Modality in Contemporary English 9783110895339, 9783110176865

This book offers original theoretical accounts and a wealth of descriptive information concerning modality in present-da

247 55 10MB

English Pages 412 Year 2003

Report DMCA / Copyright


Table of contents :
Modality in English: Theoretical, descriptive and typological issues
The semantics and pragmatics of core modal verbs
Irrealis, past time reference and modality
Modal auxiliary constructions, ΤΑΜ and interrogatives
A pragmatic analysis of the epistemic would construction in English
Towards a contextual micro-analysis of the nonequivalence of might and could
The status of emerging modal items
On two distinct uses of go as a conjoined marker of evaluative modality
Had better and might as well·. On the margins of modality?
What you and I want: A functional approach to verb complementation of modal WANT TO
Between epistemic modality and degree: The case of really
Modality on the move: The English modal auxiliaries 1961-1992
Changes in the modals and semi-modals of strong obligation and epistemic necessity in recent British English
Shall and will in contemporary English: A comparison with past uses
Pragmatic and sociological constraints on the functions of may in contemporary British English
Sociolinguistic variation and syntactic models
The role of epistemic modality in women’s talk
Double modals in the southern United States: Syntactic structure or syntactic structures?
Modal verbs in Tyneside English: Evidence for (socio)linguistic theory
Subject index
Recommend Papers

Modality in Contemporary English
 9783110895339, 9783110176865

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Modality in Contemporary English


Topics in English Linguistics 44


Bernd Kortmann Elizabeth Closs Traugott

Mouton de Gruyter Berlin · New York

Modality in Contemporary English

Edited by

Roberta Facchinetti Manfred Krug Frank Palmer

Mouton de Gruyter Berlin · New York 2003

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

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

I S B N 3-11-017686-6 Bibliographic information published by Die Deutsche


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

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

Preface Roberta Facchinetti, Manfred Krug, Frank Palmer

This volume is largely the result of a conference on contemporary English modality held at the University of Verona in September 2001, the aim of which was to foster interaction among scholars from different theoretical backgrounds with a common interest in English modality by promoting discussion, the exchange of ideas and the reporting of recent progress in the field. During the conference special attention was paid to (a) the syntax and semantics of modal elements, (b) modality and sociolinguistics and (c) modality and pragmatics. These three fields of study were discussed in detail from different perspectives, both theoretical and corpus-based: Generative syntactic theory, relevance theory and cognitive semantics each figured in various contributions; at the same time, the grammar of English modality was analyzed by means of empirical studies with data drawn mostly from synchronic corpora, both spoken and written, but also, to a lesser extent, from diachronic databases. Only a small selection of the over 50 papers presented at the conference have reached the publication stage in this book; despite differences in theoretical perspectives, terminology and topics, all of these papers offer fresh methodological impetus to a variety of current linguistic debates, converging around the following key issues: -

The semantics and pragmatics of core modal verbs; The status of emerging modal items; Stylistic variation and change; Sociolinguistic variation and syntactic models.

Each one of these areas is dealt with in a section of the present volume and all of them are framed by Frank Palmer's introductory paper which offers an overview of the most topical issues in contemporary English modality from a functional perspective. After qualifying mood as largely a semantically vacuous subcategory within the broader field of modality, being governed almost exclusively by grammatical rather than semantic rules, he recalls the three main modal categories: epistemic, deontic and dynamic, and mentions a fourth one - evidential - which appears to be dominant in languages other than English, such as native American


Roberta Facchinetti, Manfred Krug, Frank Palmer

languages and the languages of Papua New Guinea. Within this notional framework, he attempts a theoretical, typological and functional explanation - rather than positing a close answer - to a range of specific issues, including can't as the suppletive form of negative epistemic must, the deontic functions of mustn't, the role of main-verb and modal-verb negation and the role of will in interrogative sentences with second person subjects. He also discusses the relationship between these modal verbs and semi-modals which are semantically close to them - be going to, have to, need, be supposed to, had better and would rather.

The semantics and pragmatics of core modal verbs Modality is realized by linguistic items from a wide range of grammatical classes, covering not only modal auxiliaries and lexical verbs, but also nouns, adjectives, adverbs, idioms, particles, mood and prosody in speech. Despite these many possibilities, studies have been concerned more with modal verbs than with other linguistic forms, and have divided them into sub-classes according to shared syntactic and morphological features. Hence, oppositions such as "core/principal/central" vs. "marginal/semi/quasi-modals", "primary" vs. "secondary", or simply "modals" vs. "modals to a lesser extent" have been used in order to distinguish verbs which are loosely interrelated in terms of common features - such as need, dare and be going to - from those which share the same NICE properties (negation, inversion, code, emphatic affirmation). Such great terminological diversity is mirrored in the different choices made by the contributors in the papers of the present book, though the title of this section adopts the general label of "core" modals to refer to those verbs which share all of the NICE properties; these verbs are may, might, can, could, shall, should, will, would and must. Issues of grammaticalization, of semantics and pragmatics, but also, though less extensively, of syntax, are taken up by the four contributors of this first section. Paul Larreya deals with the rules governing the use of the past tense morpheme in the grammar of English modals and suggests that, irrespectively of whether the past tense morpheme is carried by a lexical verb or by a modal, semantically it always expresses a particular type of presupposition. Starting from this assumption, he claims that in all of its uses - whether "temporal", "hypothetical" or "tentative" - the past tense morpheme expresses some type of presupposed (implicit) unreality. This presupposition of unreality is either "absolute" (i.e. counterfactual), as 'ml wish I knew the answer, or "relative", as in if I won the lottery... (which



presupposes ...but I'm not very likely to win it). It may also be either "direct" (as in I wish I knew the answer), or "indirect", as in you might have broken your leg. These different types of unreal presupposition make it possible to account for different types of "temporal", "hypothetical" or "tentative" uses of the past tense in the grammar of the English verb phrase. While Larreya's theoretical discussion finds ample confirmation in a host of examples mostly from could, should, would and might, Richard Matthews mainly focuses on might and could, and opens up the scope of his analysis to the pragmatic consequences of grammaticalization on modal concepts, in order to investigate how "modalized questions" are restricted in the interpretation of open and non-open questions. Specifically, for epistemically interpretable modals, a pragmatic differentiation between modal "submissions", "dissent" and "challenges" is made, while, for deontically interpretable modals, the differentiation is between modal enquiries, modal dissent and modal challenges; finally, within dynamic interpretations, Matthews distinguishes between modal enquiries and modal challenges. He proposes that modal submissions and enquiries permit only a rather limited set of modal auxiliary expressions, but that (subjective) modality forms like might are acceptable. This suggests that it is "modalized modality", i.e. so called remote forms, that is possible in "open" interrogation. This would also support the view that "open" questions are themselves inherently modal, i.e. without an ascription to the truth-value of the propositions they contain. Stephane Gresset also focuses on the values of might and could in order to reject the general belief that might and could may be quasiequivalent or that could is replacing might as the main exponent of tentative epistemic possibility. His discussion is based on a contextual microanalysis of two samples which testify to the fact that, however close the interpretations of might and could may be, some contextual features can generally be highlighted which pave the way for one modal rather than for the other, and these contextual elements can be related to the basic meanings of can and may, i.e. "unilateral" vs. "bilateral possibility". This opposition, supported by diachronic research and grammaticalization studies showing that root modality and epistemic modality are related, accounts for both the distinctive and the common interpretations of can/could and may/might. Finally, semantics and pragmatics are shown to be closely interrelated in the paper by Gregory Ward, Betty J. Birner and Jeffrey P. Kaplan, who start from the widely acknowledged assumption that a contextually salient open proposition (OP) is relevant for the felicity of a variety of constructions. For instance, the authors suggest the sentence That would be


Roberta Facchinetti, Manfred Krug, Frank Palmer

J.K. Rowling as a possible response to Who's the British woman over there? Here the speaker instantiates the variable in the OP "The British woman over there is X" with the value "J. K. Rowling" while conveying an epistemic disposition towards the selection of that particular instantiation. Ward et al. remark in the first place that previous work on modality has failed both to recognize the role of open propositions in the use of this construction and to characterize accurately its epistemic effect. Secondly, they show that the use of epistemic would requires a previously evoked OP, instantiates the variable with a discrete member of some salient set, and conventionally implicates that the speaker has conclusive objective evidence for the truth of the proposition.

The status of emerging modal items In the past, the central English modals were the prime focus of research in modality, partly because they are identifiable in a relatively straightforward way with the help of the NICE properties. Modality, however, is a gradient (and not a binary) notion, and change within the modal system of any one language is endemic. Consequently there is a host of modal expressions in contemporary English, both verbal and non-verbal, which either already exhibit full-fledged modal features, or at least show signs of change towards more modal behaviour, i.e. of grammaticalization. Such items have recently become an additional focus of attention. It is in particular their semantic and pragmatic, but also their morpho-syntactic, properties which lead an increasing number of researchers to assume that these, too, are (emerging) modals - despite the fact that prima facie, i.e. on strict morpho-syntactic grounds on a strict morpho-syntactic level, they do not appear to be auxiliaries. While such modal expressions are still poorly understood compared to the rather well-investigated central modals, the papers in this section (as well as the contributions by Coates, Leech and Smith in later sections) represent an important step towards a better understanding of individual members of this group, and it is hoped that they can instigate further research into the dynamic and diverse class of emerging modals. Exemplary of research that adopts a wider definition of modality is the paper by Philippe Bourdin, who discusses the constructions go- V-en (as in The disease usually goes undetected until its later, more dangerous stages) and go-V-ing (as in Whenever I let him cook, he goes burning everything). In the data he presents, the verb go has been stripped of the notion of motion, wholly or in part, and of its spatial dimension; this has allowed go



to evolve into an exponent of aspectual values that are concomitants of the semantics of the un-V-en and V-ing constructions. Formally, neither instantiation of go meets any of the standard criteria of auxiliarihood. Yet both instances of go have assumed abstract, grammatical (aspectual) meanings at the expense of more concrete (spatial) ones (a process often referred to as "desemanticizaton" or "bleaching"), which is standardly associated with grammaticalization. In addition, both go constructions have taken on a modal value; indeed, the speaker is evaluating a state of affairs which is viewed as contravening the expectations one would normally have in the relevant type of situation. Still within the field of verbal modality, Keith Mitchell focuses on had better and might as well and argues in favour of a unitary analysis of the two expressions as an "inverse" (or "dual") pair, i.e. as a pair of expressions which are semantically related to each other in the same way as other pairs of modal expressions involving necessity and possibility or positive and neutral volition. Indeed, had better and might as well both contain an expression of a comparative relation: in the former case a relation of the "greater than" type and in the latter a relation of the "equal to or greater than" type - these two types of relation themselves standing in an inverse relationship to each other. Moreover, Mitchell shows that the two expressions form part of a system of options underlying the illocutionary act of giving advice, which is interpreted as involving the speaker's judgement of the comparative advantage of one course of action over another. Data from various corpora suggest, however, that the use of the two expressions is not restricted to advice-giving and that they have undergone a certain amount of semantic bleaching as well and so have taken on a more general deontic directive or decision-making function. Want to is another verbal construction that seems to enjoy incipient modal status, particularly so in informal contemporary British English speech, as testified to by Heidi Verplaetse. In her theoretical discussion she demonstrates that "volition" is a significant and highly common semantic notion that is well-embedded in the field of modality. Moreover, findings from earlier analyses reflect the private character of the mental state of volitional want to. This is supported also by syntactic and other patterns of its use, as it figures typically with first and second person subjects. Verplaetse's semantic description of want to allows her, finally, to show that the expression of volition with want to is becoming increasingly internalized in the grammar and is gaining pace in contemporary English. In a case-study of non-verbal modality, Carita Paradis tackles the various readings of really within the framework of cognitive semantics. In her corpus-based study, she identifies and explains the emergence of three


Roberta Facchinetti, Manfred Krug, Frank Palmer

different readings of really. These different readings arise through the interaction between the ontological notion of [REALITY] and the construal of really in relation to the context. She distinguishes three types: really as a marker of evidentiality (truth attesting), of subjective emphasis and of degree (reinforcement). In her semantic/pragmatic analysis of a variety of uses of really, Paradis claims that its use is conditioned by the speaker's wish to qualify an expression epistemically with a judgement of truth as perceived by the speaker. This condition, then, acts as a motivating force on the type of conceptual representations that really evokes.

Stylistic variation and change Text-type oriented approaches to English modality form another recent research tradition. Rather than looking at individual modal items and the meanings they encode in isolation, such work has identified important (con-)textual, i.e. discourse and stylistic, factors that lead to the preference of one variant over another in certain situations. Some dichotomies such as written vs. spoken, formal vs. informal, scientific vs. non-scientific, technical vs. non-technical and American vs. British - to mention only a few - testify to the wide scope of stylistic variation that is determined by formality considerations, textual purpose, interpersonal relationships and geographical distribution. As is characteristic of such approaches, the papers found in this section are all corpus-based and employ quantitative methods. What is intimately related to stylistic variation is change, as change tends to proceed by diffusion through text types. It is no coincidence, therefore, that all of the papers in this section address the role of diachrony in synchronic variation. Geoffrey Leech's paper, for instance, offers a quantitative study of how and why central modals are declining in both British and American English, while a few semi-modals are increasing. His results are evaluated in a broad theoretical framework of three "-ization processes": Americanization, grammaticalization and colloquialization. He also mentions a possible democratization process in the use of the modals, leading to a less authoritative role by the speaker or writer. In so doing he discusses the tenets according to which changes in recent English have been led by American English, and suggests that semi-modals (like be to, be going to, have to, need to and had better) are gradually emerging as a new generation (or, more technically, layer) of modals through a gradual process of grammaticalization, thus usurping some functions of the older central modal auxiliaries. His findings are based on corpus evidence from



British and American English, both spoken and written (LOB, FLOB corpora for written British English, BROWN and FROWN corpora for written American English; the SEU and ICE-GB corpora for spoken British English). They show that the decline of the central modals is particularly marked in spoken discourse and is accompanied by a tendency for modals to become more monosemous. One of the three above-mentioned "-ization processes" colloquialization - is taken up by Nicholas Smith as well, who provides further evidence for the gradual rise in frequency of semi-modals and the comparatively parallel decline in central modals mostly in the obligation/necessity domain. The data, also drawn from the LOB and FLOB corpora (British English textual samples from the nineteen sixties and nineties), provide an interesting pattern of short-term change particularly for root necessity modals, both in the groups of central modals (like must) and more marginal modal expressions (like have to and have got to). One possible reason for this, Smith suggests, is a tendency towards democratization of formal public discourse, in the sense of the elimination of overt power markers. Hence, increasing emphasis is being placed in both Leech's and Smith's papers on equality of power, or at least on the appearance of equality of power, while discourse patterns found in private conversation are making inroads into more formal text types. Maurizio Gotti's study spans a greater period of time than the two previous contributions as it compares the Present-day uses of shall and will to those occurring in previous centuries in order to trace significant evolutions in the development of their functions. Specifically, the data are drawn from a variety of sources both synchronic and diachronic, including the Early Modern English component of the Helsinki Corpus of English Texts. The use of modal meaning in contemporary English highlights significant variation in relation to the code used; indeed, he shows that in current spoken texts shall is mainly used to express prediction, while in written ones cases of obligation are as frequent as those of prediction. On the other hand, however, the prediction use of shall is increasingly being replaced by forms of will. Overall, the data show that in the course of time there has been a gradual decrease in deontic meanings, compensated by an increase in dynamic values; epistemicity shows an increase as well, although this usage remains the most limited of the three. Finally, the strong interrelationship between syntax, semantics and pragmatics with reference to stylistic variation is taken up by Roberta Facchinetti, who focuses on the factors governing the distribution of may across different text types. Her analysis of this modal in current British English leads her to show that existential interpretations of may disallow


Roberta Facchinetti, Manfred Krug, Frank Palmer

the presence of progressive or perfective aspect and that a one-to-one correlation emerges between interrogatives and deonticity on the one hand and between main verb negation of may not with epistemicity on the other. In wider semantic and stylistic terms, her study testifies to the fact that in scientific and non-scientific discourse the balancing between reporting objective data and signalling subjective evaluation is fundamental, since when presenting informational content, scholars follow specific rhetorical strategies and their stance is at least partially influenced by their audience.

Sociolinguistic variation and syntactic models Language variation is also influenced by a number of aspects of human behaviour which relate to the complex of social and cultural identity. As is well known, independently of the context of speech, speakers differing in terms of social class, ethnic group, age or sex, for example, will also exhibit differing patterns of speech. What is perhaps less well known is to what extent modal markers are exploited as identity markers. This is illustrated by Jennifer Coates, who explores the use of epistemic modal forms in the spontaneous talk of women friends, with specific reference to linguistic hedges like perhaps, I think, sort of and probably. Her data are drawn largely from a corpus of twenty spontaneous conversations between women friends and a set of ethnographic interviews with fifteen women about friendship. The results allow her to point out that epistemic modal forms have a multifunctional potential; indeed, at least four functions are widely expressed: (a) doubt and confidence; (b) sensitivity to others' feelings; (3) searching for the right word; and (4) avoidance of expert status. Bearing in mind Brown and Levinson's theory of politeness, she suggests that English epistemic modal forms are extremely useful in terms of satisfying 'face needs', since they have the effect of mitigating the force of what is said. Gender differences apart, linguistic variation is also strongly affected by regional factors. Interesting areal patterns of variation, however, pertain not only to major varieties of standard English like British and American, as discussed in the previous section by Leech and Smith; it is in particular minor, hence lesser known and often nonstandard geographical varieties which exhibit intriguing lexical, grammatical and syntactic patterns, as is shown in the last two papers of the volume. Specifically, Stephen J. Nagle focuses on the syntax of combinations of modal auxiliaries such as might could, which are a central feature of speech in the southern United States. He discusses a number of generative models, some of which are compatible, others incompatible, with Noam Chomsky's work. Nagle



stresses that the vast majority of double modal citations exhibit an ordering where the first modal (almost always may, might, or epistemic must) refers to the speaker's views on possibility or probability of the remainder of the predication. The scope of the second modal (variously can, could, will, would, should) is more directly the predicated action or state. He argues for a bi-clausal interpretation of double modals, which assumes that each modal occurs as the head of its own syntactic clause. He shows that this position offers advantages for explaining the curiosities of negation (either or both modals can be negated), perfectives (either or both modals allow perfective have) and apparent anomalies such as might didn't and might hadn't. Finally, syntax, semantics and sociolinguistic issues appear to be strongly interrelated in Graeme Trousdale's contribution on Tyneside English (TE). This variety exhibits a few well-defined differences from that of standard British English; for instance, can has a highly localized negative form cannit, while may (as a marker of both epistemic and deontic possibility) is part of the standard system, but rare in TE, where such modalities are expressed by other modal verbs. Trousdale explores such differences and investigates patterns of sociolinguistic variation (e.g. age and gender) surrounding the semantics and morpho-syntax of modal verbs in TE. The data are derived from recent recordings of twenty speakers of TE, made broadly following a social network model. He shows that a holistic approach to variation in the speech community, informed by knowledge of both sociolinguistic and formal linguistic theory, can best account for the data; specifically, the author argues that the patterns emerging from the data require an analysis which considers both internal and external conditioning factors for variation and change in the modal verb system of TE. As can be seen from this brief overview of the contributions to the present volume, theoretical ideas, descriptive information and empirical studies merge in what we hope has become a balanced book on the manifold issues related to modality in general, and to modality in contemporary English in particular. It was intended as a work that reflects the state of the art; this involves, on the one hand, displaying the diversity, but at the same time, such a work needs to reveal common new trends in order to pave the way for future research. We hope that the book will be helpful and a source of inspiration for all those who are interested in modality. December 2002 Roberta Facchinetti, Manfred Krug and Frank Palmer


Preface Roberta Facchinetti, Manfred Krug, Frank Palmer

Modality in English: Theoretical, descriptive and typological issues Frank Palmer

The semantics and pragmatics of core modal verbs Irrealis, past time reference and modality Paul Larreya Modal auxiliary constructions, ΤΑΜ and interrogatives Richard Matthews A pragmatic analysis of the epistemic would construction in English Gregory Ward, Betty J. Birner, Jeffrey P. Kaplan Towards a contextual micro-analysis of the non-equivalence of might and could Stephane Gresset

The status of emerging modal items On two distinct uses of go as a conjoined marker of evaluative modality Philippe Bourdin Had better and might as well: On the margins of modality? Keith Mitchell What you and I want: A functional approach to verb complementation of modal want to Heidi Verplaetse

Contents Between epistemic modality and degree: The case of really Carita Paradis

xvi 191

Stylistic variation and change Modality on the move: The English modal auxiliaries 1961-1992 Geoffrey Leech


Changes in the modals and semi-modals of strong obligation and epistemic necessity in recent British English Nicholas Smith


Shall and will in contemporary English: A comparison with past uses Maurizio Gotti


Pragmatic and sociological constraints on the functions of may in contemporary British English Roberta Facchinetti


Sociolinguistic variation and syntactic models The role of epistemic modality in women's talk Jennifer Coates


Double modals in the southern United States: Syntactic structure or syntactic structures? Stephen J. Nagle


Modal verbs in Tyneside English: Evidence for (socio)linguistic theory Graeme Trousdale


Subject index


Modality in English: Theoretical, descriptive and typological issues Frank Palmer



There is no doubt that the syntax and semantics of modality present one of the biggest problems for grammatical analysis and that the modal verbs of English present a number of idiosyncratic difficulties. As a result, they are particularly troublesome to those who wish to learn the language. There are very many questions that seem to need an answer, such as the following. (i) (ii) (iii)


(v) (vi) (vii)

Why is should used in I'm surprised that you should say that? Why does epistemic can't seem to be the negative of mustn 't in He must be in the office/He can't be in the office? Why is mustn't used to express an obligation not to (a prohibition) whereas German musst nicht indicates no obligation as with English neednΉ Why does You 're not supposed to do that mean "You're supposed not to do that" and I don't think he's here mean "I think he's not here"? Why is it not (any longer) possible to say When I arrived in London, I must take a taxP. Why can't we say I ran fast and could catch the bus? Why is Will you come to the party normally understood to be an invitation to the party, not a question about whether you are coming?

It is also reasonable to consider how far such questions can be answered within a typological context and how far they relate only to some idiosyncrasy of English. It is always a little dangerous to suggest that there ARE answers to such questions, for, in a sense there are never real answers in linguistic analysis. All that can be done is to state the facts and set them in some kind of systematic overall framework. This is, perhaps, not so much explanation as "explication". This paper will, therefore, set out briefly the framework within which it is possible to deal with modality and then consider some of the major issues that are involved.



Frank Palmer

Modality, mood and modal system

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. Another such sub-category is what may be called "modal system". For that reason, a clear distinction is made between mood and modality, the term "modality" being used for the wider category and "mood" for just one of the sub-categories of modality. Unfortunately, this distinction was not clearly understood in the first book on the subject published by the author (Palmer 1986), which was entitled Mood and Modality (and this title was used as the name of a conference in Albuquerque in 1992). A second edition of the book was published in 2001 and the simpler title of just Modality would have been preferable, but the publishers insisted that a second edition should have the same name as the first. So it too is entitled Mood and Modality, although it is made clear in the text that the name used for the grammatical category is simply "modality" and that there are two sub-categories of modality, mood and modal system. Mood is exemplified by the contrast between indicative and subjunctive in many classical and modern languages of Europe. A very similar contrast is made for other languages, especially in the Native American languages and the languages of Papua New Guinea in terms of "realis" and "irrealis" (see, e.g., Palmer 2001 Ch. 6). This seems, however, to be essentially only a terminological difference, possibly because the authors were not classical scholars and were unfamiliar with the notion of subjunctive, but also, perhaps, because of the way it is marked - not always as a formal feature of the verb. Modal systems are best illustrated by the functions of the modal verbs of English and possibly some other European languages. The most important distinction between mood and modal system is that mood usually (or basically, prototypically) involves a binary system, but a modal system does not. With mood 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 "non-modal" will be used for simplicity). In contrast, with modal systems, although there is usually a realis non-modal (or realis) form, which is usually unmarked, there is always a set of modal forms that form a modal system. Consider the English:

Modality in English (1)

Non-modal (realis) Modal (irrealis) (i) (ii) (iii)


They are in the office. They may be in the office. They must be in the office. They ΊI be in the office.

The unmarked non-modal form, the declarative, simply states what the speaker believes, or claims to believe, is a fact, but the three terms in the modal system simply indicate judgements by the speaker. These may be seen in terms of three types of conclusion, a possible conclusion with may, the only possible conclusion with must and a reasonable conclusion with will. A second and less important distinction is that mood is sometimes redundant in that the use of the subjunctive or of irrealis is wholly determined by the grammar and is, therefore, semantically vacuous. Thus in Latin the subjunctive is obligatory after the conjunction ut to express a result. Similar examples of an obligatory or vacuous subjunctive can be found in certain types of subordinate clause in all the modern Romance languages, including Italian. In contrast, with modal systems the terms are always contrastive and meaningful. Moreover, it seems to be the case that, to a large extent, the two types of modality are mutually exclusive: languages have either mood or modality, but not both. This is certainly the case with English, where the subjunctive has died out and the modal system has developed (see Lightfoot 1979; Plank 1984). English is almost unique in that there are very clear formal markers that distinguish its eight modal verbs. There is no need for a detailed discussion, as the facts are well known. Briefly they are: (i) (ii) (iii)

the modal verbs have all the characteristics of auxiliaries - their use in negation inversion, "code" and emphatic affirmation [what Huddleston (1976: 333) calls "Palmer's NICE properties"]; they have no third person -s form - no*cans, *mays etc.; they cannot co-occur in standard English (cf. Nagle, this volume or Trousdale, this volume, for double modals in nonstandard dialects) no *may can etc.

However, it would be a mistake to suppose that there is a neat and symmetrical system here. There are many problematic features such as the status of could, might and should and the relation of ought to to must. There is also a problem with the status of the semi-modals, especially BE GOING TO, HAVE (GOT) TO and BE ABLE TO, which are fairly recent additions to the

language (see Krug 2000), for these are often used in contexts in which modals are not available. It should be added that there is no subjunctive in


Frank Palmer

English, even though traditional scholars have seen the subjunctive in forms such as: (2)

a. b. c. d. e.

Long live the queen! Ipropose they be excluded. May he rest in peace! I'm surprised that you should think that. If I were you.

None of these is like the subjunctive with which we are familiar in, say, Latin. In the first two there is simply the uninflected form of the verb, which is also used for the infinitive and imperative. All that need be said is that this uninflected form has a number of different functions, but it is the same form throughout and not a completely different form like the subjunctives of other languages. In the next two examples there is simply a modal verb (may and should) whose function should be explained in terms of the English modal system, not in terms of mood. In the final example, were is a form unchanged for number and person, like other past tense forms such as knew (and it is the use of was elsewhere that is irregular). Modern English has a modal system, but no mood. It seems that modal verbs cannot be identified quite so easily in other European languages, though one purely syntactic feature (to do with word order) has been suggested for French (Kayne 1975: 1-27) and a different one for German (Hammer 1983: 224). In addition the modal verbs of German are marked morphologically. It seems that the potentially modal verbs of Spanish or Italian cannot be identified by any generally agreed formal marking (Spanish PODER and DEBER, Italian POTERE and DOVERE). However, they appear to share with the English modal verbs two features that are discussed in this paper - (i) they can express all three kinds of modality - epistemic, deontic and dynamic), (ii) they can be interpreted in terms of possibility and necessity. Even so, there are still problems, e.g., with the status of other forms such as French il faut and the reflexive il se peut, Spanish TEN ER QUE and hay que and Italian bisogna. Moreover, it may well be the case that mood is already on its way out in some of these languages. The subjunctive is used far less frequently in spoken German and French than as prescribed by the grammar books, and although it may be alive and well in the languages of Spain, that does not seem to be entirely true of Italian.

Modality in English 3.


The notional basis of modality

With any typological grammatical category it is necessary to recognize not only the language-specific formal features that establish the grammatical status of the category in individual languages, but also the shared cross-linguistic notional or semantic features across languages. (The term "notional" is used to avoid the issue of semantics vs. pragmatics.) Modality seems to be closely related with tense and aspect features - indeed, the acronym "ΤΑΜ" has been used to refer to the three categories that are associated (in the more familiar languages at least), with the verb - Tense, Aspect, Modality. It seems fairly clear that the notional feature associated with tense is time of the event or situation referred to and that the feature associated with aspect is the nature of the event or situation. The feature associated with modality is slightly different in that it does not relate directly to the event or situation, but to the status of the proposition that describes the event or situation. Prima facie it might seem to be concerned with whether the proposition is true (or factual) or not, but this is not entirely satisfactory. More convincing are the arguments of Lunn (1995: 430) that the essential feature of modality is that of "assertion" vs. "non-assertion". She links, in Spanish, the choice of the indicative to assertion and the choice of the subjunctive to non-assertion. She suggests that a proposition may be unworthy of assertion for three reasons: (i) the speaker has doubts about its veracity, (ii) the proposition is unrealized, and (iii) the proposition is presupposed. Her examples are the Spanish: (3)




Dudo que sea buena idea. I doubt that's a good idea. Ί doubt that's a good idea.' Necesito que me devuelvas ese libro. I need you to me to return that book. Ί need to return that book to me.' Me alegra que sepas la verdad. I'm glad that you knowthe truth. 'I'm glad that you know the truth.'

In all three the subjunctive is used in the subordinate clause, sea in the first, devuelvas in the second and sepas in the third. With the first, the speaker is not sure that what is stated is true. In the second, the speaker indicates that event of returning the book is not realized. It is not unexpected that the subjunctive is used in these. But the third example is, prima facie,


Frank Palmer

surprising. This is an example of presupposition, where both the speaker and the addressee accept the proposition (that you know the truth) as true. Now if the issue were the factuality or truth of the proposition, one would expect the indicative to be used here, for the proposition is clearly true or factual. Indeed, Givon (1994: 304) suggests that it is not merely (in his terms) "factive" or "realis", but "super realis". It is necessary, then, to explain why the subjunctive is used and Lunn provides the answer - the proposition is not being asserted by the speaker precisely because the addressee already knows and accepts it. It is assertion vs. non-assertion that is the determining factor, not truth or factuality. Similar examples of subjunctive marking for what is presupposed can be found in subordinate clauses in French and Italian. It is here that we can deal with question (i) - why should is used in I'm surprised that you should think that. For this is very much like the use of the subjunctive where there is presupposition - the proposition (You think that) is accepted by the addressee and so is not asserted by the speaker. English, like the Romance languages can use a modal (irrealis) form - but one from its modal system, not the subjunctive mood. Precisely what features are to be considered as modal or irrealis and so to be marked by the subjunctive or, in the languages with realis/irrealis marking, as irrealis, differs from language to language. Such variation is a feature of all typological categories. But there are some interesting cross-linguistic parallels. For instance, there is one Native Language of America (Chafe 1995: 357) that uses Irrealis marking with negation, prohibition, obligation, conditions, "as if', "seldom", all of which are obviously non-assertive, but also with the "Admirative" which expresses surprise. For an expression of surprise involves presupposition - both speaker and addressee accept the truth of the proposition and so is non-assertive, though its contents are factual. Similarly, verbs of surprise in Romance languages are often followed by the subjunctive. Another, but much more speculative point concerns the fact that some other languages, in both America and Papua New Guinea, mark Future as irrealis and Present and Past as realis, presumably because the past and present are known while the future is not. Can this, perhaps, be related to the fact that one way of indicating the future in English is by the use of WILL and SHALL? For these are formally (in both syntax and morphology) modal verbs. So their use might be taken to indicate that the future is treated as irrealis or modal in English too. Of course, English also has be going to, but this point actually strengthens the argument concerning the modality of WILL and SHALL, because the notional difference between WILL/SHALL and BE GOING TO can then be stated in terms of modality: WILL

Modality in English and SHALL are more conditional, less assertive than Palmer 1980: 147-159).




7 (see

Kinds of modality

There are three types of modality that can be distinguished in the modal system of English. These, following the logician von Wright (1951: 1-2), have been called "Epistemic", "Deontic" and "Dynamic". They can be exemplified in English and may be referred to as "kinds" of modality: (4)

Epistemic: They may be in the office. - They must be in the office. Deontic: They may/can come in now. — They must come in now. Dynamic: They can run very fast. -1 will help you.

Epistemic modality is concerned solely with the speaker's attitude to status of the proposition. Thus in the examples above the speaker makes the judgements that it is possible or necessary (necessarily the case that) that they are in the office. Deontic and Dynamic modality relate directly to the potentiality of the event signalled by the proposition, but of two different types, both of which may both be seen as "directive" - getting things done. Deontic modality is directive in that the event is controlled by circumstances external to the subject of the sentence (more strictly the person or persons identified by the subject). In particular, permission is given with MAY (as in the example above) and an obligation is laid with MUST. With Dynamic modality the control is internal to the subject - in the examples above it is the subject's ability to run fast with CAN and the speaker's willingness to help with WILL. The two different types of directive are very clearly shown in the two uses of CAN (deontic and dynamic). Both are concerned with enabling the subject of the sentence to act, but with Deontic the ability comes from the permission given (externally), with Dynamic the ability comes from the subject's own (internal) ability. There are many languages in which these three kinds of modality are found, and, as in English, the same forms are often used for them. But, in some languages, particularly some of the Native languages of America and the languages of Papua New Guinea, the dominant type of modality is different: there is a fourth kind, "Evidential" modality, in which, instead of making a judgement about the truth-value of the proposition, the speaker offers evidence for it. There are two main kinds of evidence - report and sensory. Sensory evidence relates to the evidence of the senses - what is seen, heard or even smelt and is found in a number of non-European


Frank Palmer

languages. Report relates to what has been or is said by individuals or by the community. There are three points to make here. First, in Europe, evidence from reports is found in Germanic languages e.g., in German with the verb SOLLEN (Hammer 1983: 231): (5)

Er soll steinreich sein. He is said enormously rich to be. 'He is said to be enormously rich.'

Secondly, Epistemic MUST and WILL have some characteristics of Evidential modality, for they signal conclusions that are based on evidence as, shown by the glosses in: (6)

a. b.

They must be in the office. (The lights are on) They ΊI be in the office. (They always are at this time)

usually indicates a conclusion based on available evidence - here, perhaps, that the office lights are on, while WILL suggests a conclusion from what is generally the case - here, perhaps, that they usually are in the office at this time. These two categories, which we might call "Deductive" and "Assumptive" are often to be found in languages with full Evidential systems involving both Report and Sensory evidence. The overall systems are different, but they appear to share some basic modal terms. Thirdly, all four kinds of modality involve non-assertion, and are, in that sense concerned with the proposition. However Epistemic and Evidential modality are concerned more directly with its status - with the reasons for its non-assertion, whereas Deontic and Dynamic are "directive", as suggested above, and concerned with the type of direction involved. Epistemic and Evidential modality might, then, be characterized as "Propositional modality" and Deontic and Dynamic modality as "Event modality". Moreover, in that sense, it might be argued that Propositional modality is the more modal of the two. On a related note, Propositional modality often develops historically out of Deontic and Dynamic modality (for instance, English MAY: "ability" or "root possibility" -> "logical possibility"; cf. Sweetser 1990; Bybee et al. 1994). This, combined with the higher degree of abstractness expressed by Propositional modality, has led some scholars (e.g. Heine et al. 1991) to claim that Propositional modality is not only more modal but also more grammatical than Dynamic and Deontic modality. MUST

Modality in English 5.


Possibility, necessity and negation

There is a problem with the negation of the epistemic and deontic verbs of possibility and necessity (CAN, MAY, MUST and NEED). The issue is, prima facie, fairly simple. If a distinction is made between the modality and the proposition it should be possible to distinguish "not possible" and "possible not" - between "it is not possible for us to go" and "it is possible for us not to go" (and similarly for "necessary"). Some languages make the distinction by placing the negative either before the modal or the main verb, but English cannot do this - the negative marker always follows the modal either as a clitic -n7 or as not. As a result English has two devices to make the distinction. First, it may use a different verb, with e.g., MAY and CAN for epistemic possibility (a similar relationship obtains between MUST and NEED for deontic necessity): (7)

a. b. c.

They may be in the office. They can't be in the office, ("not possible") They may not be in the office, ("possible not")

Secondly, it may make use of logic in that there is logical equivalence between "not possible" and "necessary not" and between "possible not" and "not necessary", and so use a logically equivalent form. Thus the negative forms in (8) are used as the negatives of MUST: (8)

a. b. c.

They must be in the office. They may not be in the office, ("possible not" for "not necessary") They can't be in the office, ("not possible" for "necessary not")

Can't here negates the modality - "It is not possible that they are in the office". In that sense it is the negative of CAN, but it also seems to function as the negative of MUST, in the sense that it is necessarily the case that they are not in the office. A possibility modal is used here, because MUST has no other way of expressing "necessary not" and because "not possible" is logically equivalent to "necessary not". But it is misleading to say that it is the negative of MUST: it is formally the negative of CAN, but used suppletively. There are many other languages where suppletion is used, but the pattern of suppletion is often different from that of English (see Palmer 1995: 461—464). This answers question (ii).

10 Frank Palmer However, in American English and in some dialects of British English must not is used where British English has can't. The only forms commonly available (for epistemic modality) are may not and must not which are most naturally taken to negate the proposition - "possible not" and "necessary not" respectively. Since again there are only two forms, there is again a need for suppletion, to supply the forms that negate the modality. With the logical equivalences, "not necessary" is being signalled by may not and "not possible" by must not. Compare the British (9b) with the American (9c): (9)

a. b. c.

They must be in the office. They may not be in the office, ("possible not" for "not necessary") They must not be in the office, ("necessary not")

There is one further point. There are some contexts where mustn 't or must not seems just possible even in the author's speech. Thanks are due to Kenji Kashino for this observation and the following examples: (10)

The restaurant is always empty. It mustn't be very good.

In both of these examples can't would be more normal, but mustn't is just possible, whereas it would not be possible in: (11) (12)

A: That woman over there - is it Mrs Wilson? B: It can't be Mrs Wilson - she died two years ago. A: Someone is knocking at the door. It may be Mary. B: It can't be Mary. She went to a movie tonight.

The context in which mustn't seems possible appears to be one in which it is important to stress "necessary not" rather than "not possible" - "it must be the case that not", rather than "it can't be the case that". So suppletion is not wholly obligatory. With deontic modals, the situation is slightly different. With possibility that expresses permission there is suppletion with needn't ("not necessary") for "possible not" ("not possible" being expressed by can V), as in: (13)

a. b. c.

You can/may come in. You needn't come in. ("not necessary" for "possible not") You can't /may not come in. ("not possible")

Modality in English


The first of the negated forms expresses "permission not to", the second "no permission". Yet there appears to be a regular and clear formal distinction between deontic "not possible" and "necessary not", since denial of permission with can't is not the same as expressing an obligation not to with mustn't. There is a clear difference between: (14)

a. b.

You can 't/may not come in. ("not possible" - no permission) You mustn't come in. ("necessary not" - obligation not to)

There is a further striking feature concerning negation and modality. It is well understood that the semantic difference between deontic needn't and mustn't is that the former expresses "not necessary" and the latter "necessary not", as in: (15)

a. b.

You needn't come in. (It's not necessary for you to come) You mustn't come in. (It's necessary for you not to come)

There is, then, negation of the modality with needn't and negation of the proposition with mustn't. There is a difference in scope of the negation. Formally (morphologically) the negative (the cliticized -n't) is associated with the modal verb, and, in an ideal world, it might be expected that formal negation of the modal verb would indicate negation of the modality. In this sense we might say that needn V is regular, with normal scope, while mustn't is irregular, and that with mustn't the negation is "displaced". This answers question (iii). This feature too is found in many other, unrelated languages, almost always with the negation of deontic necessity (see Palmer 1995: 465-468). Yet, in the Germanic languages, such displacement seems to be found only in English and Norwegian (Palmer 1997: 147). Standard German musst nicht, for instance means "needn't", not "mustn't": there is no misplacement of the negative. This displacement of the negative is also found with oughtn't and shouldn't, which express weak deontic necessity, as in the following examples, which give strong advice for the addressee not to come in: (16)

You shouldn 't/oughtn't to come in.

There is a similar situation with the semi-modals, plus some other expressions of intermediate status (which will now be considered). Usually with HAVE TO there is no displacement, with negation of the modality ("not necessary", like needn't):

12 Frank Palmer (17)

He doesn 't have to come down because he's already here.

Yet (but much less likely), You don't have to do that might be interpreted as "necessary not" like mustn't, though not if have is stressed. In contrast, with BE SUPPOSED TO there is usually displacement, with negation of the proposition ("necessary not"): (18)

You 're not supposed to do that.

This answers part of question (iv). Also relevant are the verbs HAD BETTER and WOULD RATHER. With negation there is displacement in the former, but not the latter, as in: (19)

a. b.

You hadn 't better go. (advice not to go) I wouldn't rather go. (no preference to go)

It is relevant that all the verbs with displacement are verbs expressing obligation and there are five such verbs - MUST, OUGHT TO, SHOULD, BE SUPPOSED TO and HAD BETTER. The verbs of obligation that do not have displacement are HAVE TO and NEED TO and the same is true of WOULD RATHER.

Mention should also be made of be expected to. This can be interpreted epistemically and deontically, and in both senses it seems that it is ambiguous between negation of the modality and negation of the proposition: (20)

He's not expected to come. (It's not expected that he'll come OR It's expected that he won't come) (It's not required that he should come OR It's required that he should not come)

With deontic modality at least, it seems that both senses may be quite possible in the same sentence: (21)

You 're not expected to wear a jacket, but you 're not expected to be scruffy.

The two instances here are roughly equivalent to (but weaker than) needn 't and shouldn't respectively. However, not expected to is not a form of a semi-modal verb, but the passive of EXPECT, with active forms such as:



in English


I don't expect you to wear a jacket, but I don't expect you to be scruffy.

These forms of EXPECT can be compared with the forms of other lexical verbs where there appears to be displacement. Most obvious are verbs of thinking and believing, where the feature has usually been treated in terms of "negative raising": (23)

I don't think/believe/suppose that he's here.

This answers the second part of Question (iv). This feature is found in other languages too. These verbs are all epistemic but there is one verb that is deontic with displacement/negative raising - WANT. (24)

I don't want you to come. (I want you not to come)

This is not so with e.g., REQUIRE or ORDER·. (25)

I don't require/order you to come. (NOT = I require/order you not to come)

So the behaviour of EXPECT is paralleled by that of other verbs, either epistemic or deontic, although there is no other single verb with all of its interpretations.


Two idiosyncratic features of English?

There are two features that may be idiosyncratic to English. The first is illustrated by the fact that MUST has to be replaced by HAVE TO in: (26)

a. b.

* When I arrived I must change trains. When I arrived, I had to change trains.

An obvious explanation might be that MUST has no past tense form, which may be explainable by the fact that MUST itself is historically a past tense form. However, there is more to it than that. For it is not only impossible to express deontic necessity as here with MUST, but equally impossible to express deontic possibility (permission), with MAY, even though the form might is available:

14 Frank



*When I came to his office, I might enter.

This cannot mean "I had permission to enter". The explanation, perhaps, is that the deontic modals are prototypically performative. Performatives do not merely say something - they also do something. The classic example is of the celebrity saying "I name this ship . . . ". For that is the act of naming the ship, not just a report of the celebrity's action. Similarly the prototypical function of the deontic modals is performative in the sense that it is the speaker who gives permission or lays obligation. But the expression "prototypically" is important because the speaker is not always involved. A performative act, by definition, takes place at the time of speaking (the present) and so cannot be shown as being in the past. It is for this reason that the deontic modals are not used to express past time possibility (permission) or necessity (obligation). This answers question (v). In contrast, similar restrictions do not apply to the (potential) modals of other European languages. This is supported by the fact that the epistemic modals too cannot be used to refer to past time. It is not possible to say He might be there in the sense that it is/was possible that he was there, and, again, there is no past tense form available for epistemic MUST. Of course, one can make a present judgement about the past, as in: (28)

They may/must/will have been in the office.

But these are still present acts of judgement - albeit about past events. They are not judgements in the past. The second feature is illustrated by the fact that CAN and WILL are not possible in: (29)

a. b.

*/ ran fast and I could catch the bus. *I asked them and they would come.

The reason for this seems to be that there is an implication that the event actually took place - that I did catch the bus, that they did come. It seems that, as modal verbs, CAN and WILL are inappropriate to suggest not merely the ability or willingness, but also that the event actually took place. Modality is essentially non-assertive and for an assertion the non-modal declarative (/ caught the bus, they came) is available. But the negatives are possible because the event did NOT take place: (30)

a. b.

I ran fast but couldn 't catch the bus. I asked them but they wouldn 't come.

Modality in English


However, there is no problem with (31)

a. b.

I ran fast and was able to catch the bus. I asked them and they were willing to come.

The semi-modals BE ABLE TO and be WILLING TO are not affected by a similar restriction - presumably because they are not fully modal. This answers question (vi). There is, perhaps, yet another idiosyncratic feature of English, that has already been partly noticed: the fact that the semi-modals are used in a number of situations when modals are not available. They appear, that is to say, to fill gaps created by the peculiar morphology and syntax of the modals. The two features discussed in this section show the use of HAVE TO for MUST, BE ABLE TO for CAN and BE WILLING TO for WILL, but that is not all. The modal verbs have no non-finite forms - no infinitives of participles and also cannot co-occur. Semi-modals, therefore, replace the modals in, for instance: (32)

a. b.

They must be able to come. NOT *They must can come. They may have to come. NOT *They may must come.

There are expressions in many other languages that are marginally within the modal system, as was seen with French il faut etc. 7.

Final comments

Finally, it must be said that there are many other issues that might have been discussed. For instance, there was only brief mention of the forms could, might, would and should, which are formally the past tense forms of CAN, MAY, WILL and SHALL. With the exception of dynamic could and would, these are not used for simple past time reference, but are relevant to the analysis of conditional sentences. Also, nothing has been said about the pragmatics of modality. Yet this can be important. Consider: (33)

Will you come to the party this evening?

An English speaker will not normally interpret this as a question about what is going to happen, but will take it as an invitation and might well reply "Thank you very much, I'd love to". For that reason, to ask a question about a future intention one of the following would normally be used:

16 Frank Palmer (34)

a. b.

Are you coming go the party this evening? Will you be coming to the party?

In these examples the use of the progressive with WILL rules out the dynamic sense, since, while one may like to come, it is not likely that one would like to be in the process of coming. They thus rule out any possibility of misunderstanding. This answers question (vii). For similar reasons the progressive may be used to rule out an interpretation in terms of deontic necessity with MUST. (35)

They must go to the office. They must be going to the office.

The first would normally be treated as deontic - that there is an obligation to go. Again the use of the progressive in the second makes it clear that it is epistemic, for it is unlikely that there is an obligation to be going to the office. This is not a matter of any kind of grammatical or logical system. It is rather that there are pragmatic ways of ruling out possible ambiguities and misunderstandings. Understanding the grammar, complex though it may be, is still not enough.

References Bybee, Joan, 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. Bybee, Joan and Suzanne Fleischman (eds.) 1995 Modality in Grammar and Discourse. Amsterdam/Philadelphia: John Benjamins. Chafe, Wallace 1995 The realis-irrealis distinction in Caddo, the Northern Iroquioian languages, and English. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 349-365. Amsterdam/Philadelphia: John Benjamins. Givön, Talmy 1994 Irrealis and the subjunctive. Studies in Language 18: 265-337. Hammer, A. E. 1983 German Grammar and Usage. London: Edward Arnold. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: Chicago University Press.

Modality in English


Huddleston, Rodney 1976 Some theoretical issues in the description of the English verb. Lingua 40: 331-83. Kayne, Richard 1975 French Syntax and the Transformational Cycle. Cambridge, Massachussets/London: MIT Press. Krug, Manfred G. 2000 Emerging English Modals: A Corpus-Based Study of Grammaticalization. Berlin/New York: Mouton de Gruyter. Lightfoot, David 1979 Principles of Diachronic Syntax. 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. Palmer, Frank 1986 Mood and Modality. Cambridge: Cambridge University Press. 1990 Modality and the English Modals. 2d ed. London/New York: Longman. Original edition, New York: Longman, 1979. 1995 Negation and the modals of possibility and necessity. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 454-471. Amsterdam/Philadelphia: John Benjamins. 1997 Negation and modality in the Germanic languages. In Modality in Germanic Languages: Historical and Comparative Perspectives, Toril Swan, and Olaf Jansen Westvik, 133—49. Berlin/New York: Mouton de Gruyter. 2001 Mood and Modality. 2d ed. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Plank, Frans 1984 The modals story retold. Studies in Language 8: 305-364. Sweetser, Eve 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. von Wright, Georg 1951 An Essay in Modal Logic. Amsterdam: North Holland.

The semantics and pragmatics of core modal verbs

Irrealis, past time reference and modality Paul Larreya



As is well known, the past tense forms of English modals may be used to refer to the past, as in (1)

At the age of four she could speak three languages.1

but they often express irrealis or, more generally, hypothetical meaning, as in (2)

If only I could telephone them!

Moreover (see, for instance, Palmer 1979: 29-30), not all modals can express past time reference through their past tense forms, and those that can are submitted to restrictions on this "temporal" use of their past tenses. For these reasons, it is not quite appropriate to refer to the past tense of the modals (or, for that matter, of ordinary verbs), but calling it the preterite tense may not be a better option, so I will refer to the past tense marker as the "-ED morpheme". I will say, for instance, that wrote is the association of the verb write and of the -ED morpheme, and, as regards modals, that could, might, should and would are respectively the association of the modals can, may, shall and will and of the -ED morpheme.2 The semantic description of the -ED forms of modals that I will propose is based on the following hypothesis: the fundamental meaning of the -ED morpheme is not the expression of past time reference but rather the expression of a particular type of presupposition. More precisely, I will make the claim that, in all of its uses, -ED expresses some type of presupposed unreality. The existence of a conceptual link between irrealis and past time reference is fairly obvious: in many languages, and at least in some cases, they share the same markers. (For different explanations of the relationship, s e e - among others- Bybee 1995; Dahl 1997.) This conceptual link extends, somewhat marginally, to perfective aspect, in a few uses of HAVE + -EN which I will examine briefly in section 3.3.


Paul Larreya



The concept of presupposition that I will be using is a purely linguistic concept. So, I will disregard both logical definitions and pragmatic definitions, and, for instance, I will consider that, in most contexts, (3 a) linguistically presupposes (3b): (3)

a. b.

You should have told them. You did not tell them.

Linguistic presupposition is both a category of meaning and a category of form. It has specific linguistic forms, which may be either lexical, syntactic or prosodic. The primary function of presuppositional forms is to encode that part of the information conveyed by the utterance which is already known by the hearer (or presented as being known) and which serves as a functional support for "new" information- i.e. for the assertion. So, presuppositional forms generally convey "old" (or "given") information, but in some cases (e.g. for stylistic effect), they convey "new" information. (For a more precise definition of linguistic presupposition and an account of various ways in which presuppositional forms can be used for conveying "new" information, see Larreya and Watbled 1994: 71-74.) Another important characteristic of presupposition is that it underlies meaning proper - it belongs to the implicit component of meaning - and, although very different from implicature, it resembles it in this respect. Thus, the question (4)


Would you like some more tea?

undoubtedly contains the (presupposed) information b.

You have already had some tea.

but this information is only given indirectly. It is - so to speak - underneath the surface of the utterance's meaning. In this respect, presupposition resembles implicature. Unlike implicature, however, presupposition is always "spoken": it can always be related to one or several elements of form used in the utterance; thus, in (4a), the element of form which expresses the presupposition is more. Implicature, in contrast, is generally "unspoken". In Grice's (1975: 51) famous example There is a garage round the corner (said in response to somebody saying I am out of petrol), there is nothing in the form of the

Irrealis, past time reference and modality


utterance that can be directly related to the "implicated" meaning that the garage will be selling petrol.


The semantics of the -ED morpheme

The past tense morpheme is sometimes described, from a semantic point of view, as the marker of some sort of distance between the event and the reality of the present moment. This description is, in my opinion, correct, but not sufficient. What has to be added is that the non-reality expressed by the -ED morpheme is part of the presupposition of the utterance - not of its assertion. This means (among other things) that the unreality is implicit, and that, in order to analyze the meaning conveyed by the past tense marker in an utterance, one has to be ready to look, so to speak, beneath the surface of the meaning. The unreality expressed by -ED has two further characteristics, which I will examine in instances of "hypothetical" past tense. First, the unreality is either absolute or relative. It is absolute (i.e. counterfactual here in (5)


I wish I knew the answer.

which presupposes b.

I do not know the answer.

It is relative in (6)


If I won the lottery...

which - unlike If I win the lottery... - presupposes something like b.

...but I'm not very likely to win it.4

Secondly, the unreality is either direct or indirect. It is direct in cases where the presupposed unreality concerns the event5 which is referred to by the verb or the auxiliary that bears the -ED morpheme. Example (5a) is a case of direct unreality. Its presupposition (/ do not know the answer) concerns the fact of "knowing", designated by knew - the verb which is marked by -ED. Another example of direct unreality is (2) - repeated below as (7) since the unreal presupposition it contains (/ cannot telephone them) concerns the "circumstantial possibility" expressed by the verb could·.


Paul Larreya


If only I could telephone them!

The unreality, on the other hand, is indirect in (8)

You might have broken your leg.

In most of the contexts that can be imagined for it, this sentence contains a counterfactual presupposition ("You did not break your leg"), and there is little doubt that this presupposition is expressed by the -ED morpheme which is borne by the modal MAY. The unreality concerned, however, is not that of the epistemic possibility expressed by MAY - a possibility which is seen as existing (or having existed). The unreality is that of the event (that of "you broke your leg"). In brief, the meaning of the sentence is something like "Your breaking your leg was logically possible but did not occur". What we observe in the syntax is simply a case of syntactic "raising" - not the raising of a noun phrase, or of the negation, but the raising of a presuppositional marker, which is moved from last to first place in the verb phrase - from break (your leg) to MAY. This "raising" process might be termed "presupposition raising". (Another example is I could punch your nose!, which presupposes "...but I'm not going to punch your nose".) As with all types of syntactic raising, the process is metonymic in nature, and can be seen as a form of hypallage. (For a discussion of presupposition raising and syntactic hypallage, see Larreya and Mery 1992). I will also argue that the past tense has two main varieties of use, to which I will refer under the headings temporal past tense (pleonastic though this term may sound) and hypothetical past tense. The latter is generally regarded as a kind of metaphoric extension of the former, which is considered the "real" past tense. My position is that in Present-day English there is no telling which of the two uses is the more fundamental, for neither can be considered marginal or even minor, and they are semantically akin to one another. One may think that in Old English the erosion of the morphological differences between the indicative and the subjunctive past tenses was not the only reason why they merged and gave the modern past tense - whose "hypothetical" meaning undoubtedly derives from that of the subjunctive. They probably would not have merged if there had not been something common to their respective meanings. Interestingly, Benveniste (1951) has shown that in several Indo-European languages the imperfective past tense forms seem to be derived from an optative. This suggests that the "modal" component of the meaning of these forms might be as fundamental as the "temporal" component.

Irrealis, past time reference and modality 3.1.


The hypothetical past tense

As with all presuppositional forms, the meaning has to be considered at two levels: assertion and presupposition. (i) (ii)

In the assertion of the utterance, the event is presented as theoretical (i.e. imaginary, hypothetical, etc.). In the presupposition of the utterance, the event is presented as {absolutely or relatively) contrary to reality - or, as we shall see, as contrary to some sort of reality.

There are two types of hypothetical past tense, which may be termed counter/actual past tense and tentative past tense. (The word "tentative" is used here in its most broadly accepted sense, which includes what is sometimes called "potential condition", as in, for instance, if it rained tomorrow...).

3.1. J.

The counterfactual past tense

In example (5a), which is a case of counterfactuality, the non-reality of the event is absolute: the truth-value which is considered in the assertion is exactly the opposite of the truth-value attributed to the event. In (5a), two things are expressed - at two different levels: (i)

In the assertion, the event "I know the answer" is considered from a theoretical point of view - it is the object of a wish. (ii) In the presupposition, the same event is presented as "contrary to fact" - the implicit meaning being ".. .but I don't know the answer". And, as we have seen, (5a) is a case of both absolute and direct non-reality, while in (8) the unreality is absolute and indirect.


The tentative past tense

In the case of tentativeness, the contradiction between the theoretical event and the presupposed event is not absolute - and it cannot be absolute, simply because the truth-value of the "real" event is not known. What is presupposed is that the truth-value of the event is probably (or, in some cases, possibly) different from the truth-value envisaged in the assertion so that the unreality is only relative. And, here again, the unreality may be


Paul Larreya

direct or indirect. As we have seen, example (6a) - repeated below as (9) is a case of direct relative unreality; it presupposes something like / probably will not win the lottery·. (9)

If I won the lottery...

Examples (10) and (11) below, on the other hand, are cases of indirect unreality (cases of "raising" of the presupposition, which is marked on the uppermost auxiliary or verb of the syntactic structure): (10) (11)

You could take an aspirin, (addressed to somebody who has just said that they have a headache) I wanted to ask you a favour.

In (10), the unreality (the relative unreality, to be precise) does not concern the modal which carries the -ED morpheme. This modal (could) expresses a suggestion. It is also possible to express suggestion using can, but could is often preferred because it sounds more polite. The "politeness" effect is due to the implicit doubt that -ED adds concerning the reality of the suggested action: using the -ED form (could instead of can), the speaker appears to be leaving the hearer entirely free to make their own decision. Example (11) is a case of what Quirk et al. (1985: 188) call the "attitudinal past". (Bybee [1995: 508] refers to such uses of the past tense under the appellation "[t]he so-called polite or remote uses", and very aptly remarks that they "[depend] upon the presence of a modal verb". One of her examples is I wanted to ask you a question.) In this use of -ED, it is not possible to entirely rule out the existence of a temporal component (part of the meaning might be "When I decided to come and see you, I wanted to..."), but the hypothetical meaning is undoubtedly predominant. And, once again, the non-reality does not concern the verb that carries the -ED morpheme (the verb want), or at least it only concerns it marginally. The implicit doubt that is expressed is, in fact, somewhat vague, and open to several interpretations, as it concerns (or may concern) several parts of the informational content of the subordinate clause: firstly it may concern the "asking" itself (the speaker is in some way requesting the hearer's permission to ask a favour, which in turn casts a doubt on his really wanting to ask it), and secondly it may concern the existence of the favour, or more precisely the willingness and ability of the hearer to grant it. This brief description shows that the two basic meanings of the hypothetical past tense (counterfactuality and tentativeness) have this in common: both are (a) negatively oriented and (b) presupposed.

Irrealis, past time reference and modality


3.2. The temporal past tense There are two types of temporal use of -ED, which I will call the narrative past tense and the backshifled past tense. One of the reasons why it is necessary to distinguish them is that they have different properties in the grammar of modals: the restrictions on the temporal use of could, might, should and would only apply to the former.


The narrative past tense

First of all, one thing should be pointed out concerning the "unreality" inherent in temporal uses of -ED. It is not the event itself that is unreal (as is the case with the hypothetical past tense), since it is presented as "real in the past". The unreality- whether absolute or relative- concerns the present moment, and, here again, it is situated in the presupposition of the utterance. The result is that, once again, the meaning is situated at two contrasting levels. In the assertion of the utterance, the event is presented as real (i.e. as having really existed in the past). In the presupposition of the utterance, the event is presented as unreal at the moment of speaking - or, more precisely, as having something unreal about it. This does not necessarily mean that the event has no physical existence at the present moment. Here again the implicit non-reality expressed by the -ED morpheme may be either direct or indirect: the use of -ED signals the fact that there is, in the event itself or in its context, something unreal - and we shall see that in some cases there is ambiguity as to precisely what is unreal. When the unreality concerns the event itself (in other words when the event itself no longer exists at the moment of speaking), it may be called direct. Consider the following: (12)

"Mrs Hawkins was an editor at Ullswater Press", said Ian Tooley (...). "And now we are so fortunate as to have her with us ". (Spark 1988:92)

The first clause {Mrs Hawkins was an editor at Ullswater Press) strongly suggests that Mrs Hawkins is no longer an editor at Ullswater Presswhich is confirmed by the last sentence. So the unreality expressed in (12) is direct. The question that arises at this point concerning the suggestion of "present unreality" conveyed by the past tense form of (12) is the following: is it not an implicative rather than a presupposition? For


Paul Larreya

Comrie (1985: 41-42), who discusses such examples as John was in Paris and John used to live in Paris, this type of suggestion is an implicature. He remarks that "[a]ny implicature of present non-relevance can easily be contextually cancelled". This, however, is not sufficient to prove that the "present non-relevance" expressed or implied by the past tense is not a presupposition: as Lakoff (1972: 572-573) has shown, some presuppositions can be contextually cancelled - just like implicatures. As we shall see, the possibility of cancelling certain negatively-oriented presuppositions expressed by -ED- and replacing them with different negatively-oriented presuppositions or positive assertions - arises from the fact that -ED simply signals the existence of an element of unreality. The addressee normally expects to find the precise reference of this element of unreality in the verb itself or in the clause to which it belongs, but, as we have seen, it may be elsewhere. I have suggested in Section 2 that the main difference between implicature and presupposition is that the former is generally unspoken while the latter is necessarily spoken - i.e. expressed by specific forms in the language. Thus, if somebody says I'd like some more tea, they cannot deny saying - albeit implicitly - that they have already had some tea, and the form that expresses this presupposition is more. In the same way, if somebody says John was a lawyer, they cannot deny implicitly appending an element of non-reality to the proposition John-be a lawyer·, the only peculiarity of this presupposition (one that makes it resemble an implicature) is its indeterminacy as to where exactly the non-reality lies. One of the main (and somewhat indirect) reasons why an event might be seen as non-real is the fact that its main participant(s) is/are no longer alive. (In this respect, it is hardly possible not to mention the famous controversy about the anomaly of Einstein has visited Princeton, as opposed to the generally accepted Princeton has been visited by Einstein — see e.g. Comrie 1976: 59-60.) In some cases, this gives rise to some hesitation on the part of the speaker: (13)

"She's dead?" "Yeah. Pretty brutal. You got a name for that boyfriend? (...) " "I can get it. My lady friend is - was - Ms. Dean's best friend". (James 1989: 83)

Like all presuppositions (which, in normal circumstances, are not meant to inform, and only have a functional role), this one can result in an unintentional disclosure, as in the following excerpt from a news item:7

Irrealis, past time reference and modality (14)


Zile played the role of grieving mother extremely well. (...) Police were less easily convinced (...). "A couple hours after your little girl is missing, you don't start saying she 'was' a nice girl", says Jim Leljedal of the Broward County Sheriff's office. (Newsweek; Nov. 14, 1994: 40)

The reason for the use of -ED, however, may be much more indirect. One may use -ED not because the event itself has no reality at the present moment but because the speaker sees it as part of a past situation: (15)

(16) (17)

At dinner one night (...) I met a nice boy at the Moores'. (...) Gilbert Finch I think he's called. He had a nice voice. I heard him talking to the children. (Minot [1993] 1994: 104) Newmarket lay fifty miles to the south of Norwich, and I drove there through the sunny afternoon (...). (Francis [1981] 1982: 63) "(...) He's what Hobsbawm called a primitive revolutionary. Hobsbawm was a Brit, right? " "He still is". (Wolfe [1987] 1988: 238)

The type of correction exemplified in (17) is sometimes made by the speaker himself when he realizes that his use of the past tense might be misinterpreted, and that anyway there is also reason to use the present tense - a s in (18) and (19) below. (18)


My name is Kay Fisher and at the time of this story I was thirty-two years of age. (...) I was five feet six inches tall (I still am) with dull brown hair (...) (Boyd 1993: 10) I wrote a novel called Devil of a State. (...) The novel was, is, about Brunei (...). (Burgess 1987: 431)

In (19), the present unreality implicit in was concerns the writing of the book; it is the same present unreality as that expressed by the verb wrote in the previous sentence. As to the "reality" expressed by is, it obviously concerns the book itself. But of course the hesitation or mistake may require a shift in the other direction, as shown by example (13).


The backshiftedpast tense

One of the examples of "backshifting of tenses" given by Quirk et al. (1985: 1027) is the following - numbered here as (20):

30 (20)

Paul Larreya "I am being paid by the hour", she said. ~ She said she was being paid by the hour.

As is well known, the use of the past tense is compulsory in the subordinate clause when the "reported" event is no longer in existence at the time of speech - which hardly needs an explanation, since there is reason to believe that the use of -ED was already necessary in the "original" utterance which can be assumed to correspond to the indirect reported speech. What does need to be explained is the use of the past tense in cases where the reported event is situated in the present.8 The reason for this use, however, is obvious, if one thinks of the similarity between the cases illustrated by (20) and by (15)—(17) respectively. In the latter examples, the use of -ED is motivated by a kind of metonymic process - by the fact that the event is seen as part of a situation which is no longer real. Similarly, the event which is reported in (20) is associated with a speech-situation which belongs to the past and therefore requires, for itself at least, the use of -ED. And of course there is also a more mechanical association at the level of form. What is typical of the "optional" use of the past tense is that it can generate the same type of hesitation or ambiguity as that exemplified in (18)—(19). An example is: (21)

"Didn't you say that Julian's mother lived somewhere near there? " "Yes, she did", said Deborah. "I mean she does. Not far from there (...)". (Gale [1990] 1992: 146-147)

3.3. The past tense, perfective aspect and irrealis The past tense is not, in English, the only verb form that raises the problem of the relationship between past time reference and irrealis. The perfect (at least in its infinitive form) also poses that problem. It is often closely linked to the expression of irrealis, and I will argue that it sometimes expresses irrealis - even though the role of HAVE + -EN as a marker of irrealis is very different from that of -ED. In the grammar of modality, it is not rare for "hypothetical" -ED to be combined with the perfect (i.e. with HAVE + -EN). An example is: (22)

I might have seen it (but I didn't).

In (22), and in most cases when the past tense form of a modal is combined with HAVE + -EN, the distribution of the functions seems to be quite clear:

Irrealis, past time reference and modality


HAVE + -EN fulfils a temporal function (it situates the event in the past), while -ED expresses counterfactuality or tentativeness. In (22), the role of -ED appears if one compares I might have seen it (but I didn't) with I may have seen it, and a comparison between I may/might have seen it and / may/might see it shows the role of HAVE + -EN. In some cases, however, these roles are reversed. Consider:

(23) (24)

John was to go on a trip to Italy. John was to have gone on a trip to Italy.

In (23), the -ED marker contained in was has a temporal value (it situates the intention in the past). The utterance is neither factual nor counterfactual concerning the event "John went on a trip to Italy" - it gives no information as to whether John went or not. In order to make (23) counterfactual (in order to give or rather to presuppose the information that John did not go), one has to add the marker of the perfect to it; the result is (24), which adds the morpheme HAVE + -EN to (23). So, the roles of -ED and HAVE + -EN in (23) and (24) can schematically be described as follows (schematically because the semantic construction created by the association of the two morphemes is obviously more complex): -ED fulfils the temporal function (it situates John's intention in the past), while HAVE + -EN is used as a marker of irrealis. There must be semantic reasons (which are beyond the scope of the present study) for HAVE + -EN to be able to mark irrealis in an utterance like (41), but there is also a functional reason: the -ED marker is already fulfilling one of its two potential functions (the expression of past time reference), and it cannot at the same time, without a risk of ambiguity ensuing, fulfil the other function (the expression of irrealis), so that another marker (a marker of somewhat similar meaning - namely HAVE + -EN) has to be used as a substitute. There is a third category of combined "hypothetical" use of -ED and in which it is more difficult to say which of the two morphemes expresses past time reference and which expresses irrealis. Compare: HAVE + -EN,

(25) (26) (27)

He had the key. He could open the door. He has the key. He could open the door. He had the key. He could have opened the door (but he didn't).

In (25), the -ED morpheme contained in could has a temporal meaning (as would the -ED morpheme of was in, for instance, it was possible for him to open the door). In (26), it has a hypothetical (tentative) meaning, which


Paul Larreya

appears clearly if could is replaced by can. In (27), the modal verb phrase also has a hypothetical meaning (which the addition of but he didn 't makes clearly counterfactual), but, depending on whether one compares (27) to (25) or to (26), the -ED morpheme borne by the modal could be said to have a temporal or a hypothetical meaning. And, in the same way, the function of HAVE + -EN could be judged "temporal" if one compares (27) to (22), and "hypothetical" if one compares it to (24). Examples (23)-(27) show that, at least in some cases, HAVE + -EN plays a part in the expression of irrealis and does so in association with -ED. Three further remarks can be made concerning the role of HAVE + -EN as a marker of hypothetical meaning in the English verbal system. First, HAVE + -EN cannot, in itself, express hypothetical meaning: it can only do so in association with -ED. (There is at least one notable exception, however: when followed by a perfect infinitive, as in John needn't have run, the modal need used deontically is counterfactual in that it presupposes the truth of the proposition whose "non-necessity" is asserted by the negated modal.) Second, while -ED can express either counterfactuality or tentativeness, HAVE + -EN necessarily expresses counterfactuality when it has a "hypothetical" function. Third, it is only in its infinitive form that the English perfect can be regarded as a full marker of irrealis. (There is probably a relationship between this and the fact that, as shown by Hofmann in his discussion [1976] of such forms as he is rumored to have come last Tuesday, the perfect infinitive can be a syntactic substitute for the past tense.) Most likely, it is owing to this affinity with counterfactuality that, in some cases, HAVE + -EN makes it possible to differentiate counterfactuality from tentativeness. This happens in cases where the differentiation between the two hypothetical meanings of -ED is functionally useful and where the context is not sufficient to make it so. (In such languages as English or French the main "hypothetical" form neutralizes the opposition, but some languages commonly use forms that differentiate between the two meanings — cf. Latin si venias / si venires, 'if you came'.) Thus, in if he had lived he would have taken over the leadership when Tom retires next year (from Huddleston 1977: 46), the verb phrases had lived and would have taken over are counterfactual, whereas, in the same context, lived and would take over would only be tentative (they would only express a negatively-oriented presupposition of doubt). Fleischman (1995) makes a claim which is very different from the one that I have made in this section, and which might even seem contradictory to it: she contends (1995: 519) that "[i]n many languages of the world we encounter a more than chance connection between the aspectual category

Irrealis, past time reference and modality


imperfective [my italics] and irrealis modality". The first objection one might oppose to her arguments is that she does not make sufficient allowance for the fact that, as a general rule, imperfective aspect cannot express irrealis unless it is associated with another grammatical form which expresses it more intrinsically. In practically all of her examples (with the possible exception of the Italian conditional, which requires an explanation of its own), imperfective aspect is combined with a past tense form: the verb forms which are exemplified as markers of irrealis are the past progressive (in English), the past imperfect (in French, Spanish and Italian), the conditional (in French and in Spanish, where it includes the morpheme of the imperfect), etc. - so that one may wonder why she attributes the expression of irrealis to the markers of imperfective aspect rather than· to the past tense markers. As regards the English imperfective form (namely the "progressive"), her main argument is the fact that the "politeness forms" I was hoping... / I was thinking about... / I was wondering... (exemplified on page 527) cannot be replaced by their simple-past counterparts. The force of this argument is somewhat weakened when she aptly remarks (p. 528) that in the same contexts the simple-past forms would imply "I am no longer hoping that/thinking about..."; this clearly points to the link between the "temporal" and the "hypothetical" meanings of -ED; it is tantamount to saying that, in these contexts, I hoped.../1 thought... would in some way be counterfactual, and that consequently the opposition between the two types of past (simple and progressive) more or less corresponds in this case to the opposition between the two types of irrealis (counterfactuality and tentativeness) commonly expressed by -ED. Another argument against Fleischman's hypothesis is that it does not explain the fact that in several languages (e.g. in English) the main marker of irrealis is a verb tense which is not imperfective. (In English, the simple past, which is the principal means of expression of irrealis, is obviously not imperfective in, for instance, I lived in London for ten years.) In fact, there are two elements in the semantic make-up of the forms discussed by Fleischman. Let us take, for instance, the case of the French imperfect past tense (the "imparfait"). Its fundamental meaning consists of two elements: imperfect aspect and a negatively-oriented presupposition. For various reasons, linguistic analyses tend to favour the first component, but the second is as least as important. It is a common feature of the English past tense and of the French "imparfait" - the main semantic difference between the two forms being that the former is non-progressive while the latter expresses a type of aspect which is akin to progressiveness. (For a more detailed analysis of the French imperfect and its place in the French tense system, see Larreya 1996.)

34 4.

Paul Larreya Modais and the hypothetical past tense

In most cases, the absolute or relative unreality implicitly expressed by the past tense form of a modal seems fairly obvious, as we saw in Section 2 when examining such examples as You might have broken you leg or You could take an aspirin. There are, however, a number of cases in which the presence of a presupposition of unreality is much less clear. I will briefly discuss two such cases: epistemic might and what might be termed "factual" should.

4.1. Epistemic might The contexts in which the problem of the meaning of -ED arises are those in which may and might seem to be roughly equivalent. This excludes cases in which the use of the past tense form is compulsory because there is some suggestion of "unreal" conditionality - be it ever so minute or indirect, as in (28) and (29): (28)


"Would you mind if I talked to her? " "Not at all. In fact, I'll give you a list of some of the people here who might be helpful". (Sheldon [2000] 2001: 277) It's like, censoring porno because one moron might think that's the way it goes in real life. (Branton 2000: 24)

The problem of the semantic difference between may and might is often posed in terms of degrees of certainty (of the speaker) or likelihood (of the event). The most common view is that might expresses a weaker degree of likelihood than may. Palmer (1990: 58), for instance, writes: "Might is used exactly as may is. It merely indicates a little less certainty about the possibility". For Leech (1987: 127), "[t]he effect of the hypothetical auxiliary [i.e. might], with its implication 'contrary to expectation', is to make the expression of possibility more tentative or guarded. Our team might still win the game could be paraphrased 'It is barely possible t h a t . . . ' or 'It is possible, though unlikely, that ...'.". He qualifies this judgement, however, as regards the use of might + perfect infinitive to express the possibility of an event situated in the past, about which he writes (1987: 127): "Might (...) seems to be used almost as a colloquial variant of may (= "factual possibility") without any implication of reduced likelihood". Coates goes still further in denying that there is any expression of reduced likelihood inherent in the past tense form. She writes (1983: 152): "MIGHT,

Irrealis, past time reference and modality


in my data at least, does not seem to express a more tentative meaning than MAY."

There is no doubt that in some contexts - e.g. in (30) and (31) below might does express a weaker degree of likelihood than may would: (30)


The beauty of a combinatorial system is that it generates combinations that have never before been considered but that one might want to talk about some day. (Pinker [1999] 2000: 13) "Mind you, (...) there is just that tiny chance that it might be the one, don't you think?" (Dahl [1964] 1979: 40)

And, by the same token, the replacement of may by might would, in many cases, weaken the degree of likelihood expressed by the modal: (32)

I'd suggest that what works for me may equally well work for you. (King [2000]2001: 153)

In some cases, however, contextual implicatures, or the prosody, make it difficult to judge whether might expresses a lesser or a greater degree of likelihood than may. This seems to be the case in (33) and (34):9 (33) (34)

The doctor said he can have one small drink a day, that it might even be goodfor him. (French 1994: 305) " Well, George!" he said, (...) "what do you think of it all? " "Well", I said; "in the first place - it's a damned swindle!" "Tut! Tut!" said my uncle. (...) "It's the sort of thing everybody does. After all, there's no harm in the stuff [viz. Tono-Bungay, a medicine advertised as "'the secret of vigour"] - and it may do good. It might do a lot of good- giving people confidence, f'rinstance, against an epidemic. (...)" (Wells [1909] 1961: 125)

In fact, problematic as it is, the assessment of degrees of likelihood may not provide the best key to the analysis of the may-might opposition in all cases. (It may not be relevant, either, to account for the semantic difference between epistemic might and could. On this difference, see Gresset [1999: 259-461] and this volume.) What I will suggest is that, depending on the context, and of course depending on what the speaker intends to say, the -ED marker contained in might may express two types of negatively-oriented presupposition.


Paul Larreya (i)

The presupposition may be based on the speaker's own evaluation of the degree of likelihood of the event. In this case, the meaning of It might be the case that Ρ is something like "It is possible - though not very likely - that P". (ii) It may be based on the hearer'% assumed opinion of the degree of likelihood of the event. The meaning of It might be the case that Ρ is then something like "It is possible - although you seem to believe the contrary - that P". In the second case, the "hearer-oriented" presupposition may be made explicit in the context: (35)

"(...) What have you done about your manuscripts? They could be worth something". "Unlikely, very". "They might, you never know (...) ". (Craig [1996] 1997: 317)

A further remark can be made about the implicit meaning conveyed by epistemic might through its -ED marker. This implicit meaning (added to the modal meaning proper) is essentially epistemic in nature: roughly speaking, it can be paraphrased as "...contrary to what is probable/what you seem to believe"). One may think, however, that -ED sometimes conveys (secondarily) an implicit deontic meaning which may be paraphrased as "...contrary to what you hope for/what you fear". As we shall see, a similar observation can be made about some uses of should.


"Factual "should

In most of its uses, should is to be found in contexts which are either counterfactual (as in You should be in your office at this time of day, which presupposes "...but you are not in your office") or "tentative" (as in You should give up smoking, which contains a presupposition approximately paraphrasable as "...but I'm not sure you will give up smoking"). In some cases, however, should is used in contexts which - at least apparently contain no negative implication. These contexts, which may be called factual, seem to contradict the hypothesis that -ED always expresses a presupposition of unreality. (Most "factual" uses of should concern what is often called "putative" should-see, for instance, Quirk et al. [1985: 234235]. The coincidence of the two categories, however, is only partial.) Rhetorical questions introduced by Why should... are generally counterfactual, as in:

Irrealis, past time reference and modality



"Aren 'tyou upset? ", he had asked (...). "Why should I be? (...)". (Bernard 1994: 169)

In some cases, however, they are (at least apparently) factual: (37)

"I cannot think", he said at last, "why my brother, who claims to be an artist, cannot see the immense beauty of the railways. Why should he dislike them so intensely? He is too young to be so old-fashioned". (Barnes [1995] 1996: 39)

The meaning of the rhetorical question beginning with Why should.. As in fact very different in (36) and in (37). In (36), we have an example of stylistic use of a presuppositional form: the main purpose of the utterance is to inform the hearer that "I am not upset" - although this information is only presupposed. In (37), on the other hand, the segment why should he dislike them... ? is, so to speak, under the scope of I cannot think (why)..., so that the meaning is something like it's surprising that he should dislike them so intensely. This brings us to the use of "putative" should in clauses introduced by such phrases as it's surprising/strange/odd that...One example is: (38)

I know it's a little strange, and a little bit of a contradiction, that a far-seeing place should also be a basement place, but that's how it is with me. (King [2000] 2001: 89-90)

In this case, the -ED morpheme borne by the modal expresses an implicit contradiction - which, in (38), happens to be made explicit in an incidental clause. The contradiction is between two types of logic, as it were. There is, on the one hand, the logic of what takes place in reality (which can also be seen as the logic, or will, of destiny); this is expressed by the morpheme SHALL, in the assertion of the utterance. And there is, on the other hand expressed by -ED in the presupposition of the utterance - the logic of what the addressee would expect.10 Example (39) illustrates a slightly different case: (39)

If you happen to be a science fiction fan, it's natural that you should want to write science fiction (...). (King [2000] 2001: 144)

As in (38), two "logics" come into play, but here they are not at variance with each other: the logic expressed in it's natural that...goes along with the presupposed reality. There is, all the same, an implicit contradiction


Paul Larreya

which justifies the use of the -ED marker: the reader is in some way assumed to consider the desire to write science fiction to be rather unnatural. Behre (1955: 150) clearly describes the role of the addressee in this type of implicit contradiction; he attributes the use of should after such expressions as it is not surprising that..., it is natural that... etc. to the fact that there is something which is "thought of as undesirable or unexpected to some person or persons other than the speaker (or the subject)". [The italics are mine.] In many cases, the implicit contradiction contained in the phrase it is normal that...should... is made somewhat explicit by the presence of the adverb only (It's only normal that...should...). (On the role of only and a few other adverbs in this type of context, see Arigne 1989: 208.) Three other types of "putative" should, exemplified in (40)-(42) below, can be accounted for in a similar way. Example (42), unlike the other two, is not factual, but in its semantics are approximately the same. In the three cases, what should expresses is a contradiction between, on the one hand, what happens (or may happen) and, on the other hand, what might be expected. The implicit meaning is something like "contrary to what one would expect/would have expected", and, here again, the contradiction is between two "logics" - one of which is expressed by the morpheme SHALL. In some cases, however, a deontic overtone - perceptible in (40) and (41) is added to the epistemic meaning. It can be paraphrased as "contrary to what one would hope for/have hoped for". The implicit contradiction is between two forms of "will": there is on the one hand the will of destiny, as it were, which is expressed by the modal SHALL, and on the other hand the will of the person whose words or thoughts are reported. (It is significant that one of the likely French equivalents of the modal form in (40) and (41) would be Et il a fallu que..., with the modal verb falloir expressing a deontic meaning similar to that of SHALL.) The semantic overlapping of two types of modality (epistemic and deontic) is not surprising in the case of should, which never expresses "pure" logical necessity: as is well known, it is possible to say He should be better after taking this medicine, but not (at least in a "normal" context) #He should be worse after taking this medicine. (40) (41)

That he should think she was poisoning him - oh, he was mad, he was mad (...). (McGrath 1996: 132) And who should I see in the lobby, sitting in a chair and reading a newspaper, but Mr. Earl Higgins, alias Old Cue-Ball. (King [2000] 2001:76)

Irrealis, past time reference and modality



The director had sent word that he had decided to use the day, if it should indeed be fine, to shoot one of the most important scenes in the film. (Mortimer [1957] 1988: 54)

The last type of "factual" should I will discuss is one in which the meaning is a mixture of root and epistemic necessity: (43)

I reach for the remote control, which is exactly where it should be, on the bedside table. (Tulloch [1989] 1990: 67)

This type of meaning of should (i.e. moral or logical necessity) is generally to be found in contexts that make the modal proposition either counterfactual (as in the remote control is not where it should be) or tentative (as in in future you should put the remote control on the bedside table). In (43), however, the context is (or seems) factual·, the remote control is on the table. This case is quite similar to a particular type of use of the past tense after the phrase it's about time mentioned and analyzed by Leech (1987: 123): Yet a further example of weakening is seen in a colloquial tendency to use the it's time construction in circumstances to which the implication of negative truth-commitment is quite inappropriate. One might hear, for instance, the following snatch of dialogue: A: Tiny's cooking the breakfast this morning. B: Oh good — it's about time he helped out with the cooking. It is quite evident that here the hypothetical verb helped refers to what Tiny is doing, rather than to what he is not doing. If the hypothetical form is to be given any negative force here, it must be applied to the past (Tiny's previous failure to help) rather than to the present.

There is in (43), as in Leech's example, an element of unreality which explains the use of the past tense morpheme. The speaker somewhat distances herself from the reality she is commenting upon (the presence of the remote on the bedside table), and she mentally places herself before that reality. In other words, she places herself at the moment when the person who is habitually in charge of the remote has to make the decision to put it on the bedside table or elsewhere. And, as at that moment no decision has yet been made, the use of should is still possible; so, although the meaning of -ED is here essentially "hypothetical", the "temporal" value is not entirely absent. Concerning- from a more general point of view- the use of the hypothetical past tense in the grammar of modals, a further observation can


Paul Larreya

be made: although there are no restrictions on the use of the past tense form of modals to express indirect unreal presupposition (as in You might have told me!, which presupposes "You didn't tell me"), only could and would can express direct unreal presupposition, as in (2a) - repeated below as ( 4 4 ) - and in (45), which (respectively) presuppose "I cannot telephone them" and "you probably aren't willing to listen to me": (44) (45)

If only I could telephone them! If you'd listen to me, it would make things a lot easier.

And it is probably no coincidence that could and would are also (with the exception of a few rare uses of might) the only past tense forms of modals that can be used to express a narrative past.



The main claim made in this paper is the following: the element of meaning one adds to the semantic make-up of a finite syntactic clause by appending the "past tense" morpheme to the first element of its verb phrase is entirely situated at the level of the implicit - or, more precisely, it belongs to the presupposition of the utterance. My aim has also been to show that that morpheme - whether it is borne by a lexical verb or by a modal - always expresses some type of (implicit) unreality, which can be absolute or relative, direct or indirect.

Notes 1. 2.


For greater clarity, the forms that are the object of specific commentaries are in bold type in the examples. This is farfrombeing universally accepted: might, would and should are often considered not to be the same verbs as may, will and shall. Thus, after conceding that could is the past tense form of can, Pinker ([1999] 2000: 89) writes: "Other pairs of modal auxiliaries - may-might, will-would and shall-should - began life as different tenses of the same verb, but the couples divorced long ago and might, would and should are no longer past tense forms". However, as shown by Horn (1984), some implicatures can be related to specific linguistic forms. For detailed discussions of the relation between implicature and linguistic form, see Horn (1984 and 1988); Levinson (1983: 97-166); Sperber and Wilson (1995).

Irrealis, past time reference and modality 4.


To describe this type of use of the past tense, Leech (1987: 118, 123) uses the term "negative truth-commitment" and refers to truth-values that are "contrary to assumption" or "contrary to expectation". 5. "Event" is used here in a very broad sense - not as the opposite of "state". Throughout this paper, it will denote the referent of a proposition (subject + predicate), which may be an action, a mental process, a state, etc. 6. Leech (1987: 81, 118 and passim) gives a different meaning to the adjective "theoretical". The word "factual", too, is used in this paper in a different sense from that given to it by Leech (1987: 118). 7. Concerning this type of "accidental" use of presuppositional forms, see Larreya and Watbled (1994: 73). It is a well-known fact that speakers tend to pay less attention to the presupposition than to the assertion of their utterances. 8. It is often said that in this case the use of the past tense is optional, but things are in fact more complex. On this question, see, for instance, Larreya and Rivi6re (1999: 36). 9. I have submitted (33) and (34) to 10 informants, asking them to compare the degrees of likelihood expressed by might and (potentially) by may in the same contexts. For 6 of them, might expresses a lesser degree of likelihood; 3 think that it expresses a greater degree of likelihood (and one of them adds that in It might do a lot of good he would "stress might as much as a lot')·, 1 says that might and may express approximately the same degree of likelihood. What made me think that in (33) might may express a fairly high degree of likelihood is the fact that, two pages further in the novel, the same character thus re-formulates the doctor's advice: "He says a drink will probably be good for him [...]". But of course this does not mean that might itself expresses greater certainty than may. 10. Commenting on the fact that Chafe (1995: 357) finds it somewhat surprising that in Caddo a marker of irrealis should be used in the equivalent of my goodness he knows my name!, Palmer (2001: 177) enlarges on Chafe's own explanation and very aptly suggests that the meaning of the utterance is something like "I'm amazed that he knows my name". He could obviously have added the mention of the form I'm amazed that he should know my name. The explanation I am proposing in this section for "factual" uses of should owes much to Behre's (1955) exhaustive and enlightening study. Among other things, Behre has shown the importance of two factors in the grammar of "meditative-polemic should': firstly the presence of an element of contradiction and secondly the role played by what the hearer (or reader) is assumed to think.


Paul Larreya

References Arigne, Viviane 1989 Shall et should, 6tude de modalitös. In Explorations en Linguistique Anglaise, Andri Gauthier (ed.), 153-227. Berne: Peter Lang. Barnes, Julian 1996 Reprint. Cross Channel. London: Picador. Original edition, London: Picador, 1995. Behre, Frank 1955 Meditative-Polemic Should in Modern English That-Clauses. Stockholm: Almqvist and Wiksell. Benveniste, Emile 1951 Pritirit et optatif en indo-europien. Bulletin de la Societe de Linguistique de Paris 47: 11-20. Bernard, Mary 1994 Friends and Lovers. London: Mandarin Paperbacks. Boyd, William 1993 The Blue Afternoon. London: Sinclair-Stevenson. Branton, Matthew 2000 Coast. London: Bloomsbury. Burgess, Anthony 1987 Little Wilson and Big God. London: Penguin Books. Bybee, Joan L. 1995 The semantic development of past tense modals in English. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 503-517 (Typological Studies in Language 32.) Amsterdam/Philadelphia: Johns Benjamins. Chafe, Wallace 1995 The realis-irrealis distinction. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 349-365 (Typological Studies in Language 32.) Amsterdam/Philadelphia: Johns Benjamins. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Cole, Peter, and Jerry Morgan (eds.) 1975 Syntax and Semantics 3: Speech Acts, 41-58. New York: Academic Press. Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. 1985 Tense. Cambridge: Cambridge University Press. Craig, Amanda 1997 Reprint. A Vicious Circle. London: Fourth Estate. Original edition, London: Fourth Estate, 1996. Dahl, Osten 1997 The relation between past time reference and counterfactuality: A

Irrealis, past time reference and modality


new look. In On Conditionals Again, Angeliki Athanasiadou and Rend Dirven (eds.), 97-114. (Amsterdam Studies in the Theory and History of Linguistic Science 143.) Amsterdam/Philadelphia: Johns Benjamins. Dahl, Roald 1979 Reprint. Charlie and the Chocolate Factory. New York: Bantam Books. Original edition, New York: Alfred A. Knopf, 1964. Fleischman, Suzanne 1995 Imperfective and irrealis. In Modality in Grammar and Discourse, Joan Bybee and Suzanne Fleischman (eds.), 519-551 (Typological Studies in Language 32.) Amsterdam/Philadelphia: Johns Benjamins. Francis, Dick 1982 Reprint. Twice Shy. London: Pan Books. Original edition, London: Michel Joseph, 1981. French, Mary 1994 Our Father. London: Penguin Books. Gale, Patrick 1992 Reprint. The Cat Sanctuary. London: Flamingo. Original edition, London: Chatto & Windus, 1990. Gresset, Stephane 1999 Can/may, may/might et might/could, ou l'interchangeabilitö en question. Doctoral dissertation, Universiti Paris X-Nanterre. Grice, H.P. 1975 Logic and conversation. In Syntax and Semantics 3: Speech Acts, Peter Cole and Jerry Morgan (eds.), 41-58. New York: Academic Press. Hofmann, T.R. 1976 Past tense replacement and the modal system. In Syntax and Semantics 7: Notes from the Linguistic Underground, James D. McCawley (ed.), 85-100. New York: Academic Press. Horn, Laurence R. 1984 Toward a new taxonomy for pragmatic inference: Q-based and R-based implicature. In Meaning, Form, and Use in Context: Linguistic Applications, Deborah Schiffrin (ed.), 12-42. Washington D.C.: Georgetown University Press. Horn, Laurence R. 1988 Pragmatic theory. In Linguistics: The Cambridge Survey, Frederick J. Newmeyer (ed.), 4 vols, 113-145. Cambridge: Cambridge University Press. Huddleston, Rodney D. 1977 Past tense transportation. Journal of Linguistics 13: 43-52. James, P. D. 1989 Devices and Desires. London, Boston: Faber & Faber. King, Stephen 2001 Reprint. On Writing. New York: Pocket Books. Original edition,


Paul Larreya

New York: Scribnet, 2000. Lakoff, George 1972 Linguistics and natural logic. In Semantics of Natural Language, Donald Davidson and Gilbert Harman (eds.), 545-665. Dordrecht: D. Reidel. Larreya, Paul 1996 Le temps grammatical: une question de mode. In Dynamique du Temps, Alain Suberchicot (ed.), 139-153. Clermont-Ferrand: CRLMC (University Blaise Pascal). Larreya, Paul, and Renaud Miry 1992 De la productivity syntaxique de l'hypallage. In L'ordre des Mots Domaine Anglais, Jacqueline Gueron (ed.), 143-160. Saint-Etienne: CIEREC. Larreya, Paul, and Claude Rivifcre 1999 Reprint. Grammaire Explicative de I'Anglais. London: Longman. Original edition, Paris: Longman France, 1991. Larreya, Paul, and Jean-Philippe Watbled 1994 Linguistique Ginirale et Langue Anglaise. Paris: Nathan. Leech, Geoffrey N. 1987 Reprint. Meaning and the English Verb. London: Longman. Original edition, Harlow: Longman, 1971. Levinson, Stephen C. 1983 Pragmatics. Cambridge: Cambridge University Press. McGrath, Patrick 1996 Asylum. London: Penguin Books. Minot, Susan 1994 Reprint. Folly. 2d ed. London: Minerva. Original edition, New York, Washington Square Press, 1993. Mortimer, John 1988 Reprint. Charade. London: Penguin Books. Original edition, London, Viking Press, 1957. Palmer, Frank R. 1990 Modality and the English Modals. 2d ed. London/New York: Longman. Original edition, New York: Longman, 1979. 2001 Mood and Modality. 2ded. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Pinker, Steven 2000 Reprint. Words and Rules: The Ingredients of Language. London: Phoenix. Original edition, New York: Morrow, 1999. Quirk, Randolph, Sidney Greenbaum, Geoffrey N. Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Sheldon, Sidney 2001 Reprint. The Sky is Falling. London: Harper Collins. Original edition, London: Harper Collins, 2000.

Irrealis, past time reference and modality


Spark, Muriel 1988 A Far Cry from Kensington. London: Constable & Company. Sperber, Dan, and Deirde Wilson. 1995 Relevance: Communication and Cognition. Oxford: Blackwell. Original edition, Oxford: Blackwell, 1986. Tulloch, Lee 1990 Reprint. Fabulous Nobodies. London: Picador. Original edition, London: Chatto & Windus, 1989. Wells, H.G. 1935 Reprint. Tono-Bungay. New York: Modern Library. Original edition, New York: Modern Library, 1909. Wolfe, Tom 1988 Reprint. The Bonfire of the Vanities. London: Picador. Original edition, New York: Jonathan Cape, 1987.

Modal auxiliary constructions, ΤΑΜ and interrogatives Richard Matthews

- Is that grammatical? - It'd damn well better be grammatical!



In this paper I wish to review grammaticalized modality, as expressed by English modal auxiliaries, from the point of view of its interaction with the temporal properties of finite clauses, and its interaction with the illocutionary properties of interrogatives. The first question entails the whole TAM-complex (Tense, Aspect, Modality)1 the second the pragmatics of interrogatives. These are only two out of a number of issues that arise from treating modal auxiliaries as elements in construction rather than as (free) lexical items. The exchange I have prefaced this paper with reflects how a proposition containing an aspectualized predication with a temporal ascription (Present or Non-Past) can be reacted to by means of an expression (expanded by an intensifier: damn well) of deonticity, which implicates that if the proposition is not true, draconian measures are in the offing. This is one way speakers might respond to predications offered as propositions in discourse. Another, more obvious way is to affirm them (yes), deny them (no), or to take up a position with respect to the degree of truth they express (perhaps, definitely, conceivably, etc.). If the same kind of speaker-commitment illustrated in my example is made to the question of whether the event or state-of-affairs denoted by a predication will occur or not (rather than whether the proposition is true), as in say: (1)

He'd better come if he doesn 't want to miss the fun.

then we have a recommendation for implementation or enactment. The research that this paper is based upon2 examined the questions of temporal and illocutionary properties from the point of view of various types of modality: epistemic, deontic and dynamic (in particular), these


Richard Matthews

being the types established in the linguistic literature, following a tradition in modal logic 3 . Since then, Palmer (2000) has replaced the fundamental "epistemic" vs. "deontic" parameters (of Palmer 1986) with propositional vs. event modality, propositional being subdivisible (in the first instance) into evidentiality and epistemicity, event modality subdivisible into deontic and dynamic modality. In common with a number of linguists (Householder 1971), Seuren (1969), Lyons (1977), etc. and so as to be consistent with linguistic philosophers like Hare (1970), I posited a three layered system (Matthews 1991), of utterance ascriptions: i) an illocutionary specification; ii) modality proper; iii) the nucleus (the predication). This framework was linked to temporal aspects of utterance specification, using a neo-Reichenbachian system: Illocution is linked to ts (speech-time), Modality to tr (reference time), and event/situation time (te) has the predication (Nucleus) within its scope. But the Nucleus can be aspectualized, or rather is marked for aspect. Hence, with μ standing for the point of modification, ρ standing for predication and ts, tr, te as above, we can envisage the following arrangement: Illocutionary type


[μΛ Declarative, Interrogative, etc.

[μΛ Primary tensing/ modality

Event time/ modification [μΛ

Aspect-marked Nucleus [μ t Ρ ] ] ] ] Progressive vs. non-progressive

For example: (2)

He may have been writing.

can be analyzed in terms of an Illocution "I say"; a Modality "it is possibly the case"; a temporal location with respect to reference time (tr) "anterior to now"; and an aspectualized predication: . (In fact, I have left out another modification of the nucleus: that for passive vs. non-passive). This kind of analysis is broadly consistent with, and was partly influenced by the approach in Lyons (1977). There are, however, many differences in detail. Functional Grammar of the kind developed by Dik has a compatible layered analysis of the clause, as reported in Siewierska (1991). 4 Illocution corresponds to FG's level 4, Modality to level 3, Event time/modification to FG's level 2, and Nucleus to level 1.5

Modal auxiliary constructions, ΤΑΜ and interrogatives 2.


Modal auxiliaries and meaning

The English modal auxiliaries are relatively easy to delimit, if we take as a single defining criterion their restriction to operator-only function within the finite clause6. All the items listed below are so restricted, though some, those enclosed in brackets, only fulfil the "operator-only" condition in interrogative clauses; in the case of hadn't ... better, the form itself is limited to interrogative clauses in standard usage.7 (One might have included may well, may as well, etc. as examples of complex expressions that fulfil the operator-only condition.) I have added a solidus to mayn't, because it is now hardly to be encountered8, and percentage signs to those that seem to be limited to British (or even English) English, and thus not normally available to speakers of General American English. Table 1. Modal auxiliaries with operator-only function will may can %shall must (need) (dare) ought to am/is/are to has/have got to

won't "("mayn't can't %shan't mustn't needn't daren't %oughtn't to isn't/aren't to hasn't/haven't got to

would might could should

wouldn't %mightn't couldn't shouldn't

was/were to had got to

wasn't/weren't to %hadn't got to

It is generally accepted that English modal auxiliaries have taken on or acquired many of the functions of moods which were present in older forms of the language, and match many uses of moods in other languages. Because they are more numerous, there is a greater potential for nuances of meaning. Modal expressions range from the more lexical, and semantically most transparent, for example possible in It is possible that he went, to the more grammatical, and semantically least transparent. English modal auxiliaries, because of their operator-only condition are at the grammatical end of the spectrum, markedly more so than their German or in fact general Germanic etyma. In no other Germanic language have the descendants of the old praeterito-praesentia class and their fellow travellers (e.g. will) become quite so grammaticalized.


Richard Matthews

If we accept a high degree of grammaticalization, as we must, then we have to accept its concomitant, (relative) semantic imprecision or opacity. Able and capable are relatively precise expressions, for which monosemous, context-independent semantic representations can be formulated; even their context-dependent meanings (in construction, i.e. premodifying an animate nominal, as complements of animate nominal subjects, etc.) will be seen as a particularization or metonymic extension of that semantic representation, e.g.: ABLE:

(of animate being) having power to do things (property known as ABILITY) ABLE TO vp:(of animate being) having power to enact (VP); ABLE N: (of animate [N] coextensively) having power to do things well.

Compared to this, the meaning(s) of constructions containing (modal) can are far more general and less precise. Although traditionally and popularly associated with "ability", it is also to be found glossed by "possibility", "permission", etc. It differs from able in being compatible with subjects that are inanimate or abstract. The majority of speakers reject 3a and 4a, but obviously accept 3b and 4b. (3) (4)

a. * This is able to be done. b. This can be done. a. ? * This fact is able to disprove their hypothesis. b. This fact can disprove their hypothesis.

Despite the wide range of readings associated with can, dependent in whole or in part upon "uses", a number of linguists have presumed it has an essential "nil obstat" meaning, e.g. Ehrman (1966), Papafragou (2000). Papafragou's version is expressed as: ρ is compatible with Dfactua|, i.e. the proposition is compatible with the factual domain (of discourse). My own view is that can is an assertion of possibility "it is the case that ρ is possible", as opposed to (epistemic) may, which is a predicted or speculated possibility, glossable as "it is possibly the case that p". (For greater detail, Matthews 1991: 228, 1993: 136). There are, however, modal auxiliary uses where no (productive or synchronic) relationship between readings can be established. For example: (5)

May they thrive!

Modal auxiliary constructions, ΤΑΜ and interrogatives


whose semantic relationship to: They may thrive - whether "epistemic" ("it is possible that they will thrive") or "deontic" ("they are allowed/permitted to thrive") is tenuous, to say the least. Historically, this construction was preceded by the use of mot, the obsolete present tense form that has in other constructions been replaced by must\ In view of the high degree of grammaticalization of modal auxiliaries, it is somewhat surprising that many semantic accounts have seemed to approach them as if they were lexical by employing componential analysis, componential feature analysis or componential-predicate analysis. These include works such as Anderson (1971), Leech (1969), Hermeren (1978), and a number of others. A notable exception, however, is Strang (1968), who lists a set of "moods", thus committing herself to the grammatical status of these items. Table 2. Strang's moods mood of determination mood of resolution permissive mood concessive potential compulsive conditional determinative-conditional potential-conditional

will shall may might can must would should could

This list, however, is not further elucidated and is patently a rather make-shift creation.9 It implies monosemy, a topic that I address below. Moreoever, it is not accompanied by any detailed examination of uses. One of the consequences of the lexical semantic approach is that an important part of the construction into which modal auxiliaries enter tends to get overlooked. This is the modal complement with its tense and aspect marking as well as the nature of the predicate itself and the illocutionary status of the sentence in which a modal auxiliary occurs. Another question that has divided researchers on modal auxiliaries is whether they should be treated as polysemous or univocal (monosemous). Coates (1983), Sweetser (1990), in particular have given separate semantic specifications (even if partly related specifications) for "epistemic" and "root" uses of modal auxiliaries. Ehrman (1966) argued for monosemy (basic meaning), allowing however for certain overtones (extended


Richard Matthews

meanings) in some cases. More recently, Perkins (1983) suggested monosemous specification, allowing however for differences of domain, i.e. between rational, social, and natural laws (corresponding to epistemic, deontic and dynamic, respectively). Papafragou (2000), working within a Relevance Theory framework, has argued for monosemy, expanding the pragmatic apparatus for interpretation considerably. In other words, a modal auxiliary is monosemously specified and semantically underspecified, the rest of its interpretation (or the interpretation of the utterance it is contained in) being supplied ("inferred") from the linguistic, situational and discourse context. Whereas most researchers into modal meaning have attempted to crack the whole system, Papafragou only deals with five items: can, may, must, should, and, more in passing than by focussing on it, ought to. Her argumentation is persuasive, but her data may well be too restricted and even too selective for her hypothesis to be upheld. Without referring explicitly to regularities of construction, without attempting to handle the whole system of modal auxiliaries, the polysemy vs. monosemy issue cannot be resolved. The question to address, then, is: given that the modal auxiliaries are highly grammaticalized, to what extent can their interpretations be inferred from the constructions they occur in? Or the corollary: to what extent do their interpretations fail to transcend construction?


Modal auxiliaries and types of meaning

Apart from "epistemic", "deontic" and "dynamic" modality, some further types have been differentiated in the literature. In my own work (Matthews 1991, 1993), I have distinguished: "subjunctive substitute", "epistemic", "logical", "deontic", "dispositional". "Subjunctive substitute" is a "catch-all" category for modal auxiliaries appearing in contexts that in other languages (especially Latin) would be typically subjunctive, when there was no match in interpretation with other uses of a particular modal: (6) (7) (8)

May they be happy! Should you see them, tell them to call! I'm surprised that she should think that.

Cases like (6) are cases of illocutionary modification (Cf. let in hortatives), while those in (7) and (8) are cases of ultimate grammaticalization indicating "potentialis", or more precisely, non-assertivity of the proposition.

Modal auxiliary constructions, ΤΑΜ and interrogatives


"Subjective epistemic" covers cases like: They may be coming when it is the speaker's assessment of the likelihood of that is focal. "Objective epistemic" covers cases like: It's possible they're coming, i.e. lexical modality, and: They CAN be coming, when uttered as a counter to: They can't be coming, i.e. logical modality. "Logical modality" has also been used to include irrealis "dynamic" modality forms as in Karttunen's (1972) "other worlds" example: It isn't raining in Chicago, but it could be. "Objectivized epistemic" can be used for cases in which modalized propositions are "mentioned" or "quoted", e.g. If they "may be coming", I'm going! (/^-clauses favour "deontic" interpretations and resist "epistemic" ones unless they are quoted or objective: If it's possible she's there, it's possible she's still alive.) "Subjective deontic" modality covers cases in which the speaker is the source of the deontic modality: You may go, Jeeves!', many of these will also allow a performative interpretation. "Objective deontic" modality is taken primarily to be cases like: You are permitted to go as well as: It is permitted to smoke in here, in which the deontic source is someone other than the speaker. One could talk of "objectivized deontic modality" in cases like: (9)

If they "MUST come ", we've got room for them.

where the expression MUST come is again quoted from the preceding discourse - hence the quotation marks. When we look at "dynamic" modality, there is a fairly clear distinction between forms that presuppose an external agency or dynamic source (10)

You can go there if you wish (only pragmatically distinguishable from "deontic")

and those that can presuppose internal agency: (11)

You CAN do it, you see! (cf. You're able to do it)


How DARE you say that!

Given that modal auxiliaries can cover a whole gamut of uses and thus interpretational types, let me contrast two cases, those with may and those with can.


Richard Matthews

Table 3. Non-interrogative uses of may and can Subjunctive substitute Subjective epistemic Objective epistemic Objectivized epistemic Subjective deontic Performative deontic Objective deontic Objectivized deontic External dynamic Internal dynamic Aspectual

May they be happy! They may be happy If they "may be happy",... They may do it You may go! If they "MAY do it",... ... so that they may profit -

They CAN be happy

You can go! They can do it They can be found She can dance She can hear you

Even if I have failed to capture all the non-interrogative uses here, it would still be implausible to say that may is seven-ways ambiguous, and can six-ways. Leech and Coates (1980: 85) argue persuasively for polysemy of may ("epistemic possibility" vs. "root permission") and monosemy of can ("root possibility" with gradience to "permission" and "ability").


Modal auxiliaries and temporal properties

If it is accepted that there is such a component as Tense/Aspect/Modality (ΤΑΜ) in finite verb systems, and that modal auxiliaries are (highly) grammaticalized modality expressions, then it surely follows that the behaviour of modal auxiliaries has to be viewed in the light of such a component? This is what was examined in the research I carried out as part of a project on the temporal-aspectual system of English. Briefly, I was testing the behaviour of some fifty predications selected as being representive of a variety of event notions and participant role configurations. These were the constants, assuming the predications were lexically non-ambiguous. The variables were tense and aspect marking and the presence or absence of a modal auxiliary. For example, given a predication without any kind of finite marking like , this can be tensed (Non-Past vs. Past), it may be secondarily tensed (Non-Perfect vs. Perfect), it may be aspectualized (Non-Progressive vs. Progressive) and it may be modalized (primary tense vs. Modal, i.e. may, can, will, shall, must, might, could, would, should, ought to; can't, won't, shan't, mustn't, etc.). This suggests that modal

Modal auxiliary constructions, ΤΑΜ and interrogatives


auxiliaries are solely in commutation with primary tensing. That this is not always true will be explained below. I was interested in the kind of interpretative regularity all these morphological permutations might reveal. Did modalization by a particular modal auxiliary preserve a basic ambiguity between a descriptive or narrative use of a predication (say )? John smokes is most readily interpreted as the ascription of a habit to John, i.e. a descriptive use, but a narrative use is not excluded (stage directions, etc.). John may smoke and John must smoke with "epistemic" modalization also favour the ascription of habit interpretation, but as "deontic" modalization will favour permitting or obligating to enact an instance of smoking. (As a more tenuous "dynamic" modalization (subject-internal necessity), John must smoke could be ascriptive: John must smoke, poor dear, he's addicted!) The results revealed, apart from a welter of information on the pragmatics of certain predicates and predications, the constant repetition of a certain set of temporal reference possibilities with specific tense and aspect configurations, given a fixed reading of other variables. These form one strand in a grammatically-based semantic approach to modal auxiliaries. Another strand is a division between realis, potentialis and irrealis modality,10 between presupposing predications as factual, as potentially factual, or as contrafactual. To set a standard for comparison, I take first the temporal-aspectual paradigms of predications that do not contain a modal expression. On the one hand, we have the basic narrative/descriptive tense paradigms, and on the other, the Imperative, or as I prefer to call it, the directive paradigm, which includes the use of declaratives uttered in order to change the world, e.g. Look, you're going (or else)! The patterns cover only main clauses; there are, of course, other patterns in subordinate clauses such as the so-called 'mandative subjunctive', e.g. (I insist that) she be there, and the "irrealis subjunctive" in conditional clauses, e.g. (If) she were here. Both of these, however, have alternatives in the shape of expressions with modal auxiliaries: (...that) she should be here·, and: (if) she were to be here. The readings listed below use the following terms: "narration" for sequencing of events, etc.; "description" for non-sequencing; "event" for a punctual event or act or achievement (dynamic, non-durative); "process" for telic and non-telic processes or accomplishments and activities (dynamic, durative); and "state" for qualities and states, etc. (non-dynamic, durative).11 By "dissected event" I mean that what is punctual has been opened up into either a pre-phrase (inception), e.g. She's striking a match or post-phrase (result), e.g. She was looking away; by "iterated event", I mean one that is repeated, e.g. She was hitting the dog; by "bounded state",


Richard Matthews

I mean one that, instead of being open-ended, is temporarily bound, e.g. We're living in Paris. I have taken just four representative intransitive verbs: be nice (state), sneeze and leave (events), smoke (process). Table 4. Present Narrative/Descriptive pattern: reference time is simultaneous to speech time (tj = t,)

She's nice/ smokes/sneezes/ leaves She's being nice/ smoking/ sneezing/leaving

-Past-Perf-Prog V

current narration

current description

event/holistic process

state event/process as habit ongoing process dissected/iterated event bounded state event/holistic process/ state within anterior time span ongoing process dissected/iterated event bounded state in anterior time span

-Past-Perf+Prog V

She's been nice/ -Past+Perf-Prog V smoked/sneezed/left She's been being nice/smoking/ sneezing/leaving

-Past+Perf+Prog V

Table 5. Past Narrative/Descriptive pattern: reference time is anterior to speech time (ts > t,) remote narration event holistic process

remote description

She was nice/ smoked/sneezed/left

+Past-Perf-Prog V

She was being nice/smoking/ sneezing/leaving

+Past-Perf+Prog V

ongoing process dissected/iterated event bounded state

She'd been nice/ smoked/sneezed/left

+Past+Perf-Pro26 V

event/holistic process/ state within anterior time span or at anterior time

+Past+Perf+Prog V

ongoing process dissected/iterated event bounded state within anterior time span

She'd been being nice/smoking/ sneezing/leaving

state event/process as habit

Modal auxiliary constructions, ΤΑΜ and interrogatives


A very much restricted paradigm is encountered in the case of directives. Non-imperatives are more restricted in acceptability than Imperatives: Be nice! We being nice! ?She's nice (or else)! ?She's being nice(or else)! This seems to be a consequence of the lexical nature of the predicate. Frequently, the progressive version is very much co-text-dependent, e.g. on a temporal adverbial naming a reference time. I use 'state-of-affairs' as a cover term for: ongoing process/activity; dissected/iterated event; bounded state. Table 6. Directive (Imperatives, directive uses of Pres and PresProg): reference time is not before speech time (tj < tr) . Be nice! She leaves! _ .. . . , Be smoking! She s leaving! ~ ®

_ „π .. -Past-Prog V

direction to enact event/holistic ,^ . . . process/state in Afuture _ ^, _ , , direction to realize state-of-affairs -Past+Prog . , .. . . . Λ . p V (ongoing process/activity) in future

The commonest pattern involving modal expressions is what I have called the "evaluative". This includes the majority of so-called epistemic, alethic, some dynamic and "deontic comment" uses. Subdivisions can be made for realis (can, has got to), potentialis (may, must,) and irrealis (could, might, should). What is remarkable about this pattern is that the event/process in narrative sequence is hardly attestable. Even in stage directions, the following sequence is questionable: He enters. He must stumble, because he grabs hold of the table to steady himself... This cannot be an epistemic evaluation. It is, however, possible to get a deontic interpretation; this is not evaluative, but stipulative, cf. below. Table 7. Evaluative: reference time (the point of evaluation) is simultaneous to speech time (tj = tr) She must be nice/smoke/ ?sneeze

MOD-Perf-Prog Inf

state/habit in description of current situation

She must be being nice/ smoking/sneezing

MOD-Perf+Prog Inf

"state-of-affairs" in description of current situation

She must have been nice/ smoked/sneezed

MOD+Perf-Prog Inf

She must have been being nice/smoking/sneezing

MOD+Perf+Prog Inf

event/process, state or habit anterior to current situation "state-of-affairs" anterior to current situation


Richard Matthews

The next pattern can be called "predictive", and this covers most of the futurate uses of modal expressions. What is interesting about this pattern is that futurity is fundamental with modal plus simple infinitive, but modal plus perfect infinitive and perfect progressive infinitive require a suitable adverbial or contextual setting for a future reference point to get this kind of interpretation. Thus she 7/ have gone without further contextualization will be interpreted as referring to an event completed anterior to the point of reference, which is otherwise the point of speech, whereas she'll go without further contextualization will be interpreted as referring to an event taking place at a reference point which is posterior to the point of speech. This applies in the case of will and may, but significantly not in the case of be going to, which is not, however, an operator-only modal expression as we can have: She 7/ be going to go and: She's been going to go (for hours). This means that without overt future orientation, perfect and perfect progressive cases are indistinguishable from those of evaluative. The predictive pattern also includes conditional uses of would, might, could and should. Of these, would is irrealis, while the others, I claim, are potentialis-irrealis; they can appear in both non-factual and contrafactual contexts: She might go, if she feels like it vs. She might go, if she felt like it. Table 8. Predictive: reference point is not before speech time: (ts < tr) She may be nice/smoke/ sneeze/leave

MOD-Perf-Prog Inf

She may be being nice/ smoking/sneezing/leaving

MOD-Perf+Prog Inf

She may have been nice/ smoked/sneezed/left (by then)

MOD+Perf-Prog Inf

She may have been being nice/smoking/sneezing/ leaving (by then)

MOD+Perf+Prog Inf

event/holistic process/state in future "state-of-affairs" in future (description) event/holistic process/ state completed by future reference point "state-of-affairs" terminated by future reference point

The next pattern is one that can be related to the directive pattern that we established for non-modal expressions, in which, significantly, perfect forms are excluded, while the basic aspectual distinction between non-progressive and progressive is allowed. In this third modal pattern, which I called "stipulative", the perfect is not absolutely excluded, but is dependent on appropriate adverbial support or other co-textualization. It differs from the predictive pattern, in that these perfect and perfect progressive forms do not get any kind of reading where reference time is

Modal auxiliary constructions, ΤΑΜ and interrogatives


simultaneous to speech time even without the future time adverbial. This pattern includes what have hitherto been classified as "deontic" uses of must, may and shall. Table 9. Stipulative (ts < t,) She must be nice/smoke/ sneeze/leave

MOD-Perf-Prog Inf

stipulates enactment of event/ holistic process/state in future

She must be being nice/ smoking/sneezing/leaving

MOD-Perf+Prog Inf

stipulates realization of "stateof-affairs" in future

She must have been nice/ MOD+Perf-Prog Inf smoked/sneezed/left by then She must have been being nice/smoking/sneezing/ MOD+Perf+Prog Inf leaving by then

stipulates realization of state of completion of event, etc. in future stipulates realization of state of terminated "state-ofaffairs" in future

The fourth pattern is one that I call "dispositive", and it differs from the others in that we have a tense modification of a modal expression. These resist non-perfect forms, but are susceptible to the aspect opposition: non-progressive vs. progressive. This includes so-called "dynamic" uses of can and could, and habitual and dispositional uses of will and would. This kind of predicate is subject to considerable restriction, normally allowing only subjects that denote animate beings or forces. With can/could, there is difference between an occasional-habitual and an abilitative interpretation according to whether the item is stressed or not. In some cases, it is hard to detect any difference between a dynamic and an objective epistemic/logical reading, e.g. She CAN be nice, "it is certainly possible that she's nice" and "she is known to be nice at times", but with could there is an ambiguity between a past factual and a present potentialis-irrealis interpretation. Table 10. Dispositive (t, = t,) or (t, > t,) She can be nice/smoke/ ?sneeze/leave She can be being nice/ smoking/?sneezing/leaving

-Past MOD-Perf-Prog Inf -Past MOD-Perf+Prog Inf

habit/state in current description iterative state-ofaffairs in current

She could be nice/ smoke/ ?sneeze/leave

+Past MOD-Perf-Prog Inf

habit/state in remote description

She could be being nice/ smoking/?sneezing/leaving

+Past mod-Perf+Prog Inf

iterative state-ofaffairs in remote


Richard Matthews

Finally, we have a pattern which I call "processive". This matches what has hitherto been referred to as the aspectual use of can and could. Only verbs of inert perception fit into this pattern. Again we have tensing of the modal expression, but unlike dispositive, both perfect and progressive choices are excluded. These are descriptive uses, which contrast with the narrative use of non-modalized inert perception verbs, cf. He can see her, so he calls out to her vs. He sees her and calls out to her. Table 11. Processive (ts = t,) or (ts > t,) ,, . He can see it

^ _ T r -Past MOD-Perf-Prog Inf °

process of perception in current r .. description

He could see it

+Past MOD-Perf-Prog Inf

P ™ e s s o f perception in remote description

Both the dispositive and the processive patterns allow primary tensing (±Past) of the modal auxiliary. In view of this, the modal auxiliary can be seen as being outside the modality specification; with dispositive, the modal auxiliary is in commutation with the te specification of the Nucleus, while with processive, the modal auxilary is in commutation with aspectualization of the Nucleus predication. To sum up: Strict epistemicity and alethicity and "deontic comment" can be seen as sub-types of evaluative, which could be renamed: "epistemic evaluation", "logical evaluation", and "deontic evaluation" respectively. Directive deontics, including performative deontics, can be associated with the stipulative pattern. The distinction between evaluative and predictive allows us to explain the different behaviour of epistemic interpretations of may and must, in that may fits both paradigms, whereas must is restricted to the evaluative. The distinction has sometimes been referred to as the difference between the openness or forward-lookingness of may as opposed to the closedness or backward-lookingness of must. Approaching the meaning of modal expressions from temporalaspectual paradigms does not give us a complete and exhaustive analysis of modal auxiliary construction semantics, but it does provide a principled approach, which the lexical semantic approach and the "epistemic" vs. "deontic" classification do not provide. And coupled with other distinctions such as realis, potentialis, and irrealis, and between expressions of absolute and relative degree of modality, we can establish a framework against which readings of modalized expressions can be matched. Above all, where we have a match between different readings and temporal-aspectual paradigms {can, I suggest, is a case in point) we have

Modal auxiliary constructions, ΤΑΜ and interrogatives


the basis (but not, as such, the evidence) for a pragmatic-procedural account, as suggested by Papafragou (2000).


Modal auxiliaries in interrogatives

In the preceding section, I have looked only at modality as expressed by modal auxiliaries in Declarative Main Clauses. To explore the constructional basis of modal auxiliary interpretation further, we need to know a) whether these patterns extend to interrogatives, b) whether individual modal auxiliaries are excluded in interrogatives - the case of "epistemic" may being a well-documented example. Not only are there two major types of questions: polar (which relate to predications) and particular (which relate to participants and circumstances), there are also various ways of posing questions. One way is to use an interrogative clause-structure. These are "text-book" questions. And it is forms like: Is Tom leaving? rather than Tom's leaving? or Tom's leaving, isn't he? or Tom's leaving, is he? or So Tom's leaving, eh? that I shall be considering here. And, by the same token, forms like: Who's leaving? or What's he doing? rather than: He's doing what? or He's seeing who? in the case of particular questions. The background to the posing of polar questions may vary. If the questioner is posing an unbiased question without any presupposition of a positive or negative response, this has been traditionally termed an open-question. But a questioner may have some prior belief as to whether the proposition being entertained is true or false, and in such cases this is called a non-open question. Compare: (13) (14) (15)

Is Tom leaving? (without contrastive stress: open) Isn 't Tom leaving (then)? (presupposes "I thought he was leaving": non-open) Is Tom leaving (or isn't he)? (presupposes "I thought he was leaving" - he appears not to be: non-open)

When looking at polar interrogatives containing modal auxiliaries, I found it necessary to distinguish between three cases. For "epistemic" interpretations, these were: (i) modal submission (which is analogous to open questions); (ii) modal dissent when an asserted proposition is questioned (this is like a non-open question); (iii) modal challenge when it is the modal ascription of a modalized proposition that is directly challenged, which is comparable to the second type of non-open question with contrastive stressing above. For example:





Richard Matthews A: I haven't heard from Tom this week. B: Might he be sick? (Tom's being sick is submitted for consideration) A: Tom appears to have arrived. B: Can he have? (The implicit acceptance of "Tom has arrived" is dissented from) A: Tom must have arrived. B: Must he? (The conclusion is challenged by directly questioning the modal)

Clearly any modal that can appear in a declarative clause with an epistemic interpretation can be challenged. Those where dissent is involved appear to be restricted to: can, must, and need and the semi-auxiliaries have to, be bound to. The items that seem to work for modal submissions are even fewer: might, could, would, and, somewhat less clearly, can and will. It is significant that three of these are oblique forms, or functional "tentatives".12 The nature of interrogation is itself dependent on epistemicity, in the strict sense: the questioner doesn't know the truth of ρ and asks not: is ρ the case? or is it the case that p?, but: is it possible/imaginable, etc. that p. Apart from might, could and would, can and will figure in "logical" modality and-or "dynamic" modality.13 Though negation in non-modal interrogatives seems to preclude an open interpretation, my findings suggest that certain negative modal auxiliaries can be ways of making modal submissions. (19)

A: Tom's not here. B: Mightn't he have left?

The same kind of submission seems possible with: couldn't and won't, but can V and wouldn 7 are less certain. Negated be bound to: Isn't he bound to have left is also admissible. Modal dissent is restricted to can V, couldn't, don 't/doesn't... have to, and aren 't/isn't... bound to. With the exception of mightn 7, which is usually taken to be propositional negation, all the other modal submission forms {couldn't, won't, can't, wouldn't) are cases of modal negation in declaratives. With "deontic" interpretations of modalized interrrogatives, the one that comes closest to open questions is one I call modal enquiry - it enquires about the deonticity of a course of action without any presupposition that this action has been proposed or realized. In contrast to this we also have an interpretation that enquires about the deonticity of a proposed or realized course of action, which comes closest to a non-open question; and finally

Modal auxiliary constructions, ΤΑΜ and interrogatives


one which challenges the deonticity of an action or state of affairs. These are referred to as deontic enquiry, deontic dissent, and deontic challenge, respectively. For example: (20) (21) (22)

Should I go? A: He's going. B: Should he go? A: He should go. B: Should he go?

(deontic enquiry) (deontic dissent) (deontic challenge)

It is clear that deontic enquiry is the most context-independent of these and the number of expressions involved is higher than that involved in epistemic modal submissions. These are: may, might, can, could, shall, should, ought to and possibly had better. Deontic dissent includes: can, should, must, need, ought, have got to, and semi-auxiliary have to, need to. All items tested except will/would and perhaps, for other reasons, might seem to permit a challenge interpretation. Will/would are not, in declaratives, primarily deontic (23) (24)

You will clean your kit! (implicated command) *You would clean your kit! (not as hedged command)

The first of these can receive a deontic interpretation only pragmatically. Might is also derivative, dependent on an implicature when used in a declarative. (25)

A: Well, you might open the door for me! B: Might I?

What is common to the modal enquiry interpretations is that a modal source is brought into play. There are significantly more modal expressions that do not make some kind of presupposition. Negated expressions again favour oblique or hedged modality: might... not, couldn't, shouldn't, oughtn't to, hadn't ... better, the only clear non-oblique being can't. May ... not is probably acceptable as deontic enquiry. Deontic dissent is possible with other negative modal expressions, among them: may... not, mustn 't, needn't, hasn't... got to, don 't/doesn 't... have to, don't/doesn't... need to. Since won't, wouldn't and mightn't are dubious as declarative deontics, they are also dubious as deontic challenges in interrogatives.


Richard Matthews

Of the items that could be interpreted as dispositional/dynamic, can/could, will/would, dare, need and can't/couldn't, won't/wouldn't, daren't, needn't, only will resists a dynamic interpretation under interrogation, except as a challenge: (26) (27) (28)

Will she go? (not volitional) She's going. A (not volitional) Β Will she go? A She WILL go. (and no-one can stop her) Β Will she?

A complicating factor is that a number of interrogative modal auxiliary constructions take on a special illocutionary function. In particular, this includes shall, especially in shall I/we, but also in the construction of requests. Tests for such interpretations did not reveal anything beyond the expected, which was that: will, can, would, could, won't are favoured more or less in that order - and that can't, couldn 7 and especially wouldn 't are resisted. Compare: (29)

Will Can Would Could Won't ? Can't ?* Couldn't * Wouldn't

you come in please?

Particular questions only have a participant or circumstance within the scope of the interrogative illocution. It might be expected therefore that the rest of the predication will pattern like declaratives. This turns out not to be quite the case. First, we need to differentiate between cases where the modal expression is part of the predication and those where it is external to it, e.g. (3 0)

What might she do tomorrow?

could be interpreted as: -

what (she might do what might (she do

tomorrow) tomorrow)

Modal auxiliary constructions, ΤΑΜ and interrogatives


i.e. "what is it that she might do tomorrow" or "what is it possible that she will do tomorrow". In many cases these might be cognitively equivalent under epistemic interpretations; pragmatically, they are distinguishable. Secondly, the nature of the wA-item interacts with the type of modality. I tested what, why and how. What [and presumably who(m)] request identification of a participant within the core predication; the occurrence of modal items is least extensive. How is linked to elements that expand the core predication, typically the manner of enactment, but not necessarily: (31)

How might she have gone? i) in what manner might she have gone? ii) how is it possible that she went?

//ow-questions strongly favour the dynamic and occurrential basis of situations and thus dynamic/dispositional modal interpretations. PFAy-questions are linked to contingent circumstances outside the core predication (causes, etc.). They, too, can have alternative interpretations. (32)

Why might she have done it? i) what might have been the reason for her doing it? ii) why d'you say she might have done it?

PF^-questions strongly favour the epistemic and evaluative bases of modal judgements. In addition to the relatively free constructional potential of vf/z-questions, there are also formulaic patterns: How can ... ! How could... ! How dare ... /, which can be termed exclamatives, and: Why should... ! which can be termed deontic challenges.



From the foregoing it can be said that modal interpretations are very strongly dependent on pragmatic factors. Few modal auxiliaries permit a "straight" interpretation in polar questions; other interpretations reflect assumptions and presuppositions set by context, i.e. pragmatic factors. In w/z-questions, the patterning is markedly different, the nature of the wA-item being an overriding factor. There is no way, it seems, that one can avoid the conclusion that the construction in which a modal auxiliary appears, or is compatible with, determines to a great extent the interpretation it receives. This corroborates those approaches which say that


Richard Matthews

a) the meaning, i.e. semantic specification, of modal auxiliaries should be under-determined with respect to the individual instance, b) the meaning, i.e. the pragmatic interpretation, of modal auxiliaries in construction is largely determined by the meaning, i.e. the semantics, of that construction relative to its use. Whether a modal auxiliary is univocal or ambiguous may be inherent to the auxiliary, but its interpretation in particular instances will be determined by construction, context and convention. Certain constructions are fixed, i.e. conventionalized, to such an extent that the modal auxiliary that appears in them is practically devoid of meaning, "optative-may " for example. Lexical semantic approaches are inappropriate for what is a constructional semantics issue. While I cannot come down on the side of the monosemists, because a number of modal auxiliaries are clearly ambiguous in analogous structures, it does seem to me that the "epistemic" vs. "root" or "deontic" distinction is not as such a part of the semantic specification of all modal auxiliaries. May- and m««/-constructions need to be specified in such a way that is compatible with predictive vs. inferential and possibility/permissive vs. necessitive interpretations respectively. SAowW-constructions, however, (except, of course, the "subjunctive-substitute" use of should) can be readily explained pragmatically. (33)

There should be no problem at the border.

Uttered without knowledge of the real, current situation at the border, we would have to say that it is "epistemic" or at least a "tentative prediction" potentially in contrast to will·, uttered knowing the contrary of the proposition it includes, i.e. knowing that there are problems at the border, then we have to say it is "deontic" or a comment on the mismatch between desirability/requirement and reality, i.e. what I have called "deontic comment".14 This seems to apply to ought to as well. Can, too, appears to allow a single meaning, which is then fleshed out according to context. If this kind of minimalist semantics is viable: May all your problems with modal meaning be pragmatic ones! Notes 1.

In point of fact, what is generally referred to (Chung & Timberlake 1985, inter alia) as ΤΑΜ, would be better cast as MTA for English), compare my diagram in section 1.

Modal auxiliary constructions, ΤΑΜ and interrogatives 2.


Some of the research on modal auxilaries and temporal relations was reported on in a lecture given at the Societas Linguistica Europaea meeting in Kiel 1991, but it has otherwise remained unpublished. This was funded in part by the Deutsche Forschungsgemeinschaft. The section on modality in interrogatives is based on Matthews (1991). 3. Von Wright (1951) distinguished: alethic (= aletheutic) i.e. modes of truth, e.g. necessary, possible, etc.; epistemic, i.e. modes of knowing, e.g. verified, falsified; deontic, i.e. modes of obligation, e.g. obligatory, permitted, etc.; existential, i.e. modes of existence, e.g. universal, existing, empty. He added dynamic, for modes of action, in a footnote. To these can be added, (cf. Rescher 1968), temporal (sometimes, mostly); boulomaic (= bouletic) (hoped, feared, desired); evaluative (good, bad); causal (cause, prevent); likelihood (likely, probable). 4. Those particularly associated with the development of modality (in its broader sense of incorporating mood and modal specification) are Hengeveld (1989) and Nuyts (1993,2001). 5. Modification of event time would include habituals, which have sometimes been classed as aspect, sometimes as modality; "aspectual" uses of can are in commutation with progressive aspect and can neatly be assigned to modification of the Nucleus (level 1 operator in Dikian Functional Grammar terms). 6. The much-quoted "NICE" properties (behaviour with respect to negation, interrogatives, "code", and emphasis) follow from the operator-only property. 7. Hadn't (Subject) better is the normal interrogative form; had better not the normal declarative form. Hadn't better in declaratives is extremely colloquial and attested 4 times in the BNC (3 spoken; 1 thought representation). Had better not occurs 31 times. 8. The BNC has 7 instances, 3 of them in tag-questions, i.e. strongly determined by the pattern of auxiliary repetition in such structures. 9. Similar lists were to be found 200 years earlier, e.g. White's treatise The English Verb of 1761. 10. This is a distinction that can be established in Ancient Greek and Latin, and in most Present-day Standard Average European languages in some form or another, whether synthetically or analytically. 11. There is an extensive literature on "aktionsarten". Brinton (1988) contains a useful overview (Chapter 1). Quirk et al. (1985: 200-209) is a practical application and classification under the heading "situation types". 12. There is a problem finding suitable terms for forms like: might, should, ought to, would, could. Historically they derive from past forms, but since three of them never fit into a past paradigm, I prefer the term "oblique", to distinguish them from "base" forms, even if ought to doesn't have an auxiliary base form in contemporary English. Functionally, I differentiate between "hedges" or "tentatives" (forms used to efface directness), "oratio obliqua" or "backshifted" (forms in reported speech), and "remote" (forms used in genuine past tense sequences).


Richard Matthews

13. It has frequently been noted that may, above all, cannot take on an epistemic interpretation in interrogatives, and that this function has been provided by can. Recently, Papagragou (2000) has questioned whether can is ever really epistemic, being inclined herself to say it is (dynamic) "root". 14. Unlike may and must in epistemic and deontic use, epistemic vs. deontic should cannot be differentiated in terms of a fundamental distinctive stress patterning. Differences in stress patterning of specific meanings of modal auxiliaries were recorded in Kingdon (1958), and were common currency in language teaching manuals in those days. Of more recent (theoretical linguistic) accounts of modality, only Coates (1983) has consistently addressed this question.

References Anderson, John M. 1971 Some problems concerning the modal verbs in English. In Edinburgh Studies in English and Scots, A.J. Aitken, A. Mcintosh, Hermann Pälsson, (eds.), 69-120. Edinburgh: Edinburgh University Press. Brinton, Laurel 1988 The Development of English Aspectual Systems. (Cambridge Studies in Linguistics 49.) Cambridge: Cambridge University Press. Chung, Sandra, and Alan Timberlake 1985 Tense, aspect, and mood. In Language Typology and Syntactic Description, Vol. Ill, Timothy Shopen (ed.), 202-258. Cambridge: Cambridge University Press. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Ehrman, Madeline Elizabeth 1966 The Meaning of the Modals in Present-day American English. The Hague: Mouton. Hare, Richard Mervyn 1970 Meaning and speech acts. Philosophical Review 79: 3-24. Hengeveld, Kees 1989 Layers and operators in Functional Grammar. Journal of Linguistics, 25: 127-157. Hermeren, Lars 1978 On Modality in English. A Study of the Semantics of the Modals. (Lund Studies in English 53.) Lund: Gleerup. Householder, Fred W. 1971 Linguistic Speculations. London: Cambridge University Press. Karttunen, Lauri 1972 Possible and Must. In Syntax and Semantics, vol. 1, John P. Kimball (ed.), 1-20. New York: Seminar Press.

Modal auxiliary constructions, ΤΑΜ and interrogatives


Kingdon, Roger 1958 The Groundwork of English Intonation. London: Longman. Leech, Geoffrey N. 1969 Towards a Semantic Description of English. London: Longman. Leech, Geoffrey N., and Jennifer Coates 1980 Semantic indeterminacy and the modals. In Studies in English Linguistics, Randolph Quirk, Sidney Greenbaum, Jan Svartvik, and Geoffrey N. Leech (eds.), 79-99. London: Longman. Lyons, John 1977 Semantics. Vols. 1,2. Cambridge: Cambridge University Press. Matthews, Richard 1991 Words and Worlds. On the Linguistic Analysis of Modality. Frankfurt/Main: Peter Lang. 1993 Papers on Semantics and Grammar. Frankfurt/Main: Peter Lang. Nuyts, Jan 1993 Epistemic modal adverbs and adjectives and the layered representation of conceptual and linguistic structure. Linguistics 31: 933-969. 2001 Epistemic Modality, Language, and Conceptualization. A CognitivePragmatic Perspective. Amsterdam: John Benjamins. Palmer, Frank R. 1986 Mood and Modality. Cambridge: Cambridge University Press. 2001 Mood and Modality. 2d ed. Cambridge: Cambridge University Press. Papafragou, Anna 2000 Modality: Issues in the Semantics-Pragmatics Interface. Oxford: Elsevier. Perkins, Michael R. 1983 Modal Expressions in English. London: Frances Pinter. Quirk, Randolph, Sidney Greenbaum, Geoffrey N. Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rescher, Nicholas 1968 Topics in Philosophical Logic. Dordrecht: Reidel. Seuren, Pieter M. 1969 Operators and Nucleus. Cambridge: Cambridge University Press. Shopen, Timothy 1985 Language Typology and Syntactic Description. Cambridge: Cambridge University Press. Siewierska, Anna 1991 Functional Grammar. (Linguistic Theory Guides XXI.) London: Routledge. Strang, Barbara M. 1968 Modern English Structure. 2d ed. London: Edward Arnold.


Richard Matthews

Sweetser, Eve 1990 From Etymology to Pragmatics. (Cambridge Studies in Linguistics 54.) Cambridge: Cambridge University Press. White, James 1761 The English Verb. Facsimile reproduction. (English Linguistics 135) 1969. Menston, England: Scolar Press, von Wright, Georg 1951 An Essay in Modal Logic. Amsterdam: North Holland.

A pragmatic analysis of the epistemic would construction in English1 Gregory Ward, Betty J. Birner, Jeffrey P. Kaplan



Research into the discourse functions of syntactic constructions (e.g., Prince 1986; Ward 1988; Birner and Ward 1998) has demonstrated the relevance of a contextually salient open proposition (OP) for the felicity of a variety of constructions. In our previous work (Birner, Kaplan, and Ward 2001), we have shown that epistemic would, illustrated in (1), is among this class of constructions: (1)

A: B:

Who's the British woman over there? That would be J.K. Rowling.

Speaker B's use of the modal in (1) is epistemic in the sense that it conveys her assessment of the truth of the proposition being expressed; in this case, it conveys the speaker's level of confidence in the proposition that the woman in question is J.K. Rowling. Note that, for example, the confidence level conveyed in (1) is generally lower than that conveyed by the corresponding simple present indicative ' in (2): (2)

That's J.K Rowling.

but higher than that conveyed by the epistemic evidential modals should and must, as in (3 a) and (b), respectively: (3)

a. b.

That should be J.K Rowling. That must be J.K Rowling.

In Birner, Kaplan, and Ward (2001), we argued that epistemic would·. (1) requires a salient OP in context; (2) instantiates the variable in the OP with a discrete member of some salient set; and (3) conventionally implicates that the proposition conveyed is empirically verifiable. In this paper, we refine our account of the conventional implicature involved and show that


Gregory Ward, Betty J. Birner, Jeffrey P. Kaplan

the demonstrative that in example (1) can actually have one of three distinct readings: it can be used to refer to the actual woman in question, or it can take as its antecedent either the NP the British woman over there or the variable in the OP (corresponding in [1] to the wA-expression). Finally, we will show how this fact can explain an otherwise puzzling ambiguity.


A pragmatic account of epistemic would

Previous work on modality has failed to accurately characterize the epistemic effect of would. Palmer (1990) and Perkins (1983) call epistemic would tentative; Sweetser (1982) calls it a conditional with a general, and suppressed, //^clause; and Coates (1983) declares that it expresses the predictability of some past action or state. These accounts exclude assertions like those in (4a-c), where there is clearly no tentativeness and no conditionality, and the assertion is not about the past. (4)




...I do have some answers for you. You asked about one person declaring all of the income on one property and one person taking all of the expense. The answer to that would be no. (AS in e-mail to BL, 11/14/00) B: Are you the Meredith that was listed in the Graduate Student News? M: Yeah, that would be me. (in class, 10/25/00) (A is holding a plastic bag with a fish in it.) A: Idon't know. Maybe it's in shock. B: Looks dead to me. C: That would be one dead fish. (conversation, WalMart Pet Dept., 12/29/00; token courtesy of M. Larson)

In (4a), there is nothing tentative or conditional about the answer given, nor is there anything tentative or conditional about M's response that would be me in (4b), or the assertion that would be one dead fish in (4c). All three assertions, moreover, concern the present. As we argued in our earlier work (Birner, Kaplan, and Ward 2001), what distinguishes epistemic would from the other evidential modals is the crucial role that open propositions play in its interpretation. An open proposition is a proposition with one or more underspecified elements. For instance, consider examples (5) and (6):

A pragmatic analysis of the epistemic would construction (5)


I plan to discuss several topics. What I'll discuss first is the notion of political correctness. OP: I'LL DISCUSS X FIRST (where X is a member of the set of topics) I don't usually eat dessert. Apple pie I like, but most desserts are too sweet for my tastes. OP; I LIKE X (where X is a member of the set of desserts)

In (5), the first sentence {Iplan to discuss several topics) makes salient the proposition that I will discuss these topics in some order, and in particular that I will discuss some topic first. Thus, the proposition "77/ discuss X first" is salient, where X is a variable ranging over the set of topics. The variable is what makes it an open proposition; a closed proposition would be one with no underspecified elements - that is, one that did not contain a variable. It is this salient OP in (5) that licenses the use of the wA-cleft in the second sentence (What I'll discuss first is the notion of political correctness). That is, it is the hearer's expectation that the speaker will discuss some topic first that makes it appropriate for the speaker to use this construction. Similarly, in (6) the first sentence makes salient the notion that the speaker has certain likes and dislikes regarding desserts; the salient OP "I like X", where X is a variable ranging over the set of desserts, then licenses the preposing in the second sentence (Apple pie I like, as opposed to the canonical word order I like apple pie). Without an appropriate salient OP, the preposing would be infelicitous, as in (7a): (7)

a. b.

I want this to be a great party. #Apple2 pie I like, so I think I'll serve thatfor dessert. I want this to be a great party. I like apple pie, so I think I'll serve that for dessert.

Here, the prior sentence does not make salient the notion that the speaker has various likes and dislikes toward different desserts, and therefore the preposed variant is infelicitous, as indicated in (7a). Notice that the non-preposed variant in (7b) is felicitous, since the use of canonical word order doesn't require the salience of any particular OP. Now consider again the use of epistemic would, as in (1), repeated here as (8):



Gregory Ward, Betty J. Birner, Jeffrey P. Kaplan


A: Β: OP:

Who's the British woman over there? That would be J.K. Rowling. THE BRITISH WOMAN OVER THERE ISX ( w h e r e X is a m e m b e r o f

the set of names) Here, the speaker instantiates the variable in the OP The British woman over there is Xwith the value J. K. Rowling. Such an OP must be salient for the use of epistemic would to be felicitous. To see this, compare (9) and (10): A: B: (10) A: B: OP: (9)

I wish I could marry a millionaire like Max did. #Yeah, that would be Ethel Rothschild. Which millionaire did Max marry? That would be Ethel Rothschild. THE MILLIONAIRE IS X (where X is a member of the set of names)

In (9), the OP The millionaire has name X is not salient, because the sentence is not really about the particular millionaire that Max married, but rather about A's desire to marry a person with the property of being a millionaire. The identity of Max's spouse in this context is therefore incidental and not at issue. The wA-question in (10), on the other hand, renders the OP salient by focusing its variable, and the result is a completely felicitous utterance of epistemic would. In short, the more salient the existence and identity of Max's spouse, the more salient the OP The millionaire has name X, and the more felicitous the use of epistemic would. Thus, we can conclude that the felicitous use of epistemic would requires the presence of a salient OP in the context. Moreover, the clause containing epistemic would must instantiate the variable in the OP with a discrete member of some salient set, as shown in (11): A: What is Omaha like? B: #That would be cold. A: What is the temperature in Omaha today? B: (reading newspaper weather page) Thai would be 5 degrees Fahrenheit. In (11a), speaker B's utterance fails to instantiate the variable with a discrete member of a contextually salient set because be like is too vague to

A pragmatic analysis of the epistemic would construction


evoke such a set and cold is only an ill-defined area of an infinitely varying temperature scale rather than an individuated set member. But in (lib), speaker B's response does instantiate the variable with an individuated member of a contextually evoked set, namely the set of degrees on the Fahrenheit scale, and the example is felicitous. In addition, the use of 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 (cf. Palmer's [2001] notion of epistemic deduction). This rules out, among other things, decisions, predictions, and wild guesses, as shown in (12), (13), and (14), respectively: (12)

(Host to guest at dinner party) Host: Hi Chris, come on in! Glad you could make it. Can I take your coat? Guest: Thanks. Host: Something to drink? Guest: #That would be wine, thank you.

This example is infelicitous because the illocutionary force of the guest's utterance is to request a drink, not to assert a truth-evaluable proposition. (13)

(co-workers gossiping over lunch) A: How much do you predict Jones will make with his promotion to vice-president next year? Β1: #That would be S110K, but I have no real evidence for this. B2: That would be $110K. I saw the new salary schedule.

B l ' s response in (13) is infelicitous because B's utterance encodes a prediction for which the speaker lacks objective evidence. Note that, in contrast, B2's response is felicitous, given the subsequent sentence which provides the requisite basis for his belief. (14)

A: Β: A: Β1: B2:

Guess what color I've decided to paint my living room. How should I know? Come on; just guess. #That would be blue. Ok, blue.


Gregory Ward, Betty J. Birner, Jeffrey P. Kaplan

Finally, Bl's response in (14) is infelicitous because it constitutes a guess, and as such is not based on objective evidence. B2 shows that the infelicity of B1 is not due to the guess per se, but rather to its appearance in the epistemic would construction. In contrast, reasoning and the recovery of facts are felicitous, as they require the speaker's reliance on objective evidence. That epistemic would supports a reasoned conclusion is illustrated in (15): (15)

A: There's only one card left. B: Well, I've only seen three queens, so that would be the Queen of Diamonds.

Example (15) is felicitous because a valid process of logical deduction leads to a reliable conclusion. Similarly, the mathematical calculation in (16) is objectively verifiable and thus supports the use of epistemic would: (16)

A: What's the square root of625? B: That would be 25.

Finally, (17) felicitous because an established fact about the future, like a baseball schedule, is similarly objectively verifiable: (17)

A: How many times will the Padres play the Giants next year? B: That would be eleven.

So far we have shown that epistemic would requires a salient OP, that the clause containing the modal serves to instantiate the variable in the OP with a discrete member of some salient set, and that epistemic would conventionally implicates that the truth of the resulting proposition is conclusively supported by objective evidence. Thus, the salience of the OP and the instantiation of the variable in the OP are crucial to the meaning and interpretation of epistemic would. This variable-instantiating property, furthermore, explains a certain ambiguity found in some sentences containing epistemic would. Consider (18):


A: Β:

What's your most prized possession? That would be my grandmother 's wedding ring.


MY MOST PRIZED POSSESSION IS X (where X is a member of

the set of personal belongings)

A pragmatic analysis of the epistemic would construction Here, B's response is actually ambiguous: the demonstrative that may be anaphoric either to the phrase your most prized possession or to the variable in the OP (in which case the utterance means something like "the answer to your question is 'my grandmother's wedding ring'"). These interpretations are more clearly distinguished in (19)—(21): (19)

A: B:


A: B:


A: B: A:

What was the worst experience of your life? Well, mm, that car accident I had in college was awful, but my PhD. comprehensive was a weekend take-home open-book with 14 questions over 3 subfields...that would be the worst. What are all these instruments? (pointing) This one's a sousaphone, and that would be a flugelhorn. Have you met my roommate Jill? No, but John was telling me about her. I think he said she was remarkably tall. Actually, that would be short.

In (19), that is clearly anaphoric to the NP my Ph.D. comprehensive. In (20), that has deictic reference to the instrument contextually present. In (21), however, the use of the demonstrative that is felicitous in the absence of either deictic reference or a plausible antecedent in the prior discourse; instead, that in A's response is anaphoric to the variable in the OP (i.e., JILL ISX, where X is a member of the set {tall, short}). Example (18) above, then, is ambiguous because that in B's response can take as its antecedent either the phrase your most prized possession, in which case its interpretation is analogous to that in (19), or the variable in the OP, in which its interpretation is analogous to that in (21). A reading for (18) analogous to the deictic reading seen in (20) would not be available without the ring being physically present in the context. Notice that when the antecedent of the anaphor is the variable of the OP, the time reference of the event in question is irrelevant; it may be in the past, present, or future, as seen in (22)-(24), respectively: (22) (23) (24)

A: When was your last trip to Europe? B: That would be 1992. A: When is your trip to Europe? B: That would be right now — I'm on my way to the airport! A: When is your next trip to Europe? B: That would be 2004.



Gregory Ward, Betty J. Birner, Jeffrey P. Kaplan

In (22), the trip in question is in the past, in (23) it is commencing at the time of the utterance, and in (24) it is in the future. Yet non-finite be is permitted in each case on the reading in which that is anaphoric to the variable. The reason is that the variable-instantiation is occurring at the time of the utterance; therefore, the utterance conveying this instantiation need not be marked for past or future time reference. In contrast, when the anaphor is used to refer to the event itself, the tense-marked modal must agree with the time reference of the event, as seen in (25)-(26): (25)



I had a great time the first time I traveled to Europe. That would have been 1992. b. I had a great time the first time I traveled to Europe. #That would be 1992. I bet I'm going to have a blast when I go to Europe. #That would be 2004.

In these examples, the referent of that is the trip itself. And as shown in (25), when reference is restricted to the event itself, and this event occurred in the past, perfective have is required, as shown in (25a). Infinitival be alone, as in (25b), is infelicitous. For future events, however, as in (26), English morphology provides nothing parallel to perfective have, and infelicity results.



We have shown that epistemic would requires a salient OP in context, instantiates the variable in the OP with a discrete member of some salient set, and conventionally implicates that the speaker has conclusive objective evidence for the truth of the proposition conveyed. We have also shown that there is an ambiguity associated with certain sentences containing both a demonstrative and epistemic would. In such cases, that can be anaphoric to either some entity salient in the discourse or to the variable in the OP. Moreover, we have shown how these different meanings have different ramifications for the felicity of infinitival be in sentences making reference to past and future events; only when that is taken as anaphoric to the variable is infinitival be acceptable in these sentences.

A pragmatic analysis of the epistemic would construction


Notes 1.


For assistance with database construction and coding, we thank Meredith Larson and Elisa Sneed. We also thank them, along with the members of the Fall 2000 Discourse Analysis class at Northern Illinois University, for helpful comments and discussion. Following standard practice, we use "#" to denote a sentence that is pragmatically infelicitous, i.e. one that is grammatical and meaningful, but contextually inappropriate in some way.

References Birner, Betty J., Jeffrey P. Kaplan, and Gregory Ward 2001 Open propositions and epistemic would. Paper presented at the Annual Meeting of the Linguistic Society of America, January. Birner, Betty J., and Gregory Ward 1998 Information Status and Noncanonical Word Order in English. Amsterdam: John Benjamins. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London and Canberra: Croon Helm. Palmer, Frank R. 1990 Modality and the English Modals. 2d ed. London: Longman. Original edition, London: Longman, 1979. 2001 Mood and Modality. 2d ed. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Perkins, Michael R. 1983 Modal Expressions in English. London: Frances Pinter. Prince, Ellen F. 1986 On the syntactic marking of presupposed open propositions. In Papers from the Parasession on Pragmatics and Grammatical Theory, 22nd Meeting of the Chicago Linguistic Society, (ed.), Anne M. Farley, Peter T. Farley, and Karl-Erik McCullough, 208-22. Sweetser, Eve 1982 Root and epistemic modals: causality in two worlds. In Proceedings of the Eighth Annual Meeting of the Berkeley Linguistics Society, (ed.), Monica Macaulay, et al., 484-507. Ward, Gregory 1988 The Semantics and Pragmatics of Preposing. New York: Garland.

Towards a contextual micro-analysis of the nonequivalence of might and could Stephane Gresset

Apart from a few exceptions it has been felt that, with the expenditure of sufficient effort, our knowledge of the world could be made more precise. But more recently there has been a growing awareness that the imprecision of the world might be inherent and that this should therefore be an essential component of any theory. (Jennifer Coates)1



While there is general agreement as to the distinction between the epistemic uses of might and the non-epistemic or root uses of could (ability, permission and root possibility), most authors also provide examples of an epistemic use of could, therefore raising the question of the nature of the difference between might and could in their so-called epistemic uses. Leech (1987: 120-121) suggests that it is difficult to see any difference between might and could in examples such as There could be trouble at the Springboks match tomorrow/The door might be locked already/Our team might still win the race. According to Quirk et al. (1985: 233), in There could be something wrong with the light switch and Of course, I might be wrong, "could and might have the same meaning and both express the epistemic possibility associated with may". Biber et al. (1999: 491-493) consider that "in academic prose, could, may and might usually express logical possibility", that "could and might are much more common expressing logical possibility than permission or ability", and that "in contrast to the typical functions of can, the modal could usually marks logical possibility in conversation, expressing a greater degree of uncertainty or tentativeness: That could be her/It could be anything you choose." Going over some of the exponents of epistemic possibility in English, Coates (1995: 58, 63) mentions an example of could ("The only snag is that it has been raining... and I could get held up for anything up to a week"), which, she comments, is evidence that"could is making headway as an alternative to might in the expression of (epistemic) tentative (meaning)".


Stephane Gresset

The main purpose of this paper is to discuss the position expressed by Coates (1983: 167) in the main clause of the following sentence: It does not seem implausible to suggest that, while might is becoming the main exponent of Epistemic possibility in every day spoken language and no longer expresses a more tentative meaning but is in most contexts synonymous with may2, could is filling the gap left by might and is the new exponent of tentative Epistemic possibility. This quotation calls for two remarks. The suggestion made by Coates is very tentative ("It does not seem implausible to suggest..."), probably because it is based on only 22 occurrences that she came across in her corpus (15 in the Lancaster corpus and 7 in the Survey) for her (1983) book. She adds, however, that epistemic could is "still relatively infrequent", therefore predicting that it will gain ground, which the authors just quoted seem to confirm, and particularly Biber et al. in their recent Longman Grammar of Spoken and Written English (1999: figure 6.12, 491). The position that we would like to defend is that, although could is used more and more frequently in apparently epistemic or epistemically-oriented contexts and may therefore receive an epistemic interpretation of sorts3, this does not mean that might and could are strictly synonymous, nor that their difference can be accounted for only in terms of the degree of epistemicity. It does not follow from the observation that two modals are interchangeable in a specific context that they are strictly equivalent. This simply means that what difference there is between them does not affect the meaning of the utterance to the point that it means something altogether different and may therefore cause misunderstanding. Moreover, whether one admits to the existence of an epistemic could (as many linguists do - see above) or not (as those might prefer who consider for instance the incompatibility of a "could or could not" sequence with an epistemic interpretation to be important) is less relevant to us than whether epistemic might and epistemic could are epistemic in the same way and for the same reasons. Recent research on the history of modals in English as well as in many unrelated languages and research on grammaticalization have shown that root modality and epistemic modality are related, and that epistemic modal meanings can be derived from root meanings [see Traugott (1989: 43), Sweetser (1990: 56-68), Heine (1993 : 66-69), Groussier (2000a: 95-98)]. The following hypotheses have been put forward by several authors:

The non-equivalence of might and could 1.

2. 3.


If could is replacing might as the main exponent of tentative epistemic possibility, as Coates puts it (see above), then it would follow that could and might are to be considered as more or less interchangeable. It is a commonly held position that could and might are stylistic variants of each other, which, to us, is another way of saying that they are interchangeable apart from considerations of formality. For some authors, could expresses a higher degree of probability (Larreya 1996: 29) and to others, a more tentative epistemic meaning (Coates above) than might, "tentative" meaning "more cautious", "hesitant", and consequently "less probable".

These propositions may seem to be contradictory but a very detailed analysis of the respective contextual environments of might and could shows that they are not, and this is what we are now going to undertake, bearing in mind the basic structuralist principle that any difference in form is meaningful. We hold that, however close the interpretations of might and could may be in some contexts, there always remains a difference of some kind. The linguist's task is, therefore, to bring to light, through micro-analysis, any contextual distinctive traces likely to indicate that the opposition between the two linguistic markers being studied has never been completely neutralized. While the examples analyzed in section 2 quite clearly illustrate the difference between might and could, the distinction is apparently lost or at least blurred in those examined in section 3.


A case in point: all the uses of might and could in a single text

The following article, published in the New York Times a few days before GW Bush's election, is exclusively devoted to a comparison of the possible Bush and Gore administrations, as the opening sentence makes it very clear: "Based on a few tips and heavy thumb-sucking, here is a way to compare possible Bush and Gore administrations." In other words the journalist, the well-known columnist William Safire, is indulging in speculation, and although obviously none of it is entirely groundless, he nevertheless expresses his lack of certainty as to the truth of the propositions he is putting forward, i.e. the validation of the predicative relation. We, apparently, therefore have here an occurrence of epistemic modality, which is traditionally defined as follows: Epistemic modality is concerned with the speaker's assumptions {must, should and ought) or assessment of possibilities (may, might, could and


Stephane Gresset will) and, in most cases, it indicates the speaker's confidence (or lack of confidence) in the truth of the proposition expressed. (...) (It) expresses the speaker's reservations about asserting the truth of the proposition. (Coates 1983: 18-20) It establishes a relation between the utterer and a propositional content as represented in the predicative relationship. (...) The particular value of this modality is to express the utterer's lack of certainty regarding the validation of the predicative relation (Bouscaren, Chuquet, and Danon-Boileau 1992: 40).




WASHINGTON - Based on a few tips and heavy thumb-sucking, here is a way to compare possible Bush and Gore administrations. For secretary of state, Al Gore is expected to choose Richard Holbrooke, the current UN delegate. George W. Bush has hinted that Colin Powell would be his choice. But what if General Powell prefers Health and Human Services, where his current interest lies? Then the Bush SecState would probably be Condolezza Rice, former provost at Stanford. If Mr. Bush thought Ms. Rice better suited to be his national security adviser, he might (m 1) pick Senator Richard Lugar. Meanwhile, Mr. Gore would back up Mr. Holbrooke at State with his national security adviser, Leon Fuerth. At Defense, Mr. Gore might (m 2) well sound out the Republican John McCain; he would probably decline, looking toward a presidential race in 2004. Mr. Gore might (m 3) then turn to Representative John Murthua of Pennsylvania or former Senator Sam Nunn. Mr. Bush might (m 4) also choose Mr. Nunn, a Democrat, or turn to Paul Wolfowitz or Paul Ο 'Neil. If Mr. Gore chose the former Texas governor Ann Richards for HHS, he could (c 1) then break Clinton tradition and choose a male attorney general: Jack (Tell 'Em 'Nuthin') Quinn, a former White House counsel. George W. 's first choice to head Bush Justice would be John Dcmforth, the former Missouri senator. Treasury Secretary Larry Summers, hoping to be a Gore holdover, may think his competition is the current Fannie Mae chief, James Johnson. A more intriguing Gore choice would be Bill Bradley, who showed loyalty and has Senate Finance credentials. For Mr. Bush, Paine Webber's Don Marron has the inside track, although Representative Jennifer Dunn of Washington is in the running. Mr. Gore's chief economic adviser would be Alan Blinder, and Mr Bush's is Lawrence Lindsay. Both are former Federal Reservists.

The non-equivalence of might and could


At Agriculture, Mr. Gore could (c 2) fend off a possible conservative Democratic revolt in the choice of House Speaker, if Democrats win a majority, by picking the Texan Charles Stenholm. Mr. Bush will look toward Governor Frank Keating of Oklahoma and Senator Pat Roberts of Kansas. At Education, Mr. Gore would be torn between Carolina's governor, Jim Hunt, and Representative Jesse Jackson Jr. of Illinois, whose father leads the liberal charge against Ralph Nader's surprising uprising. Mr. Bush could (c 3) reach across party lines to pick the Democrat Gerhard Casper, Stanford's ex-president, or stick with Lynne Cheney, his vice-president's better campaign half. Dick Cheney would have a cabinet pick, probably former Senator Alan Simpson of Wyoming, for Interior. Mr. Gore's Interior choice to tiptoe between development and environmental interests could (c 4) be Representative Tom Udall of New Mexico (Stu's son and Mo's nephew) or Washington's governor, Gary Locke, a ChineseAmerican. Commerce (or trade representative) for Mr. Bush -former Governor Carroll Campbell of South Carolina, and Robert Hormats. For Mr. Gore: Tom Downey. Energy might (m 5) see a Clinton holdover in Bill Richardson, a Hispanic. Mr. Bush's energy secretary may be Representative J. C. Watts of Oklahoma. At the CIA, with George Tenet becoming a diplomat, Mr. Gore could (c 5) go for Warren Rudman; Mr. Bush could (c 6) switch the FBI director, Louis Freeh, to CIA and replace him with the prosecutor Charles La Bella. (International Herald Tribune, 27/10/2000) That most modals (19 out of 23) are used in their past forms should come as no surprise: they are all modal past forms, referring to the unreal, some of them because they are located relative to an explicit or implicit //^clause, but most of them simply because of the speculative nature of the article. However, there are as many occurrences of could (6) as of might (5), in seemingly identical environments, as shown in the comparison between might I and could I, or might 2, could 2 and could 5. In other words, it is extremely unlikely that a native speaker would rule out might where could is used and vice versa* Might and could are then, technically speaking, interchangeable or quasi-equivalent and we could readily apply to could and might what Bolinger once wrote about which and that as relative pronouns: "the choice between them is so rarely a matter of life and death".


Stephane Gresset

Are might and could in "free variation" here? Could two different forms express exactly the same meaning? And should this be so, is it possible to account for the neutralization of a might/could opposition that is very clear in other contexts? It is much more rewarding, from a scientific point of view, to posit that there may, in the modal's environment, be traces or linguistic markers likely to pave the way for one form rather than for the other. More precisely, considering that our aim is to highlight the fact that could and might are not used in the same environments, we propose to focus on could, the epistemic interpretation of which is much more debatable than that of might. Let us therefore have a close look at the immediate context of the occurrences of could: could 1 is immediately followed by "then break Clinton tradition and choose...", could 2 by "fend off a possible conservative Democratic revolt in the choice of..." and could 3 by "reach across party lines to pick...". Could 4 is preceded by "Mr. Gore's Interior choice to tiptoe between development and environmental interests" and could 5, although not as conspicuous, by "At the CIA, with George Tenet becoming a diplomat, Mr. Gore...". What first springs to mind is that none of the occurrences of might has an immediate context of a similar nature. Might 1 may be preceded by an //^clause exactly like could 1, yet there is no explanatory comment such as "break Clinton tradition and...". The second, third and fourth occurrences of might are followed by adverbs ("well, also, then") which we shall come back to, and the last one {might 5) by "a Clinton holdover in...", which, although it might be taken as some sort of authorial comment, is presented here as a simple fact: "it just so happens that Bill Richardson, appointed by Clinton (as Energy Secretary), would keep his position under Gore". The reason behind the choice of might instead of may, which is used in the sentence that immediately follows ("Mr. Bush's energy secretary may be Representative..."), although not the subject of this paper, would be worth investigating here: let us simply mention that, had Gore been elected, he would most probably have preferred to "break Clinton tradition" rather than appoint "Clinton holdovers" - he has actually been severely criticized for that. Might 5 suggests that there is something unexpected, from what we know about Gore, in his choice of a Clinton holdover: the opposite would have been expected, the negative value 'not see/pick up Clinton holdovers' therefore being presented as predominant from the utterer's point of view, which apparently espouses Gore's point of view. In other words, the choices referred to by way of might are based mainly on "heavy thumb-sucking" and are neither commented on nor justified. The utterer is very careful, since he has no other evidence, and puts forward

The non-equivalence ofmight and could


hypotheses while explicitly presenting them as some options among other possible ones that can also be taken into account. See might 2, 3 and 4: might 2: "At defense, Mr. Gore might well sound out. might 3: "Mr. Gore might then turn to Representative..." might 4: "Mr. Bush might also choose..." The presence of well beside might 2 is not enough evidence in favour of might, since could well is perfectly acceptable too. However, "he would probably decline", which immediately follows the sentence with might, clearly expresses that the hypothesis put forward here ("Mr Gore - sound out the Republican John McCain") is unlikely. Now, could well is generally not used in this case, precisely because another option must then be envisaged (see "Mr Gore might then turn to..."). The relation established by then in the might 3 sentence is very similar to that established by i f . "however, if John McCain declined, then Gore might turn to X or Y". In other words, then, which is also used with could 1, does not seem to play a decisive part in the might/could distinction. As for also in the might 4 sentence, it bears on the subject and not on the object of the verb ("Mr Bush too - like Mr Gore - might choose...") and so has no incidence on what concerns us here. On the other hand, all the underlined passages accompanying could in this article serve the same purpose: they participate in the choice that the modal could expresses by referring to what is implicitly presented as being desirable to the subject of the sentence, via the utterer. It is indeed striking that all of these comments express attitudes that the subject referent obviously favours so that what is at stake is not whether things are going to turn out that way or not but whether the subject referent actually has the means to reach that end or not, whether or not he is in a position to do so. This is clearly a case of characteristics or properties of the subject, that is to say of root modality. It is perfectly in keeping with Gore's personality that he should be willing to "break Clinton tradition" (could 1), "fend off a possible conservative Democratic revolt" (could 2) or "tiptoe between development and environmental interests" (could 4), exactly in the same way that Bush would willingly "reach across party lines" (could 3). It would be somewhat strange, on the other hand, to substitute might for could in "At Agriculture, Mr. Gore might fend off a possible conservative Democratic revolt in the choice of...". It would suggest that, since the utterer is not confident as to the truth of that proposition, it may not be part of the subject referent's


Stephane Gresset

wishes, and we could then wonder whether Gore actually finds it desirable or not and consequently why he would make such a choice. In other words, with could as well as with might, the possibility that things might turn out differently cannot of course be ruled out, since they both have to do with possibility. Yet this is precisely what might indicates (p, the value associated with the modal, is taken into account, but so is explicitly p', i.e. other than p,5 may/might basically referring to bi-lateral possibility, both ρ and p' are explicitly taken into account), whereas with could, p', other than p, while not ruled out, is not explicitly taken into account {could, like can, basically refers to uni-lateral possibility, ρ only is explicitly taken into account and p' is ignored).6 We hold that there is a link between a choice that derives from the subject referent's properties or characteristics and unilateral possibility. Once the choice that is being made is justified in one way or another, then the possibility that things might be different, even if it cannot be ruled out, is less important, if important at all, and in this way may very well not be explicitly taken into account. Could, then, is used here to refer to a hypothesis which the utterer makes because he has reasons for doing so. And it is to this extent that the epistemic interpretation (which is different from an epistemic "meaning" or "value"), of could can be said to be based on the basic root meaning of can. Indeed the passages in bold type all refer to characteristics of the subject. In other words, the basic meaning of can thus accounts for the so-called epistemic interpretation of could in examples like these. The article as a whole actually makes a very consistent use of could and might, the latter being used for purely speculative hypotheses. The contrast between the last two paragraphs provides another very telling illustration. The second last (with might and may) is clearly devoted to speculations, pure (with may) or not (with might), while the last paragraph (with two occurrences of could) gives some details, though less explicit than with the preceding examples of could, as to the reasons why X or Y could be picked - the musical chairs game provides some justification of the choice put forward. We consequently come to a very simple and possibly simplistic conclusion: could and might are not synonymous. The only way to make our point clear, however, is to illustrate it with a detailed analysis of contextualized occurrences. The objective of the following section is to show that in might-could "sequences" (i.e. examples in which might and could follow each other in the same sentence or in two successive sentences with identical or similar predicative relations), the order in which they appear is significant.

The non-equivalence of might and could 3.


Might or could, or might and could? A few might-could sequences

Gilbert (1987: 160) quotes example (2) to provide evidence that might and could are interchangeable or quasi-equivalent. Indeed, the could-might sequence is prima facie acceptable. (2)

It was the last task remaining to a messy business, which might have ended better but could also have turned out worse.

We would argue that the order in which the modals appear in that sentence is not indifferent and is related to the nature of the predicative relations respectively associated with might and could. Indeed, a "messy" business is not expected to end well, even though of course a good ending, and for that matter a "better" ending, cannot be entirely ruled out. To put it differently, a better ending is clearly one possibility among others, and obviously not the most likely one. Hence the use of might, which explicitly means that there are other possible descriptions than that associated with the modal. Now the second part of the relative clause introduced by which starts with the adversative but and contains the adverb also, both words marking differentiation or otherness, therefore paving the way for a different option than that mentioned in the first part of the clause. In other words the 'turn out worse' option is presented here by the utterer as rather more expected than the first option: a messy business is, by definition, likely to end badly, not to say very badly - we have here a property or characteristic of the subject. Even though the opposite order is acceptable, it would undoubtedly produce a pragmatically different utterance: there would probably be more disappointment on the part of the utterer, as if he or she expected, for some reason, a rather good ending, however messy the business was. Could, so it seems, goes here with what is expected in the sense of justified, easily accounted for, whereas might goes with what is unexpected in the sense of what cannot be ruled out although it is only one option among others. Example (3) is a very similar example, quoted by Larreya (1996: 29), who also considers that could more or less neutralizes the opposition between material (or root) possibility and logical (or epistemic) possibility. In some contexts (such as this), could just seems to be a stylistic variant of might, with very little apparent difference. (3)

Charlie threw what was left of the offending biscuit into the garden. It might have been the biscuit, it could have been the quarrel, but most likely, I thought, it was his rheumatics that were making him so irritable.


Stephane Gresset

Here again might and could can indeed be substituted for each other. This, however, does not mean that the two utterances are pragmatically strictly synonymous; might comes first, then comes could and eventually "most likely". There is a movement from the first hypothesis to the third, and the order in which they are set (the biscuit, the quarrel, his rheumatics) most probably corresponds to the utterer's growing conviction: -


the biscuit seems to be mentioned incidentally, following from the "offending biscuit" phrase. Although the first option that comes to mind, and which, as such, must be taken into account, it is understandably a very unlikely one; the quarrel, on the other hand, seems a more relevant, and as such a more probable explanation, in the sense that a quarrel may legitimately explain Charlie's aggressive attitude (irritability may indeed logically derive from a quarrel), even though here again it is probably not the real reason.

In other words, that might should come first and be associated with the most unexpected and unlikely interpretation and that could should come second and be associated with a more relevant explanation is not indifferent at all, which in turn suggests that, should the order be inverted, the general meaning would be quite different. This is all the more interesting as Larreya, along with other linguists, mentions other examples, which to us are not very different, but in which, according to him, the distinction between might and could is much more palpable, could referring to a higher degree of probability than might as the logical possibility it expresses is based on the existence of a concrete material possibility (1996: 29). (4)


At the end Adam admitted to himself that it could be done. Smokey just might, just could have the business back in shape a month from now. (quoted by Larreya 1996 : 29) M: So what's your range at present?... Whitaker, your range? W: (pulling himself together slightly) It must be under fifteen miles. I can't get through to camp. It could be ten. It might be less. (W. Hall, The Long and the Short and the Tall)

In fact there is no reason to distinguish between these examples and examples (2) and (3). That the difference should be more or less perceptible or marked essentially has to do with the predicative relation associated with each modal, i.e. with the context. In other words, the higher degree of probability interpretation is just what it is, an interpretation, which is

The non-equivalence of might and could


perfectly compatible with the basic meaning of could but in no way an inherent feature. It may or may not surface depending on the environment. What makes this possible, however, is just the fact that, basically, could, as a form of can, does not explicitly take into account the possibility that things might turn out differently, hence the positive bias that often goes with it - as opposed to the negative bias of might. Example (6) really speaks for itself and is all the more remarkable as the stage directions add to the evidence. As a matter of fact, this passage from a play by Pinter is an outstanding metalinguistic commentary on the difference between might and could:7 (6)



(...) He examines the bedclothes. I thought these sheets didn 't look too bright. I thought they ponged a bit. I was too tired to notice when I got in this morning. Eh, that's taking a bit of liberty, isn 't it? I don't want to share my bed-sheets. I told you things were going down the drain. I mean, we've always had clean sheets laid on till now. I've noticed it. How do you know those sheets weren 't clean? What do you mean ? How do you know they weren't clean? You've spent the whole day in them, haven't you? What, you mean it might be my pong? (Re sniffs sheets.,) Yes. (He sits slowly on bed.,) It could be my pong, I suppose. It's difficult to tell. I don't really know what I pong like, that's the trouble. (H. Pinter, The Dumb Waiter)

Now, although it cannot be stated as a rule, it must be acknowledged that the might-could sequence is fairly regular. Examples (7) and (8) all bring evidence in support of the following claims: the might-could order corresponds to a specific strategy and the root/epistemic distinction cannot by itself account for the difference between might and could, while the uni-lateral/bi-lateral distinction can, and is in fact more appropriate. (7)

(...) This characteristic structure (= the commentary use of the Present) is reflected in the possibility of using the non-Progressive present to refer to complete actions, i.e. to maintain an aspect distinction just as one would in the Past tense, the only difference being that here the maintenance of the distinction is optional, since as in the Present Tense generally the Progressive can be usedfor all nonhabitual actions. Thus a film commentary might run:



Stephane Gresset "Now the villain seizes the heroine, now they drive towards the railway track, now he forces her out of the track, now he ties her to the track, while all the time the train is getting nearer.'" (All verb forms except the last are non-Progressive). Equally, the commentary could have been given throughout in the Progressive form: "Now the villain is seizing the heroine, now they're driving off (...)" One factor influencing the choice between the two possibilities is that the non-Progressive is favoured when a rapid series of events has to be commented on as they are happening. (B. Comrie, Aspect) Research data on ISD is still skimpy, since it was identified as a clinical entity only in the past decade. In their landmark 1966 book, "Human Sexual Response", William Masters and Virginia Johnson had postulated there were four phases of response in lovemaking: excitement, plateau, orgasm and resolution. But in 1979 Kaplan divided human sexual response into the three stages of orgasm, excitement and desire, which are each subject to separate problems. Some people, for instance, might feel desire for sexual intercourse without being sufficiently aroused to carry it out; others could be aroused to the point of orgasm, without feeling any real desire or pleasure. "Patients describe such experiences as similar to eating a meal when one is not really hungry," wrote Kaplan. (Newsweek, 26/10/1987)

Now follows an example of the opposite sequence, if only to make it clear that we are not dealing here with any kind of constraint, whether absolute or relative, but with, at best, affinities between a type of sequence and some line of reasoning. The could-might sequence in example (9) simply corresponds to a different approach or pragmatic strategy: (9)




Patients most often seek acupuncture to relieve painful conditions that fail to go away on their own, that do not respond well to standard treatments or that may require more hazardous remedies, like surgery or the long-term use of potent drugs. An independent panel evaluating the reports concluded that acupuncture could alleviate acute pain - for example, postoperative pain - and might also help control chronic pain, like chronic migraines, neck pain, muscle pain and osteoarthritis of the knee. The panel also found acupuncture to be highly effective in combating nausea caused by pregnancy, anesthesia or cancer chemotherapy. (International Herald Tribune, 20/11/1997)

The non-equivalence ofmight and could


Let us first note that the present/past distinction between may/might and can/could is here neutralized because the modals are part of a subordinate clause depending on the past verbal form "concluded" - compare could and might with may in the preceding independent clause. This, however, does not affect our reasoning. We would posit that what comes first here, i.e. "alleviate acute pain" is presented as based on more scientific evidence than what comes second ("help control chronic pain"), basically on the grounds that the first predicative relation is associated with could: "alleviate acute pain" is a property or an inherent characteristic of the subject "acupuncture" (capacity, ability). In the second part in might, also, like equally, announces a different example, and help (control) tends to show, as mentioned above, that the possibility expressed here may be based on less evidence than the first one. But the conjunction and also suggests, together with also, that "help control chronic pain" is simply a possibility which, although it cannot be ruled out, is less relevant to the utterer. In other words, in the might... but... could sequences that were analyzed in the examples above, might paves the way, as we put it, for something else, for a different possibility to come, introduced by but. On the other hand, could, in the same kind of contexts, does not call for anything else, and is therefore more autonomous. Whether anything should follow or not is a different matter, once it has been said that this feature is not part of the definition of could. Here again, this directly derives from the basic difference between might and could, that is to say bilateral versus unilateral possibility. And, in spite of appearances to the contraiy, the following example does not contradict the hypotheses we have just formulated: (10)




But doctors emphasized that the new drugs would not be a substitute for a flu shot, which would remain the most important defense against the disease. Instead, they might be useful for warding off the flu in those who fail to get the shot and to those who fail to respond to the shot. They also could help in years when shots are not highly effective against a particular strain. (International Herald Tribune, 26-27/9/1998) We could indeed object that there is no actual contrast here between the sentence in might, which mentions one possible beneficial effect of the new drugs, and the sentence in could, which, as also indicates, mentions another


Stephane Gresset

one. As a matter of fact, if but could be added in front of the second sentence, so could and, which means that, contrary to the examples above, the second possibility is not defined in opposition to the first one. The link between those two sentences is much weaker than that between the first two sentences, marked by "instead", which literally means "in the place of', and which we could replace by "yet" or "however". In other words "be useful for warding off the flu in..." is explicitly defined as one of several possible beneficial effects of the new drugs other than "be a substitute for a flu shot". Might here relates to what comes before, to the left, whereas the link established by could is looser and could be rendered as "Besides, they could help in years when...". The line of thought is very different from that expressed in the preceding utterances and in this passage might and could refer to altogether different discursive strategies.


Conclusions and perspectives

We hope to have shown that, however subtle the difference between might and could in the examples analyzed, some contextual features can generally be highlighted, which, if they never require one modal or the other, pave the way for (or actually have) definite affinities with one or the other. And from what we can judge, these contextual elements can always be related to the basic meanings of can and may, that is to say unilateral versus bilateral possibility. In addition to the potential of this contrast outlined in the present paper, we believe that this presentation can account for both the distinctive and the common interpretations of can/could and may/might, and as far as we are concerned here, for the presence or absence of epistemic interpretations of may/might, can, can't and could (cf. also Gresset 1999,2001). Indeed, the basic distinction we posit explains why can cannot have an epistemic reading and why can't may have one.8 As for coulcf, our hypothesis, which is basically consistent with that submitted by Larreya (this volume), is the following: the presence, in might as well as in could for that matter, of the marker -ed is due to the preconstruction of the path to the exterior of the domain or non-p as being either the case in the situation of utterance, as with counterfactuals, or preponderant from the utterer's point of view. Consequently, the only way for the utterer to consider possibility, that is, at some point, occurrences of ρ as well as occurrences of non-p, is to locate can or may relative to a point of reference disconnected from the situation of utterance, which is precisely the role played by -ed.

The non-equivalence of might and could


However, since may already refers to both ρ and non-p, the presence of -ed in might simply alters the epistemic interpretation of may in a way which explains why might may have become the main exponent of epistemic possibility (see note 2). The so-called "epistemic" interpretation of positive could, on the other hand, results from the combination of can with -ed. Gilbert (1987: 343) states that "It is the presence of -ed which enables could to receive a bilateral possibility or contingency interpretation" [see also Cotte (1988: 907)]. The process leading up to the epistemic interpretation of could is therefore different from that marked by might and this is what we have been trying to illustrate here. Might and could are not in free variation and their difference is not a matter of pure formality. It has to do with the basic difference between can and may, which explains most interpretative effects generally associated with might and could, from the negative versus positive bias to the more or less tentative degree of possibility expressed. We believe, with Gilbert (1987: 343-345) and Cotte (1988: 909-916)10, that the epistemic interpretation of bilateral possibility of could results from the combination of the basic root meaning of unilateral possibility of can with the role played by the preterite marker -ed. We agree with Coates (1980: 215) when she describes, from a purely synchronic point of view, the use that speakers make of can and may in everyday speech: The stereotype for adult speakers of English is that may expresses epistemic possibility and can expresses root possibility and they are not considered to be related. This does not in any way contradict the historical or the pragmatic and semantic links that we have underlined above, since this observation is made from an altogether different perspective; speakers of English simply do not perceive can and may as related because they feel that they use them quite differently. And based on this observation two important claims can be made. First, we believe that epistemic possibility is essentially bilateral while root possibility is essentially unilateral, and for us, permission, for example, which can be expressed by may or by can, is no exception. Secondly, what Coates states about may and can seems to us to apply to might and could as well. Even though "Epistemic could is semantically quite distinct from all other uses of could", as Coates (1983: 167) rightly puts it, epistemic could can perfectly well, like all the other uses of could, be derived from the basic root meaning of unilateral possibility that can and could have in common. While might is purely epistemic, because it refers fundamentally to bilateral possibility, could may take an apparent epistemic


Stephane Gresset

interpretation in the relevant contexts, i.e. in epistemically oriented ones. Could, then, is not purely epistemic and might and could are not strictly synonymous. Finally, Hoye (1997: 88-93) concludes from his examination of the adverbs with which could is combined that it is definitely not a central epistemic modal and that "there is little evidence (of co-occurrence) in the data that could is making any significant inroads into the epistemic territory of might." Rather than challenging our main claim that could and might in their so-called epistemic interpretations are not interchangeable, it confirms in fact that their respective epistemic uses are of a different nature and of different origins. The epistemic uses of could may therefore be more frequent, but not necessarily at the expense of might, since the contexts in which they appear are not exactly the same. Actually, the nature of the texts in which such uses of could occur or the intersubjective relations at play would require further investigation. So would the relevance of a binary root/epistemic distinction, since the other objective of this paper was to illustrate that, while two modals have a different basic meaning, one of them may, in the appropriate context, acquire interpretative effects generally associated with the basic meaning of the other modal.

Notes 1. 2.



Quotation from Coates (1983: 11). The bold lettering, underlining and italics in all the examples and quotations are ours. We will not, for lack of space, deal here with the may/might distinction, which we examine at length in our PhD (1999: chapter 6, 260-381). We support the idea that, although might may have become the main exponent of epistemic possibility in every day spoken English, it is generally not synonymous with may, as the following example, borrowed from Charreyre (1984: 55-56) illustrates: in Just look at the sky! There might be a storm soon, uttered in front of a threatening sky, it would be contextually less appropriate, Charreyre suggests, to use may instead of might, while will would be perfectly acceptable. May would convey very little information and simply express the logical, theoretical and objective possibility that such a sky is likely to bring about a storm, which is obvious. Might, on the other hand, carries some appreciation on the part of the speaker that, although a storm may not be wished for, yet it must not be ruled out. In this presentation, the "interpretation" of a modal auxiliary in a specific context is to be distinguished from its "basic meaning" or "value", which all of its uses have in common. Let it be said that, in this respect, statistics from exercises requiring speakers to fill in blanks would in our opinion not be of much interest.

The non-equivalence of might and could



Other than p, when referring to modals expressing possibility, can actually be reduced to not-p. 6. See Culioli (1990) and Groussier (2000b) for a detailed presentation of the theoretical framework we work in, Antoine Culioli's theory of enunciative operations, and Deschamps (2001) for an enunciative approach to the English modals, which our presentation is in keeping with. 7. The might/could contrast in this dialogue is analyzed in more detail in Gresset (2001:211-214). 8. See Deschamps (2001: 17-20). 9. We naturally disregard here the temporal uses of could, which can be referred back to can. 10. Gilbert (1987: 343-345) explains that where could seems to express bilateral possibility and therefore acquires a contingency interpretation, it does not refer directly to equipossibility and is in fact half-way between root modality and epistemic modality. Similarly, Cotte (1988: 909-916) states that could constructs bilateral possibility on material possibility as it is used when the hypothesis expressed by could is compatible with what the speaker knows about reality. Interestingly, Cotte adds that might may also be used for a hypothesis that is founded but that this is not part of its definition, since with might the hypothesis may just be pure speculation.

References Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 Longman Grammar of Spoken and Written English. London: Longman. Bouscaren, Janine, Jean Chuquet, and Laurent Danon-Boileau 1992 Introduction to a Linguistic Grammar of English. An UttererCentered Approach, translated and adapted by R. Flintham and J. Bouscaren. Gap: Ophrys. Bybee Joan, and Suzanne Fleischman (eds.) 1995 Modality in Grammar and Discourse. (Typological Studies in Language 32.) Amsterdam/Philadephia: Johns Benjamins. Charreyre, Claude 1984 Quand MIGHT (HAVE -EN) peut se traduire par "c 'etait comme si", Cahiers Charles V, Universite Paris VII, 6: 27-58. Coates, Jennifer 1980 On the non-equivalence of MAY and CAN, Lingua, 50: 209-220. 1983 The Semantics of the Modal Auxiliaries. London and Canberra: Croom Helm. 1995 The expression of root and epistemic possibility in English. In Modality in Grammar and Discourse, Joan Bybee and Suzanne


Stephane Gresset Fleischman (eds.), 55-66. (Typological Studies in Language 32.) Amsterdam/Philadephia: Johns Benjamins.

Cotte, Pierre 1988 Le Systeme des auxiliaires modaux dans le systeme verbal de l'anglais contemporain, Thöse de Doctorat d'Etat, Universitd de Grenoble III. Culioli, Antoine 1990 Pour une Linguistique de l'Enonciation. Operations et Representations. Gap: Ophrys. Deschamps, Alain 2001 Approche 6nonciative des modaux de l'anglais, Cahiers de Recherche T. 8. Modalite et Operations Enonciatives, collection dirigöe par Janine Bouscaren, 3-21. Gap: Ophrys. Gilbert, Eric 1987 MAY, MUST, CAN et les opörations önonciatives. Cahiers de Recherche en Grammaire Anglaise T. 3. Gap: Ophrys. Gresset, Stephane 1999 CAN/MAY, MAY/MIGHT et MIGHT/COULD, ou l'interchangeabilite en question, Ph.D. dissertation, Universite de Paris X-Nanterre. 2001 CAN/MAY et MIGHT/COULD ou l'interchangeabilite ä l'Spreuve des textes, Cahiers de Recherche. T. 8. Modalite et Operations Enonciatives, collection dirigie par J. Bouscaren, 177-222. Gap: Ophrys. Groussier, Marie-Line 2000a Subjectivisation croissante de la valeur des modaux au cours de l'histoire de l'anglais. In La Modalite et les Modaux en Diachronie et en Synchronie (Domaine Anglais), Jean Pauchard (ed), 73-101. Reims: Presses Universitaires de Reims. 2000b On Antoine Culioli's theory of enunciative operations, Lingua 110: 157-182. Heine, Bernd 1993 Auxiliaries: Cognitive Forces and Grammaticalisation, Oxford: Oxford University Press. Hoye, Leo 1997 Adverbs and Modality in English. (English Language Series.) London: Longman. Larreya, Paul 1984 Le Possible et le Necessaire. Modalites et Auxiliaires Modaux en Anglais Britannique. Paris: Nathan. 1996 Modalites et Auxiliaires Modaux en Anglais. Cours d'Agregation. Pau: Universite de Pau. Leech, Geoffrey N. 1987 Meaning and the English Verb. 2d ed. London: Longman. Original edition, Harlow: Longman, 1971.

The non-equivalence of might and could


Leech, Geoffrey N., and Jennifer Coates 1979 Semantic indeterminacy and the modals. In Studies in English Linguistics, Sydney Greenbaum, Geoffrey Leech and Jan Svartvik (eds.), 79-90. London: Longman. Palmer, Frank R. 1979 Modality and the English Modals. London: Longman. 2001 Mood and Modality. 2d ed. Cambridge: Cambridge University Press. Original edition, Cambridge: Cambridge University Press, 1986. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Sweetser, Eve E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure (Cambridge Studies in Linguistics). Cambridge: Cambridge University Press . Traugott, Elizabeth C. 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65 (1): 31-55.

The status of emerging modal items

On two distinct uses of go as a conjoined marker of evaluative modality1 Philippe Bourdin



Unbeknownst, it seems, to most reference grammars of English, be going (to) is not the only construction in which the verb go, rather than referring to motion in space, fulfils an essentially grammatical function. Thus, the following sentences exemplify two uses of go which, though well entrenched in the fabric of the language, are hardly ever mentioned, let alone documented or explained: (1)


The disease usually goes undetected until its later, more dangerous stages. (The Globe and Mail [Canadian daily newspaper], May 5, 1999: A5) There's been people starved to death in Stalingrad, so don 7 you go thinking you're something special. (BNC2, AC5 268)

One reason for this oversight is obvious: in contrast to the be going (to) construction, which encodes a fairly canonical tense/aspect category, the go of goes undetected and that of go thinking do not seem to make the sort of syntactic and notional contribution for which traditional grammar would provide any obvious ready-made label. While it is true that goes undetected has a Passive flavour to it (Kirchner 1951), the very severe restriction illustrated by (3) means that go hardly qualifies as a standard Passive auxiliary: (3)

* The disease usually goes detected at a much later stage.

I will refer to this restriction as the Negativity Constraint on go. The grammatical status of go in (2) is, if anything, even murkier. To be sure, it does pattern like the catenatives start or keep: (4)

a. b.

Don't you keep thinking you're something special. Don't you start thinking you 're something special.


Philippe Bourdin

However, the status of aspectual catenatives with regard to auxiliarihood is notoriously elusive (Brinton 1988: 65-73). And of course, it remains to be seen how the go of go thinking should be characterized in semantic, and especially modal and aspectual, terms: as will be seen below, labels like "continuative" or "iterative", useful though they are, fall short of doing justice to it.


Brief morphosyntactic characterization

The go of (1) and the go of (2) have the morphosyntactic trappings of "sub-canonical"3 grammatical markers. Both do perform the task which verbal auxiliaries, crosslinguistically, are expected to discharge: they act as "helping words" for a non-finite verbal form. They do so, however, without really meeting the strictly formal criteria of auxiliarihood - as delineated, among others, by Warner (1993: 3-13, 33-47) and Anderson (1997: 1-6). Thus, contrary to modals, they require do-support in negative sentences, questions and emphatic constructions and they are by no means restricted to finite contexts: (5)


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

... a lot of bullying does go undetected. (BNC, HVD 328) Don't you go thinking you 're something special! * This guy just doesn't can think straight. I'm afraid I can't allow such a mistake to go unpunished. (BNC, BP1 735) I don't want her to go getting any wrong ideas about what's become of me. (BNC, AC5 753) * Their son doesn't seem to can read.

The morphosyntactic issues raised by the go un-V-en construction have been tackled in part by Zimmer (1964: 35-36, 91), Siegel (1973), Hust (1977, 1978), Levin and Rappaport Hovav (1986: 625-631, 649-653) and Cowper (1995: 12-14). Those authors have explored the categorial status of the so-called "Unpassive"4 participle, but they have largely disregarded the marker go. As far as can be ascertained, the go V-ing construction, as instantiated in (2) above, has never been the object of any specific study or monograph - except, though somewhat tangentially, for Spears (1982: 865-866).

On two distinct uses of go 3.


Semantic characterization

The term "sub-canonical" is equally appropriate to a semantic description of the go un-V-en and go V-ing constructions. This is because, in contrast to such prototypical auxiliaries as the modals or even to more formally peripheral markers like be going (to), go here is subject to very severe collocation constraints which are ultimately semantic in nature.


Go un-V-en

The following sentences provide a glimpse into the workings of the Negativity Constraint governing go un-V-en: (7)

(8) (9)

(10) (11)



She 'd probably be touched if a neighbour dropped by with her mail that had gone uncollected for three days while she was bedridden with flu. (The Globe and Mail, Jan. 2, 1999: D7) The crime went without an arrest until [the test for DNA was finally carried out]. (The Globe and Mail, March 10, 1999 A8) As well, public servants, who went years without getting paid under Mr. Mobutu, have been paid only a few times by the new regime... (The Globe and Mail, Nov. 26, 1997: A20) They could simply dust off those reports to Senate that go mislaid. (Email message, June 30, 1997) The increasingly shrill warnings of a new "Red menace" by members of the country's squabbling democratic movement go largely ignored. (Maclean's [Canadian newsmagazine], Nov. 6, 1995: 34) After six terrible days, during which [they] went sleepless for 100 hours, they all made it to Elephant Island...{The New York Times Book Review, Dec. 27, 1998: 11) How does she feel about Mr. Clinton now? "Whole months go by and I don't think about the man", she said. (The Globe and Mail, Nov. 26, 1998: A22)

Mail that goes uncollected, as in (7), is, by definition, mail that fails to go anywhere: clearly, go here cannot possibly denote motion in space. In actual fact, its nearest functional equivalent would be remain, as is almost always the case with go un-V-en,5 Meanwhile, imcollected, the non-finite form to which goes provides tensed support, denotes the state resulting from an absence of process; inasmuch as it is the standard destiny of mail


Philippe Bourdin

to be collected by those to whom it is sent, the semantics of uncollected can be described in terms of counterexpectation and counternormativity.6 If we now look at examples (8) to (13), two important facts become readily apparent. One is that the various types of items and phrases with which go collocates are all characterized by the semantic feature of counternormativity: for instance, without an arrest in (8) refers to the absence of a process which is typically and normally triggered by a crime. Likewise, remaining sleepless after 100 hours, in (12), is a state that definitely goes against physiological norms. Another salient fact is that the un-V-en participle used in (7) is only the most overtly grammaticalized of the various counternormative forms with which go may collocate. This is crucial, because it means that the Negativity Constraint on go un-V-en is indeed of a semantic nature - so much so, in fact, that there is something inevitably reductionist about such convenient labels as "Unpassive auxiliary" or "go un-V-en construction". Perhaps the most telling example in this regard is (13), where the formal relationship between go - or rather go by7 - and the expression I don't think about the man is not at all one of auxiliation, but of coordination. At first glance, the Negativity Constraint would appear to be fully operative in (13). Indeed, (14), which flouts the constraint, is unacceptable: (14)

* Whole months go by and I think about the man.

However example (15),8 which also flouts the Negativity Constraint, is fine: (15)

Whole months go by and I still think about the man.

As is well-known, the adverb still carries a presupposition that the state of affairs being described runs counter to the speaker's expectations: for instance, Mary would not assert that Peter still lives in Hamburg unless there was an expectation that Peter would now live elsewhere. It is plausible to assume that the semantic property of counternormativity or counterexpectation, which is overtly common to (13) and (15), is the feature that licenses the use of go by and in both sentences. On the other hand, sentence (14) is unacceptable not because it flouts the Negativity Constraint, but because it contains no overt indication of counterexpectation regarding the state of affairs described by I think about the man. What these facts suggest is that the Negativity Constraint on go un-V-en is "merely" the formal manifestation of the underlying property of counternormativity/counterexpectation, which constitutes the defining semantic feature of the construction.

On two distinct uses of go 3.2.


Go V-ing

The go V-ing construction is subject to a distinctive semantic constraint of its own, which involves the mandatory animacy and agentivity of the grammatical subject: (16)

(17) (18) (19) (20)


(22) (23) (24)

He came over as something like a wind-up toy that was apt to go bashing itself into the nearest wall without guidance and protection. (BNC, FYY 1539) I find it very disturbing and unwelcome to have you go touching me as you do. (BNC, JYA 1131) It's all shit and blood, so don't go believing any of the rubbish on That Thing. (BNC, F9C 2717) I thought teachers didn 't go drinking in pubs. (BNC, H9D 2962) For the moment the Angevin dominions remained quiet and the new King of France was too busy elsewhere to go looking for trouble. (BNC, EFV 993) - "Andyou... you 're going to ring the police the minute I've gone, aren't you? " ... - "Now don't go flying off the handle." (BNC, HNJ 1429) Where did she go drinking out of glasses? (BNC, KBE 27) Sitting there now is okay, but if you go sitting there past midnight, you 're going to get picked up by the police. (A. Spears, 1982: 865) If you 're the little one, don't go bringing your big brother round here. (BNC, FS9 2440)

Although it is seemingly a counterexample insofar as the grammatical subject refers to an inanimate entity, (16) actually bears out the Animacy and Agentivity Constraint: the acceptability of the sentence would be in grievous jeopardy were it not for the agentive reading which the reflexive predicate bash itself confers on a wind-up toy. The Animacy and Agentivity Constraint is not without some important consequences. Thus, it is frequently the case that the go of go V-ing is ambiguous between a motional and a non-motional reading. There are also instances, such as examples (19) and (20), where both meanings are arguably coexistent or "blended", with varying degrees of fading of the motional sense. Intuitively, go drinking in (19) and go looking in (20) sit somewhere on the cline bounded at one end by purely motional phrases like go hunting and go skating, which fall outside the purview of the present study, and at the other end by phrases such as go believing in (18), go sitting in (23) and go bringing in (24), where the semantics of the V-ing


Philippe Bourdin

form precludes a motional reading of go. The very existence of such a cline is a defining property of the go V-ing construction. In contrast, the go un-V-en construction very rarely tolerates even the possibility of a two-way interpretation of go or of a "blended" reading. When we look through examples (16) to (24), two other features of the go V-ing construction stand out. It is frequently the case that the grammatical subject refers to either of the speech participants, and it is also often the case that the context is non-assertive and/or negative. The sort of vetative or admonitory contexts exemplified by (18), (21) and (24) are most especially favoured, but questions, as in (22), and hypothetical contexts, as in (23), are also environments in which go V-ing seems to flourish. There is, in short, an interpersonal quality to the go V-ing construction which stands in sharp contrast to what might be called, non-technically, the impersonal quality that attaches to the go un-V-en construction. This pragmatic effect, while echoing the semantic contrast between the animacy/agentivity characteristic of go V-ing and the stativity inherent in go un-V-en, correlates with a sharp discursive differentiation between the two constructions: go V-ing is overwhelmingly to be found in dialogues, whereas go un-V-en is typical of the sort of moralizing discourse characteristic of newspaper editorials.


Some statistical findings

The above observations are admittedly impressionistic and in need of validation through more objective investigation. It has proved useful, in this regard, to conduct a small-scale statistical survey of the British National Corpus. Two samples have been gathered, each made up of tokens of go, goes, went, going and gone in collocation with 42 distinct verbal bases. These appear in the form un-V-en (unacknowledged... unused) in Sample 1, and in the form V-ing (annoying... writing) in Sample 2; in both samples, phrases which would typically receive a motional interpretation (go unaccompanied, go driving, etc) have been systematically excluded. While in Sample 2 as many as 41% of the 118 occurrences turn out to instantiate the don't (you) go V-ing pattern, in Sample 1 occurrences in the (positive or negative) imperative are non-existent. This confirms, on the one hand, the pervasive affinity of go V-ing with vetative contexts, and, on the other hand, the complete disaffinity between go un-V-en and imperative, i.e. prototypically interpersonal, environments.9 It is equally revealing to compare instantiations of goes and went in both samples: these account for 7.3% and 27.6%, respectively, of the total number of

On two distinct uses of go


occurrences in Sample 1, but for only 0.8% and 3.3%, respectively, of occurrences in Sample 2. These findings, too, are in keeping with the impressionistic claims made above, because they suggest that go V-ing is markedly less prone than go un-V-en to occur in sentences involving a personal or temporal divorce from the speaker's hic-et-nunc.10

4. 4.1.

The interplay of evaluative modality and aspectuality Go un-V-en

As was seen above, sentences containing the un-V-en participle describe a state of affairs resulting from an absence of process, from what might be called a particular form of inertia in the way things are playing out. Radden (1996: 449-450) has shown that such a state of affairs is prototypically viewed by the speaker as contravening the normal order of things. Concomitantly, there is often a sense that a negative judgement is being passed on it.

4.1.1. Negative evaluation It would seem that go's specific contribution is to convey, or enhance, the negativity of this judgement. Thus, a town council would presumably be at greater risk of jeopardizing its chances of re-election if it chose (25a) rather than (25b) to announce its latest cost-cutting measure: (25)

a. b.

As of next month, the streets will go uncleared during weekends. As of next month, there will be no snow removal during weekends.

Likewise, (26a) would appear to describe the speaker's outrage with more force than (26b) or (26c) do: (26)

a. b. c.

There are just too many crimes in America that go unpunished. There are just too many crimes in America that are unpunished. There are just too many crimes in America that are not punished.

110 Philippe Bourdin These intuitions are confirmed by the contrast between (27a) and (27b), which are both authentic examples: (27)

a. b.

Crime is equated with punishment, but most offences are unpunished. (BNC, CBA 205) The fact that it's so systematic, going effectively unpunished and often performed by people known to the victims makes it more horrific. (BNC, CFC 1999)

In (27a) the speaker adopts the stance of a detached and objective observer, and the predicate are unpunished allows for the sentence to be construed as a factual statement. In (27b), on the other hand, the use of go un-V-en is fully in keeping with that of the intensifier so and the adjective horrific: all three items lay bare the speaker's emotional involvement in the situation. In short, speakers have a certain latitude when it comes to expressing how they feel about the breach of a norm encoded by un-V-en. The choice of go, as against be, frequently implies that they view it as invidious or offensive.11 4.1.2.


In examples like (25a), (26a) and (27b), there is deviation from existing norms as well as persistence of the deviation over a period of time. This is borne out by the fact, mentioned above, that go here has remain as its closest equivalent. It is also confirmed by examples like (28), in which go is modified by the adverbial particle on, which further highlights the protractive quality of the state of affairs: (28)

Many applicants for admission to the United States from Canada are being subjected to detention and extortionate conduct imposed by some power-mad U.S. Immigration and customs inspectors... It is particularly galling that this goes on unchallenged by the Canadian government. (The Globe and Mail, March 18, 1998: A15)

The modal contrast between be un-V-en and go un-V-en invites us to go one step further. Intuitively, an undesirable state of affairs tends to be perceived as all the more invidious as it is protracted. In very many examples containing go un-V-en, substitution by the be un-V-en construction turns out to be awkward or unnatural, precisely because it fails to give the same degree of salience to the element of persistence and protraction:

On two distinct uses of go


(29) a. Our house receives mail addressed to others quite frequently. These things matter: once, a property of ours in England went uninsured for three months; a charge-card account for a large sum incurred interest...(The Globe and Mail, Feb. 9 1999: A12) b. ? Once, a property of ours in England was uninsured for three months. Conversely, be un-V-en is the preferred construction when the notion of persistence is lacking: (30) a. With the sun shining on a warm spring day in the Russian heartland, the people strolling through the streets of Tula are unperturbed by the latest apocalyptic warnings from Boris Yeltsin's election campaign. (The Globe and Mail, May 5 1996: A16) b. ? The people strolling through the streets of Tula go unperturbed by the latest apocalyptic warnings from Boris Yeltsin's election campaign. Be is just too aspectually unstructured, and unstructuring, to yield, all by itself, the value of persistence, and it is precisely for this reason that it fails, as it actually does in (30a), to carry the evaluative weight frequently encoded by go. In other words, it is plausible to assume that the modal difference between be un-V-en and go un-V-en is a function of their differing aspectual profiles.

4.1.3. Aspectual properties The intrinsic aspectuality of go un-V-en is deserving of a much more detailed examination than can be offered here; and so is the relevance of such aspectuality to the modal import of the construction. The following sets of examples are only meant to give a rough idea of the constraints at play. In the following four pairs, the acceptability contrasts between go un-V-en and be un-V-en have little to do with the go and be markers per se. Instead, they reflect the stativity inherent in un-V-en vs. the "actional" import of the passive construction: (31)

a. b.

? They went unarrested in 1995. They were arrested in 1995.

112 Philippe Bourdin (32) (33) (34)

a. b. a. b. a. b.

They went unarrested throughout 1995. ?? They were arrested throughout 1995. The streets went uncleared for weeks. ?? The streets were cleared for weeks. ?? Throughout February, the streets went uncleared every day. Throughout February, the streets were cleared every day.

The contrast between (34a) and (34b) is particularly revealing, because it demonstrates the imperviousness of the un-V-en participle to a segmentation of the interval associated with the state of affairs into discrete units: even though the failure of the process of snow removal to materialize is a daily reality, the state resulting from this failure is not construed as a recurrent situation, but simply as one that extends continuously over time. The following pair pits go un-V-en against be un-V-en:12 (35)

a. b.

They went unarrested until 1995. ?? They were unarrested until 1995.

It was shown above that go un-V-en, in contrast to be un-V-en, highlights persistence and protraction. At the same time, it is unproblematic for go, but less so for be, to appear in an environment that overtly specifies the end of the interval associated with the state of affairs. In other words, inertia does not necessarily imply permanence. Further, permanence is not at all the same as persistence. When a given situation is viewed as persistent, the interval it spans is ipso facto being measured against what the expected or putatively "normal" interval would have been. What is at issue is the length of time between the beginning of the interval and a given moment, which may or may not be its endpoint. In other words, the background assumptions involved in the notion of persistence require two bounded intervals to be compared. Unlike be un-V-en, which favours open-ended intervals, go un-V-en is not at all incompatible with bounded intervals: hence the markedly different abilities of the two constructions, discussed above, to encode persistence. The last set of examples suggests that the aspectual profile of go un-V-en also differs from that of cognate constructions involving a verb of posture: (36)

a. b. c.

At 9 this morning, the streets lay (or: sat) uncleared. ?? At 9 this morning, the streets went uncleared. Until 9 this morning, the streets went uncleared.

On two distinct uses of go


The contrast between (b) and (c) is evidence that go un-V-en favours contexts involving temporal extension over those that focus on a moment in time. The contrast between (b) and (a) shows that go un-V-en does so more vigorously than cognate constructions involving lie or sit. 4.1.4. Persistence vs. change of state One crucial point needs to be made with regard to the conjoining of aspectual persistence and evaluative modality. It has often been remarked that go, like its purported "equivalents" in many other languages, tends to occur in sentences describing a change of state viewed negatively (Clark 1974). The usual suspects are phrases like go mad or go awry. At first sight, these stand as counterexamples to the claim made above: negative evaluation, here, seems to be conjoined with change of state, which is the very opposite of persistence in a state. When we probe a little deeper, however, a rather different picture emerges: (37) (38)


My own country, the United States, eventually will have to go metric. (The Globe and Mail, Sept. 7, 1993: Al7) So Jane Hartrick doesn 't know any sexy women with grey hair... She obviously hasn't met my wife... I agree with Ms Hartrick's husband - go natural, it's sexy. (The Globe and Mail, Dec. 2, 1993: A24) ?? go awful ?? go depressed ?? go sick

In both (37) and (38), go does function as a change-of-state, or "transformative", verb. However, the target state is not at all viewed negatively. The term "target" is being used deliberately, because these constructions usually involve the subject's volition. In this regard, there is something very suggestive about the dubiousness of the phrases in (39). These clearly involve counternormative changes of state, and yet they are hard, on the face of it, to contextualize - unless, perhaps, a situation can be imagined in which someone would deliberately choose to adopt a style that might be termed "awful" or the typical behaviour of a "depressed" person, and so on. In short, it is unwarranted to extrapolate from phrases like go mad and go awry any general rule. There is no empirical evidence at all that the pattern they are supposed to instantiate in contemporary English is productive; no evidence, namely, that go has any particular predisposition to encode uncontrolled change into a state viewed negatively.


Philippe Bourdin


Go V-ing

It has been suggested by Spears (1982: 865-866) that the negative evaluation associated with go V-ing can be characterized as one of "disapproval". Whatever label is chosen, the go V-ing construction, as instantiated in (2) as well as in (16) to (24), typically encodes the speaker's attitude towards an action which she specifically views as deviating from her own personal assumptions or expectations about what is right or desirable. While in (7) it goes against the very nature of mail, as we saw, to remain uncollected, in (2) it does not go against human nature for an individual to think that he is special; rather, it goes against the speaker's personal wishes, beliefs, or code of conduct. To put it succinctly, with go un-V-en the speaker was passing judgement on behalf of society, or even humankind at large, because the norm being contravened tended to be grounded in modes of natural or social organization which are commonly assumed to be of necessary, if not universal, applicability. With go V-ing, on the other hand, we are prototypically dealing with one individual's evaluation of what is right vs. undesirable: hence, an evaluation which tends to be situation-bound, contingent upon particular circumstances, and frequently fraught with interpersonal conflict. This being said, it seems legitimate to characterize both go un-V-en and go V-ing in terms of perceived deviation or dissonance from a standard namely from a rule, principle or convention which is considered to be fundamentally right. Also common to the two constructions is a variably high degree of attitudinal involvement on the part of the speaker. When it comes to aspect, the go V-ing construction has a hybrid quality to it which is revealed by its rough and variable functional equivalence with the catenatives start, keep on, keep, and occasionally, perhaps, with a mix of start and keep onP Again, the concept of "cline" is relevant here: when looking through examples (16) to (24), we observe fairly wide variation in the weight given to inchoation and that given to continuation and/or iteration. Inchoation, for instance, is arguably foregrounded in (20) and (21), and possibly also in (22), while it is certainly backgrounded in (23) and completely absent in (19). Continuation, on the other hand, is clearly at play in (23), and iteration in (17) and (19). What is crucial here is, on the one hand, the specific Aktionsart profile instantiated by the relevant predicative relationship, on the other the role played by go in opening up a structured interval. Within this interval, one conceivable option is to construct a class of discrete, though notionally identical, occurrences which, to use a Culiolian concept (Groussier 2000: 173), are going to be

On two distinct uses of go


scanned. This particular operation will yield the kind of iterative, and for that matter habitual, value associated with go drinking in pubs in (19). The following example is more intricate, involving as it does iteration (to catch you between husbands) "embedded" as it were within the continuous and continuative state denoted by go waiting·. (40)

I can't go all my life waiting to catch you between husbands. (M. Mitchell, Gone with the Wind, 1936; quoted in Kirchner 1952: 324)

The phrase all my life serves to define the boundaries of the interval set up by go as a verb describing motion "through time". Indeed, instances of the go around V-ing construction, where around lends overt expression to go V-ing's aspectual potential as a marker of iteration/continuation, are not infrequent: (41)

I suppose one reason we've got on so well, May, is that you don't go around resenting things andfeeling you 're wasted on ordinary mortals. (BNC, H9D 2962)

Around is to go resenting things in (41) what on was to goes unchallenged in (28): both act as "aspectual resonators". 4.3.

Towards a joint characterization

It can be inferred from the foregoing observations that in both go un-V-en and go V-ing, go functions as a conjoined marker of aspect and modality or, more precisely, of aspect and modality conjoined. There is nothing surprising about this if we assume, again in the spirit of Antoine Culioli's thinking on language, that grammatical markers have an intrinsic propensity to be transcategorial. Both constructions have been shown to encode evaluative modality. Whether the situation described by the sentence consists in a specific course of action (V-ing) or in the state of inertia resulting from an absence of process (un-V-en), go highlights its undesirability, if not its outrageousness, in the eyes of the speaker. To that extent, the function go fulfils is unmistakably modal. The aspectual import of the two constructions is subject to a greater amount of variation, and is as such more elusive. What has been at play, in each of the various instances that have come under observation, has been one of two very distinct types of aspectual value, and sometimes indeed a mixture of these two types. One stems from the notion of "motion-through-

116 Philippe Bourdin time", which is itself an avatar of go's ability to refer to undirected motion, as in go naked or let's talk as we go: it accounts for the quality of persistence/protraction and iteration/continuation typically associated with go un-V-en and go V-ing, respectively. As is well-known, go is just as capable of denoting directed motion, by definition to a site distinct from the deictic centre - and, inferentially, departure from the deictic centre or some other point of reference; hence, context permitting, the inchoative value which go V-ing sometimes takes on - or, for that matter, the use of go as a change-of-state verb, as in go metric.™

5. 5.1.

Itive verbs as modal markers: a cursory crosslinguistic excursus Itives of evaluation

It is not difficult to find instances, among Indo-European languages, of a 'go'-type, or itive, verb acting as marker of negative evaluation: (42)



[Icelandic: Indo-European, Germanic; Einarsson 1949: 146] hann for aD gifta sig he go:PAST INFIN.PRTCLE marry REFL 'He went to get married!' (... but he shouldn't have.)15 [French: Indo-European, Romance] a. Et ne va pas rater cet examen! And don't go (to) flunk that exam! 'And don't you go flunking that exam!' b. *Va reussir cet examen! Go (to) pass that exam!

[Italian: Indo-European, Romance; Brianti 1992: 165] a. Che strane cose vai a raccontare! What weird things you go (to) tell! 'What weird things you go telling!' b. Perchi vai a immaginare queste cose su di lui? Why do you go (to) imagine these things about him? 'Why do you go imagining these things about him?'

In turn, these examples are reminiscent - both morphosyntactically and semantically - of ^""-century go to V, an evaluative construction which seems to have been a precursor to go V-ing:

On two distinct uses of go 117 (45) (46)

I am sure she would not go to tell a lie of anybody. (Cameron, Pink Tippet, 1824; quoted in Visser 1969: 1400) "Dear ma 'am', uttered Nurse Gill, you'd never go to suspect her!" (H. Wood, House of Halliwell, 1890; quoted in Visser 1969: 1400)

Visser (1969: 1400) points out that the construction tends to occur in negative and hypothetical contexts, a property which it bequeathed, as it were, to go V-ing·, as well, the gloss he proposes ("be so foolish, bold, or severe as to...") is fully consistent with the modal value of undesirable deviation that is also a hallmark of go V-ing.


The aspectual variability of evaluative itives

The interplay between evaluative modality and inchoation which characterizes some uses of go V-ing is also a defining feature of aller V-er, the cognate construction in French. However, it does not take the form of conjoining per se; rather, it shows up in such collocational affinities and disaffinities as are illustrated in the following examples: (47)

a. Et s 'il te plait, ne va pas te mettre a envoyer des lettres anonymes! And please, don't go (to) start sending anonymous letters! 'And please, don't you start sending anonymous letters!' b. * Et s 'il te plait, ne va pas finir par envoyer des lettres anonymes ! And please, don't go (to) wind up sending anonymous letters!

While aller can govern an inchoative aspectualizer like se mettre (a), it does not admit of a terminative one such as finir (par). In Spanish, it is andar, the "ambulative", non-deictic counterpart to the verb go of undirected motion that seems best suited to conjoin evaluative modality and iteration: (48)

[Spanish: Indo-European, Romance; Squartini 1998: 279] jSiempre andas quejdndote! You always go about complaining. 'You keep complaining on and on!'


Philippe Bourdin

The act of complaining is perceived as all the more annoying as andar, in conjunction with siempre, describes it as being subject to "hyperbolically continuous iteration" (Yllera 1999: 3418) - i.e. to endless recurrence over a period of time which is as indefinite as is the goal of the trajectory denoted by andar when it actually refers to aimless, non-oriented motion in space (El hombre andaba por la calle, 'The man was wandering on the street').16


Other itive markers of modality

The use of itive verbs as markers of modality and/or mood is attested in a number of languages: (49)


[Drehu Austronesian, Oceanic (New Caledonia); CI. Moyse-Faurie 1983: 90] tro-a humuth la gutu OBLIGATION[ (Bourdin 1999: 215-216; 223-224). So far as itive markers go, this claim receives empirical support from such apparently diverse and wide-ranging data as the following: (55)

[Maori: Austronesian, Oceanic (New Zealand); Bauer 1993: 329, 345; Biggs 1969: 66] a. a Te.Kao kei tahaki atu ART Te.Kao LOC[] side ITIVE.PRTCLE ο Kaitaaia


Philippe Bourdin


GEN Kaitaia 'Te Kao is on the other side of Kaitaia.' i eetahi atu LOC









'[Some years there are plenty of crayfish]; other years, hardly any.' c. te puuhaa, te waata.kiriihi, me eeraa DEF milkweed DEF watercress and DISTAL.DEM:PL atu kai a te Maaori ITIVE.PRTCLE food GEN DEF Maori 'milkweed, watercress, and the other kinds of food eaten by the Maori' [Mohave: Hokan, Yuman (California); Langdon and Munro 1979: 322-323] a. nya-isvar-k i:ma-k when—sing—SAME.SUBJECT[ ^ S ^ -3 3

«f S ? -is -g

26 1 9 10 3 8 2 4 3 66

26 1 1 6 3 8 3 2 50

ö bo s ^ I -ί ·Α