual at t’ is a combination of the values ‘t-counterfactual’ and ‘t-hypothetical’. Since such a combination is different from either of the constituting values, it follows that there are actually four theoretically possible factuality values for some t-worlds, viz. [+ factual], [+ (purely) counterfactual], [+ (purely) hypothetical] and [+ not-yet-factual]. The latter three are possible realizations of ‘nonfactual’ (= [– factual]). Unfortunately, this means that the term ‘nonfactual’ is potentially ambiguous. We have to distinguish carefully between the factuality value ‘nonfactual’ and the three subvalues which it allows, viz. ‘counterfactual’, ‘hypothetical’ and ‘not-yet-factual’. In order to avoid confusion, it is better not to use the label ‘nonfactual’ in the sense of ‘notyet-factual’, as I have done in earlier publications (e.g. Declerck 1991: 105) in connection with not-yet-factual before-clauses (as in I saw him before he had seen me, which is interpreted as ‘He had not yet seen me when I saw him’ – see section 4 below).

38 Renaat Declerck As an illustration of the above claim, let us consider the following sentences: (2a) (2b) (2c) (2d) (2e)

Abraham Lincoln abolished slavery. Jimmy Carter is the President of the U.S.A. [I’d be surprised] if Fred was helping his sister right now. Joan will bake a cake tomorrow. [John promised] that he would fix the car.

In (2a), the actualization is factual in the extended objective S-world. In (2b), the actualization is (purely) counterfactual in the objective S-world. The actualization referred to in the if-clause of (2c) is (purely) hypothetical in the speaker’s S-world, because the speaker does not know whether Fred is helping his sister or not. In (2d), the actualization referred to is not-yetfactual in the speaker’s S-world.6 And the actualization of the situation referred to in the that-clause of (2e) is not-yet-factual in John’s past t-world, i.e. in the objective world holding at the time when John made his promise. It should be clear now that each of the terms ‘factual’, ‘(purely) counterfactual’, ‘(purely) hypothetical’ and ‘not-yet-factual’ can be used in two ways. Firstly, they can refer to one of the factuality values which the actualization of a situation has in a given t-world. Secondly, the four terms also apply to particular kinds of t-worlds. For example, the conditional world evoked by I would feel safer if he was with me is a subjective counterfactual S-world, because the situations that are factual in this S-world are counterfactual in the objective S-world. This means that the following conclusions have been reached: (a) There are four types of t-worlds. We can call them ‘factual’, ‘purely counterfactual’, ‘purely hypothetical’ and ‘not-yet-factual’, respectively. (b) The same four labels can be used for the four (mutually exclusive) factuality values that (the actualizations of) situations can have in given t-worlds. (c) ‘Not-yet-factual at t’ combines the ideas ‘counterfactual at t’ and ‘hypothetical at t’. However, it is incompatible with the idea ‘purely counterfactual at t’ as well as with the idea ‘purely hypothetical at t’. In sum, the factuality values that may be applicable to a given nonfactual tworld can be represented as in Figure 1.

Non-yet-factual at time t’

counterfactual purely counterfactual

39

hypothetical

not-yet-factual

purely hypothetical

Figure 1. Representation of the factuality values in nonfactual t-worlds

3. Instances of actualizations that are not-yet-factual at t As has been shown above, a situation that is referred to in the future tense is (as a rule – see section 6.2) not-yet-factual in the objective S-world or in the speaker’s subjective S-world. Similarly, a situation that is temporally located posterior to a past time, as in [John promised that] he would fix the car, is not-yet-factual in the objective world holding at that past time. ‘Not-yet-factuality at S’ can also be expressed by a modal indicative form, more specifically by a past tense or ‘conditional tense’ form. In those cases the ‘pastness’ of the tense cannot be explained in terms of location in time or purely temporal relations. (For instance, the past tense may refer to a situation that is future rather than past.) Sentences (3a–c) illustrate this (3a) It is (high) time we changed the code. (The situation of our changing the code is not-yet-factual in the objective S-world: Though it has not actualized yet, its actualization is envisaged for the future.) (3b) I’d rather you left next week. (similar) (3c) I wish it would stop raining. (The stopping of the rain is not-yetfactual in the speaker’s S-world because its actualization in a future world is envisaged as a possibility, however slight.) The reason why, in (3c), we use the conditional tense (and not the past tense, as in [3a–b]) is that, after I wish, the past tense is reserved for the expression of S-counterfactuality: (4)

I wish I had a son. (The situation of my having a son is counterfactual in the objective S-world.)

40 Renaat Declerck The tenses used to express not-yet-factuality in t-worlds that hold at a time other than S (i.e. in the past or future) are the same as those used in connection with S-worlds: (5a) It was (high) time we changed the code. (5b) I wished it would stop raining. (5c) [When we return from our holiday] it will be (high) time we changed the code. (5d) Before the end of the day you will wish that it would stop raining. Another typical context in which a not-yet-factual meaning can arise is in before-clauses in the past perfect: These can represent the actualization of the situation referred to as not-yet-factual in the t-world that holds at the time of the (actualization of the) head clause situation. For example: (6a) I saw Sam before she had seen me. (6b) [Another fifteen to twenty thousand snakes perished in the winter of 1998-9,] when a sudden frost caught them before they had descended below the frost line. (www) The (actualization of the) before-clause situation is represented here as notyet-factual at the time of the (actualization of the) head clause situation. Thus, in the first example it can be paraphrased in either of the following ways: (7a) I saw Sam at a time when she had not yet seen me. (7b) Sam had not yet seen me when I saw her. These paraphrases make it clear that the interpretation of the before-clause is neither factual nor (purely) counterfactual in the past t-world in which my seeing Sam is factual. A t-factual interpretation of the before-clause would be triggered by the past tense: (8)

I saw Sam before she saw me. (= ‘We saw each other, but I saw her first.’ Note that there is reference to two different past t-worlds here: that in which my seeing Sam is t-factual and that in which Sam seeing me is t-factual. The latter t-world follows the former in time: This posteriority relation is expressed by before.)

Non-yet-factual at time t’

41

A (purely) counterfactual interpretation of the before-clause situation would be triggered by the conditional perfect: (9a) I saw Sam just before she would have seen me [if I had not quickly turned into a side street]. (9b) I paid the electricity bill two days before the company would have cut off my supply. (The conditional perfect is possible because of the implicit condition ‘if I had not paid the bill’.) The interpretation which had seen does receive in (6a) [I saw Sam before she had seen me] is a ‘not-yet-factual at t’ reading: In the world anchored to the time of my seeing Sam, the situation of her seeing me was counterfactual but its actualization was envisaged. (As noted above, ‘envisage’ here refers to a weaker notion than prediction. It just means that I reckoned with the possibility of her seeing me.) Another case in which a not-yet-factual reading is possible concerns adverbial until-clauses and before-clauses depending on a (finite or nonfinite) clause which itself depends on a head clause whose verb is an intensional verb like intend, decide, promise… which implies that the actualization of the situation of the complement clause is posterior to the time of the intensional attitude (i.e. the intention, decision, promise, etc.) expressed by the verb. In such a context, the subjective world in which the (actualization of the) situation of the temporal clause introduced by until or before is factual is posterior to the t-world in which the making of the decision, promise, etc is factual. Compare: (10a) Meg nagged Tim until he gave in. (Both situations are interpreted as factual in the speaker’s extended S-world.) (10b) Meg decided to nag Tim until he gave in. (The until-clause situation is not-yet-factual at the time of the beginning of the nagging, which is itself not-yet-factual at the time when Meg made her decision. It follows that the until-clause situation is also not-yet-factual at the time of Meg’s decision. Note that we do not know from this sentence whether or not Tim eventually gave in.) (10c) Meg nagged Tim for three weeks before he gave in. (Both situations are interpreted as factual in the speaker’s extended S-world.) (10d) Meg decided that she would wait for some time before she replied to Tim’s letter. (The before-clause situation is not-yet-factual at the time of (the beginning of) the waiting, which is itself not-yet-factual at the

42 Renaat Declerck time when Meg made her decision. It follows that the before-clause situation is also not-yet-factual at the time of Meg’s decision. Note that we do not know from this sentence whether or not Meg eventually replied to Tim’s letter.) Other temporal clauses depending on an infinitive clause functioning as object of decided, promised, intended, etc. are interpreted as not-yetfactual, not in the t-world in which the infinitival clause situation actualizes, but in the t-world in which the decision, promise, etc. actualizes: (11) I intended to call up Bill {when he was at home / after Meg had left / while he was at his office}. (The time clause situation is factual in the extended hypothetical t-world in which my calling up Bill is factual, but both these situations – my calling up Bill and the time clause situation – are not-yet-factual at the time of my intention, i.e. they are not-yet-factual in the extended S-world in which the existence of the relevant intention is a past fact.) In this example, the ‘not-yet-factual at t’ interpretation is not induced by the temporal conjunction but by the relation of posteriority that is implicit in the meaning of the verb intend. In obsolescent English, the auxiliary should could be used to impart a not-yet-factual meaning to temporal clauses in sentences whose head clause did not involve a verb form like decided, intended, etc. The following examples are cited in Jespersen (1932: 350–352): (12a) Mr. Dombey sat down to watch them until Edith should return. (12b) She had (…) asked shelter until the storm should cease. (12c) He hastened home to be there before the guests should arrive. (12d) I had serious misgivings for the school’s future after I should have left. (12e) I was longing to see their faces when they should at last turn and see her.

Non-yet-factual at time t’

43

4. The origins of ‘not-yet-factuality at t’ and the means of expression ‘Not-yet-factuality-at-t’ always requires reference to a posterior situation. This is self-evident in connection with examples like John will kill you, I thought that John would kill you and I wish John would kill you. It is also clear in It’s high time we fixed the car, where fixing the car can only be envisaged for the future, and in I decided to tell him the truth {when / after / while} he came in, in which not only the telling but also the actualization of the time clause situation can only be interpreted as following the decision. Posteriority also plays a crucial role in before-clauses. To see this clearly, it is necessary to say something about the temporal structure that is expressed by a sentence with a not-yet-factual before-clause. Let us consider the following example: (13) I saw him before he had seen me. The before-clause here receives a not-yet-factual interpretation: The sentence can be paraphrased ‘I saw him at a time when he had not yet seen me’ or ‘He had not yet seen me when I saw him’. The temporal structure of the sentence is shown in Figure 2. (The conventions observed in Figure 2 are the following. The time-line is divided into a ‘past time-sphere’ and a ‘present time-sphere’ by a short stretch of dotted line. This dotted line in the middle of the time line represents the fact that there is felt to be a break between the two time-spheres. The Venn-diagram represents the ‘temporal domain’ – see section 2.15 of Declerck (2006) – to which the relevant times and temporal relations belong. The time of actualization of the situation referred to by the tense form which establishes the temporal domain is placed on the time line because that actualization is treated as a past fact (i.e. as factual in the extended objective S-world). A line slanting to the left represents anteriority, whereas a line slanting to the right expresses posteriority. A slanting full line represents a relation expressed by a tense form, whereas a slanting dashed line represents a temporal relation expressed by a conjunction. All times are represented by a cross (or ‘x’), irrespective of whether they are durative or punctual.)

44 Renaat Declerck

saw

speech time

x

•••

│

before

x implicit t

x had seen

Figure 2. The temporal structure of I saw him before he had seen me.

To understand the structure represented by Figure 2, we need to know that the conjunction before has developed from a phrase of the form ‘before the time that’ or ‘before then that’, variously realized in Old English as toforan þam timan þe, foran to þam timan þe and toforan þam þe – see Mitchell (1987: 379), and that the reduction to before that, and later before, has not changed this meaning. Before still means ‘before the time that’, in which the time functions as an implicit orientation time as far as the use of the tense in the before-clause is concerned. That is, in I saw him before he had seen me, the form had seen represents the seeing as anterior to an implicit orientation time (called ‘implicit t’ in Figure 2), which is itself represented as posterior to the time of the (actualization of the) head clause situation by the conjunction before. It is important to note that this is a purely temporal structure, which only involves times and temporal relations. The times are the times of the (actualizations of the) situations referred to by saw and had seen, plus the implicit orientation time inherent in the meaning of before. The temporal relations are the relations expressed by the past tense (= ‘past relative to S’) and the past perfect (= ‘anterior to a t in a past domain’), plus the posteriority relation expressed by before. There are two factors that trigger the reading that the situation of his seeing me was not yet a fact at the time when I saw him. One is the fact that the situation of his seeing me is tempo-

Non-yet-factual at time t’

45

rally related to an implicit posterior time, the other is the fact that the tense in the before-clause is ‘anteriorized’: We use had seen instead of the past tense form saw. This tense choice is important, because I saw him before he saw me is naturally interpreted as meaning that he did see me after I had seen him. 7 (The same is true of before-clauses referring to the future: To convey the not-yet-factual reading we can use I will see him before he has seen me, although there is no purely temporal reason to express anteriority. The reason for using has seen is that the unmarked interpretation of the corresponding present tense form sees in I will see him before he sees me is different, viz. ‘He will see me, but only after I have seen him first.’) Formally speaking, ‘anteriorizing’ is not the same as the phenomenon of ‘backshifting’ which occurs in indirect speech, for example, when John {died / has died} is reported as I told Bill that John had died. This kind of backshifting is no more than the phenomenon that anteriority to a past orientation time has to be expressed by the past perfect whereas anteriority to S has to be expressed by the past tense or the present perfect. By contrast, the anteriorizing which we observe in the before-clause of I saw him before he had seen me (and I will see him before he has seen me) is the phenomenon that a tense form expressing anteriority (to the implicit orientation time inherent in the meaning of before) is used instead of a tense form expressing simultaneity. The fact that an anteriorized form is used in not-yet-factual beforeclauses is not surprising. It is common knowledge that using an anteriority form where there is no purely temporal reason for using it is a widespread device to trigger epistemic modal meanings. For example, instead of saying I intended to visit him, which merely expresses that at some past time I had the intention of visiting him, we can use I had intended to visit him to express that, in spite of my past intention, I did not (and will not) visit him. Anteriorizing here triggers a counterfactual interpretation of the situation referred to in the infinitive clause. The same mechanism can also induce tentative modality, as in I wanted to have a word with you and I would be happy if she arrived tomorrow, which are tentative versions of I want to have a word with you and I will be happy if she arrives tomorrow. Needless to say, anteriorizing is essential in examples like It’s high time we fixed the car, since the ‘not-yet-factual at S’ meaning cannot be expressed by *It’s high time we fix the car. In I wish John would kill his wife the expression of posteriority creates the not-yet-factual meaning, while the anteriorizing of the marker of posteriority (will) produces a tentative interpretation, i.e. the possibility that John will kill his wife is seen as rather

46 Renaat Declerck unlikely. (The above sentence can be compared with I hope John {will / *would} kill his wife, where the second aspect of meaning cannot be expressed.)

5. ‘Strong’ versus ‘weak’ not-yet-factual interpretations We can speak of strong not-yet-factuality if, as in the above examples, the expression of posteriority is accompanied by anteriorizing of the tense form. We can speak of weak not-yet-factuality if there is posteriority but no anteriorizing, as in the following examples: (14a) I will visit them tomorrow. (The actualization of the situation is just located in a t-world that is posterior to the speaker’s S-world. This implies that my visit is not-yet-factual at S.) (14b) John promised that he would visit me the next day. (The actualization of the situation is just located in a t-world that is posterior to the past t-world in which the making of the promise is factual. This implies that John’s visit was not-yet-factual at the time of his promise.) (14c) I decided to wait until the shop opened. (similar) These examples illustrate the obvious fact that any situation whose actualization is conceived of as posterior to a time t is not yet factual at t. This is weak not-yet-factuality: The idea of ‘not-yet-factual at t’ inevitably forms part of the interpretation of the clause, but it is not the predominant aspect of meaning. Strong not-yet-factuality means that the extra mechanism of anteriorizing puts emphasis on the idea of not-yet-factuality. In other words, whereas there is weak not-yet-factuality whenever a situation is posterior to a t, there is strong not-yet-factuality only if the ‘not-yet-factual at t’ meaning is linguistically emphasized by the use of an anteriorized tense form. (In some cases this use is obligatory.) The following examples further illustrate this: (15a) “[We think] it’s time something’s done,” says Donald Skadden. (Wall Street Journal) (The sentence in the present tense only expresses weak not-yet-factuality. It just means ‘It is time to do something – a meaning which can also be expressed by It’s time something

Non-yet-factual at time t’

47

should be done. The idea that nothing has been done yet is not foregrounded.)8 (15b) It’s time something was done [about it]. (In this case strong not-yetfactuality is expressed. The idea ‘It is time to do something about it’ forms part of the interpretation, but the foregrounded aspect of meaning is the idea that nothing has been done about it yet. The combination of these two ideas naturally leads to the interpretation that something should already have been done about it, i.e. that taking measures has been postponed for too long.) For at least some speakers, before-clauses need not use an anteriorized form if a t-factual reading is ruled out by the context in which the beforeclause is used. For example, compare the following: (16a) She read the letter before I read it. (16b) She read the letter before I had read it. (16c) She tore up the letter before I {had read / read} it. Because of the posteriority relation expressed by before, the situation of my reading the letter is automatically weakly not-yet-factual in the t-world in which the (actualization of the) head clause situation is factual. The function of using the anteriorized form had read in (16b) is therefore to bring not-yet-factuality into focus, because the use of read in (16a) suggests that I read the letter too. In the latter case the speaker focuses on the t-factuality of the before-clause situation in the extended S-world rather than on the not-yet-factuality of the situation in the anterior t-world established by the head clause in the past tense. In (16c), the head clause pragmatically excludes the possibility of actualization of the before-clause situation, so that the idea of not-yet-factuality comes to the fore regardless of whether there is anteriorizing or not. This explains why the speaker can replace had read with read without creating confusion.

48 Renaat Declerck 6. Conclusion 6.1. Factuality values One conclusion from the above discussion is that, with reference to a tworld, the set of possible factuality values of (actualizations of) situations (corresponding with truth values of clauses) does not only contain the values ‘factual’ (true), ‘counterfactual’ (untrue, false) and ‘hypothetical’ (possibly true) but also the value ‘not-yet-factual’ (as yet untrue). This value, which plays a crucial role in the interpretation of all sentences of the kinds referred to in section 4, has been neglected in the linguistic literature, probably because it has no place on the traditionally recognized epistemic scale which ranges (roughly) from ‘factual’, over ‘probable’, ‘possible’ and ‘improbable’, to ‘counterfactual’. However, the value ‘not-yet-factual’ is a modal value, because it is a nonfactual value combining the ideas of counterfactuality and hypotheticality – see section 2 above. As far as I know, some of the not-yet-factuality cases discussed in section 4 have never been treated as modal. This is true of temporal clauses generally (introduced by until, when, etc.), which can be weakly not-yetfactual, and of not-yet-factual before-clauses in the past perfect, which are strongly not-yet-factual. None of the major English grammars today, viz. Quirk et al. (1985), Huddleston & Pullum (2002) and Biber et al. (1999), even mention the special use of the past perfect in (what I call) not-yetfactual before-clauses. Quirk et al. (1985: 188, 1011) call the past tense forms in It’s time we left and I wish he was here instances of the ‘hypothetical past’, which is also used in if-clauses of conditionals like I’d be happy if I was you or I’d be happy if she came. The would in I wish she would visit me more often they similarly call ‘hypothetical would’ (p.1012). Huddleston & Pullum (2002: 148) treat both [I wish] he was here and If he was here… as expressing ‘modal remoteness’ with present time reference, and treat both [I’d rather] you went tomorrow and If you went tomorrow… as expressing ‘modal remoteness’ with future time reference. The clauses in the past tense forms are called ‘modal preterite clauses’. Biber et al. (1999) do not refer to modal past or past perfect forms at all. It should be clear that the epistemic modal value ‘not-yet-factual at t’ which we have discussed in this article is quite different from the ‘modal remoteness’ expressed in ‘hypothetical’ (or ‘tentative’ – see Declerck &

Non-yet-factual at time t’

49

Reed 2001: 93–98) conditional clauses like if she came. As expressed in the title of this article, not-yet-factuality is a neglected modal concept.

6.2. The status of the ‘future tense’: tense or mood? The issue is also pertinent to the moot point whether will in Bill will arrive tomorrow is a future tense auxiliary or a modal auxiliary. Quirk et al. (1985: 213) claim that “there is no future tense in English” and that will is a “modal verb”. The modality in question is said to be prediction. Huddleston & Pullum (2002: 208) speak of “the lack of a future tense in English” and claim that futurity is a use of will that falls “within the general idea of epistemic modality” (p.188) but they do not specify what kind of epistemic modality is expressed in Bill will arrive tomorrow. This point of view is shared by many other linguists. However, our position, defended in Declerck (1991: 8–13, 2006: section 2.8), is that will arrive is a future tense form because it locates Bill’s arrival in the future. We subscribe to Lyons’ (1977: 68) and Comrie’s (1985: 9) view that tense is a grammaticalized means of locating (the actualization of) a situation in time. Linguistically speaking, will arrive does exactly this (viz. locate a situation in time) and nothing else. It is therefore a future tense form. Of course, it is true that the predicted situation is not-yet-factual at S, but this is simply because any situation that is predicted to actualize later than S is not yet factual at S. This is a question of how the objective world is rather than of how it is linguistically represented. As is true of any tense, the function of the future tense is to represent a situation as actualizing at a particular time. If the time in question is represented as posterior to S, the situation is automatically not-yet-factual at S (in the weak sense of the term). There is nothing in a future tense form to hint at the possibility of strong not-yet-factuality: There is no anteriorizing or any other formal means of focusing attention on ‘not-yet-factuality at S’. Moreover, the idea ‘not-yet-factual at S’ appears to have the strength of an implicature only. There are sentences in the future tense in which the idea is blocked (= prevented from arising) or cancelled by an element in the sentence or in its context or by our pragmatic knowledge of the world. Consider: (17a) I will still be here tomorrow. (Still blocks the implicature that I am not yet here at S.)

50 Renaat Declerck (17b) [For the moment you can’t beat me. You’ll be able to beat me when I’m fifty] because you will be ten years younger than me. (The addressee is already ten years younger than the speaker, but it is only when the speaker becomes fifty that this difference of age will play a decisive role. The future tense is used here to put the ‘temporal focus’ (Declerck 2006: chapter 11) on that future time.) Similarly, the sentence John will be in London tomorrow does not rule out the possibility that John is already in London at S (so that the sentence can be used in a context inducing this interpretation). However, it actually only predicates something of tomorrow and locates John’s being in London at that time. It does not say anything about the present. In the same way as John was in London yesterday implicates that John is no longer in London at S, John will be in London tomorrow implicates that John is not yet in London at S. (These implicatures are due to the Gricean Maxim of Relation (relevance): other things being equal, the present is more relevant to the speaker than the past or the future. So, if a situation is actualizing at S, it should in principle be referred to in the present tense. If another tense is used, the hearer has a right to conclude that the situation is not actualizing at S.)

6.3. Consequences for the definition of modality It follows from what has been argued that epistemic modality can be identified with ‘nonfactuality’. A sentence is epistemically modal if and only if the actualization of the (positive or negative) situation referred to receives the value [– factual] (= ‘nonfactual’) in the objective S-world.9 If this is the case, the actualization is factual in a modal (= nonfactual) t-world. We have distinguished three types of modal (= nonfactual) worlds: purely counterfactual worlds, not-yet-factual worlds and purely hypothetical worlds. The latter comprise different subtypes, such as ‘open possibility worlds’ (e.g. John is perhaps ill), ‘tentative worlds’ (e.g. If John came tomorrow…), etc. If this is correct, all epistemic modality is accounted for, because what is common to That must be true, That should be true, That ought to be true, That may be true, That may not be true, That might (not) be true, That cannot be true, That couldn’t be true, That needn’t be true, That will be John’s car, Would that be your car?, etc. is that these sentences all refer to a hypothetical actualization (while speaking about it in terms of different degrees

Non-yet-factual at time t’

51

of necessity or possibility). Even That must be true does not represent the actualization of the situation (of ‘that’ being true) as factual. The sentence may be a very confident conclusion, but it does not have the value [+ factual] which That is true does have. (Similarly, That cannot be true does not express counterfactuality.) The identification of modality with nonfactuality in fact also appears to be correct for modality that is not epistemic but is either ‘deontic modality’ (which has to do with the presence or absence of obligation or permission), ‘root possibility’ (as in Aerosols can explode, which means ‘It is (theoretically) possible for aerosols to explode’), ‘root necessity’ (as in Everybody must die eventually)10 or ‘dynamic modality’ (Palmer 2001) such as ability or volition (e.g. willingness). In each case the actualization of the situation referred to by the infinitive clause following the auxiliary is nonfactual in the objective S-world: None of the sentences You must swim, You may swim, You can swim, etc. represents the situation of the adressee’s swimming as factual in the objective world. In each case the actualization of that situation is envisaged as possibly factual in a hypothetical world. ‘Modal auxiliaries’ are modal in the sense that they create a hypothetical world in which the actualization of the situation of the infinitive clause is envisaged as having a factuality value. Even ‘dynamic modality’, whose modal status is often questioned, fits perfectly into this analysis: I will help you (in the sense of ‘I am willing to help you’) and I can help you (expressing my ability to help you) both fail to represent the situation of my helping you as actualizing in the objective S-world. But both will and can create a modal world, viz. a purely hypothetical world.11 Intensional verbs like think, want, expect, hope, intend, etc. are modal in the weak sense that they create a world in which the actualization of the situation referred to in the complement clause is nonfactual, more specifically not-yet-factual at t (where t is the time of the actualization of the situation referred to by the intensional verb). When such a verb undergoes anteriorizing, the idea of not-yet-factuality is supplemented with the idea (more specifically: implicature – see Declerck and Reed 2005) – of counterfactuality. Thus, I had intended to tender my resignation suggests that I did not resign, am not resigning or will not resign after all. Traditionally, only such anteriorized verb forms are treated as modal – grammars speak of a ‘modal past perfect’ – but this is simply because, unlike counterfactuality, not-yet-factuality has not been recognized as a modal value.

52 Renaat Declerck Notes

1.

This article has greatly profited from comments by Susan Reed, Ilse Depraetere, Bert Cappelle and An Verhulst. Of course, none of them is responsible for the claims that are made. 2. The term possible world is used in its usual linguistic sense. It roughly corresponds to what is called an ‘alternative (total) state of the world’ in philosophy of language – see Soames (2003: 268–275). 3. When referring to the future tense in this article we will especially be concerned with its modal implication. This does not alter the fact that we treat the future tense as a real tense – see section 6.2 below. 4. We speak of ‘most t-worlds’ because we will argue below that there are tworlds that allow a fourth factuality value, viz. the value ‘not-yet-factual at t’. 5. This claim will be more accurately formulated in section 2, where it will be argued that an actualization that is not-yet-factual at t is at the same time counterfactual and hypothetical at t. 6. It should be noted, however, that not all sentences in the future tense yield a ‘not-yet-factual at S’ interpretation. As we will see in section 6.2, this reading is an implicature which can be blocked by another constituent of the sentence (as in I will still be here tomorrow) or cancelled by contextual information or by pragmatic knowledge. 7. However, it will be pointed out at the end of section 5 that (at least for some speakers) the tense of the before-clause does not need to be ‘anteriorized’ if a t-factual reading is already ruled out by the context in which the before-clause is used. An illustration of this is He burned the letter before I read it. 8. The following is a similar example from the Wall Street Journal: It’s time business leaders and the general public learn that mankind does not rule over this natural environment 9. This conclusion may not be new (i.e. there are other linguists who subscribe to this definition), but it is certainly not uncontroversial, and what is new in this article is the claim that there are three (rather than just two) modal factuality values (i.e. factuality values in nonfactual words) – see immediately below. 10. Sentences with should (and ought to) apparently form an exception: They allow the actualization of the (positive or negative) infinitive clause situation to be factual at S, as in the following examples: (i) I should be proud of you, and I am! (www) (ii) That money should not be in the drawer. It should be in the safe. Such sentences express present necessity without a ‘not-yet-factual at S’ implication.

Non-yet-factual at time t’

53

11. Of course, I can swim can be said by someone who is actually swimming, but that does not alter the fact that the t-world created by can is a hypothetical world. That this world happens to overlap the actual world is a question of what is the case in the actual world and not of what is linguistically represented by the speaker.

References Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 Longman Grammar of Spoken and Written English. London: Longman. Comrie, Bernard 1985 Tense. Cambridge: Cambridge University Press. Declerck, Renaat 1991 Tense in English: Its Structure and Use in Discourse. London/New York: Routledge. 2006 The Grammar of the English Tense System. Berlin/New York: Mouton de Gruyter. Declerck, Renaat, and Susan Reed 2001 Conditionals: A Comprehensive Empirical Analysis. Berlin/New York: Mouton de Gruyter. 2005 What is modal about I thought that…? English Language and Linguistics 9: 311–332. Huddleston, Rodney, and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English language. Cambridge: Cambridge University Press. Jespersen, Otto 1932 A Modern English Grammar on Historical Principles. Part 4: Syntax (3rd volume): Time and Tense. London: George Allen and Unwin. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Mitchell, Bruce 1987 Old English Syntax. Vol. II: Subordination, Independent Elements, and Element Order. Oxford: Clarendon Press. Palmer, Frank R. 2001 Mood and Modality. 2nd ed. Cambridge: Cambridge University Press.

54 Renaat Declerck

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Soames, Scott 2003 Philosophy of language: overview. In International Encyclopaedia of Linguistics. Second Edition. Vol. 3, William J. Frawley (ed.), 268–275. Oxford: Oxford University Press.

Semantic ascent, deixis, intersubjectivity and modality Keith Mitchell

1. Introduction This paper argues that the three concepts listed in my title play an important – and hitherto largely unrecognised – part in the semantics of modality. I propose to show that certain recalcitrant problems in the description of modality in English may benefit from being re-examined afresh in the light of these concepts. There is general agreement among scholars that at the core of modality lie two key notions: potentiality and subjectivity. In deontic modality the speaker uses his volition, his authority or his reason to seek to bring about the occurrence (or non-occurrence) of potential situations (events, acts, processes, states, etc). In epistemic modality the speaker uses inference or evidence from his existing knowledge to try and establish or assess the potential factuality of propositions. Modality has to do, then, with potential acts and potential facts, as opposed to actual acts and facts. However, as I argue in this paper, the field of modality is also concerned with the contrast between the modally unmarked actual and the marked potential. Modality is also inherently and prototypically subjective, i.e. speakercentred. It involves “the locutionary agent’s expression of himself and of his own attitudes and beliefs” (Lyons 1982: 102). Modal utterances are speech acts in which the speaker decides what states of affairs are to obtain or occur, and the speaker expresses his conclusions about what might be true or not. Speaker-centredness, or subjectivity, in matters of person, place and time, is reflected in deixis. One thing I wish to examine in this paper is whether it makes sense to treat modality as a further, more abstract, type of deixis. In the same way that tense grammaticalises temporal deixis, and that various pronoun, determiner and adverb systems grammaticalise person- and place-deixis, so modal systems – whether inflectional, syntactic or lexical – appear to grammaticalise what we might call “epistemic deixis”,

56 Keith Mitchell i.e. proximity vs remoteness in relation to the speaker’s current state of knowledge. The difficulties that I propose to re-examine are all phenomena that have either been excluded from standard accounts of modality or been found hard to account for within existing descriptive frameworks. They are: – – – – –

existential modality clause types (syntactic mood) auxiliary do quasi-subjunctive should concessive may

2. Semantic ascent: three orders of entities I will begin by appealing to the notion that Lyons (1989), following Quine, refers to as ‘semantic ascent’, the idea that the objects or entities that we identify in the universe can be thought of as lying on an ascending scale of abstraction. Earlier (1977) he draws a similar ontological distinction between first-order, second-order and third-order entities. Physical objects are what we will call first-order entities ... they are located, at any point in time, in what is, psychologically at least, a threedimensional space ... By second-order entities we shall mean events, processes, states-of affairs, etc., which are located in time ...; and by third-order entities we shall mean such abstract entities as propositions, which are outside space and time. (Lyons 1977: 442–443)

For convenience, I will henceforth refer to the three orders of entities in abbreviated form as E1, E2 and E3. It will be noted that whereas the locations of E1s and E2s are clearly specified as space and time respectively, the domain of E3s is only negatively defined. I shall tentatively suggest that the latter, i.e. propositions, are entities whose truth (or otherwise) is what matters and that this property is located in the human mind. The property of E1s that we are concerned with is their existence or location. And, in the case of E2s, we are concerned with their occurrence or their obtention (with whether they occur, in the case of dynamic situations, and with whether they obtain, in the

Semantic ascent, deixis and intersubjectivity

57

case of static situations, i.e. states). As is implied in the previous sentence, I will follow Lyons (1977) and others, e.g. Huddleston and Pullum (2002) in using the term ‘situation’ as a cover term for events, actions, processes, states, etc. We can thus schematise Lyons’s distinction as follows: Order of entity

property

E1s are physical objects which E2s are situations which E3s are propositions which are

exist occur/obtain true

domain in in in

space time the mind

Just as everything in the first column is some kind of entity, so everything in the third column can be conceived of as some kind of space – a physical space, a temporal space and what we might call an epistemic space. A similar three-way distinction to Lyons’s is to be found in the work of the rhetorician Grierson: There are three orders of phenomena of which the human mind is always aware ...: these are (1) the order of phenomena in space, the relative position of co-existing things; (2) the order of phenomena in time, events following one another; (3) the order of thoughts in the mind, the logical dependence of one truth upon another, or, it may be, the manner in which one thought tends to evoke another even when there is no logical connection between them. (Grierson 1944: 21–22)

Note that Grierson is more explicit than Lyons in locating E3s in “the mind” but is less explicit in other respects – he talks of phenomena rather than entities, and of “things”, “events” and “thoughts” rather than physical objects, situations and propositions. Nevertheless, the parallels are clear.. The most important difference between the two scholars is in their use of the word “order”. For Lyons, “orders” are levels on an ascending scale of abstraction (hence his term ‘semantic ascent’), whereas for Grierson, “orders” are orderings (in the sense of arrangements, sequences, relationships) among entities at each level of abstraction. This is a natural preoccupation for a rhetorician and relates to the distinction between the three principal modes of discourse: description, narration and argumentation. Fundamentally, description is concerned with the spatial relations between physical objects (E1), narration with the temporal relations between situations, both static and dynamic (E2) and argumentation is a matter of the logical relationships between propositions (E3). For example,

58 Keith Mitchell (1a) Your glasses are on the table by the door. [Description: spatial relations among E1s] (1b) She opened the door and went out. [Narration: temporal relation between E2s] (1c) He was born in 1985, so he’s 20. [Argumentation: logical relation between E3s] The examples in (2) show how the same predicate, the verb lead, can express the same three types of relation. These examples also show how the three orders of entities may be expressed as NPs or what Lyons refers to as nominals (distinguishing between first-order, second-order and third-order nominals in (2a), (2b) and (2c) respectively.) This contrasts with the examples in (1) where the E2s and E3s are expressed as clauses and the relations between them are expressed as conjunctions. (2a) This path leads to the beach. (2b) His harassment of his colleagues led to his dismissal. (2c) This argument leads to the same conclusion. Our ternary distinction also provides a basis for explaining relations between lexemes derived from the same stem. Thus, for example, author is understood as the bringer into existence of an E1 (the author of the novel) whereas authority denotes either ‘(holder of) power of control over the occurrence of actions (E2)’ as in the education authorities or The UN has exercised its authority to restore peace, or ‘(possessor of) knowledge (E3)’, as in He is an authority on the subject or I have it on good authority. The verb authorise clearly takes E2 complements and the adjective/noun authoritarian denotes someone who is over-zealous in wielding authority over actions. On the other hand the adjective authoritative (as in an authoritative account of the Easter Rising) is related to the E3 meaning of authority as knowledge. Very often, too, semantic differences cross-linguistically between cognate items (‘false friends’) have an explanation in terms of levels of abstraction. Semantic change, in other words, often involves shifts across boundaries on the scale of abstraction. For example, whereas the French Il est actuellement en France has a temporal meaning and locates the state of affairs (E2) at the speaker’s time of utterance, the English He is actually in France has an unmarked epistemic modal meaning: the adverb serves to locate the proposition (E3) firmly and emphatically in the domain of the

Semantic ascent, deixis and intersubjectivity

59

speaker’s known reality. With the adverbs eventually/éventuellement the reverse seems to be the case. The English adverb has a temporal meaning, whereas the French adverb is modal (although the English noun eventuality is modal).

3. Types of modality Levels of abstraction are most obviously relevant to modality in the distinction between types of modality. Palmer’s (2001: 7) high-level terminological distinction between “event modality” and “propositional modality” neatly reflects the fact that the former (which subsumes deontic and dynamic modality) is concerned with second-order entities and that the latter (which subsumes epistemic and evidential modality) is concerned with third-order entities. To take the two major subcategories in their subjective use, we can say that in deontic modality the speaker expresses some kind of decision about the occurrence (or non-occurrence) of a potential (typically future) situation (E2) whereas in epistemic modality the speaker expresses some kind of conclusion about whether a proposition (E3) is true. In deontic modality the speaker seeks to bring about potential ‘acts’, whereas in epistemic modality he seeks to establish potential ‘facts’. Deontic modal utterances have a directive or conative function, whereas epistemic modal utterances have a speculative or verdictive function. The distinction may be tabulated as follows: DEONTIC MODALITY

EPISTEMIC MODALITY

speaker’s decisions

speaker’s conclusions

about the occurrence of situations (E2) (potential ‘acts’)

about the truth of propositions (E3) (potential ‘facts’)

function: directive/conative

function: speculative/verdictive

You can take my car

They may be at the pub

60 Keith Mitchell 4. Existential modality If deontic and epistemic modality are concerned with the speaker’s ‘take’ on potential E2s and E3s, what about E1s? I should like to suggest that it is E1s, at least mainly, that are involved in existential modality. Classic examples on which discussion in the literature has centred are: (3a) Roses can be mauve. (Palmer 1990) (3b) Welshmen can be tall. (Boyd and Thorne 1969) (3c) Footballers can be sex maniacs. (Lakoff 1972) However, the preoccupation of most scholars with this type of sentence (generic plural subject NP + can + be), albeit the most frequent type, obscures the fact that a fuller range of modals, of necessity as well as possibility, enter into this type of modality, just as they do in deontic and epistemic modality: (4a) Birds have to be bipeds. (5a) Birds can’t be mammals. (6a) Birds don’t have/need to have wings. (7a) Birds can be carnivorous.

(4b) A bird must be a biped. (5b) A bird cannot be a mammal. (6b) A bird need not/may not have wings. (7b) A bird may be carnivorous.

As is clear from the equivalence of need not and may not in (6b), the inverse semantic relation between necessity and possibility predicates also applies in existential modality as it does in the other types of modality. The (b) sentences also show that singular as well as plural indefinite generic subjects are possible, as indeed are singular definite generic subjects, e.g. The squid of the genus Loligo can be as much as two feet long (Palmer 1990: 107), though they appear to be far less frequent. The fact that the classic can sentences (1-3) are paraphrasable with the existential quantifier some is well-known, e.g. (1) = ‘Some roses are mauve’. In fact, the relationship with quantifiers is more systematic and extends to the paraphrase between necessity modals (have to/must) and the universal quantifier all (Cf. Palmer 1990: 107ff). This is not surprising in view of the fact that all/some are also a pair of inverses. Thus (4-7) may be paraphrased as (4c-7c): (4c) All birds are bipeds.

Semantic ascent, deixis and intersubjectivity

61

(5c) No birds are mammals. (6c) Not all birds have wings. (7c) Some birds are carnivorous. These may be expanded as: (4d) Every member of the set {BIRDS} is a member of the set {BIPEDS} (but not vice-versa). (5d) No member of the set {BIRDS} is a member of the set {MAMMALS}. (6d) Not all members of the set {BIRDS} are members of the set {WINGED CREATURES} (7d) Some members of the set {BIRDS} are members of the set {CARNIVORES} As these paraphrases suggest, this type of modality is concerned with the compatibility relations between two (generic) sets of entities: whether the two sets intersect or are mutually exclusive, whether one set is wholly included in the other, etc. Looked at in another way, what existential sentences do is state whether entities that belong to both sets exist or not. Note that the subject NP in each of the following refers to a hypothetical class of entity with dual set membership: (4e) (5e) (6e) (7e)

Birds which are not bipeds do not exist. Birds which are mammals do not exist. Birds which have no wings exist. Birds which are carnivores exist.

Dual set membership is even more apparent when one set is characterised adjectivally, as in paraphrases of our initial examples: (1e) Mauve roses exist. (2e) Tall Welshmen exist. These may be even more naturally paraphrased by existential theresentences of the following type: (1f) (2f)

There are such things as mauve roses. There are such things as tall Welshmen.

62 Keith Mitchell Similarly: (4f)

There is no such thing as a bird which is not a biped.

What I hope this analysis will have shown is that just as deontic modality expresses the speaker’s concern to establish the occurrence of a potential situation (E2), and just as epistemic modality expresses the speaker’s concern to establish the truth of a potential proposition (E3), so, I would like to argue, existential modality expresses the speaker’s concern to establish the existence of a potential class of objects in the universe (E1) – a potential class of entities possessing two properties. If the term ‘existential’ is appropriate for this type of modal usage, it is not so much because of the parallel between can and the existential quantifier some – which was presumably von Wright’s (1951) original motivation for proposing the term – but rather because such sentences predicate existence or non-existence of a class of entities. There appears then to be a systematic relationship between the three kinds of modality and the three orders of entities. However, there remains the important question of whether existential modality is a true modality like deontic and epistemic modality, since all our examples may be construed as unqualified assertions, as statements of fact. We will return to this question below. We must first make one modification to our account of existential modal sentences, which spoils somewhat the neat parallelism that I have suggested between types of modality and orders of entities. We need to recognise, as standard treatments do, that there are frequent examples of existential modality involving not first- but second-order entities. I have in mind such examples as: (8) (9) (10) (11) (12)

These animals can bite. The value of shares can go down as well as up. The weather can be awful.(Palmer 1990: 107) This illness can be fatal. (Leech 2004: 82) He can be very tactless/helpful. (Huddleston and Pullum 2002: 184)

As many have pointed out, these sentences allow paraphrases with sometimes rather than some (e.g. These animals sometimes bite) and that we have quantification of situations over time. Put another way, what these sentences do is state that a class of situation has been known to occur or, in

Semantic ascent, deixis and intersubjectivity

63

the case of static situations like that in (10), to obtain. Thus alternative paraphrases of (8) and (10) might be (8a) These animals have been known to bite. (10a) It has been known for the weather to be awful. This E2 variant of existential modality, like the basic E1 usage, seems also to extend to necessity predicates, but it is harder to discern a full paradigm. A possible example is: (13) Why does he always have to get drunk when he goes to a party? (= He has never been known to go to a party and not get drunk) However, this example may also be construed as dynamic modality, with the agent-controlled interpretation ‘he can’t help getting drunk’. What is more significant for our purposes is that the be known to paraphrase also extends to the basic E1 examples: (14) Mauve roses/Tall Welshmen are not unknown. (=1/2) (15) Birds which are not bipeds are not known to exist. (=4) and so on. I should like to suggest that this particular type of paraphrase reveals important implications for the question of whether there is anything modal about existential modal sentences. As we have seen, all such sentences are factual assertions. They are statements, as the paraphrases show, of empirically ‘known’ facts about hypothetical, potential classes of entity (E1 or E2) which have (or have not) been found or experienced, and hence which may (or may not) be found or experienced in future. In terms of their communicative function, utterances of such sentences may be characterised as ‘didactic’ – they are frequent in teaching and in academic texts (cf. Huddleston 1971: 297–305) or they often count pragmatically as warnings: most of the examples cited are most easily contextualised as occurring in one of these categories. They seek to apprise the addressee of the existence or occurrence of potential classes of entity that they might think unlikely to exist or occur. Such utterances anticipate and seek to remedy perceived gaps in the addressee’s knowledge or they seek to correct the addressee’s perceived or anticipated misconceptions. In other words, they seek to re-

64 Keith Mitchell pair disparities between the speaker’s and the addressee’s states of knowledge. They involve what Stein (1992: 143) refers to (in dealing with a different modal phenomenon – see below) as ‘epistemic contrastiveness’: a contrast or discrepancy between two states of knowledge – here the gap between the speaker’s and the addressee’s current states of knowledge. (Let us note, however, that there are other possible types of epistemic contrast which we will examine below.) There is modality here but not in the usual narrow sense of the speaker wrestling with propositions whose truth is unknown to him, but in the sense that the speaker is taking account empathetically of what he perceives as doubt or ignorance on the addressee’s part. The modality we have here, then, is intersubjective rather than purely subjective – it involves the addressee’s as well as the speaker’s current state of knowledge. So we can perhaps summarise existential modality as follows: The speaker seeks to establish in the hearer’s mind the actuality of a potential class of entity: the known existence (or non-existence) of a class of E1, or the known occurrence (or non-occurrence) of a class of E2. There is a potential objection to be answered here. How is a sentence like (8) These animals can bite different from the unmodalised These animals bite? I seem to remember that when health warnings first appeared on cigarette packets they said Smoking can kill; now they say Smoking kills. I believe this is because now everyone is assumed to be fully aware of the health risk – all that is needed is a reminder. Previously there was assumed to be a potential discrepancy between the knowledge of the government’s health authorities and that of (at least some) smokers. I have just received a new university staff identity card accompanied by a letter which says: Take care of your card. Replacements can cost £6. I don’t think they are telling me that the replacement cost may vary, that some replacements cost £6 but others are cheaper. I think the can here is to be understood as a tactful way of saying ‘You may not realise this but we draw your attention to the fact that we charge £6 for replacements’.

5. Deixis In §1 I suggested that modality might be considered as ‘epistemic deixis’. Again here, Lyons’s three orders of entities will be relevant. Deixis may be defined as a basic mode of identifying entities in the universe of discourse with reference to features of the situation of utterance:

Semantic ascent, deixis and intersubjectivity

65

the participants (speaker and addressee), the place and the time (see e.g. Fillmore 1997). Standard treatments of the area accordingly recognise and describe three subtypes of deixis: person deixis, spatial deixis and temporal deixis, which are typically grammaticalised in personal pronoun systems, systems of demonstratives and tense systems respectively (and are also realised in certain lexical choices). I wish to suggest that there is an additional grammatical category which qualifies as deictic, namely mood/modality (in all its grammatical guises) which grammaticalises in various ways the distinction between whether a proposition is believed or known by the speaker to be true (and can be asserted as fact) or not (and falls short in varying degrees of being assertable). Just as systems of person and spatial deixis provide for the expression of proximity to, and remoteness from, the location of utterance, and just as tense systems enable speakers to express varying degrees of proximity/remoteness in relation to the time of utterance, so systems of mood and modality enable us to express how a proposition relates to the speaker’s and hearer’s state of knowledge and belief at the time of utterance – from the potentially true to the counterfactual. If E1s (physical objects, including persons) can be located by reference to the speaker/hearer and the physical space(s) they occupy, and if E2s (events and states) can be located by reference to the speaker’s/hearer’s time of utterance, why should it not be possible for E3s (propositions) to be located by reference to whether they are part of the speaker’s/hearer’s current state of knowledge? My emphasis on both speaker and hearer in the previous paragraph is important because, as is clear from standard accounts of deixis, the speaker often adopts the hearer’s viewpoint in personal, spatial and temporal deixis, producing instances of what has come to be called ‘empathetic deixis’. Here are some classic examples, presented without commentary because of space limitations: (16) Dentist: Is this the tooth that’s troubling you? Patient: No, that one’s OK; it’s the one next to it that’s painful. (17) (Mother to child) Daddy’s gone to live with another lady. (18) (Graffiti) Kilroy was here. (19) I’ll come and see you when you go into hospital. (20) (Note left on kitchen table) I’ve gone to the butcher’s – back by 11.30.

66 Keith Mitchell (21) (Card left in letter box) Your meter reader called at 9.35 am today but you were out. I argued above that what is modal about existential modal utterances is that the speaker empathises with the addressee’s state of knowledge – with the addressee’s ignorance, doubt, uncertainty with regard to a proposition whose factuality is clear to the speaker himself/herself. I believe this is needed to explain a number of other phenomena which are hard to accommodate in an account of modality that focuses entirely on the speaker. The remainder of this paper will be devoted to these.

6. Syntactic mood (clause-type) Epistemic deixis seems to be what the grammatical category of mood realises. The term mood is most typically associated with inflectional distinctions, such as the obsolescent contrast in English between indicative (realis) and subjunctive (irrealis). But it has also been applied, rightly in my view, (e.g. by Halliday 1985) to the syntactic contrast between what are more commonly termed ‘clause types’. The syntactic system of mood or clause type in English can be analysed as having five major terms. Each of the categories is associated with a characteristic use as an illocutionary act. Category of clausetype/syntactic mood

Characteristic use (illocutionary act)

Example

i. ii. iii. iv. v.

Statement Closed question Open question Exclamatory statement Directive

You are generous. Are you generous? How generous are you? How generous you are! Be generous.

Declarative Closed interrogative Open interrogative Exclamative Imperative

(Huddleston and Pullum 2002: 853) It is a well-known fact that the associations between syntactic category and illocutionary act are not strict one-to-one correlations, hence the term ‘characteristic’. The variations are comprehensively covered in Huddleston and Pullum (2002: 854ff). The simplified picture given above suffices for our present purposes.

Semantic ascent, deixis and intersubjectivity

67

What concerns us here is that the illocutionary acts most characteristically associated with clause types (ii-v) are modal in that only (i) is used for unqualified assertions. Clause types (ii-iv) are related to epistemic modality and (v) to deontic modality. Questions and exclamations are intersubjective epistemic illocutionary acts involving the negotiation of propositional knowledge and belief between speaker and hearer, whereas directives are deontic and involve the speaker negotiating with the hearer to bring about (or prevent) the occurrence of a situation. What I wish to emphasise again here is the involvement of both speaker and hearer, the element of negotiation. Indeed Halliday stresses this by referring to the function of (syntactic) mood within the grammar as the “interpersonal function”. It is significant that there are languages that have complex formal systems containing both traditional modal categories and illocutionary categories comparable to those expressed in English by syntactic mood: see for example the “knowledge and belief clitics” of Ngiyambaa (Donaldson 1980), listed in Palmer (2001: 16). These include Assertion, Categorical assertion, Counter-assertion, Exclamative, Ignorative and Imperative in addition to the more obviously modal Evidential, Irrealis, Purposive and others. A similarly mixed set of categories occurs in Tonkawa (Palmer 2001: 82). Palmer characterises the knowledge and belief clitics of Ngiyambaa as “at best, marginally modal”. I would disagree, and suggest that these interpersonal illocutionary categories are very centrally modal – indeed they are superordinate modal categories when seen in relation to more specific attitudinal categories involving notions such as necessity, possibility, volition, etc. There is a danger that English-speaking (and more generally, European) writers on modality start from the assumption that the most central modal meanings are those associated either with the set of modal auxiliary verbs, namely the specific categories I have just listed (see e.g. van der Auwera and Plungian 1998). The clause-type system of English is treated as something entirely separate, largely because it is a syntactic rather than an inflectional or lexical phenomenon, in spite of the fact that semantically the marked members of the system indicate that the propositional content of the clause is something not factually assertable, something whose epistemic status is unknown or unclear to the speaker. Huddleston, nevertheless, in the course of an exhaustive treatment of clause type, acknowledges that modality is a “closely related field” (Huddleston and Pullum 2002: 880).

68 Keith Mitchell To focus on just the epistemic illocutionary acts, non-modal declaratives used as statements are the unmarked, default option: they are unqualified assertions of fact, a speaker’s expression of a commitment to the truth of the propositional content. From an interpersonal viewpoint, the speaker additionally presumes that the hearer does not share this knowledge. Interrogatives used as questions are clearly marked and are the reverse of statements: the speaker does not know whether the proposition is true, but presumes that the hearer does know and seeks the hearer’s commitment. The epistemic and interpersonal nature of both statements and questions is thus fairly self-evident. The exclamative clause type, characterised by an initial degree-phrase containing what or how followed typically by SV word order is formally a curious hybrid of interrogative and declarative. It is used to make exclamatory statements, i.e. statements modified by an exclamatory component which “gives them a strong subjective quality”. They “express the speaker’s strong emotional reaction or attitude to some situation”, notably that some aspect of the situation is “remarkable” or surprising (Huddleston and Pullum 2002: 922). What is remarkable, as is indicated by the degreemodifiers what/how, is the unspecified degree of some property in the propositional content. And indicating that something is remarkable or surprising is to say something about its epistemic status: I perceive it as true but I find it hard to accommodate it into my existing knowledge/beliefs. I am wary of unqualified acceptance because the propositional content conflicts with my expectations determined by my present state of knowledge. I therefore distance myself from total commitment and use a clause-type that is not straightforwardly declarative. Here again we have “epistemic contrastiveness”, but this time the contrast is between the speaker’s new knowledge and his knowledge hitherto. There is nonetheless an interpersonal component in the form of an appeal (sometimes made explicit) for agreement or confirmation from the hearer Directives, the illocutionary act typically performed by uttering an imperative, are, as we have said, deontic speech acts. Directive is a superordinate term, subsuming various more specific speech acts designed to get the addressee to act: imposing obligations and prohibitions, granting permission and exemption, advising, inviting, etc. These more specific directives may be spelt out by using various modal auxiliary (and nonauxiliary) verbs. Using an imperative, on the other hand, expresses a plain directive shorn of any expression of necessity, possibility, volition, advantage, etc and neutralises these distinctions, leaving the specific force of the

Semantic ascent, deixis and intersubjectivity

69

directive to be inferred from context. For example, Bring along your partner may be construed either as obligation or permission, and Don’t bring a bottle may be a prohibition or an exemption. Semantically and pragmatically, therefore, this clause-type is the superordinate term of the system of deontic modality, which is usually associated exclusively with uses of modal verbs.

7. Auxiliary do : another modal auxiliary? This is an appropriate point to reconsider the auxiliary use of the verb do and its status among the auxiliaries of English. Since its main use is in distinguishing between clause types (and in expressing negation) and therefore in the performing of both deontic and epistemic illocutionary acts, there is a case for considering it as another modal auxiliary. This is by no means an original suggestion. Diver (1964: 330) includes do in his modal “system of hypothesis” and offers the following “scale of likelihood”: do must should may can

‘certain’ ‘very likely’ ‘more than likely’ ‘less than likely’ ‘possible’

However, Diver seems to be concerned here only with what is usually termed ‘emphatic’ do. We will return to this below. First, let us note that auxiliary do shares important grammatical properties with the modal auxiliaries: – Like them (with the exception of ought) it is an intransitive catenative verb that takes bare infinitival complements. The only intransitive verbs that take bare infinitival complements are the modal auxiliaries; even the transitive verbs that license such complements (causative make, let, have, perceptual see, hear, feel, plus experiential know) are all arguably modal in meaning – the causatives being deontic and the others being evidential.

70 Keith Mitchell – Do is morphologically defective in the same way as the modal auxiliaries in having only finite forms for past and present tense, except that it does (!) have a 3rd person singular present tense form in -s. – It cannot take a(nother) modal auxiliary as its complement. Like the modals it can take the non-modal auxiliaries be and have as complements, but only when it is used deontically, i.e. in imperative clauses. Diachronic evidence also suggests that do is closely related to the modals. According to Warner (1993: 221), the tendency to use auxiliary do, at least in interrogatives, became firmly established in the period 1475–1525, which is also when “the cluster of changes sharpening the coherency and isolation of modal group verbs” took place. Warner comments: “The coincidence of date here strongly suggests that developments in modals and in do are interconnected”. The epistemic and deontic nature of do is also related to its etymology. Non-auxiliary, lexical do is a transitive verb and is dynamic with an agentive subject. It is a verb which denotes an act. Auxiliary do, on the other hand, is intransitive and is compatible with both static and dynamic situations as well as with non-agentive subjects and is used in utterances where we are establishing whether something is a fact or not. The noun fact derives from the Latin factum, the supine form of the verb facere ‘do’. A fact is something ‘done’ or ‘achieved’, something that has taken its place in reality, in our experience, in the domain of what we know to be true. It is something viewed as accomplished, settled, established beyond dispute. The Shorter Oxford English Dictionary (SED) defines a fact as ‘something that has really occurred or is the case; hence, a datum of experience, as dist[inct] from conclusions’. The noun act also denotes something ‘done’ or ‘performed’ and derives from the supine form actum of the verb agere. However, whereas fact denotes a E3 and is what Lyons calls a third-order nominal, act in its basic use denotes only a E2. Interestingly, the noun act has an extended meaning of ‘(documentary record of) a legislative decision’, and two similar meanings are associated with the deverbal noun deed. But the related adverbs indeed and actually are similar to in fact in having ‘ascended’ to epistemic meanings: all three are used to reinforce or emphasise the speaker’s assertions. There are thus various reasons for viewing factual do as the unmarked member of the set of epistemic modal auxiliaries, contrasting with the non-

Semantic ascent, deixis and intersubjectivity

71

factual meaning of the others, which express what the SED refers to as ‘conclusions’ – estimations and inferences involving concepts such as necessity and possibility. Its unmarked, factual status perhaps explains why do, unlike the other modals and the subjunctive, retains the -s form for the 3sg present tense. Epistemic uses of auxiliary do include not only its use for denials and questions in negative and interrogative clauses but also its ‘emphatic’ use for what Stein calls ‘epistemic contrastiveness’, as in She does make mistakes. This could be uttered, with stressed does, as a counter-assertion, to disagree with an interlocutor’s previous utterance such as I don’t think she’s ever put a foot wrong. It may also be used to signal agreement with a previous utterance: A: She’s terribly error-prone. B: Yes, I agree, she does make mistakes. Or it may be used concessively, to acknowledge that the proposition is true before adding some mitigating information: She does make mistakes, but not very serious ones or She does make mistakes, but that’s not a reason for sacking her. (This use is similar to concessive may, dealt with below.) Or it might be uttered as an exclamation of surprise: That new secretary – she does make some pretty ghastly mistakes, doesn’t she? We should perhaps add that interrogatives and negative-interrogatives can also have the illocutionary force of exclamations: Boy, does she make mistakes! Doesn’t she make some ghastly mistakes! In all these there is an indication of the proposition being contrary to someone’s expectations – either the hearer’s or the speaker’s own. Diver (1964: 331) captures this insight by contrasting emphatic do and the plain unmodalised indicative (She makes mistakes) in terms of “a questioned certainty” versus “an unquestioned certainty”. Auxiliary do in imperatives, as we have said, behaves somewhat differently: both negative don’t and ‘emphatic’ do can be prefixed to any affirmative imperative clause structure – first- and third-person imperatives with let(‘s) as well as basic second-person imperatives. This applies even to clauses with be and to clauses with overt third-person subjects, where do/don’t (and not *does/doesn’t) are used. (22a) Do be quiet. (23a) Do somebody please shut the door. (24a) Do be warned – this stuff is potent. (25a) Do let’s invite John.

(22b) Don’t be so silly. (23b) Don’t any of you come complaining to me. (24b) Don’t be tempted to have another drink. (25b) Don’t let’s quarrel over this.

72 Keith Mitchell (26a) Do let me make myself plain.

(27a) If anyone objects, do let him say so now

(26b) Don’t let me worry you, but you ought to show that lump to a doctor. (27b) Don’t let there be any misunderstanding.

The negative imperatives (b) function as directives to refrain or desist from an act. The affirmative emphatic do-imperatives (a) function as insistent directives – requests, offers, invitations, suggestions, advice, etc where refusal or reluctance on the part of the addressee is encountered or anticipated. Since imperatives are characteristically used as directives, a deontic illocutionary act, factuality and knowledge of propositions (E3s) are not what is at issue here. What is being negotiated is the actualisation of a potential situation (E2): what we might think of as the conversion of an agendum into an actum.

8. ‘Quasi-subjunctive’ should In this section and the next, we will take a brief look at two uses of modal auxiliaries which are hard to explain without some recognition of the part played by epistemic contrastiveness and empathetic epistemic deixis: the epistemic use of ‘quasi-subjunctive’ should, and concessive may. By ‘quasi-subjunctive’ should is meant its highly grammaticalised use in complement clauses, where ought to cannot be substituted. There is a deontic (‘mandative’) use and an epistemic (‘evaluative’) use: (28) I insist/It is essential/I’m very keen that you (should) be present at the meeting. (29) I’m surprised/It’s surprising that he should feel offended. We shall be concerned here only with the latter use, as in (29), which occurs in the complement clause of a wide range of predicates of evaluation, for example. (30) It’s funny/strange/odd you should say that: I was just thinking the same thing myself. (31) I’m disappointed that he should have changed his mind. (32) It’s incredible that we should both have the same birthday.

Semantic ascent, deixis and intersubjectivity

73

(33) It shocks me that there should be such poverty in such a rich country. Many evaluative predicates, like those in (29-33), express the speaker’s difficulty in believing the proposition in the complement clause and the modal should indicates that the speaker distances himself from it epistemically. These predicates indicate that the propositional content is remarkable, surprising, and thus remote from the speaker’s expectations, so that the modality we see here is similar to that found in exclamative clauses. However, there are other predicates such as those in (34-37), for which this account does not suffice: (34) It’s normal that you should be feeling a bit tired. (35) It’s natural that she should be upset. (36) It’s understandable that he should have lost his temper. Here the propositions in the subordinate clauses are something that the speaker accepts as fact. I would like to suggest that the modal marking once again indicates that there is an interpersonal, inter-subjective factor operating here. The speaker is empathising with the hearer’s incomprehension of the situation described in the proposition, with his doubts about its rightness, and the utterance is designed to allay those feelings. These sentences have strong affinities with a subclass of what Dancygier (1998: 61) calls ‘non-predictive’ conditional sentences, and can indeed be paraphrased as such: (37) If you’re feeling a bit tired, that’s normal. (38) If she’s upset, it’s natural. (39) If he lost his temper, it’s understandable. Typically if-clauses carry a modal meaning: the speaker does not know whether the proposition is true (or in counter-factual conditions, knows that it is not true) but invites us to imagine that it is true – in order to spell out the consequences. In (37-9) something different is going on. The speaker himself accepts the proposition in the protasis as true but empathises with the hearer’s puzzlement. These sentences say something like ‘If you’re surprised/mystified that X is the case, let me reassure you’. A similar use of conditional sentences is found in:

74 Keith Mitchell (40) If I’m opening the window, it’s because it’s getting very stuffy in here. (41) If Mary is late this morning, it’s because she had a dentist’s appointment. The empathetic force of the protasis in (40) and (41) becomes clear if we expand if to the empathetically explicit If you’re wondering why... Palmer (2003: 6) argues that the reason why quasi-subjunctive should is used in sentences like (29–37) is because the propositional content is not asserted but presupposed: it is understood that both the speaker and the hearer accept the proposition as factual. Here Palmer follows Lunn (1995) who argues that non-assertion is what triggers the use of the subjunctive in Spanish, and that presupposition is one of three kinds of non-assertion. However, this does not square with standard accounts of presupposition. There are classic presupposition-triggers (see e.g. Levinson 1983: 179) which do not license the use of should in the complement clause: (42) I’ve just remembered/realised that it is (*should be) your birthday today. (43) I’m sorry to hear that your dog has (*should have) died. (44) I’m glad you’re (*should be) feeling better today.

9. Concessive may The use of may in such examples as (45) She may be very beautiful but she’s as thick as two bricks. (46) You may have finished your work but that doesn’t mean you can go home early. is again often cited as an example of a modalised sentence which is used to express a proposition that the speaker accepts as true. Palmer again explains this in terms of presupposition (2001: 31). I would prefer, however, to view this instance of modality in interpersonal terms as an expression of epistemic contrast. It is called ‘concessive’ because the speaker concedes, admits that the proposition is true. It has the meaning of ‘I grant (you) that ...’, which is a kind of epistemic permission: ‘I permit this to be a fact’. If there is reluctance on the part of the speaker it is not reluctance to accept

Semantic ascent, deixis and intersubjectivity

75

the proposition as true but a reluctance to accept it as the whole truth: there is more to say that is true and that is what the speaker goes on to do with the coordinated clause beginning with but. The paraphrase with nevertheless/nonetheless is instructive here: She’s very beautiful, I agree; nevertheless/nonetheless she’s as thick as two bricks. This signals that the second proposition is ‘no less true’ than the first. The truth of one proposition does not invalidate the truth of the other. This is clearly signalled by the French n’empêche in one translation of (45): Elle est sans doute très belle; n’empêche qu’elle est bête comme mes pieds. The verb empêcher, whose normal meaning is ‘prevent (the occurrence of a situation [E2])’ here takes on an epistemic meaning with a propositional complement [E3]. Concession is a part of negotiation. In (45) speaker and hearer are cooperating to negotiate a mutually satisfactory description of a person, composed of agreed propositions. This seems to be parallel to the negotiation of action: You may borrow my camera but please be careful with it.

10. Conclusion I have tried in this paper is to suggest a number of lines along which we might expand discussion of modality so as to make sense of certain phenomena that are hard to fit in to the apparatus of traditional accounts. I have suggested that Lyons’s three orders of entities might provide a framework for accommodating existential modality alongside deontic and epistemic modality. Also I have suggested that subjectivity, taken simply as speaker’s viewpoint alone, needs to be opened up into intersubjectivity, and that modality in actual use is essentially an interpersonal phenomenon involving the relationship between speaker and hearer. Again taking my cue from Lyons’s tripartite ontological distinction, I have suggested that it might be valuable to consider modality as ‘epistemic deixis’. And I have suggested that syntactic mood (clause type) deserves equal consideration alongside inflectional mood and the system of modal auxiliaries (and other modal lexemes) as a realisation of modal meaning. This in turn led to the suggestion that we might plausibly extend the modal auxiliary system to include do. This paper has been essentially speculative and programmatic. All of the topics I have touched upon require and deserve a good deal more investigation than I have been able to offer here and it would not be surprising if many of my suggestions proved unfruitful in the light of further research.

76 Keith Mitchell My purpose has been to provoke and stimulate – to sow a few seeds and fly a few kites in the hope of suggesting a broadening of the agenda for all of us working in the complex, fascinating and seemingly inexhaustible field of modality.

References Boyd, Julian, and James P. Thorne 1969 The semantics of modal verbs. Journal of Linguistics 5: 57–74. Dancygier, Barbara 1998 Conditionals and prediction: Time, Knowledge and Causation in Conditional Constructions. Cambridge: Cambridge University Press. Diver, William 1964 The modal system of the English verb. Word 20: 322–352. Grierson, Herbert J.C. 1944 Rhetoric and English Composition. Edinburgh: Oliver and Boyd. Halliday, Michael A.K. 1985 An Introduction to Functional Grammar. London: Edward Arnold. Huddleston, Rodney D. 1971 The Sentence in Written English: A Syntactic Study Based on an Analysis of Scientific Texts. Cambridge: Cambridge University Press. Huddleston, Rodney, and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Lakoff, Robin 1972 The pragmatics of modality. Papers from the 8th regional Meeting of the Chicago Linguistic Society. 229–246. Leech, Geoffrey N. 2004 Meaning and the English Verb. 3rd edition. London: Longman Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. Lunn, Patricia V. 1995 The evaluative function of the Spanish subjunctive. In Modality in Grammar and Discourse, Joan Bybee, and Suzanne Fleischman (eds.), 419–449. Amsterdam/Philadelphia: John Benjamins. Lyons, John 1977 Semantics 2 . Cambridge: Cambridge University Press. 1982 Deixis and subjectivity: Loquor, ergo sum? In Speech, Place and Action: Studies in Deixis and Related Topics, R.J. Jarvella. and W. Klein (eds.), 101–124. Chichester/New York: John Wiley.

Semantic ascent, deixis and intersubjectivity

1989

77

Semantic ascent: a neglected aspect of semantic typology. In Essays on Grammatical Theory and Universal Grammar, Douglas Arnold, Martin Atkinson, Jacques Durand, Claire Grover, and Louisa Sadler (eds.), 153–186. Oxford: Clarendon Press. Palmer, Frank R. 1990 Modality and the English Modals. 2nd edition. London: Longman. 2001 Mood and Modality. 2nd edition. Cambridge: Cambridge University Press. 2003 Modality in English: theoretical, descriptive and typological issues. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank Palmer (eds.), 1–17. Berlin/New York: Mouton de Gruyter. Stein, Dieter 1992 Do and tun: A semantics and varieties based approach to syntactic change. In Internal and External Factors in Syntactic Change, Marinel Gerritsen, and Dieter Stein (eds.), 131–155. Berlin/New York: Mouton de Gruyter. van der Auwera, Johan and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. Von Wright, George H. 1951 An Essay in Modal Logic. Amsterdam: North Holland. Warner, Anthony R. 1993 English Auxiliaries: Structure and History. Cambridge: Cambridge University Press.

Degrees of modality Raphael Salkie

1. Introduction This paper sets out a new account of modality as a typological category, and applies it to English.1 Previous studies have argued for a prototype approach to modality, but have not constructed such an account consistently. Section 2 sets out the need for a prototype analysis, and section 3 offers a set of criteria which define the core of the category of modality. We examine several English modal auxiliaries with respect to the criteria, concluding that some of them frequently fail to meet the criteria and are therefore not core members of the category – in other words, they have a low degree of modality. Section 4 locates modality within a wider framework. In section 5 we revisit two problems in the analysis of English modals, arguing that the proposed framework sheds new light on them. Section 6 criticises the alternative analysis put forward by the Cambridge Grammar of the English Language (Huddleston and Pullum 2002). We conclude that the framework needs to be tested on other languages, but that it appears to be useful for English.

2. Why modality is elusive What do linguists mean by the term “modality”? In the literature we find four different approaches: (a) A broad interpretation. Palmer’s extensive survey, for instance, says that “modality is concerned with the status of the proposition that describes the event” (2001: 1). The weakness of this approach is that the description is very vague, so that what it includes and excludes is unclear. (b) A narrow definition. Van der Auwera and Plungian, in their typological study, say that they “propose to use the term ‘modality’ for those semantic domains that involve possibility and necessity as paradig-

80 Raphael Salkie matic variants”, a definition which they describe as “relatively restricted” (1998: 80). This strategy has the virtue of starting with domains that virtually everyone recognises within modality; unfortunately, though, it excludes some semantic domains, such as volition and evidentiality, which many scholars would also want to include. (c) A list of sub-types of modality. An example is Downing and Locke, who write: “modality is to be understood as a semantic category which covers such notions as possibility, probability, necessity, volition, obligation, and permission” (2002: 382). This gives an indication of the domain, but fails to set clear boundaries. (d) Avoidance of the problem by focusing on a set of expressions, such as modal verbs in English; modality is then taken to include whatever these expressions mean. Coates (1983), for instance, analyses the meaning and use of English modals in corpus data, without attempting any definition of the term modal. The difficulty with the domain of modality is that there is broad agreement about some central members of the class, but disagreement about some of the candidates for inclusion. It is not possible to provide an acceptable definition with clear boundaries: instead, we need a clearly defined core but fuzzy boundaries. This suggests that a prototype approach is more promising: we should specify the criteria which demarcate the core, and allow for peripheral members of the category which do not meet all the criteria. Items that meet all or most of the criteria will be said to have a high degree of modality, while those in the periphery will have a low degree. It is common practice in typological studies to treat grammatical categories as having a core and a periphery, as noted by standard handbooks such as Comrie (1989: 37–38 and 106–110) and Croft (1990: 66–67 and 124– 127). The prototype strategy was explicitly proposed for English grammar by Huddleston (1984: 72), and is adopted in a number of recent grammars of English, such as Greenbaum (1996: 92) and Biber et al. (1999: 59). It is applied to modality in English by Huddleston and Pullum (2002), who say that “modality … is not sharply delimited or subdivided, so that we shall need to make reference to the concept of prototypical features and to allow for indeterminacy at the boundaries of the categories” (2002: 172). We shall argue below that Huddleston and Pullum’s specific analysis is problematic, but their general strategy seems to be correct.

Degrees of modality

81

The prototype approach to grammatical categories is sometimes linked with work in cognitive psychology by Rosch (cf. Rosch and Lloyd 1978). More commonly, though, a commitment to prototypes is primarily a recognition of the need to allow for different degrees of category membership, and this is the situation with the work in typology and English grammar just cited. The consequences of adopting a prototype approach will therefore vary from category to category, depending on the criteria for membership. There is no reason, for example, to expect core members of a category to be more frequent, acquired earlier by children, or less liable to change over time. What we would expect is for the criteria to have implications for the analysis of individual languages, and for cross-language comparison.

3. Criteria for modality We propose four criteria for the typological category of modality, the first two quite traditional, the second two more problematic. The criteria are semantic and pragmatic, the assumption being that cross-linguistic categories have to be characterised in this way. Other properties of modal expressions, particularly their morphosyntactic behaviour, will vary from language to language.

3.1. Possibility and necessity We shall adopt the proposal cited above from van der Auwera and Plungian as a starting point, treating it as just one criterion among others, rather than the only one. As they note, the view that possibility and necessity are central to modality is widespread and traditional. Their position is supported by Kiefer, who claims that “the relativization of the validity of sentence meaning to a set of possible worlds” is “the essence of modality” (1994: 2515). If we take this criterion seriously, then it excludes from the core the category of dynamic modality, proposed by some analysts to cover ability can and various uses of will in English (Palmer 2001: 9). These uses of can and will do not invoke other possible worlds: we shall look at some examples in sections 3.6 and 5.1. Since several scholars have suggested that dynamic modality may not be a type of modality at all (e.g. Palmer

82 Raphael Salkie 1990: 37; Papafragou 1998: 2), assigning dynamic uses of modals to the periphery seems reasonable.

3.2. Epistemic and deontic At the heart of many analyses is the distinction between epistemic modality, which refers to “judgements about the factual status of the proposition” (Palmer 2001: 8), and deontic modality, involving attitudes to “acts performed by morally responsible agents, e.g. obligation and permission” (Papafragou 2000: 3). It is natural to take these two domains as our second criterion: items which express either of them will have a high degree of modality, while items which do not will be nearer the periphery. This criterion will assign to the periphery dynamic modality (which already failed the first criterion) and alethic or logical modality (Lyons 1977: 791). Future time uses of will also fail this criterion: we will look at some examples in section 3.6.

3.3. Subjectivity This term, introduced by Lyons (1977: 797), has been used in several different ways in the analysis of modality, and remains controversial (cf. Verstraete 2001, Timotijevic, this volume). Verstraete conceives of subjective modality as bringing into existence “a particular position of commitment [by the speaker – RS] with respect to the propositional content of the utterance” (2001: 1517). He calls this type of commitment “modal performativity”, distinguishing it from the traditional use of the term performativity which links it with illocutionary force (and for which he proposes the term “interactive performativity”). Let us suppose that core instances of modality are subjective in this sense, making this our third criterion for modality. In a recent paper, Narrog (2005: 169-171) argues that subjectivity is not a useful basis for defining modality because it is difficult to define precisely, and because it does not apply to all uses of items which are normally taken as core instances of modality. If subjectivity was the sole basis for characterizing the category of modality the latter criticism would be a reasonable one, but in the prototype framework proposed here it is just one criterion among others. Including subjectivity reflects the views of many

Degrees of modality

83

scholars who have studied modality; using it alongside other criteria avoids giving it too much weight. We return below to Narrog’s first criticism. Verstraete is one scholar who explicitly argues that uses of English modal verbs vary with respect to subjectivity: he argues that epistemic modals are always subjective, dynamic modals are never subjective, while deontic modals sometimes are and sometimes are not (2001: 1525). If we take subjectivity as a criterion for modality, then epistemic modals in general will have a higher degree of modality than dynamic ones, with deontic modals in between. Consider these authentic examples, of which the first two and the fifth are taken from Verstraete (2001): (1)

…on my return from Mogadishu last week Amnesty confirmed that the preacher for whom it had campaigned had actually been released some time in nineteen eighty-nine. The flood of letters must have had some impact after all.

(2)

You seem to be seeking to destroy yourself in some way, but you must not include me in your plan of action.

(3)

Sometimes the hills were so steep that, despite our driver's haste, the horses could only go slowly. I wished to get down and walk up them, as we do at home, but the driver would not hear of it. “No, no,” he said. “You must not walk here. The dogs are too fierce.” (Bram Stoker, Dracula)

(4)

Macmillan English Dictionary for Advanced Learners The most up-to-date dictionary CD-ROM for learners of English (…) I've been unable to install the CD-ROM on my computer. What have I done wrong? (…) If you are using Windows® 2000 or Windows® XP, only users in particular groups can install programs. For Windows® 2000, you must be a member of the Administrators group or the Power User group to install the CD. For Windows® XP, you must be a member of the Computer Administrators group to install the CD. (User Guide)

84 Raphael Salkie (5)

But to reach orbit an object must accelerate to a speed of about 17,500 miles per hour (28,000 kilometers per hour, called satellite speed or orbital velocity) in a horizontal direction; and it must reach an altitude of more than 100 miles (160 kilometers), in order to be clear of the atmosphere.

Verstraete argues that epistemic uses of must, as in (1), always involve the speaker’s commitment to “the propositional content of the utterance”, while for deontic uses this varies. It appears that the speaker is the source of the obligation in (2), but this is only partially the case in (3), where the constraint is a result of the circumstances, and the driver’s words are somewhere between reporting an obligation and giving advice. In (4) the User Guide is endorsing an obligation imposed by a third party, rather than by a social or ethical norm. Cases like (5) are more peripheral still: the necessity here derives from the laws of physics and the speaker is simply reporting it. We represent the four instances in table 1. Table 1. Different uses of must Example 2 3 4 5

Deontic (social or ethical norm)

X X

Subjective
X
X

We now take a closer look at subjectivity, responding to Narrog’s point that it needs to be made more precise. As defined by Verstraete it is clearly a pragmatic notion, since it crucially involves the speaker. A possible refinement of our third criterion would be to add to it as a statement about pragmatics: modal expressions which involve pragmatics will have a higher degree of modality than those which do not. Specifically, we propose that core instances of modality involve “primary pragmatic processes” in the sense of Recanati (2004: 21). These are processes which are required for a sentence to have a complete propositional meaning in a particular context: the clearest examples involve indexical expressions such as personal pronouns and tenses, which must have their reference assigned from the context. A sentence such as I live in Belgium only has a full propositional meaning after the application of the pragmatic processes of assigning a reference to the subject pronoun and the time of speaking. Recanati calls this type of pragmatic process saturation, because it involves filling in a

Degrees of modality

85

slot in the meaning of a sentence which would otherwise be empty and therefore uninterpretable. Any process which involves identifying the speaker will be a primary pragmatic process in this sense. Secondary pragmatic processes, in contrast, involve inferences or other extensions from the propositional content: they have a much wider range and are typically optional.2 It is evident that core instances of modality, such as (1) and (2) above, involve a similar process of saturation, because they require the speaker to be identified in order to be interpretable. Peripheral instances like (5) do not require saturation in order to interpret the modal. All three examples involve the modal must, however, which highlights a key difference between our analysis and the framework put forward by Papafragou (1998, 2000). She makes a distinction between may and must, which in her system always involve a primary pragmatic process (specifically, saturation), and can and should, which never do (1998: 14; 2000: 43). To analyse example (5), Papafragou would have to argue that it is analysed in a parallel way to (2): just as in (2) we need to invoke an (unspecified) social or ethical norm, in (5) we need to bring into play the domain of laws in physics. So far, we agree. Papafragou would also have to argue that invoking the domain of physics involves a primary pragmatic process of saturation which includes assigning a referent to the speaker. There is no reason to accept this, in my view: I would argue that sentence (5) fails the pragmatics part of the subjectivity test. In the framework proposed here, unlike that of Papafragou, modals are not tested on the criteria as isolated words, but in specific sentences. A modal may pass the subjectivity test in some uses, but fail it in other instances. If we refine our third criterion in this way, the question arises of the relation between the modal expression and the rest of the sentence that contains it. Many studies of modality simply take for granted that a distinction has to be made between the (pragmatic) modal part of an utterance and the (semantic) residue. Various ways of representing this distinction have been proposed.. In generative studies, the distinction is often made in the syntax (cf. Wurmbrand 1999, Butler 2003), while other work distinguishes between the modal expression and “the propositional content of the utterance” (Verstraete 2001: 1517, cited above), or assumes a distinction between the modal expression and the “embedded proposition” Papafragou (2000: 43) as part of a semantic and pragmatic analysis. In some cases it is easy to recognise this division, particularly where negation can apply either to the modal or the proposition:

86 Raphael Salkie (6) (7)

You must not go. (p is negated – “it is necessary that you not go”) You need not go. (modal is negated – “it is not necessary that you go”)

In other instances, the distinction is less clear: (8) (9)

John can’t swim. There won’t be any trouble at the next match.

Here the negation seems to have scope over the whole sentence, since we can paraphrase (8) and (9) as: (8') (9')

It is not the case that John can swim It is not the case that there will be trouble at the next match

Another aspect of our third criterion, then, is whether there is a sharp division between the meaning of the modal expression and the meaning of the rest of the sentence; that is, whether or not there is a complete proposition which can be identified separately from the modal. Summarising, the third criterion has three components: Subjectivity (i) commitment by the speaker (ii) primary pragmatic processes. (iii) a sharp distinction between the modal expression and the propositional content. It is hard to decide whether these are three separate criteria or so closely connected that we should take them together. For ability uses of can, the three criteria yield the same result. We have already noted that “dynamic” can ranks low on part (iii); for this use of can we can accept Papafragou’s claim that no primary pragmatic process is involved, so it fails (ii); and it is easy to see that speaker commitment is not involved in interpreting ability can. In this case, then, the three criteria all work in the same direction, consigning ability can to the periphery. We return to can in sections 3.6 and 5.1 below. For will, part (ii) of the subjectivity criterion has to be treated separately from the other two. Papafragou does not discuss will, but in its straightforward future time uses like (9) it lacks speaker commitment and forms part

Degrees of modality

87

of the proposition (on the reasonable assumption that time reference is part of the truth-conditional content of a sentence, since John will arrive has different truth conditions from John arrived); hence it fails parts (i) and (iii). Future time will passes part (ii), however, since time reference is indexical, but this is not the reason why core instances of modality pass part (ii): it just so happens that primary pragmatic processes are involved in time reference. Thus future time will meets one part of the third criterion but fails the other two, which suggests that they should in fact be regarded as separate.

3.4. Extremes of the modality scale Our final criterion takes account of the fact that modal expressions can often be located on a scale. This is clearest for epistemic examples, where the factual status of a proposition may be judged to be just possible at one extreme, or highly certain at the other. In between we find expressions of probability or moderate certainty, for example: (10) John has possibly arrived by now. (11) John has probably arrived by now. (12) John has almost certainly arrived by now. The two extremes have well-known logical properties with respect to negation. They also have pragmatic properties which are shared with other scalar expressions (cf. Levinson 2000: 83). The logical and pragmatic properties of modal expressions at intermediate points on the scale, such as probably in (11), are more variable. Let us tentatively take location near one of the extremes of the scale as a criterion for modality. We thus treat may and must as near the core. The English auxiliaries can, could and might also meet this criterion, as does need at the opposite extreme. As for will, it is close to the extreme occupied by must in its epistemic use, as in this example: (13) John will have arrived by now. In its other main uses, however, such as future time (9), propensity (14) and volition (15), there is no relevant scale on which to locate the meaning of will:

88 Raphael Salkie (14) They’ll often go for days without speaking to each other. (15) Will you please open the door? In these uses, will fails to meet the fourth criterion. It could be argued that the fourth criterion follows from the first one, since possibility and necessity are extreme points on a scale. We shall keep it separate here, however, in the hope that further research may clarify this question.

3.5. Summary of the criteria Core members of the modal category meet these four criteria: A. They express possibility or necessity. B. They are epistemic or deontic. C. They are subjective, involving (i) commitment by the speaker. (ii) primary pragmatic processes. (iii) a sharp distinction between the modal expression and the propositional content. D. They are located at one of the extremes of a modal scale.

3.6. Applying the criteria Consider now some examples in relation to the criteria (those labelled BNC are from the British National Corpus): (16) I don’t know for sure but there may be milk in the fridge. (17) Nursing and medical staff may also need psychological support to cope with the intense nature of the treatment and the uncertainties in outcome. (BNC EE8 800) (18) I can track him, and he won't know. (BNC CM4 2289) (19) It can be cold in Stockholm. (20) An analysis of the results should shed light on the workings of the Northern Ireland labour market. (BNC HJ0 18810)

Degrees of modality

89

(21) Your essay is an argument, and you should present different sides of that argument. (BNC HXH 1408) Example (16) meets all the criteria, whereas (17), which does not express speaker uncertainty but generalises over the members of the class of nursing and medical staff, fails criterion (Ci) because there is no special speaker commitment. It also fails criterion (Cii), because unlike (16) there is no need here to invoke the knowledge of the speaker in order to interpret may: to interpret (17) we simply have to assume certain properties of nursing and medical staff, exactly as we do in interpreting the non-modal Nursing and medical staff also need psychological support. However (17) passes (Ciii), since the proposition can be negated separately from the modality (that is, Nursing and medical staff may also not need psychological support means “It is possible that nursing and medical staff do not need psychological support”). We conclude that (17) has a slightly lower degree of modality than (16). Turning to (18), can here has a low degree of modality because it fails criteria (A), (B) and all of (C) (admittedly this example involves reference to the speaker, but that is due to the first person pronoun, not the modal). Example (18) fails criterion (A) because it does not involve reference to other possible worlds. Some analysts have captured this insight by including “factuality” in their analysis of can (cf. Papafragou 2002: 43; Larreya and Rivière 2005: 89), but thereby they are effectively conceding that this use of can fails criterion (A), since factuality is the very antithesis of our first (and least controversial) criterion for modality. In short, this common use of can fails the first three criteria for modality, and thus seems to have a very low degree of modality. Against this claim one might give these arguments: (a) can shares the morphosyntactic properties of auxiliaries like may, must, etc. (b) there are semantic similarities between (non-modal) ability and (modal) possibility. (c) can and may are similar in other uses; for example they are both used for permission. (d) In negative and interrogative contexts, can has an epistemic use (e.g. That can’t be true and Can that really be my brother?)

90 Raphael Salkie Argument (a) is only relevant if it can be demonstrated that members of this morphosyntactic class of items share semantic properties – and much of this paper has argued that this is not the case. Argument (b) is more important, and it is clear that in many instances, can is used in English with a more general sense of possibility rather than just ability. My view is that these more general uses are pragmatic inferences from the ability sense, and do not indicate that can meets criterion (A). On argument (c), Papafragou (2002: 58) gives a subtle analysis of differences between can and may in their permission uses which follow from the factual nature of can versus the wider domain of evidence invoked by may: as just noted, this effectively concedes that permission can is not modal (in our terms, it fails criterion A). As for the fourth argument, we can again cite Papafragou (2002: 78–79), who gives reasons against treating such examples as epistemic. Example (19) shares all these properties of (18), except in respect of criterion (Ci). There is an element of subjectivity in (19): by choosing not to say It is cold in Stockholm, the speaker indicates that her utterance is based on previous knowledge about that city, not current knowledge of its temperature at the time of speaking. Compare It may be cold in Stockholm, a prototypical epistemic judgement like (16) which is a statement about the likelihood of the actual temperature being low, either now or in the future. Hence (19) is a weak assertion which stays in the realm of potentiality, and expresses a low degree of speaker commitment to the proposition It is cold in Stockholm. This gives (19) a higher degree of modality than (18), but a lower degree than core instances of epistemic may. Turning now to should, most analysts would treat (20) as epistemic. Note, however, that in prototypical instances of epistemic modality like (16), a body of evidence is brought to bear on the likelihood of a proposition being true. The evidence normally includes general facts, not just the facts of the specific situation expressed by the proposition: in (16), for instance, the evidence might include knowledge that fridges often contain milk, or that on previous occasions milk was located in this fridge. In (20), the body of evidence does not include such general facts, but simply the facts of the situation in question. The effect of using should in (20) is to make a prediction about real future events which is less confident than a prediction using will in its peripheral future time use. This example does not have the crucial generalised context that is characteristic of core epistemic modality. Papafragou claims that should in all its uses “relies quite heavily on the sort of structured knowledge humans typically possess about the normal

Degrees of modality

91

course of events” (2000: 62). She proposes that this is a restriction on the body of evidence that can be brought to bear on sentences containing should, and that the semantics of should is otherwise identical to that of must. While the claim about the normal course of events is correct, I would argue that this is not a small difference between should and must but a fundamental one. If so-called epistemic should in fact stays firmly in the real world, then we are not dealing with the domain of possibility and necessity at all: as the quotation from Kiefer in section 2.1 makes clear, these domains invoke other possible worlds. Thus (20) fails criterion (A), and we analyse it as saying “In the normal course of events, an analysis of the results will shed light on the workings of the Northern Ireland labour market”, yielding the implication of a moderately confident assertion. Turning now to (21), we can argue that this example too fails criterion (A). Its meaning too refers to “the normal course of events”: its construal as imposing a moderate obligation comes from the interaction of that meaning with the context. Many of the relevant aspects of the context will be of the same kind that results in a deontic sense of sentences with must; but should arrives at its obligation interpretation via a different route from must. We revisit these matters in 5.2 below.

4. Modality and irrealis A characterisation of modality should situate it in relation to other grammatical categories. In this section, we make some suggestions about this wider picture, drawing on Palmer (2001). Suppose we take as the default type of sentence a factual description of an event or state of affairs in front of the speaker’s eyes. We can use the general term realis for this type of sentence, and any sentence which departs from the default will be called irrealis. Languages which make this distinction in their grammar will be said to instantiate the category Reality Status. For languages such as Manam (Papuan – Palmer 2001: 147–148) this top-level distinction may be the only expression of Reality Status: every finite verb has to be specified morphologically as realis or irrealis. Exactly where a language draws the line seems to vary: Manam treats past, present and habitual as realis, while future, imperatives, counterfactual and sequences of habitual are marked as irrealis. Palmer states that simple past and present time are always realis (2001: 168), but that some semantic domains appear as realis in some languages and irrealis in others. An example is future, which is irrealis in

92 Raphael Salkie Manam but comes under realis in Caddo (Caddoan, Midwestern US – Palmer 2001: 169). This combination of an invariant core and a fluctuating periphery argues for a prototype approach to Reality Status, like the one we have adopted here for modality. Within the irrealis category, we propose that a language can choose one or more of three subcategories. A. The sentence is asserted or not asserted. We shall call this category mood, with a binary division between indicative and subjunctive (cf. Palmer 2001: 3–4) B. The sentence can be marked as one whose truth is not in doubt, or marked to indicate judgements based on hearsay, personal experience, etc. The usual term for this category is evidentiality. C. A language can refer only to the real world, or can invoke other possible worlds. This is the (first criterion for the) category of modality. We thus treat modality as a sub-category of irrealis, within the top-level category of Reality Status. It is useful to contrast this with the different taxonomy proposed by Narrog (2005). He first proposes that terms used in the literature, such as “(realis vs.) irrealis”, “factuality” and “validity”, are essentially equivalent, and indicate “whether a state of affairs is ‘actual’ or ‘existent’ at a specific point or interval of time” (2005: 183). He gives reasons to choose “factuality” as the best term, and defines modality as follows: Modality is a linguistic category referring to the factual status of a state of affairs. The expression of a state of affairs is modalized if it is marked for being undetermined with respect to its factual status, i.e. is neither positively nor negatively factual. (2005: 184)

Narrog thus equates our category of Reality Status with modality. I think that there are two problems with this system. Firstly, it takes one attribute of modality and makes it the only defining criterion, whereas there are good reasons for taking more than one criterion. Secondly, it fails to capture the importance of real versus possible worlds as distinctive of modality, as opposed to the wider realis-irrealis distinction in languages like Manam. The notion “possible worlds” is crucial to another difference between Narrog’s system and the one proposed here. He argues that evidentiality is

Degrees of modality

93

a type of modality, giving as an example the evidential use of the German modal verb sollen: (22) Er soll bei einem Unfall umgekommen sein. ‘He allegedly died in an accident’ Narrog comments: The factuality of the proposition (He died in an accident) is undetermined, just as it is in the case of epistemic modality. The difference is that while in epistemic modality the proposition is undetermined with respect to its factuality relative to the world of knowledge and beliefs of the speaker, with evidentiality it is undetermined relative to sources of information other than the speaker. (2005: 188)

My argument against this is that examples like (22) do not invoke other possible worlds, whereas epistemic modals do. If I say He may have died in an accident, I am invoking other worlds in which He died in an accident is not true. The evidential in (22) does not invoke other possible worlds, but takes a view about the real one. Narrog is right to point out similarities between evidentiality and modality, but conflating them goes too far.

5. Some consequences of the prototype approach We have argued that the English auxiliaries can, should and will often have low degrees of modality. This section revisits two controversies in the analysis of English modals and attempts to show that the approach proposed here sheds new light on them. The first involves can in contrast with may, and the second compares should and must. We intend to deal with will in a separate paper.

5.1. Can, may and the real world Leech (2004: 82–83) examines these two sentences: (23) The road may be blocked. (24) The road can be blocked.

94 Raphael Salkie

Leech says that (23) expresses “factual possibility” while (24) is an instance of “theoretical possibility”: example (24), he claims, “describes a theoretically conceivable happening”, whereas (23) “feels more immediate, because the actual likelihood of an event’s taking place is being considered”. He contrasts two likely contexts: (23') The road may be blocked by floodwater (“that possibly explains why our guests haven’t arrived” – dialogue between husband and wife expecting visitors). (24') The road can be blocked by police (“and if we do this, we might intercept the criminals” – said by one detective to another). What Leech says about the different contexts is exactly right, but his explanation is wide of the mark. We have argued above that “factual modality” is a contradiction in terms; and we have also accepted the claim in earlier work that can has factuality as part of its meaning. How does Leech come to put things the other way round, treating may as factual, rather than can? The reason is that epistemic may takes a proposition (The road is blocked) and expresses a judgement about the likelihood of this proposition being true in the real world, as compared to other possible worlds. With can, on the other hand, no other possible world is invoked: example (24) (in a context like 24') contains a single proposition about a current capability of the police. Leech seems to consider that the capability in (24) is further removed from the real world than the comparison with other possible worlds in (22). That is not, however, the relevant point.3 In (24) the capacity exists in the real world and no other world is invoked, whereas (23) compares the real world with other possible worlds. Thus (24) has a high degree of modality (it meets all the criteria) while (23) is marginal (it fails all of them). (For more on Leech’s methodology, and a comparison with other analyses of can and may, see Salkie 2001: 578–580).

5.2. Should and must – Rivière’s problem Rivière (1981) discusses these examples: (25) You live in Oxford, you must know Professor Fen then.

Degrees of modality

95

(26) You live in Oxford, you should know Professor Fen then. (27) You know Professor Fen, you must live in Oxford then. (28) *You know Professor Fen, you should live in Oxford then. [OK in obligation sense] (29) *John is a brilliant student, he must pass his exam easily. (30) John is a brilliant student, he should pass his exam easily. The asterisk on (28) indicates that should here cannot have the sense it has in (26), where it appears to be a weaker version of epistemic must, but only the sense of “you are obliged to live in Oxford”. The main part of Rivière’s problem is to explain why (28) is not acceptable in this sense, while (26) is; a secondary part is to explain why (29) is unacceptable in contrast to (30). I have argued in earlier work that Rivière’s proposed solution is not workable (Salkie 2002: 85–87). In the framework put forward here, should has the sense of “in the normal course of events”, and has a low degree of modality since it fails criteria (A), (B) and (D). In the right pragmatic context, we interpret “normal course of events” as carrying the stronger sense that one event doesn’t just follow another but is the normal consequence of another. The word then in (26) invites us to apply this stronger interpretation here. Thus (26) does not express an epistemic judgement but a statement about a normal consequence of living in Oxford. This is very different from (25), which expresses a judgement of the likelihood of the proposition You know Professor Fen being true, given the evidence in the context (which includes the first clause). Hence should and must are both possible in this context, but for different reasons. In (27), the speaker once again makes an epistemic judgement, but the crucial example (28) cannot express the notion “normal consequence”, because living somewhere is not normally a consequence of knowing someone. Hence (28) cannot be interpreted as parallel to (26), and another sense of should is required for the sentence to be acceptable. Turning to the secondary part of Rivière’s problem, we can easily interpret (30) in terms of “normal course of events”. As for (29), this seems to be unacceptable because of a lexical restriction on epistemic uses of must in English, which cannot normally refer to future time. (The fact that epistemic may, and translation equivalents of must in French and German, are

96 Raphael Salkie not so restricted suggests that we are dealing here with an idiosyncrasy of must in English). The framework set out in this paper thus enables us to find an alternative solution to Rivière’s problem.

6. Degrees of modality in other work In his paper on the future tense controversy, Huddleston (1995: 421) introduces the notion that some uses of English modals have a low degree of modality, because the additional element of meaning which the modal adds to the sentence is small. The notion is amplified in the Cambridge Grammar of the English Language (Huddleston and Pullum 2002 – henceforth CGEL). This section tries to show that the notion as they use it is incoherent, because it groups phenomena which have nothing in common, and redundant, adding nothing to an analysis of modality. CGEL proposes that degree of modality “has to do with the extent to which there is a clearly identifiable and separable element of modal meaning” (179). This sounds similar to our criterion (Ciii), but in developing the idea, CGEL uses only one test: whether there is an appreciable difference in meaning between pairs of clauses where one member is unmodalised, the other modalised. They give four types of example, the first one involving may, which they claim adds a distinct element of meaning in (32) compared with (31): (31) They know her. (32) They may know her. In examples like (33) and (34), however, the modal adds little to the meaning: (33) Strange as it seems, I know you. (34) Strange as it may seem, I know you. (CGEL, 179) There is hardly any difference between the first clause of (34) and its nonmodal counterpart in (33), so may is said to have a low degree of modality in (34). In their second type of example, CGEL applies the same reasoning to should, which in their view has a high degree of modality in (36), where

Degrees of modality

97

it is clearly distinct from (35), but a low degree in (38), which barely differs from (37): (35) Your passport is in the drawer. (36) Your passport should be in the drawer. (37) It’s odd that he is so late. (38) It’s odd that he should be so late. Before commenting on these examples, we should note that similar data with can are analysed in Kjellmer (2003), who observes that with performative verbs, verbs of perception and verbs of mental processes, can appears to add little to the meaning: (39) Detective Superintendent Mike Morgan, leading the inquiry, said: ‘I can assure you he is not a suspect.’ [cf. I assure you] (40) I can just see a soldier beating a man with a stick. [cf. I just see] (41) I don’t think they particularly enjoy the reading log [and] to a certain extent I can empathize with that. (Kjellmer 2003: 164-6) [cf. I empathize] Kjellmer argues that can does in fact add to the meaning here, but the important point for our purposes is his demonstration that this use of can is mostly restricted to first person singular constructions and to particular verbs (2003: 149–153). What we have here is a type of partially fixed idiomatic construction with a slot that can only be filled by a limited class of lexical items (compare “It was a(n) X’s dream”, where X usually has to be an agent noun like promoter or scientist). Returning to the “X as it may seem” construction in (34), there is little doubt that it is idiomatic in the same way. Evidence for this comes from the 48 instances of “X as it may seem” in the BNC. In nine of these examples X is the word strange, and a further 21 contain an adjective which is similar in meaning, most commonly surprising, incredible and unlikely. In 11 of the remaining examples the adjective has a negative connotation (daft, arduous, haphazard, etc), or is in a negative context (e.g. not so easy as it may seem), leaving only three instances with an adjective such as sensible in a neutral or positive context. Examples like (38) are not as fixed as this construction with may, though this near-vacuous use of should is restricted to subordinate clauses

98 Raphael Salkie which complement evaluative predicates like odd, strange, etc. and their positive counterparts like natural and interesting. Furthermore, both the examples with supposedly low-degree may and those with should are marginal in other ways: they are formal, rather archaic, and rare except in certain genres. It would not be worth inventing a special term for these marginal examples unless there was a reason to apply the term also to central uses of modals. These uses of may and should are similar to the reduction of meaning sometimes called “semantic bleaching” which often occurs when a content word evolves into a function word. They should be dealt with under this heading, rather than as a reduction in meaning that is specific to modal expressions. CGEL gives a third type of example, arguing that low degree modality also applies to some core uses of will: (42) She is one year old tomorrow. (43) She will be one year old tomorrow. (CGEL: 179) They note that “there is no effective difference” between (42) and (43) (CGEL: 190). These examples are very different, however, from those involving may and should. In the first place, (43) is not marginal or idiomatic. More importantly, in (34) and (38) the modals have lost some of the meaning that they have in their central uses, but here the point is surely not that will has lost some of its usual meaning: rather, it is the present tense in (42) which under certain conditions has acquired a future interpretation. The small difference between (42) and (43) arises because the non-modal (42) has taken on some of the futurity meaning which is usually expressed by will. If we wanted a term for this we should call it “acquired modality” of the present tense, not “low degree modality of will”. The last type of example that CGEL uses to illustrate low degree modality again involves will, and is illustrated in (44): (44) Ed is Tom’s father and Tom is Bill’s father, so Ed will be Bill’s grandfather. CGEL notes, correctly, that “the evidence for [the conclusion] is so strong that I could equally well have used unmodalised is” (2002: 189). They give a parallel example in the section on epistemic must, and make a similar remark about the small difference between the modalised and unmodalised version (2002: 181). Here again, though, we surely have a very different

Degrees of modality

99

type of phenomenon from the other three examples. When there is a small difference between epistemic must / will on the one hand, and unmodalised assertions on the other, the reason is simply that in contexts like (44), claiming “p is almost certainly true” is not very different from just saying “p”. In other words, what CGEL calls “high strength” modality (near the necessity end of the modality scale) is often close to making an unmodalised assertion. That is a rather banal fact about epistemic necessity, and does not motivate introducing a separate notion of “degree of modality”. To sum up, cases where there is a small difference in meaning between a sentence with a modal and its unmodalised counterpart occur for three completely different reasons: A. B.

C.

The modal loses some of its meaning (may and should in some marginal uses). The unmodalised sentence has acquired some of the same meaning as the sentence with the modal (futurate use of the present tense vs. will). Deductions based on strong evidence, and expressed using epistemic must or will, do not differ greatly in strength from unmodalised assertions.

The notion of “degree of modality” in CGEL thus appears not to cover a single phenomenon but three unrelated ones. Furthermore, these phenomena can each be explicated by analytic notions which are independently needed, so the proposed new notion adds nothing useful to the analysis of modality. In this paper we have treated the notion “degree of modality” as a more complex one, involving several different criteria instead of the single one used in CGEL. We might add that a key motivation for introducing the notion in CGEL and in Huddleston (1995) was to support the claim that will is a marker of modality, not of tense. In subsequent work we hope to show that will should be treated as both a non-prototypical modal and a non-prototypical tense, thereby resolving the controversy about its category status.

100 Raphael Salkie 7. Conclusion We have set out an account of modality using the prototype approach to grammatical categories, a strategy which is widely accepted by typologists and by English grammarians. Using four criteria for membership of the category, we claimed that some auxiliaries in English (e.g. must and may) have core uses which meet all the criteria and peripheral ones which do not. Other auxiliaries such as can, should and will usually fail one or more of the criteria. We have tried to show that the notion “degree of modality” can be broken down into a small set of criteria, and that the proposed framework sheds light on a range of English data. We argued that the alternative prototype approach to the notion “degree of modality” in CGEL is misconceived. Any account of a typological category needs to show its value in more than one language. In future work we plan to apply the prototype framework to modal expressions in French and German. For English, at least, treating different instances of modal expressions as varying in the degree to which they match the prototype seems to be a useful grammatical tool.

Notes 1. I am grateful to Jelena Timotijevic, Heiko Narrog, and an anonymous Mouton reviewer for comments on an earlier version. None of them is responsible for any errors. 2. Recanati argues that several processes which are often taken as secondary are in fact primary, which sets him at odds with other theorists of the semanticspragmatics interface. Processes like saturation, however, are recognised by many schools of thought, even if they disagree about other aspects of pragmatics (cf. Levinson 2000: 177; Carston 2002: 185ff.). 3. The passive voice probably contributes to Leech’s faulty analysis. Example (23) describes a property of the police (in a context such as [23']), while (22) makes a judgement about a property of the road. If the examples had been We can block the road versus We may block the road, the example with can would perhaps have felt “more immediate” and factual than the tentative modalised one with may.

Degrees of modality

101

References Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and E. Finegan 1999 Longman Grammar of Spoken and Written English. London: Longman. Butler, Johnny 2003 A minimalist treatment of modality. Lingua 113: 967–996. Carston, Robyn 2002 Thoughts and Utterances. Oxford: Blackwell. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Comrie, Bernard 1989 Language Universals and Linguistic Typology. 2nd ed. Oxford: Blackwell. Croft, William 1990 Typology and Universals. Cambridge: Cambridge University Press. Downing, Angela, and Philip Locke. 2002 A University Course in English Grammar. London: Routledge. Greenbaum, Sidney 1996 The Oxford English Grammar. Oxford: Oxford University Press. Huddleston, Rodney 1984 Introduction to the Grammar of English. Cambridge: Cambridge University Press. 1995 The case against a future tense in English. Studies in Language 19: 399–446. Huddleston, Rodney, and Geoffrey Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Kiefer, Ferenc 1994 Modality. In The Encyclopedia of Language and Linguistics, Volume 5, Ronald Asher (ed.), 2515–2520. Oxford: Pergamon Press. Kjellmer, Goran 2003 A modal shock absorber, empathiser/emphasiser and qualifier. International Journal of Corpus Linguistics 8: 145–168. Larreya, Paul, and Claude Rivière 2005 Grammaire Explicative de L’anglais. 3e édition. Paris: Pearson Longman France. Leech, Geoffrey 2004 Meaning and the English Verb. 3rd Edn. London: Longman. Levinson, Stephen 2000 Presumptive Meanings. Cambridge, MA.: MIT Press.

102 Raphael Salkie Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Narrog, Heiko 2005 On defining modality again. Language Sciences 27: 165–192. Palmer, Frank 1990 Modality and the English Modals. 2nd edn. London: Longman. 2001 Mood and Modality. 2nd edn. Cambridge: Cambridge University Press. Papafragou, Anna 1998 Inference and word meaning: the case of modal auxiliaries. Lingua 105: 1–47. 2000 Modality: Issues in the Semantics–Pragmatics Interface. Amsterdam: Elsevier. Recanati, François 2004 Literal Meaning. Cambridge: Cambridge University Press. Rivière, Claude 1981 Is should a weaker must? Journal of Linguistics 17: 179–195. Rosch, Eleanor, and Barbara B. Lloyd (eds.) 1978 Cognition and Categorization. Hillsdale, MI: Lawrence Erlbaum. Salkie, Raphael 2001 Grammatical explanation in Britain and France. In Langage et Référence:Mélanges offerts à Kerstin Jonasson à l’occasion de ses soixante ans, Hans Kronning, Coco Norén, Bengt Novén, Gunilla Ransbo, Lars–Gören Sundell, and Brynja Svane (eds.), 573–582. (Studia Romanica Upsaliensia 63.) Uppsala: University of Uppsala. 2002 Probability and necessity in English and German. In Information Structure in a Cross–linguistic Perspective, Hilde Hasselgård, Stig Johansson, Bergljot Behrens, and Cathrine Fabricius–Hansen (eds.), 81–96. Amsterdam: Rodopi. Timotijevic, Jelena Another look at modals and subjectivity. (In this volume). van der Auwera, Johan and Vladimir Plungian. 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. Verstraete, Jean–Christophe 2001 Subjective and objective modality: Interpersonal and ideational functions in the English modal auxiliary system. Journal of Pragmatics 33: 1505–1528.

Degrees of modality

103

Wurmbrand, Susi 1999 Modal verbs must be raising verbs. In WCCFL 18: Proceedings of the 18th West Coast Conference on Formal Linguistics, Sonya Bird, Andrew Carnie, Jason D. Haugen, and Peter Norquest (eds.), 599– 612. Somerville, MA.: Cascadilla Press. .

Another look at modals and subjectivity Jelena Timotijevic

1. Introduction The term subjectivity has been used in several different ways in studies of modality. This paper argues for a new treatment of the notion of subjectivity in the analysis of English modals. Section 2 gives a brief survey of different approaches to subjectivity in the literature on English modals, where we also examine how subjectivity has been applied to modality in German. In section 3 we argue that the concept of subjectivity should be linked to the relationship between semantics and pragmatics in the interpretation of modals. We propose that uses of modals should be called subjective if they involve primary pragmatic processes in the sense of Recanati (2004). Further, we will argue that the relationship is not the same for all modals, but varies for each modal.

2. Different treatments of subjectivity There are a number of different treatments of subjectivity in the literature on modals. In this section we will examine them in turn. The pioneering work in this field is represented by Lyons (1977) who argued that epistemic and deontic modals can both be subjective or objective. That most modals are subjective is the position taken by Palmer (1990). We then look at the relationship between subjectivity and syntax through the generative approach, mainly in German (Grebe et al 1973, but also see Butler 2003 for English modals), as well as investigate the notion of subjectivity being an inherent property of some modals, as suggested by Larreya and Rivière (1999). Nuyts’ (2001) analysis revisits this issue by arguing for an interpretation of subjectivity as an evidential dimension, or that it is conceived as a type of speaker commitment (Verstraete 2001).

106 Jelena Timotijevic 2.1. Epistemic and deontic modals can both be subjective or objective The subjective/objective distinction in the analysis of English modality was originally made by Lyons (1977), who argued that both epistemic and deontic modals can be either subjective or objective. He firstly differentiates between epistemically modal and non-modal statements, the latter being characterised by “unqualified assertion”, which means that the speaker commits herself to what she asserts [my italics], but she does not lay claim to knowledge in the utterance itself, i.e. she is not asserting an epistemically modalised statement. (1977: 797) Epistemically modalised utterances on the other hand are those in which the speaker “explicitly qualifies [her] commitment to the truth of the proposition expressed by the sentence [she] utters”. This qualification is usually evident in the verbal component. (1977: 797) Lyons uses this distinction to support his subjective/objective argument: in subjective epistemic statements the speaker subjectively qualifies her commitment in the state of affairs, which points to her own uncertainty. Objective interpretation points to the speaker’s [inherent] knowledge of a particular situation being possible, in Lyons’ terms the speaker could present that possibility as an objective fact. (1977: 798). So, for epistemic modals Lyons looked at examples like: (1)

Alfred may be unmarried. (1977: 797) Subjective interpretation: the speaker knows nothing about the number or presence of any unmarried people in the community. He subjectively commits to the possibility of Alfred being unmarried, based on his own assumptions. Objective interpretation: the speaker knows about the number of married people in the community, which is not merely a belief. He therefore knows that there is a possibility of Alfred being unmarried.

Let us now look at (2). (2)

Alfred must be unmarried. (1977: 797) Subjective interpretation: in every day use of language it would be more natural to interpret (2) in terms of subjective epistemic modality, because the speaker subjectively qualifies her commitment to the

Modals and subjectivity

107

factuality of the proposition, rather than expressing an unqualified, or categorical “I-say-so”. Subjective modal utterances are statements of opinion, tentative inference, etc. rather than statements of fact (1977: 799). Objective interpretation: if we consider a community of ninety people where we know that thirty of those people are unmarried, and we subsequently identify exactly twenty nine people that are unmarried, except for Alfred’s marital status. The speaker is committed to the factuality of the information he/she is giving to the addressee, and can therefore be justified in claiming to know that the interpretation of (2) as “Alfred is unmarried” must be true in some sense.

As the pioneering work in this field, Lyons’ analysis is a good starting point, but it is not clear that his distinction above has consequences for grammar. It is questionable whether English, or any other language, distinguishes grammatically between objective and subjective epistemic modality.

2.2. Most modals are subjective Palmer (1990) argues that epistemic and deontic modality are almost always subjective. The function of epistemic modals is to make judgements about the possibility or necessity that something is or is not the case, i.e. epistemic modality is that of propositions rather than states, events or actions (1979: 41). He claims that epistemic modals are normally subjective, where the epistemic judgement rests with the speaker. With deontic or performative modality, the speaker has a role of giving permission, laying an obligation, or making a promise or threat. Further evidence for the performative nature of deontic modality is in the assertive and negative forms where the act takes place in the moment of speaking, for, by their nature, performatives cannot be in the past (1979: 59). Palmer introduces a third type of modality, dynamic modality, under which he groups ability and volition. Whereas epistemic and deontic modality are “related to the speaker”, dynamic modality is not, and may not be modal at all (1990: 7).

108 Jelena Timotijevic Palmer seems to suggest a close relationship between subjectivity and modality, but does not develop this relationship further in his work.

2.3. Subjectivity and syntax In some treatments of modals, epistemic uses are treated as subjective, deontic ones as objective. The most explicit analysis of this kind has been for German. For example, both Grebe et al (1973) and Griesbach (1986) claim that: a) b)

c) d)

Epistemic uses are “subjective” and indicate speaker involvement/commitment. Most other modal uses, primarily deontic ones, are “objective” and indicate no speaker commitment. Objective modals signal a relationship between a subject and a predicate. If a modal can be put in the past tense, the meaning is objective. The “evidential” uses of sollen and woollen, which have no direct counterpart in the English modals, are subjective.

Two interpretations of the modal können, which in English corresponds to both can and may, are discussed in Grebe et al (1973: 69 and 72-73). The modal können in sentences (3a) and (3b) is “objective”; können in (4a) and (4b) is “subjective:” (3a) Er kann bald sterben. (=Es ist möglich, daß er bald stirbt.) ‘He can die soon.’ (‘It is possible that he dies soon.’) (3b) Aus diesen sozialen Spannungen kann durchaus ein Streik entstehen. (= Es ist durchaus möglich, daß aus diesem sozialen Spannungen ein Streik entsteht.) ‘A strike can develop from social unrests.’ (‘It is perfectly possible that a strike can develop from social unrests.’)

(4a) Du kannst das Geld verloren haben. ‘You may have lost the money. (‘It is possible that you have lost the money. ’)

Modals and subjectivity

109

(4b) Das könnte ein gutes Geschäft werden. ‘This could be a good business deal. ’ (‘It is possible that this is a good business deal. ’) Grebe et al claim that we are dealing with possibility interpretations in both instances (examples in (3) and (4)), although we note that example (3a) can also be seen as an instance of ability interpretation of können (He will be able to ‘let go’). Examples (3a) and (3b) indicate, however, that the conditions and circumstances make possible for something to happen, as well as signal a relationship between a subject and a predicate, hence objective modal uses. Subjective interpretations in (4a) and (4b) express the speaker’s personal opinion and viewpoint. Here, too the emphasis is on syntax: the authors argue that the grammatical structure modal + the infinitive of the main verb, signals subjective modal meanings. We can see that in Grebe et al’s analysis there does not seem to be very much difference between the objective and subjective interpretations. More recent work on German modals challenges this type of analysis (e.g. Zifonun et al 1997: 1886). The authors argue against the classification of modal uses in two categories, that of subjective epistemic and objective non-epistemic modality. Their analysis is similar to that of Nuyts (2001) (see below) where Zifonun et al draw a distinction between intersubjective “accessibility” to the knowledge of a particular situation in question and existing knowledge which can be proven, in other words the subjective/objective distinctions depends on the quality of the evidence available (1997: 1886). On the one hand we are dealing with examples of historical and factual data, everyday knowledge, etc., and on the other it is the speaker’s own ‘access’ to the knowledge of some kind, supposed observation, subjective insight, superstition, and so on. The idea that objective (i.e. deontic) modals express a relation between a subject and a predicate is similar to the assumptions behind much generative work on English modals. Deontic modals are treated as two-place predicates, whereas subjective (epistemic) modals are one-place predicates. This idea has a long history in generative work, going back at least as far as Ross (1969). The syntactic consequence of this analysis is that epistemic modals are raising verbs while deontic modals are control verbs. A recent version of this analysis is found in Butler (2003). We do not have the space to discuss these approaches in detail. We note, however, that Wurmbrand (1999) argued that they are not sustainable, and that both deontic and epistemic modals are represented by a rais-

110 Jelena Timotijevic ing structure rather than a control structure. She claims that the properties of subjects in modal constructions are determined by the lower verb not the modal, and that rather than the syntax, it is the context (pragmatics) that plays a crucial role in distinguishing deontic and epistemic modalities.

2.4. Subjectivity as an inherent property of some modals Larreya and Rivière (1999: 80-81) propose that some English modals are inherently subjective while others are “neutral.” They argue that can, will, have to and ought to are neutral, while may, must, shall and should are always subjective. An example of neutral interpretation is example (5): (5)

Mary can swim.

Larreya and Rivière claim that the modal judgement expressed by can in (3) is normally a result of a simple observation. Contrary to this, modal judgement can be expressed as a personal opinion of the speaker, his/her will or wish, which, in Larreya and Rivière’s terms, corresponds to subjective modal meaning, as in (6): (6)

You may smoke.

2.5. Subjectivity as an “evidential” dimension Nuyts’ (2001) analysis is restricted to epistemic modals. He argues for an interpretation of subjectivity as an evidential dimension, which can, but need not be, expressed at the same time as the epistemic qualification. His work is based on the analysis of corpus data of spoken and written Dutch and German, covering modal adverbs, modal adjectives, mental state predicates and modal auxiliaries.

Modals and subjectivity

111

His proposal on the nature and status of subjectivity is based on a convincing reanalysis of Lyons’ (1977) definition of subjective and objective modality. As we have seen in 2.1 Lyons makes this distinction based on whether the speaker subjectively or objectively qualifies her commitment to the factuality of the proposition, in other words example (1) above may indicate that the speaker is uncertain about the fact that Alfred is unmarried (subjective interpretation), or it may indicate, as Nuyts puts it, “…[…] that there is mathematically computable chance that Alfred is unmarried…[…]…” based on what the speaker knows about Alfred and his community (objective interpretation). (2001: 385) Nuyts suggests, however, that speaker-involvement does not seem to be the only parameter in determining the subjective/objective distinction. In his reassessment of Lyons’ view, on the one hand, the dimension of subjectivity is characterised by the quality of evidence, and on the other hand we are dealing with the status of that evidence drawn from either (a) the speaker’s own knowledge and conclusions, or (b) from a larger group of people who share the same conclusion(s). Nuyts uses terms ‘subjectivity’ and ‘objectivity’ to account for the difference in the quality of evidence. The latter distinction is analysed under the terms ‘subjectivity’ and ‘intersubjectivity’ (see below). In line with the above, he reanalyses Lyons’ example (1) to show that two separate semantic dimensions are concerned: one which involves the speaker’s evaluation of the probability of the state of affairs (epistemic evaluation), and the second which is the speaker’s evaluation of the quality of that evidence which fits under what is traditionally labelled evidentiality. (2001: 386) In his analysis of the corpus data Nuyts argues as follows: – Modal adverbs like probably appear neutral in terms of subjectivity. Depending on the context, they may involve both subjective or nonsubjective evaluations (2001: 389). – Similarly, context dependent modal adjectives such as probable can express both subjective and non-subjective meanings. – Modal auxiliaries appear neutral for subjectivity. Nuyts (2001: 392) here draws on Lyons’ analysis of the same sentence appearing in two different contexts. In conclusion, Nuyts’ argument suggests that the dimension of subjectivity is inherent to some epistemic expressions, but not to others. He also sug-

112 Jelena Timotijevic gests that the term “objectivity” should be replaced by “intersubjectivity”, since it covers cases where the evidence is generally accepted, for example, in the scientific community, as in (7) (2001: 389): (7)

Alle Sterne in einem solchen Sternhaufen sind sehr wahrscheinlich etwa gleichzeitig aus einer gemeinsamen großen Gaswolke entstanden. ‘All stars in such a cluster very probably developed at about the same time out of a single large cloud of gas.’

Subjective interpretations are those where the evidence is available only to the speaker. Nuyts proposes a useful account of subjectivity in relation to epistemic modals, but does not offer an analysis of subjectivity that also applies to deontic modality, which could have been a useful tool in analysing different modals’ behaviour.

2.6. Subjectivity as degree of speaker commitment Verstraete criticises previous accounts of subjectivity, and proposes that subjectivity should be analysed as the degree of commitment by the speaker to the propositional content of the utterance (2001: 1517). He argues that English epistemic modals are always subjective, while deontic modals can be either subjective or objective. Dynamic modals are always objective (2001: 1525). A consequence of this analysis is that objective modals can be tensed, while subjective ones cannot. Verstraete argues that epistemic modal uses cannot express the speaker’s past judgement (thus example (9) below is a tentative use of might). Dynamic modals of volition and ability occur in the past tense (e.g. past ability in (8)), as well as some deontic interpretations, mainly objective deontic modals which express the existence of past necessity (as in example (10)). (8)

(9)

Gillespie positioned himself for a challenge but before he could move in for the tackle Hughes had driven the ball high past Grobbelaar from 25 yards. Well, she might be coming to Clare’s party

Modals and subjectivity

113

(10) In fact Old Trafford had to wait another 20 minutes but then two goals arrived together, like buses. (examples are from the ICE-GB corpus, cited in Verstraete 2001: 1524): For Verstraete deontic modality can be either subjective or objective. He argues that subjective and objective deontic uses differ in interrogatives. In subjective cases, the speaker asks whether the “interlocutor” is committed to a deontic necessity – compare (11) with (12) and (13) (Verstraete 2001: 1521): (11) What we want is for the right honourable gentleman to use the full weight of his office. We are getting tired of a cosmetic approach an oversanguine approach. There is a crisis and he must act now. (12) Must the minister act now? (13) “You’ve got to be there by nine o’clock in the morning at the latest. You’ll be crossing the main refugee routes. Shouldn’t be too bad”. “Must I leave my platoon, sir? At this moment?” “Stop arguing and get down there. It’s nobody’s fault but yours that you speak fluent German. You know perfectly well every linguist’s name is listed.” (COBUILD) Objective deontic interpretations do not undergo this shift. They are interpreted in terms of the truth of the proposition and the existence of necessity, as in example (15), which could be paraphrased as: “do you think it is true that brake shoes must be changed in sets of four?” (2001: 1522): (14) Brake shoes must always be renewed in sets of four. (15) Must brake shoes always be renewed in sets of four? (16) First-time buyers should ask: Who is responsible for a title search and abstract? Who will provide title insurance? Must the termite inspection be paid before closing (in which case you will need a receipt), or can it be paid at settlement? (COBUILD) To sum up, epistemic modality is always subjective for Verstraete, while deontic modality can be either subjective or objective. Dynamic modality is always objective. This is an advance over earlier work, but Verstraete’s claims have several weaknesses:

114 Jelena Timotijevic (a)

(b)

(c)

He ignores the body of work represented by Larreya and Rivière, in which some modals are always subjective and some are always objective. Verstraete rejects Lyons’ approach (2001: 1525) because the distinction that Lyons makes is “not reflected in the linguistic behaviour of [modals]” (by linguistic he means “syntactic”). But there is no reason to assume that the semantic-pragmatic behaviour of modals is always necessarily reflected in their syntax. He tries to connect the meaning and use of modals with their syntactic behaviour (e.g. co-occurrence with tenses). Here it is my view that the issue is the relationship between semantics and pragmatics, not syntax.

3. Subjectivity is a pragmatic notion In this section we will use the work by Recanati (2004) to capture the idea of “more pragmatics”. Our analysis will then investigate how Papafragou (2000) accounts for the distinction between subjective and objective modal uses. We will argue for a new treatment of subjectivity as a pragmatic notion by modifying Papafragou’s (2000) proposal. Finally, we will revisit the issue of monosemy and polysemy and why it is relevant here. We have examined several different ways in which the term “subjectivity” is used in analysing modal verbs. We propose that the central issue here is actually the balance of semantics and pragmatics in the interpretation of modals. Uses of modals that have been called “subjective” crucially involve the speaker and thus involve more pragmatics than uses which are “objective”. This idea, implicit in Verstraete’s work, can be made more specific by using the notion “primary pragmatic process” from Recanati (2004). He argues that the truth-conditional interpretation of utterances is pragmatic to a large extent, since a number of pragmatic processes play an important role in determining what is said (2004: 20-21). Recanati calls them “primary pragmatic processes”, an example being saturation, in which indexical expressions, such as tenses and personal pronouns, have their reference assigned from the context. “Secondary pragmatic processes” like free enrichment, are optional. Compare these two examples: (17) He is tall (18) Everybody went to Paris

Modals and subjectivity

115

In (17), we do not have a complete proposition, with determinable truth conditions, until the context has enabled us to assign a referent to he. In (18), we have a complete proposition which is true if every existing person went to Paris. The proposition actually communicated in context is most likely “everybody in a particular group went to Paris”. The process by which we arrive at this proposition is “not linguistically but pragmatically required”. In (17) we need saturation, while (18) is an instance of free enrichment. Recanati’s distinction between primary and secondary pragmatic processes is applied to modality by Papafragou (2000) who claims that may and must require saturation whereas can and should undergo enrichment. In her analysis may, can, must and should each have a single meaning, with pragmatics accounting for the interpretations these words have in particular contexts, which, although implicit in Papafragou’s framework, is what accounts for the distinction between subjective and objective modal uses. Using p for the embedded proposition and treating the context as a set of propositions in a domain D, Papafragou claims that the logical relationship between the two can be entailment or compatibility. So, the proposed meanings are (2000: 43):

MAY CAN MUST SHOULD

p is compatible with the set of all propositions in D (D is unspecified) p is compatible with the set of all propositions in D (D is factual) p is entailed by the set of all propositions in D (D is unspecified) p is entailed by the set of all propositions in D (D is normative)

Because the domain D is specified for can and should, there is no need for saturation to specify the domain. For may and must, on the other hand, the hearer needs to search for a domain, so for these modals the utterance is incomplete until a relevant domain has been identified. Although Papafragou argues at some length that the distinction between saturation and enrichment is genuine (2000: 44-47), she does not show that the distinction has empirical consequences in the different behaviour of the two types of modal. She does, however, suggest that can encodes potentiality (2000: 48), and that should involves knowledge of “the normal course of events” (2000: 62). Suppose that we accept her analyses of the “saturation” modals may and must, but reanalyse the other two as follows: CAN:

p is potential

116 Jelena Timotijevic SHOULD:

p is part of the normal course of events

The potentiality of can sometimes resides in the capacities of a person (ability sense), or it may be part of a broader domain of facts, giving the possibility sense. A process of free enrichment will identify the relevant domain in a particular context. Similarly, the context will suggest for should which events are relevant and what counts as “normal” for them. Compare now some examples of can (19-20) and some of may (21-22) which illustrate why can is objective and may is subjective, taken from the INTERSECT corpus: (19) In addition, you can specify scanner settings to control such things as brightness, document quality, resolution, and more. (20) The effects of fear on both the body and brain can turn it into an asset when it is consciously used to its best effect in putting on a performance. (21) We went on a little, then: “It’s no use going any further,” said Marceline; “these orchards are all alike; possibly at the other end of the oasis they may be a little larger ... (22) In most cases you will need to do this 5 - 10 times (only once or twice for soft foods such as tomatoes or hard-boiled eggs, while hard foods, such as carrots, may need longer). Our claim is that (19) and (20) are complete: they refer to a potential, and the hearer can, if she wants, try to reconstruct the speaker’s communicative intention by working out where the potential is located. Thus (19) can be taken as an instance of “ability” can, whereas in (20) we have the likely interpretation of the “sometimes” sense where a broader domain of facts is relevant. Examples (21) and (22), on the other hand, merely refer to the possibility of a proposition p being true – they assert no more than that p is compatible with a domain of propositions. Without knowing what that domain of propositions is, we cannot identify the type of possibility. The search space is wider than for can, since potentiality has to be located in the real world, while possibility can invoke other possible worlds. We thus have to ask what type of evidence the speaker drew on in order to assert the possibility, and this can either be genuine uncertainty as in (21), or the fact that members of a class vary in their other properties as in (22), based on knowledge of (in this case) foodstuffs. Although can in (20) and may in

Modals and subjectivity

117

(22) have similar “sometimes” interpretations, different processes are involved in the two cases: for can we invoke the real world, whereas for may we have to involve speaker knowledge. We conclude that the “subjective” modals may and must are interpreted differently from the objective can and should.

3.1. Monosemy vs. polysemy Here we point to the close link between pragmatics/semantics relationship in the interpretation of modals (Recanati 2004) and monosemy/polysemy distinction as argued by Papafragou (2000). In this section we note another difference between the two types of modals, which is the relationship between their different uses. We base our ideas on the above two accounts to further explain subjectivity as a pragmatic notion and argue that subjective modals have a different relationship between their uses to deontic modals, which will show that “subjective” modals have a different type of pragmatics to the “objective” ones. Whether modals are polysemous or have a single abstract meaning is an issue that has long been debated in the literature. It is linked to the question of the balance of semantics and pragmatics in the analysis of modals: theories which treat modals as monosemous tend to give modals a skeletal semantics and make pragmatics and the context do more of the work of interpretation. Papafragou, for instance, argues against analyses of the English modals such as those by Palmer (1990) who treats them as ambiguous, and Sweetser (1990) who treats them as polysemous. Some interesting discussion of this issue can be found in Palmer (1979, 1990), and Coates (1983). Palmer concedes that “there are many examples [of should] where it is difficult to be sure whether we have dynamic or epistemic modality” (1979: 49), for instance: (23) Should only take three days for the survey report to be in to the building society. (24) So he should be around sort of between half past two and half past three. He also finds it hard to separate dynamic and deontic should (1979: 69). Coates refers to Palmer’s discussion but says that she will “treat the root and epistemic meanings of should and ought as two distinct categories”

118 Jelena Timotijevic (1983: 80). She proposes that there are three types of indeterminacy (1983: 14-18): gradience, in which different meanings shade into each other (e.g. ability vs. possibility can); ambiguity, where it is not possible to decide which of two distinct meanings is intended (e.g. He must understand that we mean business); and merger, where the two meanings involved are not mutually exclusive (Coates gives an example with ought which is similar to (23-24)). One of her reasons for separating the root and epistemic meanings of should and ought is that they involve ambiguity or merger, which presuppose two different categories (1983: 80). In his later book, Palmer comments that the distinction between gradience and merger is arbitrary (1990: 22). What is noteworthy is that it is can which Coates uses to illustrate gradience, and should which exemplifies merger. Genuine ambiguity is represented by must: the meaning of her example He must understand that we mean business is resolvable given a specific context, whereas the cases of gradience and merger are not – the indeterminacy remains no matter how much context is added. This is what we would expect if a sentence with must has an incomplete meaning until the context has done its work, but after this the different interpretations are distinct; while examples with can and should have complete meanings. With certain examples of can and should, however, it is difficult or unnecessary to enrich them in ways that enable us to make the clear distinctions that are possible with must. In other words, the “saturation” modals like must have distinct meanings, while can and may do not. We thus propose that the “subjective” modals like must and may have a different type of pragmatics from the “objective” ones like can and should. This has consequences for the way they are interpreted (cf. 19-22) and the relationships between their different uses (23-24). With can and should the different senses merge into each other, whereas with must and may they are clearly distinct. It is not surprising that the examples that Groefsema uses (1995: 57) to argue against polysemy also involve can: it is harder to argue that must and may are monosemous.

3.2. Is should subjective or objective? Our analysis is similar to Larreya and Rivière, who as we saw in 2.4 treat some modals as inherently subjective and some as objective. For space reasons we cannot compare our division of the English modals with theirs

Modals and subjectivity

119

in detail, but we can note one difference: they treat should as subjective, whereas in our treatment it is objective. We shall argue now that their reasons for calling should subjective are not convincing. Larreya and Rivière base their argument for should being subjective on the claim that it is shall + past tense, and shall is a subjective modal. Consider these examples (1999: 81): (25) Shall I close the window? (26) Will I close the window? The authors argue that in (25) shall is subjective because it expresses personal opinion, or the will or wish of the hearer, whereas in (26) it has an objective necessity interpretation existing independently of the hearer. The relationship between shall and should is analysed in detail by Huddleston (1995). Huddleston argues that the only cases where should is the past tense of shall are those where should is a formal counterpart of would, and they are restricted to the first person. Some of the examples he offers are (1995: 409): (27a) (27b) (27c) (27d)

We shall soon have a new president. I shall easily finish before she returns. If they offer me the job I shall certainly accept. I shall like to read it.

(28a) (28b) (28c) (28d)

Only two months later we should have a new president. I knew I should/*shall easily finish before she returned. If they offered me the job, I should/*shall certainly accept. I should like to read it.

(We disagree with Huddleston, however, when he says that (28d) is a common use of should, as I would like to read it is much more normal and common in every day communication). These uses of should are instances where it has little modal meaning, and is close to a tense marker. Turning to the main modal uses of should Huddleston clearly shows that those examples are not the past tense of shall. Consider the following (1995: 410): (29) You should work harder. (30) The next road on the left should be King Street.

120 Jelena Timotijevic (31) If you should experience any difficulty, please let me know. (32) It is essential/desirable that he should be told. Huddleston maintains that modal should is not possible with past time meaning and that it does not occur as the counterpart of shall. Example (29) has a deontic interpretation, whereas example (30) is epistemic probability with no comparable use of shall. For (31) Huddleston suggests “an open condition” interpretation (1995: 410). For should be in (32), Huddleston says that it “alternates” with the use of be, which is tenseless here; there is no analogous use of shall. Thus there are reasons to doubt the assumption that should is the past tense counterpart of shall, an assumption which is fundamental to Larreya and Rivière’s claim that should is subjective.

4. Conclusion We have proposed a new treatment of the subjective/objective distinction in terms of the balance between semantics and pragmatics in determining the meaning of modals. We argued that previous account of subjectivity have a number of weaknesses. We claim that subjectivity is a pragmatic notion, and that it applies to certain modals but not to others. We have linked subjective modal meanings to primary pragmatic processes like saturation, whereas less subjective modal meanings involve free enrichment. This has implications for the interpretation of different modals, and for the relationship between their different senses. A monosemy analysis looks more promising for objective modals. In future work we plan to analyse the implications of this work cross-linguistically.

References Butler, Jonny 2003 A minimalist treatment of modality. Lingua 113: 967–996. Coates, Jennifer 1983 The semantics of the modal auxiliaries. London: Croom Helm.

Modals and subjectivity

121

Foley, William, and Robert Van Valin 1984 Functional syntax and universal grammar. Cambridge: Cambridge University Press. Grebe, Paul, Helmut Gipper, Max Mangold, Wolfgang Mentrup, and Christian Winkler 1973 Der Grosse Duden: Band 4. Grammatik der deutschen Gegenwartssprache. Bibliographisches Institut Mannheim, Wien/Zürich. Dudenverlag. Griesbach, Heinz 1986 Neue deutsche Grammatik. Langenscheidt. Groefsema, Marjolein 1995 Can, may, must and should: A Relevance theoretic account. Journal of Linguistics 31: 53–79. Huddleston, Rodney 1995 The case against a future tense in English. Studies in Language 19: 399–446. Larreya, Paul, and Claude Rivière 1999 Grammaire explicative de l’anglais. Nouvelle édition. London: Longman. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Nuyts, Jan 2001 Subjectivity as an evidential dimension in epistemic modal expressions. Journal of Pragmatics 33: 383–400. Palmer, Frank 1979 Modality and the English Modals. London: Longman. 1990 Modality and the English modals. 2nd ed. London: Longman. Papafragou, Anna 2000 Modality: Issues in the semantics-pragmatics interface. Amsterdam: Elsevier. Recanati, François 2004 Literal meaning. Cambridge: Cambridge University Press. Ross, John Robert 1969 Auxiliaries as main verbs. Studies in Philosophical Linguistics Series I. W. Todd (ed.), 77–102. Evanston: Great Expectations Press. Salkie, Raphael Degrees of modality. (In this volume). Sperber, Dan, and Deirdre Wilson 1995 Relevance: Communication and Cognition. Oxford: Blackwell. 2nd. ed. Sweetser, Eve 1990 From etymology to pragmatics. Cambridge: CUP.

122 Jelena Timotijevic Verstraete, Jean-Christophe 2001 Subjective and objective modality: Interpersonal and ideational functions in the English modal auxiliary system. Journal of Pragmatics 33: 1505–1528. Wurmbrand, Susi 1999 Modal verbs must be raising verbs. WCCFL 18 Proceedings, S.Bird, A. Carnie, J. Haugen, and P. Norquest (eds.), 599–612. Somerville, MA: Cascadilla Press. Zifonun, Gisela, Ludger Hoffmann, and Bruno Strecker 1997 Grammatik der deutschen Sprache. Berlin: Walter de Gruyter.

For a topological representation of the modal system of English Alain Deschamps and Lionel Dufaye

1. Introduction The aim of this paper is to offer a general representation of the system of English modals based on a linguistic framework known as “Théorie des opérations énonciatives”. The framework posits sequences of operations and sets of relations, both linked with cognitive processes. An overview of Antoine Culioli’s theoretical background for this framework can be found for instance in Groussier (2000), in which the main concepts of the theory are explained. The theory was originally developed in French, and giving it an English form often implies using English metalinguistic terms that are not always widely used, or to which we give specific definitions. In this paper, we propose a general abstract form for each modal (a schematic form) and show that the apparently diverse meanings and uses of each modal are based on chains of linguistic operations (schemes) determined by a few predefined parameters. Once the abstract form has been postulated and the necessary basic parameters defined, no extra metalinguistic operations will be needed to account for the extreme variety of meanings and uses of modal auxiliaries in authentic examples.

2. Basic concepts

2.1. Levels of linguistic analysis The required definitions underline the idea that linguistic operations belong to different levels (notional, predicative or enunciative): The basic concept is that of notions (sets of physico-cultural properties) giving rise to notional domains with an Interior (I) possessing those properties, an exterior (E) having none of the properties, a frontier (F) not totally different from I but having different properties, a centre

124 Alain Deschamps and Lionel Dufaye (corresponding to prototypes or high degree), and a totally disconnected position IE from which these different points or areas can be constructed. The predicative relation (< a r b >) corresponds to a set of three notions (2 arguments and a predicate) which initially have no particular order. The predicative relation is by itself a new complex notional domain provided with the same partition as the basic notions themselves. The third level is that of enunciative operations in which predicative relations are endorsed by an enunciator or speaker and submitted to sets of operations giving them situational coordinates (subjective and spatiotemporal) based on and originating in the discourse situation. < a r b > ∈ Sit (T/S). Complementary operations complete the system of situational determination: qualification (properties, intersubjective relationships, evaluation or valuation) and quantification (spatio-temporal relations and occurrences).

2.2. Linguistic analysis of modality Concerning modality, we will be content with using only two areas (Interior and Exterior): the frontier is not necessarily taken into account. If the predicative relation is validated, or stated as being the case for a specific speaker, we are dealing with assertions or assertive modality. To simplify, the utterance may take a positive or a negative value: “it is the case that < a r b > is the case” = I or “it is the case that < a r b > is not the case” = E. For instance, I have lost my keys or I haven’t lost my keys. This means that only one of two values is taken into account and the other value is not considered – the other value (or in French “altérité”, translated as otherness) is rejected. If I is the case, then E is left aside and reciprocally. This is the pattern found in assertive modality, but in a number of instances, no definite value can be given to the predicative relation: it is impossible to state that I is the case or E is the case, because we are outside the plane of validation (i.e. neither asserted positively nor asserted negatively). Then we have to build a completely exterior point from which a bifurcation with a “path” leading to I and a path leading to E can be constructed, both values being situated in the plane of validation. This disconnected position called IE serves as a basis for establishing the prospective validation of one of the two values.

Topology and the English modal system 125

2.3. The construction of otherness If we consider first that the path leading to I is taken into account, it appears that there are three different ways of apprehending the complementary scenario E (hence three ways of construing “otherness”): – only I is envisaged and E is not considered at all (I in all cases), – I is envisaged but the path leading to E is opened too, (I among other values), – I is envisaged and the path to E is blocked (I and nothing else). Thus, otherness is either simply disregarded or explicitly considered or altogether denied. This corresponds to the three patterns shown in table 1. Table 1. Three types of otherness 1 I

2 E

I

3 E

I

E

\

\/

\/

IE

IE

IE

In addition, it must be acknowledged that E is always far more complicated than I, the exterior being a representation of the lack of I (void of I) simply meaning that I cannot be constructed. This will be found later in such examples as he can’t swim in which can’t swim does not correspond to anything positive. A second possible definition of E is that of the linguistic complement of the notion which might be given a positive definition, as in antonymic relations (raw / cooked, open / shut, refuse / accept or catch / miss). A third possibility is illustrated by qualitative differences (sing / shout for instance, yellow / red), in which E stands for ‘qualitatively other than I’. Since the construction of otherness rests on a polar opposition, we can expect that IE – the vantage point from where otherness is constructed – is compatible with either a qualitative representation (linked with notional properties, evaluation, valuation, subjectivity and intersubjectivity) or a quantitative representation (quantification and construction of occurrences related to spatio-temporal parameters). More intuitively, we could say that the Qualitative component is active whenever the speaker evaluates the event either as good or bad (i.e. what the speaker wishes or doesn’t wish) or as compatible or incompatible (e.g. “he can(not) swim”, “for this to

126 Alain Deschamps and Lionel Dufaye work properly the grates must be hot”, etc.). The Quantitative component, on the other hand, is active whenever we consider whether an event has taken / is taking / or will take place. We have to assume that otherness depends on qualitative grounds (properties) or on quantitative grounds (validation of occurrences) which allows us to combine two representations of otherness, one linked with qualitative determination (hence Qlt or quality) the other linked with quantitative determination (hence Qnt or quantity). Since the construction of otherness is double, two separate bifurcation patterns with two neutral points have to be postulated, both related to one of the parameters of the enunciative origin (i.e. deictic coordinates: Speaker + Time of utterance). This dual character is always present in all modal phenomena. The combination results in a set of 9 possible representations of modal forms, as shown in table 2. Table 2. Heuristic combinations

1 I E

2 I E

I E

3 I E

I E

I E

\/ \/

\ \

\/ \/

IE Qlt

IE Qlt

IE Qlt

IE Qnt

IE Qnt

IE Qnt

____________________________________________________ 4 5 6 I E

I E

I E

I E

I E

I E

\/ \

\ \/

\/ \

IE Qlt

IE Qlt

IE Qlt

IE Qnt

IE Qnt

IE Qnt

____________________________________________________ 7 8 9 I E

I E

I E

I E

I E

I E

\/ \/

\ \/

\/ \/

IE Qlt

IE Qlt

IE Qlt

IE Qnt

IE Qnt

IE Qnt

Topology and the English modal system 127

It appears that two of these patterns (7 and 9) are contradictory: you can’t establish at the same time that there is a path to E (in one component) and no path to E (in the other component). Thus, only seven combinations are theoretically meaningful. This doesn’t mean that languages are bound to have modal markers corresponding to each division of the table but that such patterns are likely to correspond to modal representations with or without modal auxiliaries.

2.4. Degrees of prominence Next, a second parameter proceeds from the same basic pattern: a different weight can be attributed to the qualitative variable and to the quantitative one, with three possible degrees of prominence or preponderance: – Either the Qlt parameter is prominent, noted Qlt / (Qnt), – or the Qnt parameter is prominent noted, (Qlt) / Qnt, – or both are equally weighted noted Qlt / Qnt or (Qlt) / (Qnt). When Qnt is prominent, priority is given to the construction of occurrences, whereas when Qlt is prominent, the stress is laid on properties, valuation or intersubjectivity. When both parameters are equally weighted, the two kinds of determination are given equal importance. It will be shown later that in so-called epistemic values, the Qnt parameter is insisted on whereas in root values Qlt is prominent or equally weighted (equal preponderance).

2.5. The relation between Qnt and Qlt A third parameter has to be added to the system, which bears on the way Qlt and Qnt are related. We mentioned earlier that the two neutral points (IE belonging at once to two different paths) at the basis of the bifurcation have to be located in relation to the enunciation origin which is double (a spatio-temporal origin T on the one hand, a subjective origin S on the other hand). Chains of relations serve to locate each neutral point with respect to the origin, by means of three possible relations: identification, differentiation or disconnection (respectively noted =, ≠ or ω).

128 Alain Deschamps and Lionel Dufaye If both parameters are directly identified to the origin, it follows that the speaker builds a notional property and locates its occurrences in relation to himself or herself. This appears for instance in epistemic modalities or in so-called future values of will. When a relation of differentiation is used the qualitative parameter is related to the speaker or enunciator, but the quantitative parameter is related to the co-enunciator or co-speaker (=addressee), responsible for the prospective validation of I or E. This will explain so-called deontic examples of the modals, where the co-speaker is likely to validate or not a predicative relation whose qualitative delimitation is endorsed by the speaker. Last, the two location relations can be disconnected, the speaker assuming the qualitative side of the predicative relation but having no part in its possible validation which is disconnected from the situation of enunciation (dispositional values such as sporadic can, will or may). Thus, the main idea of this representation is that modal values are not built on a continuum, nor on sets of disconnected values (with or without a central value) but on sequences of predetermined and retrievable operations based on fundamental parameters: – the construction of otherness (the bifurcation pattern), – the prominence of qualitative and quantitative parameters, – the location relation linking these two parameters to the origin of utterance.

3. English modals 3.1. The schematic forms of the modals If we go back to table 1, we are now in a position to give each English modal an invariant schematic form or abstract representation, filling five squares of the general pattern. That distribution is shown in table 3 with three symmetrical patterns corresponding to may, will and must and two non-symmetrical representations for shall and can. It appears that squares 7 and 9 have been previously excluded and that 4 and 8 have been left empty on grounds that might be explained in terms of incompatibilities. The five remaining patterns correspond to the set of modal auxiliaries

Topology and the English modal system 129 Table 3. Operating combinations

1 I E

2 I E

I E

3 I E

I E

I E

\/ \/

\ \

\/ \/

IE Qlt

IE Qlt

IE Qlt

IE Qnt may

IE Qnt will

IE Qnt must

___________________________________________________ 5 6 I E

I E

I E

\ \/

\/ \

IE Qlt

IE Qlt

IE Qnt can

May Will Must

I E

IE Qnt shall

double bifurcation with paths leading to I and to E. two paths leading to I but no paths opened leading to E. two paths leading to I and two paths leading to E blocked (otherness excluded).

The other two auxiliaries have non-symmetrical patterns: Can

Shall

a qualitative representation identical to that of will (only the path to I is defined) while its quantitative representation is similar to that of may (both I and E can be reached). a qualitative representation identical to that of must (the path to I is open, the path leading to E is blocked) while its quantitative representation is similar to that of will (only the path to I is defined).

130 Alain Deschamps and Lionel Dufaye 3.2. Implications of the formal representation The assumed representation must first account for all the possible values generally attributed to each modal in all kinds of different contexts. It must also allow us to deal with the similarities, quasi-equivalences, compatibilities and incompatibilities of the modals and non-existing patterns, that is it must explain why commutations between modals are permitted (for instance shall and will, can and may, will and can, shall and must, will and must) as predicted by the theoretical model itself, or prohibited (shall and can, shall and may, can and must, may and must…). Table 3 predicts some of the compatibilities and incompatibilities. For instance, it appears that shall and will have a similar Qnt pattern and then similar values when the Qnt parameter is activated (future values for instance). A comparison between 1 and 5 in the table shows that certain similarities exist between may and can on the one hand (permission or possibility); comparing 2 and 5 highlights those between will and can on the other (dispositional values). A similar representation of must and shall leads us to expect some kind of compatibility between these two modals. Finally, a certain likeness is expected between will and must (epistemic values) when one path only is considered in its qualitative value (the path leading to E is either blocked or not taken into account). On the other hand, it appears that some patterns are incompatible, notably in all cases when an explicit bifurcation is confronted with a pattern with a forbidden path: can / must, may / must, can / shall, may / shall are never found in similar contexts with approximating values nor, for similar reasons, do may and will commute. A second consequence of this representation is that it should explain why some modals never appear (in standard English) in some contexts: can for example is not found in epistemic positive contexts (preponderance of Qnt): (1)

*He can be waiting for us.

The reason is that with epistemic possibility, neither the positive nor the negative scenarios can be ruled out (Qnt) because neither is inconceivable (Qlt). Thus the negation has to bear on the predicate both quantitatively and qualitatively, which is impossible with can since there is no path towards E in the Qlt component. On the other hand, epistemic can is compatible with interrogation and negation since these operations bear on

Topology and the English modal system 131

can itself, not on the predicate (see also section 4.1 for a comparison with will): (2)

Can he be waiting for us? or he can’t be waiting for us.

It appears too that interrogations are also rare with may and must in epistemic contexts, because scanning the two values of the notion has no meaning if both paths are open or if one path has already been blocked. A third implication of the pattern is that the usual distinction between possible (two paths) and necessary (one path open one path blocked) is insufficient since the system does not rest on a twofold distinction but allows for a third case, that corresponding to the case when one path is open and nothing is said about the second. This configuration has common points with the other two: no scenario is ruled out (as is the case with “possibility”); one scenario only is taken into account (as is the case with “necessity”). The model must also account for the historical evolutions of the system of modals and for non-standard variants such as double modals. This representation also has to explain the various uses of each modal according to: – The bifurcation pattern – The preponderance relationship – The location relationship

3.3. The example of will The example of will will serve to exemplify the processes but the same can be done for each modal. If we look at the pattern corresponding to will, when Qlt is preponderant it will give rise to values in which only one path is taken into account, generally reflecting a positive value but the occurrence per se is not located in relation to the Time of utterance: This tent will sleep up to 6 people. If Qnt is predominant, the construction of occurrences becomes primary with two distinct cases. In the first, the speaker knows that the validation has already taken place but is not in a position to assert I or E. By choosing one path without considering the second, the speaker means that he or she

132 Alain Deschamps and Lionel Dufaye can’t or doesn’t want to choose the negative value. I is then retained as highly probable: That will be the police; quick! If, on the contrary the example refers to an instance of future validation, will indicates that I is selected as the only path and expected to be validated. This corresponds to contextual uses in which the qualitative dimension is left aside, giving will characteristics not very different from those of a future tense (possibly marked by the contraction of the modal into ‘ll). If equal prominence of Qlt and Qnt is the case, no preponderance appears, and a property is built from previous occurrences giving way to potential new occurrences. These so-called dispositional values may be illustrated by such examples as: (3) (4) (5)

He will drive with the brake on. Oil will float on water. Boys will be boys.

The same detailed review can be made for each modal1 explaining its values and showing that no central value has to be advocated and that no clear-cut distinctions amounting to distinct lists of values are necessary. The various contextual values assumed by the modals result from sequences of operations but the variables are limited in number and cannot be extended at will. A stable and relatively poor system of representation explains a nearly boundless corpus of examples and values but no extra parameters need to be added in the process. It appears next that extra operations can be performed on this basic pattern giving rise to more diversity and more intricate constructions. Negation, interrogation, stress, and temporal disconnection (past forms) must be dealt with in a general treatment of modality.

3.4. Modals and negation A single illustration of the reliability of the reference system will serve to exemplify the reasoning. Negation of modal auxiliaries is generally brought back to two possibilities, negation of predication and negation of modality, a dichotomy which is clearly insufficient: for instance, in an example like You mustn’t stay here, the negation bears both on modality (I

Topology and the English modal system 133

forbid you to stay here) and on the predicate (It is necessary for you not to stay here: i.e. only the negative scenario is acceptable). If we start from table 1, we see that negative patterns can be directly built from the original shape of each schematic form, without any previous operations, as shown in table 4. Table 4. Negative patterns 1 2 I E I E

\

/

IE

IE

can’t

won’t shan’t

3 I

4 E

I

E

\/

\/

IE

IE

may not

mustn’t shan’t

If a single path is open to I, it is either possible (1 in table 4) to state that the path is blocked and that there is no way to I (he can’t swim) or (2) to invert the pattern and direct the only path opened to E, nothing being said of the path leading to I (I won’t come). With the second type of bifurcation, no real inversion of the pattern is possible but the negative form (3) will result in giving different weights to the two paths. Instead of putting more weight on I (with E open too) more weight will be given to E (epistemic may not) although both paths remain open. He may have forgotten is neutral as regards forget / not forget but he may not have forgotten will insist on not forget (contradicting a previous construction of forget). Hence the use of may or may not in order to eliminate any possible preponderance in the pattern. With the third type, inverting the pattern will open the path to E and close the path to I (4). This is the way mustn’t is usually said to work (you must[not do that]). Other ways of building negative forms can be found in which a previously activated pattern is deconstructed, giving way to intersubjective relations. Once more we may start from the bifurcation pattern and look at the various possibilities. If a bifurcation pattern with two paths open has been previously set up, negation will amount to replacing the pattern by a pattern with a blocked path.

134 Alain Deschamps and Lionel Dufaye (6)

Books may not be taken out of the library.

It is interesting that semantically equivalent examples are possible without any previously defined patterns: (7) (8)

Books cannot be taken out of the library. Books must not be taken out of the library.

If the original pattern has an open path and a closed path, it may be questioned, explaining the use of needn’t as a negation of must. Both paths are reopened but the pre-construction of must will put more weight on E. It appears that this situation requires inter-subjective relationships, the speaker and the co-speaker using different modal patterns. If a single path is open leading to I, negation will amount to questioning the pattern and opening the path to E (from will to won’t, not very different from refuse). The same starting point may also be questioned and correspond to closing the path to I, without any attention being paid to E, as in: (9) He will have left it in the room. (10) He can’t have left it in the room. This brief survey of negative constructions illustrates the complexity of negation operations not easily definable in a two way system which only allows for negation of the predication or negation of the modal. Very often negative forms will be combined with non-neutral stress patterns, adding an extra parameter to the interpretation of modal values.2

4. The formal system of modals 4.1. Common operations guarantee neighbouring values In the previous section, we assumed a model in which each modal auxiliary can be described as the marker of two operations (quantitative and qualitative). It will now be argued that each operation is shared by two different auxiliaries. In other words, we will put forward a formal paradigm that is consistent with the idea that modal auxiliaries form a system – i.e. a paradigm of markers that are connected. To put the question

Topology and the English modal system 135

on a concrete footing, let us take the example of MUST and SHALL. Although these two auxiliaries are not interchangeable in many contexts, they are virtually equally acceptable in others. For instance: (11) He must /* shall have lost your address. (12) He must / shall do the washing-up. In the first example, the quantitative operation is foregrounded in the sense that, as in any epistemic context, we are concerned with the spatiotemporal extension of the event; in everyday language we say that an “event takes place”, which conveys approximately the same idea.3 In epistemic contexts, SHALL cannot be substituted for MUST. In the second example, which forces a deontic reading, it is not so much the question of whether the event will take place that matters, but rather the fact that the speaker evaluates the event as a desirable scenario. In other words, it is the qualitative operation that is foregrounded when MUST and SHALL are – to a certain extent – interchangeable. To account for such overlapping zones, we need to hypothesise that MUST and SHALL share a common operation. More technically, we will say that MUST and SHALL are connected within the modal system. If we take a look at the schematic representations put forward above we observe that there is indeed a qualitative operation that is common to MUST and SHALL (see Table 3: the relevant part is reproduced here):

I E

I E

I E

I E

\/ \

\/ \/

IE Qlt

IE Qlt

IE Qnt

SHALL

IE Qnt

MUST

By merging these two representations we obtain a string of operations that corresponds to a subpart of the whole system where SHALL and MUST connect:

136 Alain Deschamps and Lionel Dufaye

I E

I E

I E

\

\/

\/

IE

IE Qlt

IE

Qnt

SHALL

Qnt

MUST

But SHALL has another neighbour in the system, which is WILL. Indeed, table 3 predicts that WILL and SHALL must have an overlapping zone since they share the same quantitative operation:

I E

I E

\

\

IE Qlt WILL

I E

I E

\/ \ IE Qnt

IE Qlt

IE Qnt

SHALL

This quantitative operation means that SHALL and WILL can both express the coming to existence of an event and disregard any contrary scenario. In other words, if someone says “I will/shall do the washing-up”, it never implies that “the washing might also not be done”. All they are saying is that the event “will” take place without paying attention to potential contingencies. This is why the diagram shows only one path (to the positive value = Interior). We can thus add another chunk to the string by merging WILL and SHALL:

Topology and the English modal system 137

I E

I E

I E

I E

\

\

\/

\/

IE

IE Qnt

IE

IE

Qlt

Qnt

Qlt

WILL

SHALL

MUST

Table 3 also predicts that WILL must share a qualitative operation with CAN. This idea may seem counterintuitive at first sight since WILL and CAN are not usually associated as a “pair” of markers. However, examples such as the ones that follow are in keeping with our hypothesis: (13) This bucket can/will hold 2 gallons. (14) This tent can/will sleep up to 20 people. As shown in the diagram, the formal operation is more or less the same as the one shared by WILL and SHALL: one path leading to the positive value. The difference depends solely on the qualitative dimension of the operation shared by WILL and CAN. For instance, in an example like This tent can/will sleep up to 20 people, it is the properties of the tent that are described regardless of any actual situation so that it is the qualitative aspect that is preponderant in such contexts (compatibility = Qlt). A further piece can thus be added to the formal paradigm:

I E

I E

\/ \ IE Qnt

IE Qlt CAN

I

E

\ IE Qnt

WILL

I E

I E

\/

\/

IE

IE

Qlt SHALL

Qnt MUST

138 Alain Deschamps and Lionel Dufaye There remains one last step to complete the string: we have to introduce MAY and merge its formal representation with that of CAN so that the whole system will eventually look as follows:

I E

I E

I

E

\/ \/ \ IE Qnt

IE Qlt

MAY

I E

I E

I E

\

\/

\/

IE

IE

IE

IE

Qlt

CAN

Qnt WILL

Qlt SHALL

Qnt MUST

There is however one problem as regards this last shared operation, for MAY and CAN do not seem to commute in quantitative contexts. Let us take an epistemic context such as (15)

She may have lost your address.

It would be impossible to substitute CAN for MAY. (16)

*She can have lost your address.

This phenomenon can be accounted for by the fact that CAN always requires the activation of its qualitative facet. In other words, CAN has the characteristic of being “quality oriented” so that it never appears in purely quantitative contexts. Is there any observable evidence that MAY and CAN share a common operation? In fact, some contexts do seem convergent with the formal approach developed here: (17) Aspirin may cause allergic reactions. (18) Aspirin can cause allergic reactions.4 In both examples, the modality conveys some extensional contingency of there being situations where the drug causes allergic reactions and situations where it doesn’t. Although it is backgrounded, the quantitative

Topology and the English modal system 139

component is active and thus provides a straightforward explanation of why the two markers can commute in a number of contexts with minor semantic differences.

4.2. Differential operations guarantee contrastive values Hitherto, it has been argued that modal operations are “polymorphic”, in the sense that they can be shared by two different modals, thus explaining the commutability of certain auxiliaries in particular contexts. This commutability is always relative however, for any two modals can never be perfectly synonymous, no matter what the context may be. In other words, the selection of one modal rather than another always implies a semantic nuance. To account for this contrastive phenomenon we need to consider the fact that each modal is associated with two operations but shares only one operation with another modal. For example, two markers that share a common quantitative operation necessarily have different qualitative operations. Conversely, two markers that share a common qualitative operation necessarily have different quantitative operations. Hence, two markers can never be totally similar, although they can be partly similar. Let us return to MUST and SHALL for a concrete illustration of this principle: (19) He must do the washing-up. (20) He shall do the washing-up. As far the qualitative aspect is concerned, there is no difference between SHALL and MUST in this context: in both cases the positive scenario is valued as the “good” scenario whereas the negative scenario is ruled out (it is represented by a “crossed-out path” between the level of representation IE and the level of non-validation E):

140 Alain Deschamps and Lionel Dufaye

I E

I E

I E

\

\/

\/

IE

IE Qlt

IE

Qnt

SHALL

Qnt

MUST

To provide a simpler (although clumsier) paraphrase, we could say that in both cases: the speaker wants “the addressee to do the washing up” and does not want “the addressee not to do the washing up”. From a quantitative point of view however, the implications are different. MUST takes the negative scenario into account and rejects it. Whereas with SHALL, the negative scenario is not even envisaged (no path towards the negative zone). Consequently, SHALL will be much more assertive than MUST as regards the actualisation of the event. SHALL does not only state that “I want you to do the washing up” (qualitative aspect) but also that “the washing up will be done” (quantitative aspect). This quantitative facet is a lot weaker with MUST since the negative scenario, even though rejected, is taken into account as a potential case. Here is another example to prove the point. We can either say: (21) Lions can be dangerous. or (22) Lions may be dangerous. But only CAN is acceptable if the subject is singular: (23) A lion can be dangerous. (24) ?A lion may be dangerous.

Topology and the English modal system 141

The difference between the first 2 examples is minor. With CAN, the predicate is presented as a potential attribute of lions. We see from the formal representation that only the positive side of the domain is envisaged so that the plural lion-S refers to lions as a class of animals, one of whose notional properties is to be potentially dangerous. With MAY on the contrary, the qualitative component explicitly includes the negative counterpart as a possible scenario so that both and are presented as potential attributes. By taking into account both the positive and the negative scenarios, MAY activates the Quantitative component, that is to say the class of situations where is either validated or not validated. Consequently, the plural lion-S refers not only to the class of lions but also to the class of situations. This might explain why “A lion may be dangerous is odd unless we add something like “…if you creep up on it”. (25) A lion can be dangerous. (26) ?A lion may be dangerous. (27) A lion may be dangerous if you creep up on it. In the case of A lion can be dangerous the predicate is presented as a potential attribute of any lion. However, it would be strange to say that a lion, no matter which one, is both potentially dangerous and notdangerous, unless we envisage situational backgrounds (positive and negative: if you creep up on it) onto which this two-fold qualitative representation can be mapped. In other words, the quantitative component is activated with MAY, but remains backgrounded with CAN.

5. Conclusion Negative and interrogative clauses, levels of stress and preterit forms of the modals can be dealt with within the same theoretical framework, which would be too lengthy to talk about and will not be done here. But the main idea is that some sort of linguistic calculation of the different values of the modal in authentic examples is possible if we start from an invariant form for each modal with a few basic operations requiring previously defined parameters.

142 Alain Deschamps and Lionel Dufaye

Notes 1. 2. 3.

4.

Cf. Deschamps (2001) or Gilbert (2001). Cf. Rouskov-Low (2001). Remember that an operation is said to be Qualitative (or Qlt) whenever the speaker evaluates the event as either good/bad or compatible/incompatible; whereas an operation is said to be Quantitative (or Qnt) whenever we consider whether an event has taken / is taking / will take place. See Gresset (2001) for analyses of similar examples.

References Bouscaren Janine, Jean Chuquet, and Laurent Danon–Boileau 1992 Introduction to a Linguistic Grammar of English. Translated and adapted by Ronald Flintham, and Janine Bouscaren. Paris: Ophrys. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Cotte, Pierre 1988 Le Système des auxiliaires modaux dans le système verbal de l’anglais contemporain. Thèse de doctorat d’état, Université de Grenoble III. Culioli, Antoine 1990 Pour une Linguistique de l’Énonciation: Opérations et Représentations, Tome 1. Paris: Ophrys. 1994 Continuity and modality. In Continuity in Linguistic Semantics, Catherine Fuchs, and Bernard Victorri (eds.), 21–32. Amsterdam: John Benjamins. 1995 Cognition and Representation in Linguistic Theory. Michel Liddle (ed.). Amsterdam: John Benjamins. 1999a Pour une Linguistique de l’Énonciation: Formalisation et Opérations de Repérage, Tome 2. Paris: Ophrys. 1999b Pour une Linguistique de l’Énonciation: Domaine Notionnel, Tome 3. Paris: Ophrys. Deschamps, Alain 1998 Modalité et construction de la référence. In La Référence 1: Statut et Processus, Travaux Linguistiques du CERLICO n°11, Nicole Le Querler, and Eric Gilbert (eds.), 127–145. Rennes: Presses Universitaires de Rennes. 1999 Essai de formalisation du système modal de l’anglais. In Les Opérations de Détermination, Quantification / Qualification, Alain

Topology and the English modal system 143 Deschamps and Jacqueline Guillemin-Flescher (eds.), 267–285. Paris: Ophrys. 2001 Approche énonciative des modaux de l’anglais. In Modalité et Opérations Enonciatives, Cahiers de Grammaire Anglaise n°8, Janine Bouscaren, Alain Deschamps, and Lionel Dufaye (eds.), 3– 21. Paris: Ophrys. Dufaye, Lionel 2001 Les Modaux et la Négation en Anglais Contemporain. Paris: Ophrys. Gilbert, Eric 1987 May, Must, Can et les Opérations Enonciatives, Cahier de Recherche en Grammaire Anglaise n°3. Paris: Ophrys. 2001 Vers une analyse unitaire des modalités. In Modalité et Opérations Enonciatives, Cahier de grammaire anglaise n°8, Janine Bouscaren, Alain Deschamps, and Lionel Dufaye (eds.), 23–99. Paris: Ophrys Gresset, Stéphane 2003 Towards a contextual micro–analysis of the non–equivalence of MIGHT and COULD. In Modality in Contemporary English, Roberta Fachinetti, Manfred Krug, and Frank Palmer (eds.), 81–102. Berlin/New York: Walter de Gruyter Groussier, Marie-Line 2000 On Antoine Culioli’s Theory of Enunciative Operations. Lingua 110.3: 157–182. Larreya, Paul 1984 Le Possible et le Nécessaire, Modalités et Auxiliaires Modaux en Anglais. Paris: Nathan. 1997 Notion et opérations modales: pouvoir, devoir, vouloir. In La Notion, Claude Rivière and Marie-Line Groussier (eds.), 156–166. Paris: Ophrys. Leech, Geoffrey 1987 Meaning and the English Verb. London/New York: Longman. Palmer, Frank R. 1974 The English Verb. London: Longman. 1979 Modality and the English Modals. London: Longman. 1986 Mood and Modality. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Press. Rouskov–Low, Jennifer 2001 A prosodic approach to modality. In Modalité et Opérations Enonciative: Cahier de Grammaire Anglaise n°8, Janine Bouscaren, Alain Deschamps, and Lionel Dufaye (eds.), 255–282. Paris: Ophrys.

Epistemic might in the interrogative Gilbert Ghio

1. Introduction Modals are traditionally divided into root and epistemic modals. While the root modals are compatible with the interrogative form it is generally accepted that the epistemic modals are incompatible with such a form. For example one cannot say : *May he have missed the bus? or *Must he have had an accident? because the predicate following an epistemic modal stands both for the positive and negative value of this predicate (e.g. It may rain implies It may rain or not rain) and in the interrogative one can question either one value of the predicate or the other but not both simultaneously. A. Deschamp (2001: 14) says : We can also see that the interrogative form is excluded with epistemic may and must ….Only a scanning operation 1 on two values taken separately in turn and on a par can justify a question. Only will and can are compatible with these prerequisites. [translation mine]

and F.R. Palmer (1987: 56): We seldom question epistemic modality. Only the modality can be questioned and can is used : ‘Can they be on holiday?

But if Can they be on holiday ? is acceptable so is Might it rain this afternoon?. How then can we account for examples where might is possible but may is not although they are semantically equivalent? (1)

But for the time being, it gives Japanese firms a small advantage. Might their biggest advantage, however, lie in their most inscrutable institution of all, the Tokyo stockmarket? (BNC ABJ 3838)

(2)

And yet the Vietnamese government is hoping. Might this be the beginning of the end of America’s trade embargo against Vietnam, a

146

Gilbert Ghio

hint that diplomatic relations will be established between the two countries? (BNC ABJ 1075) (3)

Yet these side effects will exacerbate powerful deflationary forces in the real economy, thus delaying the recovery which investors are so eagerly anticipating. Might the stockmarket ‘price-keeping’ operation backfire after all? (BNC CRC 2863)

(4)

Only in size and restricted contents (advertising goods, charities, insurance, unit trusts) does junk mail differ much from the newspaper or magazine as a medium. Might advertising be regarded as a mass medium on similar grounds? (BNC CRY 185)

In this paper I intend to show that the epistemic form might is compatible with the interrogative because of the hypothetic context. I will first analyse the examples to show that we are really dealing with epistemic modals, that a pre-construct (see 2.2 for this term) exists in each case and that, in the interro-negative, the negation bears on the predicate, not on the modal. I will then suggest a topological representation of the linguistic operations at work and finally attempt an explanation. My explanation fits well into the theoretical framework of A. Culioli, as explained by M-L. Groussier (2000: 157-182). Within this framework, it reads as follows.

2. Analysis 2.1. The epistemic status of might The majority of the examples contain stative verbs (be, lie) which favour an epistemic interpretation. That we are dealing in these cases with epistemic modality is also made explicit in some examples by the co-occurrence of markers found in epistemic contexts such as : perhaps, possible that, wonder whether : (5)

And he thought Imran ‘unapproachable’ during the world cup. Might he perhaps have been happier staying in Lahore to help in his father’s spare-parts business? (BNC CU1 1722)

(6)

In order to make that pattern more obvious, the statistician then draws through the points a ‘best line’ – that is, a line which repre-

Epistemic might in the interrogative

147

sents the general trend, even though some of the points lie well clear of it. Might it not be possible that ours is indeed the only Universe, and that it represents the ‘best line’ through all the trillions of probabilities? (BNC CET 1557) (7)

He stopped for a moment, the crumpled sheet of prescription pad on his knee, and wondered whether Donald was really worth all this effort. Might it not be simpler just to let him go? (BNC ASS 17994)

We also find the aspectual markers have + en and be + ing which favour an epistemic interpretation too, as in (5) above or in (8). (8)

It occurred to him as he walked that he had not seen Taheb since their visit to the palace compound. Might she be expecting him to visit her, or at least to send her a message? (BNC H84 2502)

2.2. The existence of a pre-construct2 These examples also contain adverbs such as yet, but, however, only which are the mark of an adversative context, i.e. they signal that the speaker starts from the opposite value (negative) of the predicate : (1)

But for the time being, it gives Japanese firms a small advantage. Might their biggest advantage, however, lie in (… ) the Tokyo stockmarket?

(2)

And yet the Vietnamese government is hoping. Might this be the beginning of the end of America’s trade embargo…

(3)

Yet these side effects will exacerbate powerful deflationary forces in the real economy, thus delaying the recovery which investors are so eagerly anticipating. Might the stockmarket ‘price-keeping’ operation backfire after all?

(4)

Only in size and restricted contents (…) does junk mail differ much from the newspaper or magazine as a medium. Might advertising be regarded as a mass medium on similar grounds?

From these examples we can infer the following pre-constructs :

148

Gilbert Ghio

(1’)



(2’)



(3’)

< the stockmarket operation - not backfire >

(4’)

< Junk mail - not be a mass medium >

Sometimes the trace of the opposite value (this time the positive one) is to be found after the modal : (9)

(9’)

There are many cases in the Digest in which there is a confusion between legacy and trust. Might this suggest that the difference was no longer regarded as very important? (BNC B2P 1490) < The difference between legacy and trust – be important >

Sometimes the change in point of view is not as explicit but can be traced in a sentence. (10) And this Thiercelin who was persecuting him, now he came to think about it, much resembled young Yakovlev, whom he had slain in sport all those years ago .Might it be him in disguise, returned to spy on him and seek revenge? (BNC B20 1350) (10’) now he came to think about it → before he thought the opposite (i.e. < it – not be Yakovlev >). In each case a pre-constructed value, opposite to that introduced by might, can be traced. This pre-construction of the opposite value of the predicate shows that the utterer works on the two opposite values of the predicate (x and non-x) and not those of the modal (might/mightn’t). In this case, could, which would imply that the value introduced by the modal is opposed to all other possible values (x1, x2, x3…) couldn’t be substituted to might without a change of meaning. As P. Cotte (1988: 914-915) says : Could does not allow a return to x but derives x from what grounds it” [the extra-linguistic world] and “With could the alternative is positive (x1, x2, x3…) and not x versus non-x. [translation mine]

Epistemic might in the interrogative

149

For example, in (4) could would be a root or dynamic modal meaning ‘would it be possible for advertising to be regarded as a mass medium’, as opposed to something else, whereas might means ‘Is it the case that advertising…’, as opposed to ‘not the case ’ i.e. ‘not regarded’, the negative value. Might serves to re-introduce the positive value ‘regarded’, which was excluded in the pre-construction. As S. Gresset (1999) says about might in his contrastive study of COULD and MIGHT : I consider x as a non-precluded value because for me non-x [pre-construct] is the case. [translation mine]

It is the introduction of an alteration in the situation signalled by the aforesaid adverbs and implying a change of point of view from the utterer which makes x (the positive value) possible again. This value is constructed not simply as possible but as ‘not-impossible’ or ‘not-precluded’ as underlined by such expressions as ‘after all’ (3), ‘on similar grounds’ (4). This is corroborated by the translation into French where the negative interrogation is needed to account for the fact that the two opposite values of the predicate are taken into account by the utterer (whereas could would be rendered by the simple interrogative form), the negation being the trace of the negative pre-construct. (1’’) Leur plus grand avantage ne pourrait-il pas résider… ? (2’’) Ceci ne pourrait-il pas être le début… ? (3’’) L’opération de maintien des prix de la Bourse ne pourrait-elle pas se retourner… ? (4’’) Les publicités ne pourraient-elle pas être considérées comme… ? Indeed many examples in the corpus are in the interro-negative in English.

150

Gilbert Ghio

2.3. The scope of the negation In the case of the negative interrogation, might serves to re-introduce the positive value previously barred on the plane of validation (because the negative one has been chosen) while at the same time the negative value is being questioned on the plane of representation 3 : (11) <something missing> t about the rest of it?” asked the Substitute, lighting a cigarette, “Up there. Mighn’t the girl be hidden on the mountain?” BNC CJX 107 (12) To hate yourself is to hate your body. Might not that message have been taken up by my immune system? (BNC B19 111) (13) Once the Bible began to be approached as historical material, open to critical historical study, it was natural that the NewTestament, and in particular the life and personality of Jesus himself, should come to be re-examined. Might not Jesus himself have been rather different from the picture given of him in the gospels and the subsequent teaching of the church? (BNC CL6 242) (14) But all the nuclear powers except China have suspended nuclear testing for the time being. Might this not be the moment to give nuclear disarmament a push, by keeping the test–ban promise? (BNC CRA 122) F.R.Palmer (1987: 108-109) says about such examples : The modality is negated by can’t (‘It is not possible that…’), the proposition by may not (‘It is possible that… not…’ ).‘May/Might not John be in his office?’ = isn’t it possible that John is in his office?

but he goes on : This could, however, also be taken to be a question (especially with stress on not) about the negative proposition : ‘Is it the case that John may/might not be in his office?’[emphasis mine] But there is no way, except with a tag, of expressing questions about the negation of the modality (John can’t be in his office, can he?).

indicating that the negation bears on the predicate, not on the modal.

Epistemic might in the interrogative

151

When the negation is placed after the modal, it is only because of the length of the subject : the girl (11), that message (12), Jesus himself (13). When the subject is short, it (6, 7), this (14), the negation is always near the predicate Besides, epistemic modals, because they deal with both values of the predicate at the same time (see introduction), cannot be negated 4. Finally it must be mentioned that it is the value of the predicate which is considered as hypothetical, not the modality. It is necessarily hypothetical because its opposite (pre-construct) has been validated as true to the extralinguistic world. Thus an adequate paraphrase would not be : *Would it (not) be possible that it is…? but rather : Is it (not) possible that it is…? or, as F.R. Palmer suggests ( 1987: 108-109) : Is it the case that it might (not) be? To sum up this analysis we can say that, because of the epistemic status of might, the speaker can re-introduce a value previously excluded in the pre-construct, that what is questioned is the positive or negative value of the predicate, not that of the modality and that it is this value which is regarded as hypothetical and not the modality itself. Having established the epistemic value of the modal, the existence of a pre-construct and the scope of the negation we can now try to explain why epistemic might is compatible with the interrogative form whereas may isn’t.

3. Topological representation In all the examples studied above the utterer is working on three different planes : the plane of validation which corresponds to the reality of things ( the state of things as they are, what ‘is the case’) and where the preconstruct is situated, a hypothetical plane from which he can consider the other value of the predicate without excluding its opposite (pre-construct) and a plane of representation (at the level of the notion) where both values are present so that he can question one without excluding the other. With epistemic might he can play on both values of the predicate, positive (I) and negative (E) and at the same time isolate one to question it as shown on Fig. 1 and Fig. 2. 3.1. The interrogative form (2)

And yet the Vietnamese government is hoping. Might this be the beginning of the end of America’s trade embargo against Vietnam, a

152

Gilbert Ghio

hint that diplomatic relations will be established between the two countries? (BNC ABJ 1075)

E

Plane of validation (pre-construct)

I

might

E

I ?

Hypothetical plane (I biased)

Plane of (I,E) representation

Figure 1. Interrogative

The speaker starts from a negative pre-construct : <no hope> which implies < not the end of America’s trade embargo >. This value is pre-asserted 5 on the plane of validation. Hence the opposite value (I, positive) cannot be considered on the same plane. The ‘unreal’ preterit might enables him to consider it on a hypothetical plane without excluding the negative one (E, dotted line on Fig.1). Then he takes us to the plane of representation (that of the notion 6) where both values (I and E) co-exist. From there he can question I and choose either I or E. The positive one (I) is biased because it is considered twice, once with might and once with the interrogation.

Epistemic might in the interrogative

153

3.2. The interro-negative form (13) Once the Bible began to be approached as historical material, open to critical historical study, it was natural that the New Testament, and in particular the life and personality of Jesus himself, should come to be re-examined. Might not Jesus himself have been rather different from the picture given of him in the gospels and the subsequent teaching of the church? (BNC CL6 242)

Plane of validation (pre-construct)

E

I

might E

? Hypothetical plane (I biased)

I Plane of

(I,E)

representation

Figure 2. Interro-negative

The presupposition here is that the Bible cannot be wrong which implies <Jesus – not different from the picture given in the Gospel>. Here again this value is pre-constructed and therefore pre-asserted on the plane of validation. This time the speaker starts from the image of E on the hypothetical plane. Its opposite, I, is not textually present and cannot be considered from the plane of validation (crossed line on Fig.2). Might enables him to re-introduce this value previously barred. Then he takes us to the plane of representation where he can question E bringing us back to I or E. I is biased because it is re-introduced twice, once with might and once with the questioning of E.

154

Gilbert Ghio

4. A combination of three factors 4.1. Unreality What makes epistemic might and the interrogative form compatible is the hypothetic status of the utterance. Here might is equivalent to a conditional clause. In (2), for example, we can say (15). (15) Might this be the beginning..? = And what if this was the beginning? In both cases the speaker situates himself on a plane different from the plane of validation. This enables him to re-introduce one value while not excluding its opposite, as shown above. A.Culioli (1990: 149) says about hypothesis and the conditional mode : All contradictions are suspended. The utterer can imagine every possibility : he can consider what is not asserted and assert it on a hypothetical mode. [translation mine]

It is because one value, either I or E, can be asserted (even if it is on a hypothetical mode) that it can then be questioned by a return to (I,E) on the plane of representation while the other value is still present on the hypothetical plane.

4.2. An epistemic value The pre-construct being a pre-asserted value, it precludes its opposite. It is an alteration in the situation (marked by yet, but…) which enables the speaker to reconsider it by re-introducing the opposite value, I. But he can do it only on a hypothetical plane and with the help of an epistemic form which enables him to retain both values (I and E) , otherwise the preconstructed value E would be asserted and the question would no longer be possible. I is re-introduced either directly from the plane of validation or from the image of E on the hypothetical plane (in the interro-negative). S. Gresset (1999) says about the use of might versus that of could : …x is put forward as the value that the utterance aims at constructing and which is the value favoured by the utterer who cannot exclude non-x which he knows is the case. [translation mine]

Epistemic might in the interrogative

155

The bias on I points to the value the speaker favours without being able to assert it because its opposite E has been pre-asserted in the pre-construct on the plane of validation.

4.3. The interrogative form All the examples are in narrative contexts, not in dialogue. The question is a rhetorical one which the subject asks himself (Cf. (7) : he wondered whether). There is no answer because no answer is needed since a value of the predicate (the negative one, pre-construct) has already been validated as true to the extra-linguistic world. In the case of the interrogative form, starting from E (pre-construct) the utterer moves to I only to question it in (I,E) which brings him either to I or E. In the case of the interro-negative form, starting from E (image of the pre-construct on the hypothetical plane), he questions it, which brings him back to I again, but E is still present. It is this playing on the two opposite values of the predicate in a circular movement without ever being able to stop on one which conveys the hesitations of the subject.

5. Conclusion It is only with might in this particular context that such a phenomenon can be found. It is the result of the combination of a pre-construction and two scanning operations (epistemic modality and interrogative form) on the two opposite values of the predicate (x and non-x). These operations serve a double purpose : to convey the hesitations of the speaker and at the same time point to the value which he favours (bias on I) but cannot assert because its opposite E has been pre-asserted in the pre-construct. This is possible only if the utterer can consider both values of the predicate simultaneously and question one at the same time. In order to do this, he must situate himself on a hypothetical plane (because one value has already been chosen in the pre-construct) from which he can select one value without excluding the other and then question it by a return to the plane of representation (I,E). This he can do with might but not with could or may. A consequence of this bias on the positive value of the predicate (I) in the interrogative and interro-negative is that, in these examples, might and

156

Gilbert Ghio

might not are equivalent from a semantic point of view as is attested by the translation into French which makes use of the interro-negative in both cases. But the linguistic operations involved in English are slightly different.

Notes 1.

2.

3.

4. 5.

6.

A scanning operation consists in considering every possibility without choosing one in particular because none is suitable, or the speaker doesn’t know which is suitable, or all are equally suitable. The Yes/No questions and the hypothesis are typical of this operation bearing on the two areas (I, interior and E, exterior) of a notional domain (see this term, note 5). A pre-construct is a predicative relation (operation which consists in relating a subject and a predicate) which is not actually worded in the context (generally before) but can be inferred from it. In “This job is not really interesting, yet I do like it” the pre-construct is “I don’t like this job” A. Culioli distinguishes three planes in relation to modality : 1) The plane of representation where the speaker situates himself in order to operate his choice of modality concerning the notional domain (I,E) (see this term, note 5). 2) The plane of validation where the speaker decides if the predicative relation (see this term, note 2) is true or false according to reality. 3) The hypothetical plane which is distinct from the plane of validation ecause the predicative relation placed on this plane by the speaker does not have to be true or false according to reality. The modal choices made by the speaker depend only on his imagination, without reference to the state of things as they. Except can in such examples as ‘He cannot not go’ where the epistemic status is disputable. A value (of the predicate) is said to be asserted when it is in conformity with the reality of things. Hence it is possible to speak of a positive or negative assertion according to whether what is referred to is true or false. A notion like TABLE for example, is made up of all the properties in common to all forms of TABLE. Around this notion can be built a notional domain composed of an Interior, I (what TABLE refers to) and an Exterior, E (what TABLE does not refer to). There are also borderline cases (not really TABLE but something like it) which form what is referred to as the Boundary or Frontier.

Epistemic might in the interrogative

157

References Cotte, Pierre 1988 Le système des auxiliaires modaux dans le système verbal de l’anglais contemporain. Thèse de Doctorat d’Etat, Université de Grenoble III. Culioli, Antoine 1990 Pour une linguistique de l’énonciation. Opérations et représentations. Tome 1. Paris : Ophrys. Deschamps, Alain 2001 Approche énonciative des modaux de l’anglais , Cahiers de Recherche 8: 3–21. Gresset, Stéphane 1999 CAN/MAY, MAY/MIGHT, MIGHT/COULD, ou l’interchangeabilité en question. Thèse, Université de Paris X Nanterre. Groussier, Marie-Line 2000 On Antoine Culioli’s theory of enunciative operations. Lingua 110: 157–182. Palmer, Frank R. 1987 The English Verb. London: Longman.

MAY in concessive contexts Jean-Claude Souesme

1. Introduction According to some linguists1, the modal may can have a concessive interpretation, apart from its basic root and epistemic values. We shall first contend that this modal has no concessive value as such, but that the context only is responsible for it: for instance, in order to give a sentence a concessive value, it is necessary to introduce but before the second clause after an independent one bearing may: (1)

Borrowing is a positive thing, lending negative. To be a borrower may be uncomfortable, but it is also exhilarating, instructive, lifelike.

But is the true marker that gives the utterance its concessive interpretation since it reveals the crossing over the boundary between the interior and the exterior of a notional domain2 as we explained in Souesme (1995). Secondly, in adversative contexts introduced by although, though, etc, these adverbs are enough to introduce the idea of a concession, so the presence of the modal itself does not play any real part in the matter. Though introduces a value (P) which does not exclude another one (Q) while Q’ was supposedly implied by P. As mentioned in the Collins Dictionary, though is used: – to add your opinion to a statement that you know is true and that your opinion cannot affect. – to prevent someone thinking something that may follow logically from what you have just said, but is not true. We will also contend that may is not the result of an enunciative operation3 between two co-utterers: the enunciator alone is responsible for the introduction of may, it is part of his enunciative strategy. Nor does may come as the result of the projection of the root value of permission on the enunciative plane: on the contrary, we will contend that its value is closer to the epistemic one.

160 Jean-Claude Souesme But what is its true value ? Some examples seem to indicate that when the modal may is used, its value is very close to the epistemic one, not to say properly epistemic: (2)

‘Goodbye. I’m really disappointed in you, John. You may be very clever – I wouldn’t know about that – but you have a lot of growing up to do.’ (Coetzee, Youth)

As a matter of fact, there often exists an opposition between the first part of the sentence in which the statement is apparently non-assertive and the second one in which the utterer definitely takes a stand in favour of the proposition; we can even notice the use of adverbs like certainly, or the use of do as the marker of the utterer’s position in favour of the positive value of the second predicate: (3)

It is like the weakness in the character of a great man, which may make him less admirable but certainly makes him more interesting. (Somerset Maugham, Honolulu)

(4)

What will cure him, if it were to arrive, will be love. He may not believe in God but he does believe in love and the powers of love. (Coetzee, Youth)

(5)

The channel claims peak-hour viewership of up 500,000 Italian households. The numbers may be small compared to the major networks, but they do contain a fabulous little secret. (Time, November 10, 2003) Let us now compare these two examples:

(6)

Love may be blind, but that does not last.

(7)

Love is all fine and dandy, but we all know that hate makes for better copy. (Time, November 10, 2003)

In example (7), the utterer juxtaposes two statements about two different notions which, according to him, stand in opposition to each other, and a comparison is made between love and hate, whereas in example (6), the utterer makes a statement about the properties of love: while he is not too sure about the truth value of the first one, he is quite definite about the sec-

May in concessive contexts 161

ond one. So we might be tempted to consider that may has an epistemic value here. We will none the less consider that a distinction has to be made between the usual and proper epistemic value of the modal and the value it has when occurring in a sentence with but; yet it is a subtle one: only through the punctuation can we make a distinction between the two. The mere presence of a full stop instead of a comma enables the reader to recognize the true epistemic value of may: (8)

It may not be easy to maintain such bravado amid the danger that Iraq’s police face. But many Baghdad policemen believe things will soon get better. (Time, February 23, 2004)

We also noticed that, in the presence of a comma, may will retain its epistemic value if, and only if, it is followed by may not: (9)

The Kremlin mayor may or may not love the winner, but that’s not the point. (Time, May 24, 2004)

The following example is quite revealing in this respect: (10) There may have been ancillary benefits from this show of force: Libya has given up its nuclear ambitions; Iran may, or may not, be doing the same. But the situation in Iraq is chaotic... (Time, February 23, 2004) The first may in may have been is definitely non epistemic as the next sentence about Lybia shows, whereas the second is, as is revealed by the presence of may not. We have just pointed out that in example (10), the first occurrence of may is non epistemic, since the truth-value of the statement is unquestionable. We found a great many examples of that sort, and the most obvious one is: (11) The sun may rise in the East, but it sets in the West. May cannot be considered as epistemic either in: (12) Bush’s ratings may be slipping, but we live in a terrorist age, and 2004 may be the first election in decades in which polling patterns in

162 Jean-Claude Souesme May end up predicting nothing about November. (Time, May 24, 2004) Bush’s ratings were slipping at that time, as the polls showed on the same page of the Time issue. Likewise, the utterer does not deny the absence of validation of the relative predication4 in: (13) Jesus Christ, I may not be any Cocteau or Brecht, but at least I’m serious. So we agree with Larreya and Rivière (2005: 99) when they claim that the utterer does not question the truth-value of the predicative relation in which may is inserted. However, in some other contexts, we would be tempted to say that the utterer refuses to endorse the validity of the predicative relation himself, or he simply refuses to debate upon it, or he seems to take it as granted. In the following examples, the presence of the adverb also seems to imply that the utterer does not express his doubt about the validity of the first predicative relation; he seems to go as far as conceding it to his implicit co-utterer / reader: (14) Drinks cost up to $7 each and last orders are at 4:30 a.m. It may be a sinfully late hour, but it’s also just in time for morning prayers. (Time, May 17, 2004) (1)

Borrowing is a positive thing, lending negative. To be a borrower may be uncomfortable, but it is also exhilarating, instructive, lifelike.

(15) It may have been awkward to have cameras meet troops when they landed, but wasn't it also appropriate? (Time, December 28, 1992) The same remark would apply to Quirk et al.’s example: (16) If he’s poor, he’s (at least) honest. which they alter into: (17) He may be poor, yet he’s also honest.’

May in concessive contexts 163

However, if the second part of the statement is also true, it should normally imply that the first one is as well. So may should not be considered as a means to express some sort of doubt upon the validity of the predicative relation, all the more so as in the following example, the utterer immediately gives the proof of what he has just asserted with may: (10) There may have been ancillary benefits from this show of force: Libya has given up its nuclear ambitions; Iran may, or may not, be doing the same. But the situation in Iraq is chaotic… (Time, February 23, 2004) A third option envisaged by Lapaire and Rotgé (1992: 487) would be to consider that whether the predicative relation is validated or not does not matter. What does matter is what is stated in the second part of the sentence, as in examples 18 and 19: (18) She may be the greatest soprano of any century, but we have no way of judging. (The Sunday Telegraph Magazine, 22-3-1981) (19) I’ve been listening to this stuff for years. It’s sheer crap. He may be intelligent but he’s got a screw loose somewhere. So, to sum it up, comments and appreciation would vary from one example to another, and differ whether a co-utterer is implicitly implied or not: in some cases, the utterer does not take a definite stand upon the validity of the predicative relation, which could be glossed as it may be true, etc, while in others, he seems to imply I will not argue about that, since it is beside the point; we shall then posit that may serves as a verbal marker opening the way to the validity of another predicative relation which is not impeded by the validity or validation of the previous predicative relation to which may has been applied. But how does this relate to the usual interpretations of the modal?

2. The fundamental value of MAY Let us start with what we consider to be the fundamental value of may. When using this modal, the utterer posits one value (which we will call I, as being the interior of the notional domain but the other values of the notional domain (F and E respectively corresponding to the frontier and the

164 Jean-Claude Souesme exterior of the notional domain) are not ruled out. Fundamentally, MAY marks a bifurcation between two paths, towards two possible values: though one path is opened which was previously barred, the second path still remains valid. This applies when may is used with its root value. When you say: You may go out it means that you can either stay or go. Stay in is still a possibility. With may, the utterer takes into account the two values p (= go out) and p’(= not go out) of the predicate go out. With the epistemic interpretation, one value has been chosen, but this does not imply that any other value belonging to the linguistic complementary5 of the notional domain is not valid as well. This remark is valid too when may is the marker of a possibility as in: (20) Let’s have a quick lunch so that we may make an early start this afternoon. which means that a new path will thus be open, so that the co-utterer will be left with two options, either leave early or not. In the present case, we are working on the validation of two different predicates applied to the same subject. So we are working on two different predicative relations. We may then consider that may plays about the same role, in so far as it announces that one value supposedly excluded before (as being part of the enunciator’s pre-construction), or antagonistic to the first, will be introduced by but in the second part of the utterance. As a matter of fact, but reveals the crossing over the frontier towards the exterior of the notional domain, since what was supposedly considered as excluded is now posited as certain. Several markers such as yet, at least, are quite revealing of this crossing over the frontier between the interior and the exterior of the notional domain: (16) If he’s poor, he’s (at least) honest. [He may be poor, yet he’s also honest.] (borrowed from Quirk et al) (13) Jesus Christ, I may not be any Cocteau or Brecht, but at least I’m serious. (21) Literature may not be as noble as mathematics, but at least there is nothing about literature that intimidates him. (Coetzee, Youth)

May in concessive contexts 165

(22) Blasphemy may no longer be an enforceable crime. But we can at least show that it doesn’t pay. (23) Hotel art may never improve, but at least in Paris, hotel food is becoming artistry itself. (Time, April 26, 2004) Under such considerations, it no longer seems necessary to call for a hypothetical or implicit co-utterer, as Bouscaren or Larreya among others have stated6.

3. Contexts in which MAY is used Before trying to explain why may has to be used in such examples, we shall consider the various contexts in which may is used with this still undefined value. Most occurrences are to be found when it is all about states or properties of the subject which are deemed incompatible: (24) He may be only nineteen, but he is dependent on no one. (Coetzee, Youth) Among other things, for any enunciator, “being only nineteen” usually implies “being dependent on one’s parents”. So the second clause introduces another property of the subject which does not stand in conformity, or in accordance with, what has been implied by the notional property mentioned before and the same remarks would apply to examples (25) to (27). (25) Caroline may not be the mysterious, dark-eyed beloved he came to Europe for, she may be nothing but a girl from Cape Town from a background as humdrum as his own, but she is, for the present, all he has. (Coetzee, Youth) (26) He should have taken the initiative. ‘I am not what I may seem to be,’ he should have said. ‘I may look like a clerk, but in reality I am a poet, or a poet to be.’ (Coetzee, Youth) (27) Burchell and his men may be dead, and their wagons turned to dust, but they really lived, their travels were real travels. (Coetzee, Youth)

166 Jean-Claude Souesme But may is also used with the same value when two events7, or states of affairs, are considered as vying with each other: (28) London may be stony, labyrinthine, and cold, but behind its forbidding walls men and women are at work writing books, painting paintings, composing music. (Coetzee, Youth) (21) Literature may not be as noble as mathematics, but at least there is nothing about literature that intimidates him. (Coetzee, Youth) (29) He may have his own vision of London, but there is nothing unique to that vision. (Coetzee, Youth) (30) Subtitle: Forget about flu – you’re more likely to get sick from a pet TITLE: Bird flu and SARS may make the headlines, but if you’re really worried about getting an animal disease, you need look no farther than the fluff ball curled up on your pillow. (Time, March 15, 2004) (31) Winnie’s story may evoke images of the Third World workshop, but her reality takes place much closer to home. (K. Franklin, Australian Broadcasting Corporation, 2003) A third possibility is when a state of the subject stands in contradiction with a situation or state of affairs, whose existence the utterer considers as definite: (32) If Pialat is well hated by his colleagues in France (...) he insists in every new film on going his own, deeply unconventional way. With “Under the Sun of Satan”, he has made a film that audiences may hate or love but which they will not be able to ignore. The example provided by Lapaire and Rotgé: (33) He may be rich but I still don’t want to marry him. shows that the second state of affairs does not stand in conformity with what is implied by the notional property mentioned before. Still marks the refusal to cross over the frontier implied by He is rich (implying: he is marriageable).

May in concessive contexts 167

Another context is when a property of the subject stands in contradiction with a present property considered as the result of a past event: (34) Afghanistan’s interim government may be inexperienced, but it’s already learned one valuable skill: how to raise money. (Time, April 12, 2004) We also came across the opposite relation, that is to say, between a state of the subject resulting from a past event and another property of the subject: (35) Bill Belichick may have revolutionized American football, but he could walk through the streets of Seoul or Sao Paulo unrecognized. (Time, May 24, 2004) Sometimes, a present resulting state of the subject is considered as standing in opposition with a present situation, or state of affairs: (36) Continental governments may have put family-friendly policies in place, but it seems only a small minority of Europe’s new dads feel they can afford to stay at home. (Time, May 10, 2004) (37) BUSH MAY HAVE RAISED A HUGE SUM FOR his re-election campaign, but money can’t buy competence or credibility with the voters. (Time, April 5, 2004) A situation referring to a present state of the subject can be considered as encompassing a property that is definitely different from the one mentioned in the following clause: (38) He may be living in England, but it is certainly not by invitation of the English working class. (Coetzee, Youth) In some other examples, may is used when referring to a past situation in which two properties are considered at odd with each other: (39) It may have been awkward to have cameras meet troops when they landed, but wasn't it also appropriate? (Time, Dec. 28, 1992)

168 Jean-Claude Souesme 4. Why MAY has to be used in such contexts In order to better understand the role played by may in such contexts, it is necessary to go back to the predicative level itself. In all the examples studied up to now, we have noticed that the second part of the sentence brings in an alteration among the notional properties of one of the predicative notions used either as a subject or as a predicate. It is then necessary to destabilize8 the predicative relation and the only possibility to do so is to go back to the notional level in order to recapture the notional properties involved and reject those that are not9. The modal may is fundamentally up to doing so, to de-actualizing the predicative relation in a way, since its fundamental function is to leave the two paths open towards the interior and the exterior of a notional domain: none of them is barred. As a consequence, in the contexts presently studied, thanks to the use of may, the second part of the sentence is being heralded as introducing, or referring to, a property which has been considered by the enunciator as foreign to the notional conformity of the previous predicative relation. By notional conformity, we mean that a property like “being dependent on one’s parents” (cf. example (24)) stands in conformity with what everybody understands when we say “be only nineteen”. Thus we can easily see why there is a plurality of meanings concerning the first predicative relation: if the enunciator wants to assert Q, he is obliged to de-actualize P in order to reach the notional level and consider the notional properties of the constituents of the predicative relation. As a consequence, being obliged to destabilize the predicative relation means he cannot endorse its validation. The first relative predication ( in example (39) for instance) being thus de-actualized with the use of may, the second one ( in the same example) will by contrast stand as asserted and endorsed by the enunciator, but this is more an “effet de sens” than anything else.

5. A false equivalence between MAY and THOUGH 5.1. may in subordinate clauses introduced by though We might also consider that all the clauses with may presently under consideration can be altered into subordinate clauses beginning with though or although. This is probably why some linguists have considered may as

May in concessive contexts 169

concessive. But this argument is not valid as it does not answer the question why may is also used in clauses introduced by though, although, which as such is but one more proof that may has no concessive value in itself. Gauthier (1997) and Filippi (1997) have already studied the part played by such conjunctions as far as concession is concerned. When though, although are used alone, two events are opposed to each other: an assertion (called P) implies another assertion (called Q’) as the expected value, while Q is the right one. Though then introduces one value, P, which does not exclude Q as was supposedly implied by P; but in such contexts, we are faced with two separate and disconnected events: (40) She wore a thick fur coat, though I could not see her properly. (borrowed from G. Ranger) The use of may in concessive clauses is not redundant at all then; it implies that, according to the utterer, there exists an alteration among the notional properties of the subject: (41) For though the tape may have been Fakhet’s final testament, it’s hardly the last we’ll hear from the terror groups that may have helped him. (Time, April 19, 2004) (42) Though a handful of U.S. troops may be responsible for Abu Ghraib, it is the thousands of servicemen and –women who are in Iraq and who, like the troops from the Rock of the Marne, may be going back there soon who have to face the rage the prison abuse has stoked. (Time, May 24, 2004) (43) Though the revelations may come as a surprise, the commission’s attention has moved steadily toward Iran in recent months. (Time, July 26, 2004) In the last example for instance, the use of may is justified by the fact that it is not just two events as such that are brought together, but what the first proposition implies concerning the resulting state of the Commission; what these revelations imply is considered to be at variance with the present notional property of this Commission (= their attitude toward Iran) to be derived from its behaviour during the last few months. Likewise, in the following example:

170 Jean-Claude Souesme (44) Although Miss Copley may be a very old friend, Sir Louis is a very influential one and the Gallery needs influential friends. (A. Wilson, Realpolitik) the use of may is due to the fact that a property of Miss Copley’s implied by the first part of this sentence (proposition P) is at odds with a property implied by the second statement. May enables the co-utterer to recreate a notional domain10 out of the property be a very old friend, and here again the second part of the sentence is being heralded as introducing, or referring to, a property of another subject which has been considered by the enunciator as foreign to the notional conformity of the first notional domain. Things get clearer when we consider the context of the short story: Sir Louis who wishes to come on a Tuesday, but Tuesdays are always booked by Miss Copley. For the co-enunciator, Miss Copley’s property be a very old friend implies have precedence over, but according to the enunciator, the fact that Sir Louis is a very influential person necessarily implies have precedence over Miss Copley, so that his visit should not be postponed; it should take place on the very day Miss Copley usually comes and visits her friend. If may had been left out, the sentence would have been rather difficult to understand: we would be induced to take both events (i.e. Miss Copley being a very old friend, and Sir Louis being a very influential one) jointly into account, and since it would then be impossible to consider a notional domain out of Miss Copley is an old friend, it would then be difficult to see why we would have to acknowledge the fact that Miss Copley is an old friend before asserting that Sir Louis is a very influential person.

5.2. May in subordinate clauses of concession beginning with an adjective Another type of concessive constructions is to be found when an adjective is placed before the conjunction: (45a) Genius though she may be, she is quite unassuming. Here again, the presence of may can easily be explained away if we consider that when the modal is used, we no longer remain on the quantitative level, which means that of actual events, but we are directly situated on the qualitative level, that is that of the notional properties of the subject, hence

May in concessive contexts 171

the necessary use of the simple present in the second part of the utterance. By contrast, in Quirk et al’s example: (45b) Genius though she was, she was quite unassuming. we refer to a specific situation in which the validations of two different predicative relations are taken into consideration and may is no longer required. Let us now consider (46a). (46a) Rich though he may be, he is not a spendthrift. The second part of the utterance introduces a property of the subject which does not conform to the notional property mentioned before. If we now change though into as, we have to alter the rest of the sentence altogether, as we have to mention the fact that what is logically expected is contrary to reality: (46b) Rich as he may be, he could be a spendthrift. If we now delete may, we then have to introduce another property of the subject which is in keeping with the first. (46c) Rich as he is, he can buy any yacht he likes.

5.3. May in subordinate clauses of concession beginning with a verb The same distinction can be established: may is not to be used when a specific situation is mentioned: (47) Fail though I did, I would not abandon my goal. (Quirk et al. 1985: 1098) But when it comes to taking into account the properties of the subject, may becomes necessary again: (48) He would like to have a taste of Angst, if only to know what it is like. But, try though he may, he cannot find anything: in his heart that he can recognize as Angst. (Coetzee, Youth)

172 Jean-Claude Souesme (49) With his lofty unconcern for mere living, Henry James exerts a strong pull on him. Yet, try though he may, he cannot feel the ghostly hand of James extended to touch his brow in blessing: James belongs to the past: by the time he himself was born, James had been dead for twenty years. James Joyce was still alive. (Coetzee, Youth) By way of a comparison, when the verb is followed by as instead of though, such as in: (50) Work as he may, he won’t be able to pass. we still remain on the qualitative level, and as E. Gilbert put it, whatever the qualitative value retained from the gradient11 on work, any validation of the predicative relation will always bar the way to the validation of the second predicative relation. In that context, may acts as a scanning operator12 of the various possibilities, always opening up a new path to a new value. In such contexts, might is possible when properties of the subject are at stake in a past specific situation; but the preterit has a temporal value, hence the use of could in the main clause: (51) Try as I might, I could not understand what he meant. (52) And try as they might, the critics could not paint this mild-mannered, bicycle-riding, leftist vegetarian as a corporate apologist. (Time, April 26, 2004)

5.4. May in subordinate clauses introduced by while May is also responsible for the concessive value of while when it comes to an opposition between notional properties: (53a) While Dutch police may overlook drug buys, the French do not. (Time, Dec. 28, 1992) (54) Yet while he may not always look or act the part, John Browne is an oilman’s oilman. He became CEO of British Petroleum (BP) in 1995… (Time, April 26, 2004)

May in concessive contexts 173

When may is used, it means an alteration is being introduced either concerning two different subjects in relation to the same predicate, as in the first example, or concerning two different predicates in relation to the same subject as in the second example, and while takes on its contrastive value. If we deleted may as in example (53a) (53b) While Dutch police overlook drug buys, the French do not.(Time, Dec. 28, 1992) we would remain on the quantitative level, that is, we would take into consideration two specific events; this means we would remain on the temporal plane and as a matter of fact, while would have its temporal value of simultaneity. We must now add that when might is used in clauses introduced by while, it necessarily follows that it has an epistemic value: (55) While it might be tolerable to wait two hours to board an aircraft, that’s just not practicable for any mass transit system. (Time, March 22, 2004) As has already been mentioned by many linguists (including Bouscaren, Larreya and Rivière, Lapaire and Rotgé, Leech, Quirk) the use of the preterit means that the value is from the start considered as fictive, but the utterer still takes it upon consideration in order to show that even if the predicative relation were validated, it would not bar the path to the validation of the second predicative relation. And the same explanation applies to the following examples: (56) “One might feel shame at such a home, but artists come here, and even under this ceiling we have wonderful evenings.” (National Geographic, Feb. 1991) (57) The experience might be more comfortable on a warm summer’s evening with light still in the sky. But you couldn’t beat it for atmosphere the way it was that winter night. (58) Self-absorbed as Jackson might be, he may not have noticed that the world’s opinion of him has diminished and soured. (Time, December 1, 2003)

174 Jean-Claude Souesme 6. Conclusion We must admit that we have come across a large number of these uses of may in recent issues of Time, and in one of Coetzee’s last novels, Youth, as if it were a new fashionable rhetorical device. It may be so, though we did not make any statistics; nor did we study this phenomenon from a diachronic point of view. Nonetheless, we may consider that it is an effective tool as part of an utterer’s enunciative strategy; may has a true linguistic function in concessive or adversative contexts: it allows the utterer to place himself on the qualitative level only. May then serves as a verbal marker opening the path to the validity of another predicative relation which is not impeded by the validity or validation of the previous predicative relation to which may has been applied ; when it is used in specific contexts, the second clause of the sentence is then heralded as introducing, or referring to, a property of the subject which is considered by the enunciator (and would normally be by any co-enunciator as well) as foreign to the notional conformity of what has been implied by the notional property mentioned before. As a general rule, we might say that in such contexts, may announces that one value, supposedly excluded according to the enunciator’s pre-construct because it is at variance with what the first predicative relation implies, will be introduced by but in the second part of the utterance.

Notes 1. Cf Larreya and Rivière (2005 : 89) and Bouscaren and Chuquet (1992 : 4142). 2. We all have a mental image of what a notion like BORROW for example stands for. It is made up of the properties common to all forms of BORROWING. Around this notion can be built a notional domain composed of an Interior (= what BORROW really means) and an Exterior (= what BORROW does NOT mean or imply). But there are also border line cases (= not really BORROW, but something akin to it somehow) which form what is called the Boundary or Frontier. For further explanation, see Culioli (1995). 3. Enunciative operations are those that take place on the enunciative plane, and as such they refer to the construction of the utterance, which is when the utterer locates the predicative relation relative to the situation of utterance itself. 4. A predicative relation is the result of an operation which consists in choosing a subject and a predicate; for instance in: Harry Potter come and live here! the utterer simply considers the idea of coming and living here (= the predicate) concerning Harry Potter (the subject). This is what is called the predicative

May in concessive contexts 175

5. 6. 7.

8. 9. 10. 11.

12.

level. At that stage, the utterer just refers to the idea of Harry Potter coming and living here; no link has been established yet between the subject and the predicate. The utterer does not validate the relation : a predicative relation is said to be validated only when the utterer considers that the subject and the predicate ‘go together’, that ‘it is the case’, as in : Harry Potter comes and lives here next week. As the example shows, note that it does not necessarily mean that the event itself has actually taken place. See note 2. See note 1. The word ‘event’ is a general term used in Culioli’s theory for the referent of a proposition ; therefore it can refer to any kind of situation, state of affairs, or actual event if any: it is totally independent from the nature of the process involved (either a state verb, an action verb, etc.). By this we mean that the utterer no longer considers the relative predication as being validated. See note 2. See note 2. A gradient can be applied to a notion like WORK for instance, when we refer to the qualitative level, that is when we consider the quality of the properties involved; in the present case, the gradient would go from produce very little work to work very hard. Typically, a scanning operation consists in considering every possibility in one given context without choosing one in particular.

References Bouscaren, Jeanine, Jean Chuquet, and Laurent Danon-Boileau 1987 Grammaire et textes anglais – Guide pour l’analyse linguistique, Ophrys, Gap. Bouscaren, Janine, Jean Chuquet. and Laurent Danon-Boileau 1992 Introduction to a Linguistic Grammar of English – An utterer-centered approach, translated and adapted by R. Flintham, and J. Bouscaren. Gap: Ophrys. Culioli, Antoine 1985 Notes du Séminaire de DEA 1983–1984, prises par Jean-Claude Souesme, Poitiers, Paris, Département Recherches Linguistiques de l’université de Paris 7. 1995 Even though, even if, as though, as if. In Cahiers Charles V 19 : 85– 91) (also available in Culioli. Pour une Linguistique de l’Enonciation, Tome III, Gap: Ophrys). 1995 Cognition and Representation in Linguistic Theory, Texts selected, edited and introduced by M. Liddle, translated with the assistance of

176 Jean-Claude Souesme John T. Stonham. Amsterdam and Philadelphia: Benjamins, [translation and adaptation of Culioli 1985]. Filippi, Catherine 1997 Though et la relation de concession. In La Notion, Gap: Ophrys, 91– 98. Gauthier, André 1997 Even, just et le domaine notionnel. In La Notion, Gap: Ophrys.121– 128. Ghio, Gilbert 1996 Can, May, les modaux du possible en anglais contemporain. Thèse de doctorat, Université de Nice-Sophia Antipolis. Gilbert, Eric 1987 May, Must, Can et les Opérations Enonciatives. Cahiers de Recherche, Tome III, Gap, Ophrys. Groussier Marie-Line 2000 On Antoine Culioli's theory of enunciative operations. Lingua 110: 157–182. Groussier, Marie-Line, and Claude Rivière 1996 Les mots de la Linguistique. Gap: Ophrys. Lapaire, Jean-Rémi, and Wilfrid Rotgé 1992 Linguistique et Grammaire de l’Anglais. Toulouse : Presses Universitaires du Mirail. Larreya, Paul, and Claude Rivière 2005 Grammaire Explicative de l’anglais. 3rd edition. London: Longman. Leech, Geoffrey 1987 Meaning and the English Verb. London: Longman. Quirk Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Ranger, Graham 1998 Les constructions concessives en anglais. Cahiers de Recherche, n° spécial, Gap: Ophrys. Souesme, Jean-Claude 1995 But, marqueur de passage de frontière et ses traductions en français . In Relations discursives et traduction, M. Ballard (ed), Lille : Presses Universitaires de Lille: 157–186. 2003 Grammaire Anglaise en Contexte. 2nd edition. Gap: Ophrys. 2008 100 Fiches de Grammaire Anglaise. 2nd edition. Rosny, Bréal.

When may means must: deontic modality in English statute construction Ross Charnock

1. Introduction As well as rules which forbid certain actions or impose certain duties, all modern legal systems include power-conferring rules, which grant rights without imposing norms of behaviour. The distinction between obligations and rights is commonly expressed in statutory language, as in ordinary speech, through the use of modal verbs. In English statutes, shall is normally used to impose an obligation, while may is generally used to confer a discretionary or enabling power. This use is illustrated uncontroversially, for example in “A local authority may within their district [...] construct a public sewer [...].” (Public Health Act 1936 s15[1]) [my emphasis]1 The modal verb is here seen as conferring a right, rather than as imposing an obligation. The question occasionally arises as to whether in particular circumstances the holder of the power, in this case the local authority, could be legally obliged to act under the relevant statute, and if so whether and in what sense enabling expressions like may could then be said to impose a duty. Various courts have held that in certain circumstances there may indeed be an obligation to exercise a statutory discretion. In such cases, the statutory may is said to have a coercive meaning similar to must. Surprisingly, this is the first sense given under ‘may’ in Stroud’s Judicial Dictionary (Stroud: 1986). This reference book gives as synonyms various other legal expressions normally used to grant a discretionary power, including ‘it shall be lawful’, and ‘shall hereby have power’. Where shall occurs in such contexts, it does not have the effect of creating a new legal obligation or duty, but functions as part of an enabling expression to create a new legal power. Linguists commonly accept that expressions literally signifying permission can be used to imply an obligation. However, they would be reluctant to accept the more radical claim implicit in the relevant judgments, that, under certain conditions, may can be understood, in its literal meaning, as including not just a permissive but also a coercive sense. This conclusion is

178 Ross Charnock reached through different forms of legal analysis, illustrated here in Julius 1880, Re Shuter (2) 1959 and Anns 1977.

2. Judicial analysis 2.1. Julius v Oxford (Bishop)2 The leading English case for the proposition that the literal meaning of the legal may includes not just the permissive, but also the coercive sense is Julius v Oxford (Bishop) 1880, in which numerous precedents are cited. The authority is seen to derive from Alderman Backwell’s case, decided as long ago as 1683, by Lord Keeper North in the Court of Chancery. Backwell’s case depended on the interpretation of the Bankrupts Act 1571, which enacted (ch7, s2) that where complaints were made in writing against “a person being bankrupt”, the Lord Keeper “shall have full power and authority” to grant a “commission of bankruptcy”. When the Exchequer was closed in 1676, Alderman Backwell’s creditors were obliged to petition for a declaration of bankruptcy. As the expression used in the statute was only empowering, and did not impose an obligatory duty, the Lord Keeper was able to delay matters for seven years (apparently with the complicity of the government), before finally declaring that: [T]hough the words in the Act of Parliament were, that the Chancellor ‘may grant a Commission of Bankrupt’ [sic], yet that ‘may’ was in effect ‘must’, and it had been so resolved by all the judges. And the granting of a commission was not a matter of discretion; but that he was bound to do it. (Alderman Backwell’s case 1683, per North LK, as cited by Lord Blackburn in Julius 1880)3

Later courts were quick to make use of the new, wide interpretation, in order to reach a just result in appropriate cases. A rule of construction4 was established in Reg v Tithe Commissioners 1852. This case concerned the interpretation of the Tithe Act 1842, which empowered commissioners to confirm agreements for commutation of tithe in particular circumstances. It was held that, although the statute purported to confer a power, the Commissioners were bound, where the conditions were met, to accept the relevant agreements. It was stated as a rule of law that:

When may means must

179

The words undoubtedly are only empowering but it has been so often decided as to have become an axiom, that in public statutes words only directory, permissory or enabling may have a compulsory force where the thing to be done is for the public benefit or the advancement of public justice (Reg. v Tithe Commissioners 1849, per Coleridge J).

The coercive interpretation is here considered as axiomatic. This rule of construction was reconsidered in Julius v Oxford (Bishop) 1880. Here, Dr Frederic Julius accused his local rector of “unauthorised deviations from the ritual of the Church, and the use of unauthorised vestments”. The bishop declined to act on this complaint as, in his view, legal proceedings would tend to “cover all persons concerned in them with ridicule, and [to] bring the Church itself into some contempt”. Dr Julius, however, considered that under the Church Discipline Act 1879, the Bishop was under a duty to issue a commission. The relevant section of the Act reads: In every case of any clerk in holy orders in the United Church of England and Ireland who may be charged with any offence against the laws Ecclesiastical [...], it shall be lawful for the Bishop of the Diocese within which the offence is alleged or reported to have been committed, on the application of the party complaining thereof, or if he shall think fit of his own pure motion, to issue a commission to five persons [...] for the purpose of making inquiry as to the grounds of such charge or report. (Church Discipline Act 1879 s3) [my emphasis]

The expression ‘it shall be lawful’ appears to grant a right, rather than to impose an obligation. However, as this was a public statute, and the “thing to be done” was for the public benefit, the Court of Queen’s Bench followed the rule given by Coleridge J in Reg. v Tithe Commissioners, the judges agreeing unanimously that the expression should be understood in this context as imposing an obligation. Coleridge J’s rule was presented as an “established canon of construction”, according to which such words as ‘it shall be lawful’, in statutes of the relevant class, had a “settled meaning”, imposing on the holder of the discretion a duty to exercise it. However, when the case reached the House of Lords, this decision was reversed, again unanimously. In a series of concurring judgments, their Lordships took the opportunity to clarify the law. They agreed, first, that, although may did not itself occur in the Statute, the expression ‘it shall be lawful’ should indeed be taken as equivalent: “it is not inaccurate to say that the words conferring a power are equivalent to ‘may’” (Julius v Oxford (Bishop) 1880, per Lord Blackburn.)

180 Ross Charnock However, they also considered that the rule applied by the Queen’s Bench was wrongly formulated. In the opinion of the Lord Chancellor, the relevant precedents showed only that: [...] where a power is deposited with a public officer for the purpose of being used for the benefit of persons who are specifically pointed out, and with regard to whom a definition is supplied by the legislature of the conditions upon which they are entitled to call for its exercise, that power ought to be exercised, and the court will require it to be exercised” (Julius v Oxford (Bishop) 1880, per Earl Cairns LC).

Lord Blackburn pointed out that the precedents all concerned applications made by those whose private rights required the exercise of a legal power: “The enabling words are construed as compulsory whenever the object of the power is to effect a legal right”. This new formulation of the rule seems more restrictive than the earlier version; the coercive interpretation is no longer considered as “axiomatic”, but is instead allowed only when the power granted is for the benefit of a specific person, or class of people expressly designated (“persons specifically pointed out”), and who have the right to demand the exercise of the discretion (“who were entitled to call for it”). In the result, it was held that, in this particular case, the Bishop was under no legal obligation to proceed with the enquiry. However, the judgments confirm that, in appropriate contexts, language purporting to confer a discretion can be read as imposing a duty. In such cases may will be given a coercive interpretation, equivalent to must: Where there is such a duty, it is not inaccurate to say that the words conferring the power are equivalent to saying that the donee must exercise it. (Julius v Oxford (Bishop) 1880, per Lord Blackburn.)

A similar conclusion is reached through a different form of legal reasoning in Re Shuter (2) 1959. The reasoning in this case was based on the established rules of statute construction, which involve (defeasible) presumptions.

When may means must

181

2.2. Re Shuter (2) Re Shuter concerned the interpretation of the Fugitive Offenders Act 1881, dealing with the extradition of prisoners to British colonies. Section 7 of this act provided that: If a fugitive committed to prison under the Act has not been taken out of the Jurisdiction within one month after committal, a superior court, on application by the fugitive, after due notice, may, unless sufficient cause is shown to the contrary, order the fugitive to be discharged out of custody. (Fugitive Offenders Act 1881 s7) [my emphasis]

Shuter had been apprehended and detained in England. He was wanted in Kenya, where he was alleged to have committed various offences. As one of the police officers involved was on leave until 16 August 1959, the magistrate arranged for an escort back to Kenya on that day, knowing that the maximum period of one month would be exceeded (by just one day). As Shuter had not been taken out of the jurisdiction within one month, he petitioned to be discharged. The question to be decided was whether may, in s7 of the Act, should be construed as permissive or obligatory. The Chief Justice examined the different possible readings. He first considered the assumption that the word was intended as permissive, and that it simply gave the judge the right to make the order. If that discretionary reading was adopted, there would then be two possible interpretations of the text. First, the words ‘unless sufficient cause is shown to the contrary’ may also be understood as discretionary, that is as allowing the judge a supplementary power. However, the words would then be redundant. In the judge’s words: “if the discretion is completely at large and ‘may’ means ‘may’, then it would be quite unnecessary to have the words “unless sufficient cause is shown to the contrary”. (Re Shuter (2) 1959, per Parker CJ.) Alternatively, the words ‘unless sufficient cause [...]’ may be taken as limiting the judge’s discretion so that, if there is sufficient cause, then he no longer has the right to discharge the prisoner: “[A]nything may be taken into consideration, save only that, if sufficient cause is shown to the contrary, then the discretion is cut down to that extent.” However, this leads to a contradiction: “It is a case where what is sufficient cause in any particular case may well depend on opinion and discretion.” As neither of these alternative interpretations was acceptable, Parker CJ concluded that, in this section of the Act, the word may could not be interpreted as discretionary.

182 Ross Charnock Having rejected the discretionary reading, Parker CJ went on to consider the coercive interpretation, noting that if may is interpreted as imposing an obligation, then the judicial discretion will be restricted to the interpretation of the expression ‘sufficient cause’. On that interpretation, the judge would be under a duty to order the release of the prisoner (unless sufficient cause was indeed shown to the contrary). In the opinion of the Chief Justice, this was the correct reading: It seems to me that in ordinary language, s7 is saying that there is no jurisdiction for holding a man beyond one month unless sufficient cause is shown [...]. The natural meaning is ‘shall, unless sufficient cause is shown to the contrary, order the fugitive to be discharged out of custody’” (Re Shuter (2) 1959, per Parker CJ).

In the result, however, Parker CJ chose to exercise his discretion in refusing to order the discharge, holding that in this case there was indeed “sufficient cause to the contrary”. The coercive interpretation of may, in s7 of the Fugitive Offenders Act 1881, should be compared with the later, permissive, interpretation of the same word in s6 of the same statute, in R v Brixton Prison (Governor) ex parte Enohoro 1963. Section 6 reads as follows: Upon the expiration of fifteen days after a fugitive has been committed to prison to await his return, or if a writ of habeas corpus or other like process is issued with reference to such fugitive by a superior court, after the final decision of the court in that case, 1) [...] a Secretary of State [...] may, if he thinks it just, by warrant under his hand order that fugitive to be returned to the part of Her Majesty’s dominion from which he is a fugitive. (Fugitive Offenders Act 1881, s6) [my emphasis]

Chief Enohoro was wanted in Nigeria. Although the Home Secretary thought it just to order his return to that country, he delayed ordering the extradition in order to give the House of Commons an opportunity to express a view. It was submitted that the Home Secretary had thus (illegally) surrendered his statutory discretion to a third party. This argument depended on the presumption according to which words used within a single statute should be interpreted as far as possible in the same way on each occurrence. It had been established in 1959 (in Re Shuter, above), that may was to be understood in s7 as coercive. The word should therefore be interpreted in the same way in s6. Following this interpretation, the Home Secretary should be considered as obliged to order the extradition, once he had

When may means must

183

decided it was just to do so. In not doing so immediately, he had failed to comply with the Act. The judge should therefore order the release of the prisoner. Enohoro was heard by Parker CJ, the same judge as in Re Shuter (2). He accepted that the word should normally be interpreted in the same way in the different parts of the statute. However, he held that the presumption did not apply in this particular case, as two different uses could be distinguished. In s7, the legislative intention was to preserve the liberty of the subject by preventing detention for an unduly long period; whereas in s6, the intention was to safeguard the rights of the prisoner by preventing an order from being given for his return until fifteen days had passed. This safeguard would be ineffective if the order was considered mandatory at the end of this period of reflection. The judge was therefore entitled to hold that the word may did not have the same meaning in s6 as in s7 of the statute: For my part, I am quite satisfied that that argument ultimately depends on whether it is right to treat the word ‘may’ in s6 as mandatory [...] I am quite satisfied that in s6 it cannot be read in that sense. Section 7, as I have already said, is dealing with the liberty of the subject, and the prevention of his being detained for an unduly long period. After thirty days he must be discharged unless sufficient cause is shown. Section 6 on the other hand, so far as time is concerned, is not saying that something must happen after a certain time, but is saying that something shall not happen before a certain time has expired [...] and, for my part, I can see no reason for giving ‘may’ a mandatory reading in that connexion. R v Brixton Prison (Governor) ex parte Enohoro 1963 per Parker CJ.)

Parker CJ therefore held that the Home Secretary was not obliged to order the extradition immediately, and that the fact that he delayed giving the order could not therefore justify the release of the prisoner. As has been seen, different reasoning was adopted in Julius and in Re Shuter to justify the coercive reading of discretionary language. A third approach, derived from common law rules, is taken in Anns v Merton LBC 1977.5

2.3. Anns v Merton LBC Following the precedent established in the leading case of Donoghue v Stevenson 1934, the English courts are entitled, under the common law, to find

184 Ross Charnock a “duty of care” even in the absence of statutory or contractual obligations, and to declare liability in negligence where damage is caused by a failure to respect that duty. The courts have found a duty of care in an ever wider range of situations. The extent to which this common law duty can be found in addition to, or in parallel with, statutory powers was discussed extensively in Anns v Merton London Borough Council 1977. The facts were as follows. According to the Public Health Act 1936, a local council had the power to inspect the foundations of building projects before allowing completion. Merton LBC did not carry out a proper inspection of a project involving the construction of a number of maisonettes, and failed to notice that the foundations were too shallow. The lease-holders suffered substantial loss when their maisonettes started to subside, and demanded compensation. When the case reached the Lords, the Council’s defence was that, under the terms of the Act, there was no obligation to carry out an inspection at all. The statute stated in s61(1): “Every local authority may, and if required by the Minister, shall make bylaws for regulating [...] the construction of buildings”, and in s61(2): “bylaws made under the section may include provisions as to the giving of notices, the deposit of plans and the inspection of work”. [my emphasis] The House of Lords had to decide whether, in spite of the explicit distinction between may and shall in s61(1), the defendant council was under a duty of care to the plaintiffs. The principal judgment was given by Lord Wilberforce, who held that, as the damage was foreseeable, there was a prima facie duty of care. He pointed out that the public authorities “must, and in fact do, make their discretionary decisions responsibly and for reasons which accord with the statutory purpose”. They were therefore “under a duty to give proper consideration to the question whether they should inspect or not”. Lord Wilberforce seems to have considered that the greater the statutory discretion allowed, the greater the common law responsibility: “It can safely be said that the more ‘operational’ a power or duty may be, the easier it is to superimpose upon it a common law duty of care”. He therefore concluded that, although the statutory language conferred a power, and not a duty, the local authority was nevertheless liable under the common law rules: “If they do not exercise their discretion in this way they can be challenged in the courts.” The Anns decision has since been overruled, and can no longer be seen as an authoritative statement of the English law. In Murphy v Brentwood

When may means must

185

1990, Lord Bridge complained that the Anns judgment seemed to impose stricter obligations on the local council than on the negligent builders: He [Lord Wilberforce] cannot, I think, have meant that the statutory obligation to build in conformity with the bylaws by itself gives rise to obligations in the nature of transmissible warranties of quality. If he did mean that, I must respectfully disagree. I find it impossible to suppose that anything less than clear express language such as is used in s1 of the 1972 Act would suffice to impose such a statutory obligation.” (Murphy v Brentwood 1990 HL, per Lord Bridge)

However, the interpretation of the statutory may as compatible with an obligation derived from the common law duty of care, as suggested in Anns, does not appear to have been definitively rejected.

3. Linguistic issues In comparison with ordinary usage, legal adjudication may seem specialised and artificial. The legal approach to the interpretation of statutory terms may, therefore, appear at first sight to be irrelevant to the objectives of linguistics. However, the law remains an important repository of significant examples of authentic language use, often involving explicit references to the interpretative process. These should not be ignored. Further, although both the law and the rules for its interpretation have developed empirically, as cases have arisen, the resulting analysis is both coherent and systematic. As it is not presented as a general semantic theory, the legal account entails no claim regarding the correct approach to be adopted in other linguistic fields. However, precisely because it does not correspond to traditional semantic analysis, legal interpretation often raises questions of interest to linguists. The legal approach to deontic modality in particular raises issues relevant both to the problem of performativity and to the ongoing debate between the “literalists” and the “contextualists”.

3.1. Performativity When a judge interprets may as imposing an obligation he does not (usually) see himself as simply substituting one word for another in an attempt to correct what he considers to be a mistake on the part of the legislature.

186 Ross Charnock On the contrary, he claims to be applying the law in its literal interpretation, as intended by Parliament. However, if the legislative draftsman really intended to impose an obligation, then his apparent preference for may, in place of shall, requires explanation. One possible explanation depends on a specialised, performative reading of the legal shall. Legislation may be considered as essentially performative in nature, to be analysed, not as a simple (constative) statement of the law, but rather as a set of speech acts (commands) which have the effect of creating new norms and rules of behaviour.6 Clearly, in English statutes, the modal verb shall, used felicitously to create a new duty or obligation, does not simply indicate what will happen. Nor (in spite of the presumption against retrospective application) should it be understood as a prediction of what the law will be in the future. It is better analysed in this context, if not as an explicit performative verb, at least as a kind of performative marker.7 This use of shall is confirmed in numerous examples, often involving explicit commands or instructions to the judge, rather than to the general public. Many statutes use the word to ensure that words are interpreted in a certain way, as in the Theft Act 1968, s1(1): “[...] ‘thief’ and ‘steal’ shall be construed accordingly”. The judge may be directed to find particular articles obscene, as in the Obscene Publications Act 1959, s1: “an article shall be deemed to be obscene if [...]”. Or he may be obliged to reach a particular verdict, as in the Offences Against the Person Act 1861, s16): “A person who without lawful excuse makes to another a threat [...] shall be guilty of an offence [...].” [my emphasis] The performative function explains the occurrence of shall in negative contexts, where it is often used in preference to ‘may not’, normally considered to be logically equivalent and more natural. A particularly celebrated example is found in the First Amendment to the United States Constitution: “Congress shall make no law [...]” [my emphasis]. In such contexts, shall does not simply denote the absence of a right but is used performatively to impose a legal interdiction. The performative view may also explain the common use of shall as part of permissive expressions like ‘it shall be lawful’; in such cases the modal verb may be taken to mark a (performative) change in the law.8 According to the linguistic theory of speech acts (Austin:1962), performative utterances may be inappropriate or infelicitous, but they cannot be considered false. Indeed, explicit performatives are said to be self-verifying, in the sense that, once the verb ‘to promise’, for example, has been used performatively, it cannot be denied that a promise has been made

When may means must

187

(even though it may have been insincere). In the same way, once a positive duty has been created through the use of the statutory shall, that duty must inevitably arise in the circumstances defined in the Act.9 The quality of inevitability associated with the legal shall helps to explain why the legislator often prefers to use a permissive expression when the intention is to create a legal duty in circumstances which cannot be exhaustively defined, or which may not occur. It may be observed that shall does not usually occur in contexts where it is necessary to specify exceptions to a rule.10 In Anns, Lord Wilberforce explained why what he considered to be an obligation was expressed, in the Public Health Act 1936, “in terms of functions and powers rather than in terms of positive duty”. In his view, because it was necessary to “strike a balance between the claims of efficiency and thrift”, it was neither feasible nor desirable to limit the circumstances in which there would be a duty to exercise the statutory discretion. The legislator therefore preferred may to shall. Similarly, in one of the examples given by Lord Blackburn in Julius, a judge “must” pass sentence in case of a guilty verdict. Yet, as the accused is presumed to be innocent until proven guilty, such a verdict cannot be considered as inevitable. The legal obligation is therefore presented merely as a judicial power. The performative account of the legal shall thus accounts for the fact that even where the intention is to impose an obligation or duty, the legislator may still prefer to use may. It also explains why it seems impossible to interpret shall as may, as would be predicted if the legal concept of obligation was simply assumed to be vague enough to tolerate the substitution of one word for another. A second linguistic issue raised by the legal analysis of may concerns the nature of literal meaning.

3.2. Literal interpretation and contextualism It is common in ordinary language for an obligatory meaning to be ascribed to permissive expressions. Such uses are normally explained through pragmatic inference. Everyday examples often involve the modal verb can, as in ‘You can just stop doing that’, intended as an order or an instruction. Similarly, what is presented literally as a choice may be understood as an obligation if it is clear that it is the only possibility. This reading is illustrated in ‘Tell her she can take a taxi’, in the context of a mother-in-law’s telephoned request for a lift from the station at an inconvenient time. The con-

188 Ross Charnock textual understanding may also be based on non linguistic inference, based for example on ethics, as is implicit in the French saying noblesse oblige. According to this principle, if you have the means to help others, then there is a moral obligation to do so.11 It is tempting to suppose that the various legal interpretations of the statutory may can be accounted for in a similar way. In Anns 1977, the power conferred by statute was said to imply a responsibility, and thus, by inference, to impose a common law duty. In Re Shuter (2) 1959, the permissive reading was seen to lead either to redundancy or to contradiction; it was therefore more coherent to prefer the coercive interpretation. On first impression, such reasoning may be thought to correspond to a simple application of Gricean conversational maxims. However, this pragmatic approach does not correspond to the legal view. On the contrary, English judges are expected to adhere as far as possible to “the grammatical and ordinary sense of the words” (per Lord Wensleydale in Grey v Pearson 1857). In exceptional cases, they may invoke a “Golden Rule”, which allows courts to avoid inconsistent, absurd, or (quaintly) inconvenient results, by interpreting words, not with their “ordinary signification”, but by “putting on them some other signification, which, though less proper, is one which the court thinks the words will bear” (per Lord Blackburn in River Wear Commissioners v Anderson 1877). There are several reasons for this “literal” approach to judicial interpretation. First, the (speaker’s) intentional meaning is necessarily subjective and thus in principle unverifiable. Second, pragmatic interpretation is context dependent, the same expression being understood in different ways on different occasions. Such a flexible approach to statute construction would be judicially unacceptable. Indeed, if the content of the rule were allowed to vary according to particular circumstances, the judges would find themselves adjudicating on a case-by-case basis. They could then no longer be said to be following a rule at all. For these reasons, where no alternative meaning is available, the “literal meaning” is given precedence, even where there is a clear conflict between what was said and what appears to have been meant. This is the essence of what is known as the “literal rule”. According to this rule, although the judges are supposed, in principle, to be giving effect to the intention of the legislator, their interpretation must depend, in practice, on the words used, taken in their “ordinary and natural sense”, rather than on any more pragmatic understanding. As was stated by Tindall CJ in the Sussex Peerage case 1844: “The words alone do, in such a case, best declare the intention

When may means must

189

of the lawgiver.” English judges are therefore reluctant to take the meaning of may, and other judicial power-conferring expressions, to be contextually implied. In Re Shuter (2) 1959, for example, Parker CJ was careful to say, not that Parliament meant to impose an obligation (notwithstanding the use of may in the statute), but rather that the natural and ordinary meaning of the text under analysis corresponded to shall. The approach taken in Julius 1880 is particularly intriguing. The claim here was that, that according to precedent, words conferring a power could be understood as saying that the power must be exercised. The enabling words in the statute could therefore be construed as compulsory. In that case, the judges seem to have presented the coercive sense of may as part of its literal meaning. It may be possible to accomplish this simply by adding this meaning to the restricted number of fundamental senses available in appropriate contexts. Such a solution need not be considered theoretically impossible, as there is in any case no general agreement among linguists on how many basic senses should be distinguished. However, not only does such a polysemic approach appear methodologically unsound, it is incompatible with the judges’ refusal to consider the term as ambiguous at all. In his judgment in Julius, the Lord Chancellor declared, contrary to the suggestion made in the lower courts, that power-conferring expressions equivalent to may should not be considered ambiguous: “The words ‘it shall be lawful’ are not equivocal. They are plain and unambiguous” (Julius v Oxford (Bishop) 1880, per Earl Cairns LC). Similarly, for Lord Blackburn: “I do not think that the words ‘it shall be lawful’ are in themselves ambiguous at all.” In order to follow the literal rule, and to avoid interpreting words differently in different cases, English judges seem to assume that the “ordinary” or “natural” sense of statutory terms can be clearly established. They speak in this context of the “true and correct” meaning. However, as all meaning depends as least to some extent on contextual features, the idea of a definitive, fixed meaning appears impossible in principle. Even on the most basic, lexical level, it is commonly observed that words have different senses according to context of use. The word ‘operation’, for example, is understood differently in mathematical, military or surgical contexts. Although these meanings may be derived by analogy from a more general sense, they each appear to be treated as primary within the relevant speech community. It is therefore clear that the so-called “literal rule” cannot itself be taken literally. In this context, the judges speak (oxymoronically) of the “literal

190 Ross Charnock meaning in context” and are traditionally willing to include the context of “the whole of the statute”. In order to preserve the appearance of objective adjudication, they have adopted a wide definition of the concept of literal meaning, and applied the rule in different ways in different fields of law. In statute construction, since Pepper v Hart 1993, it has been acceptable to go beyond the text itself, in order to take account of external evidence in the form of parliamentary debates where necessary to ascertain the legislative intention. In contract law, and in the interpretation of wills, there is ongoing debate concerning the extent to which the “surrounding circumstances” may be admissible as an aid to interpretation where the sense appears ambiguous.12 Bennion (1993: 2), a distinguished legislative draftsman, supposes that English judges will normally follow the literal rule; yet he refuses to consider this incompatible with contextual understanding. He prefers what he calls the “informed interpretation rule”, according to which “the interpreter is to infer that the legislator, when settling the wording of an enactment, intended it to be given a fully informed, rather than a purely literal, interpretation (though the two usually produce the same result)”. In the most recent judicial pronouncement on the subject at the time of writing, Lord Hoffmann, in Kirin-Amgen v Hoechst Marion Roussel 2005, went so far as to express doubt as to whether the literal rule was ever strictly observed, allowing only that such principles “used to be applied in legal interpretation (at any rate in theory)”. Kirin, an intellectual property case, was concerned with the wording of a patent. A purely literal interpretation, remorselessly applied, would have unintended consequences, leading to a lack of protection for the invention. Lord Hoffmann relied on a European protocol to justify a more purposive construction, holding that the court should be entitled to take account, not just of the literal meaning, but also of what would be understood in the particular context by someone skilled in the art. This implied that meaning should be considered as being: [...] highly sensitive to the context of and background to the particular utterance. It depends not only upon the words the author has chosen but also upon the identity of the audience he is taken to have been addressing and the knowledge and assumptions which one attributes to that audience (KirinAmgen Inc v Hoechst Marion Roussel HL 2004, per Lord Hoffmann)

Lord Hoffmann here rejects the traditional literal approach, now considered as a historical, legal fiction. However, he is careful not to base his interpretation on an unverifiable, intentional meaning. Instead, the semantics is

When may means must

191

relativised to a specific speech community, in this case a community of scientists. Clearly, this approach cannot be applied directly in the field of statute construction, as statutes are in principle addressed to the entire population (within the jurisdiction); it does, however, imply that legal understanding should take account of shared knowledge in the relevant circumstances.13 This approach to legal interpretation appears to correspond in practice, not to a purely literal meaning, but rather to the contextualist definition of “what is said”. On the contextualist account, as proposed by philosophers of language, there is no pre-existing, literal meaning available as a basis for consensual understanding. Instead, the semantic meaning depends on the relevant background features. The semantic and pragmatic levels of meaning are no longer clearly separable. This approach to semantics was not developed as a theory of modality in particular, but applies rather to sentence meaning in general. It derives from the work of for example Waismann (1951) on open texture, Putnam (1975) on stereotypical meaning, or Searle (1979) on literal meaning, all of whom deny the existence of a fixed, literal meaning, independent of background assumptions. In his extended discussion of the contextualist debate, Recanati (2004: 146 et seg) speaks, not of a specific, literal meaning, but rather of “semantic potential” within the theory of what he calls “Meaning Eliminativism”. Although, by convention, English judges still claim be basing their interpretation on “literal meaning”, it seems that they have always taken account of at least some elements of the context when determining “what is said”. Indeed, certain common yet strangely obscure legal turns of phrase, like “the meaning is wide enough to include [...]”, or “[...] is within the terms of the statute”, seem to assume a wide, contextualist view. The claim made in Julius 1880, for example, is precisely that the meaning of the deontic may is broad enough to include the coercive sense. In a striking parallel with the term used by Recanati (2004), Lord Selbourne, one of the judges in Julius, described the power-conferring expressions used in the Church Discipline Act 1879, as “potential”, declaring them to be the same “whether there is or is not a duty or obligation to use the powers which they confer.”

192 Ross Charnock 4. Conclusion: occasion-sensitivity and open texture The apparent rejection of literal or acontextual meaning as a basis for semantic understanding may seem at first sight implausible. This philosophical view of meaning may be expected to raise formidable theoretical difficulties. However, it also has the advantage of avoiding equally serious problems associated with the more conventional literalist view. It is well known, for example, that attempts to derive conversational meanings from a basic literal sense, through the machinery of Gricean implicature, often appear counter-intuitive in cases where speakers fail to notice linguistic ambiguities and are apparently unaware of literal incoherence (Travis 1991).14 A particularly strong version of the contextualist theory is found in Travis (1989), who shows that the semantics of ordinary expressions must be established not just by reference to the surrounding context, but also to shared knowledge of the underlying purpose of the discourse. He thus introduces the notion of occasion-sensitivity into the consensual understanding of “what is said”, going so far as to treat basic truth-values as dependent on the discursive context. To take a single, particularly mundane, example, the sentence ‘There is milk in the fridge’ would be true where there is an unopened carton of milk in the fridge, and the interlocutor intends to prepare cornflakes for breakfast; but false if there is just a puddle of milk on the bottom shelf. On the other hand, where a wife is reproaching her husband with not having cleaned the fridge properly, the opposite will be the case, as in this context a carton of milk will no longer count as milk. Neither interpretation involves any reference to a putative “literal meaning”, presumably involving a fridge full of nothing but “wall-to-wall” milk. Yet the occasion-sensitive meaning is still considered to be part of the semantics of the utterance. Although this consensual (non literal) understanding of “what is said” depends to some extent on the context, it is still to be distinguished from more pragmatic, implicit meanings, which depend on specific intentions, and which might here involve invitations to prepare breakfast, instructions to buy more milk, or orders to clean the fridge properly. The fact that similar examples can be multiplied indefinitely means that it is necessary to reject the simplistic assumption according to which meaning is either literal or pragmatic. The contextualist view does not imply a rejection of semantics in favour of pragmatic inference. Instead, the two levels are seen as interdependent.

When may means must

193

On this view of semantics, there is no requirement to explain any particular use by reference to a supposedly more basic sense. Indeed, no sense is considered more basic than another, since, in appropriate contexts, the relevant meanings should all appear equally natural. This consensual approach corresponds well to the legal analysis of power-conferring expressions. Neither in Julius, nor in Re Shuter (2) is there any attempt to derive the coercive meaning of may from a more basic, permissive sense. The reasoning in Anns was based on non linguistic inference; based on the common law. The contextualist approach has the further advantage of being compatible with Hart’s (1961) less radical theory of open texture, widely accepted by legal philosophers as an account of the semantics of legal rules. According to the theory of open texture, all substantive terms are subject to a “penumbra of uncertainty”,15 illustrated through the different possible interpretations of the invented rule, “No vehicles in the park”. Hart (1961: 126) gives the example of a motor car as an uncontroversial example of a “vehicle”, and “bicycles, airplanes, roller skates” as less clear cases subsisting in the penumbra of the general term. The question as to whether the penumbral meanings should be considered as literal or pragmatic is not addressed. Hart’s account of legal interpretation recalls the linguistic theory of prototypes (Rosch 1983). Applied to the semantics of the statutory may, this approach would involve taking the core meaning of the modal verb as permissive and the coercive sense as peripheral, that is as part of the penumbra of uncertainty. Again, the coercive meaning need not be derived from a core discretionary sense, any more than a peripheral penguin would have to be derived from a prototypical robin before it could be said to count as a bird. Although Hart assumes that judges will normally apply the law unproblematically, following the literal rule, he admits that they will inevitably be required to exercise their discretion in “hard cases”, where the legislative meaning is less clear. He is, however, reluctant to accept the level of indeterminacy implicit in the overtly pragmatic approach of the American “legal realists”. He considers such arguments for rule-scepticism as “incoherent” (1961: 136), and points out (1961: 126) that “[t]here will indeed be plain cases, constantly recurring in similar contexts, to which general expressions are clearly applicable”. On the other hand, he is forced to admit (1961: 126) that these plain cases, “where the general terms seem to need

194 Ross Charnock no interpretation”, are simply the most familiar ones, “where there is general agreement in judgments as to the application of the classifying terms”. For a contextualist linguist, the fact that a context is familiar does not mean that it no longer has any semantic role to play in consensual understanding. On the contrary, even Hart’s “plain cases” are seen, from this point of view, to depend on general, background assumptions regarding the meaning of “what is said”. Hart’s theoretical account of the semantics of legal rules thus appears to correspond to the judicial account of interpretative reasoning as presented in individual judgments. In spite of persistent claims on the part of judges to be following a literal rule of interpretation, it is apparent that the judicial concept of meaning is in fact fundamentally contextualist. The availability of a coercive interpretation of may (and other permissive expressions), as observed in the legal context, further suggests a possible extension of the contextualist theory into the field of modality. English judicial practice may thus be considered as providing empirical support for the contextualist view.

Notes 1.

English statutes are divided into sections, here section 15. By convention, subsections are given in brackets, here subsection (1). Where necessary, further subdivisions are given in separate parentheses. They are published by Her Majesty’s Stationery Office (HMSO): London, and collected in Halsbury’s Statutes, or in Halsbury’s Laws of England (4th edition 1973), London: Butterworth, 52 volumes, with cumulative supplements. 2. Cases are identified using the names of the parties and the year of publication (between square brackets). This usually, though not invariably, corresponds to the date of judgment. References also include the Appeal case number (AC) or details of the published report. A convenient source is the All England Reports (Butterworths: London) (All ER). The volume number is given where necessary to the left of the name of the publication, with the page number to the right. Early cases (1220 to 1865) may be found in the English Reports (Edinburgh: Green, 1932, 180 volumes). These references give the date of judgment (between round brackets) and the (abbreviated) name of the editor of the original publication. Cases from 1997 onwards are available on <www.bailli.org>. 3. Even older precedents are cited in the notes to Backwell as authority for the proposition that “[w]henever a statute directs the doing of a thing for the sake

When may means must

195

of justice or the public good, the word may is the same as shall”. [my emphasis] 4. The legal term for ‘interpretation’ is derived from the verb ‘to construe’. 5. English law is made up of a combination of common law rules, decided by the judges, and statutes passed by Parliament in the form of legislation. Although Parliamentary legislation takes precedence, the common law rules remain valid unless and until they are either overruled by the judges themselves, or overtaken (that is expressly contradicted) by statute. 6. This positivist analysis of the law in terms of general commands was proposed by J. Austin (1832) 7. The performative analysis of the legislative shall is discussed in Garzone (2001). As this performative function does not normally appear in ordinary language, and other meanings are possible even in the legal context, it has been suggested that, in the interest of clarity, the word should no longer be used in statutes, and that it should be replaced for example by must. On this question, see Williams (this volume). 8. There is a sense in which may could also be said to have a performative function, insofar as it is also used to create new legal rules. But, precisely because no command or obligation is involved, the nature of the speech act is less clear. On the relation between performativity and normativity in judicial discourse, see Charnock (forthcoming). 9. It should, however, be noted that in the section on “Can saying make it so?”, Austin (1962: 7-11), appears reluctant to consider illocutionary acts as changing states of affairs in the external world, or as creating (non-linguistic) obligations. Austin’s performative utterances are simply acts or events, in contrast to the “social acts” discussed by Reinach (1913), whose ontological universe included such things as “promisings”. On this point, see Laugier (2004). 10. Apart from formulaic phrases like “unless a contrary intention shall appear”, commonly found in meanings clauses, the use of ‘shall’ with ‘unless’ is rare in statutory language. However, this collocation does occur, for example in the Wills Act 1838 s18(4)(b): “[A]ny other disposition in the will shall take effect also, unless it appears from the will that the testator intended the disposition to be revoked by the marriage.” [my emphasis] The fact that it is much easier to find examples where ‘shall not’ occurs with ‘unless’ is no doubt due to the strong collocation of ‘unless’ with ‘not’ rather than with shall. 11. In certain precisely defined circumstances, French law imposes not just a moral but also a legal obligation on a person to do what he can to help others; penalties may be imposed for “non-assistance à personne en danger”. 12. See Charnock (2006) on literal interpretation and the treatment of ambiguity in legal adjudication. 13. The same judge had already proposed a significant loosening of the literal rule in the field of contract in Mannai Investment v Eagle Star Life Assurance

196 Ross Charnock 1997, and in Investors Compensation Scheme v West Bromwich Building Society 1998. 14. The proposed rejection of literal meaning brings out a point of disagreement between Grice and Austin. Grice’s (1989) theory of conversational implicature presupposes the existence of a literal meaning to which the machinery of implicature may be applied. Indeed, a particular understanding cannot be said to be pragmatically derived unless the relevant words correspond literally to something else. Austin (1962), on the other hand, considered contextual use itself as constitutive of meaning. On this question, see Travis (1991). 15. Hart’s use of the term ‘penumbra’ may have been taken from Russell (1923), where the word is used in relation to “vagueness”.

References Austin, John 1995 The Province of Jurisprudence determined. (Rumble, ed.) Cambridge: Cambridge University Press. Original edition 1832. Austin, John L. 1962 How to Do Things With Words. Oxford: Oxford University Press. Bennion, Francis 1993 Hansard – help or hindrance? A Draftsman’s View of Pepper v Hart. Statute Law Review (Winter), 1–21. Charnock, Ross 2006 Forms and functions of ambiguity in English common law adjudication”. In Gotti, M. and J. Flowerdew. J (eds.) Studies in Specialized Discourse. Berne: Peter Lang. 239–261. (Linguistic Insights 41.) forthcoming. Overruling as a speech act: performativity and normative discourse. Journal of Pragmatics (special issue on “Speech acts in legal language”). Garzone, Giuliana 2001 Deontic modality and performativity in English legal texts. In Gotti, M., and M. Dossena (eds.) Modality in specialised texts, Bern: Peter Lang, 153–173. Grice, H. Paul 1989 Studies in the Way of Words. (Collected Papers). Cambridge, Mass: Harvard University Press. Hart, Herbert L.A. 1994 The Concept of Law. 2nd ed. Oxford: Clarendon. Original edition 1961.

When may means must

197

Laugier, Sandra 2004 Performativité, normativité et droit. Archives de Philosophie 67: 505–524. Putnam, Hilary 1975 Mind, Language and Reality. Cambridge: Cambridge University Press. Recanati, François 2004 Literal Meaning. Cambridge: Cambridge University Press. Reinach, Adolf 1913 Die apriorischen Grundlagen des bürgerlichen Rechts. [English translation in Crosby, J.F., (ed.) Aletheia, 1983, 1–142.] Rosch, Eleanor. 1983 Prototype classification and logical classification – the two systems. In New Trends in Cognitive Representation: Challenges to Piaget's Theory, Ellin K. Scholnick (ed.), 111–144. Hillsdale NJ: Erlbaum. Russell, Bertrand 1923 Vagueness. In The Collected Papers of Bertrand Russell, Vol. 9:147–154. London: Routledge [1992]. Searle, John R. 1979 Literal meaning, In Expression and meaning. Cambridge: Cambridge University Press. Stroud, Frederick 1986 Stroud’s Judicial Dictionary of Words and Phrases. (5th edition). James, John, S. (ed.). London : Sweet and Maxwell, 5 vols. Travis, Charles 1989 The Uses of Sense: Wittgenstein’s Philosophy of Language. Oxford: Clarendon. 1991 Annals of analysis. Mind 100: 237–264. Waismann, Friedrich 1951 Verifiability. In Logic and Language (1st series), Anthony Flew (ed.), 119–112. Oxford: Blackwell.

Cases cited Anns and others v Merton Borough Council [1977] AC 728 Alderman Backwell’s case (1683) 1 Vern 152 Grey v Pearson [1857] All ER 21 Investors Compensation Scheme v West Bromwich Building Society [1998] 1 WLR 896 Julius v Oxford (Bishop) [1880] 5 AC 214

198 Ross Charnock Kirin-Amgen Inc v Hoechst Marion Roussel [2005] 1 All ER 667 Mannai Investment v Eagle Star Life Assurance [1997] AC 749 Murphy v Brentwood District Council [1990] 2 All ER 908 Pepper (Inspector of Taxes) v Hart [1993] 1 All ER 42 R v Brixton Prison (Governor) ex parte Enohoro [1963] 2 All ER 477 Reg v Tithe Commissioners (1852) 14 QB 459 River Wear Commissioners v Anderson [1877] All ER 1 Re Shuter (2) [1959] 3 All ER 481 Sussex Peerage case (1844) 8 ER 1034.

Statutes Bankrupts Act 1571 Church Discipline Act 1879 Fugitive Offenders Act 1881 Obscene Publications Act 1959 Offences Against the Person Act 1861 Public Health Act 1936 Tithe Act 1842 Theft Act 1968 Wills Act 1838

Legal English and the ‘modal revolution’ Christopher Williams

1. Introduction There is a joke (albeit not a very funny one) about modals involving a Frenchman on holiday in England; he was swimming off the Brighton coast when he suddenly got into difficulties and started shouting “I will drown and no one shall save me”. So no one did. How was the poor man supposed to know about the subtleties of shall, that it has deontic force in the second and third person? It has been by far the most commonly used modal auxiliary for at least 600 years in prescriptive legal English, and it will feature large in this paper. The language of the law has commonly been portrayed – not just in English – as one of the most conservative types of discourse. And yet, as a result of deliberate drafting policy, over the last fifteen years or so there has been something of a revolution in the choice of finite verbal constructions in some legislative texts in certain English-speaking countries. It is the nature, extent and impact of this ‘modal revolution’ that I wish to explore here.

2. Calls for reforming legal language For centuries it has been acknowledged that legal language tends to be abstruse and often incomprehensible to the layperson. Illustrious statesmen and writers such as Thomas Jefferson and Charles Dickens have poked fun at (and sometimes poured scorn on) the legal profession in this regard. Some, including Jeremy Bentham, adhere to the ‘conspiracy theory’, namely that the abstruseness of legal language is the result of an exclusionary policy which obliges ordinary citizens to rely on a lawyer’s skills to interpret a legal text for them. During the 1960s consumer movements gathered strength calling for greater respect for the demands and rights of such ordinary citizens, and by the 1970s the Plain Language Movement – originating in the United States but soon spreading to England and to all major

200 Christopher Williams English-speaking countries – had become sufficiently well-rooted and vociferous in its calls for overhauling legal English that it began to penetrate fringes of the legal establishment and academia.

2.1. Abolishing shall Various aspects of legal language were called into question, and reforms were proposed to reduce sentence length, the use of the passive and of nominalization, to remove Latinisms and to replace obscure, archaic lexical items with terms closer to the language of everyday life.1 Moreover, as regards the finite verbal constructions used, the modal auxiliary most strongly associated with legal English – shall – came in for unprecedented criticism. It was considered to be “ubiquitous, imprecise, and royal sounding” (Mowat: 1994), “a ‘dead’ word never heard in everyday conversation” (Lauchman 2002: 47). In short, to quote the title of one of the many articles on the subject, it was deemed that ‘Shall must go’ (Asprey 1992). By the late 1990s shall had indeed been banished from a number of prescriptive legal texts, particularly in the major English-speaking countries of the southern hemisphere – Australia, New Zealand and South Africa. To give an idea of the extent of the changes brought about I need merely repeat (see Williams 2005a; 2005b) the case of South Africa’s Interim Constitution of 1994 which contains 1288 cases of shall, 0 cases of must, and 265 cases of may, whereas the definitive text of the Constitution drafted less than three years later contains 0 cases of shall, 414 cases of must, and 274 cases of may. Such a striking distribution shift in modals can only be described to my mind as revolutionary, a clear case of language engineering carried out, in this particular case, thanks to the influence of the Canadian lawyer and Plain Language exponent Phil Knights who was a member of the Constituent Assembly which actually drafted the 1997 Constitution. In the northern hemisphere shall-free legislative texts can also be found to a limited degree in Canada, while in the United States and the United Kingdom, the countries which first spawned the Plain Language Movement, and in the major international organizations such as the United Nations, the International Labour Organization and the European Union, legislative drafters have tended to cling to tradition. However, for example, within the Translators Service of the EU there is a ‘Fight the Fog’ campaign aimed at bringing the language of officialdom closer to the citizen.2 (Fog, of course, is the metaphor for the entire English legal system in

Legal English and the ‘modal revolution’

201

Charles Dickens’s novel Bleak House.) Nevertheless, the EU authorities do not envisage doing away with shall which can still be found abundantly in the English version of the EU Constitution, with over 1700 instances of shall in the 2004 draft. I therefore decided to examine more closely the choice of finite verbal constructions used in those legislative texts where shall has been deliberately removed and to compare this shall-free corpus with the contemporary situation of prescriptive texts in English seen at a world level where, generally speaking, among the modal auxiliaries shall still reigns supreme. 3. Comparing ‘Shall-free data’ and ‘World data’3 My shall-free corpus consists of legislative texts from Australia, South Africa and New Zealand, constituting a total of over 160,000 words. One of the texts is exceptionally lengthy, namely the New South Wales Local Government Act 1993, and the South African Constitution 1997 itself is over 30,000 words. I then compare the results of this corpus with what I have called the ‘World data’ which I have compiled from 36 prescriptive texts coming from a variety of sources (the US, the UK, the EU, the UN, the ILO, Australia, Canada, Ireland, South Africa and New Zealand) constituting a total of over 140,000 words. The results are shown in Figure 1. The most significant differences in the ‘Shall-free data’ with respect to the ‘World data’ would appear to be as follows: – –

– –

a rise of almost nine percent in the use of the present simple, from over 43 percent in the ‘World data’ to 52 percent in the ‘Shall-free data’; a massive increase in the use of must, from just over three percent in the ‘World data’ to well over 13 percent in texts where shall has been done away with; a significant rise in may in the ‘Shall-free data’, up from over 13 percent in the ‘World data’ to over 18 percent; a major increase in the use of the semi-modal be to, which constitutes less than one percent in the ‘World data’ but more than quadruples in the ‘Shall-free data’.

One striking feature about both sets of data is the predominance of just three finite verbal forms, with the present simple, may and must accounting for almost 84 percent of the total in the ‘Shall-free data’, and the present

202 Christopher Williams

Shall-free data present simple may must be to present perfect past simple would present progressive will can

% 52.0 18.2 13.7 4.3 4.1 3.3 1.0 0.6 0.6 0.5

per 1000 words 29.3 10.3 7.7 2.4 2.3 1.9 0.5 0.3 0.3 0.3

World data present simple shall may present perfect past simple must should will be to would

% 43.5 22.6 13.3 4.7 3.6 3.3 2.2 1.6 0.9 0.8

Per 1000 words 21.5 11.2 6.5 2.3 1.8 1.6 1.1 0.8 0.4 0.4

Figure 1. Percentages and occurrences per 1000 words of the 10 most frequently used finite verbal constructions of the ‘Shall-free data’ and the ‘World data’

simple, shall and may accounting for over 79 percent of the total in the ‘World data’. Another remarkable feature is the primacy of the present simple in both sets of prescriptive legal texts, a characteristic ignored by most scholars of legal language who generally concentrate on the ‘classic’ modal auxiliaries. This predominance can partly be explained by the fact that I have taken into consideration all finite verbal constructions, in both main clauses and subordinate clauses and, as is well-known, legal English abounds in subordinate and embedded clauses. Nevertheless, even if we ignore the use of the present simple in subordinate clauses, we still find a huge number of cases where it is adopted in main clauses, for example when establishing broad questions of principle, as in: (1)

This Act binds the Crown in right of New South Wales and, in so far as the legislative power of Parliament permits, the Crown in all its other capacities, except to the extent to which this Act otherwise provides (New South Wales Local Government Act 1993, Chapter 1[4])

or in so-called ‘definition provisions’:

Legal English and the ‘modal revolution’

(2)

203

In this Act […] “the city” means the City of Nottingham […] “vehicle” means a vehicle intended or adapted for use on the road. (Nottingham City Council Act 2003)

or when used performatively as in: (3a) There is hereby established a South African National Academy of Intelligence as an Organisational component in terms of the Public Service Act, 1994 […] (South Africa Intelligence Services Act 2002 Chapter 5[1]) (3b) The Security Council […] Unequivocally condemns in the strongest terms the horrifying terrorist attacks which took place on 11 September 2001 in New York, Washington D.C. and regards such acts, like any act of international terrorism, as a threat to international peace and security (UN Security Council Resolution 1368 of 12 September 2001). I take the view that – with just a few exceptions – all finite verbal constructions in main clauses in prescriptive legal texts are to be considered as being modal, irrespective of whether or not they contain a so-called ‘modal’ auxiliary. In other words, even when the present simple is used as in this clause from the South African Constitution of 1997 (4)

The national flag of the Republic is black, gold, green, white, red and blue, as described and sketched in Schedule 1 (Section 4, South Africa Constitution 1997)

is clearly has prescriptive rather than descriptive value, precisely because of the principle-establishing function of the clause and its collocation in a legally binding text. Therefore, as I have argued elsewhere (Williams 2004a: 242), the present simple should be considered as the ‘normative indicative’ in prescriptive legal English, just as it is in, say, French or Italian. Returning to the differences between the ‘Shall-free data’ and the ‘World data’, how do we account for the almost nine percent increase in the adoption of the present simple in texts where shall has been eliminated? There are several reasons for this increase, one being because some of the more traditionally-minded drafters still sometimes use shall in subordinate clauses where it is purely ornamental and merely conjures up ‘the flavour

204 Christopher Williams of the law’ without actually conveying any particular meaning, as in the opening clause of this sentence (5)

Before any permit required by this chapter shall be issued, the annual fee required therefore shall be paid into the state treasury […] (Missouri Revised Statutes, Chapter 312: Nonintoxicating beer).

However, my impression is that such cases are much less frequent in prescriptive texts these days than they were a few decades ago. Where there would seem to be a significant increase in the use of the present simple instead of shall is in main clauses when laying down general principles or conditions where the sense of obligation is weak or practically non-existent, frequently where there is a non-human subject. For example, if we return to quote 4) on the South African national flag, we can see that in the Interim Constitution of 1994 it says (6)

The national flag of the Republic shall be the flag the design of which is determined by the President by proclamation in the Gazette (Section 2(1) of South Africa Interim Constitution 1994).

Other cases where shall was changed to the present simple between 1994 and 1997 include the following: (7a) Section 1(1) The Republic of South Africa shall be one, sovereign state (Interim 1994) (7b) Section 1 The Republic of South Africa is one sovereign democratic state […] (Constitution 1997) (8a) Section 2(2) The national anthem of the Republic shall be as determined by the President by proclamation in the Gazette (Interim 1994) (8b) Section 4 The national anthem of the Republic is determined by the President by proclamation (Constitution 1997) (9a) Section 9 Every person shall have the right to life (Interim 1994) (9b) Section 11 Everyone has the right to life (Constitution 1997). There are literally hundreds of such alterations between the Interim version and the definitive version. I shall be giving further examples of changes made between the two texts since they offer us a means of verifying in de-

Legal English and the ‘modal revolution’

205

tail how this ‘modal revolution’ was undertaken, whereas with most shallfree texts, such as the New South Wales Local Government Act 1993, we have no previous text to compare it with and can only make an inspired guess as to which constructions would have been drafted using shall if it had not been explicitly banned. Perhaps even more significant than the increase in the use of the present simple is the increase in the frequency of must, up from a lowly three percent in the ‘World data’ to over 13.5 percent in shall-free texts. This is particularly striking in view of the general decline in the use of must in British, American and Australian English as a whole, particularly of deontic must, as outlined by Leech (2003), Smith (2003), and Collins (1991). Clearly, must is generally considered by drafters wanting to avoid shall as being the closest equivalent to shall when expressing an obligation. Equally clearly, the metafunction of the legislative text will strongly determine the overall frequency of occurrences of given verbal constructions. For example, in the New South Wales Local Government Act 1993, which is essentially concerned with regulating the administrative powers of local authorities, must accounts for 9.4 percent of all finite verbal constructions, whereas it constitutes a staggering 28 percent of all finite verbal constructions in the South Africa Constitution 1997, a document which represents not just a linguistic revolution but above all a political revolution in marking a definitive break with the horrendous apartheid regime that had ruled until the early 1990s, setting out the new rights and duties of the nation’s multiracial and multilingual community. In this particular context, the decision to adopt on a massive scale a modal auxiliary frequently used to express strong obligation in everyday conversation as opposed to the rather archaic, legalisticsounding shall would seem to be highly appropriate. The following is a particularly clear example of how the Interim Constitution of 1994 was modified in accordance with Plain Language principles in the 1997 Constitution: (10a) Section 4(1) This Constitution shall be the supreme law of the Republic and any law or act inconsistent with its provisions shall, unless otherwise provided expressly or by necessary implication in this Constitution, be of no force and effect to the extent of the inconsistency (Interim 1994)

206 Christopher Williams (10b) Section 2 This Constitution is the supreme law of the Republic; law or conduct inconsistent with it is invalid, and the obligations imposed by it must be fulfilled (Constitution 1997). Besides the elimination of shall in the 1997 text and its replacement by the present simple where laying down a general principle and by must where laying down an obligation, we can also see how the latter text has been drafted so as to avoid the syntactic discontinuity (a typical feature of legal language, as Bhatia (1993: 112) among others has pointed out) present in the Interim text, where shall is separated from the main verb by an embedded clause, and how the second half of the sentence has been rephrased, and also considerably shortened, so as to change its negative meaning (“shall […] be of no force” etc., with a negative embedded clause wedged in the middle) into a positive meaning (“must be fulfilled”) using the modal of strong obligation. This is in line with the Plain Language Movement’s suggestion that negative structures should be avoided as a rule because they may cause confusion and tend to be more difficult to process as information. Such is the preponderance of must in the text that the drafters occasionally seem to fall into the same habit used by traditional drafters with shall, by inserting it even in subordinate clauses – here in an ‘if’ clause – where it would appear to convey no discernible modal function: (11) Section 73(5) A Bill passed by the National Assembly must be referred to the National Council of Provinces if it must be considered by the Council. A Bill passed by the Council must be referred to the Assembly (Constitution 1997). The five percent rise in the use of may in shall-free texts is certainly attributable in part to the fact, sometimes overlooked by linguists, that shall not is frequently used – much more often than may not – by traditional drafters to indicate prohibition. Just to give one example, the Consolidated Version of the Treaty on European Union of 2002 contains 18 cases of shall not but not one of may not, e.g.: (12) Abstentions by members present in person or represented shall not prevent the adoption of decisions referred to in paragraph 2 (Article 7(5) of the Consolidated Version of the Treaty on European Union 2002).

Legal English and the ‘modal revolution’

207

By contrast, here is one of the 84 instances in which may not is used in the New South Wales Local Government Act 1993. It also illustrates that syntactic discontinuities can still be found in shall-free texts: (13) A person may not, in respect of the same ward, be enrolled more than once in a roll of electors (Section 304[1] of the New South Wales Local Government Act 1993). Another major jump in terms of frequency concerns the semi-modal be to which rises from less than 0.7 percent in the ‘World data’ to over 4.2 percent in the ‘Shall-free data’, beating all other modals except may and must. The use of this semi-modal is specifically encouraged by the Australian Office of Parliamentary Counsel (2000) in cases where there is “the need to use a gentler form” of obligation, as in Section 158 of the New South Wales Local Government Reform Act 1993: (14) A draft local approvals policy is to consist of three parts. Part 1 is to specify the circumstances (if any) in which (if the policy were to be adopted) a person would be exempt from the necessity to obtain a particular approval of the council (Section 158 of the New South Wales Local Government Reform Act 1993). In my opinion, in such cases it is highly likely that shall would be used instead of be to by most drafters not obliged to avoid shall as part of their drafting policy.

4. Conclusions Finally, what are the possible long-term consequences of this ‘modal revolution’ in legal English, also in terms of how easily it could be exported to other English-speaking countries, particularly the United States, given that the highest concentration of English speakers is to be found there. The question is a complex one because shall is perceived in the US – much more than, say, in the UK – as being an anachronism. Hence calls for the abolition of shall in American legal English continue to grow. At the same time deontic must is only rarely used in everyday communication in the US. One therefore wonders how a massive influx of instances of must in American legal texts would be perceived by legal practitioners and by the

208 Christopher Williams general public, especially in the long term. On the other hand, if you abolish shall from legally binding texts, how else do you unequivocally convey obligation in legal English? It would also appear to be highly unlikely that the UK, the ‘mother’ of legal English in its various guises throughout the world, will decide to do away with shall in the near future. Anyone familiar with the enactment clause that precedes any piece of legislation in the UK today, even in the newly devolved assemblies of Scotland, Wales and Northern Ireland, will know that it reflects the language patterns and values of late medieval society. And the repeated use of shall in legislative texts, together with other discoursal and linguistic features such as binomials (e.g. “null and void” or “unless and until”), multinomials (e.g. “adjudged, ordered and decreed”) and formulaic expressions, such as “the truth, the whole truth, and nothing but the truth”, contributes in conveying what Danet has termed the “poetization” of legal language which distances it from everyday conversation and makes it more ‘majestic’ and authoritative, or old-fashioned and alienating, depending on one’s perspective, to the ordinary citizen. It is no coincidence that the English-speaking countries most prepared to experiment with innovative forms of legal language are those on the periphery, as it were, in terms of their political and historical role in the world, countries less tied down by the past and by conceptions of grandeur and self-importance. Moreover, it would appear that, so far, the introduction of shall-free legislative texts has not led to any noticeable increase in cases of litigation in the courts. On the other hand, if the predominant legal culture in a country is still reluctant to make radical changes, it would be counterproductive to insist on bringing about a ‘modal revolution’. In the end, a pragmatic approach would seem to be preferable to any dogmatic imposition. The case of shall in legal language is rather like that of the monarchy and the House of Lords in British institutional life: you feel deep-down that their days are probably numbered, that sooner or later they will be done away with. But not just yet.

Notes 1.

The question of the proposed reforms of legal language and the Plain Language movement is discussed in more detail in Williams (2004b) and Williams (2005a).

Legal English and the ‘modal revolution’ 2. 3.

209

At europa.eu.int/comm/translation.en/ftlog/ For a more detailed analysis of the data, see Williams (2005a).

References Asprey, Michèle 1992 Shall must go. The Scribes Journal of Legal Writing. 3/79: 79–83. Australian Office of Parliamentary Counsel 2000 Plain English Manual, 19. Available on the web at http://www. opc.gov.au/about/draft_manuals.htm Bhatia, Vijay 1993 Analyzing Genre: Language Use in Professional Settings. London: Longman. Collins, Peter 1991 The modals of obligation and necessity in Australian English. In English Corpus Linguistics: Studies in Honour of Jan Svartvik, Karin Aijmer, and Bengt Altenberg (eds.), 145–165. London: Longman. Danet, Brenda 1984 Legal Discourse. In Handbook of Discourse Analysis. vol. 1: The Disciplines of Discourse Analysis, Teun A. van Dijk (ed.), 273–291. London: Academic Press. Facchinetti, Roberta, Manfred Krug, and Frank R. Palmer (eds.) 2003 Modality in Contemporary English. Berlin/New York: Mouton de Gruyter. Lauchman, Richard 2002 Plain Language: A Handbook for Writers in the U.S. Federal Government. Available on the web at www.mindspring.com/~rlauchman/PDFfiles/ PLHandbook.PDF. Leech, Geoffrey 2003 Modality on the move: the English modal auxiliaries 1961–1992. In Facchinetti, Krug, and Palmer (eds.), 223–240. Mowat, Christine 1994 Buddhist, Running, and Plain Language in Calgary (Parts One and Two). Michigan Bar Journal, July and August issues. Available at: www.michbar.org/committees/penglish/columns/94_aug.html. Smith, Nicholas 2003 Changes in the modals and semi-modals of strong obligation and epistemic necessity in recent British English. In Facchinetti, Krug and Palmer (eds.), 241–267.

210 Christopher Williams Williams, Christopher 2004a Pragmatic and Cross-cultural Considerations in Translating Verbal Constructions in Prescriptive Legal Texts in English and Italian. In Intercultural Discourse in Domain–Specific English, special issue of Textus, Christopher N. Candlin. and Maurizio Gotti (eds.), 17: 217– 246. 2004b Legal English and Plain Language: an introduction. ESP Across Cultures 1: 111–124. 2005a Tradition and Change in Legal English: Verbal Constructions in Prescriptive Texts. Bern: Peter Lang. 2005b Vagueness in legal texts: is there a future for shall? In Vagueness in Normative Texts, Vijay Bhatia, Jan Engberg, Maurizio Gotti, and Dorothee Heller (eds.), 201–224. Bern: Peter Lang.

Posteriority in expressions with must and have to: a case of interplay between syntax, semantics and pragmatics1 An Verhulst

1. Introduction: a meeting ground between modality and time A lot has been written on the labels with which we try to capture modal meanings, the most popular labels being epistemic and root modality. A topic that has received less attention is the expression of temporal information in modal utterances. The following quotes show that observations in which modality and time meet are often restricted to occasional and rather vague references: [B]oth the present and the past tense forms of the modals share abnormal frames of time reference and may be used to refer to the present and the future: ‘They could well be leaving on the next ferry’; ‘Perhaps you might like to get in touch with the committee now. (Hoye 1997: 75) [Some] modals must co-occur with the perfect in order to communicate past time. [E.g.] ‘You should have cleaned your room.’ … ‘Chandra could have done it.’ … In some cases, however, a semi-auxiliary is required to communicate past time. … To communicate strong obligation in the past, a speaker can use the past tense of have to – ‘Thomas had to clean his room last night.’ (Berk 1999: 144, emphasis mine)

This paper focuses on the time/modality interface with a twofold aim. The first is of a terminological nature. A moment’s reflection on the precise meaning of the present, past or future orientation of modal utterances raises some questions. For instance, the ‘past times’ referred to in the quote by Berk refer to two different types of past time. The difference can be explained using the model of the English tense system outlined by Declerck (1991). One of the basic insights of this theory concerns the distinction between, on the one hand, the establishment of a temporal domain by the

212

An Verhulst

location of a situation in a (present, past or post-present) temporal sector and, on the other hand, the temporal relations (of simultaneity, anteriority or posteriority) expressed in this domain. For example, in He told me that he had bought the doll the act of telling is situated in the past time-sphere, while there is a relation of anteriority between the buying and the telling. This distinction can be applied to utterances with a modal auxiliary, which consist of an auxiliary with a modal meaning (M) and a residue (the infinitive clause P).2 On the one hand, there is the temporal reference of the modal meaning and, on the other hand, there is the temporal relation between the modal meaning and the time of the situation referred to by the infinitival complement. For example, a suitable paraphrase of Thomas must have cleaned his room is: ‘I confidently conclude now that Thomas cleaned his room’, in which must locates the necessity in the present, while there is a relation of anteriority between the necessity and the infinitive have cleaned. Thus, with respect to the examples by Berk (1999) quoted above, what is past in Thomas had to clean his room is the modal meaning, while in You should have cleaned your room it is the action expressed by the complement that is past.3 Putting both these utterances under the heading Using the Perfect to Express Past Modality (Berk 1999: 144) is bound to cause confusion. We need a framework like the one referred to above for a correct understanding of the temporal information in modal utterances. By way of example, the present analysis centres on posteriority in utterances with must and have to, which is especially interesting for expressions of root necessity, in which the situation in the complement is often situated in the future (cf. Berk 1999: 105; Coates 1983: 235; Krug 2000: 55; Larreya 1984: 108; Palmer 1986: 97). For instance, (1) can be paraphrased as ‘it is necessary now that you will clean the room’. (1)

You must clean the room. 4 (posteriority, root)

The second aim of this paper is to get a better insight in the ways in which this temporal relation comes about. More particularly, in what follows I will consider three factors, i.e. the weight of particular aspectual features of the infinitive clause, of pragmatic information and of adverbials of time and frequency.

Posteriority with ‘must’ and ‘have to’

213

2. Structural, semantic and pragmatic development of posteriority 2.1. The aspectual nature of infinitive clauses A first factor that is of importance with respect to temporal relations is the aspectual nature of the infinitival complement. To understand this, compare (1) with (2), which has a similar structure, though there is simultaneity. (2)

He must be very intelligent (simultaneity, epistemic)

Simultaneity between M and P is only possible if the infinitive clause contains a homogeneous situation. (Non-)homogeneity is an aspectual feature that concerns the internal constituency of a situation (cf. Declerck 2006: 45). An example of a homogeneous situation is “working in the garden”. The subparts of the situation are of the same kind as the entire situation. As a result, a homogeneous situation can be used to refer not only to the whole situation but also to a subpart: if ‘I worked in the garden’ is a true description of what you did from 10-12, it is also a correct representation of what you did from 10-11. Thus, in (2) be very intelligent denotes a homogeneous situation. 5 By contrast, in (1) clean the room is heterogeneous: no part of the situation is itself an instance of cleaning the room. Every stage of the cleaning is an instance of cleaning part of the room. This information is important for the interpretation of (1) and (2). The situation referred to by must (i.e. the obligation as an illocutionary act) is punctual: it coincides with the moment of speaking. Consequently, there can only be simultaneity if it is possible to focus on a subpart of the situation denoted by the infinitive clause. This is possible in (2), but not in (1), in which clean the room encompasses the situation as a whole. It is important to note that homogeneity is not only determined by the lexical composition of the verb phrase (lexical aspect); it can also be determined by grammatical features (grammatical aspect). For instance, the situation denoted by be cleaning the room in (3) is homogeneous due to the use of a progressive form. (3)

She must be cleaning the room. (simultaneity, epistemic)

In other words, the aspectual nature of the infinitival clause appears to play a crucial role in the kind of temporal relation that is established. As a

214

An Verhulst

final illustration, consider the ambiguity of (4). The two most plausible readings of this instance differ with respect to the interpretation of the infinitival clause. (4)

My boss must be giving a very important speech. (ambiguous: simultaneity / posteriority, epistemic)

On the first reading, be giving has progressive aspectual meaning, in which case it denotes a homogeneous situation and there is simultaneity (e.g. My boss must be giving a very important speech as even the PM is in the audience, taking notes). On the second reading, be giving has no aspectual meaning but denotes a pure (arranged) future (e.g. My boss must be giving a very important speech, for even the PM called to say that she will come). As a result, be giving a very important speech denotes a heterogeneous situation and there is posteriority.

2.2. The future-orientated nature of necessity Some studies have distinguished between root and epistemic modality on the basis of formal or semantic patterns. Thus, Coates (1983) argues that “an Epistemic modal has a 100 per cent association with features such as Progressive aspect … or Stative verb” (246). Later research (e.g. Papafragou 2000; Wärnsby 2004) has put into perspective these claims. For instance, Papafragou (2000) offers convincing arguments for her claim that properties of the verbal complement in modal utterances, “rather than being structurally determined by the type of main-clause modal (root or epistemic), correspond to pragmatic aspects of the interpretation process.” (101) With regard to my observations on aspect (cf. 2.1) this means that aspectual features are but one factor that determines temporal relations. For example, in (5) the complement denotes a homogeneous situation, while the temporal relation depends on the interpretation of be happy. (5)

The children must be happy. (Tregidgo 1982: 84) (ambiguous: simultaneity, epistemic / posteriority, root)

If be happy describes a present state, there is simultaneity, and the utterance will be interpreted in terms of epistemic necessity. By contrast, if be happy is to be interpreted as a strongly desired result, as the future reali-

Posteriority with ‘must’ and ‘have to’

215

sation of this state, there is posteriority; the utterance will then be interpreted in terms of root necessity. The posteriority in the interpretation of (5) in terms of root necessity is due to the logical fact that a necessity often implies the expectation of fulfilment. This expectation is not inherent in the semantics of the modal auxiliary; it is just part of a speaker’s general knowledge and is activated when interpreting an utterance with a modal auxiliary that can express a necessity. The temporal information of (5) can then be described as follows. Must locates the necessity in the present. The infinitive, be, is tenseless because it is a non-finite form (cf. Declerck 1991: 34). This means that it cannot express a temporal structure; it can only indicate a temporal relation. The reference time of this relation is clarified by the context. Thus, the temporal relation between must and be happy can be considered as ‘pragmatic posteriority’, as it comes about through the expectation of future actualisation of the necessity. 6 Yet, while agreeing with Papafragou’s conclusion that context is crucial to the decoding of modal utterances (cf. Papafragou 2000: 64, 103), I think we should be careful not to overestimate the workings of pragmatics. One of the examples Papafragou gives to illustrate the importance of contextual information is (6), which can be interpreted in terms of root necessity. (6)

I must be the best chess player there is. (Papafragou 2000: 102) (ambiguous: simultaneity, epistemic / posteriority, root)

The interpretation is described as follows: “be (the best chess player) … is understood as marking the beginning of a state (one in which the speaker is the best chess player). This inchoative interpretation corresponds to an achievement, and achievements are activities.” (Papafragou 2000: 102). However, in my view, it is not accurate to conclude that there is a “reinterpretation of be … as [an] activity predicate” and that “the complement of the modal denotes an activity” (Papafragou 2000: 102-103, emphases mine). As in (5), the idea of achievement in (6) arises from the temporal relation of posteriority between the modal meaning and the complement rather than being denoted by the verbal complement. More specifically, the dynamic aspect lies in the implication of action generated by the relation of posteriority: for a future situation to be the case some action has to be realised (‘it is necessary that I will be the best chess player; therefore, I must improve my playing skills and insight in the game’). In other words, the complement clause (be the best chess player) still denotes a state (semantic

216

An Verhulst

level), while the implication refers to a dynamic situation (pragmatic level). These two levels should not be mixed up; pragmatics can only supplement lexical meaning, not replace it. 7 To sum up, this analysis explains why the temporal relation in necessity examples with an infinitive clause that denotes a homogeneous situation can be ambiguous between simultaneity and (pragmatic) posteriority. A final example of this is given in (7). (7)

The committee must be looking at new ways to realise tax cuts. (ambiguous: simultaneity, epistemic / posteriority, root)

In the two most likely interpretations of this instance the progressive infinitive be looking has aspectual meaning. Either be looking describes a present situation, in which case there is simultaneity and the utterance will be interpreted in terms of epistemic necessity, or the utterance describes the necessity of a future event of be looking, in which case (7) will be interpreted in terms of root necessity. In the last section I will look at the role of adverbials in the interpretation process.

2.3. Adjuncts of time and frequency Example (8) contains a temporal adverb, which specifies the time of the expected realisation of the obligation. (8)

You have to clean the room tomorrow. (posteriority, root)

The adverb has scope over the infinitival complement: tomorrow does not denote the time at which the obligation is made, but the time of the content of the obligation. In examples (9a-b) the adjunct plays a more important part. (9 a) gets an epistemic interpretation: ‘I confidently conclude now that she is back’. The addition of a temporal adverb in (9 b) brings about a different temporal relation. In the default interpretation this is accompanied by a change in the type of necessity: (9b) can be paraphrased as ‘she is required to be back by ten’. (9a)

She must be back. (simultaneity, epistemic)

Posteriority with ‘must’ and ‘have to’

(9b)

217

She must be back by ten. (posteriority, root)

This shows that there must be some link between the presence of adjuncts with future time reference and root necessity. When it comes to describing the precise nature of this link, Hoye (1997), which discusses the interplay between adverbs and modality, remarks that “adjuncts … do not in their various semantic roles impinge in any obvious way on modal meanings. However, … the adjuncts of time … interact with the modals in a complex way. The nature of the relationship between the modals and tense and the implications this has for co-occurrence with time adjuncts is largely determined by their epistemic or deontic status.” (152-3) In other words, in Hoye’s view, the type of necessity determines the interpretation of the temporal adjunct. I think it is more correct to state the reverse; as (9a-b) illustrate, temporal adjuncts do influence the type of modal meaning, be it in an indirect way: temporal adjuncts can bring about one or the other temporal relation, which in turn typically occurs in certain types of necessity. In the remainder of this section I will be looking at the meaning of the adverb always, whose scope is different from that of the adjuncts in the previous examples. In (10) a speaker utters what he considers to be a rule: (10) I believe that the dress code is crucial. The rules say gentlemen must always wear a jacket with a tie. (Cobuild) (posteriority, root) It is the time frame of the obligation that is interesting here. The sentence has generic reference, which arises from the presence of a generic subject and of the frequency adverb always.8 This could lead to the conclusion that there is simultaneity between must and wear. When the rule was first made, posteriority was clear: ‘from now on, gentlemen are obliged to wear a jacket and a tie’. Yet, as the rule became conventionalised (generalized), this relation of posteriority sunk back into simultaneity.9 However, if one takes into account the distinction between, on the one hand, the temporal relation between the modal meaning and the complement and, on the other hand, the temporal reference of the auxiliary, this view appears not to be correct. The presence of always turns the obligation into a general rule, but this does not change the temporal relation between the modal meaning and the infinitive clause. The temporal structure of (10) can be clarified by understanding the meaning of always as an indicator of unlimited frequency. In their discus-

218

An Verhulst

sion on frequency adjuncts Huddleston and Pullum (2002) remark that “with bounding [frequency] adjuncts it is a matter of how many times, [e.g. twice], while with non-bounding [frequency] adjuncts it is characteristically a matter of how often [e.g. regularly, always]” (714). Yet, they add that in cases of non-bounding frequency quantification an unexpressed restriction is often recoverable from the context (Huddleston and Pullum 2002: 718). For instance, I always handwash my blouse “does not entail that I spend my life handwashing the blouse. … [Its meaning can be captured by the following paraphrase:] ‘on all occasions when I wash this blouse, I do so by hand’.” (Huddleston and Pullum 2002: 718) Similarly, (10) can be paraphrased as ‘men have to wear a jacket with a tie whenever they are supposed to behave like gentlemen’. The necessity is thus valid in present, past and future situations which fulfil a contextual requirement. The meaning of (10) can then be described as follows. Firstly, the use of the present tense means that the obligation expressed by must concurs with the moment of speaking, just as in (1); next, there is a temporal relation of posteriority between the auxiliary and the infinitival complement; finally, the use of always does not imply that the situation denoted by the complement (the wearing) is simultaneous with the modal meaning (the necessity). Rather, always has scope over the whole of the modal auxiliary and the complement, as appears from the abstract paraphrase of examples like (10): ‘whenever one is confronted with a situation x, y has to be carried out by person z’; or, taking into account the temporal relation: ‘in situation x ([action y by person z] = posterior to modal meaning)’. This analysis again shows (cf. 2.2) that it is vital to respect the semantic information of syntactic units. It will be clear, then, that in the following claim temporal and modal meanings are confused: “because of the semantic association of M [= medial position in the sentence] with modality and degree, even adverbs which purportedly express different sense relations, such as time, take on a hint of these.” (Hoye 1997: 149) Hoye gives the following examples: Pupils must at all times wear school uniform and You should always remember to bring your key. Thus, in the latter instance always is considered as a temporal adverb with a modal shade, while in fact, it is no more than a generic marker; there is nothing in its semantics that could invite a modal meaning. Just as always in (10) does not transform the temporal relation between the auxiliary and its complement, must does not affect the meaning of always.

Posteriority with ‘must’ and ‘have to’

219

3. Conclusion The gist of this paper is that studying the temporal information in modal utterances puts into perspective the existence of verbal patterns, such as the typical occurrence of progressive, perfective and stative infinitives in epistemic modality. I propose that a more unified account of the processing of modal utterances can be reached by decoding the temporal information from a semantic and pragmatic point of view. To illustrate this point I have looked at the development of posteriority in instances with must and have to. The argument centres around three main factors in the realisation of this temporal relation. Firstly, I have pointed to the significance of the aspectual (homogeneous or heterogeneous) nature of the infinitive clause (cf. [12]). Secondly, I have shown that the posteriority between the modality and the infinitival complement can be merely pragmatic (e.g. [5]). Thirdly, I have mentioned the importance of adverbs (cf. [8-9]). Furthermore, I have emphasised that it is vital to take into account the existence of syntactic and semantic ‘units’. As such, I have argued that pragmatics can supplement but not substitute for lexical aspects of the infinitival complement (cf. [5-6]). Further, the analysis of (10) revealed that the temporal structure of modal utterances can be quite complex. Finally, it should be clear that this paper is meant as an impetus to the exploration of the area of time and modality. For example, an extension of the present discussion to other modal auxiliaries would be useful. In addition, a study of the expression of other temporal relations than posteriority in utterances of root necessity should further improve our understanding of this modal subfield.

Notes 1.

2.

3.

I would like to thank the Fund of Scientific Research – Belgium (Flanders) for making possible this research. Further, I am grateful to Renaat Declerck and Ilse Depraetere for their insightful comments on an earlier draft of this article. This distinction has not gone unnoticed in the literature. It has especially been referred to in observations on the negation of modal auxiliaries. For instance, Palmer (1990: 34–35) talks about ‘modality’ and ‘event’ (in non-epistemic modality) or ‘proposition’ (in epistemic modality). With respect to the examples by Hoye (1997) quoted above the modal meaning is twice situated in the present (‘It is possible that…’), but whereas in They could well be leaving on the next ferry the situation in the complement refers

220

4.

5. 6.

7.

8. 9.

An Verhulst to the future (‘… they will leave on the next ferry’), in Perhaps you might like to get in touch with the committee now it is situated in the present (‘… you like to get in touch with the committee now’). The observations in this paper are based on data from the Cobuild-corpus, the World Wide Web (abbreviated as ‘www’) and scientific literature. The source of examples that I adapted will not be given. In fact, states are always homogeneous since a state does not change throughout its duration (cf. Declerck 2006: 51). Though a more detailed discussion about the different types of modality is beyond the scope of this paper, it should be noted that the future orientation referred to here is especially clear in instances of so-called ‘performative necessity’ (Coates 1983: 33, 38), e.g. (1), in which a speaker requires a hearer to do something. When a speaker imposes an obligation on a hearer, the speaker expects that the action will be carried out, while the hearer knows that this is expected by the speaker. (Whether or not the hearer actually intends to or eventually refuses to carry out the action is linguistically irrelevant; what matters is that the communicative act of commanding something includes the expectation of future performance.) It is important to note that Papafragou’s focus on pragmatics follows from her adherence to Relevance Theory (Sperber and Wilson 1986/1995). The basic principle of this system is that an utterance communicates a presumption of its own optimal relevance. As a result, contextual factors play an important role in interpretation processes. In the light of this framework Papafragou proposes that “may, can, must and should all have a unitary semantics, with pragmatics accounting for the interpretations these words have in particular contexts.” (Salkie 2002: 716) Apart from frequency, always can also indicate duration, as in I’ve always been afraid of spiders. This interpretation was suggested to me during a discussion about example (10) at the Second International Conference on Modality in English, Pau, 2-4 September 2004.

References Berk, Lynn M. 1999 English Syntax. From Word to Discourse. Oxford: Oxford University Press. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm.

Posteriority with ‘must’ and ‘have to’

221

Declerck, Renaat 1991 A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha. Declerck, Renaat (in collaboration with Susan Reed, and Bert Cappelle) 2006 The Grammar of the English Tense System. (The Grammar of the English Verb Phrase. Vol.1). Berlin: Mouton de Gruyter. Hoye, Leo 1997 Adverbs and Modality in English. London: Longman. Huddleston, Rodney, and Geoffrey K. Pullum (eds.) 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Larreya, Paul 1984 Le Possible et le Nécessaire: Modalités et Auxiliaires Modaux en Anglais Britannique. Paris: Nathan-Recherche. Palmer, Frank R. 1990 Modality and the English Modals. 2nd ed. London: Longman. Original edition 1979. 1986 Mood and Modality. Cambridge: Cambridge University Press. Papafragou, Anna 2000 Modality: Issues in the Semantics-Pragmatics Interface. (Current research in the semantics/pragmatics interface 6.) Oxford: Elsevier. Salkie, Raphael 2002 Review of Anna Papafragou (2000), Modality: Issues in the Semantics-Pragmatics Interface. Journal of Linguistics 38: 716–718. Sperber, Dan, and Deirdre Wilson 1995 Relevance: Communication and Cognition. 2d ed. Oxford: Blackwell. Original edition 1986. Tregidgo, Philip S. 1982 Must and may: demand and permission. Lingua 56: 75–92. Wärnsby, Anna 2004 (De)coding modal utterances: the case of English must and Swedish måste. Paper presented at the Second International Confernce on Modality in English, Pau, 2–4 September 2004.

Using the adjectives surprised/surprising to express epistemic modality Christiane Rocq-Migette

“Modality could (…) be defined as the grammaticalization of speakers’ (subjective) attitude and opinions.” (Palmer 1986: 16). These can be expressed through a wide range of linguistic means. The corpus-based study presented here is focused on the use of be surprised/be surprising. These predicates basically refer to a feeling, the reaction of a speaker to an event or situation, and so the modality they convey is attitudinal. But where if-clauses complement It would be surprising / I would be surprised, the interpretation is mostly epistemic: thus the Longman Dictionary of English Language and Culture (1992: 1335) gives I wouldn’t be surprised if he changed his mind as equivalent to he will probably change his mind. What will be studied here is how this interpretation can be derived through the use of the modal would1, past tense indicating modal remoteness in the if-clause, and the meaning of the lexical items themselves. The part played by negation2 will also be commented on, as well as some problematic features. Almost all examples are taken from the British National Corpus (from now on referred to as BNC).

1. Syntactic arguments, clause order, semantic roles: points common to and differences between be surprising / be surprised In both cases, the if-clause limits the scope of the adjectival predicate. But its syntactic function is different depending on whether surprising or surprised is used:

1.1. Surprising Surprising belongs to the set of adjectival predicates taking extraposed that- or to-clauses. If-clauses are also extraposed. is cataphoric, the if-

224 Christiane Rocq-Migette clause being in fact the logical subject, which expresses the external cause of the surprise. The experiencer is not expressed explicitly: (1)

It would be surprising if education escaped being profoundly influenced by secularism.(The Effective Teaching of Religious Education, B Watson, BNC)

This could be rephrased, if the clause order was grammatical: (1’)

* would be surprising.

1.2. Surprised Surprised is originally a past participle, hence a passive meaning. The clause that follows it is the complement of the adjective and could be analysed as the original grammatical subject of an active form. Examples of an active form are attested (with the if-clause extraposed again):

(2)

It would surprise me very much if we agreed to a boycott,’ said one Italian analyst. (Daily Telegraph, electronic edition, 1992, BNC)

Both active and passive forms make it explicit that the speaker is the experiencer. But by using the be-surprised form, the speaker brings into focus and thus emphasises the fact that it is his or her personal opinion that is expressed. (3)

‘We will look very closely into the suggestion that the operation was cancelled because we were short-staffed, I would be very surprised if that was the case,’ he said. (Northern Echo, Leisure pages, BNC)

These differences between the two structures have consequences for the type of discourse in which each is used: with surprising, the evaluation is not explicitly attributed to anybody but reasons for the opinion expressed are present in the context, inferences are made. The great majority of the examples from the BNC occur in argumentative contexts, often in academic discourse or press articles. The speaker or writer has a point to

Surprised / surprising and epistemic modality

225

make, gives arguments and concludes with the sentence containing would be surprising as in (4) below. The structure can often be found in written texts. (4)

No one has attempted a census of the fauna, but there are many common birds and the vegetation must supply food for many insects, so it would be surprising if the insectivores and small rodents were not colonising, and thereby attracting predators. The soil fauna may also be developing…(New Scientist, 1991, BNC)

When I’d be surprised is used, the speaker gives his personal position, as said before, and thus is more involved. No justification is needed, though it is sometimes given. The structure belongs to direct speech, to spoken language. The analysis which follows will only be relevant for uses of be surprised in the first person (singular or plural).

2. It would be surprising if / I would be surprised if: a conditional construction Though the type of structure described here is not a prototypical example of a conditional sentence such as , it contains some basic features of the prototypical conditional construction: –

–

A relation of entailment between p and q is asserted, q being the matrix clause containing surprising/surprised. In other words, if p is true3, q is necessarily true as well. Though the clause order in sentences expressing an assessment is very regularly q if p, what comes first logically is p. The relation of entailment is first established by the use of the conjunction if (its other function being to introduce a hypothesis). Will expresses that the feeling is subsequent to the event but this order is logical as well as temporal: surprise appears as a reaction to the event, so it both comes after and is caused by it. Thus the modal, and not only if, expresses the logical link of consequence from the event (external cause) to the feeling. This value is not created by the specific association of an event and a feeling but is part of the possible meaning of will as has been shown in different studies4.

226 Christiane Rocq-Migette –

The point of view of the speaker concerning the possibilities of actualisation of p and consequently of q is marked by the tenses employed. Modal remoteness, which expresses the presupposition5 of the speaker concerning his hypothesis, seems to be a characteristic feature of the structure: the past tense is used for the modal (would) in the matrix and also for the verb in the if-clause in standard examples: the event and the reaction to it are thus presented as contrary to the speaker’s expectations. However, a negation can be used in the if- clause, thus changing ‘contrary to the speaker’s expectation’ into ‘in keeping with’ the speaker’s expectation’. The negation can also occur in the matrix clause and theoretically the presupposition should be different. I would be surprised if… not is not the same as I would not be surprised if …

3. Possible combinations 3.1. I would be surprised if / It would be surprising if + positive polarity in the dependent clause (5) Only 44, and with 16 years to go before the official NatWest retiring age, this is, he agrees, unlikely to be his last job. ‘How long it is right for both me and the bank to do the job is not something we are considering at present,’ he says, ‘but I would be surprised if I was still here in 16 years’ time. (Daily Telegraph,1992, Business section, BNC) (6)

It is not possible to say that these families were typical of all Caribbean families in London. However, it would be surprising if their behaviour were completely atypical. (London Jamaican, Sebba Mark, 1993, Longman, BNC)

In both examples, the hypotheses, in (5) and in (6), are presented as being contrary to expectation through the use of the past tense form. Thus what is meant is I don’t expect to be here in 16 years’ time and it is not likely for their behaviour to be completely atypical. Example (6) is another example of the use of would be surprising in an argumentative context in which the speaker wants to make a point. What is being discussed is typi-

Surprised / surprising and epistemic modality

227

cality, atypicality. In the if-clause if their behaviour were completely atypical, the modal remoteness expressed by were bears on the adverb completely; not completely atypical is the expected feature; with the use of the negative adjective, this is a way of saying that a degree of typicality is expected. Though there is no syntactic negation in the subordinate clause, this example shares the characteristics of the structures studied in what follows.

3.2. I would be surprised if … not / It would be surprising if…not (7)

I would be surprised if it wasn’t still possible to go through ten years of education with only the faintest idea of what has been happening in the world, even your own country. (So very English, Rowe Martha, 1990, BNC)

(8) Considering that the body depends on sensitive electrical interactions to function, it would be surprising if these diverse electrical frequencies did not have some effect on us, whether detrimental or otherwise. (Homeopathy for everyone, Gibson R. and S. Gibson, 1987, Penguin, BNC) Here, the hypothesis combines modal remoteness with negation, the result being that it is felt to be according to expectation: What is meant in (7) is I expect it to be still possible… and in (8) it is likely for these diverse electrical frequencies to have some effect on us. The positive expectation is shown through the use of some in (8). The combination of modal remoteness with negation is similar to the use of a double negation leading to an assertion.

3.3. I would not be surprised if / It would not/hardly be surprising if (9) Secretary Janet Pike, 25, of Gateshead, said: ‘I’m a big Daniel DayLewis fan and I wouldn’t be surprised if this made him a Hollywood heart-throb.’ (Today, 1992, BNC)

228 Christiane Rocq-Migette (10) It would hardly be surprising if so many organisations, each with a finger in a large but nonetheless limited pie, sometimes disagreed on how the money should be spent. (Management Today, 1991, BNC) With these matrix clauses and the use of modal remoteness, contradictory forces are at play: The past tense form indicates that the hypothesis is unlikely, but the negation in the matrix then expresses that it is not contrary to expectation. The result is that the literal meaning is it is possible that, which seems somewhat weaker than the structures which were analysed previously. This raises the question of a scale in the expression of degrees of expectation, but in discourse the difference can be felt to be slight6. This is shown by the equivalent given for I wouldn’t be surprised if he changed his mind by the Longman Dictionary of English Language and Culture, already mentioned in the introduction and repeated here: he will probably change his mind. He will possibly change his mind would be closer to the literal meaning.

4. From reference to a feeling to a prevailing epistemic interpretation The possibility of having an epistemic interpretation for surprised (feeling surprise) and surprising (causing surprise) is due to the meaning of the referent: the concept to which surprise refers belongs to the category of emotions and feelings; but contrary to happiness or sadness which are purely emotive, surprise is created by something which is external to the speaker. I am surprised cannot be used without a situational or textual context giving the source of the feeling. Thus we have two possible aspects of evaluation in the meaning of the lexical items surprised and surprising. Which aspect prevails depends on the status of the event presented as causing surprise. This can be seen if we compare factive examples and counterfactual examples on the one hand with, on the other hand, contraryto-expectation presuppositions which were analysed previously. (11)

Where the Elgin marbles are concerned, I am surprised that the present inhabitants of the Greek mainland want anything to do with them. (The Daily Telegraph 1992, World affairs material, BNC)

Surprised / surprising and epistemic modality

(12)

229

To put this another way, we can predict what is going to happen in most of the situations in which we find ourselves because we have expectations that certain rules will be followed. For example, we would be surprised if the police did not wear a particular uniform or if a traffic warden did not book us for parking on a yellow line. (Introductory Sociology, Sheard K. 1992, BNC)

With that-clauses or with contrary-to-fact presupposition, the status of the event is not a subject of discussion (or at least it is not presented as such). It is felt to be part of the shared knowledge of the speaker and his addressees: in (11), Greek people are known to want the Elgin Marbles back; in (12), the police are known to wear a particular uniform and the traffic wardens to book us for parking on a yellow line. The emotive aspect is then more relevant. With if-clauses and presupposition expressing ‘contrary to expectation’, the status of the event, as assessed by the speaker, comes to the foreground. The prediction feature becomes central in the information which is imparted. The epistemic aspect is more relevant this time. Cotext lends support to this idea; an argumentative environment has already been mentioned for surprising; and when I would be surprised if is used, such words such as likely / unlikely can sometimes be found in the linguistic environment (see example [5]).7

5. Problems This seems to be an elaborate but consistent system. However, surprising phenomena can be discovered when a large corpus such as the BNC is studied:

5.1. Discrepancies in the use of tenses Examples in which the if-clause contains a present tense rather than a past tense (though would is used in the matrix) are too frequent to be considered as marginal. However, the examples are far more frequent with I would be surprised if than with it would be surprising if: 46 examples out of 114 for the first compared with 6 out of 60 for the second.8.

230 Christiane Rocq-Migette (13)

I would be surprised if wheatear, ring ouzel, chiff chaff and possibly Sandwich tern and sand martin are not reported somewhere in our region by next weekend. (Northern Echo, Applied Science pages, BNC)

(14)

I’d be very surprised if he manages to keep to the 1.5 per cent increase in public sector pay. (Today, 1992, BNC)

(15)

It would be surprising if benchmarking attracts as much media attention in five years’ time as it does now. But it is undoubtedly a useful management tool and, in a competitive economy, it always will be. (Accountancy, 1992, BNC)

It seems to be essentially a feature of spoken English, found in most cases when there is an explicit reference to the future: in (13), by next week-end, and in (15), in five years’ time; in (14), the reference to the future is not so explicit but is implied by keep to. Possibly, as the whole sentence expresses the expectation of the speaker (whether it is positive or negative as regards the event which is referred to), the present tense of open hypotheses is then used. We would then have a kind of notional concord, if we borrow the concept which Quirk et al. (1985:757-8) use for Subject-Verb concord when there is a discrepancy for semantic reasons9. Here the discrepancy lies in the syntactic concord of verbs in a matrix clause and its subordinate. Not contradictory with what precedes is the hypothesis that we are faced with a progressive lexicalisation of I would (not) be surprised if, that is to say a change from a free conditional structure to a set phrase. I would (not) be surprised if… then would become a lexical item simply meaning I think this will not happen, or I think this will happen with possible variants in the use of tenses. The phenomenon could also be explained by non-standard usage, but many occurrences take place in perfectly standard contexts. Another element to be taken into account is that this discord in the use of tenses more generally occurs with appreciative conditional structures such as it would be better if (and not in prototypical uses10): (16)

I think it would be better if you’re there. ( borrowed from Ford C. and S. Thompson, 1986: 36811)

Surprised / surprising and epistemic modality

231

5.2. Double negation: would not be surprised if …not Huddleston & Pullum (2002: 845) list these occurrences among the standard variety of pleonastic negation. (17) below is borrowed from their examples: (17)

I wouldn’t be surprised if it didn’t rain

(18)

I shouldn’t be at all surprised if he didn’t fancy me, thought Jenny. Perhaps not though, I seem to fancy every man I meet fancies me at the moment. (A Clubbable Woman, Hill, R. 1987, BNC)

Rain is expected in (17) and ‘he’ is expected to fancy ‘Jenny’ in (18), as is shown by the following sentence: Perhaps not though (=Perhaps he did not fancy me, though), which is expressed to contradict I shouldn’t be at all surprised if he didn’t fancy me. Thus the events referred to with a negative subordinate clause are expected here just as the events were expected in examples (9) and (10) with the same matrix but no negation this time in the subordinate clause. The trouble is that we can have a negation inside the if-clause which is not pleonastic. The use of double negation is then a source of ambiguity and context is necessary to understand what is meant: (19)

‘Who’s gone, Mr Multhrop?’ asked Rose, suddenly aware that the hotel seemed remarkably empty. ‘Lionisers. Forty of them have booked out. And now all the others have disappeared. I shouldn’t be surprised if we don’t see them again. Just send for my things, they’ll say. It’s happened before.’ (Murder makes an entre, Myers A., 1992, BNC)

(20)

When James asked his brother Nicholas for a turn on his new bike, Nicholas got off and helped James on to the bike – but his mother made no comment, and James rode off without a word of thanks. It would not be surprising if Nicholas didn’t share his things next time round. (Discipline; a Positive Guide for parents, Herbert M. 1989, BNC)

In (19) I shouldn’t be surprised if we don’t see them again means that the absent guests are not expected again. The epistemic modality then bears on

232 Christiane Rocq-Migette a negated event: <we – not see them again>. In (20) ‘Nicholas’, a boy who was not thanked at all for sharing a toy on one occasion, is not expected to share again. The scope of the modality expressed through It would not be surprising is a negated event again, . Though we have the same forms as in (17) and (18) the interpretation of the polarity is the reverse. No pleonastic use of the negation similar to the one found in examples (17) and (18) has been found with would not be surprising so far. Since it seems to be quite common with I wouldn’t be surprised if, it could suggest that this is another possible feature specific to spoken language.

6. Conclusion Though this article has been centered on the use of would be surprised if / would be surprising if to express epistemic modality, not all occurences of these structures can be analysed in this way: the event which is referred to in the subordinate clause has to be non-certain. When this condition is not met, conditional sentences still express a feeling instead of possibility. The meaning brought about by the combination of modal remoteness and negation raises a first question about their semantic relation, and a second one about the possibilities it offers to express varying degrees of probability. The two structures are not equivalent: the syntactic differences linked with the opposition -ing/-ed give the possibility of expressing different standpoints and are used for different purposes by the speaker: argumentation or expression of a personal point of view. A close study of a corpus also brings out specificities of spoken language in the use of tenses and pleonastic negation.

Notes 1. Should sometimes occurs as a variant for the first person: see examples (18) and (19). 2. Not occurs in its full form or in its weak form. The difference in use, if any, has not been studied here. 3. ‘True’ or ‘not true’ only refers to what is presented as such by the speaker. 4. The logical value of will was highlighted by Binnick (1971, 1972) when he compared the uses of will and be going to. In many examples, will is used when

Surprised / surprising and epistemic modality

233

there is "a conditioning factor"; thus "a more contingent future" is expressed with the modal auxiliary (1971: 41-42). In the second part of his study, Binnick (1972: 5) writes: " Given a future clause with will and some clause which completes the sentence containing that future clause, there is a logical relation, which …we have characterized as causal." Palmer (1979) also shows that will can have a conditional interpretation when he compares 'Don't sit on that rock. It'll fall.' with 'Don't sit on that rock. It's going to fall.': "The first sentence would always be taken to mean that it will fall if you sit on it; the likely interpretation for the second is that it is going to fall anyway." (1979: 112-113). Larreya (1984: 188-192, 207-212) emphasizes the relation of entailment which will helps to establish. Most of the time, this relation is closely linked with the expression of futurity, but then the future is relative to the hypothesis and not to the time of speech. However, will can also be used for a purely logical link: comparing 'If it is fine, he is at the seaside now.', with 'If it was fine, he would be at the seaside now.', Larreya stresses that the lexeme WILL used in the second example in association with the preterit cannot express temporal reference; but he disagrees with the idea that it could only be a 'dummy' to carry the preterit marking unreality since will can be inserted in the apodosis expressing the consequence of an open hypothesis, for instance in 'If he has a British passport, he is/will be British.' Will here clearly expresses a logical link. This is not an exhaustive review of all the studies setting out this analysis of will. 5. ‘Presupposition’ is used here as referring to a linguistic concept: “Linguistic presupposition is both a category of meaning and a category of form. It has specific linguistic forms (…)” (Larreya 2003: 22). 6. There seems to be some analogy with negation raising, which can occur with non-factive verbs such as believe or think (Horn 1986). 7. Such expressions as likely, unlikely, probably, presumably, I am sure…, were found in the immediate context in 15 out of 114 examples with would (not) be surprised. In the remaining examples, the prediction feature concerning the possibility or probability of the event is also present but implied instead of being explicit. 8. No statistical study is aimed at here, but these figures show that this use is not marginal at all for would (not) be surprised. 9. In such cases, the noun group is singular but refers to several elements in the extra-linguistic world: decade for example can be found either with syntactic concord or notional concord. The choice depends then on what the speaker has in mind. The reverse phenomenon is also possible: Ten dollars is all I have left. 10. A discrepancy in the presuppositions expressed in the protasis and apodosis through verbal markers only occurs in specific cases, such as a tentative polite use of the modal past which does not necessarily occur in both clauses (e.g:If it is necessary to convey a message, I would ask you to do so through a mes-

234 Christiane Rocq-Migette senger, Ishiguro). There may also be two hypotheses, the hypothetical one which triggers the use of would being implied by the preceding cotext (e.g. But we have to tell. If they never find who did it, if he kills again, we’d never forgive ourselves. – P.D. James) In this case, there is in fact no discrepancy, the underlying hypothesis for the second sentence being If we didn’t tell (RocqMigette 1997: 65-66) 11. Ford and Thompson did not discuss the problem of tenses in this article though, but clause order.

References Binnick, Robert I. 1971 Will and Be going to. In Papers from the 7th regional Meeting, Chicago Linguistic Society, 40–52. 1972 Will and Be going to II. In Papers from the 8th regional Meeting, Chicago Linguistic Society, 3–9. Ford, Cecilia E., and Sandra A. Thompson 1986 Conditionals in discourse: a text–based study from English. In On Conditionals, Elizabeth C. Traugott (ed.), 353–372. Cambridge: Cambridge University Press. Horn, Laurence, R. 1989 A Natural History of Negation. Chicago, Il.: The University of Chicago Press. Huddleston, Rodney, and Geoffrey K Pullum 2002 The Cambridge Grammar of the English Language, Cambridge: Cambridge University Press. Larreya, Paul 1984 Le possible et le nécessaire, modalités et auxiliaires modaux en anglais britannique. Paris: Nathan. 2003 Irrealis, past time reference and modality. In Roberta, Facchinetti, Manfred Krug, and Frank Palmer (eds.), Modality in Contemporary English, 21–45. Berlin: Mouton de Gruyter. Palmer, Frank R. 1979 Modality and the English Modals. New York: Longman. 2nd ed. London/New York: Longman, 1990. 1986 Mood and Modality. Cambridge: Cambridge University Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman

Surprised / surprising and epistemic modality

235

Rocq-Migette, Christiane 1997 L’expression de la condition en anglais contemporain, comparaison avec quelques formes du français. Thèse de doctorat, Université Paris 13. (Published by Atelier National de Reproduction des Thèses, 1998).

Commitment and subjectivity in the discourse of a judicial inquiry Juana I. Marín-Arrese

1. Introduction Meanings expressed by linguistic elements such as epistemic modals include both content and procedural aspects (Blakemore 1987; Traugott and Dasher 2002). Epistemic modals contribute the notion of differing degrees of certainty to the meaning of the utterance/sentence, and are also indexical of the speaker/writer’s subjective attitude. Evidential qualifications also indicate speaker/writer’s subjective/intersubjective attitude as well as their assessment of the validity of the information on the basis of its evidential source, yielding different values in the degree of commitment of the speaking self towards the information proffered (Wilett 1988; Sanders and Spooren 1996; Mushin 2001; Plungian 2001, Marín-Arrese 2002, 2004 inter alia). The notion of subjectivity, since its original formulation by Benveniste (1966) as the capacity of the speaking self to view him/herself as subject of enunciation, has been amply discussed in the literature from various perspectives: the expression of self and of the speaker’s subjective belief state or attitude toward the proposition (Lyons 1977; Traugott 1995); the degree of salience of the subject of conception or some other facet of the ground, or the extent to which the information is implicitly grounded in the perspective of the speaker (Langacker 1991, 2000, 2002); and the degree to which the speaker assumes personal responsibility for the evaluation of the information or shares this responsibility with others – ‘intersubjectivity’ – (Nuyts 2001, 2002). From a rhetorical and communicative view, these modal and evidential resources involve interactional meanings between speaker/writer and actual or potential interactants. Mushin (2001: 58) makes the point that evidentiality involves not only reference to the source of information but also speaker’s ‘epistemological stance’: “Speakers are motivated to adopt a particular epistemological stance partially on the basis of their source of

238 Juana I. Marín-Arrese information, but also on the basis of their rhetorical intentions, on how they want their utterance to be understood and treated in the moment of the interaction”. White (2003: 260) has also convincingly argued that by the use of these wordings, for which he uses the term ‘heteroglossic engagement’, “the textual voice acts first-and-foremost to acknowledge, to engage with or to align itself with respect to positions which are in some way alternatives to that being advanced by the text”. The use of there resources to express speaker’s enunciational stance, the stance that a speaking person assumes in his/her ‘enunciator’ role, may reflect an experiential position (I see/hear...), or an epistemic position (I conclude/know...) (Brandt 2004: 5-6). Speaker’s basic enunciational position, the elementary deictic position which signals “the here-and-now speech act” (I have to say ...), may also be deviated to adopt a position of ‘aphony’, by which “the speaker emphatically withdraws or refrains from investing in the utterance (I am not saying that X...)” (Brandt 2004: 7). This paper aims to explore the semantic domains of epistemic modality and evidentiality, involving experiential, cognitive and communicative dimensions (Marín-Arrese 2004). The paper addresses the issue of speaker stance, viewed in terms of the experiential, epistemic or communicative position adopted by the speaker. The paper also explores the expression of speaker’s commitment to and responsibility for the textual information, through the study of the parameters of evidentiary validity and degree of certainty, or alternatively whether the speaker adopts a position of aphony, disclaiming knowledge and responsibility for the information. Subjectivity and intersubjectivity is explored in terms of the interaction of two parameters: 'salience or overtness of the role of the conceptualizer', which refers to the degree of explicitness, implicitness or opaqueness of the role of conceptualizer, and the parameter 'personal vs. shared responsibility' (Langacker 1991; Nuyts 2001; Marín-Arrese 2003). 1 This paper presents the results of a study based on the Hearing Transcripts of The Hutton Inquiry2 (August-October 2003), which investigated the circumstances surrounding the death of Dr David Kelly. In the Statement by Lord Hutton – 28 January 2004 – he sets out the basic issues which counsel addressed in their examination and cross-examination of witnesses: –

the preparation of the dossier of 24 September 2002 and the inclusion of the reference to Weapons of Mass Destruction

Commitment and subjectivity in a judicial inquiry

–

– –

239

Dr Kelly’s meeting with Mr Gilligan in the Charing Cross Hotel on 22 May 2003 and Mr Gilligan’s broadcasts on the BBC Today programme on 29 May 2003 the decisions and actions taken by the Government after Dr Kelly informed that he had spoken to Mr Gilligan on the 22 May 2003 the factors which may have led Dr Kelly to take his own life

I have examined the transcripts of some of the participants more directly involved in the events and have grouped them according to their institutional, social or private domains: Government (AB: Anthony Blair, AC: Alastair Campbell, JS: John Scarlett), BBC (AG: Andrew Gilligan, SW: Susan Watts, GH: Gavin Hewitt) and Family (JK: Janice Kelly, RK: Rachel Kelly, SP: Sarah Pape). The expression of speaker's enunciational stance and subjectivity/intersubjectivity is expected to vary depending on the different categories of participant involved in the communicative event. It is hypothesized that there will be: (a) A higher presence of expressions of experiential stance in the Family group, since the participants are more emotionally involved in the event. Conversely, the Government group and the BBC will tend to adopt an epistemic enunciational position. (b) A higher presence of negative polarity utterances (aphony) in the Government group, since this group will tend to adopt a strategy of detachment and disclaim responsibility for the events investigated; (c) A higher degree of commitment to the information in the Family group, due to their greater involvement and the fact that they have no direct or indirect responsibility in the events. (d) A higher presence of expressions of explicit personal responsibility in the Family group, associated with more involvement and investment in the utterance, whereas the Government group will tend to adopt a more intersubjective stance (shared responsibility), and the BBC a more neutral stance (more impersonal and objective position). This paper thus explores the following features in the discourse of these witnesses: (i) The identification, classification and quantification of the linguistic resources for the expression of personal evidentiality and epistemic modality in the corpus of transcripts, which will provide an indication

240 Juana I. Marín-Arrese of speaker’s enunciational stance: Experiential position associated with a more emotional stance or Epistemic position typically perceived as more logical and objective, or the degree to which the speaker invests or refrains from investing in the utterance (‘aphony’, Brandt 2004). (ii) The similarities and differences in the presence and patterning of the parameters of speaker commitment (evidentiary validity and estimations of certainty) in the discourse of the different categories of participants: Family, BBC, Government. (iii) The similarities or differences in the parameters of subjectivity: Degree of salience of the subject of conception and explicitness of speaker role; degree to which the speaker assumes personal responsibility for the evaluation of the evidence (subjectivity) or whether the assessment is ‘potentially’ shared by others (intersubjectivity). Section 2 addresses the description of the parameters of speaker commitment. In section 3, I discuss the notions of subjectivity and intersubjectivity. Section 4 presents the results of the corpus study, and the conclusions are provided in the final section.

2. Speaker commitment 2.1. Evidentiality and modality The borderline between the domains of evidentiality and epistemic modality is not always clear in the literature. As Dendale and Tasmowski (2001: 341-2) note, three relations can be found in recent studies: “disjunction (where they are conceptually distinguished from each other), inclusion (where one is regarded as falling within the semantic scope of the other), and overlap (where they partly intersect)”. The first position is illustrated by the notion of evidentiality as restricted to the identification of the source and means whereby information is available to the speaker/writer. De Haan (1999:85) argues that there is a semantic distinction between the two domains: Epistemic modality EVALUATES evidence and on this evaluation assigns a confidence measure to the speaker’s utterance. [...] An evidential ASSERTS that there is evidence for the speaker’s utterance but does not interpret the evidence in any way.

Commitment and subjectivity in a judicial inquiry

241

A broader conception of evidentiality includes both the source of information and an estimation of its reliability (Chafe 1986). For some, evidentiality is regarded as the superordinate category (Matlock 1989). Others, like Palmer (2001), consider that evidentiality should be subsumed under the domain of modality. Evidential together with epistemic modality would represent two subsystems of propositional modality. The position that the two domains partly intersect is held by van der Auwera and Plungian (1998), who note that the interface between the two domains is that of inferential evidentiality and epistemic necessity. In cross-linguistic terms, inferential evidentials involving conclusional processes (equivalent to English must) are considered a case of epistemic necessity. Plungian (2001: 354) also notes that epistemic modality is the domain where there is overlap between evidential and modal values, and notes: While an evidential supplement can always be seen in an epistemic marker, the opposite does not always hold: not all evidential markers are modal in that they do not all necessarily imply an epistemic judgement.

Further arguments are those proposed by Fitneva (2001), who argues for a distinction between ‘speaker-attitude markers’, and ‘source-of-information markers’, which involve a ‘co-constructed’ (Duranti 1993) evaluation of the information, allowing the hearer to assess its validity. Thus, while in both cases the reliability of the information is qualified, their use and purpose differs: Speaker attitude is used when the speaker is in a position to competently decide on the reliability of information; source of information is used when the speaker-hearer dyad is better off to negotiate the reliability of the information. (Fitneva 2001:405)

Following Fitneva (2001) and Plungian (2001), I maintain the distinction between speaker-oriented reliability judgements expressed by epistemic modal markers, and assessments of the validity of the information coded by evidential markers. Modals indicate the speaker’s subjective estimations of the reliability of the information, while leaving speaker perspective implicit. Evidentials may explicitly indicate that the information proffered is presented from a particular perspective (typically the speaker/writer and/or hearer/reader in the case of personal evidence), but may also opaquely involve both speaker and hearer.

242 Juana I. Marín-Arrese 2.2. Epistemic modality and degree of reliability Epistemic modality concerns speaker’s assessment of the communicated proposition and his/her attribution of different degrees of certainty regarding its truth (van der Auwera and Plungian 1998, Palmer 2001 inter alia): inferred certainty or necessity (must, cannot); probability (will, would, should); possibility (may, might, could). As in previous studies, I will distinguish three points on the scale of reliability: high, medium and low. I have grouped linguistic resources found in the corpus into two categories: modal auxiliaries and modal markers, which include predicative adjectives, sentence adverbs and some nominal expressions. (a) High certainty: Necessity Modals: must, cannot. Modal markers: bound to, certain, certainly, (high) chance, sure, for sure, surely (b) Medium certainty: Probability Modals: should, will, would, ought Modal markers: likely, likelihood, probably, probable, probability. (c) Low certainty: Possibility Modals: may, might, could Modal markers: maybe, perhaps, possibly, possible, possibility, unlikely. Negative expressions indicating aphony include: I am not sure, We are not sure, I am not certain, ... Gotti (2003: 286) observes that must suggests that there is only one possible conclusion drawn from the observable facts, whereas will seems to indicate a possible conclusion or a reasonable explanation drawn from “generally known facts”. In similar terms, Ward et al. (2003: 75) hold that “epistemic would conventionally implicates that the speaker believes she or he has conclusive objective (that is, empirical or logical) evidence for the truth of the proposition encoded in the utterance”. Some modal adjectives and adverbs seem to refer to conclusional processes (bound to), whereas others tend to express speaker’s confidence based on personal knowledge or common sense (I am sure, certainly).

Commitment and subjectivity in a judicial inquiry

243

(1)

No, I mean, as I say, he brought it up. He commented on it. So he [[must]] have believed that he was in a position to comment on it. (AG)

(2)

This you [[will]] not be able to read but I can: “Material that was put in during the final week.” (GH)

(3)

That was consistent with our overall objective. So they would have felt – they [[would]] have been conscious of that search, as it were, for further releasable information. (JS)

(4)

I am pretty [[sure]] it was in the Charing Cross Hotel. I am pretty sure all my meetings with him were. I certainly know the last two were. (AG)

(5)

That is [[probably]] correct because I would have got him on the mobile number. (GH)

(6)

Not that he was definitely, but that there was a much higher [[chance]] or that there was a very serious chance that he was. (JS)

In the realm of possibility, the distinction between conclusions based on evidence or estimations based on general knowledge or experience is neutralized or blurred, since there is little difference between “absence of real confidence” and “absence of good grounds for a conclusion” (Palmer 2001: 34-35). (7)

I think certainly it came to Jonathan and I [[may]] have scanned my eye over it myself, but I cannot absolutely recall that. (AB)

(8)

Well, it seemed possible, [[maybe]] more than possible that it did not tell the full story. (JS)

Langacker (1991) argues that the notions of potency and ‘evolutionary momentum’ – the tendency for reality to evolve along specific preferential paths – are crucially involved in modality. He notes that the locus of potency lies in the way the speaker, as primary conceptualizer assesses ‘the world and its evolution’: “given how reality has been evolving up through the present, what is the likelihood of it continuing to evolve in such a way

244 Juana I. Marín-Arrese as to ‘reach’ the target process?” (Langacker 2000: 309). With respect to the characterization of the epistemic senses of modals, a distinction is made between ‘projected reality’, in which “the future course of reality can be projected with considerable confidence”, and ‘potential reality’, where there is nothing to bar reality from following alternative paths from those which are preferential in terms of its predictable evolution (Langacker 1991: 277-78). Langacker (1991: 275) also makes a distinction between modal uses that “pertain to the future”, and others that refer to “present situations that are not yet part of known reality”. In its epistemic value, must is “limited to a present-time construal”, and situates the designated process in projected ‘known reality’. The notion of potency is here interpreted as the force of evidence which leads the speaker to the most likely conclusion. Epistemic may allows for both present and future uses. In its present-time epistemic use, the designated process is situated in potential ‘known reality’. The key notion is “the speaker’s knowledge of present reality”: modal may underscores the notion that the speaker does not yet know whether a particular situation is part of reality or non-reality. In his characterization of future-time epistemic senses, Langacker (1991: 278) notes that may situates “the designated process in the realm of potential reality”, that is, in view of the speaker’s present conception of reality there is nothing to impede its evolution such that it will lead to the occurrence of the process. Conversely, “will places the designated process in the realm of projected reality”; unless some unforeseen factor mediates, reality will follow its predictable path such that the process will take place. In cases where will is used to refer to a present situation, there is a subtle contrast between must and will, which derives from this distinction between present and future time systems, in that “must conveys immediacy, whereas will suggests a certain epistemic distance”. Modal will implies that the prediction based on present knowledge might be altered by new information, whereas must indicates that “confirmation is regarded as virtually inevitable” (Langacker 1991: 280). The status of will and would with regard to their epistemic values is controversial. Palmer (1990: 57) notes that “where there is reference to a future action, it is difficult and sometimes impossible to distinguish epistemic will from the will of futurity”. In the case of would the paraphrase might be “It would be reasonable to conclude that...” (Palmer 1990: 58). Whereas epistemic possibility may in both its present and future senses refers to potential reality, will (and would) refers to projected reality. In

Commitment and subjectivity in a judicial inquiry

245

this study I have limited my selection to those instances of will involving a “central-epistemic use”, where the value is “more a matter of assumption, very often with a suggestion of future confirmation” (Huddleston and Pullum 2002: 189). Most of the instances of epistemic would found in the corpus involved assumptions about past events, so in that sense we might say that, in a similar way to must, it situates the designated process in projected ‘known reality’, though suggesting a greater epistemic distance.

2.3. Evidentiality and evidentiary validity The evidential system concerns the specification of the source of evidence available to the speaker, and, in some systems, also an indication of his/her attitude towards the reliability of the information. Different subdivisions of the domain of evidentiality can be found in the literature (Chafe 1986; Willett 1988; inter alia). Plungian (2001: 352ff) distinguishes between the parameters “direct vs. indirect” sources of information that the speaker has for P (P being the described situation), and “speaker’s personal involvement” (personal vs. mediated evidence), and makes the following threeway distinction: (a) Direct evidence, which includes visual (“P, and I see/saw P”) and sensoric (“P, and I perceive(d) P”) evidence, as well as an endophoric category to express “speaker’s inner state” (“P, and I feel(felt) P”, e.g. I know); (b) Reflected evidence, that is, personal indirect access to the evidence through “synchronous inference” (“P, because I can observe some signs that P”), “retrospective inference” (“P, because I can observe some traces of P”), or “reasoning” (“P, because I know Q, and I know that Q entails P”); (c) Mediated evidence, indirect access to the information through some other person’s statements or beliefs: Quotative (“P, because I was told that P”). Distinctions between sources of evidence should not be interpreted in absolute terms, as Fitneva (2001:406) points out: Perceptual information is processed and molded in the mind, so there is some cognitive or inferential process involved. Conversely, inferential information starts with a perceptual stimulus. A better way to think about the linguistic markers for source of information is that they emphasize an aspect of the acquisition of the information. There is a fuzzy boundary between perception and cognition that might or might not be codified in language.

Evidential markers reflect different degrees of speaker commitment towards the information expressed. Perceptual evidence has traditionally

246 Juana I. Marín-Arrese been associated with a higher degree of validity, whereas evidence based on reasoning suggests a lower degree of validity, reportive evidence being the least valid (Willet 1988). Mushin (2001: 66) notes that ‘personal experience epistemological stance’ (I notice, I see, ...) always tends to imply a high degree of confidence in the validity of the information. However, the adoption of an ‘inferential epistemological stance’ may reflect different degrees of speaker commitment regarding the validity of the information depending on the process involved (That means vs. It seems). A distinction should also be made between speaker’s degree of confidence and estimations of validity which might be intersubjectively shared. The validity of direct personal evidence might be accepted and go unchallenged by the hearer/reader to a greater or lesser extent, depending on the context (e.g. the validity of the speaker as expert source, etc.). As will be seen below, validity is here seen as the syncretism of values in the parameters ‘source of evidence’ (experiential, cognitive, communicative’) and ‘directly accessed vs. indirectly inferred’ information (Marín Arrese 2004). Also relevant in the estimations of validity, from a discourse perspective, is the cooccurrence of verbal and non-verbal markers in the utterance, which in some cases increase speaker’s estimations of the degree of validity, and in other cases function as hedges. A significant parameter of evidentiary validity also concerns the choice of polarity. As noted above, negative polarity utterances, according to Brandt (2004: 7), reflect an enunciative position of ‘aphony’, whereby the speaker expresses non-commitment (I do not know, I cannot recall, ..). Negatively asserted information is thus flawed in evidentiary validity, since the speaker expressedly refrains from assigning any validity to the information. (9)

No, I do [[not recall]] being given a name at that time. I [[cannot recall]] when I first heard the name. I mean, it may have been in these telephone conversations. It may not have been. (AB) 3

The following domains will be distinguished, on the basis of what aspect is emphasized in the acquisition of the information, and/or how the speaker conceptualizes her/his access to the information, including cases of metaphorical meaning-shift (cf. Plungian 2001):

Commitment and subjectivity in a judicial inquiry

Personal Direct Experiential I see Cognitive I think Communicative I state

Personal Indirect It seems I conclude That suggests

247

Mediated They see They believe They say

Figure 1. Parameters of Evidentiality

Since the aim of this paper is to explore the expression of speaker commitment and subjectivity in discourse, I have restricted my study to evidential markers of ‘personal involvement’ (Plungian 2001). The study of mediated evidentiality, the Quotative, is beyond the scope of the present paper. Experiential stance: This category includes markers that emphasize ‘the perceptual aspect’ of the acquisition of the information (Fitneva 2001), indicating that the speaker has direct personal sensory access to the evidence, or that the evidence is experientially available to her/himself and (potentially) also to the addressee (feel, find, hear, notice, see,...). (10) On the 45 minutes specifically, I also [[saw]] that the language changed a bit. (AG) (11) I felt during the conversation that I was not going to have too much time with him. I got the [[sense]] he was on the move in New York. (GH) Indirect perceptual markers emphasize the inferential process in the acquisition of information on the basis of observable results; that is, “the evidence is presented as a sign, or a direct proof, for the claim” (Sanders 1999: 478) (Appear, indicate, look (as if/like), seem, show, sound (like), ...). (12) They are senior Parliamentarians, they are experienced people, all of them, and they had [[shown]] quite clearly, for example over the report they did on the Bali bombing, they were perfectly prepared to be independent. (AB)

248 Juana I. Marín-Arrese (13) He was mentioned in the headlines. I cannot remember in exactly what terms. I felt that it [[sounded]] as though he was not being rubbished, that he was not being sort of held to be a guilty party. (SP) (14) We had seen things appearing about it in the papers, in February of 2002 they were going to publish a dossier. After about six weeks it [[seemed]] to transpire they were not going to publish it then after all. (AG) Some cases involve metaphorical meaning shift from perception to speaker’s knowledge and beliefs. In languages like English, in which evidentiality is realized implicitly by lexical verbs, we often find, in the diachronic development of perception verbs, semantic extension to cognitive evidential uses. Matlock (1989: 216) notes that the metaphor KNOWING IS SEEING motivates and structures these meaning-shifts, on the basis of “speaker’s deduction based on the perception of the end results of an event”. Other such structuring metaphors are: BELIEVING IS FEELING, BELIEVING IS FINDING. (15) So I [[saw]] this very much as a list of points from him entirely up to me to respond to or not as I saw fit. (JS) (16) I wanted to say, contrary to some of the interpretations that were put on my evidence of yesterday in today’s newspapers, I [[feel]] that there were significant differences between what Dr Kelly said to me and what Andrew Gilligan has reported that Dr Kelly said to him. (SW) Within this category, the following scale of validity has been distinguished: (a) High: demonstrate, find, glimpse, hear, note, notice, overhear, recognise, see, show. (b) Medium: appear, indicate, fear, feel, form (the view), look, reflect, seem, sound, struck (me), take it, take out of, view. (c) Low: get the feeling, get/give/have the impression, sense, transpire.

Commitment and subjectivity in a judicial inquiry

249

Negative expressions indicating aphony include: I was not aware, I did not see, It did not look, I did not hear, ... Cognitive stance: The markers included here underscore the cognitive basis of the information proffered by the speaker (‘endophoric’ markers, Plungian 2001). They include predicates of belief and/or general knowledge where “the speaker voices personal views belonging in the realm of strictly individual experiences or attitudes” (Nuyts 2001:122) (Believe, doubt, imagine, know, remember, think). (17) My brother replies: “I have only met Susan Watts on one occasion, which was not on a one-to-one basis, it was at the end of a public presentation.” I [[know]] from conversations that we have had in the past that he very much used the word to mean what it meant. (SP) (18) That was just after the first Gulf War I [[believe]]. (JK) Indirect cognitive predicates underscore access to the information as a result of inferential processes (generalization, deduction, ...). This is parallel to what Fitneva (2001:409) terms ‘conclusional’ information. In these cases, “the relation between claim and evidence is established indirectly, via a reason: a relation of consequence” (Sanders 1999: 478) (Assume, conclude, presume, ...). (19) I was not (aware), but I, you know, would have thought it perfectly natural that the MoD had to prepare to field inquiries. I [[assume]] they had been doing that for several days. (AB) (20) I [[presume]] it must have been. (AC) (21) I think it was David that told me that, yes, but I cannot be absolutely sure. I [[suppose]] it could have been Jonathan but I think it was David. (AB) The scale of validity is the following: (a) High: conclude, decide, (not) doubt, know, learn, realise, recall, recollect, remember.

250 Juana I. Marín-Arrese (b) Medium: assume, believe, consider, expect, judge, (make) the assumption, presume, think. (c) Low: doubt, guess, imagine, occur (to me), suppose, suspect. Negative expressions indicating aphony include: I do not know, I cannot remember, I do not recall, ... Communicative stance: In this category I have included examples of performative use (I have to say...). These are cases which explicitly indicate speaker’s basic enunciational position (the “here-and-now speech act”, Brandt 2004: 7). Also included are expressions of attribution of the information to self (I said, I suggested, ...). (22) But again I must [[stress]] it was he who brought up the 45 minutes. (AG) (23) Well, I have [[explained]] in my witness statement that I had an absolute duty of confidentiality to Dr Kelly. (AG) Examples involving meaning-shifts from the domain of verbal communication to the domains of knowledge and belief (i.e. BELIEVING IS SAYING) or to the domain of perception (i.e. PERCEIVING IS TELLING) have also been found. In most cases they involve inferential conclusional meanings (i.e. That says something, suggests, tells us, ...). (24) Other explanations, for example, might be that the Government simply had not noticed my reports, which [[means]] that fewer people maybe listen to the Newsnight programme. (SW) (25) Well, that would [[suggest]] that. I am not aware that I replied to that e-mail. (AC) (26) Not specifically, no. Again, I [[would say]] that he was passing information to me that was not sensitive in any way, not operational information. (SW) The scale of validity in this category is the following: (a) High: agree, (would certainly) say, (must) stress.

Commitment and subjectivity in a judicial inquiry

251

(b) Medium: interpret, mean, read, suggest, (carry the) suggestion, tell, understand. (c) Low: (would) characterise, (should/would/could) say, (difficult to) tell. Negative expressions indicating aphony include: I cannot say, I am not suggesting, you can never tell, I would not say, ... Evidential non-verbal: These markers also make reference to experiential (clearly), cognitive (no doubt) and communicative (categoric) processes. In this category, I have included sentence adverbs (obviously), predicative adjectives (it is clear), and relational and existential constructions involving nominals (my knowledge is, I did have suspicions, there was a sense). Again, a scale of validity is distinguished for the examples found: (a) High: aware, clear, clearly, confident, conclusion, conscious, my experience, my knowledge, no doubt, obviously, palpable, my recollection, tangible. (b) Medium: apparent, apparently, assumption, my belief, indication, my memory, perception, my thought, my understanding, my view, predictably, presumably. (c) Low: feeling, my guess, my impression, my suspicion, suspicions, sense. Negative expressions indicating aphony include: not aware, not categoric, unclear,... (27) ..., I had never known him to be as unhappy as he was then. It was [[tangible]]. (JK) (28) I did have [[suspicions]] that the dossier was not accurate because simply of the fact that – that weapons had not yet been found in Iraq, ... (AG) As mentioned above, the different modes of access to the information reflect different values on a scale of speaker commitment. Direct experiential evidence might seem to be afforded higher evidentiary validity by the hearer/reader than cognitive or communicative evidence. At the same time within each category we also find different scalar values, indirect inferen-

252 Juana I. Marín-Arrese tial uses would appear to rank lower in evidentiary validity. I would thus tentatively suggest a continuum along the following lines:

High

Medium

Low

Experiential: I saw It seems You get the feeling Cognitive: I know I think I suspect Communicative: I have to say That suggests Figure 2. Evidentiary validity

3. Subjectivity Subjectivity has been identified by Langacker (1991, 2002) as the extent to which the information is implicitly grounded in the perspective of the speaker, in contrast to objectivity, whereby the subject of conception or some other facet of the ground is explicit and salient. Traugott (1995) and Traugott and Dasher (2002:22) are concerned with ways in which the SP/W’s point of view is explicitly encoded (subjectivity), or with “the explicit, coded expression of SP/W’s attention to the image or ‘self’ of AD/R in a social or an epistemic sense” (intersubjectivity). Nuyts (2001:34) reframes the notion of subjectivity in terms of “the quality and/or the status of one’s evidence for an epistemic judgment”. He conceives the dimension of subjectivity vs. intersubjectivity as the degree to which the speaker assumes personal responsibility for the evaluation of the evidence (subjectivity) or whether the assessment is ‘potentially’ shared by others (intersubjectivity). I will here attempt to characterize subjectivity/intersubjectivity drawing on these notions.

3.1. Salience of conceptualizer role Langacker’s (1991, 2000, 2002) conception of subjectivity is explained on the basis of perceptual notions. In a situation of ‘optimal viewing arrangement’, the ‘viewer’ or subject of conception remains ‘offstage’, thus being implicit and construed as maximally subjective, whereas the entity which

Commitment and subjectivity in a judicial inquiry

253

functions as object of conception is placed ‘onstage’ and is thus salient and objectively construed. In contrast, in a situation of ‘egocentric viewing arrangement’, the viewer goes ‘onstage’, thus becoming more salient and objectified. The speaker serves the role of subject of conception and source of the predication, but may also become an object of conception as a participant within the predication. In conceptual terms, when the speaker functions solely as conceptualizer, his/her role is maximally subjective; s/he is not designated by any linguistic expression nor implicitly evoked. When explicitly mentioned, as in some evidential expressions (I suppose, We think), the speaker is maximally objectified since in addition to the role of conceptualizer, s/he is also the salient object of conceptualization. With epistemic modals, the speaker as primary conceptualizer is evoked without being explicitly mentioned, so that the role of the conceptualizer is less prominent and more subjectified. In the case of inferential evidentials (it seems, that means), the presence of the speaking subject is opaque; there is covert reference to the conceptualizer, and the role of the conceptualizer is almost maximally subjective. Langacker (2000: 350) notes that with expressions such as it seems, the conceptualizer may be “only potential” or “is construed generically or in a generalized fashion”. For Sanders and Spooren (1996), degree of subjectivity is best viewed in terms of the degree to which the speaker’s active consciousness is foregrounded (high subjectivity: I think; semisubjective: may, must; nonsubjective: It appears). Sanders (1999) observes that though the conceptualizer is objectified in evidential expressions which explicitly designate the speaker, these expressions also maximally foreground the conceptualizer’s estimations of the validity of the information, so that in effect what is expressed by the sentence is maximally subjectified. Similarly, in the case of epistemic modals, subjectivity is created by evoking the speaker’s “active consciousness” (Sanders 1999: 473); the conceptualizer is objectified to a certain extent in that s/he functions as an implicit point of reference within the predication, and what is expressed by the sentence is subjectified, though to a lesser extent. With inferential evidentials, where the speaker’s active consciousness is further backgrounded, the conceptualization seems to be more objectively construed than in the case of modals. A distinction thus needs to be made between the explicitness vs. implicitness of the role of the conceptualizer and the degree to which the information designated is presented as maximally or minimally subjective. Drawing on these notions, a continuum may be identified in the dimension

254 Juana I. Marín-Arrese of subjectivity in terms of the parameter ‘salience or overtness of the role of the conceptualizer’, from cases where the conceptualizer is part of the object of conceptualization and is thus encoded as the explicit source of the evaluation, to those where the conceptualizer is implicit and non-salient, and those where the role of the current speaker as source of the evaluation is opaque: Explicit

Implicit

Opaque

I think ...

That may ...

It seems ...

Figure 3. Salience of the role of conceptualizer

3.2. Subjective vs. intersubjective evidentiality Sanders and Spooren (1996: 246) note that whereas I-embeddings “explicitly encode the speaker/writer’s personal limitation of the validity of information” (personal responsibility), in the case of experiential evidentials (It seems), “the commitment to the validity of the information is shared or at least potentially shared by the speaker/listener and other participants” (non-subjective or intersubjective responsibility) Similarly, Nuyts (2002: 447) characterizes the dimension of subjectivity vs. intersubjectivity as involving “an indication of whether the speaker is solely responsible for the epistemic evaluation of the state of affairs – i.e., subjectivity – or, alternatively, whether she shares this evaluation with others, possibly including the hearer – i.e., intersubjectivity”. Evidential expressions not encoding the speaker may be interpreted as expressing either personal or shared responsibility. Experiential evidentials (it seems) suggest shared responsibility, whereas cognitive and communicative evidential expressions (that means, that suggests) seem to indicate that the speaker is responsible for the evaluation, though they do not rule out that the information might be ‘potentially’ shared, since the speaking subject is opaque. Nuyts (2001: 206) also holds the view that modals should be considered evidentially neutral with respect to (inter)subjectivity. He argues that the “(inter)subjective meaning element is not inherent in the modal as such, but only arises due to contextual conditions”. I would contend this view, since both must and will often have evidential nuances which might lend them an intersubjective value. And though the inferential or conclusional

Commitment and subjectivity in a judicial inquiry

255

nuance is blurred in epistemic may, this does not necessarily imply that it should be considered neutral in terms of subjectivity. The role and status of non-verbal markers regarding the dimension of subjectivity is not clear in the literature. Nuyts (2001: 206) considers that modal adverbs (possibly) are neutral in terms of their (inter)subjectivity, whereas modal adjectives (it is possible that) suggest a more intersubjective value. He also notes that predicative adjective constructions (I am sure, I am aware, ...) constitute “strongly idiomaticized constructions to express a subjective evaluation” (Nuyts 2001: 68). One might argue that evidential adverbs and adjectives (obviously, it is clear that) may be considered evidentially intersubjective, since these expressions ‘potentially’ evoke shared access to perceptual evidence, whereas modal adjectives and adverbs seem to lack this intersubjective evidential value, and are more akin to modal auxiliaries in evoking speaker’s assessments. I have accordingly included both verbal and non-verbal markers in the quantitative analysis, since they all contribute to configure speaker’s stance in terms of their responsibility for the evaluation. In the characterization of subjectivity, I have considered the interaction of two parameters: ‘salience or overtness of the role of the conceptualizer’, which refers to the degree of explicitness, implicitness or opaqueness of the role of conceptualizer; and the parameter ‘personal vs. shared responsibility’. The interaction of these parameters yields a four-fold distinction: (a) Explicit personal responsibility: The current speaker is overtly the sole source of the evaluation. Evidential predicates: I saw, I think, I would say,... Modal or evidential adjectives: I am sure, I am aware, ... Other evidential expressions: My guess is, My understanding was,... (b) Explicit shared responsibility: The speaker overtly presents the evaluation as shared with the interlocutor or third persons. Evidential predicates: We all know, as you can see,... Modal or evidential adjectives: We are sure, we are clear, ... Other evidential expressions: Our perception/view was, ... (c) Implicit personal responsibility: The speaker is solely responsible for the evaluation, s/he is the implicit subject of consciousness. Modal auxiliaries: May, will, must,... Modal adjectives and adverbs: It is possible, perhaps,... (d) Opaque personal/shared responsibility: The speaker functions as conceptualizer, but the qualification is based on evidence which is tacitly

256 Juana I. Marín-Arrese shared with others or ‘potentially’ accessible to the interlocutor or audience. Evidential predicates: That shows, It seems, It was judged, That means, That suggests,..... Evidential adjectives and adverbs: It is clear, Obviously, It is palpable,... Other evidential expressions: There was a sense, there was a feeling, That carried the suggestion, ...

4. Corpus Study 4.1. The texts I have worked with a selection of the Hearing Transcripts of The Hutton Inquiry (1 August – 13 October 2003). The witnesses, the dates and the texts selected are described in Table 1 below. The study focuses on the following dimensions, discussed in the previous two sections: (i) Speaker’s commitment to the information proffered: a. Degree of reliability b. Degree of evidentiary validity (ii) Subjectivity/Intersubjectivity: a. Explicit personal responsibility b. Explicit shared responsibility c. Implicit personal responsibility d. Opaque personal/shared responsibility The texts were examined in the first instance to identify examples of evidential and modal qualifications. An electronic search using Monoconc was carried out to ensure that all the instances present in the texts had been identified. The examples found were then analysed and tagged according to the dimensions identified above. The data were submitted to further analysis for the quantitative results. The results were tested for significance, using the Chi-Square Test, at a value of significance p< 0.05. The Null hypothesis, H0, posits that there is no relationship between the row variable

Commitment and subjectivity in a judicial inquiry

257

and the column variable (i.e. the participants and the dimensions of modality and evidentiality, or subjectivity). Table 1. The Hutton Inquiry: Private/Public Domains and Witnesses DOMAINS F A M I L Y B B C

G O V E R N M E N T TOTAL

NAME & DATE Janice Kelly 1 Sept. Rachel Kelly 1 Sept.

RELATION/POSITION/ROLE Wife

WORDS 9,148

Daughter

8,632

Sarah Pape 1 Sept. Andrew Gilligan 12 August

Sister

7,906

Susan Watts 12-13 August Gavin Hewitt 13 August Anthony Blair 28 August Alastair Campbell 19 August John Scarlett 26 August

Defence and Diplomatic Corre- 17,551 spondent of the Today programme on BBC Radio 4 Science Editor of BBC Newsnight 10,580 Special correspondent for the BBC’s 10 o’clock News Prime Minister

4,347 13,829

Prime Minister’s Director of Communications and Strategy

18,735

Chairman of the Joint Intelligence Committee (JIC) & Head of the Intelligence and Securities Secretariat in the Cabinet Office

13,567

104,295

4.2. Results and discussion Table 2 shows raw figures and ratio per thousand words of the use by different speakers of modal and evidential (experiential, cognitive, communicative) expressions, and the results of the different groups.

258 Juana I. Marín-Arrese Table 2. Modal and evidential expressions and private/public domains MOD. & EVID. JK RK SP Family

MODAL N R 72 7.87 53 6.14 61 7.71 186 7.24

EXP. N R 66 7.21 83 9.61 37 4.68 186 7.24

COG. N R 90 9.84 90 10.43 119 15.05 299 11.64

COMM. N R 11 1.20 14 1.62 31 3.92 56 2.18

TOTAL N R 239 26.12 240 27.80 248 31.37 727 28.30

AG SW GH BBC

115 51 26 192

6.55 4.82 5.98 5.91

64 55 36 155

3.65 5.20 8.28 4.77

164 9.34 73 6.90 51 11.73 288 8.87

81 38 16 135

4.61 3.59 3.68 4.15

424 217 129 770

AB AC JS Govern.

94 106 89 289

6.79 5.66 6.56 6.26

125 104 47 262

9.04 5.55 3.46 5.68

255 257 106 614

119 74 62 254

8.60 3.95 4.57 5.50

589 42.60 533 28.45 297 21.89 1,419 30.76

TOTAL MEAN R

667 6.39

603 5.78

18.44 13.72 7.81 13.31

1,201 11.51

445 4.27

24.16 20.51 29.67 23.71

2,916 27.96

The three groups4 share a preference for an epistemic enunciational position (modals+cognitive evidentials), which is consistent with the communicative event. The mean figures for evidential expressions show that there is a higher tendency for the Family group to use perceptual forms, indicating an experiential enunciational stance. This is particularly the case in the discourse of Rachel Kelly (RK), which reflects higher emotional involvement. The Family group also shows comparatively high use of epistemic modals, and of cognitive evidentials. This group has the lowest figures for communicative evidentials. The results (experiential+epistemic stance) are consistent with the private nature of the group. The Government group shows the highest ratio for the use of cognitive evidentials, which together with the relatively high ratio of epistemic modals, seem to point to a preferred epistemic enunciational position. The figures for communicative evidentials (I have to say) are also the highest in this group, which is consistent with their political role, and their aim to explicitly underscore the validity of their statements. Lastly, the BBC group shows the lowest figures for modals, and for experiential and cognitive evidentials compared to the other two groups. Nonetheless, if we take the figures for modal and

Commitment and subjectivity in a judicial inquiry

259

cognitive evidentials, we might conclude that this group is also characterized by an epistemic enunciational position. The figures for communicative evidentials are also relevant; the results for this group (epistemic+marked communicative enunciational stance) are parallel to those of the Government group. It has been observed that total commitment to truth is zero-marked in most languages, reflecting the workings of our cultural models regarding knowledge, whereby information is assumed to be true unless otherwise indicated (Matlock 1989; Bybee et al. 1994). In general terms, Susan Watts (SW) is the speaker who makes least use of modal and evidential expressions (R= 20.51); the higher proportion of zero-marking in her discourse would seem to reflect higher commitment to truth and a more objective stance. This is in sharp contrast with the discourse of Anthony Blair (AB), who uses the highest proportion of modalizations (R= 42.60), reflecting a lesser degree of commitment to truth.

4.2.1. Speaker commitment Tables 3 and 4 show results (raw figures and ratio per thousand words for the different groups) pertaining to the dimension of speaker commitment in the three groups: estimations of degree of reliability of the information and degree of evidentiary validity. Also included are the figures for negative expressions indicating lack of commitment concerning reliability (I am not certain, I am not sure) or evidentiary validity (I did not see, I cannot remember, I cannot say,...). Overall there is a constant distribution in the figures for Speaker Commitment in the three groups: speakers seem to favour expressions of low reliability and medium evidentiary validity. As regards the expression of degree of certainty5, the distribution is quite similar in the three groups. The three groups show higher figures in the category of low certainty. The figures for the Government are the closest to the mean ratio. The BBC group deviates from the mean in showing fewer instances of high and medium certainty; conversely, it shows the highest ratio for low certainty. The most tentative speaker is found in the Family group; Mrs Kelly (JK) shows the highest figures for aphony regarding the parameter of speaker's certainty (I am not sure, I am not certain).

260 Juana I. Marín-Arrese Table 3. Speaker commitment: reliability REL.

HIGH N R

MEDIUM N R

LOW N R

NEGAT. N R

TOTAL N R

JK RK SP Family

16 5 21 42

1.75 0.58 2.66 1.63

11 22 15 48

1.20 2.54 1.89 1.87

23 24 24 71

2.51 2.78 3.03 2.76

22 2 1 25

2.40 0.23 0.13 0.97

72 53 61 186

7.87 6.14 7.71 7.24

AG SW GH BBC

35 5 4 44

1.99 0.47 0.92 1.35

20 13 7 40

1.13 1.22 1.61 1.23

51 29 13 93

2.90 2.74 2.99 2.86

9 4 2 15

0.51 0.37 0.46 0.46

115 51 26 192

6.55 4.82 5.98 5.91

AB AC JS Govern.

23 27 24 74

1.66 1.44 1.77 1.60

25 31 17 73

1.81 1.65 1.25 1.58

36 43 38 117

2.60 2.29 2.80 2.53

10 5 10 25

0.72 0.27 0.74 0.54

94 106 89 289

6.80 5.66 6.56 6.26

TOTAL MEAN R

160 1.53

161 1.54

281 2.69

65 0.62

667 6.39

The figures for evidentiary validity6 in Table 4 indicate greater differences among the groups. The Family group shows ratios which are close to the mean in all the parameters, except for the use of negative expressions where average results are lower than the mean. Within the group, there is considerable difference between Sarah Pape (SP), Dr. Kelly's sister, who makes very frequent use of negation (R= 4.17), indicating lack of investment in her declarations, and Rachel Kelly (RK), his daughter, whose discourse is characterised by the expression of medium to high evidentiary validity and few instances of negation. The BBC group also shows a distribution of evidentiary validity which is close to, though slightly below, the mean, except for low evidentiary validity. Particularly noticeable is the use of expression high evidentiary validity on the part of Andrew Gilligan (AG) and Gavin Hewitt (GH), in contrast with Susan Watts (SW), who tends to use expressions of medium evidentiary validity.

Commitment and subjectivity in a judicial inquiry

261

Table 4. Speaker commitment: evidentiary valididity VALID.

HIGH N R

MEDIUM N R

N

JK RK SP Family

37 78 56 171

4.04 9.04 7.08 6.66

100 95 90 285

10.93 11.00 11.38 11.09

6 4 8 18

0.65 0.46 1.01 0.70

24 10 33 67

2.62 1.16 4.17 2.61

167 187 187 541

18.25 21.66 23.65 21.06

AG SW GH BBC

124 36 42 202

7.06 3.40 9.66 6.22

120 82 38 240

6.84 7.75 8.74 7.39

10 15 5 30

0.57 1.42 1.15 0.92

55 33 18 106

3.13 3.12 4.14 3.26

309 166 103 578

17.60 15.69 23.69 17.80

AB AC JS Govern.

169 12.22 110 5.87 64 4.72 343 7.43

232 203 87 522

16.78 10.83 6.41 11.31

16 13 5 34

1.16 0.69 0.37 0.74

78 101 52 231

5.64 5.39 3.83 5.01

495 427 208 1,130

35.80 22.79 15.33 24.50

TOTAL MEAN R

716 6.86

1,047 10.04

LOW R

82 0.79

NEGAT N R

404 3.87

TOTAL N R

2,249 21.56

The Government group presents considerable differences in the parameters of evidentiary validity. Both Anthony Blair (AB) and Alastair Campbell (AC) make extensive use of negative expressions (aphony), AB showing the highest ratio (R= 5.64). Conversely, in the discourse of AB there is frequent recourse to expressions of high evidentiary validity (obviously), which are also indexical of intersubjectivity. Particularly striking within this group are the total figures for evidentiary validity. AB’s discourse is characterised by the profuse use of evidential expressions (R= 35.80), which would indicate low commitment to truth. Within this group, there is a significant difference between the discourse of John Scarlett (JS) and that of the other two members. This is perhaps attributable to differences in role: AB and AC assume a political role, whereas the role of JS as a member of the Secret Sevices might be deemed to be more professional. JS presents the lowest ratio of expressions of evidentiary validity (R= 15.33). In this respect, it is relevant to point out that his results for speaker commitment (both reliability and evidentiary validity) are similar to those of Susan Watts (SW), whose degree of in-

262 Juana I. Marín-Arrese volvement and responsibility in the events investigated is considerably lower than that of Andrew Gilligan (AG), within the BBC group.

4.2.2. Subjectivity/intersubjectivity Table 5 presents the results for the parameters of Subjectivity7: Expressions which are indexical of explicit personal responsibility (I am sure, I saw, I think, I have to say), explicit shared responsibility (We are not sure, We noticed, As you can see, We all know, We can say), implicit personal responsibility (may, perhaps, it is possible, ...), opaque personal/shared responsibility (It seems, it was noted that, that meant, obviously, it was palpable,...). Table 5. Subjectivity/Intersubjectivity Explicit (SP) N R

Explicit (Sh.) N R

Implicit (SP) N R

N

JK RK SP Family

147 129 171 447

16.07 14.94 21.63 17.40

5 4 5 14

0.55 0.46 0.63 0.54

49 51 59 159

5.36 5.91 7.46 6.19

38 56 13 107

4.15 6.49 1.64 4.16

239 240 248 727

26.12 27.80 31.37 28.30

AG SW GH BBC

269 148 89 506

15.33 13.99 20.47 15.58

17 3 2 22

0.97 0.28 0.46 0.68

98 47 24 169

5.58 4.44 5.52 5.20

40 19 14 73

2.27 1.79 3.22 2.25

424 217 129 770

24.16 20.51 29.67 23.71

AB AC JS Govern.

381 377 199 957

27.55 20.12 14.67 20.74

53 18 10 81

3.83 0.96 0.74 1.75

82 100 72 254

5.93 5.34 5.31 5.51

73 38 16 127

5.28 2.03 1.18 2.75

589 42.60 533 28.45 297 21.89 1,419 30.76

SUBJ

TOTAL MEAN R

1,910 18.31

117 1.12

582 5.58

Opaque R

307 2.94

TOTAL N

R

2,916 27.96

Overall, the highest figures are found in the explicit expression of the role of speaker/conceptualizer, where the speaker as an objectively construed participant is given maximal salience. These results are consistent

Commitment and subjectivity in a judicial inquiry

263

with the event and activity type: testimony of witnesses in a judicial inquiry. The Family group shows the highest ratios for the use of expressions where the conceptualizer is implicit, and also those where it is opaque. The results that differ in this group are those of Sarah Pape (SP), who makes frequent recourse to explicit expressions of subjectivity (R= 21.63). The discourse of SP is thus characterised by explicit subjectivity and personal responsibility. In keeping with their professional role, the figures for explicit subjectivity and personal responsibility are on average the lowest in the BBC group, except for Gavin Hewitt (GH) (R= 20.47). This group also has the lowest ratios for implicit and opaque subjectivity. The only discordant feature is the relatively high figures for overt shared responsibility in the discourse of Andrew Gilligan (AG) (R= 0.97), reflecting a greater tendency to overtly share responsibility with other participants involved in the events. In the Government group, the overall results for explicit subjectivity, involving both personal and shared responsibility, are the highest. One should, however, recall that a high proportion correspond to negative expressions (I do not know, I do not remember). Particularly interesting in this group is the discourse of Anthony Blair (AB), which shows the highest ratio (R= 3.83) for expressions of explicit shared responsibility (we). This points to a strategy of diffusion of responsibility. Lower results, though still comparatively significant (R=0.96), in this parameter are the figures for Alastair Campbell (AC). It is relevant to observe in this respect, that the three witnesses with the highest ratios for overt shared responsibility (AB, AC and AG) are the ones who have perhaps been assigned by public opinion the greater burden of responsibility surrounding the events investigated in this inquiry. Another parameter which contributes to the mystification of personal responsibility is the use of expressions of opaque personal/shared responsibility (obviously). Again, the discourse of AB shows a ratio (R= 5.289) of use which is well above the mean.

5. Conclusion In this paper I have tried to account for the relationship between the semantic domains of evidentiality and epistemic modality and the dimensions of speaker commitment and subjectivity, in terms of a set of related parameters. Degree of speaker commitment is conceived as a complex dimension

264 Juana I. Marín-Arrese involving epistemic assessments of certainty regarding the information proffered as well as assessments of evidentiary validity. Subjectivity has been characterized both in terms of the degree of salience attached to the role of conceptualizer (explicit, implicit, opaque), and to the dimension of subjectivity-intersubjectivity involving the contrast between personal vs. shared responsibility. The corpus study based on the transcripts of The Hutton Inquiry has confirmed our hypotheses in general terms. The Family group is characterized by a more experiential stance in comparison to the other groups, which seems to reflect their higher emotional involvement. The Government group, and the BBC, on the other hand, would be characterized by an epistemic enunciational position. Regarding the dimension of speaker commitment, a relevant feature is the high presence of negative polarity utterances in the Government group, especially in the case of Anthony Blair (AB) and Alastair Campbell (AC). This clearly suggests a stance of detachment, disclaiming knowledge and responsibility for the events investigated. AB also adopts a strategy of shared responsibility, which contributes to mystify his personal responsibility in the process of decision-taking. Again the figures for low reliability judgements in the Government group evince a position of tentativeness with respect to the information proffered. With respect to the dimension of subjectivity, a higher presence of explicit subjectivity expressions was observed in the Government group, and to a lesser extent in the Family group. On average these two groups show the highest frequencies for personal responsibility, which might be associated with more personal involvement in the events. Also relevant are the results for intersubjectivity; AB shows the highest frequencies for the explicit expression of shared responsibility, as well as for opacity in the expression of responsibility, which reflects his desire to diffuse the assignment of responsibility on the part of public opinion. Within the groups, it is also interesting to comment on the specific positions of some of the members. Results have shown that the participant role that a particular member has adopted irrespective of their group domain has obvious consequences for the discourse. One clear example is the discourse of Sarah Pape (SP) (Family group), who adopts the role of an expert witness, more akin to her professional role as doctor than to her private role as Dr. Kelly’s sister. This is reflected in the values for certain parameters: high reliability expressions (R= 2.66), and expressions of explicit personal responsibility (R= 21.63), both well above the mean. An interest-

Commitment and subjectivity in a judicial inquiry

265

ing parallel is observed between John Scarlett (JS) (Government) and Susan Watts (SW) (BBC) for the various parameters: both witnesses, in keeping with their professional role, show the lowest ratios of modalized expressions in their discourse, which would seem to reflect higher commitment to truth and a more objective stance.

Notes 1.

2. 3.

4.

5.

6.

7.

This paper presents results of research funded by the Comunidad de Madrid (Research project: La variación lingüística en géneros y ámbitos discursivos en inglés y en castellanos. Ref.: 06/HSE/0272/2004). The Hearing Transcripts were accessed on the Hutton Inquiry web site www.the-hutton-inquiry.org.uk. The files are covered by crown copyright. In his discussion on transitivity and the coding of events, Langacker (1991) makes a distinction between processes pertaining to the physical domain, the mental domain, and the domain of social interaction. The Chi Square test has been calculated on the basis of results of the ratios per 100,000 words, since the number of words of the texts in the each group differed considerably. The distribution for modality and evidentiality in the three groups is significant. There is very strong evidence against the null hypothesis (Chi-square= 199.76, Df: 6, p#0.000); there is association between group and choice of stance mode. The observed values are not due to random variation. The distribution for the expression of reliability in the three groups is significant. There is strong evidence against the null hypothesis (Chi-square= 26.34, Df: 6, p#0.000); there is association between group and choice of degree of certainty. The distribution for the expression of validity in the three groups is significant. There is strong evidence against the null hypothesis (Chi-square= 91.94, Df: 6, p#0.000); there is association between group and choice of validity. The distribution for the expression of subjectivity/intersubjectivity in the three groups is significant. There is strong evidence against the null hypothesis (Chisquare= 138.58, Df: 6, p#0.000); there is association between group and choice of dimensions of subjectivity/intersubjectivity.

References Benveniste, Émile 1966 De la subjectivité dans le langage. In Problèmes de Linguistique Générale, 258–66. Paris: Gallimard. [Original edition 1958]

266 Juana I. Marín-Arrese Brandt, Per Aage 2004 Evidentiality and Enunciation. A Cognitive and Semiotic Approach. In Perspectives on Evidentiality and Modality, Juana. Marín Arrese (ed.), 3–10. Madrid: Editorial Complutense. Bybee, Joan, Perkins, Revere and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the languages of the world. Chicago: University of Chicago Press. Chafe, Wallace 1986 Evidentiality in English conversation and academic writing. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe and Johanna Nichols (eds.), 261–72. New York: Ablex. Dendale, P., and Tasmowski, L. 2001 Introduction: Evidentiality and related notions. Journal of Pragmatics 33: 339–48. Duranti, Alessandro 1993 Intentions, self, and responsibility: An essay in Samoan ethnopragmatics. In Responsibility and Evidence in Oral Discourse, Jane H. Hill, and Judith T. Irvine (eds.), 24–47. New York: Cambridge University Press. Fitneva, Stanka A. 2001 Epistemic marking and reliability judgements: Evidence from Bulgarian. Journal of Pragmatics 33: 401–20. Huddleston, Rodney and Geoffrey Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Givon, Talmy 1982 Evidentiality and epistemic space. Studies in Language 6: 23–49. Gotti, Maurizio 2003 Shall and will in contemporary English: A comparison with past uses. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank Palmer (eds.), 267–300. Berlin: Mouton de Gruyter. de Haan, Ferdinand 1999 Evidentiality and epistemic modality: Setting boundaries. Southwest Journal of Linguistics 18: 83–101. Langacker, Ronald 1991 Foundations of Cognitive Grammar, Vol. II. Stanford: Stanford University Press. 2000 Grammar and Conceptualization. Berlin: Mouton de Gruyter. 2002 Deixis and subjectivity. In Grounding: The Epistemic Footing of Deixis and Reference, Frank Brisard (ed.), 1–28. Berlin: Mouton de Gruyter.

Commitment and subjectivity in a judicial inquiry

267

Lyons, John 1977 Semantics, Vol. 2. Cambridge: Cambridge University Press. Marín–Arrese, Juana 2002 Evidence and Commitment in Discourse: A comparable corpus study. Paper presented at the III Congreso Internacional de la Asociación Española de Lingüística Cognitiva (AELCO). 15–17 May 2002, Valencia: Universidad de Valencia. 2003 Subjectivity, intersubjectivity and epistemic/evidential qualifications in newspaper discourse: A comparable corpus study. Paper presented at the 8th International Cognitive Linguistics Conference (ICLC–8). 20– 25 July 2003. Logroño: Universidad de La Rioja. 2004 Evidential and epistemic qualifications in the discourse of fact and opinion: A comparable corpus study. In Perspectives on Evidentiality and Modality, Juana Marín Arrese (ed.), 153–84. Madrid: Editorial Complutense. Matlock, Teenie 1989 Metaphor and the grammaticalization of evidentials. Berkeley Linguistics Society 15: 215–25. Mushin, Ilana 2001 Evidentiality and Epistemological Stance. Amsterdam: John Benjamins. Nuyts, Jan 2001 Epistemic Modality, Language, and Conceptualization: A Cognitive– Pragmatic Perspective. Amsterdam: Benjamins. 2002 Grounding and the system of epistemic expressions in Dutch: A cognitive–functional view. In Grounding: The Epistemic Footing of Deixis and Reference, Frank Brisard (ed.), 433–66. Berlin: Mouton de Gruyter. Palmer, F.R. 1990 Modality and the English Modals, 2nd edn.. London: Longman. 2001 Mood and Modality, 2nd ed.. Cambridge: Cambridge University Press. Plungian, Vladimir 2001 The place of evidentiality within the universal grammatical space. Journal of Pragmatics 33: 349–357. Sanders, José 1999 Degree of subjectivity in epistemic modals and perspective representation. In Issues in Cognitive Linguistics, Leon de Stadler, and Christoph Eyrich (eds.), 471–489. Berlin: Mouton de Gruyter. Sanders, José and Spooren, Wilbert 1996 Subjectivity and certainty in epistemic modality: A study of Dutch epistemic modifiers. Cognitive Linguistics 7: 241–64.

268 Juana I. Marín-Arrese Traugott, Elizabeth 1995 Subjectification in grammaticalisation. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein, and Susan Wright (eds.), 31–54. Cambridge: Cambridge University Press. Traugott, Elizabeth and Richard Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan and Vladimir Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. Ward, Gregory, Betsy Birner and Jeffrey Kaplan 2003 A pragmatic analysis of the epistemic would construction in English. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank Palmer (eds.), 71–9. Berlin: Mouton de Gruyter. Willett, Thomas 1988 A cross–linguistic survey of the grammaticalization of evidentiality. Studies in Language 12: 51–97. White, Peter 2003 Beyond modality and hedging: a dialogic view of the language of intersubjective stance. Text 23: 259–284.

Hearsay adverbs and modality Agnès Celle

Abstract This paper proposes an enunciative analysis of three modal adverbs (reportedly, allegedly, supposedly). It is shown that these adverbs, unlike epistemic adverbs such as obviously or probably, do not express the speaker’s point of view as regards the truth-value of the utterance. By referring to some original assertive source whose existence is far from being clear, they introduce a fictitious point of view with which the speaker may or may not identify. In this way s/he remains on the safe side vis à vis the addressee, who may in turn form a judgement of his / her own or even be forced to react. Assertion is suspended until conclusive evidence is provided by facts or by the addressee. These adverbs may be regarded as disassertive rather than epistemic since they indicate that the speaker does not commit himself to the truth of his / her utterance, whatever the pragmatic purpose might be.

1. Introduction The aim of this paper is to determine whether English hearsay adverbs belong to the realm of epistemic modality. Three adverbs are examined, namely reportedly, allegedly and supposedly. We take as our starting-point Greenbaum’s (1969) study of adverbs, which provides semantic and syntactic criteria that locate these adverbs in the same semantic sub-set. We attempt to show that whatever the scope of hearsay adverbs, modality always has a superordinate status. The nature of the modality at play is then discussed. Unlike epistemic adverbs, hearsay adverbs do not express the speaker’s judgment, but rather his / her lack of commitment to the truth of his / her own utterance. In this paper, we argue that the function of hearsay adverbs is to detach the speaker from his / her utterance, which neutralises assertion and prevents him / her from forming a judgment of his / her own. In the last section, a comparative account of these adverbs is presented.1

270 Agnès Celle 2. Scope of hearsay adverbs and status of modality In Greenbaum’s classification of adverbs, reportedly, allegedly and supposedly are termed “attitudinal disjuncts”. In the sub-set they belong to, adverbs have two syntactic correspondences: “That CLAUSE is PARTICIPLE BASE” and “It is PARTICIPLE BASE (that) CLAUSE”. Greenbaum (1969: 94-95; 224) holds that the following transformational rule accounts for the derivation of these adverbs: “It is Ved (that) S → Ved + ly S”. The scope of disjuncts extends over an entire clause or sentence, as opposed to adjuncts which are integrated within a clause where they modify a word or a word-group. This being said, adverbial scope is not a well-defined concept, a fact which complicates the distinction between word-modification and sentence-modification. For instance, in the examples below there is a difference in terms of scope between (A), where adverbs modify an adjective, a noun-phrase or a prepositional phrase, and (B), where conversely adverbs clearly apply to the whole sentence: A. (1) An African-American suspect, reportedly unarmed, was shot and killed. (http://www.cnn.com/2000/US/03/01/bronx.shooting/index.html) (2)

It [neo-liberalism] has been imposed on the post-socialist economies as the (allegedly) best means of rapid system transformation, economic renewal, and re-integration into the global economy. (N. Fairclough, Analysing Discourse, p. 5)

(3)

If neither of the pair can keep law and order in the areas supposedly under Palestinian administration, Hamas and other Islamists will certainly seek to impose their own. (The Economist, August 7th 13th 2004, p. 32)

B. (2) Reportedly the government has vowed to pass Bill C-5 legislation despite the objections of some of its very own backbenchers, objections that have surfaced as the ferocity of our opposition has mounted against the bill over the last few months. (TransSearch2) (5)

Allegedly there is a separate box sitting somewhere into which the money goes and into which someone could come and tap the account. (TransSearch)

Hearsay adverbs and modality

(6)

271

Supposedly, defences of this sort are needed to protect American territory against a missile attack by rogue states such as Iran or North Korea. (M. Klare, Le Monde Diplomatique, English Edition July 2001)

Nevertheless, while adverbs certainly do modify a word or a word-group in examples under (A), “such ‘non-sentential’ uses are more apparent than real”, as Perkins (1983: 91) has pointed out. Moreover, the paraphrase suggested by Greenbaum for “attitudinal disjuncts” is also applicable here and may indeed contribute to a better understanding of the function of these adverbs: A. (1’) It is reported that the suspect was unarmed. (2')

It is alleged that neo-liberalism is the best means of rapid system transformation, economic renewal, and re-integration into the global economy.

(3')

It is supposed that the areas are under Palestinian administration.

Broadly speaking, this correspondence might be viewed as confirmation of the superordinate status of modality, whatever the position of hearsay adverbs. Hearsay adverbs may be integrated within a clause that they qualify3 in the same way as epistemic modals may be integrated within the clause they modify. But hearsay adverbs have a greater mobility than modals, being able to precede a word or a word-group as in (A), while remaining “peripheral in clause structure”, in Quirk’s words (Quirk & al. 1972: 421). They have a predicative role in the sense that they are always construed as qualifying a predicative relation, as shown by the paraphrases in (1’), (2’) and (3’). In utterances such as those under (B), however, hearsay adverbs qualify the truth-value of the predicative relation as a whole. It should be 4 stressed that it is only in this case that the French conditional may correspond to a hearsay adverb: (4')

Le gouvernement se serait promis d'adopter le projet de loi C-5 en dépit des objections soulevées par certains de ses propres députés, objections qui ont été formulées au moment où s'accentuait notre opposition au projet de loi ces derniers mois. (TransSearch)

In the case illustrated under (A), on the other hand, the hearsay adverb qualifies the validity of the relation established between a noun and an

272 Agnès Celle adjective or a prepositional phrase. It indicates to what extent the speaker is justified in qualifying the noun the way s/he does. The relation between the noun and its modifiers is thus justified solely by virtue of hearsay. In both (A) and (B), hearsay adverbs qualify the link between the speaker and a predicative relation. Likewise, hearsay adverbs may also be incident on the relation between a main clause and an adverbial clause. This allows for the speaker not to commit himself or herself to a causal relation or a purpose-oriented relation as in the following examples: (7)

Mr. Fontaine is now facing eight charges of bribery and fraud for allegedly funnelling over $600,000 from the addictions foundation and away from those very children. (TransSearch)

(8)

We now know that he has scheduled a press conference for 1 p.m. today, supposedly to transfer the failed registry to the office of the Solicitor General. (TransSearch)

By remaining peripheral to the predicative relation they qualify, hearsay adverbs may virtually modify the link between the speaker and any phrase or clause. It is noteworthy that this flexibility is typical of English modal adverbs in general. There exists no equivalent in French, the conditional being limited to a sentential function by which the speaker specifies that s/he disclaims responsibility for his or her utterance5. 6 The peripheral nature of hearsay adverbs is further illustrated by their resistance to negation. Indeed, these adverbs always lie outside the scope of negation: (9)

Yet supposedly Canada is not involved in Bush's war. (TransSearch)

(10) Although I'm fine, I'm supposedly not allowed to wrestle for another few months. (*http://www.thesun.co.uk/article/0,,3-2003270713,00.html (Document Dated: 1 January 2004) (11) DFO reportedly did not receive copies of the auditor's management letters. (TransSearch) The paraphrase test used above works here as well, making it clear that the verb phrase that is negated is that of the complement clause: “it is supposed that Canada is not involved in Bush’s war; it is supposed that I am not allowed to wrestle for another few months; it is reported that DFO did not receive copies of the auditor’s management letters.” The reason why Greenbaum’s paraphrase is still viable with negative utterances is because

Hearsay adverbs and modality

273

hearsay adverbs cannot be negated. None of these examples can be glossed with a negative paraphrase: * it is not supposed that / not reported that. In this respect, hearsay adverbs differ from other modal adverbs that may have two different functions depending on their position in a negative utterance. For instance, possibly may be negated when preceded by the negation, as Hoye (1997:146-147) has demonstrated: (12) She can’t possibly be reached at home. (Hoye, 1997: 146) (13) Possibly, she can’t be reached at home. (Hoye, 1997: 146) In (12), the adverb is integrated within the verb phrase where it serves to reinforce the negated modal, which can be glossed by “it is impossible to reach her at home”, as noted by Hoye (1997:146). By contrast, the adverb in (13) lies outside the scope of negation and modifies the entire clause, which may be paraphrased by “It is possible that she can’t be reached at home”. Both the position and the function of the adverb illustrated in (12) are incompatible with hearsay adverbs which cannot be preceded by the negation: (9’)

?Yet Canada is not supposedly involved in Bush’s war.

(10’) ?Although I’m fine, I’m not supposedly allowed to wrestle for another few months. (11’) *DFO did not reportedly receive copies of the auditor’s management letters. (1’’) *An African-American suspect, not reportedly armed, was shot and killed. As noted above, mobility is a notable feature of modal adverbs that sets them apart from other modal expressions such as adjectives or modal auxiliaries. However, allegedly, reportedly and supposedly are not as mobile as other modal adverbs such as possibly or obviously. Although modal adverbs may theoretically occur in any position, actual corpus data mainly drawn from the Canadian Hansard and from journalistic discourse7 show that reportedly, allegedly and supposedly favour medial position. Compared with the rest of the corpus, examples with adverbs in initial position as in B are exceptions. As shown by the statistics, medial position amounts to more than 90%.

274 Agnès Celle Table 1. Position of hearsay adverbs in parliamentary debates Canadian Hansard Initial position

Reportedly 100 examples 5%

Allegedly 100 examples 3%

Supposedly 100 examples 10%

Medial position

95%

97%

90%

Table 2. Position of hearsay adverbs in journalistic discourse Journalistic discourse Initial position Medial position

Reportedly 200 examples 0% 100%

Allegedly 120 examples 0% 100%

Supposedly 50 examples 2% 98%

Before attempting to provide an explanation of this phenomenon, I will first clarify the meaning of these adverbs.

3. Semantic definition of the sub-class 3.1 Hearsay adverbs as “attitudinal disjuncts” Greenbaum (1969: 94) defines attitudinal disjuncts as follows: “they express the speaker’s attitude to what he is saying, his evaluation of it, or shades of certainty or doubt about it.” Greenbaum (1969: 203) then distinguishes between adverbs that express conviction (definitely, certainly, admittedly, undoubtedly, decidedly etc) and those that express some degree of doubt. In the latter case, adverbs “can be roughly ordered according to the degree of doubt expressed, with doubtless, presumably, likely and probably at one end implying least doubt and conceivably at the other implying most doubt. Several imply that other people may be convinced of the truth of what is being said, e.g. conceivably and arguably. Some, e.g. allegedly and reportedly, imply in addition that what is being said has been stated previously by others.” Greenbaum (1969: 122) also notes that these adverbs differ from most other attitudinal disjuncts in that they cannot be modified, unlike likely or probably for instance: *very allegedly, *very reportedly, *very supposedly.

Hearsay adverbs and modality

275

Although Greenbaum’s classification is certainly correct, it is questionable whether the concept of doubt is a valid criterion, all the more so as allegedly, reportedly and supposedly are all incompatible with adverbs that modify the degree of doubt such as most, highly or very. Moreover, given that these adverbs all have past participle stems, reference to a source might be regarded as fundamental and not simply additional.

3.2. Evidentiality Might it then be possible to separate the quotative function from the epistemic judgment, given that these adverbs also qualify the truth-value of the sentence? In the approach adopted by Lyons (1977: 797), hearsay adverbs are considered to involve epistemic modality in so far as “Any utterance in which the speaker explicitly qualifies his commitment to the truth of the propositions expressed by the sentence he utters, whether this commitment to the truth is made explicit in the verbal component [...] or in the prosodic component, is an epistemically modal, or modalized utterance.” However, different views have been advanced more recently, particularly with respect to evidentiality. As it is, Lyons’ definition of epistemic modality echoes Chafe’s definition of evidentiality. According to Chafe (1986: 262), evidentiality is not limited to the expression of evidence. Evidentiality in the broad sense involves “attitudes toward knowledge.” In Chafe’s opinion, hearsay adverbs denote a mode of knowing; that is, they ‘qualify knowledge as having been acquired through language rather than direct experience.’ (Chafe 1986: 268). These definitions of epistemic modality and evidentiality in the broad sense overlap, which is evidently problematic from a semantic point of view. In an attempt to draw a line between the two categories, Palmer (1986: 70-71) argues that some systems are ‘predominantly evidential, e.g. that of Tuyuca, while others, e.g. that of English, are predominantly judgments.’ Some other languages have a mixed system. However, there is not always a clear-cut distinction between evidentials and judgments, given that ‘some languages that have a predominantly judgment system may include quotative within that system’: There are languages in which the speaker may indicate the strength of his commitment to what he is saying, not in terms of possibility and necessity but in terms of what kind of evidence he has. [...] It is worth noting here, however, that even European languages may use these modal markers as

276 Agnès Celle evidentials. German, for instance, uses the modal verbs sollen and wollen to express what everyone says and what the subject of the sentence claims, respectively. [...] This is not a matter of judgment, still less of possibility and necessity, but relates to the commitment of the speaker to the truth of what he is saying. He indicates that his responsibility is limited by the fact that some one else said it. [...] It would be foolish to deny the name of ‘epistemic’ to such features, for not only are they clearly concerned with the speakers’ knowledge and belief, though a little more indirectly, but also they often occur in the same formal system as Judgments.8 (Palmer 1986: 53)

While acknowledging that the use of quotative devices is “not a matter of judgment”, Palmer nonetheless relates these devices to the system of epistemic modality, which can hardly be conceived of as independent of judgment. To resolve this apparent contradiction, the concepts of epistemic modality and assertion need to be clarified.

3.3. Restrictions on epistemic modality In English, several adverbs may serve a quotative function. It should be stressed, however, that these adverbs do not freely co-occur with epistemic modals and adverbs. In this respect, they differ from other modal adverbs such as apparently, clearly, evidently and obviously, which collocate with epistemic modals, as Hoye (1997: 105; 168) has shown: (14) Surely the run must be nearly over now. (quoted by Hoye 1997: 200) (14') *Allegedly / *reportedly / *supposedly the run must be nearly over now. (15) Evidently there must still be other receptors, able to respond to changes of light intensity in the wall of the diencephalon. (quoted by Hoye 1997: 192) (16) Moreover, its extent must obviously increase with the reactivity of the aromatic substrate. (quoted by Hoye, 1997: 169) In (15) and (16), the epistemic judgment draws upon the speaker’s perception of the current state of affairs. Undeniably, epistemic modality and inference are conceptually related, which supports Guimier’s claim that

Hearsay adverbs and modality

277

inference and epistemic judgment are the defining characteristics of modal adverbs. Guimier (1986: 256) argues that inference drawn from evidence establishes a link between real world and discourse, allowing thus the speaker to form an epistemic judgment about a possible world. In (15) and (16), evidently and obviously qualify the truth-value of the utterance by reinforcing the speaker’s confidence in the truth of the proposition. As Hoye (1997: 168) puts it, “These emphasizers convey the notion that such is the evidence available to the speaker that no other conclusion can possibly be drawn.” As far as co-occurrence with modals is concerned, hearsay adverbs also differ from nominalized speech act verbs. Deverbal nouns such as report and comment allow for the source to be identified and for a modal judgment to be expressed in the complement clause: (17) [Headline] Bush's military views reportedly shifting. By James Gerstenzang AUSTIN, Texas (Los Angeles Times) – President-elect George W. Bush, who questioned the extent of the U.S. commitment in Europe during the campaign, sounded more cautious Monday about reducing the American military role there, according to a senior senator. The report by Sen. Carl Levin (D-Mich.) and comments by other members of a bipartisan delegation of defense experts from the Senate and House suggest that as he comes face to face with the responsibilities of being commander in chief, Bush may be reluctant to make dramatic shifts in U.S. defense policy and spending priorities. But he showed no inclination to step back from his support for a national missile defense program. (http://www.cnn.com/2001/ALLPOLITICS/stories/01/09/latimes.mi litary/index.html *Document Dated: 18 June 2004) The verb suggest in (17) indicates that the complement clause cannot be taken as a verbatim quote, but rather as the speaker’s interpretation of the report and comments. With the modal may, the speaker assesses how likely Bush’s reluctance is, leaving the choice open between the value selected and the complementary value. In this case, may cannot co-occur with reportedly: (17') ?Bush may reportedly be reluctant to make dramatic shifts in U.S. defense policy and spending priorities.

278 Agnès Celle If, however, the epistemic judgment originates from the reported speaker and not from the speaker, co-occurrence with may is possible: (18)

In the six months since its premiere, Mel Gibson's The Passion of the Christ has grossed more than $370 million in the U.S., making The Passion the second top grossing movie of 2004, trailing only Shrek 2. This week, the DVD and VHS versions of the film will be released. Distributor Twentieth Century Fox reportedly may ship as many as 15 million copies to retailers. (http://www.csicop.org/scienceandmedia/passion-of-christ)

Interestingly, the modal that was found to co-occur most frequently with these adverbs is will. Will is closer to assertion than may or must in the sense that no alternative to the future occurrence is taken into account9. (19) [Headline] NIH reportedly to release guidelines allowing research on embryo cells. Because current law forbids the use of federal funds to harm a human embryo, federally funded scientists may not extract the cells from human embryos for research purposes. That job would have to be done by privately funded researchers who could then pass the useful cells to federally supported scientists, creating a separation of responsibility for the embryo's destruction that opponents have criticized as morally meaningless, according to the Post report. The Post said the new guidelines will insist that federal research be conducted only on cells taken from frozen embryos that were destined to be discarded anyway. They also reportedly will disallow payments to embryo donors and will preclude donors from specifying who should receive their embryo's stem cells. Those provisions aim to discourage the creation of a market in embryo cells and block any incentive for a woman to create fresh embryos just to provide a potential treatment for a sick relative. (* http://www.cnn.com/2000/HEALTH/08/23/stem.cells/index.html Document Dated: 18 June 2004) (20) I find it offensive that members across the way and others who are opposed to this bill, rather than stick to the real issues in the bill and what it deals with, try to create these bogeymen about the fact that the government supposedly will come and take all their guns, which is absolute nonsense, and they know that. (TransSearch) In these examples, it is not the speaker who is making predictions. In fact, the modal judgment expressed by will is made by the reported speaker and

Hearsay adverbs and modality

279

is simply being quoted by the speaker, as in (18). A comparison may be drawn between the conditions of use of modals here and the restrictions on their use in if-clauses. As in if-clauses, modals may co-occur with hearsay adverbs if they express “objective epistemic modality” and not “subjective epistemic modality”, according to the distinction made by Lyons (1977: 805). “Objective epistemic modality” results from back-reference and reported speech, the judgment originating from an origo other than the speaker. In the case of will in if-clauses, Close (1980) argues in a similar way that the speaker is not “making a prediction to assumed future actuality”, but “assuming predictability”10. With hearsay adverbs as with ifclauses, modals have to express the reported speaker’s judgment. This constraint is reflected in the position of hearsay adverbs, which more often than not precede the whole verb phrase. They tend not to be inserted between the modal auxiliary and the verb base, a position that is more easily filled by adverbs that express the speaker’s own point of view. The distinction made by Lyons between “objective epistemic modality” and “subjective epistemic modality” is an important one. However, as Nuyts (2001) has shown, this distinction has an effect on the speakerhearer relation and might be more adequately handled in terms of intersubjectivity vs. subjectivity. He defines intersubjectivity as follows: Intersubjectivity means that the information (and the epistemic evaluation of it) is generally known, and hence is not new (or surprising) to speaker and hearer(s). (Nuyts 2001: 396)

Indeed, the modals which co-occur with hearsay adverbs refer to shared knowledge and not to the speaker’s own point of view. As hearsay adverbs conflict with subjective evaluation, they may be regarded as addresseeoriented. However, not all modals are able to erase the speaker’s point of view when combined with hearsay adverbs. With must, for instance, the reported speaker’s point of view cannot take precedence over the speaker’s, at least when epistemic modality11 is at play. This might explain why our database did not yield a single instance of epistemic must co-occurring with hearsay adverbs and why the following utterances are not felicitous: (18') ?Distributor Twentieth Century Fox reportedly must ship as many as 15 million copies to retailers.

280 Agnès Celle (19') ?They also reportedly must disallow payments to embryo donors and must preclude donors from specifying who should receive their embryo's stem cells. When combined with a hearsay adverb, must is prone to express an obligation imposed by the reported speaker, which does not make sense in the contexts of these examples. Hearsay adverbs seem unable to conflate reference to a source and the speaker’s own epistemic judgment, which a priori might be a problem for Guimier’s definition of modal adverbs, as his definition implies both inference and epistemic judgment. The tests conducted so far seem to indicate that the latter is not compatible with hearsay adverbs, unless it originates from the reported speaker. However, Guimier (1986: 256-257) stresses that the two components of modal adverbs are not necessarily equally weighted. One may even completely take precedence over the other. He gives the example of visibly, which he considers purely inferential, and contrasts it with probably, which he considers purely epistemic. Along the same line of reasoning, it might be argued that the speaker’s epistemic judgment is overshadowed by the quotative function in the case of hearsay adverbs. And yet, reference to another speaker’s point of view does have a modal effect, given that these adverbs inevitably qualify the truth-value of what the speaker is saying. More specifically, these adverbs signal the speaker’s lack of commitment to what s/he is saying. Therefore, the paraphrases provided by Guimier to capture the meaning of purely inferential adverbs are not applicable to hearsay adverbs: (21) Gertrude visibly didn’t listen. Paraphrases: “From what I could see, I can assert that…; what I could see allows me to assert that…” (Guimier 1986: 253) Even if these paraphrases are adapted for hearsay adverbs (“what I heard allows me to assert that...; from what I heard, I inferred that”) they remain inadequate because hearsay does not lead the speaker to infer anything. Nor does it allow him / her to produce an assertion.

3.4. Disassertion By and large, these adverbs refer to what other speakers have said or claimed and thus introduce a point of view different from the speaker’s.

Hearsay adverbs and modality

281

However, they should not be taken to express indirect speech, there being no original speech act. Unlike indirect speech, hearsay implies a fictitious situation12 disconnected from the situation of utterance where the reported speaker is in fact a fictitious origo. Hearsay adverbs tend to appear in medial position because the speaker indicates on what belief and on what grounds his / her utterance is predicated. What is at stake is the assertive status of the utterance rather than the link between a state of affairs and an epistemic judgment. The nature of the speaker’s relation to alternative representations13 implied by hearsay adverbs might be used as a criterion for distinguishing them from other modal adverbs that freely collocate with epistemic modals. In (15) and (16), the epistemic judgment conveyed by the modal must and the modal adverbs is based on the same modal operation, that is, an inferential process that eliminates alternative values. Obviously and evidently may easily occur sentence-initially because they establish a link between the speaker’s perception of a state of affairs and his / her conclusion drawn from external evidence, as Guimier (1986: 253-255) has demonstrated. As a result, the speaker’s commitment to his / her own utterance is reinforced. In (18), (19) and (20), the modal judgment is reversed. The adverbs do not eliminate alternative values, but rather introduce a point of view different from the speaker’s, whose confirmation is postponed. Assertion is suspended until the view expressed is confirmed by facts or by the addressee. Reference to alternative representations, that is, to some speaker’s prior or even fictitious speech, may be used by the speaker as a pretext to avoid committing himself / herself. The speaker thereby acts as a mere locutor and not as an enunciator, according to the distinction made in the enunciative framework. As Culioli (1995: 92) points out, producing an assertion involves taking responsibility for it and for all its possible consequences: To be able to assert, to produce an assertion, I must declare it publicly […]; but there also needs be the commitment of someone who takes responsibility, who guarantees your statement or who wishes to assert something despite your position. If it is simply an instance of locution, you do not really have assertion. For there to be responsibility, one must guarantee what one states. In an institutional setting, the guarantor could be sanctioned if what he guarantees does not materialize. If at some time one asserts – in the fullest sense of the verb – that a certain event will have such and such a consequence and it does not follow, if one produces an assertion or a theorem and others realize that it cannot be reproduced experimentally, or if others can demonstrate that one’s theorem is erroneous, as one has guaranteed it, this can possibly have repercussions; sanctions may be taken. Now we see the importance of not having merely a locutor.

282 Agnès Celle In an assertive utterance, alternative values are eliminated by the speaker who commits himself to the truth of the proposition without explicitly marking his / her conviction. This speech act has an effect on the intersubjective relation since it leaves no room for polemic. By contrast, hearsay adverbs point to some polemic issue and involve the addressee as a potentially discordant anchor point. In news reporting, the speaker takes the precaution of presenting information as second-hand to disclaim any responsibility. S/he thereby acts as a mere locutor, leaving it to the reader to form opinions and judgments of his / her own. In parliamentary debate and more generally whenever verbal exchange is possible, this subjective dissociation may take on a pragmatic value to spark off discussion or even to criticise widely-held views. In this type of discourse, hearsay adverbs may appear sentence-initially (see figure I) precisely to question the source of information and elicit the addressee’s response. It should be stressed that in such dialogical contexts, the point of view that is presented as fictitious by hearsay adverbs may coincide with the addressee’s. While epistemic adverbs basically express the speaker’s judgment by specifying his / her position relative to alternative representations, hearsay adverbs do not. Instead, they leave it up to the addressee – reader or hearer – to form a judgment. Hearsay adverbs might therefore be considered to be both disassertive and addressee-oriented.

4. Specific meaning of each adverb 4.1. Reportedly Let us now move on to the specific meaning of each of the adverbs under discussion. The Longman Dictionary of Contemporary English (hereafter LDOCE) provides the following definition for reportedly: “according to what some people say”. Of all three adverbs, reportedly is the most neutral, because it simply indicates that the speaker detaches himself / herself from the assertive source. Reportedly frequently occurs in headlines to sum up a news report as in (17) and (19), signalling right from the start that the report is not to be interpreted by the reader as the journalist’s own report but as reported speech, even in the absence of a framing clause. This signal allows syntactic flexibility within the body of the article, where reported

Hearsay adverbs and modality

283

speech is mostly paratactic. Hypotactic reported speech is less frequent and generally follows an instance of paratactic reported speech, as in (22), where reported speech is first paratactic and accompanied by a reporting clause in inquit position that qualifies the statement. It is only then that indirect speech is introduced by the framing clause a White House official confirmed: (22) U.S. reportedly OKs funds for Iraqi opposition (CNN) – The United States has given an Iraqi opposition group permission to use U.S. funds to collect information on alleged violations of human rights and United Nations resolutions by the government of President Saddam Hussein, U.S. officials say. A White House official confirmed that final approval for the Iraqi National Congress to use the money inside Iraq was given this week by the Treasury Department. The move marks the first policy decision by President George W. Bush on how to treat the Iraqi opposition. (*http://www.cnn.com/2001/US/02/02/us.iraq/index.html *Document Dated:18 June 2004) The definition of the LDOCE is partly confirmed by these examples, which can all be paraphrased by “according to what some people say.” However, this definition provides no explanation for the modal status of these utterances. Indeed, reportedly marks indirect attribution, which cannot be equated with pure indirect speech, since the original speech act is not referred to. Only information is retained and passed on to the reader. It has been noted in the analysis of (17) that reported speech goes hand in hand with the speaker’s interpretation, which is marked by modal verbs and adverbs that are related to the situation of utterance. By contrast, reportedly does not translate the original speech act into indirect speech, which explains why it is not compatible with modals and adverbs that are indexed to the speaker and to the situation of utterance. Reportedly does not indicate any degree of doubt or certainty, even if such connotations may arise from the text. In (22), for instance, the verb confirm takes it as established beyond doubt that the proposition is true. In fact, this position on the truth-value of the proposition is limited to the reported speaker a White House official. Metalinguistic comments added as an afterthought suggest that the speaker is in a position to adopt a range of

284 Agnès Celle different attitudes to what s/he is saying. S/he may even hope that the proposition is not true as in (23). (23) I would like to take this opportunity to say that I hope the person who was reportedly infected – I hope she has not been – will recover very quickly. (TransSearch) S/he may also lack information as in (24): (24) Then we read stories, one as recently as in today's paper, reportedly, and I say reportedly because all the facts are not in yet, but reportedly about a $2 million party for the top brass in the military. (TransSearch) In both cases assertion is, as it were, suspended until facts are established. Reportedly does not cast any doubt on the proposition, but simply indicates that the speaker chooses to distance himself / herself from the assertive source. Reportedly expresses no epistemic judgment, the speaker being only held accountable for attributing information to other speakers, even if the sources are not identified.

4.2. Allegedly According to the LDOCE, allegedly is “used when reporting something that people say is true, although it has not been proved”. While reportedly simply detaches the speaker from his / her source, allegedly detaches him / her from the whole propositional content, for s/he cannot say whether the proposition is true. Hence the paraphrase “but I don’t believe it” suggested by Ramat (1996: 288), as opposed to “someone told me so” for reportedly. The “but I don’t believe it” paraphrase seems appropriate for example (2). Similar expressions may appear in the form of a metalinguistic comment as in the following examples: (25) If we allow the United States to strike Iraq, because it allegedly has evidence, which is far from convincing as far as we are concerned, we could never oppose China striking Taiwan, as an example. (TransSearch) (26) He was referring to two things Sharpe had allegedly written. I say allegedly because I do not believe he necessarily wrote them. (TransSearch)

Hearsay adverbs and modality

285

Here again, the modal meaning of the adverb needs to be clarified. In this respect, the difference with reportedly is twofold. Firstly, allegedly specifies that the speaker refuses to commit himself / herself to the truth of the proposition, while there is simply no speaker’s commitment with reportedly. Secondly, reportedly indicates speech attribution, even if the source is unidentified, while allegedly does not necessarily imply an assertive source, hearsay serving as a fictitious pretext in a number of cases. In this respect, example (27) may be contrasted with (29) and (22): (27)

[Headline] US Marine officer investigated for allegedly slurring gays in e-mail WASHINGTON (CNN) – A Marine Corps lieutenant colonel is under investigation for derogatory comments he allegedly made on an internal e-mail about gays serving in the Marine Corps. In passing along the Pentagon's guidance by way of e-mail, however, Melton allegedly injected personal bias toward homosexuals before redistributing the memo internally. (* http://www.cnn.com/1999/US/12/14/dont.ask.email/index.html * Document Dated: 18 June 2004)

The body of the article in (27) is interpreted as the journalist’s report without attribution, which explains why allegedly has to be repeated throughout the article. By using allegedly, the speaker does not seek to detach himself / herself from the source of allegations. Instead, s/he disclaims responsibility for the validity of each accusation until facts are established by the investigation or the trial. This does not mean, however, that the speaker does not believe these accusations and the paraphrase “I don’t believe it” is less felicitous here than in (2), (25) and (26). It should be stressed that unlike reportedly, allegedly is extensively used in reports of criminal cases. As noted by Palmer (1986: 73), “it is important for the writer to report what has been said in criminal cases without being caught by the laws of libel.” Here again, the speaker is not casting any doubts on the proposition. As Huddleston & Pullum point out (2002: 769), a difference has to be made between the adverb and the corresponding predicative element “it is alleged”: (28) Max had allegedly falsified the accounts. (Huddleston & Pullum 2002: 769) (28') It was alleged that Max had falsified the accounts. But I’m sure he hadn’t.

286 Agnès Celle According to Huddleston and Pullum, “the fact that the qualification is expressed adverbially rather than by a predicative element serves to background it, and make it less accessible to denial. The verbal version [...] allows the expression of a contrary view (but I’m sure he hadn’t), but (28) does not accommodate that kind of continuation.” The reason for this difference might be that the predicative element introduces an instance of reported speech for which only reported speakers are accountable. Allegedly, by contrast, qualifies the speaker’s statement by destabilising his / her own assertion. This difference is illustrated in the following example14: (29) “Charlotte Dard suffered occasional blackouts,” Dr. Lanier explains. “Or at least this is what I was told at the time.” “Who told you?” “Her sister. It appears,” he goes on, “or let me qualify this by saying it is alleged, that she suffered from retrograde amnesia…” “I certainly would think her family would know that, unless no one was ever home.” “Problem is, her husband Jason Dard's a rather shady character. Nobody around here knows much, maybe nothing about him, except he's rich as hell and lives on an old plantation. I wouldn't call Mrs. Guidon a reliable witness. Although she certainly could be telling the truth about her sister's condition prior to her death.” “I've read the police narrative, which is brief. Tell me what you know,” Scarpetta says. After a coughing bout, Dr. Lanier replies, “The hotel where she died is in a not-so-nice part of the city, in my jurisdiction. A housekeeper found her body.” “What about blood tests? In the paperwork you sent me, all I got were postmortem levels. So I don't know whether she might have had the elevated GGTP or CDT associated with alcohol abuse.” “Since I first contacted you, I have managed to track down premortem blood test results, because she was in the hospital about two weeks before her death. Misfiled, I'm embarrassed to say. I've got a particular clerk I'd pay heaven and earth to get rid of. But she's the sort to

Hearsay adverbs and modality

287

sue for one thing or another. The answer to your question is no – no elevated GGTP or CDT.” “In the hospital for what?” “Tests after her most recent blackout. So, obviously, she had one of these blackouts two weeks before she died. Again, I say allegedly.” (P. Cornwell, Blow Fly, p. 282-283) Being questioned by Kay Scarpetta, Chief Medical Examiner, Dr Lanier realises that nothing can be taken for granted and feels the need to qualify his statements. By using the predicative element it is alleged that, he is alluding to what her sister told him. The initially assertive statement Charlotte Dard suffered occasional blackouts is thus presented as her sister’s version of events, which may not be the true one. In contrast, the use of the adverb allegedly at the end of the excerpt does not involve reporting, which explains why the adverb is not interchangeable with the predicative element. So and obviously both indicate that the speaker first draws conclusions from external evidence in a statement that expresses his spontaneous reaction, that is, he believes that she had a blackout before she died. However, he then uses allegedly to downtone his own statement, though evidence seems to point to Charlotte Dard’s having had a blackout before she died. The reason why the speaker qualifies his statement is to remain on the safe side vis-à-vis the hearer. Therefore the paraphrase to be suggested here is “I don’t take responsibility for my belief” as opposed to “I don’t take responsibility for that belief” in (2). Allegedly may thus detach the speaker not only from other speakers’ claims, but also from his / her own discourse. In the latter case, hearsay provides a mere pretext for the speaker to avoid committing himself / herself to the truth of a predicative relation that the addressee is likely to find fault with.

4.3. Supposedly To cite the LDOCE, supposedly is “used when saying what many people say or believe is true, especially when you disagree with them”. Supposedly refers to a belief that is shared by many speakers and possibly by the speaker him/herself. Starting from this definition, the modal function of this adverb can be explained from an enunciative point of view. In fact, supposedly restricts the belief it denotes to a detached representation of

288 Agnès Celle reality, which blocks identification with what is or was actually the case. The utterance is viewed as valid in theory, but theory and practice cannot be reconciled. In the following example, supposedly clearly functions as a disassertive marker, the circumstances of Darnley’s murder being still unclear: (30) It was rumoured that Mary was in love with a tough, rough Scottish nobleman called Bothwell. She supposedly asked him to kill Darnley, her feeble English husband, so that they could run away together. (F. Macdonald, p. 92) Supposedly refers to a plausible scenario that seems to have been corroborated by the fact the Mary Stuart ran away with Bothwell. It corresponds to a logical argument. However, even if this scenario might coincide with the true story, the presence of the adverb disallows the following paraphrase: “indeed she asked him to kill Darnley, her feeble English husband, so that they could run away together.” The predicative relation <she – ask – him – to kill Darnley> is viewed as a logical argument detached from reality and cannot be identified with facts. In other words, alternative scenarios cannot be elimininated, which blocks assertion. Because it refers to a detached representation of reality, supposedly is often used when the actual state of affairs does not meet the speaker’s expectations. This explains why supposedly is frequently, but not necessarily, combined with conjunctions such as but, however, and yet: (31) With this release, though, Symantec has expanded the scope of Anti-virus. The program is supposedly now able to deal with newer terrors like spyware and keystroke-logging programs, as well as annoyances such as adware and diallers. However, as the configuration options on these new triggers effectively amount to ‘on’ or ‘off’, the user is reliant on Symantec’s definitions of the latest spyware / adware / dialler threats being up to date – but our tests were none too reassuring on this point. (Personal Computer World, March 2004, S. Colvey, p. 68) In (32), the competence attributed to the designer reflects some other point of view than the speaker’s. It is viewed as a theoretical ability contradicted by events, a top designer being expected to manage more than seven sales. The discrepancy between the designer’s expected proficiency and the results indirectly calls into question his being a top designer:

Hearsay adverbs and modality

289

(32) Astonishingly, the supposedly top designer managed a total of just SEVEN sales, with staff outnumbering customers in every shop. (http://www.mirror.co.uk/news/allnews/content * Document Dated: 1 January 2001) A sense of denial may emerge from there being a gap between the speaker’s representation and what actually happens or will happen, as in (20), (32), or in (33): (33) That being said, I invite the member from the Canadian Alliance to read my speech tomorrow because history contradicts some of the arguments that he put forward regarding the fact that the Acadians were supposedly given a last chance to take the oath of allegiance. (TransSearch) Because situational evidence does not concur with a norm shared by speakers, this norm itself becomes debatable. The hearer is thus indirectly led to infer from situational evidence that his / her norm does not hold. In the context of this debate, this conflict is exploited pragmatically by the speaker in an attempt to persuade the hearer that his arguments are wrong.

5. Conclusion While modal adverbs such as obviously or evidently eliminate alternative representations in an inferential process that allows the speaker to form an epistemic judgment, reportedly, allegedly and supposedly introduce an alternative representation, allowing two states of affairs – a fictitious one and the actual one – to coexist. These adverbs are modal in that they specify the speaker’s lack of commitment to the truth of his / her utterance and might therefore be termed disassertive rather than epistemic. At the same time, these adverbs are addressee-oriented, since it is up to the addressee to draw conclusions, the speaker pretending not to have a personal point of view or being reluctant to express it. With reportedly, the responsibility for the proposition is attributed to some source that is distinguished from the speaker. With allegedly, reference to an assertive source may be purely fictitious in order for the speaker to disclaim responsibility for the proposition until facts are established. With supposedly, a theoretical representation – whether it be the speaker’s logic or an accepted norm – is presented as removed from reality. However accepted this norm might be, it is shown not to be borne out by actual facts. Therefore, this adverb may be used to

290 Agnès Celle undermine socio-cultural norms in an indirect way, implying disagreeement without explicitly expressing it. Crucially, these adverbs legitimise the speaker’s utterance vis à vis the addressee by taking account of the other’s viewpoint. At the same time, non-commitment may become a strategy for the speaker to suggest disagreement without ever facing the other in a direct way, the fictitious plane functioning as a buffer zone between speaker and addressee.

Notes 1.

2. 3. 4. 5.

6. 7. 8.

I wish to thank the anonymous reviewers for their constructive comments on an earlier version of this study, which helped me improve it. Any remaining errors are my sole responsibility TransSearch is both a database of translations and a bilingual concordancer. The type of adverbial scope in question here corresponds to Guimier’s “incidence exophrastique” (Guimier 1986: 217–219; 276–277). The French conditional is analysed from a contrastive point of view in Celle (2006: 74–92). Although contrastive analysis is not the aim of this paper, it should be pointed out that hearsay adverbs operating at the noun level or at the inter-clausal level are a problem for translation into French. Modality will either be left out in French or translated using a modal verb in a comment clause (il paraît, semble-t-il). These two tendencies are illustrated by the following examples: – Constable Jimmy Ng of Richmond, a Mountie, was killed recently by an allegedly speeding driver who was formally charged with alleged street racing. Le policier Jimmy Ng, de Richmond, agent de la GRC, a été tué récemment par un conducteur qui conduisait trop vite et qui a été officiellement inculpé pour course dans les rues. (TransSearch) – (7’) M. Fontaine fait face à huit chefs d'accusation de subornation et de fraude. Il est accusé d'avoir détourné plus de 600 000 dollars de la fondation pour la lutte contre les dépendances et d'en avoir ainsi privé ces enfants. (TransSearch) – (8’) Nous savons maintenant qu'il tiendra une conférence de presse à 13 heures aujourd'hui, afin d'annoncer, semble-t-il, le transfert du registre raté au bureau du Solliciteur général. (TransSearch). See Quirk & al. (1972: 423): “They [disjuncts and conjuncts, as opposed to adjuncts] can appear initially before a negative clause.” Unless stated otherwise, all these examples were extracted using WebCorp. The italics are mine.

Hearsay adverbs and modality

291

9.

May and must do imply an alternative value, albeit in an opposite way. Any alternative value is ruled out with must, while the complementary value cannot be ruled out with may. 10. This distinction is examined in more detail in Celle (2004: 212–213). 11. Deontic must is perfectly acceptable with hearsay adverbs, as in the following example: Even schoolchildren are scrutinized for dissident tendencies. They reportedly must carry a 17-page Interior Ministry identity form that, in addition to documenting the carrier's name and current and former addresses, lists assessments of political attitudes and participation in communist youth activities. (http://www.heritage.org/Research/LatinAmerica/BG1579. cfm&term= reportedly&case=case) 12. The concept of fictitious situation is borrowed from Culioli (1995 : 115 ; 125) : “We construct a locator, which I have called fictitious locator Sit01, detached from the time-based locating system. We construct an image of the enunciator-subject S0 and of T0 that has the following properties : it is constructed with respect to Sit0 : Sit01 is in fact (S01, T01)”. Another plane is produced, disconnected from the situation of utterance. The fictitious locator enables the enunciator to dissociate himself / herself from the locutor ,thus not to commit himself / herself to the utterance. 13. Assertion is considered to constitute the most basic level of modality in the enunciative framework. While other types of modality all involve otherness, i.e. an alternative representation, assertion eliminates it. This view is inspired by Culioli (1990: 207–210): “[…] for an assertion to be produced, we must eliminate alternative values (French altérités) so that we shall construct a loop by identifying the notional representation of an event, attached to a predicative relation, with the representation of what actually is the case, as defined by the asserting subject. […] Consequently, the metalinguistic representation of an assertion will be: , and no other value than is to be identified with the state of affairs located relative to Sit(uation); in other words ‘ is the case, and other –than is not the case’” 14. My thanks to Françoise Doro-Mégy, from the University of Paris 8, for pointing out this example to me.

References Celle, Agnès 2004 Future time reference in the conditional protasis in English and French: a corpus–based approach. In Practical Applications in Language and Computers, Barbara Lewandowska–Tomaszczyk (ed.), 209–217. Bern: Peter Lang.

292 Agnès Celle 2006

Temps et modalité : l’anglais, le français et l’allemand en contraste. Bern: Peter Lang. Chafe, Wallace 1986 Evidentiality in English Conversation and Academic Writing. In Evidentiality: the linguistic coding of epistemology. Advances in discourse processes, Wallace Chafe, and Johanna Nichols (eds.), 261– 272. Norwood: Ablex. Close, Reginald A. 1980 Will in if–clauses. In Studies in English Linguistics for Randolph Quirk, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (eds.), 100–109. London: Longman. Culioli, Antoine 1990 Pour une linguistique de l’énonciation, opérations et représentations, Tome 1. Paris: Ophrys. 1995 Cognition and Representation in Linguistic Theory. Edited and introduced by Michael Liddle. Translated with the assistance of John T. Stonham. Amsterdam: John Benjamins. Greenbaum, Sidney 1969 Studies in English Adverbial Usage. London: Longman. Guimier, Claude 1986 Syntaxe de l’adverbe anglais. Lille: Presses Universitaires de Lille. Hoye, Leo 1997 Adverbs and Modality in English. London: Longman. Huddleston, Rodney, and Geoffrey Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Longman Dictionary of Contemporary English. 2003 London: Longman Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. Nuyts, Jan 2001 Subjectivity as an evidential dimension in epistemic modal expresssions. In Journal of Pragmatics 33: 383–400. Palmer, Frank 1986 Mood and Modality. Cambridge: Cambridge University Press. Perkins, Michael 1983 Modal Expressions in English. London: Frances Pinter. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1984 A Comprehensive Grammar of the English Language. London: Longman.

Hearsay adverbs and modality

293

Ramat, Paolo 1996 “Allegedly, John is ill again”: stratégies pour le médiatif. In L'énonciation médiatisée, Zlatka Guentchéva (ed.), 287–298. Louvain/Paris: Peeters.

Sources of corpus examples Cornwell Patricia 2003 Blow Fly. Timewarner Paperback. Fairclough Norman 2003 Analysing Discourse. Textual Analysis for Social Reseach. London: Routledge. Le Monde Diplomatique English Edition. Macdonald Fiona 2004 The History of Scotland for children. Bath: Parragon. Personal Computer World. The Economist. TransSearch: http://www.tsrali.com/ Corpus data from the World Wide Web were extracted using WebCorp.

When Yes means No, and other hidden modalities Claude Rivière

If it were done when ‘tis done, then ‘twere well / It were done quickly. Macbeth [I, vii, 1]

1. Introduction Saying yes when you mean no, or the opposite, is not remarkable in discourse: it is the common device of irony. It is more surprising to find the same device at the syntactic and semantic heart of the language to express time location and to measure the duration of events. The constructions that will be considered here do not include any terms that may be considered as negative in any way. It is the way the terms are put together that yields the negative meaning, even though it can’t be linked to any single term. The constructions are the notorious it BE … since..., as in: (1)

He admits: ‘It is three years since I’ve scored for England’. And it BE … before:

(2)

It was eighteen days before Margaret finally decided.

These constructions with paradoxical polarity will be compared to two other types of constructions: (a)

positive-polarity constructions that express time location:

(1a) I last scored for England three years ago. (2a) Margaret finally decided eighteen days later. (b)

negative-polarity constructions that express duration:

(1b) I haven’t scored for England for three years.

296 Claude Rivière (2b) Margaret didn’t decide for eighteen days. The originality of these constructions is still more obvious when they are compared to the irrealis, which also implies the opposite of what is stated. The irrealis differs from the constructions under consideration in two ways, which are in fact connected. First, the verb form used for the irrealis must be the ‘fictional’ (also ‘modal’) preterit, and secondly, as a consequence, the event that is explicitly stated is placed in a fictional space while the implied event is in a real space, at the same point in time . The constructions studied here state two events in the same real space, but they are, of course, ordered in time. The exploration of the hidden negation will lead to a consideration of other modal phenomena and, more specifically, to the effects of the modality of expectation, another hidden modality, and to the role of the adverb only.

2. The construction it BE … since… 2.1. Description The construction it BE [duration] since [ P ] is ambiguous between two readings: time location and measurement of duration. (A)

Time location. An event p is located in time by measuring the distance between time Tp (time of the event) and time T1 (time of the point of view), which may be T0 (time of speaking) as in the following example:

(3)

The cause of recovering Greenham Common for the people has been taken up by the author of Watership Down, Mr Richard Adams, who knew the area before it was requisitioned during the second world war.’I don’t see why this splendid open space should not be restored just because it’s 50 years since it was taken in a national emergency,’ he said. BNC: The Guardian, electronic edition of 1989 12– 31.

Here the duration is used as an instrument towards time location, and this interpretation can (theoretically) be paraphrased with an ago phrase:

When Yes means No

297

(3a) This splendid open space was taken in a national emergency 50 years ago. (B)

Measurement of duration. This is the reading where the expressed polarity is positive while the intended polarity is negative.

(4)

If you must know, I eat rabbits, wild birds, chickens from the farm and young sheep from the hills. I like children when I can find them. But they’re usually too well locked in at night. It’s two months since I tasted child meat.’ BNC: “Saki” and Rosemary Border, Tooth and claw (1991).

The point of this utterance is not to locate in time the occasion when the ogre ate child meat, but on the contrary to measure the duration of his not eating child meat. This way of measuring does not mention the negative event to be measured, it only mentions the positive complementary, which is used as the basis for the calculation. As in the previous case, p is first located in time, by using a phrase that expresses duration. Then, it is used as the left boundary (start) for a length of time during which p is no longer valid. In other words, the (indirect) time location of p becomes the direct measurement of p’, so that it is possible to measure something that is never mentioned and that does not exist. This reading, which does not mention p’, can be paraphrased with an utterance that explicitly measures p’ with a for phrase: (4a) It’s two months since I tasted child meat. = I haven’t tasted child meat for two months. The equivalence is represented in figure 1. The device used here by language is curiously reminiscent of the graphic illusions described by Gestalt theory (see Ninio 1998): an object that has no outline, which is only a void or a gap between other objects, is still clearly defined and visible. The type of measuring used by language could be termed ‘void’ or ‘gap’ measuring (in French: mesure en creux).

298 Claude Rivière since

EMPTY ( = not taste)

taste child meat

two months

NOW

not taste child meat

for two months

NOW

Figure 1. Equivalence it BE … since … / not V for

2.2. Choosing between the two readings (the repeatable/unique event criterion) The general principle is as follows: The reading is that of time location when p is stable, that is unique and non repeatable. Conversely, the “negative” reading, measurement of p’, stems from the referential instability of p, that is when p is not such that it is unique and unrepeatable. An event that cannot be repeated cannot lead to a negative reading (example 5): (5)

IT IS 500 years since Columbus discovered America — and across the Atlantic they are celebrating in true Hollywood style. BNC: The Daily Mirror (Aug–Oct 1992).

After the non repeatable discovery of America by Columbus, it is impossible to consider measuring what would be the time (extending up to now) during which he hasn’t discovered America. This is necessarily a case of location of the event. On the other hand, when the event is repeatable the negative reading appears, as in (6):

When Yes means No

(6)

299

This autumn the last of my children is off to university and I’m wondering if it’s realistic to consider entering the job market, since it’s 29 years since I worked. BNC: She magazine (Oct 1989).

The event is repeatable, but it is not repeated. The occurrence of the event at a certain point in time opens the period during which it is not repeated, even though it could be.

2.3. The factors that determine the stability of the event The factors that determine the stability of the event are varied: adverbs, verb form, adjectives, noun phrase determination. The adverb last is a safe marker of an event that could be repeated (unstable) but is not: (7)

If it is some time since you last exercised on a regular basis, spend a few weeks following a gentle stretching and warming-up routine. BNC: Ursula Markham, Your four point plan for life, (1991).

The present perfect in the since clause also leads to the negative reading: (8)

I’m happy. It’s been weeks since I’ve felt like this. Liza Cody, Backhand (1991).

This makes it incompatible with events that are unanimously recognised as unique: (9)

* It is over two hundred years since Mozart has died.

For a detailed explanation, see Rivière 2003. Noun phrase definiteness is decisive. Thus the presence of any, which, as an operator of ‘scanning’ (Culioli’s parcours) 1, makes the reference of the NP unstable, and leads to the instability of the event: (10) It’s a long time since anyone brought me flowers. BNC: Eamonn McGrath, The Charnel House. (1990).

300 Claude Rivière On the other hand, the ambiguity of the indefinite article makes for uncertainty, as in example (11): (11) It is 50 years since an unknown researcher discovered a miracle drug. If the article in (11) is specific (a certain unknown researcher, a certain miracle drug), the event is necessarily unique and is located in time: 50 years ago an unknown researcher discovered a miracle drug. If, on the other hand, the article is non-specific (an indeterminate unknown researcher, an indeterminate miracle drug), the event is repeatable, but hasn’t been repeated and this interpretation could be paraphrased by No unknown researcher has discovered a miracle drug for 50 years.

3. The construction it BE … before… 3.1. Description 2

At first sight, the construction it BE … before… looks like a symmetrical image of it be... since..., ... see example (12): (12) He added this warning “We, in the trade, feel that many people think that this is a new version of the whole Bible. It is, of course, only the New Testament: it will be many years before the Old Testament, and the Apocrypha are available.” LOB corpus In this example, the event the Old Testament and the Apocrypha are available is located in time through the measurement of the distance up to the point of view (in this case identical with the time of speaking). The difference with the it BE … since construction is, of course, that the event is in the future relative to the point of view. However, as with it BE … since, the point of the utterance is not to specify when the event will occur, it is rather to mention the length of time during which the event won’t be occurring. The “negative” meaning is brought out by the possibility of a negative paraphrase: (12a) The Old Testament and the Apocrypha won’t be available for many years.

When Yes means No

301

This is confirmed by the fact that the other paraphrase (plain location), though it is theoretically available, is unlikely in the context: (12b) * The Old Testament, and the Apocrypha will be available in many years. / many years from now. This paraphrase, however, becomes possible if the adverb only is inserted: (12c) The Old Testament, and the Apocrypha will only be available in many years / many years from now. The construction is also possible with the preterite: (13) BARTON’S MEN CUT the telegraph wires in half a dozen places, carrying away whole sections to make repairs more difficult. It was over an hour before their escape was discovered, but still the news that Barton was free flashed across the central portion of the state. Brown Corpus In this case, since the event of discovering is not located relative to the time of speaking , the event is not projected into the future: it is always an established fact. Again, what is put forward is not the location of the event but the duration of the non-event, plus a judgement on the duration (long or short). The acceptability of the paraphrases is the same: (13a) (13b) (13c)

Their escape was not discovered for over an hour. * Their escape was discovered over an hour later. Their escape was only discovered over an hour later.

So, whatever the tense used, it is always the duration of the non-event that is stressed. There is no ambiguity between two readings.

3.2. Asymmetry between past and future; the repeatable event criterion Since there is no ambiguity and the “negative” meaning is always apparent, the next question is the relevance of the unique event criterion. The answer is that in this case the criterion becomes irrelevant: the nature of the event is indifferent.

302 Claude Rivière It is the different position of the event relative to the point of view that causes the difference between the since and the before constructions: the prospective and the retrospective views are, unsurprisingly, not symmetrical, and the interval between the point of view and the event takes on different values. With a retrospective view, the distinction repeatable / unique is relevant because after a repeatable event (which has not been repeated) there is the absence of the event, which may be represented as follows: P ‘

P’

T0/T1 Figure 2. Retrospective view : virtually repeatable events

The dotted events are “ghost” events (virtual events), which might have occurred but didn’t. On the other hand, if the event is not repeatable, this zone is completely empty, no ghost events are conceivable. With a prospective view, the possible repetition of an event is pointless because repetition can only occur after the first occurrence, and the point of view is before the first occurrence. On the other hand, whatever the nature of the event, the space before the event is available for occurrences which do not take place:

P’

P

T0/T1 Figure 3. Prospective view : possible events that do not occur

When Yes means No

303

Figure 3, like figure 2, shows a zone with possible events that do not occur. That is the important feature the retrospective and prospective views have in common, and what causes the negative reading. However, the apparent symmetry of the two diagrams is deceptive and it hides the effects that result from the position of the point of view relative to the event.

4. Modal effects The it BE…since… construction with its time location reading is often linked to modal judgements that may be expressed by adverbs such as only or already, or none at all. The judgements may thus express either that the distance is too long or that it is too short. Example (14), with just, is a case where the distance is thought short by the speaker: (14) This was a major achievement,’ commented David Wootton, managing director of LASMO Nederland.’It is just 21 months since we received approval to go ahead with the project, and we are delighted that we have started up production within a few weeks of a date that many thought was unrealistic!’ BNC: Pipeline.u.p.

In example (15), now more than 12 months tilts the balance towards a ‘too long’ reading: (15) Does the Prime Minister recall that it is now more than 12 months since he first started dithering about the election date, since when 750,000 people have lost their jobs? BNC: Hansard extracts 1991– 1992]. Without an explicit marker, the assessment (shorter or longer than the personal norm of the speaker) may go either way, or there may be no assessment at all. Example (16) does not seem to contain any assessment: (16) IT IS 500 years since Columbus discovered America — and across the Atlantic they are celebrating in true Hollywood style. BNC: The Daily Mirror (Aug–Oct 1992). Example (17) is more complex:

304 Claude Rivière (17) LITTLE Laura Davies was spending her first full day back at home in Britain today. It is five months since she flew to America for a lifesaving bowel and liver transplant. Four-year-old Laura bubbled with health yesterday as she arrived at the family home in Eccles, Greater Manchester to cheers from neighbours. BNC: Liverpool Echo Daily Post. (1993). The second sentence may be seen as a factual explanation of the first sentence, but, if one goes on reading, in the light of what follows (bubbled with health), the same sentence takes on new meaning: the duration (five months) is seen as short, Laura has made a quick recovery. Thus, modal judgments are made clear through the confrontation with elements of the context, either in an explicit way or, more subtly, by comparing fragments of the text and the pre-conceived ideas they convey.

Things are different with a prospective view, and it BE….before…. The modal assessment is rarely explicit, and it is always “longer than could be expected”. This is confirmed in two ways: a. b.

the paraphrase with only that stresses the modal meaning: the event occurs later than expected. the cases with a negation in the main clause: the negation is then always associated with the adverb long.

(18) In view of the weather conditions, we’d maintained a full head of steam, so it wasn’t long before we were under way and steaming towards the other ship at our best speed. LOB corpus (19) It cannot be long before other major Commonwealth nations such as Canada, New Zealand and the former British colonies of the Caribbean follow suit’. This use of negated long shows that any positive term that appears in the same position means in fact “long”. Not long is the negation of any positive term, and, conversely, any positive term is a variation on long. The explanation of this – and of the negative reading – lies in the action of a hidden modality linked to the prospective view. With a prospective view, the event is expected (by the speaker). I will therefore call this modality expectation. When expectation is at work, the event must occur as

When Yes means No

305

early as possible, and any distance between the point of view and the event goes against expectation and is seen as a delay. For instance: (20) The results of photo-finishes were declared there in a couple of minutes, whereas at Epsom on Wednesday it was nearly 10 minutes before the judges could work out who won the fifth race. [The Independent,1989] For the speaker, there is a norm such that the result of a race immediately follows the finish, so any delay in the announcement is seen as too long compared to the norm. Such an utterance cannot be adequately paraphrased with a plain mention of time location: (20a) * At Epsom on Wednesday the judges worked out who won the fifth race after nearly 10 minutes. On the other hand, adding only makes this acceptable: (20b) At Epsom on Wednesday the judges only worked out who won the fifth race after nearly 10 minutes. It must be noted that the expected event is not necessarily favourable, for example: (21) With this medicine, it will be four hours before you feel any pain. This can also be paraphrased with only: (21a)

You’ll only feel pain in four hours’ time.

Feel pain, which is unfavourable, is expected but, thanks to the medicine, is supposed to occur later than the “norm”. At this point, we have two hidden modalities: (a) in both constructions, the negation, which does not result from the same causes in the two cases; (b) in the it is…before… construction, the modality of expectation. The modality of expectation deserves a closer look, especially as there exists a third construction that combines an inversion of polarity with expectation, the not V for… construction.

306 Claude Rivière 5. More evidence for the modality of expectation 5.1. The construction not V for… Another construction that we have already come across (not V for…) gives support to the modality of expectation. Here are two examples: (22) The two women thought so little of it at the time that they did not discuss it for a week; did not write down any account of it for three months; did not publish it to the world until ten years later. LOB corpus (23) But the Government has said public money will not be available for at least 15 years. OXFORD–A corpus The construction measures a negative event whose beginning may be either the time of speaking or some point in the past. This is not, however, only a measurement: when in the future there is a strong implication that the nonevent is brought to an end by the occurrence of the event, and when in the past, it is more than an implication. Positive paraphrases that locate the event are possible, but they are only really adequate with the adverb only (the b examples): (22b) (23b)

They only discussed it a week later. Public money will only be available in 15 years time.

Another paraphrase is, of course, available with it BE…before… (the c examples): (22c) (23c)

It was a week before they discussed it. It will be 15 years before public money is available.

So, we have come full circle. The two constructions can be represented by two complementary diagrams:

When Yes means No

307

It was a week before they discussed it. P P’

T1 They did not discuss it for a week P P’

T1 Figure 4. Equivalence it BE….before…. / not V for…

An interesting example shows how close the meanings of the two constructions are. The two constructions are used in succession, first in the headline then at the beginning of the article (24) [headline] Jail rioters may not go to trial for two years [beginning of article] HOME Office sources have indicated that it could be two years before charges against the rioters at Strangeways prison come to trial. The news came as a huge police investigation began into the 25-day siege at the Manchester jail. [The Times (Apr 1990)] The modal effects of not V for… are the same as those of it BE… before, though the modality of expectation applies here to an event which, curiously enough, is not explicitly stated. Consequently, the event is seen as further in time than the speaker’s norm allows. Here is another example: (25) He [Louis XVI] also lacked the sensuality of the earlier Bourbon monarchs: his marriage was not consummated for seven years, and he had to have a ligament cut before he could have a proper erection. [The Independent (1993)] Here, even though the speaker knows what eventually happened (the marriage was consummated), he/she applies a norm according to which the

308 Claude Rivière consummation of a marriage closely follows the marriage. Any delay in the actual occurrence of the projected event is then too long. The usual paraphrases can be given: (25a) (25b)

His marriage was only consummated after seven years. It was seven years before his marriage was consummated.

The implication of the event through the mention of its negation is so strong that the construction can be used in cases where, out of context, the negation of the event seems nonsensical: (26) Columbus did not discover America for two years. The meaning, of course, is not that Columbus spent some time not discovering America or failed to discover America over a period of two years. This utterance takes on meaning if the speaker, placing his point of view in the past and before the event, expects the event, even though fictitiously.

5.2. Only and its paradoxes Expectation can be seen at work in the utterances with only which I have been using as acceptable paraphrases throughout this paper. Compare the effect of only in the following examples, in the future and in the preterite: (27a) (27b) (28a) (28b)

The Old Testament will only be available three months from now. I’ll only see him on the 21st. The Old Testament has only been available for three months. I only met him in 1999.

These examples display two curious features. In the (a) examples, when the view is prospective, the meaning is (as expected) that the length of time is longer than expected. When the view is retrospective, however, the distance between the event and NOW is shorter than might have been expected. See figure 5. Moreover, in the (b) examples, only behaves in an unusual way (see Declerck 1994 for another account of this fact). Normally, when only is applied to a series of numbers, it selects the smallest number: I only have ten euros, not eleven. Here – in examples 27 and 28 –

When Yes means No

309

the 21st is selected to the exclusion of a smaller number, e.g. the 20th , and in the same way selecting 1999 excludes 1998, or any previous year. It is the combination of the modality of expectation with only that accounts for those paradoxes. Only selects a term out of a series, in the present case a series of time points available for the occurrence of an event. In accordance with the modality of expectation, any expected event must occur as early as possible. Now, as we have seen, the real event is preceded by ghost events (that might have occurred), but only selects the real event, even though it stands further in the series (the 21st rather than the 20th). When the view is prospective, the event is thus (quite normally) further from the point of view. However, when the view is retrospective, expectation is still at work: the actual event again occurs later than the ghost events (later than expected), and, again, is selected by only (1999 rather than 1998). In this case, the event is then closer to the point of view than expected.

A expected event

A real event

T1

b expected event

B real event

Figure 5. Only, prospective and retrospective prospective view: T1–B (real) longer than T1–b (expected) retrospective view: T1–A (real) shorter than T1–a (expected)

5.3. The importance of syntax and its interpretation The construction have not V-en for…, which has been used as a paraphrase of it is…since… throughout this paper, deserves some closer attention. It offers an intriguing contrast with not V for… (with the future or the easerit), since it does not imply p through the mention of p’: I haven’t worked for years does not imply I worked years ago as opposed to Columbus did not discover America for two years which implies Columbus discovered America two years later. This is negative evidence for the modality of expectation. Expectation is linked to a prospective view, which is not part of the have not V-en for…, construction. The negative event is viewed

310 Claude Rivière from a point later in time and the meaning is that given (quite regularly) by the perfect: the interest is focused on the (present) point of view and not on the onset of the negative state. At the same time, and as a consequence of the absence of expectation, the have not V-en for…, construction does not involve any modal values and the paraphrase with only is not acceptable: *I only worked years ago. This may not be thought a sufficient explanation for the exceptional properties of have not V-en for… as it is…since…(with retrospective view) does imply p’ through the mention of p. The answer lies in the relation between the syntax and the interpretation of the syntax. This paper has been dealing with two types of constructions, A and B, as shown in Table 1, which also displays their other properties, including the modality of expectation. Table 1. Syntax and interpretation

A One clause

B two clauses

1. prospective view, expectation Columbus did not discover America for two years p’ located relative to duration p’ stated, p implied chronological sequence: p’ – (p) It was two years before Columbus discovered America. Duration located relative to p p stated, p’ implied chronological sequence: (p’) – p

2. retrospective view I haven’t worked for twenty years p’ located relative to duration no implication only p’ It is twenty years since I’ve worked. duration located relative to p p stated, p’ implied chronological sequence: p – (p’)

The B constructions can only state a positive event and can imply a negative polarity. It cannot be otherwise, as a consequence of their syntactic structure. Their syntax must be compared with that of the A constructions. Through their syntax (NP Negative VP for NP = duration) the A constructions locate a negative event relative to a duration. This translates as the association of duration to the event. The syntax of the B constructions (it BE NP = duration since / before positive S), on the other hand, represents the location of a duration relative to an event. While the location of an event relative to a duration is an obvious operation, the opposite may not be quite as clear. When locating a duration relative to an event, the locating event sets a definite boundary (onset or endpoint) to the duration. This requires that the event be positive, as only a positive event can be a

When Yes means No

311

point in time. The B constructions, therefore, can only mention positive 3 events . The it BE…since construction can then imply the opposite polarity thanks to the repeatable event effect and mesure en creux (“gap” easureing), an effect that cannot apply to have not V-en for… (A type) which is the banal measuring of an event en plein, and is not different from its positive counterpart have V-en for… (I have worked for twenty years), for which no one would ever imagine the possibility of a negative implication (? I didn’t work before that). Table 1 also shows the conditions that make it possible for the A constructions and the B constructions to be paraphrases of each other. Type A and type B differ in a complementary way at several levels: – – –

Order of the operations of location Stated polarity Number of clauses

The fact that the B constructions are made up of two clauses enables them to invert the order of the operations of location: duration located relative to event instead of event located relative to duration. The stated polarity is also inverted (positive instead of negative), so that, eventually, the duration is located relative to a positive event, while in the A constructions a negative event is located relative to the duration. The double inversion results in the relation of paraphrase between the two types.

6. Conclusion The first goal of this paper was to study the (paradoxical) case of positive constructions that mean the opposite of what they state explicitly, the negation being then a hidden modality. The interpretation of these constructions – that is the chain of operations that leads to a negative reading – depends on various factors that are – not unexpectedly – different when the view on the event is prospective and when it is retrospective. When the view on the event is retrospective, the nature of the event is essential: is it repeatable or unique. This criterion, however, is of no value when the view is prospective. This led to the discovery of another hidden modality, expectation, whose existence is confirmed by a consideration of the not V for construction and of the behaviour of the adverb only.

312 Claude Rivière The next step would be to consider the use of the constructions in discourse. How far are constructions that are paraphrases of each other adequate substitutes of each other in a given context? The competing constructions should be considered from the point of view of discourse coherence, and more specifically in relation to the informational structure of utterances.

Notes 1. Scanning: “an operation in which the utterer scans one by one all the elements in a class without singling out any particular element (scanning consists in running over the whole domain without being willing or able to pick out one (or more) distinguished values. (…) The scanning operation posits that each element in the class is identical with respect to the property of being able to validate the relation in question: the class is considered as being homogeneous.” Bouscaren et al. (1992: 109). See also: Culioli (1990: 121,171 and 183). 2. Declerck 1995 deals with a related construction, it BE… until, from a different perspective. 3. I have come across one case of a negative event in this construction: Almost dark now, and the bombers would be taking off from Fenton Bishop, heaving and thrashing into the air, into the same sun that had reddened Loch Ardneavie. How long would it be, Jane demanded silently of the darkness, before she was able to speak of Rob or think of Rob and not weep inside her? It was nearly two months since her call-up papers had come; two months since S-Sugar had not come back. She had been angry at first, then disbelieving, but finally she had gone to the aerodrome to see the padre, begging him for news of Rob. Elizabeth Elgin, All the sweet promises (1991), BNC. In this exceptional example, not come back can be conceived of as a point because the plane’s not coming back at the expected time is (for the female character) an event that can be precisely placed in time. It is, in any case, an instance of time location and not the measurement of the duration of an event.

When Yes means No

313

References Bouscaren, Janine, Jean Chuquet, and Laurent Danon-Boileau 1992 Introduction to a Linguistic Grammar of English – An UttererCentered Approach, translated and adapted by Ronald Flintham, and Janine Bouscaren. Gap: Ophrys. Culioli, Antoine 1990 Pour une Linguistique de L'énonciation: Opérations et Représentations, Tome 1. Paris: Ophrys. Declerck, Renaat 1994 The only/already puzzle: a question of perspective. Cognitive Linguistics 5–4: 307–350 1995 The problem of not…until. Linguistics 22: 51–98 Ninio, Jacques 1998 La Science des Illusions. Paris: Odile Jacob. Rivière, Claude 2003 La construction it is… since…: repérage, mesure et stabilité référentielle. Anglophonia 14: 21–35

Modality and the history of English adhortatives Manfred Krug

1. Aims and issues Seppänen (1977) provides an early treatment of the relationship between let and the English auxiliary system. More recently, adhortative let’s has become a paradigm case of grammaticalization (cf. Hopper and Traugott 2003; Krug 1994; van der Auwera and Taeymans 2004). While most recent studies concentrate on present-day variation, this paper’s focus is on the history of English adhortatives, including the history of their negation patterns. Let constructions will take centre stage, but older layers – i.e. the preterite present verb uton and adhortatives expressed by subjunctive mood – will figure in the discussion too. In addition, the scope of the present paper is wider in that it also considers first person singular and third person let imperatives. The ensuing analysis will shed some new light on old issues in grammaticalization theory, in particular as regards the role of pragmatic inferencing in semantic change, and the debate on layering vs. gap filling. Before we can enter the data analysis proper, however, we need to consider the place of adhortatives and imperatives in current thinking on modality.

2. Adhortatives, imperatives, mood and modality Let constructions of type (1) and (2) are the focus of the present paper. Both belong to the class of adhortatives (or hortatives), whose more explicit (but less elegant) designation is imperative of the first person plural inclusive. (1) (2)

Let’s go! (1st person imperative plural = (ad)hortative) Let us go! (2nd person imperative: ‘Allow us to go.’ or adhortative above)

316 Manfred Krug As is well known, the source construction – given in (2) – is ambiguous in present-day English because it has retained the older 2nd person imperative reading (‘Allow us to go’); the contracted form let’s, by contrast, does not allow for this interpretation and thus is unambiguously adhortative. To clarify the relation between adhortatives and the domains of mood and modality, some terminological issues need to be addressed in this first part of the paper. The most general category into which adhortatives fall is the pragmatic speech act category of directives (cf. Matthews 2003). Directive is a discourse function encompassing all utterances which serve to elicit some action from the addressee(s). Directives can be realized grammatically through imperatives (such as Open the window!), lexically (e.g. please) or phonologically. For instance, declarative sentences (such as You are going to pick up grandma at the station) or questions (such as Can you pick me up at the station?) are directives, given the appropriate context and intonation. Directives therefore include both direct and indirect speech acts. The central criterion for classification as directive is the illocutionary force of an utterance, in other words the pragmatics of the speech act – whether or not the speaker’s purpose is to get the addressee(s) to do something. Intermediate between the broadest category of directive and the narrowest category of adhortative is that of imperative. As stated above, imperatives are defined as expressing a directive through grammatical means (e.g. Crystal 2008: s.v. imperative), though this will shortly be seen to be ambiguous. In a large number of languages, imperatives are realized through the absence of a subject; in addition, the prototypical English imperative (the second person imperative) contains the base form of the verb, as in Be quiet! In some approaches, such second person imperatives are called orders or commands, and in narrow definitions of imperatives, the three terms are synonymous. However, under a wider (and I believe more useful) definition, imperatives are not confined to second person imperatives, for there exist also first person imperatives (adhortatives) and third person imperatives (or jussives). In English, two of these – first and third person imperatives – can occur with or without let. In other words, a common way to express imperatives in English is by means of a periphrastic construction involving let. For convenience, Quirk et al.’s (1985: 832) overview of imperatives is reproduced below.

Modality and the history of English adhortatives

317

Table 1. Summary of forms of imperatives (Quirk et al. 1985: 832) 1st PERSON without subject without let with subject with let

2nd PERSON

3rd PERSON

–

Open the door.

–

–

You open the door.

Someone open the door.

Let me open the door.

–

Let someone open the door.

Let’s open the door.

Why should imperatives be considered to fall within the domain of modality? Part of the answer lies in the fact that imperative is traditionally regarded as one of three moods, the remaining two moods being indicative and subjunctive. However, to answer the question just posed in a principled and crosslinguistically applicable way, it is not enough to classify imperative as a subcategory of mood. Instead, it seems useful to distinguish further between syntactic and inflectional mood (as does Mitchell, this volume). Though the terms are largely self-explanatory, a short explanation is as follows: Inflectional mood is mood expressed inflectionally; syntactic mood, by contrast, refers to a clause type or sentence function. On this view, all English imperatives – including periphrastic constructions involving let or modal verbs – are instantiations of syntactic mood, while only second person imperatives of the type Be quiet! are instantiations of inflectional mood. The second part of the answer for why imperatives should be considered within treatments of modality lies in a recent reconsideration of the relationship between mood and modality. In Palmer (2001, 2003), modality is seen as the superordinate category comprising all instantiations of mood and, with minor qualifications, every utterance that falls short of a plain declarative or statement of a fact. Witness Palmer’s most recent position on the distinction between mood and modality: The investigation of a large number of languages suggests that what has traditionally been called ‘mood’ is only one type of grammatical sub-category within a wider grammatical category [viz. that of modality]. … With mood

318 Manfred Krug there is a feature that is, so to speak, either on or off. This might be described in terms of ‘modal’ and ‘non-modal’, the indicative being non-modal and the subjunctive being modal. (The terms ‘realis’ and ‘irrealis’ for this general distinction are, perhaps, preferable – so that the indicative is realis and the subjunctive is irrealis, but the more familiar ‘modal’ and ‘nonmodal’ will be used for simplicity). (Palmer 2003: 2) Prima facie it [modality] might seem to be concerned with whether the proposition is true (or factual) or not, but this is not entirely satisfactory. More convincing … [is] that the essential feature of modality is that of ‘assertion’ vs. ‘non-assertion’. (Palmer 2003: 5)

Adhortative and jussive let constructions are appeals for future action. According to Palmer’s definition, they would be classified as ‘modal’ (or ‘irrealis’). The preceding considerations used semantic and functional arguments for the inclusion of let constructions within the realm of modality. Further arguments in favour of classifying all imperatives within the domain of modality can be gleaned from comparing English to other languages and their ways of expressing imperatives for different persons and numbers. A synthetic language like Latin, for instance, realizes all of its imperatives through inflectional mood: two of them – jussives and adhortatives – through the use of subjunctive (inflectional) mood;1 second person imperatives, by contrast, are realized through imperative (inflectional) mood: (3) (4) (5) (6) (7)

i ‘Go!’ (2nd ps. sg.) ite ‘Go!’ (2nd ps. pl.) eamus ‘Let’s go’ (adhortative, 1st ps. pl. inclusive) eat ‘He shall go’/ ‘Let him go’ (jussive, 3rd ps. sg.) eant ‘They shall/may go’/ ‘Let them go’ (jussive, 3rd ps. pl.)

Latin examples are instructive in so far as the English equivalents sometimes express the same function inflectionally, as in the second person, while at other times the English equivalents require periphrastic constructions, be they let or modal verb constructions. If we adopt an onomasiological perspective, which starts out from a function and investigates all expressions used for that function, then there is no justification for considering modal verb constructions like He shall go! as falling within the domain of modality, while classifying a rough semantic equivalent (Let him go!) as falling outside this domain.

Modality and the history of English adhortatives

319

Even these few examples from just two related languages suffice to show that for a definition of modality to be crosslinguistically valid and applicable, it must be able to encompass three such closely related functional types as first, second and third person imperatives under a single heading. This can only be achieved if imperative and subjunctive inflectional mood fall within a common superordinate category. It is suggested here, in line with recent approaches, that this superordinate category be (or, to illustrate this point: should be) labelled syntactic imperative mood, which in its turn is to be considered as part of the broader domain of modality.2 In other words, since subjunctive inflectional mood in various languages is used to express imperative function, i.e. syntactic mood, there remains little reason to exclude the same syntactic mood (as found in let adhortatives and jussives) from discussions of modality in English. Perhaps the best way of summarizing and justifying the present classification is by clarifying the following relationship of entailment: a) b)

Subjunctive and imperative inflectional mood are both subcategories of syntactic mood. Syntactic mood (when ‘on’, i.e. when not indicative/realis) is a subcategory of modality.

It follows that both inflectional and syntactic imperative mood are subordinate categories of modality. This can be represented in set theory as (8) or schematically as in Figure 1 below. (8)

adhortative ⊂ imperative ⊂ syntactic (non-indicative) mood ⊂ modality

The entailment relation reads as follows: If A is a proper subset of B, and if B is a proper subset of C, then A is also a proper subset of C. Applied to the present case, which has a fourth set, the relationship is as follows: Adhortatives form a proper subset of imperatives. These are a proper subset of mood expressions. Expressions of syntactic mood, finally, are a proper subset of modality expressions. Adhortatives are thus a proper subset of modal expressions.

320 Manfred Krug

adhortative

imperative

syntactic mood

modality

Figure 1. Adhortatives, imperatives, mood and modal expressions (represented within set theory)

Needless to say, essentially the same argument as for let adhortatives goes for let imperatives of the type Let me/Lemme tell you something, and likewise for jussives (Let him/them/’em open the door). Each of these is ambiguous between second person imperatives on the one hand, and first and third person imperatives respectively, on the other. Both constructions, however, deserve more detailed investigations than can be offered within the scope of the present paper.

3. The history of English adhortatives – a gap filled? The following chapter sketches the semantic, pragmatic and grammatical changes that adhortative let constructions underwent since their first occurrence in Middle English. Their present-day variation has become a popular topic since 1993 due to the publication of the first edition of Grammaticalization (Hopper and Traugott 2003, 2nd ed.), which included a seminal sketch of adhortative let’s. Follow-up studies include Krug (1994), van der Auwera and Taeymans (2004) and De Clerck (forthcoming a, b).3 The historical details, however, though outlined in Hopper and Traugott (1993 [2003]: 10-13) and Traugott (1995: 36-37), are still largely unknown. It is this gap that the present chapter intends to fill. Old and early Middle English had two options for expressing adhortatives: (i) inflectional subjunctive mood, analogous to the Latin example (5) above and illustrated for English in (11) or (19) below; and (ii) a periphras-

Modality and the history of English adhortatives

321

tic construction involving a preterite present verb: uton ‘let’s’, which later eroded to uten and ute. Preterite presents belong to a class of verbs from which all modern central modals – like may, must, can etc. – originate.4 The fact that uten belonged to the same paradigm adds another structural argument for the inclusion of adhortatives in the domain of modality. As regards the older subjunctive adhortative, it was argued in the previous section that inflectional mood is a subcategory of modality. Hence, the shift in adhortatives from subjunctive inflectional mood to a periphrastic (let) imperative presents no theoretical problem: As imperatives, both constructions are expressions of syntactic mood, and thus modal in nature. The preterite present verb uten fell into disuse in late Middle English (Warner 1993: 186). Two examples are given below, the second being the latest attestation in the Oxford English Dictionary (hereafter OED).5 (9)

c1000 Ags. Gosp. Matt. xvii. 4 ʒ yf þu wylt, uton wyrcean her þreo eardung-stowa. ‘If you want to, let’s make here three dwelling places’

(10)

c1275 Passion of our Lord 173 in O.E. Misc. 42 Ariseþ vp… and vte we heonne go. ‘Get up [2nd ps. IMP]... and let’s go from here’

An important question in grammaticalization theory is whether or not new constructions fill gaps (see the discussion in Hopper and Traugott 2003: 126). According to Hopper and Traugott (2003: 11) and Traugott (1995), the let adhortative has been available since the late 14th century. Since the last attestations of uten date from the late 13th century, this might seem to be a case where a new construction enters the language due to a functional gap. The next section will explore this issue in more detail.

3.1. Semantic-pragmatic changes of let Traugott (1995: 37) presents as the earliest clear example of adhortative let a passage from Chaucer’s Canterbury Tales. For convenience, her example is reproduced as (11) below. It illustrates both the old (2nd ps. imperative) and the new (adhortative) let constructions. Importantly, the new let construction is coordinated with a subjunctive adhortative, which underpins the adhortative reading.

322 Manfred Krug (11)

Com doun to-day, and lat (2nd ps. IMP) youre bagges stonde… What, lat (ADHORT) us heere a messe, and go (SUBJUNCT ADHORT) we dyne. ‘Come down today and let your bags stand (leave your bags)… Let us hear a mass and let us go dine.’ (c. 1387 Chaucer, CT, Shipman’s Tale 1410)

I will argue in what follows that while the first clear cases of adhortative let indeed date from the 14th century, there are precursors which motivate the emergence of let adhortatives through the exploitation of pragmatic inferences. Let us, for this purpose, consider the following examples: (12)

a1225 Leg. Kath. 812 Scheoteð forð sum word, & let us onswerien. ‘Shoot forth (utter) some word and let us answer.’

(13)

c1250 Death 261 in O.E. Misc. 184 Lete us hatie þat woh and luuie þat riʒ te. ‘Let us hate the evil and love the right.’

The context of (12) suggests temporal sequentiality of the two clauses, and it is therefore most probably an instance of the older permissive/causative construction, with the permissive ‘allow to’ reading being more likely. Example (13), by contrast, is ambiguous. Uttered in a religious context, it has two addressees: If we go by the book, one is second person singular (God/Christ/Trinity). This addressee suggests the older permissive/causative reading, with the more likely reading here being causative ‘make us hate/love’. The second addressee in (13) is the group of believers (e.g. the community in church, prayers etc.) of which the speaker is part. Hence there is a secondary first person inclusive addressee, which yields a backgrounded adhortative reading. Such examples are not uncommon. Compare (14), which explicitly addresses Christ, but implicitly the community of all Christians: (14)

c1340 Cursor M. 23208 (Trin.) Crist let vs neuer be in þat fere. ‘Christ, let us never be in that companionship.’

Modality and the history of English adhortatives

323

Some forty years later, i.e. in Chaucer’s work, clear cases of adhortative let constructions are legion. Nevertheless, some of these examples can be interpreted as permissive 2nd person imperatives: (15)

c1374 Chaucer Troylus ii. 222 lat us daunce, And cast youre widewes habit to mischaunce!

(16)

c1374 Chaucer Troylus v. 395 Let vs speake of lustie life in Troy…

(17)

c1384 Chaucer H. Fame 294 Let us speke of Eneas…

(18)

c1386 Chaucer Melibeus 383 But now let us speke of the counseil…

(19)

c1386 Chaucer Man of Law’s T. 855 Lat vs stynte of Custance but a throwe, And speke we [SUBJUNCT ADHORT] of the Romayn Emperour. ‘Let’s stop (talking) about Constance for a moment and let’s speak of the Roman Emperor’

Example (19) is especially interesting, because – like (11) above – it coordinates adhortative let us with a subjunctive adhortative. It is interesting for a second reason, in that it suggests – together with examples (16) to (18) and the earlier (12) – that verbs of speaking played a seminal role in the change from 2nd person permissive let us to 1st person adhortative let us. On second thoughts, this is not as surprising because a speaker uttering “Let us speak” performs two speech acts simultaneously: One is asking his/her interlocutors for permission to speak. On the other hand, this is a context in which pragmatic inferences can easily arise because while it is polite to ask for permission to speak, often the turn taking conventions are such that a change of speaker is indicated by the context anyway. The second speech act, arising by pragmatic inferencing, is thus to express the intention to speak on behalf of a group which the speaker is a representative of. In other words, asking for permission to speak is not always required and thus to a certain extent redundant, triggering pragmatic inferences, which can become semanticized over time. Thus, pragmatic enriching in combination with verbs of speaking is likely to have motivated

324 Manfred Krug the semantic ambiguity of present-day let us constructions. Such a development can be integrated into a minimally modified version of Traugott’s (2000) model of semantic change by invited inferences given in Figure 2. The modifications concern exclusively the transfer or extension of the model from lexemes to constructions. Utterance-type meaning Conventionalizing of implicatures SP/AD constraints on weighting of implicatures (preferred uses, saliency, relevance, subjectivity, etc.) Utterance-token meaning SP exploits implicatures innovatively

(semanticization) Stage I. Coded meaning

Stage II. New coded meaning

Construction → Meaning 1

Construction → Meaning 1 + Meaning 2

let us

permissive/ causative

permissive/ + adhortative causative

Figure 2. Adaptation for constructions of Traugott’s (2000) model of semantic change by invited inferences (cf. also Traugott and Dasher 2002: 38)

The development sketched above has implications for the question raised at the beginning of this section, viz. whether we are dealing with an instance of gap filling. It was argued above that pragmatically enriched let us constructions such as (13), dating from 1250, are available at a time when uton is not yet obsolete (see the last OED attestation, quoted as (10) above, which dates from 1275). Hence this cannot be a genuine case of gap filling. However, the answer is not a straightforward rejection of a gap filling scenario either. I would take a cautious stance, not only because first and last OED citations are not identical to emergence or extinction of a word or construction (though the OED is unparalleled in any other language and generally offers a fairly good approximation). More important is that before a construction dies out, it obviously decreases in frequency. There can be little doubt, uton/ute(n) was obsolescent in the 13th century. The sub-

Modality and the history of English adhortatives

325

junctive – the second possibility for expressing adhortatives – was a recessive option too, though by no means as near extinction as uton. The rise of adhortative let us/let’s, then, is probably best seen as an example of layering, and as part of a more general drift towards analyticity in English, since an analytic construction replaces an older inflectional one, the subjunctive. True, analytic let us equally ousted another analytic construction (uton). Two factors seem likely to have played a role in this development: One is the ever increasing homogeneity of the modal verbs and the concomitant loss of non-deontic or non-epistemic preterite present verbs (cf. the loss of witan ‘know’; Warner 1993); the other is a higher degree of synchronic transparency in adhortative let us constructions, where a common verb (rather than an obsolescent, eroding and thus opaque one as in uton > uten > ute) is followed by an obligatory (rather than an optional) first person pronoun. In addition, the incoming use of let adhortatives ties in with a typological trend of the English language: an increasing dispreference for subjectless constructions, as is also exemplified by the roughly contemporaneous decline in impersonal constructions.6 While it is true that us (as marked for objective case) is prima facie not the syntactic subject of the adhortative construction, it nevertheless is the logical subject and (understood) agent of the adhortative. Syntactically it can be analysed as an apokoinou constituent, i.e. as both the object of the let clause and the subject of the adhortative. Recall a simple example: (2)

Let us go.

For one thing, case-marked pronominal forms in subject (or subject complement) position are quite common in English. Compare: (20)

Who wants to go? – Me.

(21)

Who’s there? – (It’s) me/us.

One might object that oblique forms in subject or subject complement position tend to be stressed forms and post- rather than preverbal. While the latter point is granted, the answer to the former is that us was not lenited when let adhortatives emerged. It cliticized only 200 years later (see next section). Structurally more similar to adhortative Let us go are raising constructions like (22), traditionally known as ACI (or accusative with infinitive) constructions; or nonstandard but common zero relatives like (23):

326 Manfred Krug (22)

I saw him go.

(23)

It was me left the room.

Of these (22) is particularly similar to adhortative let us because the casemarked pronoun functions at the same time as the object of the preceding verb and as the subject of the following verb. In summary, this section has suggested that two syntactic and pragmatic contexts facilitated the transition of let us constructions from second person permissive/causative imperatives to first person inclusive imperatives (i.e. adhortatives). The first is their frequent early occurrence in religious contexts, where the speaker addresses both God/Christ and the Christian community at the same time. The second is the early preponderance of verbs of speaking after let us constructions. It was argued that in such contexts speakers both ask their interlocutors for permission to speak and at the same time express their intention to speak for a larger group. The overall development is one from a main verb (permissive/causative) to a modal verb construction (adhortative). Cognates of let have undergone similar, though not identical, developments in other Germanic languages, for instance in German and Dutch (on the latter, see van der Auwera and Taeymans 2004). Notice, however, that of the so-called ‘main’ verb uses, permissives would qualify as quasi-modal or emerging modal constructions in more recent approaches (e.g. Traugott 1997; Krug 2000; van der Auwera and Taeymans 2004) because it has deontic semantics and takes infinitival complements.

3.2. Formal and semantic changes of us With let’s, Lehmann’s parameter of attrition (i.e. loss of phonological substance, cf. Appendix) is evident from the encliticization of us. As is argued by Hopper and Traugott (2003: 10-13), and as will be illustrated in some more detail in what follows, cliticized us is also losing substance on the semantic side. It is no longer restricted to first person plural, but slowly proceeds to all persons and numbers. Examples will be provided not only from spontaneous spoken English but also from books and magazines (which involve planned production and proof-reading), so that they cannot be easily dismissed as performance errors. In a first step, however, I shall treat phonological erosion.

Modality and the history of English adhortatives

327

Phonological erosion from us to ’s has semantic and syntactic implications: As soon as let is followed by clitic ’s rather than full form us, syntactic reanalysis is complete, since the older ‘allow us to’ reading is no longer available. Compare: (24)

1584 Lyly Sapho II. iii. 108 To th’ Tap-house then lets gang, and rore.

(24)'

To th’ Tap-house then let us gang, and rore.

Example (24) has only the adhortative reading, while (24)' is ambiguous in that it has the additional reading ‘Allow us to go to the tap house and revel boisterously.’ Contracted adhortative let’s ∼ in the late 16th century is not uncommon and found with different authors: (25)

1584 Lyly Sapho IV. iv. 33 Vulcan My shag-haire Cyclops, come, lets ply Our Lemnion hammers lustily.

(26)

1588 Shakespeare Loves Labours Lost V. ii. 228 If you denie to dance, let’s [~ lets] hold more chat.

(27)

1590 Lodge Euphues Gold. Leg. (1609) 54 Lets haue a little sport with him.

It was seen in the previous section that unambiguous adhortative let constructions are attested from the late 14th century. The examples above show that it takes 200 years until the first attestations of clitic forms can be found. I have argued elsewhere (Krug 2000: 177) that phonological erosion is a direct consequence of increasing discourse frequency (cf. similarly Bybee 2003). By inverse implication, one can conclude from the first phonologically reduced and thus more closely bonded occurrences of let’s that adhortative let constructions are likely to have seen drastic frequency gains in the 16th century. This assumption is confirmed by OED frequency counts for let us and lat us occurrences, normalized per 10,000 citations. The graph shows that the text frequency of let adhortatives tripled in the middle of the 16th century. 7

328 Manfred Krug 8 7

instances per 10,000 quotes

6 5 let/lat us per 10,000 quotes

4 3 2 1 0 1351-1375 1376-1400 1401-1425 1426-1450 1451-1475 1476-1500 1501-1525 1526-1550 1551-1575 1576-1600 25-year period

Figure 3. Let us in the OED from 1350-1600

In spoken present-day English, cliticization of us is near-categorical. In the spoken BNC, 245 let us constructions compare with 4148 let’s constructions. Put differently, 94.4% of all us occurrences are cliticized. The remaining 5.6% include a number of ‘allow us to’ readings. In other words, there is little stylistic differentiation for the adhortative. No matter the level of formality in spoken British English, with the notable exception of religious formulae like Let us pray (cf. Note 12), the adhortative is almost always let’s rather than let us. This development exemplifies Hopper’s (1991) principle of split (or divergence), which predicts that lexical and grammatical uses of a given construction will diverge formally over time. In this case, adhortative let’s contrasts with permissive/causative let us. Erosion does not stop with the cliticization of us. As Hopper and Traugott (2003: 12) have shown, clitic ’s can stand for the entire let’s construction in present-day English. This seems to be particularly common with highly frequent sequences, resulting in forms like ’sgo, ’sfight, ’sdo it. Allowing for some simplification, the overall development of let us constructions is the following: (28)

full verb construction

> auxiliary construction >

let us ‘allow us to’ speak >

let’s go

>

clitic/affix ’sgo

Modality and the history of English adhortatives

329

3.3. Details of the generalization and desemanticization of us Starting from Hopper and Traugott’s (2003: 10-13) analysis, the data to be presented shortly will confirm their assumption that us is currently being generalized. The desemanticization of us and, in particular, clitic ’s will be demonstrated by an extension of use beyond first person plural reference. Let’s refers to first person plural but is supported by lexical information The least exciting hint at semantic loss is lexical emphasis indicating that let’s denotes inclusive first person plural. Common items are each, both (of us), you and me, you and I, and indeed us. On semantic grounds, this would be redundant if enclitic ’s in let’s had retained the full semantic load of us. One might object that let’s each and let’s both (of us) are just common reinforcers or that they are merely substitutes for over-formal let us both/each. But that still raises the question of why emphasis or explication is necessary in the first place. Additional information seems especially superfluous where let’s itself is clearly dual, for example, when there is no third person present. There is ample evidence of this in spoken and written components of the Bank of English Corpus (BEC), both American and British components:8 Support from both and each (29)

Ta-ta let’s both of us hope they locate whatever it is they’re looking for. (Brit books 0111)

(30)

But inherent in some relationships is a kind of blanket lying policy; let’s both pretend to be other people – infinitely cheery, allgenerous, ever-sexy – so our love is never threatened. (magazines 0375)

(31)

Hansen: I don’t know # Let’s both open up and see what we get, OK. (Am radio 29/11/92)

(32)

So let’s each say her [sic!] mea culpa and start the year 1884 under new auspices. (Am books 530)

(33)

Silvia said, “let’s each go to different ones and compare notes later.” (Am books 433)

330 Manfred Krug Support from you and me, you and I, me and you Three options of lexical support – you and me, you and I, me and you – are particularly likely to occur where the speaker emphasizes the dual use of let’s. Here the emphasis is generally stronger than in the above examples, although you and me etc. – unlike both (of us) – are not logically restricted to dual number: (34)

Their idea of couth is saying anything other than, “Hey kid, let’s me and you boogie” (Brit books 0037)

(35)

“Tell you what,” said D’or: “Let’s you and me go see this Laurie person at the boys’ camp.” (Am books 0455)

(36)

“I’m ready whenever you are.” “Then let’s you and me go out to the garage. Ernie, don’t...” (Am books 0462)

(37)

“Let’s you and me sit down and talk.” (overheard in the USA)

(38)

“... He’s been phoned and he’ll come in and take charge of the medical end of things. Let’s you and me examine the body first.” (Brit books 0037)

In examples (39) to (42) let’s is followed by you and I. This sometimes reflects attempts at a slightly more formal way of expressing the same kind of emphasis as the oblique case you and me. Somewhat ironically, however, given the right intonation (or punctuation, i.e. in parentheses), oblique case pronouns are grammatically well-formed according to prescriptivist rules, while subjective case is not. Imagine a scenario in which more than two people are present. In such cases you and me can be analysed syntactically as an apposition, expressed through a verbal clarification that could be accompanied by a gesture such as pointing to just one interlocutor and the speaker himself: (35)'

Let’s – you and me – go see this Laurie person at the boys’ camp.

The illusion of more formal registers can be seen from example (39), which ridicules European bureaucrats and can thus be expected to mock a rather formal level of Eurospeak. (Notice the name of the variety, Delorean, termed after a former president of the European Commission.)

Modality and the history of English adhortatives

(39)

331

Delorean: “All right, well then let’s you and I talk in the next day or two and we’ll plan to get together in the next five days.” (Brit books 0099)

Example (40) shows that speakers vacillate between let’s you and I and let you and I. Of these, let you and I is clearly hypercorrect, at least if one assumes that let has retained its transitive character. (40)

Let(’s) you and I do it! (Quirk et al. 1985: 338)

Probably less formal, but unequivocally dual are the following two examples: (41)

From his tone he might just as well have said, ’Let’s you and I go to bed.’ (Brit books 0113)

(42)

Skinner said, “let’s you and I put our relationship onto a better foundation and here’s to it.” What did he mean? (Am books 0519)

Slips of the tongue, false starts or syntactic reanalysis completed? Perhaps, however, let has not retained its transitive character for all speakers. For the fact that let’s can be followed by subjective case pronouns (you and I) is not without syntactic consequences, both as regards the syntactic status of let’s and as regards the status of the following pronouns. In examples like (39) to (42) let’s can be analysed as having completely lost its verbal status. Let’s takes the position of an adverbial and is no more than an introductory modal particle, as the following clause must be interpreted as a main clause. On this analysis, a number of let’s occurrences followed by subjective case pronouns need no longer be interpreted as false starts, e.g.: (43)

Let’s we’ll go and do the shopping. (spo 1074)

(44)

Let’s we’d better go and supervise hadn’t we. (spo 0772)

(45)

...shall we? Let’s we’ll go go back to that. (spo 0826)

332 Manfred Krug (46)

I suppose I can conclude here but I mean let’s I shall get back to the [name] Street... (spo 0742).

A potential counterargument to the above dismissal of a false-start-analysis could be put forward on the grounds that all four examples above contain a modal (or, in the case of had better, an emerging or quasi-modal) verb pointing to the unfulfilled action at speaking time and performing a discourse function similar to that of adhortatives. Yet it has been pointed out that new meanings evolve in heavily redundant contexts (e.g. Traugott 2000: 289f). A comment on the redundancy of the above contexts seems indicated: The fact that let’s is functionally almost identical to notions expressed by modal verbs and modal constructions – volitional/intentional/futural we’ll and deontic we’d better – lends another notional argument for the inclusion of adhortatives within the modal domain. The ultimate step towards modal particle status would be for let’s to occur in a construction like ??Lets we go and do the shopping, but in the currently available corpora (except the internet, which I explore elsewhere, cf. Krug 2007) I have not been able to locate an example of this kind. Notice, however, that this would be parallel to the path that ute(n) took when it increasingly adopted reinforcing we in the 12th century (cf. Note 6). From a uniformitarian perspective it is therefore not unlikely that let’s will follow a similar path as its predecessor. The above analysis has focused on the syntactic status of let’s. Semantically the most intriguing example within first person plural use would be a repetition of us, as this would be pleonastic if ’s in let’s carried its full semantic weight. Prima facie, (47) seems to be a relevant example: (47)

Ray: OK, we’ll go bumping along here. Hansen: OK # Take care # Thanks for your call # Let’s us know what happens. (Am radio 30/09/90)

However, a different interpretation is preferable. For one thing know is pragmatically odd in adhortatives. For another, the context – the end of a telephone conversation – suggests a phrase ‘Let us know what happens.’ Much more likely is therefore a false start let’s for let, due to strong entrenchment: The high string frequency of let us not only leads to automatic erosion and cliticization of us but also to false starts of the type above.

Modality and the history of English adhortatives

333

Let’s refers to first person singular When let’s refers to the speaker alone it becomes semantically identical with let me or lemme. This use is particularly prominent with verbs that initiate an action of mutual involvement of speaker and hearer, as can be seen in Quirk et al.’s example of first person singular imperative, which they describe as “very colloquial” (1985: 830): (48)

Let’s give you a hand.

Similar greeting scenes in the London Lund Corpus indicate that such situations seem to favour this use of let’s because the speakers may transfer the reciprocal action from the extralinguistic to the syntactic level. There is one illustrative example in the London Lund Corpus in which let’s indubitably refers to the speaker only. The two participants talk about the piano of B’s mother. B asks A for information about the piano’s quality. (49)

B A B

oh it’s that bad is it# [əm] – well# [lemi] let’s put it like this# – [əm] – – the piano at your mother’s is bad# [m] (S.7.3g.5)

Before this conversation the information was only available to speaker A, who is reluctant to spell the bad news, which can be derived from his extensive use of fillers and pauses. The phonetic transcription [lemi], i.e. lemme or let me, explicitly marks that it is only A who ‘puts it like this’. He may correct himself using the inclusive let’s to demonstrate his empathy. But this is merely a rhetorical device, conscious or subconscious, and it does not alter the fact that let’s in this example semantically only refers to a single speaker himself. A related phenomenon is widespread use of us for me (cf. Quirk et al. 1985: 350f., esp. p. 351, notes a, b). Let’s refers to second person For reference of let’s to second person, Hopper and Traugott cite examples of encouraging a child and one conversation between adults (2003: 11): (50)

Lets [sic!] wash your hands.9

(51)

Lets eat our liver now, Betty.

334 Manfred Krug (52)

Lets you go first, then if we have any money left I’ll go.

Hopper and Traugott suggest that “the semantic steps proceed by small steps,” that is from first to second to third person (2003: 11). This order, however, is only a formalization according to a grammatical pattern. In the present context of adhortatives, it is important that from the perspective of the speaker, it makes psychological sense to include first we (i.e. all those for whom s/he speaks), then you (i.e. all those addressed) and finally all others (third persons, i.e. those remote from both speaker and addressees). As was seen in the religious contexts quoted at the beginning of this chapter, the transition from first to second person is most inconspicuous when there is an overlap between the speaker’s party and the addressees. In conclusion, the extension from first to second to third person indeed seems to be a natural order of progression for let’s.10 Let’s referring exclusively to third persons has not yet been reported. More frequent is the combination of second and third person after lets. Example (53) quoted below is from Hopper and Traugott (2003: 11), who ascribe this use to Midwestern American speakers.11 (54) is an identical structure, quoted from an American novel. (53)

Lets you and him fight.

(54)

Well, we were being set up all the time, it turned out. The FBI was in there like some little guy in a bar going let’s-you-and-him-fight. (Pynchon 1990, Vineland, pp. 232-33)

Finally, one example from the spoken BEC can be cited, which at first sight suggests a use of let’s with second and third person pronouns: (55)

M03 M02 F02

Have they got to have the mentality of adults? So that let’s you let’s you and (name of male speaker) out. Oh no no no. (spo 1167)

However, as no verb follows let’s, a different interpretation is more convincing. This is to assume a spelling mistake let’s for lets; the structure intended by the speaker is thus probably So that lets you and X out in the sense ‘that exempts you’. I would therefore dismiss the example from considerations of imperative let constructions.

Modality and the history of English adhortatives

335

Let’s refers to third person: Spread from generic and indefinite pronouns? As noted above, let’s referring exclusively to third persons has to my knowledge not yet been reported. In the Bank of English Corpus, there are at least a number of examples in which let’s is followed by generic one or universal indefinite pronouns such as somebody, someone, everybody and everyone. For their syntactic behaviour (such as reference by his or their; verb inflection in the simple present tense) these can be counted among third person pronouns. The following three examples illustrate this use of lets with explicit reference to non-first person subjects: (56)

This is the U.S. Coast Guard, and we are exercising our right to board and conduct a safety inspection. Let’s everybody stay where we can see you, please... (scanned Brit books 0886)

(57)

If you get bit, you’re bit on your own # Let’s everybody pray12 at this time. (Excerpt of people praying) (Am radio 30/11/92)

In addition to these corpus data, some further anecdotal evidence can be presented: (58)

Let’s everybody get their sarnies now! (overheard at Birmingham University, England)

Universal indefinites such as everybody and generic one can be motivated by their affinity with inclusive first person reference. If everyone (typically everyone present) is seen as belonging to the speaker’s party, then everybody is virtually synonymous with us. I can provide no more than one nonpronominal third person lets example: (59)

Lets people know that teachers work in the vacation. (overheard in the USA by Christian Mair, personal communication)

Clearly, people shows a close semantic affinity to everybody: semantically, people as well as everybody and generic one include first, second and third person singular and plural. Note, for example, that some discourse analysts, e.g. Winter (1977), attribute nouns such as thing to a category of ‘anaphoric connectors’, which is closely connected with pronouns; people could be seen as its human plural equivalent. Note further that Halliday and Hasan label people a ‘general noun’ (1976: 274), which is “a border-

336 Manfred Krug line case between a lexical item (member of an open set) and a grammatical item (member of a closed set).” Traditionally, however, people is analysed grammatically as third person plural. This, then, represents the ultimate step of progression according to Hopper and Traugott’s developmental path outlined above. The ensuing example, finally, can no longer be explained away on semantic grounds as it is both semantically and grammatically a clear third person use: (60)

You stand back and hold their coats and say, ‘Let’s him and him fight.’ You kind of just watch. (Am radio 29/08/92)

However, while such uses are attested, third person examples beyond indefinite pronouns are rare and seem almost confined to the idiomatic expression lets him and X fight. Whether we are actually dealing with a spreading use can not be answered in the positive at the present stage. It is certainly not yet a productive pattern (cf. also De Clerck forthcoming a, b).

3.4. Summary of developmental paths To integrate all examples thus far invoked into a grammaticalization framework, let us briefly recall Hopper and Traugott’s (2003:7) cline of grammaticality: (61)

content item > grammatical word > clitic > inflectional affix

The entire construction Table 2 presents the major stages in the evolution of adhortative let constructions. The development of us in let’s The sixth stage above has been labelled monomorphemic (Hopper and Traugott 2003: 12), since cliticized ’s has shifted from morpheme to phoneme status. As noted earlier, such constructions are consequently spelt without apostrophe by Hopper and Traugott. A particle label has been

Modality and the history of English adhortatives

337

Table 2. Historical trajectory of adhortative let constructions Stage

(i)

(ii)

(iii)

(iv)

(v)

(vi)

Example (modern English equivalents) Syntactically transparent transitive main (or premodal) verb construction, 2nd ps. Imperative 2nd ps. imperative (permissive/ causative) let us constructions inviting inferences related to adhortative, esp. in religious contexts Construction with independent (pre-) modal verb let plus free proform us; ambiguous between (i) and adhortative; typical early complements: verbs of speaking Unambiguous modal adhortative let construction, as coordinated with subjunctive adhortative Rapid growth of modal adhortative construction: Encliticization of us with reference to first person plural inclusive Modal particle: Extension of reference to non-first person plural (generalized imperative for all persons, still rather rare)

Let us (‘allow us to’) go.*

Attested since Old English

Let us (‘make us’) hate the evil and love the right. 13th c. Let us speak of lusty life in Troy. (‘Allow us to speak’ or ‘Let’s speak’)

Let us pause of Constance for a moment and speak we (SUBJUNCT ADHORT) of the Roman Emperor. Let’s go.

14th c.

14th c.

16th c. Let’s wash your hands. Let’s everybody stay where they are. Let’s you and I do it.

20th c.

* Less common than permissive ‘allow us’ is the causative ‘make us’ reading. This does not have pre-modal semantics but nevertheless serves an auxiliary function.

variously attached to adhortative let’s in the literature before (e.g. Quirk et al. 1985; Biber et al. 1999; van der Auwera and Taeymans 2004). I would like to reserve this label for constructions where us has ceased to function as a morpheme, i.e. where let’s is no longer compositional and thus no longer functions as an adhortative but as an imperative for persons other than 1st ps. plural inclusive. In a slight alteration of Hopper and Traugott (2003: 12), the general development which us has undergone in let us can be depicted as follows:

338 Manfred Krug (let) us

>

(let)’s

>

(let)s

which illustrates the following movement on the cline of grammaticality: free grammatical morph

>

clitic

>

phone

The historical details presented above lend full support to the sketch given in van der Auwera and Taeymans (2004: 243). A passage from that paper can indeed serve as a partial summary of the present paper: In English, the first person inclusive plural pronoun turned towards the imperative, it cliticized, it then seems to have allowed strengthening with us, you and I or you and me, a pattern which further allowed variations with you and him, bare you, and marginally even they and them, and me. In this process let’s became a particle.

4. Let adhortatives: History of negation patterns There is hardly any time lag between the first unambiguous affirmative let adhortatives and negated ones. Both are attested in Chaucer’s works from the late 14th century. The first negated adhortatives with encliticized us, however, are much later: They date from the early 17th century, which, too, is roughly contemporaneous with affirmative let’s constructions: (62)

c1390 Chaucer Man of Law’s Tale, Prol. 32 Lat vs nat mowlen thus in ydelnesse.

(63)

1538 Starkey England I. iv. 123 Let us not entur into thys dysputatyon now, but ... dyffer hyt to hys place.

(64)

1605 Shakespeare Macbeth II. iii. 141 Let’s not consort with them.

(65)

…, come, let’s not despair; the flesh is strong. (ARCHER, BrE, 1675duff.D1)

Perhaps surprisingly, it takes some 350 years until the first instances of DO negation with adhortative let can be found in the OED (example 66), and it takes another 130 years until clitic ’s is first recorded in that pattern (69).

Modality and the history of English adhortatives

339

This first example of don’t let’s from the OED, however, can be antedated by half a century with the help of the ARCHER Corpus (example 68): (66)

1738 Swift Polite Conv. iii. (D) No candles yet… don’t let us burn daylight.

(67)

But don’t let us conclude too hastily; hear the other side. (ARCHER, BrE, 1753foot.D3)

(68)

…, don’t let’s talk of that (ARCHER, BrE, 1820serl.D5)

(69)

1869 Galaxy June, 831 Don’t let’s get outside of more’n a bottle apiece, and that plain whiskey.

4.1. Negation patterns of let adhortatives in present-day English Apart from the negation patterns mentioned above, there exists another variant in present-day English, let’s don’t. It was added (as a colloquial U.S. American English structure) to the OED as late as 1993 and this fact suggests that it was rare in the first half of the 20th century, despite a first American English citation from 1939 (OED Online, s.v. do): Let’s don’t think about a lot of Chinese (Dorothy Parker, Here Lies, 33). Negation patterns are also interesting syntactically, because they offer information on the word class status of let and let’s, but this discussion is beyond the scope of the present paper. Suffice it here to say that let’s vacillates between particle, main verb and auxiliary status (see Krug 2007 for details). As certain negation strategies were expected to correlate with regional variation (cf. Quirk et al. 1985: 148, 831), I compared several spoken and written British and American English registers with regard to their use of negated adhortatives. The results are displayed in Table 3.

340 Manfred Krug Table 3. Negation patterns of let’s in the Bank of English Corpus let’s not

let us not

don’t let’s13

let’s don’t

don’t let us14

do not let us

do not let’s

let’s do not, let us don’t, let us do not

382

103

54

0

8

3

0

0

79

10

15

0

1

1

0

0

91

23

26

0

2

0

0

0

newspapers

68

41

7

0

1

2

0

0

(37m) magazines and journals (45.5m)

144

29

6

0

4

0

0

0

American

127

24

1

8

1

0

0

0

98

16

1

6

1

0

0

0

29

8

0

2

0

0

0

0

British (136m) spoken (26.5m) books (27m)

(31m) spoken (21m) books (10m)

It emerges very clearly that in both British and American English of the 1980s and 1990s, let’s not is by far the most common negation strategy. Let us not is a somewhat more formal option, but rather infrequent in spoken language. This suggests that the split between adhortative let’s and permissive second person imperative let us, which was found for affirmative patterns in section 3.2 above, is almost complete in spontaneous speech with negation patterns too. Two negation patterns seem largely confined to one side of the Atlantic Ocean: Don’t let’s (together with don’t let us15 and do not let us) accounts for a sizable minority of negated adhortatives in British

Modality and the history of English adhortatives

341

English but is almost nonexistent in the American components. Let’s don’t, by contrast, appears to be restricted to American English. It is clear that these findings represent no more than rough regional tendencies. Even though the Bank of English Corpus was at the time of consultation the largest accessible corpus of current English,16 some of the above figures are too small to allow for firm conclusions. Elsewhere (Krug 2007) I compare internet data, which yield a more differentiated picture.

5. Concluding remarks After a theoretical discussion placing adhortatives within the current view of modality, this paper has detailed the historical development of English adhortative let constructions. It was argued that certain syntactic and pragmatic environments motivated the transition from a permissive second person imperative to the current dominant adhortative function. On the basis of historical data, this paper has adopted the position that, rather than representing an instance of gap filling, the developments in the realm of English adhortatives reflect a typical case of layering. Moreover, a spurt in the use of the construction was evidenced for the 16th century. Phonological erosion was argued to be a direct concomitant of the observable drastic frequency gains. As far as negation patterns are concerned, it was seen that even the largest corpora currently available to the linguist yield rather low figures for a medium- to low-frequency construction, although a rough sketch could be drafted: Both British and American English use overwhelmingly let’s not, while don’t let’s seems to be a predominantly British strategy and let’s don’t an exclusively American one. More fine-grained analyses and the investigation of regional varieties other than British or American English thus remain issues for future research.

Appendix: Parameters of grammaticalization (Lehmann 1995 [1982]: 164)

parameter integrity

weak grammaticalization bundle of semantic features; possibly polysyllabic

paradigmaticity item participates loosely in semantic field paradigmatic variability

-

→ strong grammaticalization → few semantic features; oligo- or monosegmental

- paradigmatic- → small, tightly integrated ization paradigm

free choice of items according to communicative intentions

structural scope item relates to constituent of arbitrary complexity

process attrition

obligatorification

→ choice systematically constrained, use largely obligatory

-

condensation → item modifies word or stem coalescence → item is affix or even phonological feature of carrier

bondedness

item is independently juxtaposed

-

syntagmatic variability

item can be shifted around freely

-

fixation

→ item occupies fixed slot

Modality and the history of English adhortatives

343

Notes 1 2.

3.

4.

5. 6.

7.

I am not aware of a classification where subjunctive is not regarded as an inflectional mood. It is clear that subjunctive inflectional mood serves also functions other than imperative in Latin (as well as in the modern Romance languages), so that there is no one-to-one relationship between subjunctive and imperative mood. However, it is also clear that in non-imperative functions, subjunctive mood in Latin almost always expresses modal notions too, e.g. irrealis or potential events and actions, consequences or purposes. There are only few uses of the Latin subjunctive for which a modal motivation is difficult to establish. (Noteworthy are for instance subordinate clauses in which the subjunctive occurs after the conjunction cum meaning ‘when’ or ‘after’.) The discussion of present-day variation in this section is based largely on the analysis presented in Krug 1994, my unpublished MA-thesis. In the meantime, De Clerck (forthcoming a, b) has independently come to similar conclusions. Although historical dictionaries and grammars do not generally include uton among the preterite present verbs, I here follow the classification and arguments given in Denison (1993: 318) and Warner (1993: 94-99, 188-195). In addition to the modern modals, the preterite presents included modern marginal modals like dare and ought. Of the central modals, only the case of will is debatable. Warner (1993: 186) classifies it as marginally belonging to the class of preterite present verbs in Old English and as a full member in Middle English. Again, Denison (1993: 296, see there for morphological detail) holds essentially the same position. Unless otherwise stated, the examples in this section were found through full text searches in the OED on CD-ROM (1999) or the OED Online (2000ff.). The trend towards overt subjects can even be observed for obsolescent ute(n), which died out around the end of the 13th century but was usually reinforced by we from the end of the 12th century onwards (cf. example (10) in this paper; for earlier examples, see Bosworth and Toller 1898: s.v. witon; OED s.v. ute). It seems, then, that ute(n) was relegated to modal particle status during its last century of existence.` This search does not differentiate lexical ‘allow us to’ from adhortative constructions, because the context given in the OED often does not suffice to clearly tell between these sometimes ambiguous constructions. However, since the (more) lexical ‘allow us to do’ construction is more stable than the incoming adhortative, one can attribute most of the observable increase to the adhortative function. This overall trend was confirmed by cross checks across the centuries.

344 Manfred Krug 8.

9. 10.

11. 12.

13. 14.

15.

16.

The components selected for the present analysis totalled 167m words (see Table 3 for details). Unless otherwise stated, the data in this section are from the Bank of English Corpus (my bold throughout). As noted above, they drop the apostrophe where let’s does not refer to first person plural. Cf. van der Auwera and Taeymans (2004) on similar developments in Dutch. Johan van der Auwera (p.c.) pointed out to me that in some Slavonic languages the order of extension is from third to second to first person. While I do not know the details of these adhortatives’ history, I presume that they do not start out from a first person inclusive pronoun, so that their development is not directly comparable to the case of English let’s. It seems that second and third person uses of lets are indeed favoured in American English, but that requires a separate comparative study. One might have assumed that a contraction let’s would never make it into church language (except into the use of swinging priests, perhaps). There is no Let’s pray... in recordings of sermons in either of the British spoken corpora that I have consulted (LLC, BEC). There are, however, two British written examples in religious language: “We need to have an open heart to be able to see God in others. Let’s pray for the grace we need. Pray for us so that...” (scanned Brit books 0881) and “Now let’s pray and talk to the Father quickly. And then I want you to turn to somebody and say, Behold, God is my salvation.” (Brit books 0853). There are, of course, several instances of let’s in nonreligious language, e.g., “Let’s pray she did go to Prescott Hill. That’s what she told Thomas.” (Brit books 0853). Non-adhortatives have been excluded, e.g., “...whether we like it or don’t let’s hope we’ve seen the winner” (spo 0680). Non-adhortatives have been excluded, e.g. second person imperatives (e.g. Don’t let us down), main and subordinate clauses such as (if) they don’t let us... Don’t let us is possible, if rare, as a first person adhortative in British English. It is more frequent as a second person imperative when let has retained transitive verb status. This again attests to a near-complete split between adhortative and permissive let constructions. As stated above, the components selected for the present analysis totalled 167m words. This is over 50% larger, for instance, than the 100m word British National Corpus, which has a further disadvantage for investigations of this kind in that it has no American component.

Modality and the history of English adhortatives

345

References Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 The Longman Grammar of Spoken and Written English. London: Longman. Bosworth, Joseph and Thomas Northcote Toller 1898 An Anglo-Saxon Dictionary. Oxford: Oxford University Press. Bybee, Joan L., William Pagliuca, and Revere D. Perkins 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Crystal, David 2008 A Dictionary or Linguistics and Phonetics. 6th ed. Oxford: Blackwell. Davies, Eirlys 1986 The English Imperative. London: Croom Helm. De Clerck, Bernard Forthcoming a. The inter–dependence of synchronic variation and diachronic change: The case of let’s in American and British English. Paper given at Ghent, Nov. 2003. Forthcoming b. Let’s: Diachronic development, synchronic variation and implications for future grammaticalization processes. Paper given at Leuven, Feb. 2005. Denison, David 1993 English Historical Syntax. London: Longman Halliday, Michael A.K. and Ruquyia Hasan 1976 Cohesion in English. London: Longman. Haspelmath, Martin 2004 On directionality in language change with particular reference to grammaticalization. In Up and Down the Cline: The Nature of Grammaticalization, Olga Fischer, Muriel Norde, and Harry Perridon (eds.), 17–44. (Typological Studies in Language, 59). Amsterdam: Benjamins. Heine, Bernd 1993 Auxiliaries: Cognitive Forces and Grammaticalization. Oxford: Oxford University Press. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press.

346 Manfred Krug Hopper, Paul J. 1991 On some principles of grammaticization. In Approaches to Grammaticalization. Vol I, Elizabeth Traugott and Bernd Heine (eds.), 17– 35. Amsterdam: Benjamins. Hopper, Paul J. and Elizabeth C. Traugott 2003 Grammaticalization. 2nd ed. Cambridge: Cambridge University Press. Original edition 1993. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Krug, Manfred 1994 Contractions in present–day English: A corpus-based study of brachychronic language change. MA–thesis Exeter, England. 2000 Emerging English Modals: A Corpus-based Study of Grammaticalization (Topics in English Linguistics, 32). Berlin: Mouton de Gruyter. 2007 Modern methodologies and changing standards in English linguistics. In Proceedings of the 30th International AEDEAN Conference, María Losada Friend, Pilar Ron Vaz, Sonia Hernández Santano, and Jorge Casanova (eds.), CD-ROM. Huelva: Huelva University Press. Lehmann, Christian 1985 Grammaticalization: synchronic variation and diachronic change. Lingua e Stile 20: 303–318. 1995 Thoughts on grammaticalization [Lincom Studies in Theoretical Linguistics, 1]. Munich: Lincom Europa. Original edition 1982. Matthews, Richard 2003 Modal auxiliary constructions, TAM and interrogatives. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank Palmer (eds.), 47–70. Berlin: Mouton de Gruyter. Oxford English Dictionary (OED) on CD–ROM/Online 1999ff. James Murray et al. (eds.). Oxford: Oxford University Press. Palmer, Frank R. 2001 Mood and modality. 2nd ed. Cambridge: Cambridge University Press. Original edition 1986. 2003 Modality in English: Theoretical, descriptive and typological issues. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank Palmer (eds.), 1–17. Berlin: Mouton de Gruyter. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, Jan Svartvik 1985 A comprehensive grammar of the English language. London: Longman.

Modality and the history of English adhortatives

347

Seppänen, Aimo 1977 The position of let in the English auxiliary system. English Studies 58: 515–529. Traugott, Elizabeth C. 1995 Subjectification in grammaticalisation. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein and Susan Wright (eds.), 31–54.Cambridge: Cambridge University Press. 1997 Subjectification and the development of epistemic meaning: The case of promise and threaten. In Modality in Germanic Languages: Historical and Comparative Perspectives, Toril Swan and Olaf Westvik (eds.), 185–210. Berlin: Mouton de Gruyter. 2000 Semantic Change: An overview. In The First Glot International State-of-the-Article Book: The Latest in Linguistics, Lisa Cheng and Rint Sybesma (eds.), 385–406. Berlin: Mouton de Gruyter. 2003 Constructions in grammaticalization. In Handbook of Historical Linguistics, Brian D. Joseph and Richard D. Janda (eds.), 624–647. Oxford: Blackwell. Traugott, Elizabeth C. and Richard B. Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan and Martine Taeymans 2004 Let’s, in English and in Dutch. In An International Master of Syntax and Semantics: Papers Presented to Aimo Seppänen on the Occasion of his 75th Birthday, Gunnar Bergh, Jennifer Herriman, and Mats Möberg (eds.), 239–247. (Gothenburg Studies in English, 88). Göteborg: University of Gothenburg. Visser, Frederikus Th. 1963–73 An Historical Syntax of the English Language. 3 vols. Leiden: Brill. Warner, Anthony R. 1993 English Auxiliaries: Structure and History (Cambridge Studies in Linguistics, 66). Cambridge: Cambridge University Press. Winter, Eugene O. 1977 A clause–relational approach to English texts: a study of some predictive lexical items in written discourse. Instructional Science 6: 1– 92.

On the “great modal shift” sustained by come to VP1 Philippe Bourdin

1. Control-come vs. Raising-come When the verb come combines with infinitival clauses, its behaviour is, on the face of it, closer to that of the verb want than to that of its deictic counterpart go: (1a) (1b)

Mary wanted to see Sam. Mary wanted to like Sam.

(2a) (2b)

Mary came to see Sam. Mary came to like Sam.

(3a) (3b)

Mary went to see Sam. * Mary went to like Sam.

Clearly, substituting like for see has dire consequences in (3), which it does not have in either (1) or (2). It is just as clear that while nothing much happens to the meaning of wanted as a result of this substitution, the impact on the syntax and semantics of came is far from innocuous. Indeed, standard diagnostic tests confirm that (2a) and (2b) involve two very different constructions. Thus, complementing came by the prepositional phrase to Paris can be easily done in (2a), but is worse than problematic in (2b): (4a) (4b)

Mary came to Paris to see Sam. * Mary came to Paris to like Sam.

This is evidence that the argument structure of came is different in the two sentences. Passivization of the infinitival clause provides additional evidence to that effect: (5a) (5b)

Sam came to be seen by Mary. Sam came to be liked by Mary.

350 Philippe Bourdin While (2b) and (5b) have the same truth conditions, (2a) and (5a) clearly do not: (2a) entails that it was Mary who “came” and (5a) that it was Sam. In short, the NP Mary in (2a) is in a predicative relationship with both came and see Sam, which is another way of saying that the infinitival clause is an adjunct clause of purpose and (2a) an abridged version of (6): (6)

Mary came in order to see Sam.

Came in (2b), on the other hand, is what is known in the generative tradition as a “raising” verb – specifically a “subject-to-subject raising” (SSR) verb, with Mary being the semantic subject of like Sam and the syntactic subject of came. This is another way of saying that the NP Mary has been moved from its initial position as subject of like Sam in an underlying structure which looks roughly like (7): (7)

[ ___ came [ Mary to like Sam ] ]

As a raising verb, come utterly fails to select its syntactic subject. Thus, in (8), existential there is obviously selected by exist in the embedded clause, not by came:2 (8)

There came to exist in Roman as in Greek portraiture a visual code in which expression was given to moral qualities. (British National Corpus AR4 330)

Also, because it does not denote motion in space, come as a raising verb is perfectly compatible with verbs with which, as a spatial verb, it would have been in a relation of antonymy or quasi-synonymy – namely go in (9) and arrive in (10): (9) (10)

‘But how did you come to go your separate ways?’ (BNC AC2 1937) … so how it was that the Gypsy Kings, carrying large musical instruments and followed by a television crew, came to arrive at the dinner table unmaimed, must remain a mystery for ever. (BNC G2E 277)

For ease of reference, I will use the expression ‘Control-come’ to refer to the come to VP construction, as it is instantiated in (2a), to distinguish it

The “great modal shift”of come to VP

351

from the verb come alone. And I will use the expression ‘Raising-come’ to refer to the come to VP construction as it is instantiated in (2b). The semantic contrast between (2a) and (2b) is, at first blush, just as clear-cut and robust as the syntax. To put it in a nutshell, sentence (2a) entails that Mary “came” to a site which is identified with the deictic centre: it is in fact the only thing that the sentence asserts, since the event of seeing Sam is merely contemplated. In short, came is the centre of an autonomous predication. Sentence (2b), on the other hand, describes Mary’s emotional attitude to Sam; whatever contribution Raising-come brings to the meaning of the sentence, it certainly does not refer to motion in space. Indeed, sentence (2b) in no way entails that Mary “came”: intuititively, came seems to qualify in some way what the sentence asserts, which is that “Mary liked Sam”. It seems uncontroversial, therefore, to argue that Control-come involves a spatial interpretation of come and a purposive interpretation of to, with come bearing the assertive weight of the sentence. Raising-come, on the other hand, involves a non-spatial interpretation of come and a non-purposive interpretation of to, with the assertive weight of the sentence being borne by the infinitival clause. In other words, the subject of Control-come refers to an individual whose purpose is to bring about the state of affairs described by the infinitival clause; this volitional feature is, I will argue, absent from sentences instantiating Raising-come. To that extent, the contrast between the two constructions is ultimately of a modal nature and the “great modal shift” alluded to in the title of this paper essentially refers to the fading of volitionality that Raising-come entails, in contrast to Control-come. Before trying to delineate the semantics of Raising-come, I will address two formal issues which may have some semantic relevance. One has to do with the nature of the contrast between the Control construction and the Raising construction: is it strictly binary and absolute or does it allow of a grey zone? The other issue is the morphosyntactic versatility of Raisingcome. I will presently deal with other subject-to-subject raising verbs which may be usefully compared with come in those respects.

2. Refining the contrast The verb come is not alone in entering a control structure as well as a subject-to-subject raising construction. Various authors have dealt with aspectual verbs like begin and continue and also with the evaluative verbs prom-

352 Philippe Bourdin ise and threaten. What is very distinctive about come is that the syntactic contrast between the two constructions is associated with a clear semantic contrast involving to – namely the contrast between purposive to and a use of to which appears to be semantically opaque, if not vacuous.3 As far as can be ascertained, the only verbs behaving like come in that respect are be going and grow as instantiated in (11) and (12) and in (13a) and (14a), respectively: (11)

I’m going to mail this letter.

(12)

There’s going to be a major storm.

(13a)

The sand is so hot and dry that in many places no bush or clump of grass can grow to provide a signpost from which to take bearings. (BNC F9F 1446)

(14a)

If a worker grows to like Picasso, he will be more easily corrupted than if he were to like Gozzoli. (BNC FTW 1065)

The semantic and syntactic contrast between (11) and (12) is straightforward and it illustrates what traditionally serves as the paradigm example of grammaticalization in English: if (11) is understood as a reply to Where are you going?, the infinitival clause refers to the purpose of the motion event; in (12), on the other hand, there is no motion in space and there is no purpose. The behaviour of grow in (13a) and (14a) is just as interesting, even though there has been very little descriptive work done about it. (13a) describes both the increase in size and maturity which defines growing as an organic process and the outcome, rather than the purpose, of that process. Whatever semantic function grow and to fulfill in (14a), it is clear that neither the organic process of growing nor the result of that process are at issue. Hence the acceptability of (13b) as a paraphrase of (13a) and the unacceptability of (14b): (13b) No bush or clump of grass can grow to the point of providing a signpost from which to take bearings. (14b) * If a worker grows to the point of liking Picasso…

The “great modal shift”of come to VP

353

Now, the contrast between Control-grow and Raising-grow is not absolute. The following examples would seem to fall between the cracks: (15)

A glimpse in the mirror, or the well-meant comments of an old family friend, may remind us that we are growing to look like our parents. (BNC BLW 1004)

(16)

“The wind of change”, that phrase coined by Harold Macmillan in 1960, grew to have a more general application than just to Africa, or just to politics. (BNC BLW 1004)

Grow does refer here to some sort of development – physical and moral in (15), metaphorical in (16); this points towards a Control interpretation. However, the predications of which are growing in (15) and grew in (16) are the centre are ancillary, at best, to those headed by look and have, respectively: this is suggestive of a Raising interpretation. In the case of come to VP, likewise, the contrast is by no means strictly binary. Sentences like (17a) and (18a) suggest, rather, that prototypical Control-come and prototypical Raising-come are the polar endpoints of a continuum: (17a)

The flashback concluded with her mother coming downstairs to find her husband dead and Mary sitting in a corner. (J. KabatZinn, 79)

(18a)

This is quite a nice send-off for Kimmel, whose show on Comedy Central featured girls jumping up and down on trampolines. He enters late-night television where many before him – hello, Chevy Chase – have come to fail. (The Virginian-Pilot and The LedgerStar, Norfolk (Virginia), Jan. 24, 2003; in Factiva)

The immediate co-text leaves no doubt that there is motion, real or metaphorical, on the part of the individual denoted by the subject: the motion is very real in (17a), while in (18a) late-night television is metaphorically conceptualized as a location which stand-up comedians may or may not enter. In fact, as (17b) and (18b) demonstrate, come here could stand on its own as the centre of an autonomous predication: (17b)

The flashback concluded with her mother coming downstairs.

354 Philippe Bourdin (18b)

He enters late-night television, where many before him have come.

There seems to be no justification, therefore, in denying come the status of a Control verb. On the other hand, to in (17a) and (18a) cannot possibly be interpreted as a purposive connector; rather, the infinitival clause refers to the result of the subject’s coming, or at the very least to the event immediately adjacent to the coming (Groussier 1981: 62). This would explain why (17c) and (18c) provide acceptable, if perhaps stylistically clumsy, paraphrases: (17c)

The flashback concluded with her mother coming downstairs and finding her husband dead.

(18c)

It is a place where many good coaches have come and failed.

Interestingly, come to V here is also equivalent to end up V-ing, an equivalence which, as will be shown presently, can also be characteristic of Raising-come: (17d)

Her mother ended up finding her husband dead.

(18d)

It is a place where many good coaches have ended up failing.

What the paraphrases (17d) and (18d) reveal is the factuality of the infinitival clause in (17a) and (18a) – a state of affairs which is typical of Raising-come constructions as opposed to those involving Control-come. It may be recalled, for instance, that sentence (2b) – Mary came to like Sam – asserted that Mary did end up liking Sam, whereas sentence (2a) – Mary came to see Sam – in no way asserted that she saw him. To sum up, come to VP in (17a) and (18a) exhibits some of the syntactic features characteristic of Control-come constructions. On the other hand, there is no denying that the assertive force that attaches to the infinitival clause is characteristic of Raising-come constructions. Finally, the connector to is neither the purposive to of (2a) nor the semantically vacuous to of (2b); rather, it functions like a marker of temporal adjacency.

The “great modal shift”of come to VP

355

3. On the morphosyntactic versatility of Raising-come Examples (19) to (21) show that Raising-come may occur in both the have + PP and be + ing forms as well as in generic statements: (19)

Another historian notes that women have not come to rival men, despite the equality established through legislation (!). (BNC CLW 553)

(20)

Towns, therefore, were coming to assume greater importance as places of defence…. (BNC EDF 1298)

(21)

For too many people, these principles translate into cultural and therefore ethical relativism. The result is that people come to believe that all morality is a matter either of law or one’s personal and non-arguable opinions. (The Globe and Mail, Feb. 26 2004, A20)

There is also plenty of evidence that Raising-come tolerates various types of modalization. A sample is provided in examples (22) to (27), which involve such modals as can, be to, deontic must and have to as well as may, both in its epistemic sense and its hortative sense: (22)

Of course the distinction drawn above only remains if we assume that it is possible for us to understand a proposition which we would or could never be justified in believing or could never come to know to be true. (BNC F9K 110)

(23)

At any rate, Barth himself was to come to make exactly the opposite criticism of the book, which he later used to call ‘my wellknown false start’. (BNC CL6 1060)

(24)

To teach effectively… the teacher must come to appreciate and understand the lifestyle of the children in his or her care. (BNC CLY 1347) Again, there is no direct attribution; the child reader has to come to know the conventions which are operating… (BNC EF8 1378)

(25)

356 Philippe Bourdin (26)

She expressed a wish that they should be sent to one of her nieces, May Bailey, ‘in the hope that she may come to care about poetry’. (BNC APR 857)

(27)

May we come to know again the trust and faith of little children. (BNC ALH 259)

The predicative relationship of which Raising-come is part may be negated, as in (19) and (22), or it may be the target of an open question. As far as unembedded wh-questions involving Raising-come are concerned, a cursory search in the BNC has turned up over thirty sentences headed by how, but none headed by why: (28)

How did you come to be Alan’s employee, when he was initially yours? (BNC H8T 792)

There is evidence, however, that embedded why-questions provide a suitable context for Raising-come: (29)

I have no idea why it has come to bear that name. (BNC G39 972)

There is one restriction which Raising-come would seem to share with all other subject-to-subject raising verbs: its incompatibility with positive commands. For instance, it is impossible to imagine a situation in which (30) could possibly be uttered on a non-spatial reading of come: (30)

* Just come to write your masterpiece, why don’t you?

(31a)

?? Don’t come to forget how kind and generous they have been to you! (31b) ? Don’t ever come to forget how kind and generous they have been to you! The slight difference in acceptability between (31a) and (31b) suggests that vetative contexts (warnings not to do something) may provide a marginally suitable environment for Raising-come, preferably if the injunction makes it explicit that the time must never come when the kindness and generosity will be forgotten.

The “great modal shift”of come to VP

357

Finally, examples (32) to (35) demonstrate that the verb of the infinitival clause can itself be in the be + ing or have + PP form and also that it can be passivized: (32)

How did Faith Gauld… come to be crushing grapes with her feet in front of Prince Edward? (BNC K99 104)

(33)

I was rather going to ask how you come to have accepted a post which must necessarily be curtailed before too long. (BNC HGV 1935)

(34)

So there would be no way in which different regions in the early universe could have come to have had the same temperature as each other… (BNC H78 1087)

(35)

Through human rights education, basic human rights can come to be considered universally applicable and will, as a matter of course, be universally respected. (BNC CFH 640)

The range of environments in which Raising-come can occur is therefore quite wide. It is comparable with the collocational range of subject-tosubject raising verbs such as seem, tend or turn out, except that these are probably more resistant to modalization: (36)

? At any rate, Barth himself was to seem to make exactly the opposite criticism of the book.

(37)

?? To teach effectively the teacher must tend to appreciate and understand the lifestyle of the children in his or her care.

(38)

? May we turn out to know again the trust and faith of little children.

On the other hand, the morphosyntax of Raising-come is different from that of the aspectual SSR verbs begin and continue, on two counts. First, of course, because the verb embedded under begin or continue can be either in the infinitive or in the ing-form. Second, because when it is in the infinitive, it is probably more constrained, aspectually, than when the embedding

358 Philippe Bourdin verb is Raising-come. For instance, example (39b) sounds very bad, while the acceptability of (39a) and (40a) is somewhat dubious: (39a) (39b)

?? It’s beginning to have rained a bit much to my taste. * When it continues to have rained for three days solid, you will know it’s time to pack up and leave.

(40a)

?? She’s certainly the kind of girl who will stand up to him if he begins to be acting like a spoiled brat. Why would she marry you if you continue to be acting like a spoiled brat?

(40b)

These formal differences between Raising-come and Raising-begin or Raising-continue are worth mentioning because, as will be seen below, Raising-come is often described in the literature as an aspectual marker. It would also be instructive to compare the morphosyntax of Raisingcome with that of be going. Space constraints, however, prevent me from doing so.

4. The semantics of Raising-come 4.1. Aspectual considerations Those grammarians and linguists who have attempted to identify the semantics of Raising-come have invoked such aspectual notions as “ingressiveness” (Poutsma, 1926: 299; Visser, 1969: 1393), “resultativeness” (Kirchner, 1952: 114), or “resultativeness” and “gradualness” conjoined (Kruisinga, 1931: 225; Friederich, 1970: 46-47). Jespersen (1931: 355-6; 359; 371) offers a temporal account, in terms of “futurity” and what he calls “after-past time”. Finally, a number of linguists have suggested that the semantics of Raising-come pertains, at least tangentially, to the field of modality – in particular, Curme (1931: 397), Groussier (1978: 42–3) and Nicolle (2002: 64–66). What is remarkable is the lack of consensus among these various authors, whether the account they offer is of an aspectual or a modal nature. There is little doubt that the semantic function of Raisingcome is as elusive as it may on occasion appear tenuous, so much so indeed that one would look in vain for any gloss, let alone any definition, in the

The “great modal shift”of come to VP

359

OED or, for that matter, in such major grammars of reference as Quirk et al. (1985) or Huddleston & Pullum (2002). Let us consider, first of all, the notion of “ingressiveness”. To be sure, it is not difficult to find examples in which begin could substitute for come; neither is it difficult, however, to find examples in which come alternates with a terminative marker such as end up. For instance, substitution by end up is possible in (41), while in (42) substitution by begin is possible as far as the first occurrence of Raising-come is concerned: (41a)

How did this greatest and most awe-inspiring of the great cats come to make the East its special domain? (BNC CK2 12) (41b) How did this greatest and most awe-inspiring of the great cats end up making the East its special domain? (42a)

An increasing number of archivists and a few historians are coming to believe that a major change has taken place in the manner in which human society creates the evidence which will be used by the historians who, in the future, come to write about the late twentieth century (Morris, et. al. 1992). (BNC JOV 3257) (42b) An increasing number of archivists and a few historians are beginning to believe that a major change has taken place… The second occurrence in (42a), however, is neither ingressive nor terminative. Rather, its semantic contribution is very close to that of the predictive will appearing in the superordinate clause and it could indeed be replaced by predictive will: (42c)

Human society creates the evidence which will be used by the historians who, in the future, will write about the late twentieth century.

Likewise, the concept of “futurity” is certainly not irrelevant to the use of come in (43) and (44a): (43) (44a)

This will be of great help to both of us when we come to continue his therapeutic treatment. (BNC C9W 289) Indeed, as workers come to recognise that the corporatist game is rigged against them… they will rebel against it. (BNC G3L 927)

360 Philippe Bourdin (44b)

Indeed, workers will recognise that the corporatist game is rigged against them, and as they do so, they will rebel against it.

As Jespersen would have pointed out, adverbial clauses of time constitute a syntactic context that disallows the use of predictive will. And indeed a paraphrase involving predictive will is at times readily available, as in (44b). There are plenty of examples, however, in which Raising-come just does not have predictive force – if only because it is itself modalized by predictive will, as in (45): (45)

But perhaps it will come to be thought by his readers that these successive attitudes to the autobiographical, and to plain speaking, in art are equally valid, equally reversible. (BNC A05 1276)

The concept of “gradualness” is just as problematic. It is quite true that Raising-come is frequently modified by the adverb gradually: (46)

However, he gradually came to see that his fear of God was beneficial because it kept him from falling into sin. (BNC ALK 105)

It is just as easy, however, to find examples which challenge the claim that gradualness is intrinsic to the semantics of Raising-come. Thus, the adverb rapidly in (47a) and the phrase at once in (48a) rule out any hint of gradualness: (47a) (48a)

It was upon serfdom that the critics of Tsarism rapidly came to focus their attention. (BNC EA6 738) I at once came to appreciate why it is that chemistry has so largely lost its appeal to the young. (BNC ALW 1417)

4.2. Raising-come as a marker of subsequentiality Interestingly, (47a) and (48a) can easily be paraphrased, respectively, by (47b) and (48b), in which Raising-come has simply been deleted: (47b)

It was upon serfdom that the critics of Tsarism rapidly focused their attention.

The “great modal shift”of come to VP

361

(48b) I at once appreciated why it is that chemistry has so largely lost its appeal to the young. We are dealing here with a feature of Raising-come which has been observed by Kirchner (1952: 114) and Visser (1969: 1393). There is a cluster of environments in which deletion of Raising-come would not substantially alter the truth-conditional meaning of the sentence, let alone jeopardize its acceptability. I will refer to such uses of Raising-come as “tenuous” uses. They are illustrated by (47a) and (48a), as well as by (49) and (50a): (49)

I had got no further by the time I came to have my first business meeting with Mr Farraday during the short preliminary visit he made to our shores in the spring of last year. (BNC AR3 46)

(50a)

But while Mr Smith campaigned well, he may in hindsight come to be blamed for Labour’s catastrophic defeat. (BNC AK2 402)

“Tenuous” does not mean the same as “pleonastic”. Even when it is deletable, Raising-come often has the effect of removing ambiguity or indeterminacy, for instance when it is in the scope of a modal marker. Thus, (50a) means much the same as (50b), which suggests that the function of come to is to rule out a permissive reading of may while ruling in a subsequential interpretation of the infinitival clause: (50b) It may be that Mr Smith will later be blamed for Labour’s catastrophic defeat. Suggestive as it is, the concept of tenuousness does not play a central role in the semantics of Raising-come. Subsequentiality, on the other hand, does. What Raising-come always does, I would claim, is to configure the state of affairs denoted by the infinitival clause as the outcome of a preparatory process, or at the very least, as an occurrence that puts an end to a latency period. This means that those grammarians who have invoked the notion of “resultativeness” are on to something – just as Rowlands (1966: 66) is on to something when she characterizes Raising-come as “[conveying] that an event is the culmination of things that have gone before”. Following Rowlands, I will, from now on, refer to the state of affairs denoted by the infinitival clause as the “culminative process” and I will refer to the

362 Philippe Bourdin preparatory process spanning the latency period as the “pre-culminative process”. Tenuous uses of Raising-come correspond to a backgrounding of the pre-culminative process. In topological terms, this may take the form of a “contraction” of the latency period. Thus, in (47a), its duration is evaluated as short, while in (48a) the adverb at once configures it as spanning little more than an instant. Non-tenuous uses of Raising-come, on the other hand, correspond to a foregrounding of the pre-culminative process. This is what happens in examples like (46), where the adverb gradually points to the incremental nature of the preparatory process, as well as in examples (51a) and (52a), in which the chain of events leading up to the culminative process is explicitly brought into focus, and in which, quite tellingly, Raising-come cannot be deleted: (51a)

(51b)

(52a) (52b)

That is to say, the fit between meanings and perceptions is a matter of social convention, just as much as the historical process by which each language comes to have its own particular set of word forms for the meanings which it uses. (BNC HPY 48) * … just as much as the historical process by which each language has its own particular set of word forms for the meanings which it uses. How did you come to be an actor? (BNC A06 2197) * How were you an actor?

Likewise, in (53), it is reasonable to assume that the adverbial phrase in this way qualifies the process that led to the first sighting of Neptune, rather than the sighting itself: (53)

It was in this way that Galle came to make the first sighting of the planet now known as Neptune. (BNC FBE 917)

Let us consider, finally, the semantic contrast between (54a) and (55a): (54a)

Our former neighbour came to dig a hole in our garden.

(55a)

Our former neighbour came to dig holes in our garden.

The “great modal shift”of come to VP

363

Why is it that out of context, the most natural interpretation of (54a) involves Control-come, while that of (55a) involves Raising-come? The infinitival clause refers to a single occurrence of hole-digging in (54a) and it is not immediately obvious how this individuated occurrence could have the status of a culminative process. (55a), on the other hand, describes multiple occurrences; these can easily be construed as constitutive of habitual behaviour, and in the real world a habit just never comes from nowhere. The qualitative change which it represents is typically brought about by a chain of events and marks the endpoint of a latency period. Thus it is because it lends itself readily to a subsequential interpretation that came in (55a) is easily construed as an instance of Raising-come. This does not mean that the Control interpretation of (54a) and the Raising interpretation of (55a) are cast in stone. (54b) is evidence that with proper contextualization the infinitival clause can be construed as denoting a culminative process so that came can be understood as an instance of Raising-come. (54b)

Things got so bad with our former neighbour that he came to dig a hole in our garden.

Conversely, the insertion of a temporal specification in (55b) has the effect of making a Control interpretation more plausible: (55b) Every night, our former neighbour came to dig holes in our garden. In terms of Culioli’s théorie des opérations énonciatives, and specifically his theory of qualification (Qlt) and quantification (Qnt),4 it is noteworthy that a Raising interpretation is correlated with a qualitative determination: in (54b), the emphasis is on the “meaning” of the hole-digging, on what it “says” about how the neighbour is perceived by the speaker. A Control interpretation, on the other hand, is correlated with a quantitative determination: in (55b), the emphasis is on the hole-digging as a set of temporally situated and potentially countable occurrences.5

364 Philippe Bourdin 4.3. On the modal status of the culminative process As a verb denoting motion in space, come essentially describes a trajectory whose endpoint is identified with the deictic centre. Depending on the cotext, there may be foregrounding of the trajectory, as in (56), or of the endpoint, as in (57): (56)

It took them three days to come all the way from Liverpool.

(57)

The mailman usually comes at 9.

Raising-come is characterized by a semantic duality which can be traced to these two meaning components. Simply put, foregrounding of the trajectory is correlated with emphasis on the pre-culminative process, while foregrounding of the endpoint is correlated with emphasis on the culminative process. There remains a crucial question: what happens, in Raising constructions, to the deicticity of come? To ask this question is in fact to raise the issue of the modal effect triggered by Raising-come. When one goes over the various examples that have been adduced so far, there emerges one thread that links them all – namely a sense that the syntactic subject of come refers to a salient participant6 whose intentions are simply not at issue. It is as if the participant were brought to the culminative process by a chain of events or a set of circumstances without much consideration being given to any specific intent on her or his part. Two further examples will illustrate this point: (58a)

How did you come to learn Navajo?

(59a)

In London he gradually took over the Underground system and came to control every line except the Metropolitan. (BNC EDN 2051)

In both (58a) and (59a), the culminative process is of the agentive type. Agentivity, however, is not at all the same thing as intentionality. This becomes clearer when one compares (58a) with (58b) and (59a) with (59b): (58b) (59b)

Why did you learn Navajo? He came to London to control the Underground system.

The “great modal shift”of come to VP

365

It has already been observed that when Raising-come occurs in a direct question, it is typically headed by how, rather than by why. Someone who asks (58b) wants to know what motivated the co-speaker to learn Navajo, in other words what her purpose was. Someone who asks (58a), on the other hand, is requesting information pertaining to the circumstantial pathway that led the co-speaker to learn Navajo: the speaker wants to know, so to speak, the whole “story”, the whole “script”. In addition, the How did you come to VP? pattern is very frequently associated with speaker evaluation (typically astonishment, incredulity) – so that we are not very far, with (58a), from a quasi-exclamation like (58c): (58c)

How could it be that you learned Navajo?

In short, asking How did you come to VP? may be another way of saying How did it come to that? The confrontation between (59a) and (59b) is even starker. Sentence (59b) involves a Control construction. It does not entail, let alone assert, that the salient participant actually gained control of the Underground system. Rather, it asserts that there was an event of motion in space on his part and that it was for the express purpose of controlling the Underground system. Sentence (59a), on the other hand, asserts that he gained control of every line but one: what is crucially missing is any assertion, or even any suggestion, that he had intended things to be that way. Indeed, it is as though come’s shift from the domain of space and time conjoined – in (59b) – to that of time alone – in (59a) – was correlated with a shift in the scope of assertion – a shift which, depending inter alia on the semantics of the verb in the infinitival clause, involves the suppression of intentionality or its backgrounding. Here is the “great modal shift” mentioned in the title of this paper. It can be argued that there are two factors that conspire to create the shift away from intentionality. One contributing factor is quite simply the formal contrast between Control constructions and Raising constructions. There exists between them a contrapuntal relationship, so that Raising-come confers on the connector to a non-purposive value which is the mirror-image of the purposive semantics that is characteristic of to in the Control construction. Semantic reanalysis here is the necessary correlate of syntactic reanalysis. The other factor is the deicticity intrinsic to the semantics of spatial come, specifically the operation whereby the goal of motion is identified

366 Philippe Bourdin with the deictic centre.7 It is the speaker who selects the deictic centre – in the default case her own location – and it is also the speaker who carries out the identification. There is no reason to think that these crucial semantic attributes of come simply vanish into thin air when it happens to denote motion through time alone, rather than motion in space and time. Basically, my claim is that come never ceases to be a marker of identification. As far as the Raising construction is concerned, I would make the further claim that the salient participant is construed as a moving entity and the infinitival clause as an abstract location: what come signifies is the identification of the salient participant’s final position with that abstract location. Again, it is none other than the speaker who carries out the identification; and it is as if such explicit intrusion inevitably amounted to a downgrading of the salient participant’s own role in reaching her abstract destination. In other words, it is from the very deicticity of come that the backgrounding of intentionality characteristic of Raising-come ultimately proceeds. One method for testing this hypothesis is to compare the semantics of come in Raising constructions with that of get. Such comparison is relevant because the essential difference between the two verbs lies precisely in the non-deicticity of get. Example (60a) is representative of sentences in which the salient participant’s intent is simply not at issue. (60a)

The more I grew up, the more I got to like her. (BNC BMM 241)

In examples (61a), (62a) and (63a), on the other hand, the co-text makes it quite clear that the salient participant’s involvement in the culminative process is the product of deliberate and purposive action on his, her or their part: (61a)

He diagnosed that his concentration on getting to know the clergy prevented him from moving among the wider community as he should. (BNC A68 2009)

(62a)

Diana, on the other hand, wanted to spend time really getting to know her husband. (BNC ECM 1377)

The “great modal shift”of come to VP

(63a)

367

There are other problems teachers have to face up to in practice; getting to learn how to use the programs and how to fit them into their schemes, for instance: both require much thought and time. (BNC H88 1582)

Interestingly, Raising-come substitutes more easily for Raising-get in (60a) than it does in (61a), (62a) and (63a): (60b)

The more I grew up, the more I came to like her.

(61b)

? He diagnosed that his concentration on coming to know the clergy prevented him from moving among the wider community as he should.

(62b)

? Diana, on the other hand, wanted to spend time really coming to know her husband.

(63b)

? There are other problems teachers have to face up to in practice; coming to learn how to use the programs and how to fit them into their schemes, for instance…

It would seem that Raising-come here is slightly odd and that Raising-get is better equipped to encode as culminative a process which is so explicitly represented as involving deliberate intent on the part of the salient participant. It is just as significant, finally, that when the culminative process most definitely eludes any intent on his part, come is markedly better than get: (64a)

Later in the film, McNamara refuses to discuss the responsibility he might bear for the damage wreaked in Vietnam. Nor does he want to talk about why, though he came to regret the war, he failed to speak out against it after he left office. (F. Kaplan)

(64b)

?? Nor does he want to talk about why, though he got to regret the war, he failed to speak out against it after he left office.

It is logical to infer from such comparisons that deicticity is crucial to the suppression or backgrounding of intentionality which is a hallmark of Raising-come.

368 Philippe Bourdin 5. Concluding observations From the foregoing arise two issues in need of further elaboration. One is the diachronic relationship between Control-come and Raisingcome. Example (65) is representative of the earliest examples of Controlcome to be found in Old English, while (66), in Middle English, is the earliest attestation of Raising-come provided by Visser: (65)

Cuom forthon sunu monnes to soecanne & to hælenne. ‘The son of man came forth to seek and to heal.’ (10th century, Lindisfarne Gospels, Luke, 19, 10; ex. quoted in Visser [1969: 1392])

(66)

To-day…Ihc wulle do pruessse, For thi luve, in the felde Mid spere and mid schelde. If ihc come to lyve Ihc schal the take to wyue. (c1225, King Horn 556; ex. quoted in Visser [1969: 1394])

We are dealing here with a time lag which is significant enough to support an assumption that has been made throughout this paper: namely, Raisingcome is derived from Control-come via a process of syntactic reanalysis. This would imply that the “great modal shift” that has been argued for corresponds to a diachronic reality. Whether it qualifies as an instance of grammaticalization, however, is debatable. Semantically, the loss by to of its purposive meaning is certainly typical of the sort of “bleaching” which is conventionally associated with grammaticalization processes. The semantic shift undergone by come makes the case for grammaticalization just as compelling: after all, the strengthening of the temporal component and the correlative fading of the spatial component have also been crucial to the auxiliarization of be going, a paradigm case of grammaticalization. On the other hand, one can hardly gloss over the reality that the verb come in Raising constructions exhibits none of the morphosyntactic attributes of a grammatical marker, specifically an auxiliary. In short, it is as if “semantic” grammaticalization had proceeeded without leaving any formal imprint. The modal status of Raising-come is another topic well worth pursuing. Pierre Cotte (2004) and Heidi Verplaetse (2004) have made a persuasive case for including volition, and therefore intentionality, in the semantic “map” covered by modality. This is also, essentially, the position taken by Palmer (2001: 131–135), Krug (2002) and Hengeveld (2004: 1193–1196)

The “great modal shift”of come to VP

369

among others: for a thoroughgoing discussion, see Verplaetse (2003: 152– 155). Now, granting intentionality the status of a modality in its own right raises the issue of whether the suppression or backgrounding of intentionality encoded by come also counts as a type of modality. Example (67) suggests that the semantics of come’s cognate in Norwegian raises just the same question, and so does that of French venir in examples like (68): (67)

han kom til å knuse den gamle to INFIN break DEF old he come:PAST kinesiske vasen Chinese vase ‘He accidentally broke the old Chinese vase.’ (Norwegian; ex. quoted from Nicolle [2002: 58])

(68)

c’ est alors qu’ un chevalier it be:PRES then that INDEF knight vint à passer to pass.by come:PAST ‘It was then that a knight passed by.’ (French; ex. quoted from Bourdin [2005: 273])

The question ultimately resolves to this: is “modality in absentia” still modality?

Notes 1.

2.

3.

I gratefully acknowledge support from Glendon College (York University, Toronto) and the Social Sciences and Humanities Research Council of Canada in the form of travel grants awarded in April 2004. I am also grateful to Lara Bourdin for contributing her intuitions as a native speaker of Canadian English. Henceforth, bold type will be used in the examples to highlight each occurrence of come (or other main verbs under discussion) as well as the verb of the infinitival clause. There exists an extensive body of literature on the semantics of to + infinitive, including Groussier (1981), Cotte (1982a and 1982b), Chuquet (1986), Girard (1999), Verspoor (1999) and Boulonnais (2004). Haspelmath (1989) offers a diachronic/typological perspective on to’s shift from purposive to nonpurposive meaning.

370 Philippe Bourdin 4.

5.

6.

7.

The essentials of Culioli’s theory are presented in Groussier (2000), which is a useful gateway to Culioli’s published work (1990, 1999a and b). As far as the concepts of Qnt and Qlt are concerned, I am referring here to the account offered by Deschamps (1999: 270): “La quantification est liée au spatio-temporel et à la construction d’occurrences situationnelles, la qualification relève de la construction des propriétés, des relations intersubjectives et de la position de l’énonciateur.” [“Quantification pertains to the anchoring in space and time and the construction of situation-bound occurrences, while qualification has to do with the construction of properties, intersubjective relations and the position of the speaker.”] Chuquet and David (2003: 229) compare the semantics of came to the conclusion with that of concluded. They argue that the concatenation of came and to tips the scale towards a qualitative (Qlt) determination, whereas concluded strikes a balance between the quantitative (Qnt) and the qualitative (Qlt). The concept of ‘salient participant’, to refer to the individual or entity denoted by the raised subject in a construction involving subject-to-subject raising, is borrowed from Langacker (1995: 58-59). In Antoine Culioli’s théorie des opérations énonciatives, identification is one of the three values that the operator of “location” (repérage) may take: a term X may be identified with a term Y, it may be differentiated from it or it may be disconnected from it (Groussier 2000: 168–9). To take a straightforward example, 1st person pronouns involve identification with the speaker, 2nd person pronouns involve differentiation and 3rd person pronouns disconnection. For a useful terminological and conceptual clarification of “identification” vs. “identity”, see Larreya (2003).

References

Boulonnais, Dominique 2004 TO et les infinitives : l’hypothèse de la transcendance prépositionnelle. In La Contradiction en anglais, Claude Delmas (ed.), 55–90. (Travaux du C.I.E.R.E.C. 116). Saint–Étienne: Publications de l’Université de Saint–Étienne. Bourdin, Philippe 2005 Venir en français contemporain : de quelques fonctionnements périphrastiques. In Les Périphrases verbales, Hava Bat–Zeev Shyldkrot and Nicole Le Querler (eds), 261–278. (Linguisticae Investigationes Supplementa 25). Amsterdam/Philadelphia: John Benjamins.

The “great modal shift”of come to VP

371

Chuquet, Hélène and Caroline David 2003 Come : point de vue et perception dans le récit. In Verbes de parole, de pensée, de perception. Études syntaxiques et sémantiques, Jean Chuquet (ed.), 203–235. Rennes: Presses Universitaires de Rennes, and Poitiers: MSHS–Université de Poitiers. Chuquet, Jean 1986 To et l’infinitif anglais. (Cahiers de recherche en linguistique anglaise). Paris: Ophrys. Cotte, Pierre 1982a Autour de TO. In Apports français à la linguistique anglaise, 57–80. (Travaux du C.I.E.R.E.C. XXXV). Université de Saint–Étienne. 1982b TO, opérateur de dévirtualisation en anglais. Modèles linguistiques 4: 135–149. 2004 Modality and volition. Paper presented at the Second International Conference on Modality in English (ModE2), Pau. Culioli, Antoine 1990 Pour une linguistique de l’énonciation. Opérations et représentations (Tome 1). Paris: Ophrys. 1999a Pour une linguistique de l’énonciation. Formalisation et opérations de repérage (Tome 2). Paris: Ophrys. 1999b Pour une linguistique de l’énonciation. Domaine notionnel (Tome 3). Paris: Ophrys. Curme, George O. 1931 Syntax (Vol. III of A Grammar of the English Language). Boston: Heath & Co. Deschamps, Alain 1999 Essai de formalisation du système modal de l’anglais. In Les Opérations de détermination. Quantification / qualification, Alain Deschamps and Jacqueline Guillemin–Flescher (eds), 269–285. (Collection L’homme dans la langue). Paris: Ophrys. Friederich, Wolf 1970 Die infiniten Formen des Englischen. Infinitiv, Gerundium, Partizip. 3rd ed. Munich: Verlag Akademische Buchhandlung. Girard, Geneviève 1999 TO et la linéarité. In Langage et linéarité, Pierre Cotte (ed.), 51–58. Villeneuve–d’Ascq: Presses Universitaires du Septentrion. Groussier, Marie–Line 1978 Sur les verbes come et go en anglais contemporain. T.A. Informations 1: 22–41, and 2: 33–56. 1981 La préposition TO devant l’infinitif en anglais contemporain. In Analyse des prépositions. Troisième colloque franco–allemand de

372 Philippe Bourdin linguistique théorique, Christoph Schwarze (ed.), 40–66. (Linguistische Arbeiten 110). Tübingen: Max Niemeyer. 2000 On Antoine Culioli’s theory of enunciative operations. Lingua 110: 157–182. Haspelmath, Martin 1989 From purposive to infinitive – A universal path of grammaticization. Folia Linguistica Historica 10: 287–310. Hengeveld, Kees 2004 Illocution, mood, and modality. In Morphology. An International Handbook on Inflection and Word–Formation (Vol. 2), Geert Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas (eds), 1190–1201. (Handbücher zur Sprach– und Kommunikationswissenschaft 17.2). Berlin/New York: Walter de Gruyter. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Jespersen, Otto 1931 A Modern English Grammar on Historical Principles (Part IV: Syntax. Third Volume). London: George Allen & Unwin, and Copenhagen: Ejnar Munksgaard. Kirchner, Gustav 1952 Die zehn Hauptverben des Englischen im Britischen und Amerikanischen. Halle: Niemeyer. Krug, Manfred 2002 A path to volitional modality. In English Historical Syntax and Morphology, Teresa Fanego, María José López–Couso and Javier Pérez-Guerra (eds), 131–155. (Amsterdam Studies in the Theory and History of Linguistic Science. Series IV, Current Issues in Linguistic Theory 223). Amsterdam/Philadelphia: John Benjamins. Kruisinga, Etsko 1931 A Handbook of Present-Day English. Part II: English Accidence and Syntax. Groningen: P. Noordhoff. Langacker, Ronald W. 1995 Raising and transparency. Language 71: 1–62. Larreya, Paul 2003 Identité et identification, Cycnos 21:1. http://revel.unice.fr/cycnos/document. html?id=17 Nicolle, Steve 2002 The grammaticalisation of movement verbs in Digo and English. Revue de Sémantique et de Pragmatique 11: 47–67.

The “great modal shift”of come to VP

373

Palmer, Frank R. 2001 Mood and Modality. 2nd ed. (Cambridge Textbooks in Linguistics). Cambridge: Cambridge University Press. Poutsma, Hendrik 1926 A Grammar of Late Modern English. Part II: The Parts of Speech. Section II: The Verb and the Particles. Groningen: P. Noordhoff. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rowlands, Evan C. 1969 Teach Yourself Yoruba. London: Hodder and Stoughton. Verplaetse, Heidi 2003 What you and I want: A functional approach to verb complementation of modal want to. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug and Frank Palmer (eds), 151– 189. (Topics in English Linguistics 44). Berlin/New York: Mouton de Gruyter. 2004 Volition: a proposal for subclassification within the borders of modality. Paper presented at the Second International Conference on Modality in English (ModE2), Pau. Verspoor, Marjolyn 1999 To infinitives. In Issues in Cognitive Linguistics, Leon De Stadler and Christoph Eyrich (eds), 505–526. (Cognitive Linguistics Research 12). Berlin/New York: Mouton de Gruyter. Visser, Frederic Th. 1969 An Historical Syntax of the English Language (Part Three, First Half). Leiden: E. J. Brill.

Sources of examples British National Corpus (BNC): http://sara.natcorp.ox.ac.uk/lookup.html Factiva: http://www.factiva.com/ The Globe and Mail [Canadian daily newspaper]. Kabat-Zinn, J. 1990 Full Catastrophe Living. New York: Delta Trade Paperbacks [Bantam Doubleday Dell Publishing Group Inc]. Kaplan, F. 2003 The Evasions of Robert McNamara: What’s true and what’s a lie in The Fog of War? http://www.slate.com/id/2092916

List of contributors Philippe Bourdin York University Toronto, Canada & University of Paris X Paris, France [email protected]

Pierre Busuttil University of Pau Pau, France [email protected]

Agnès Celle University of Paris VII Paris, France [email protected]

Ross Charnock University of Paris IX Paris, France [email protected]

Alain Deschamps University of Paris VII Paris, France [email protected]

Renaat Declerck Catholic University of Leuven Kortrijk, Belgium [email protected]

Lionel Dufaye University of Paris VII Paris, France [email protected]

Gilbert Ghio University of Nice Nice, France [email protected]

Manfred Krug Otto Friedrich University of Bamberg Bamberg, Germany [email protected]

Paul Larreya University of Paul Pau, France & University of Paris XIII Paris, France [email protected]

376 List of contributors Juana I Marín Arrese Complutense University of Madrid Madrid, Spain [email protected]

Jean-Claude Souesme University of Nice Nice, France [email protected]

Keith Mitchell University of Edinburgh Edinburgh, U.K. [email protected]

Jelena Timotijevic University of Brighton Brighton, UK [email protected]

Claude Riviére University of Pars VII Paris, France [email protected]

Johan van der Auwera University of Antwerp Antwerp, Belgium [email protected]

Christiane Rocq-Migette Université Paris XIII Paris, France [email protected]

An Verhulst Catholic University of Leuven Kortrijk, Belgium [email protected]

Raphael Salkie University of Brighton Brighton, UK [email protected]

Christopher Williams Bari University Bari, Italy [email protected]

Author index Austin, J. 186, 195, 196 Asprey, M. 200 Beneviste, E. 237 Berk, L. 211, 212 Bhatia, V. 206 Biber, D. 48, 80, 337 Binnick, R. 232 Blakemore, D. 237 Bosworth, J. 343 Boulonnais, D. 369 Bourdin, P. 7 Bouscaren, J. 3, 165, 173, 174 Boyd, J. 60 Brandt, P. 238, 240, 246, 250 Busuttil, P. 27 Butler, J. 85, 105, 109 Bybee, J. 327 Carston, R. 100 Celle, A. 6, 290, 291 Chafe, W. 241, 245, 275 Charnock, R. 5, 195 Chuquet, J.174, 369, 370 Close, R. 279 Coates, J. 16, 80, 117, 118, 212, 214, 220 Collins, P. 205 Comrie, B. 49, 80 Cotte, P. 148, 368, 369 Croft, W. 80 Crystal, D. 316 Culioli, A. 3, 5, 6, 123, 146, 154, 156, 174, 175, 281, 291, 299, 363, 370 Curme, G. 358

Dancygier, B. 73 Danet, B. 208 Dasher, R. 237, 252 David, C. 370 Declerck, R. 2, 35, 37, 43, 48, 49, 211, 213, 215, 219, 220, 308, 312 Dendale, P. 240 Denison, D. 343 Depraetere, I. 219 Deschamps, A. 5, 142, 145, 370 De Clerck, B. 320, 336, 343 De Haan, F. 240 Diver, W. 69, 71 Donaldson, T. 67 Downing, A. 80 Dufaye, L. 5 Duranti, A. 241 Facchinetti, R. 7 Filippi, C. 169 Fillmore, C. 65 Fitneva, S. 241, 245, 249 Ford, C. 234 Friederich, W. 358 Gauthier, A. 169 Gordon, D. 18 Ghio, G. 5 Gilbert, E. 142 Girard, G. 369 Gotti, M. 242 Grebe, P. 105, 108, 109 Greenbaum, S. 6, 80, 269, 270-272, 274, 275 Gresset, S. 142, 149, 154 Grice, H. 50, 188, 192, 196

378 Author index Grierson, H. 57 Griesbach, H.108 Groussier, M. 3, 123, 146, 354, 358, 369, 370 Guimier, C. 276, 277, 280, 281, 290 Halliday, M. 66, 67, 335 Hart, H. 193, 194 Hasan, R. 335 Haspelmath, M. 369 Hengeveld, K. 368 Hopper, P. 315, 320, 321, 326, 328, 329, 333, 334, 336, 337 Horn, L. 233 Hoye, L. 217-219, 273, 276, 277 Huddleston, R. 48, 49, 57, 63, 66, 68, 79, 80, 96, 99, 119, 120, 218, 231, 245, 285, 286, 359 Jespersen, O. 42, 358, 360 Kiefer, F. 81, 91 Kirchner, G. 358, 361 Kjellmer, G. 97 Krug, M. 7, 212, 315, 320, 326, 327, 332, 339, 341, 343, 368 Kruisinga, E. 358 Lakoff, G. 18, 60 Langacker, R. 237, 238, 243, 244, 252, 253, 370 Lapaire, J. 163, 166, 173 Larreya, P. 1, 9, 11, 12, 25-27, 89, 105, 110, 114, 118-120, 162, 165, 173, 174, 212, 233 Lauchman, R. 200 Laugier, S. 195 Leech, G. 93, 94, 100, 173, 205 Lehmann, C. 326, 342 Levinson, S. 74, 87, 100 Locke, P. 80 Lunn, P. 74

Lyons, J. 2, 32, 33, 49, 55-58, 64, 75, 82, 105-107, 111, 114, 237, 275, 279 Marín-Arrese, J. 6, 237, 238, 246 Matlock, T. 241, 248 Matthews, R. 316 Mitchell, K. 2, 44, 317 Mowatt, C. 200 Mushin, I. 237, 246 Narrog, H. 82-84, 92, 93 Nicolle, S. 358 Nuyts, J. 105, 110-112, 237, 238, 249, 252, 254, 255, 279 Palmer, F. 16, 21, 59, 60, 67, 74, 79, 81, 82, 91, 92, 105, 107, 108, 117, 118, 145, 150, 151, 212, 219, 223, 241-244, 275, 276, 285, 317, 318, 368 Papafragou, A. 2, 82, 85, 86, 89-91, 114, 115, 117, 214, 215, 220 Perkins, M. 271 Plungian, V. 67, 79, 81, 237, 241, 242, 245, 246, 249 Poutsma, H. 358 Pullum, G. 48, 49, 57, 67, 68, 79, 80, 96, 218, 231, 245, 285, 286, 359 Putnam, H. 191 Quine, W. 56 Quirk, R. 48, 49, 173, 230, 271, 290, 316, 333, 337, 339, 359 Ramat, P. 284 Recanati, F. 2, 84, 100, 105, 114, 115, 117, 191 Reed, S. 35, 49 Reinach, A. 195 Rivière, C. 3, 6, 89, 94-96, 105, 110, 114, 118-120, 162, 173, 174, 299

Author index 379

Rocq-Migette, C. 6, 234 Rosch, E. 81, 193 Ross, J. 109 Rotgé, W. 163, 166, 173 Rouskov-Low, J. 142 Rowlands, E. 361 Russell, B. 196 Salkie, R. 2, 27, 94, 95, 220 Sanders, J. 237, 247, 249, 253, 254 Searle, J. 191 Seppänen, A. 315 Smith, N. 205 Soames, S. 52 Souesme, J. 5, 159 Sperber, D. 220 Spooren, W. 237, 253, 254 Stein, D. 64 Stroud, F. 177 Sweetser, E. 11, 27, 117 Taeymans, M. 315, 320, 326, 337, 338, 344 Tasmowski, L. 240 Tellier, A. 21, 23, 28 Thompson, S. 234 Thorne, J. 60 Timotijevic, J. 2, 82 Toller, T. 343

Traugott, E. 11, 237, 252, 315, 320, 321, 324, 326, 328, 329, 332-334, 336, 337 Travis, C. 192, 196 Tregidgo, P. 214 van der Auwera, J. 27, 67, 79, 81, 241, 242, 315, 320, 326, 337, 338, 344 Verhulst, A. 6 Verplaetse, H. 368, 369 Verspoor, M. 369 Verstraete, J. 82-85, 105, 112-114 Visser, F. 28, 358, 361, 368 Von Wright, G. 62 Waismann, F. 191 Ward, G. 242 Wärnsby, A. 214 Warner, A. 70, 325, 343 Watbled, J. 26 White, P. 238 Willett, T. 237, 245, 246 Williams, C. 5, 200, 203, 208, 209 Wilson, D. 220 Winter, E. 335 Wurmbrand, S. 85, 109 Zifonun, G. 109

Subject index abduction (abductive) 14, 15 actualization 32-44, 46, 48-52, 140, 215, 226 adhortative 7, 315-341, 343, 344 adverb 6, 216-218, 255, 269-275, 276, 277-291, 299, 301, 303, 304, 306, 308-311 ambiguity 117, 118, 189, 190, 192, 214, 216, 300, 320, 322, 324, 337 antecedent 14-17, 22, 27, 28 aphony 238-240, 242, 246, 249251, 259, 261 apodosis 233 argumentation 57, 58, 232 be able to 9, 109 be expected to 12 be going to 21, 27, 352, 368 be supposed to 12 be sure to 12 because 11, 27 before-clause 2, 6, 7, 37, 40-45, 47, 48, 52, 300-303, 306-308 bleaching 98, 368 Caddo 92 can 2, 5, 26, 27, 60, 62, 64, 81, 8587, 89, 90, 93, 94, 97, 100, 108, 110, 115-118, 128-131, 134, 137, 138, 141, 145, 156, 187, 202, 321 causation 12, 17, 322, 324, 326 certainty 27, 87, 237, 238, 240, 242, 243, 259, 264, 274, 283 clause type 3, 7, 56, 66-69, 75

cliticization 325, 327-329, 332, 336, 338 cognitive stance 247, 249, 250, 258 come 7, 349-370 commitment 82, 84, 86, 88-90, communicative stance 63, 247, 250251, 258, 259 conceptualizer 238, 243, 252-255, 262-264 conditional (construction) 35, 38, 39, 41, 49, 73, 154, 223-232 consequence 95, 225, 281 contrastiveness 139, 290 control-come 7, 349-351, 353, 354, 363, 365, 368, 369 could 5, 87, 148, 149, 154-156, 172, 242 counterfactual 1, 2, 4, 32-41, 45, 4852, 65, 73, 91 culminative process 361, 363-367 dare (to) 343 deixis 2, 55, 64-66, 72, 126, 366, 367 disassertion 269, 280-282, 288, 289 do 2, 56, 69-72, 75, 160, 338-341 doubt 64, 66, 73, 89, 106, 111, 116, 163, 193, 274, 275, 283-285, 300 Dutch 110, 326, 344 ellipsis 12, 14-16, 27 encliticization 326, 337, 338 entailment 115, 225, 319 enunciational position 6, 7, 238-240, 250, 258, 259, 264

Subject index 381

enunciative level 3, 4, 123, 124, 159, 165, 174 erosion 321, 325-327, 332, 341 evidentiality 2, 80, 92, 93, 108, 110, 111, 237-239, 240, 241, 245-252, 256-264, 275, 276 expectation 7, 215, 231, 296, 304311 experiencer 224 experiential stance 238-240, 246249, 258, 264 external volition 18, 19, 22, 23 factuality values 2, 20-23, 32-52, 87, 89, 92-94, 111 fictitious locator 281, 282, 285, 289291 free enrichment 115, 116, 120 French 58, 59, 75, 96, 149, 156, 188, 271, 290, 369 future 2, 16-17, 20, 27, 28, 31, 3537, 39, 43, 45, 48-50, 52, 63, 86, 87, 90-92, 95, 96, 98, 99, 128, 132, 186, 211, 212, 214-217, 220, 230, 244, 300, 301, 318, 332, 358, 359 ‘gap’ measuring (mesure en creux). 297, 311 German 1, 17, 93, 96, 105, 108-110, 112, 276, 326 Gradient 11, 118, 172, 175 Gradualness 360, 362 grammaticalization 7, 49, 55, 65, 223, 315, 321, 336, 342, 352, 368 habitual 16-17, 26, 91, 363 have (got) to 6, 26-27, 110, 211-221 hearsay 269-291 hortative 315 hypothetical 2, 4, 31-39, 44, 48-53, 61, 63, 151-156, 165, 225 identification 127, 239, 366, 370

imperative 68, 71, 72, 91, 315-344 implication 13-16, 27, 28, 50, 63, 131, 140, 169, 215, 216, 308, 324, 327 implicature 18, 19, 49-51, 192, 196, 324 implicit 11, 18, 41, 44, 45, 238, 253256, 262-264 inference 246-252, 276, 280, 281, 323 ingressiveness 358, 359 intensional verb 35, 41, 42, 51 intentionality 7, 41, 42, 45, 188, 332, 364-369 intersubjectivity 55, 67, 73, 75, 109, 111, 112, 125, 127, 133, 134, 237240, 252, 254-256, 261-264, 279, 282 irrealis 2, 91-93, 296, 318, 319, 343 Italian 203 jussive 316, 318-320 Latin 318, 343 law 5, 177-198, 199-210 let's 7, 315-341, 344 lexicalization 230 may 1, 2, 5, 6, 56, 74, 75, 85, 87, 89, 90, 93, 94, 96-100, 108, 110, 115118, 128-131, 133, 134, 138, 140, 141, 145, 151, 156, 159-176, 177198, 200-202, 206, 207, 242, 244, 253, 255, 277, 278, 291, 321, 361 Manam 91, 92 meaning 4, 18-20, 22, 23,29, 67, 70, 71, 73, 75, 84-87, 96-99, 108-111, 114, 115, 117-120, 149, 179, 182, 191, 211, 212, 214-219, 223, 225,232, 295, 300, 301, 304, 308, 324, 363 contextual 185, 192

382 Subject index literal 5, 177, 178, 185, 187, 189192, 196, 228 measurement of duration 295-298, 306 metonymy 12, 22 Middle English 321-324, 337, 338, 343, 368 might 1, 5, 87, 112, 145-157, 172, 173, 242 modal particle 343 modal past 48, 233 modal past perfect 48, 51 modal remoteness 48 modal square 9, 10 modality definition 32, 50, 51, 79, 80, 92, 223, 319 degree of modality 2, 79-103, 218 deontic 1, 10, 13, 51, 55, 59, 60, 62, 67, 69, 75, 82-84, 88, 105-110, 112, 113, 117, 120, 128, 135, 177199, 205, 217 dynamic 51, 59, 63, 81-83, 107, 112, 113, 117, 149 epistemic 1, 6, 11-13, 27, 45, 4850, 55, 58-60, 62, 67, 69, 75, 8284, 87-99, 105-113, 117, 118, 120, 127, 128, 130, 131, 133, 135, 145157, 159-161, 164, 173, 213-219, 223-235, 237-245, 258, 263, 269, 275, 276-280, 281 existential 2, 56, 60-64, 66, 75 hidden 6, 7, 295-313 implicative 1, 10, 13-17, 21, 27 in absentia 369 intersubjective 64 objective 26, 107, 111 physical 10

problematic 1, 10, 12, 13-15, 27 subjective 82, 88, 106-108, 111 modalization 1, 9, 10, 24-27, 259, 355 a posteriori 24-26 a priori 24-26 constative 24-26 counterfactual 24-26 evaluative 24-26 qualified 24-26 simple 24, 25 monosemy 2, 114, 117, 118, 120 must 2, 5, 6, 17, 18, 26, 83-87, 89, 91, 93-96, 98-100, 110, 115, 117, 118, 128-131, 133-140, 145, 177198, 200-202, 205-207, 211-221, 242-245, 253, 254, 279, 280, 291, 321 narration 57, 58 necessity 1, 9, 23, 26, 51, 60, 63, 67, 69, 71, 79-82, 88, 91, 99, 107, 112, 113, 119, 131, 212, 214-218, 220, 241, 242, 276 need (to)27, 87, 134 negation 6, 69, 86, 87, 132-134, 149-153, 223, 226-228, 260, 272, 273, 305, 338-341 double negation 9, 227, 231, 232 negation raising pleonastic negation 231, 232, 361 neutral orientation 23 Ngiyambaa 67 non-certain non-factual 20, 22, 35, 37, 40, 48, 50-52, 70, 71 Norwegian 369 notional concord 230 notional conformity 168, 170, 174 notional domain 3, 4, 123, 125, 156, 159, 163-165, 168-170 not-yet-factual at time t 2, 31-53

Subject index 383

obligation 17, 28, 69, 80, 82, 84, 107, 177-187, 189, 205-208, 213, 216-218, 220, 280 Old English 1, 21, 23, 44, 321, 337, 343, 368 opaque 35, 238, 253-256, 262-264, 325, 352 orders of entities 2, 56-65, 70, 72, 75 otherness 4, 124-129 ought (to)24, 52, 69, 72, 110, 117, 118, 242, 343 perceptual 245, 247, 255 performativity 82, 107, 185-187, 195, 203, 220, 250 permission 80, 82, 89, 107, 159, 177, 178, 181, 182, 187, 188, 193, 322-324, 326, 340, 341, 361 polarity 226, 227, 232, 239, 246, 264, 295, 305, 310, 311 polysemy 2, 114, 117, 118, 189 possibility 1, 9, 10, 12, 13, 17-19, 26, 48, 51, 60, 67, 69, 71, 79-82, 87-91, 94, 106-109, 116, 118, 130, 131, 164, 232, 242-244, 276 factual possibility 37 theoretical possibility possible worlds 31, 32, 50, 52, 9194, 116, 277 posteriority 31, 35-37, 39-47, 49, 211-221 potential reality 55, 59, 60, 62, 63, 90, 116, 141, 244, 343 pre-culminative process 362, 364 predicative level 123, 151, 155, 168, 174-175 predicative relation 3-5, 124, 128, 156, 162-164, 168, 171-174, 271, 272, 288, 350, 356 prediction 17, 37, 40, 41, 49, 132, 186, 229, 233, 244, 278, 279 presupposition 74, 153, 226, 228, 229, 233

primary pragmatic process 114, 115, 120 probability 48, 80, 87, 111, 120, 232, 242, 243 projected reality 244 protasis 233 prototype 2, 79-81, 90, 92-96, 100, 124, 193 purposive to 7, 350-352, 354, 365, 368, 369 qualification 1, 110, 124-132, 134142, 170, 172, 174, 175, 286, 287, 363, 370 quantification 63, 124-132, 134-142, 170, 173, 218, 239, 363, 370 raising-come 7, 349-370 reliability 241-245, 259-261, 264 repeatability (of an event) 285, 298, 299, 301-303, 311 reported speech 278-283, 286 responsibility 237-240, 254-256, 262-264, 276, 281, 282, 285, 289 resultativeness 358, 361 root modality 1, 11-13, 118, 127, 145, 149, 159, 164, 212, 214-217 S-world 32-52 salient participant 253-255, 264, 364-366, 370 saturation 84, 85, 100, 114, 115, 118, 120 scanning (parcours) 3, 299, 312 scanning operator 3, 156, 172, 175, 299 scope 24, 150, 151, 217, 218, 270274 sculan 22, 23 semantic ascent 55-59, 70 semantic change 21, 58, 315, 324, 326-328, 368

384 Subject index shall 5, 6, 20, 22, 23, 27, 28, 110, 119, 120, 128-130, 133, 135-140, 177, 184, 186, 187, 189, 195, 199210 should 2, 24, 25, 42, 52, 56, 72-74, 85, 90, 91, 93-100, 110, 115-120, 202, 242 simple modalization 24, 25 source of evidence 245, 246 Spanish 11, 74 speaker commitment 6, 105-108, 111-114, 238, 240-252, 256, 259261, 263, 264, 269, 272, 275, 276, 280-282, 285, 287, 289, 290 speaker’s expectations 226-230, 288 speech-act verb 11, 27 spoken language 225, 232, 340 stance: 237-240 strong not-yet-factual interpretation 46, 47 subjectivity 2, 6, 33, 35, 39, 41, 55, 59, 75, 82-87, 90, 105-122, 125, 188, 237-240, 247, 252-256, 262264, 279, 282, 324 subjunctive 66, 74, 92, 315, 318321, 323-325, 337, 343 subsequentiality 360-363 sufficient condition 15, 16 syntactic mood 7, 56, 66-69, 75, 317, 319 t-world 32,52 temporal clause 2, 39, 42-45, 48, 50 to + infinitive 369

Tonkawa 67 truth 63, 65, 68, 70, 71, 87, 106, 107, 114, 242, 259, 261, 264, 276, 282, 285, 287, 289 truth-value 4, 13, 14, 16, 32-35, 160162, 192, 269, 271, 275, 277, 280, 283 Tuyuca 275 unreality 154 until-clause 2, 41 uton / uten / ute 315, 321, 324, 325, 332, 343 validity 4, 92, 124-128, 131, 132, 141, 150-156, 162-164, 168, 171175, 238, 240, 241, 245-252, 256, 259-261, 264 volition strong, subjectal, weak 1, 12, 17-24, 28, 51, 55, 67, 80, 87, 112, 332, 368 weak not-yet-factual interpretation 46, 47 will 5, 16-24, 26-28, 49, 81, 82, 8690, 93, 98-100, 110, 128-134, 136138, 145, 202, 225, 232, 233, 242245, 254, 278, 279, 343, 359, 360 willan 21-23 would 119, 202, 223, 225, 226, 229232, 242-245