Modals and Periphrastics in English: An Investigation into the Semantic Correspondence between Certain English Modal Verbs and Their Periphrastic Equivalents [Reprint 2012 ed.] 9783110958904, 9783484303393

Citation preview

Linguistische

Arbeiten

339

Herausgegeben von Hans Altmann, Peter Blumenthal, Herbert E. Brekle, Gerhard Heibig, Hans Jürgen Heringer, Heinz Vater und Richard Wiese

Paul Westney

Modals and Periphrastics in English An investigation into the semantic correspondence between certain English modal verbs and their periphrastic equivalents

Max Niemeyer Verlag Tübingen 1995

To Cynthia

Die Deutsche Bibliothek - CIP-Einheitsaufnahme Westney, Paul: Modals and periphrastics in English : an investigation into the semantic correspondence between certain English modal verbs and their periphrastic equivalents / Paul Westney. - Tübingen : Niemeyer, 1995 (Linguistische Arbeiten ; 339) NE: GT ISBN 3-484-30339-5

ISSN 0344-6727

© Max Niemeyer Verlag GmbH & Co.KG, Tübingen 1995 Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt Einband: Hugo Nädele, Nehren

Contents

1

Introduction 1.1 Some existing proposals 1.2 Some basic descriptive problems 1.3 The present investigation 1.4 Data and the variety question 1.5 Methodological issues 1.6 Organization

1 1 3 6 7 8 10

2

Periphrastic forms 2.1 Introduction: establishing semantic relatedness 2.2 The classification of auxiliaries and modals 2.3 Criteria for periphrastics 2.3.1 Syntactic criteria 2.3.1.1 Tests of syntactic behaviour 2.3.1.2 Restrictions on the ordering and combination of i t e m s . . . 2.3.2 Semantic criteria 2.3.2.1 Lack of subject-auxiliary restrictions 2.3.2.2 Voice neutrality 2.3.2.3 Idiomatic status 2.3.2.4 Modal meaning 2.3.2.5 Epistemic-non-epistemic alternation 2.3.2.6 Conclusion 2.3.3 Phonological criteria 2.4 Conclusion

11 11 14 18 18 19 22 27 27 29 29 30 30 33 33 35

3

Approaches to this area 3.1 Introduction 3.2 Types of modal meaning and related terminology 3.3 Subjectivity and objectivity 3.3.1 Introduction 3.3.2 Various approaches 3.3.2.1 Semantic structure: modal and propositional elements . . 3.3.2.2 Point of view 3.3.2.3 Empathy perspective 3.3.2.4 The observer and the observed 3.3.3 Subjectivity, objectivity and modal-periphrastic alternation 3.3.4 Deontic source and deontic target 3.3.5 Some relevant descriptions

38 38 40 44 44 46 46 51 52 53 54 59 63

VI 3.4

3.5

3.6 3.7 3.8 3.9 4

The generative framework 3.4.1 Scope phenomena 3.4.2 Restrictions associated with subjectivity Pragmatic approaches 3.5.1 Various approaches to word meaning 3.5.2 Pragmatic interpretation 3.5.3 Processing constraints 3.5.4 Tense, time and modality

67 70 73 76 77 79 82 84

The 'actuality hypothesis' and related proposals Style and variety Markedness Conclusion

85 88 90 93

M u s t , have ( g o t ) t o , n e e d (to) and b e b o u n d t o 4.1 Introduction 4.2 Framework of this discussion 4.2.1 Some basic problems 4.2.2 General approach 4.3

4.4

Variety and style 4.3.1 Must vs have (got) to 4.3.2 Epistemic vs deontic uses 4.3.3 Have got to as an informal item Deontic uses of must, have (got) to and need (to): present tenses, affirmative . 4.4.1 First person subjects 4.4.1.1 First person subjects with must 4.4.1.2 First person subjects with have to

4.4.2

4.4.3

4.4.4

94 94 95 95 97 99 99 102 102 103 104 104 108

4.4.1.3 First person subjects with have got to 4.4.1.4 First person subjects with need (to) 4.4.1.5 Conclusion on first person subjects Second person subjects

109 Ill 112 113

4.4.2.1 Second person subjects with must 4.4.2.2 Second person subjects with have to 4.4.2.3 Second person subjects with have got to 4.4.2.4 Second person subjects with need (to) 4.4.2.5 Conclusion on second person subjects Third person subjects 4.4.3.1 Third person subjects with must 4.4.3.2 Third person subjects with have to 4.4.3.3 Third person subjects with have got to 4.4.3.4 Third person subjects with need (to)

113 117 119 120 121 121 121 124 125 126

4.4.3.5 Conclusion on third person subjects Conclusion

126 126

VII 4.5

4.6

4.7 4.8

5

Other aspects of the deontic uses of must, have (got) to and need (to) ...

127

4.5.1

Future reference

128

4.5.2

Habitual reference

131

4.5.3 Past reference 4.5.4 Non-affirmative forms 4.5.5 Non-finite forms Epistemic uses of must, have (got) to and need (to) 4.6.1 Present and future reference 4.6.2 Past reference

134 138 141 142 143 147

Be bound to General conclusions 4.8.1 General findings 4.8.2 Review of proposals in the literature 4.8.2.1 Major proposals 4.8.2.2 Some individual proposals

149 151 151 154 154 156

Other modal-periphrastic sets

162

5.1 5.2

Introduction Should, ought to, be supposed to and had better 5.2.1 Should and ought to 5.2.1.1 General issues 5.2.1.2 Compaxing the items

162 162 162 162 164

5.2.2

Be supposed to 5.2.2.1 Be supposed to compared with should/ought to 5.2.2.2 Specific issues

175 176 177

5.2.3

Had better 5.2.3.1 Had better compared with should/ought to 5.2.3.2 Specific issues

181 181 182

5.3

5.2.4 Conclusion Will, be going to, is to and be willing to 5.3.1

5.3.2

Will and be going to 5.3.1.1 Major analyses and major issues 5.3.1.2 Style and formality 5.3.1.3 Temporal orientation 5.3.1.4 The conditionally or contingency of will 5.3.1.5 The objectivity of be going to 5.3.1.6 Conclusion Will and is to

5.3.3 5.3.4

Will and be willing to Conclusion

185 185 »

186 186 189 192 198 200 201 202 204 205

Vili 5.4

6

Can and be able to 5.4.1 The semantic range of be able to 5.4.2 Suppletive uses of be able to 5.4.3 Could and actuality 5.4.4 Style and formality 5.4.5 Conclusion

Conclusions

Bibliography

206 206 208 209 212 213 214 220

1.

Introduction

The profusion of complex verb forms with a quasi-auxiliary status (eg, have got to, used to, be about to) is a striking feature of present-day English. Since the true modals (understood in a strict sense as can, could, may, might, must, shall, should, will and would) are morphologically defective, it is commonly assumed that a number of these complex, periphrastic forms (hereafter called 'periphrastics') fulfil suppletive syntactic roles: in some cases, there is fairly general agreement about this role (eg, with must, there is the past form had to for the wanting *musted\ with can, the infinitive to be able to for *to can; and with may/can (permission), the participle being allowed to for *maying/*canning); in other cases, this is not so (eg, with will, the infinitive to be going to for *to will; and with should, the past form was supposed to for *shoulded). The semantic aspects of this situation are rather more obscure. A close relation between some modal-periphrastic (hereafter ' M - P ' ) pairs is standardly accepted (eg, musthave (got) to, should-ought to), between others there is less agreement as to the exact nature of the relation (eg, will-be going to), while in certain cases a periphrastic is not obviously relatable to any one modal (eg, is to). In cases where a close semantic relationship is posited, this is commonly presented, whether implicitly or explicitly, as one of 'synonymy' (however that term may be understood). Nevertheless, certain specific semantic - or pragmatic - distinctions have often been argued for. Three well-known examples can be mentioned here: first, the association of some form of speaker involvement or orientation with must, in contrast to have (got) to; second, the relation of the explicit expression of the performance of an action to the use of was able to, in contrast to that of could; and third, various hypotheses to account for distinctions in use between will and be going to. Such matters have not always been assigned great importance. Thus, distinctions between the must-have (got) to pair are treated by scholars such as Poutsma (1904) and Kruisinga (1925), and, generally, in more detailed pedagogical materials; but they are ignored in the Grammar of contemporary English (Quirk et al 1972) (however, this is rectified in Quirk et al 1985), and they have no established place in theoretical accounts of the auxiliary verbs.

1.1

Some existing proposals

There have been few proposals for semantic distinctions applicable to M - P sets, as opposed to individual pairs. Three very broadly comparable approaches can be singled out here. Firstly, the general notion that modals are maximally unmarked items, both syntactically and semantically (Bouma 1975, and, especially, Perkins 1983), provides one

2

obvious basis for such a contrast; the saméis suggested by the statement by Joos (1968:21) that "modals have purely privative meanings, quasi-auxiliaries1 are additive too". Secondly, Haynes (1967:212), in an examination of modals and catenatives 2 , claims a basic contrast between the sets as follows: (1.1)

The modal system is a system of hypothesis in which the speaker considers and arrives at statements of probability. The catenative system expresses autonomic responses to external circumstances.

Both sets of items are here seen as being concerned "with statements that cannot be true or false; however, while modals are presented as speaker-oriented, the validity of assertions expressed by catenatives is specifically "contingent upon some other state or action" (op cit:25). Such comments can be related to the claim sometimes made that modals are inherently 'subjective', and periphrastics 'objective'. There is some similarity between these two approaches, in that both suggest that the semantic content of periphrastics is more highly specified, and they will be reflected in the discussion below at various points. The third approach, which I shall refer to as the 'Larkin/Lakoff proposal', is relatable to the second, but makes the most specific claim in this area and will now be investigated in some detail. Lakoff (1972:239) claims that the simple modals and their periphrases "two parallel sets of forms" - have normally been treated as "perfect semantic equivalents" (the only exceptions cited being Larkin 19763, and Binnick 1971 and 1972), and suggests that the contrast between related pairs could be generalized as follows: (1.2)

when the speaker agrees with, or takes upon himself, the atomic meaning of the modal, he can use the simple modal form. Otherwise he must use the periphrastic variant.

The proposed set of pairs discussed by Lakoff is as follows:

1

2

3

modal

periphrastic

must

have

may

be allowed

will

is to

should

be supposed

to to to

This term is applied by Joos to any lexical verb that takes an infinitive complement; in fact, however, he concentrates on eight commonly occurring items including, for example, be going io and have io (1968:20-30), and these correlate closely with the periphrastics that will be discussed here. The term 'catenative' is used in this account in a broad sense that subsumes all the items that will be treated as periphrastics here. Lakoff quotes this paper in its original version of 1969. I shall refer to the later, extended version (1976).

3 Lakoff explicitly restricts the set of pairs as above, and reports failing to find a parallel distinction between can and be able to (1972:240).4 To illustrate the claimed contrast between modals and periphrastics, I shall take Larkin's discussion of the must-have to pair, as in the following: (1.4)

My girl must be home by ten.

(1.5)

My girl has to be home by ten.

regarding which Larkin states (1976:392) that (1.6)

(1-4) implies that the speaker "goes along with" the prohibition that the sentence states, while (1.5) is neutral in this regard.

He subsequently makes this more explicit (393): (1.7)

the speaker of a must sentence identifies himself in some way with the source of the compulsion.

Despite Lakoff's statement to the contrary, these claims do reflect widespread feelings about some M-P pairs; however, the Larkin/Lakoff proposal is uniquely wide in application and explicit in an area where vague or qualified comments have been more normal. This hypothesis does not seem to have been taken up by other scholars working in the field of modality in English (understood in a sufficiently broad sense to include periphrastic exponents), apart from brief comments of qualified approval by Palmer (1990:81)5, and of rejection by Bouma (1975:324-325); it will therefore be discussed in some detail here (briefly in 1.2, and more extendedly in 3.5 6 ).

1.2

Some basic descriptive problems

In this section I shall first discuss some evident difficulties with the Larkin/Lakoff proposal, then go on to some implications of certain other proposals that have been made in the

5

6

In a later discussion of this issue (1981) she offers a rather different list, as follows: modal periphrastic must have to may be allowed to can be able to should be obliged to The discrepancies between this set and that in (1.3), with its explicit exclusion of can-be able to, are not explained, and remain baffling. However, the listing in (1.3) is argued for at some length, so that version will be assumed in the discussion below. Palmer's Modality and the English modals will normally be referred to in its second edition (1990), although in many cases the relevant comments appeared in the first edition (1979a). Bold type is used to indicate chapters and sections in the text, and normal type to indicate examples, quotations, etc: thus 3 . 2 refers to Chapter 3, section 2, while (3.2) refers to quotation 2 in Chapter 3.

4 field of M-P alternation, and finally comment on specific pedagogical problems that arise here. Any literal association of modals with speaker involvement, in the sense of initiation or support, and of periphrastics with its absence, as set out in (1.2), (1.6) and (1.7), cannot be descriptively correct. I shall illustrate this with three examples using must and have to: these items are chosen here since they are commonly used to exemplify the general question (cf (1.4)-(1.7) above), and thus, effectively, constitute the paradigm case. As can be seen, adequate contextual material is essential for discussion of these matters: (1.8)

she'd just made up her mind that she must leave him and go off . . . (Svartvik k Quirk 1980:286-287/562-565)7

(1.9)

Where do we go now? We sit and wait. We see if any nations are prepared to cough up enough troops for the safe zones (which do not include territories disputed by Muslims and Croats, and where the fighting will almost certainly continue). We must also sit idly by as Serbia and Croatia slip into economic anarchy, with all the dangers that this implies. (The Observer, 30.5.1993:22)

(1.10)

Β

what he doesn't realize is that not everybody else can work quite as hard as he can, and you just have to say A yea Β well, I'm singing tonight or something A yes Β if he says you've got a seminar (CEC:130-131/207-215)

In (1.8) the speaker does not in any way initiate or support the 'necessity' denoted by must (and had to is in any case available for such uses in reported speech); in (1.9) must reports something that the author in no way wishes, but is, on the contrary, opposed to; and in (1.10) have to clearly presents the speaker's advice. According to any theory associating must with speaker authority and have (got) to with its absence, the alternative items would be more appropriate 8 in each case; it is, therefore, quite clear that statements such as that in (1.7) require some kind of qualification. One possibility to consider is whether (1.7) can be saved with a rather freer interpretation. Thus, in (1.8) it might be claimed that the subject's perspective overrides the speaker's, or, alternatively, that she must leave him is a reported version of I must leave him which is available to present the subject's/original speaker's viewpoint. In (1.9) the 7

8

This work (A corpus of English conversation: hereafter referred to as 'CEC' or 'the corpus') is used as the major illustrative source in the present study; cf discussion of data in 1.4. The references are to page numbers (here 286-287) and tone units (here 562-565). The texts are slightly simplified for convenience of reference; prosodie marking is given only in special cases. Individual speakers are indicated only when more than one is involved, eg, Ά', 'B', 'a' (always as in the original). I am grateful to Jan Svartvik for permission to use the Corpus of English conversation in this simplified form. It will be explained below in 1.5 how terms like 'possible', 'appropriate' and 'normal' are to be understood in such contexts.

5 author might be assuming the perspective of the circumstances that determine the inactivity in question, and this would then overrule his own feelings about its undesirability. Finally, in (1.10) ß ' s advice to A might be based circumstantially rather than personally, thus justifying the use of have to. However, such freedom of interpretation of the speaker's role in the use of the items in question would stretch the original claim, as in (1.2) or in (1.7), rather far, and, crucially, make it difficult to see the necessity for the claimed interpretations of (1.4)-(1.5). The challenge of establishing the nature of the distinction in use between must and have to becomes more complex if data such as the following are considered (from Coates 1983:54): (1.11) (1.12)

*I must get up at 7 am every day. 9 I have to get up at 7 am every day.

Such an incompatibility between the use of must and habitual reference is sometimes argued for. Whether this is absolute, which would make (1.11) ungrammatical, or merely a contextually determined tendency, it is unclear how such a distinction, which would associate have to, but not must, with the potential for habitual reference, should, or can, be related to the speaker-based distinction claimed in (1.2). To move to other items, and other proposals, it has been argued (eg, Palmer 1990:194196) that in the following pair: (1.13) (1.14)

*I ran fast, and could catch the bus. I ran fast, and was able to catch the bus.

an implication 1 0 of actuality attaches to was able to, but not to could, making (1.1) odd, and that this can be generalized to a certain extent both for these items and for other M - P pairs. Again, it is unclear how such a distinction could be related to a basic speaker-related contrast in use. These examples will suffice here to indicate that there is a considerable descriptive problem involved in determining the nature of (a) such distinctions in use as can be shown for a particular M - P pair, and (b) their applicability to other such pairs. More seriously, it is far from clear exactly which modals and periphrastics are to be paired. For example, Lakoff's list in (1.3) includes will-is to as a, pair; yet will and be going to are more usually related together, and is to is not normally treated as a modal-related periphrastic. Further, while must and have to are regularly treated as a pair, have got to is rarely considered, other than as a colloquial variant of have to (eg, Palmer 1990:114), much as have got is treated as a colloquial variant of have; however, to return to (1.11) and (1.12), have got to may sometimes behave more like must than have to, supposedly making the following ungrammatical:

10

Judgements of g r a m m a t i c a l l y will normally be retained from the relevant sources; cf similar cases in (1.13) and (1.15) and the discussion in 1.5. (1.11) is not actually given by Coates, but is implied in her argument (cf (4.122) in 4.5.2). Palmer uses the term 'implication', but 'implicature', in a Gricean sense, might be appropriate (cf Palmer 1990:192,194).

6 (1.15)

*I've got to get up at 7 am every day.

(Coates 1983:54). As a final example, if should-be supposed to is taken to be a valid pair, as in (1.3), it is unclear how ought to, which is widely treated as a synonym of should, relates to them. Descriptive problems of such kinds are reflected in the pedagogical aspects of this area. It is striking that a fair amount of the discussion of relevant questions has arisen in the context of the teaching and learning of English as a foreign language; for example, the account by Haynes (1967), referred to in 1.1, is explicitly motivated by pedagogical problems. General learning difficulties with the English modals - together with related periphrastics - are frequently reported. Thus, Quirk discusses errors in this area that are "legion and apparently ineradicable" (1981:160-161); while in a survey of learner English, dealing with problems encountered by native speakers of 19 different languages or language groups, difficulties with the appropriate use of sets of modals and periphrastics, such as must-is to/have to, may-be allowed to, shall/will-be going to, shall/should-be supposed to, and can-be able to are a recurrent theme (Swan & Smith 1987:passim).

1.3

The present investigation

In view of the kinds of problems illustrated above, it is proposed here to investigate the possibility that there is a systematic distinction or set of distinctions to be observed in the use of modals and related periphrastics. This will involve the exploration of the use of specific sets of supposedly related items, with particular reference, as in all the relevant literature, to must and have (got) to, and the attempt to establish general principles applicable across all the paired or grouped items in question. This account is primarily descriptive in intention, although a number of theoretical issues will be involved, and is strictly concerned with the nature of M-P alternation. It is thus not centrally concerned with the syntax and semantics of the set of English modals, an area very thoroughly investigated over the past two decades (notably, Coates 1983, and Palmer 1990); nor with modality in general, whether as manifested in English (Perkins 1983) or in language generally (Palmer 1986), or as a semantic/logical concept (eg, Lyons 1977:Ch.l7). This restriction of focus means that many issues that may seem central to the study of modals receive little, if any, mention. Finally, the historical aspects of the development of periphrastics in relation to the recategorization of modal verbs (eg, Lightfoot 1979:Ch.2, Plank 1984, and Hopper & Traugott 1993:45-48), however revealing they may be, lie outside the concerns of a synchronic presentation; on the other hand, account is taken of recent developments and ongoing change wherever this is found relevant. Like the great majority of studies in this area, this description is primarily at sentence level, although extended texts will often be used to provide adequate contextualization. A discourse level of analysis will, however, be seen to be ultimately essential for an understanding of the items under discussion.

1.4

Data and the variety question

The vaxiety of English discussed here is 'standard British English' (BE), as described in Quirk et al (1985) and accounts specifically concerned with the modals in BE, especially Coates (1983), Perkins (1983) and Palmer (1990). The Corpus of English conversation (Svartvik & Quirk 1980), referred to in Footnote 7 (where the conventions of citation are explained), is used as the major source of data 11 ; otherwise, data are drawn from the following sources: written material (especially journalism and fiction), occasional spoken data, the relevant literature (naturally occurring, wherever possible), and intuition. Written material is primarily 'contemporary', in the sense of deriving from the last half century, but occasionally includes older texts where either there is no reason to assume its non-validity, or it illustrates recent development in the language. A few novels that are frequently quoted from are included in the Bibliography. It is worth emphasizing the desirability both of using real data where possible, and of presenting this in adequate contexts. A lot of discussion and theorizing in this general area has been based on material that is invented or non-contextualized, often both; there will be plenty of evidence of this later in the present work, and it is illustrated above in, for example, (1.4)-(1.5) and (1.11)-(1.15). There is the further, related question to what extent native-speaker intuition is an ideal source. These issues will be discussed in 6. Some of the literature cited is American in origin, and must be assumed to reflect American English (AE). This could, in principle, be problematic, since there are clearly some differences between BE and AE in the use of the modals and periphrastics, as has been demonstrated by, especially, Coates L· Leech (1980); on the other hand, since British and American scholars, or, more exactly, scholars dealing primarily with BE and AE respectively, regularly cite one another without evident restriction in this respect, it must be concluded that differences are not generally felt to be of significance. My strategy here will be to assume that, in general, all statements may be equally relevant, whether they appear to reflect AE or BE; but for maximal consistency, BE data will be used as far as possible, and findings will be taken to be primarily applicable to BE. However, some evident discrepancies between the two major systems will be commented on. Some data on Australian English are also taken into account briefly. Apart from potential variety distinctions the possibility of differences according to style must also be allowed for. This concerns two basic contrasts: first, that between speech and writing, and second, that between informal and formal use. (Obviously, these parameters are complex, and tend to overlap to the extent that speech is typically informal and writing typically formal.) The possibility of stylistic variation of such kinds has not often been given systematic attention in this field (but cf Coates & Leech 1980, Coates 1983, and Hermerén 1986). While some of the discussion will, of necessity, be neutral as to style, relevant variation will be taken account of wherever possible. This general issue will be taken up in detail in 3.7. 11

This corpus has the advantage of being a body of publicly available data reflecting present-day spoken BE. Since it is part of the Survey of English Usage ('SEU'), it is also used to some extent both by Coates (1983) and by Palmer (1990), and a small number of texts discussed by one or other of those scholars will recur here.

8

1.5

Methodological issues

Some remarks as to method must be made here. Establishing distinctions between M - P pairs (or sets) involves (a) decisions regarding polysemy and synonymy, and (b) judgements of acceptability. Two prominent problems in these fields are that many of the items to be compared are commonly simply treated as synonymous or 'equivalent', while it is rarely the case that one can confidently categorize a particular use of an item as ungrammatical. These two related questions will be discussed in turn. The variety of uses of the modals, in particular, has to be treated in terms of either polysemy or homonymy. Polysemy has been much more commonly chosen, but that solution leaves the question of relating the different 'meanings' together; this issue will be discussed in some detail in 3.5.1-2. When it comes to the 'semantic overlap' between M - P pairs, a specific approach to synonymy is required; and this matter will now be discussed. Cruse (1986:Ch.l2) posits a scale of synonymy between, at the upper end, 'full synonymy', where two items display "equinormality in all contexts" (op cit:268), and, at the lower end, 'cognitive synonymy', which can, roughly, be equated with truth-conditional equivalence. This approach derives from the Firthian treatment of word meaning as a product of a word's collocational possibilities (cf, especially, Firth 1957:194-196, and Carter h McCarthy 1988:32-38). In detail, this approach works as follows: (1.16)

It is taken as axiomatic . . . that every aspect of the meaning of a word is reflected in a characteristic pattern of semantic normality (and abnormality) in grammatically appropriate contexts. . . . every difference in the semantic normality profile between two items betokens a difference of meaning. (Cruse 1986:15-16)

This means that a pair of full synonyms must be equally 'normal' in all conceivable contexts. Such an approach to word meaning bypasses any semantic-pragmatic distinction, although it would be reasonable to draw the limits of semantics at cognitive synonymy, and treat items that are more than cognitively, but less than fully, synonymous as distinguishable pragmatically. The possibility of drawing a distinction between semantic and pragmatic aspects of meaning and use in the area under discussion is rarely discussed explicitly; an exception in the literature is Haegeman (1989), who seeks to treat the pair will and be going to as truth-conditionally - ie semantically - equivalent, and only distinguishable pragmatically. It is, in fact, rarely, if ever, possible to show clear-cut semantic distinctions between M - P pairs (cf 2.1 and 3.5 on these issues). The demonstration of non-total synonymy between a pair of items, or rather between a pair of sentences including contrasting items, involves judgements of acceptability, the other issue to be discussed in this section. It is, as was noted above, rarely the case that one can absolutely characterize a specific use of an item as unacceptable or ungrammatical 1 2 ; the problem can be illustrated by reconsidering examples (1.11) and (1.13). The 'bad' 12

Although the terms 'unacceptable' and 'ungrammatical' can, and no doubt should, be distinguished, there will be no explicit distinction between their use here: the strictly practical reason for this is that cases where acceptability may be doubtful do not normally seem to raise issues of grammaticality, while questions of ungrammatically do not normally arise.

9 (1.11) becomes 'better' with the addition o f . . . this week, for example; (1.13), by contrast, is a rare case where there may be, at least at first sight, no obvious context which would make the sentence acceptable. In general, '?' will be used to indicate simply that there is something odd about a sentence as it stands (which means, informally, in its 'most obvious' or default interpretation in such context as it has), and categorical judgements of grammaticality will mostly be avoided (but acceptability judgements will normally be retained from their original sources; cf Footnote 9). 13 This matter is, in any case, perhaps better treated in terms of 'relative normality'. (1.16) suggested that any difference of meaning between two items will be demonstrable within a context, where one item will be shown to be preferable to or 'better' than the other. This is illustrated by testing for relative normality between potential synonyms, as in the following pairs (Cruse 1986:269): (1.17)

This is a scandalous waste of money.

(+)

(1.18)

This is an outrageous waste of money.

(—)

(1.19)

He looks almost Chinese.

(+)

(1.20)

He looks nearly Chinese.

(—)

The point in such examples is that (1.18) and (1.20), as they stand, are likely to be found acceptable, and hence treated as possible, even if slightly odd; but when a choice between them and (1.17) and (1.19), respectively, is available, the first examples axe preferable (or more appropriate). This does not mean that, for example, in the frame This is a(n) ... waste of money, scandalous is necessarily the 'best' item; it is merely better than outrageous. Hence this type of judgement or elicitation is for relative normality; and ' + ' simply means 'better, here', and '—' 'worse, here'. Thus, a judgement shown as '—' has no necessary relation to '?' or '*', although they are likely to correlate to some extent. Following the discussion in 1.4, a context of use may also include an indication of variety and style. In this study, explicit testing of this kind will be used on occasion; more generally, any judgement of 'normality' or 'relative normality' will presuppose such a procedure. An important further complication arises here with our material: in some cases comparison is made between three forms (eg, must, have to and have got to), so that a relative normality judgement may have to be split into, for example, two ' + ' s and one '—', if two forms appear markedly more normal than a third. Further, it may happen that of three items A, Β and C, A and Β are both preferable to C, but a distinction in preferability between A and Β can subsequently be established. The discussion in 1.5 has pointed to a contextual approach where real or 'recoverable' linguistic and/or extralinguistic data may contribute to the elucidation of texts, and thus of the meaning potential of individual items. Within such an approach, the possible futility of discussing examples like (1.4)-(1.5), simply as they stand, becomes obvious.

13

Within one tradition of corpus-based descriptive grammar, a time is anticipated when "the asterisk of the 'impossible sentence' will become a thing of the past" (Francis 1993:156).

10

1.6

Organization

This work will be organized in the following way. In Chapter 2 the problem of the establishment of a set of periphrastic or quasi-modal items 14 will be dealt with, and various approaches to showing semantic/pragmatic relationships between pairs or groups of modals and periphrastics will be discussed. This will lead to the selection of items for subsequent detailed treatment. In Chapter 3 various arguments used for establishing systematic relationships among groups of items will be critically reviewed, and the theoretical approach to be adopted in the ensuing descriptive section set out. Must and related items are then discussed in detail (Chapter 4), and, partly in the light of the findings, the other items selected are dealt with next (Chapter 5). In Chapter 6, finally, the findings are summarized and some implications for the theories and proposals discussed in Chapter 3 considered; in addition, further more general issues that arise in the course of the study are discussed.

14

'Quasi-modals' will be used here as a cover label for items that lie between the status of full/true modals and that of full/main/lexical verbs, partly because it appears to be the term most widely used (eg, Shopen 1971, Hakutani k Hargis 1972, Chapin 1973, Coates 1983, and Perkins 1983). Other terms used in this area include 'quasi-auxiliaries' (eg, Joos 1968, Palmer 1974, and Bolinger 1980), 'semi-modals' (Palmer 1990), 'semi-auxiliaries' (Kajita 1968), 'quasi-auxiliary modal expressions' or 'non-auxiliary modal expressions' (Perkins 1983), and 'lexical auxiliaries' or 'lexico-modal auxiliaries' (Downing L· Locke 1992). The term 'catenatives' (as used, for example, by Quirk et al 1985, or by Twaddell 1963 and Haynes 1967) covers a markedly broader area (cf Footnote 9 in 2). 'Periphrastic items' or 'periphrastics' are to be understood here to denote any complex items, including quasi-modals, that are claimed to be relatable to full modals; cf definition in 2.1.

2.

Periphrastic forms

2.1

Introduction: establishing semantic relatedness

The term 'periphrastic' will be used to denote a complex verb that, ideally, displays the following three features: (a) grammaticalization; (b) idiomaticity; (c) semantic relatedness to a central modal auxiliary. These features require some clarification. Feature (a), grammaticalization, requires that periphrastics have in common a set of syntactic and semantic features which together suggest the existence of a grouping of some significance but with less than categorial status (cf discussions of grammaticalization by Palmer 1986:3-7, Croft 1990:230-244, and Hopper & Traugott 1993:Ch.l). One aim of this study is to determine to what extent, and how, this can be shown to be the case, and it will be the general topic of 2.3. Feature (b), idiomaticity, requires that the meaning of a complex form is not simply a function of the meanings of its components. This is intended to distinguish periphrastics from simple paraphrases (or paraphrastic forms). This will be investigated in 2.3.2. Feature (c), semantic relatedness, is the basic constraint here. Establishing truthconditional equivalence should be an appropriate way of showing whether an M - P pair are semantically equivalent (cf 1.5), and the possibility of doing this will now be explored. It will be shown that while adequate demonstration of equivalence can be very difficult to achieve, an informal assumption of such equivalence between M-P pairs and sets is a necessary condition for any such investigation (and in the literature this is standardly taken as self-evident). I shall start with the items must and have got to, commonly taken to be roughly equivalent (must and have to have already been noted as, effectively, the paradigm case for the investigation of M-P relatedness; for convenience, have got to rather than have to is used in the present discussion). One test for such equivalence checks compatibility in two conjoined clauses; conjunction.will be indicated by a dash between the clauses in question (and is not used for this purpose here since it contributes to oddity in the sentences): (2.1) (2.2)

This work's got to be finished by April - it must be. This work must be finished by April - it's got to be.

Both (2.1) and (2.2) appear fine, suggesting compatibility between must and have got to. Compatibility in both directions is required here, as the following data for have got to and ought to show: (2.3) (2.4)

This work ought to be finished by April - it has got to be. ?This work has got to be finished by April - it ought to be.

12 Here, (2.3) is fine, but (2.4) is odd. The oddity of (2.4), but not of (2.3), arises from the fact that in a pair of semantically related conjoined clauses, the second can either repeat or increase the value of the first, but not reduce it, unless with a qualifier such as at least.1 Since has got to is 'stronger' than, or has a higher scalar value than, ought to, but not vice versa, (2.3) is normal, since the second clause can be seen as reinforcing the first, but (2.4) is not, since the second clause is neither equivalent to, nor stronger than, the first. As a further test, assertion of one item should not be compatible with negation of the other: (2.5)

*This work must be finished by April, but it hasn't got to be.

(2.6)

*This work has got to be finished by April, but it mustn't be.

(I ignore the complication that negation works in different ways in these sentences - in (2.5) the modality is negated, but in (2.6) it is the proposition.) To complete this picture, we can consider the effect of negation with have got to and ought to conjoined: (2.7)

*This work has got to be finished by April, but it oughtn't to be.

(2.8)

This work ought to be finished by April, but it hasn't got to be.

(2.7) is odd, but (2.8) is normal. The reason for the possibility of (2.8) lies in the effect of scalarity between the two items, referred to above, which allows one to say Ά , but not B' when Β has a higher scalar value than A. The combined findings for must and have got to, as shown in (2.1)-(2.2) and (2.5)-(2.6), and for have got to and ought to, as shown in (2.3)-(2.4) and (2.7)-(2.8), constitute substantial evidence for the semantic equivalence of the first pair, but not of the second. To take this question further, this time in relation to three items involving two cases of potential equivalence - will, be willing to and be going to (will being related to both the other items) - we can consider the following (here and is used as the link between the conjoined items): (2.9)

I'll help you, and I'm willing to.

(2.10)

I'm willing to help you, and I will.

(2.11) (2.12)

I'll help you, and I'm going to. ?I'm going to help you, and I will.

(2.9) and (2.10) point to the equivalence of will and be willing to, parallel to (2.1) and (2.2) for must and have got to. For will and be going to, (2.11) similarly points to equivalence, but (2.12) is rather odd - the reason for this will be considered in a moment. The evidence of negation is as follows:

Cf the discussion of sentences like "I don't just think so - I know so", as showing incompatibility between the verbs contrasted, and thus a higher scalar value for the second, in Caton (1966:especially 59); cf Westney (1986).

13 (2.13) (2.14)

*Π1 help you, but I'm not willing to. ?I'm willing to help you, but I won't.

(2.15)

*I'll help you, but I'm not going to. 2

(2.16).

*I'm going to help you, but I won't. 3

Of these, only (2.14) is less than fully anomalous. The combined evidence for the pairs of items is as follows: for will and be willing to only (2.14), and for will and be going to only (2.12) fails to provide full support for equivalence. We must now consider why (2.12) and (2.14) do not conform to the general patterns. I shall take (2.12) first. It was said above that when two semantically related clauses are conjoined, the second item is either equivalent in strength to the first or reinforces it; thus, while (2.3), with ought to followed by has got to, was normal, (2.4), with has got to followed by ought to, was not. In the case of (2.12), since (2.11), which reverses the positions of the items in question, is normal, it could be hypothesized that be going to is in some way 'stronger' than will·, and in so fax as be going to is generally felt to be 'present-oriented' as opposed to 'future-oriented' will, and present time is more immediate or real than future time, this might appear a valid argument. (For discussion of will and be going to, cf 5 . 3 . 1 . )

In the case of (2.14), where am willing to is followed by won't, a possible reason for the non-total anomaly of the conjunction might be that will, at least when negated, is felt to be stronger, or semantically wider, than a statement of positive willingness, and thus to override it. A further issue arises in this discussion of (2.9)-(2.16): if the evidence for semantic equivalence is valid, then one could ask whether, since both will and be going to, and will and be willing to, are equivalent pairs, be going to and be willing to are also equivalent. Conjoining does not show any anomaly: (2.17) (2.18)

I'm going to help you, and I'm willing to. I'm willing to help you, and I'm going to.

In the tests for contradiction, the following emerges: (2.19) (2.20)

?I'm going to help you, but I'm not willing to. I'm willing to help you, but I'm not going to. 4

In (2.19) there is some oddity, though hardly a clear contradiction, while (2.20) seems possible. This would suggest that the evidence for or against the equivalence of be going to and be willing to is simply inconclusive. The essential difference between this pair of 2

3

4

A similar sentence is used by Palmer (1990:196), as evidence that will, unlike be willing to, implies actuality. Truth-conditional equivalence between will and be going to sentences is claimed by Haegeman (1989), on the grounds that one cannot simultaneously deny a will sentence and affirm a be going to one, and vice versa. A similar sentence is used by Palmer (1990:196), as evidence that be willing to, unlike will, does not imply actuality (cf Footnote 2).

14 items and the others we have been examining seems simply to be that there is no reason to assume that be going to and be willing to axe equivalent, whereas the other pairs have often been discussed precisely in this connection. To conclude this discussion of feature (c), then, the requirement of semantic equivalence between an M - P pair will be assumed to be demonstrable, but a basic proviso must be added to the effect that only items that are commonly assumed to be closely related semantically should be taken into consideration.

2.2

The classification of auxiliaries and modals

Before deeding with the classification of periphrastics, it will be necessary, first, to set out criteria for the classification of auxiliaries, and, within them, of modals, and then go on to deal with items that are, in various ways, intermediate between modal and full verb status. In this section I shall discuss auxiliaries and modals, and the possibility of defining a set or sets of items intermediate in status between modals and full verbs, and then in 2 . 3 I shall go on to consider possible criteria for a set of periphrastic items. On the basis of that discussion I shall draw conclusions in 2 . 4 for the subsequent account. The following set of syntactic criteria has been widely accepted as a test of auxiliary status (cf, especially, Palmer 1987:14-21): (2.21)

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

negation: the existence of -n't negative forms, as in shouldn't but *workn't·, inversion: the possibility of occurrence in first position, before the subject, as in should you? but *work you?; 'code': the possibility of recurrence without a full verb, as in so should we but *so work we; emphatic affirmation: the possibility of use, when stressed, in affirmation of a denied or doubted statement, as in but he Should! but *but he * works! (for but he 1does work!)

These criteria, often referred to as the 'NICE' properties, define the primary auxiliaries (be, have, do) and, allowing for a few anomalies, the secondary auxiliaries or modals (can, could, may, might, shall, should, will, would, must, ought to, dare, need). This list of modals can be narrowed down to a 'pure' central group, which would exclude the last three. These three items have, in particular, the following characteristics: ought to (at least in standard B E ) standardly requires to with the infinitive, while dare and need also occur as lexical verbs; in addition, used to, with considerable restrictions, may show NICE features, but is usually excluded (Palmer terms it "very marginal", 1987:170). Beyond these items, there are verbs commonly treated as being semantically related to modals, like is to, be going to and had better, which all start with a primary auxiliary, and in that respect display auxiliary characteristics. Obviously, there has to be some way of deciding how a limit can be set to complex forms of this kind, since while items such as

15 those mentioned are felt to be fixed units, there are also cases of a less obviously idiomatic status, eg, be willing to and be allowed to. A relatable, but rather more detailed, treatment of this question is provided by Quirk et al (1985:121-127). The criteria for auxiliaries here are as follows: (2.22)

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

(g) (h)

operator in negation with not (eg, cannot)·, negation and verb contraction (eg, isn't, 'we); inversion of subject and operator (eg, will she?); emphatic positive (eg, 11will try); operator in reduced clauses (eg, can you? - no, I can't); pre-adverb position: positional option available for frequency adverbs and disjuncts like certainly (eg, she would never/certainly believe that story but *she believed never/certainly that story); quantifier position (eg, the boys will all be there but *the boys play all there); semantic independence of the subject, manifested in three different ways: (i) of subject-auxiliary restrictions (eg, the bus ought to be here at 5 but *the bus hopes to be here at 5); (ii) possibility of existential iAere-constructions (eg, there used to be a school but *there hoped to be a school); (iii) active-passive correspondence (eg, thousands of people will meet the president — the president will be met by thousands of people; but thousands of people hope to meet the president φ the president hopes to be met by thousands of people).

The first five items here correspond very largely to the NICE properties listed in (2.21): (a) and (b) cover negation, together with general auxiliary contractions (eg, we've, we'll), (c) inversion, (d) emphatic affirmation, and (e) code. Of the remaining three, (f) and (g), which concern the position of operators in relation to adverbs and quantifiers, will not be considered here; however, (h), which, unlike the other items, explicitly involves semantic issues, will be returned to below in the discussion of various possible features for periphrastic forms - in 2.3.2.1 for sub-item (i), 2.3.1.1 for (ii) and 2.3.2.2 for (iii). Assuming that the auxiliaries as a whole can be satisfactorily listed, the modale cam then be defined as a subset according to the following two syntactic criteria (Palmer 1987:26): (2.23)

(a) (b)

no -s form: thus *he shoulds; no non-finite forms: thus *to should, *shoulding, *shoulded; hence no co-occurrence: thus *I should can go.

Quirk et al (1985:127-128) have the following criteria for the characterization of the modals:

16

(2.24)

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

construction with the bare infinitive (eg, they must go but they ought to go); finite functions only; no 3rd person singular inflection; abnormal time reference (eg, he might return next May).

Criteria (a)-(c) together correspond to (2.23). (d) is not simple: this refers to the fact that past forms of modals can be used in ways not available to other verbs - thus, it was possible he will return next May does not correspond to the standard epistemic sense of he might return next May. On the other hand, the form in question, might, is not itself regularly available for past time reference (thus *he might return yesterday, to parallel he may return today/tomorrow). This means that it is doubtful how far forms like might and should can be treated as past tenses of may and shall, respectively, except in specific uses, such as in reported speech. The modals do, in fact, generally have anomalous potential for time reference (eg, with must, that must be done later in the summer, where must has future reference; but with have to, that has to/will have to be done later in the summer, with restrictions on these uses, suggesting that must has wider temporal reference). This criterion for modals will be relevant to the comparison of certain M-P pairs below. Finally, we must consider the question whether it is at all possible to define a set or sets of items intermediate in status between auxiliaries (including modals) and lexical verbs. This is very obviously a confusing area, above all on account of overlapping criteria and variable usage, and some accounts have emphasized the non-categorial nature of any classification, and have seen the question in terms of gradience between two relatively clearly defined poles. Quirk et al (1985) provide the most extensive description, but first I shall mention two other relevant accounts. Huddleston (1980), in an attempt aimed at descriptive clarification, sets out syntactic data for 37 items ranging from auxiliaries to catenatives such as begin and try, according to 30 parameters; in toto, since variation in use is allowed for, the overall picture this produces is very complex, and no attempt is made to sub-categorize the items under discussion. Rather more simply, Bolinger (1980) takes a set of ten items, ranging in order from should, an unquestionable auxiliary, through a group with roughly decreasing claims to auxiliary status - ought to, used to, got to, be supposed to, have to, be going to, want to and try to to regret to, the last having none of the features associated with auxiliaxies. Bolinger is particularly interested in the gradience of the items under discussion, and in the potential state of transition of the intermediate items (between should and regret) towards full auxiliary status. 5 The ten items are plotted on a grid according to nine criteria, seven syntactic, one semantic and one phonological. The syntactic criteria are close to those that lie behind the organization of (2.25), to be discussed below. The semantic criterion is the availability of items for "epistemic, aspectual, or modal meaning"; modal, including epistemic, meaning will be discussed in 2 . 3 . 2 . 4 - 5 , but aspectual meaning will not concern us here. The phonological feature is typified by the 'wanna' pronunciation of want to (the

T h e notion of shifting category membership is also applied by Shopen (1971), in relation to sequences such as go get and come sleep (these represent a process in A E that is hardly prominent in current BE).

17 status of this item being the basic motivation of Bolinger's study); this will be returned to in 2.3.3. T h e scheme set out by Quirk et al (1985:136-148) involves four distinct categories between central modals and full verbs, listed as (b)-(e) below 6 ; although these categories are listed separately, they are treated as, ultimately, constituting a gradient: (2.25)

(a)

central modals

(b) (c)

marginal modals modal idioms

(d)

semi-auxiliaries

(e) (f)

catenatives main verbs

can, could, may, might, shall, should, will, would, must dare, need, ought to, used to had better, would rather, would sooner, be to7, have got to; 'less common idioms': would (just) as soon, may/might (just) as well, had best have to, be able to, be about to, be bound to, be going to, be obliged to, be supposed to, be willing to; also: be apt to, be due to, be likely to, be meant to, (be unable to, be unwilling to) eg, appear to, keep + -ing

To consider these sub-categories in turn, (a) is the set of items that fully meet the NICE criteria, as in (2.21), while (b) contains the four items that do so restrictedly. (c) is a group of items that are auxiliary-like in that their first element is an auxiliary, and modal-like in that they have no non-finite forms 8 , and thus cannot co-occur with modals. Presumably the modal idioms, like the preceding two categories, axe to be considered a closed set. Group (d) are also idiomatic structures, formed of be χ to (χ being an adjective or participle, with the sole exception of about), apart from the item have to. These items differ crucially from the modal idioms in that they do have non-finite forms and can cooccur with modals. A striking feature of this syntactically based classification is the fact that the items have to and have got to are listed in different sets, although semantically they are always treated as close, if not synonymous. The boundaries of class (d) are not clear, and the last two items are introduced as doubtful members; the reasons for this, and the status of other possible members, will be discussed in 2.3.1. Catenatives, group (e), are more obviously an open class, and can simply be treated as main or lexical

6 7

8

Regarding varying terminology for these categories, cf Footnote 14 in 1. This item is normally considered to lack non-finite forms, such as *be to itself, and thus is more reasonably termed is to (eg, Palmer 1987:160-161, and 1990:4-5). There are always surprises. For example, would rather is standardly considered to have no non-finite forms, yet consider the following: She says she doesn't but she says it in a way that makes you rather she'd say she did a bit. (Amis 1991:43) Rather, having presumably lost its would, functions here infinitively, but the sentence is not odd at first sight/hearing.

18 verbs that enter into particular syntactic structures, while having certain syntactic and semantic similarities to semi-auxiliaries.9 In the following section I shall examine a number of syntactic, semantic and phonological features that might contribute to the definition, from the set of intermediate items, of a set of periphrastic items, with specific reference to the three requirements listed at the beginning of 2.1. I shall use as the basic point of reference the classification by Quirk et al as set out in (2.25)(c) and (d), with two modifications: (1) the 'less common idioms' under (c) will be omitted from consideration on the simple grounds that they axe noncentral; and (2) need (to) (this denotes the forms need and/or need to) and ought to will be retained from the 'marginal modals', since they are commonly associated with must and should, respectively, while the other marginals dare and used to axe not associated with any standardly accepted modal meanings (cf 2.3.2.4). The items to be investigated are, therefore, the following: (2.26)

marginal modals

modal idioms

semi-auxiliaxies

need (to) ought to

had better would rather would sooner is to have got to

have to be able to be about to be apt to be bound to be due to be going to

be likely to be meant to be obliged to be supposed to be willing to be unable to be unwilling to

The maxginal items need (to) and ought to will not be included in the investigation in 2.3, but axe among the items selected in 2.4. The modal idioms will also not always be included in the discussion in 2.3, for structural reasons. Have to is an obvious anomaly among the semi-auxiliaxies.

2.3 2.3.1

Criteria for periphrastics Syntactic criteria

It is impossible to locate any single syntactic criterion for the definition of a naturallooking class of 'intermediate' items (which will now be referred to in general as 'quasimodals'); but there appeax to be two types of evidence which might support the isolation of a set of such items with a kind of sub-categorial status. Firstly, there are certain details 9

The term 'catenative' is used as in the text by, for example, Matthews (1981:Ch.8), Huddleston (1984:Ch.5.6) and Palmer (1987:Ch.9), but has been used rather differently in the American tradition, eg, by Twaddell (1963) and Haynes (1967). For Twaddell, being catenative is "a property of a construction, not of given verbs" (1963:22), and he applies the term variously to items termed 'catenatives' in the text above as well as to 'decaying modals' such as ought to and to items such as be going to.

19 of syntactic behaviour which may serve to distinguish a set of quasi-modals from full lexical verbs, for example resistance to the adjunction of a òy-phrase or to segmentation. Secondly, certain proposals have been made with regard to the ordering and combinatory possibilities of specific sets of such items. These will now be discussed in turn. 2.3.1.1

Tests of syntactic behaviour

The tests of syntactic behaviour will be taken first. These will involve the semi-auxiliaries and, only very limitedly, the modal idioms. Five tests will be applied to this set of items: these relate to the omissibility of being when initial and of to when final, the possibilities of inserting a qualifier before, or agent after, the central item in a be - to structure, and the availability of existential there constructions. Clearly, judgements of 'possibility', etc, that are involved here are not necessarily categorical, and there is considerable interpersonal variability, but it is assumed that the cumulative set of findings will have some indicative value. (1) The first test concerns the possibility, with be + adjective/participle + io-infinitive structures, of dropping being when in initial position in a sentence. Quirk et al (1985:144) suggest that the availability of this option would indicate that an adjective/participle was being used in a normal, literal sense, while its non-availability (ie, its resistance to segmenting) would point to a special, idiomatic sense, and be typical of a semi-auxiliary. They cite the contrast between the following: (2.27) (2.28)

Compelled to take stern measures, the administration lost popularity. ?Bound to take stern measures, the administration lost popularity.

as evidence that be bound to, unlike be compelled to, forms a syntactic unit; and they use the following contrast: (2.29) (2.30)

Unable/Unwilling to resist, Matilda agreed to betray her country. ?Able/?Willing to resist, Matilda declined to betray her country.

to justify assigning doubtful status to be unable to and be unwilling to among the semiauxiliaries in (2.25)(d). Rather similarly, Downing & Locke (1992:316) distinguish a set of 'lexical auxiliaries', one subset of which has the structure be - io-infinitive. They assume a test such as is used here, and their findings, to the extent that the same items axe involved, correspond to those in (2.32) below, apart from their exclusion of be likely to from the 'possible' class. If this test is applied to all the semi-auxiliaries (except, of course, have to), whether in the frames used in (2.27)-(2.28), in (2.29)-(2.30), or in the following: (2.31)

- to leave the party early, they felt perfectly satisfied.

the result is as follows (in tests (l)-(2) it is difficult to determine degrees of nonacceptability, so 'very odd or impossible' is used as a blanket phrase):

20 (2.32)

possible in initial position without being

very odd or impossible in initial position without being

be be be be

be be be be be

likely to obliged to unable to unwilling to

able to about to apt to bound to due to

be going to be meant to be supposed to be willing to

(2) We shall next consider omission of final infinitival to in the use of these items. The contrast in question is shown in the following pair; (2.33)

He's keen to leave, but isn't able.

(2.34)

*He's keen to leave, but isn't about.

Any distinction demonstrable here should point to the degree to which a semi-auxiliary is idiomatic and thus not segmentable; the two modal idioms that end in to axe also included in this test. The following shows the findings, within the frame He's keen to leave, but isn't/hasn't -, modified in some cases as noted: (2.35)

possible in final position without to

very odd or impossible in final position without to

be be be be be

is to have got to have to11 be about to be apt to14 be bound to

able to obliged to unable to10 unwilling to12 willing to13

be due to be going to be likely to be meant to be supposed to

(3) The third test concerns the possibility of qualifying the adjectival item in semiauxiliaries by, for example, very; other items that function similarly include more, most and quite. (This cannot, of course, be applied to have to.) The contrast in question is shown in the following pair: (2.36)

He's very willing to organize people.

(2.37)

*He's very able to organize people.

In this kind of context, the items can be divided in the following way: (2.38)

10 11 12 13 14

As As As As As

in in in in in

accept very qualification

do not accept very qualification

be be be be

be able to be about to be bound to be due to be going to

apt to likely to unwilling to willing to

He's keen to leave, but is un able. *Ee's keen to leave, but doesn't have. He's obliged to leave, but is unwilling. He's obliged to leave, but isn't willing. *He's always keen to leave early, but isn't apt.

be be be be

21 (4) The next question concerns the possibility of the insertion of an agent before the to in these items. Obviously we are here restricted to semi-auxiliaries that include a passive participle-like form (be bound/meant/obliged/supposed to). The possibilities of inserting an agent phrase with by are as follows (epistemic and non-epistemic senses are in turn illustrated for the items be bound/meant/supposed to, and glosses are provided for these items to distinguish these senses; be obliged to does not have this ambiguity): (2.39a) (2.39b)

He's bound to give the right answer. *He's bound by his knowledge to give the right answer, ('he's certain')

(2.40a)

If you believe this, you're bound to act on it.

(2.40b)

If you believe this, you're bound by your conscience to act on it. ('you're obliged')

(2.41a)

He's meant to arrive at six.

(2.41b)

*He's meant by his friends to arrive at six. ('he's expected')

(2.42a)

The coffee set was meant to be a present.

(2.42b)

The coffee set was meant by his friends to be a present, ('it was intended')

(2.43a)

This is supposed to be the only solution.

(2.43b)

This is supposed by some people to be the only solution, ('this is believed')

(2.44a)

You're supposed to register any change of address.

(2.44b)

*You're supposed by the authorities to register any change of address, ('you're required')

(2.45a)

He was obliged to settle for early retirement.

(2,45b)

He was obliged by his employers to settle for early retirement.

The items be bound to, be meant to and be supposed to each have one case where agent insertion is unacceptable - (2.39b), (2.41b) and (2.44b), while no such restriction is evident with be obliged to. The fact that the three first items allow agent insertion in the other sense - once epistemic and twice non-epistemic - complicates the issue; however, it simply appears that with these items one use is more firmly established and thus shows more evidence of idiomaticity than the other. This evidence suggests a distinction between be bound to, be meant to and be supposed to, and the other forms under discussion. (5) The last test concerns the possibility of constructions with existential there. Quirk et al (1985:127) mention this as one of a small number of ways in which auxiliaries, in general, may display their semantic independence of the subject (cf (2.22)(h)(ii)). The examples used are:

22 (2.46)

There used to be a school on the island.

(2.47)

*There hoped to be a school on the island.

If this test - using the frame There - be a school on the island - is applied to both the modal idioms and the semi-auxiliaries, the following emerges (in this case, a fairly clear distinction between 'possible', 'doubtful' and 'impossible' items can be made, although, ultimately, we must assume a gradient here) 15 : (2.48)

use of items in existential iAere-construction: doubtful possible had better is to have got to have to be about to be apt to

be be be be be be

bound to due to going to likely to meant to supposed to

be able to be obliged to be unable to

impossible would rather would sooner be willing to be unwilling to

Conclusion. General consideration of the findings of these tests suggests that the items listed below show a marked degree of idiosyncracy in their syntactic behaviour. The criterion adopted here is that they fail to behave 'normally' (ie, non-idiosyncratically) on at least three of the five tests. The sole exception among the semi-auxiliaries here is have to, which differs from all the other items in not starting with be and not having a potentially adjectival/participial middle element, and thus in only being eligible for two of the tests. The modal idioms are very restrictedly available for such tests, but their syntactic idiosyncracy is obvious. (2.49)

2.3.1.2

modal idioms

semi-auxiliaries

had better would rather would sooner is to have got to

have to be able to be about to be apt to be bound to

be due to be going to be meant to be supposed to

Restrictions on the ordering and combination of items

In this section I shall examine the possibility of defining a class of quasi-modals in terms of ordering and combinatory restrictions within sentence structure. This will refer closely to the accounts in Hakutani & Haxgis (1972) and Chapín (1973), which appear to be unique in this area. It should be emphasized that our concern here is primarily to establish a set of quasi-modals, rather than to determine the restrictions that may apply to them. (1) Hakutani h Hargis (1972) seek to define a set of 'quasi' items and accommodate them within a modified version of the traditional auxiliary phrase-structure rule, as in Chomsky 15

Postal's list of 'R-triggers' that permit raising (1974:292-294) is also relevant here; this includes six of the items given as 'possible' in (2.48).

23

(1957:111). While this attempt must be considered a failure, it usefully highlights some of the prominent descriptive problems in this area. The proposed expansion of the Aux rewrite rule is as follows: (2.50)

Aux — • Tense (M)(Q) n (have+en)(Q) n (be+ing)(Q) n

where 'M' denotes modals and 'Q' quasi-modals (Hakutani & Hargis 1972:32016). This rule is shown to generate not only such relatively simple sentences as the following (the part of the Aux structure following 'Tense' is in each case shown in brackets) (315): (2.51) (2.52) (2.53)

John may have to go. (M, Q) He was to have been there. (Q, have+en) Mary has to be leaving. (Q, be+ing)

but also more complex examples such as the following (318-319): (2.54) (2.55)

I might be able to be going to have to have been about to be having to give up. (M, Q, Q, Q, have+en, Q, be+ing, Q) I might be going to be going to have to have been about to be having to resign. (M, Q, Q, Q, have+en, Q, be+ing, Q)

These examples suggest that the rule in (2.50) deals correctly with at least some of the positional and combinatorial possibilities of modals and quasi-modals. Thus, the following, which contravenes the ordering of elements in (2.50) in various ways, is obviously ill-formed: (2.56)

*The children have to might being able to have slept. (Q, M, be+ing, Q, have+en)

There are, however, a number of very serious problems with this (and no doubt any such) proposal. Three areas will be discussed: (a) the membership of the categories M and Q, (b) selectional restrictions required for Q, and (c) the basic validity of a rule like the Aux rule in (2.50). Membership of M and Q. The category M includes the central modals, together with the marginal modals, as in (2.25)(b), and some of the modal idioms listed under (2.25)(c). Q includes five of the items under (2.25)(d), together with is to. Here a few difficulties arise. If is to is treated as a quasi-modal, then sequences like ought to be to or used to have been to would have to be grammatical, which is wrong. It is not, in any case, clear why is to should be assigned in this account to Q rather than to M, since all the examples in the text have it in the first position after Tense, where it could, according to (2.50), fill either M or Q. 17 16 17

Page references in the ensuing discussion are to this paper unless noted otherwise. Is to is rather problematic among the 'intermediate' items, since it lacks non-finite forms while being the only modal idiom in (2.25)(c) to begin with the verb io be, which is not so restricted. Cf Footnote 7 above.

24 There is a further problem with have got to, which is lacking from both lists. Presumably this item is treated in this account (as it usually is) as a variant of have to; but if this is the case, it would result in the acceptance of non-finite forms such as to have got to, which, like be to (from is to), are standardly considered missing. Finally, be bound to is also lacking from Q, whereas be able to, be unable to and be going to are included on the grounds that "the words able, unable and go (as used in the constructions) are semantically differentiated from their respective lexical items" (314). This would suggest that be able to, etc, are idiomatic units, but not be bound to, which seems an odd conclusion. Questions of this kind point to descriptive inadequacies, but no less to the inherent difficulty of establishing any sort of clear-cut syntactic category of quasi-modal. Selectional restrictions for Q items. It is stated that "the degree of [semantic] acceptability decreases if a quasi-modal containing be, like be going to, is selected in place of have to" (317). This is a strange claim, since, first, all the Q items listed contain be except have to, and, second, 'reduced acceptability' is illustrated in the text by examples like: (2.57)

John might have been being to run.

which is surely impossible, whether the being is interpreted as part of the progressive group might have been being ( ... silly, etc), or - as is presumably intended - as a form of is to, which is generally agreed to have no non-finite forms. Further, the examples given above as (2.54) and (2.55) are defended in the following ways: (2.54) is "a grammatical sentence, however extremely rare it is", while (2.55) "from a semantic point of view . . . cannot be rejected simply as 'illogical' or 'meaningless'" (320). But these are hardly satisfactory ways of dealing with obviously very odd sentences. These cases highlight the question of the grammaticality of such complex sentences in general. It does not seem plausible that there can be either any purely syntactic constraints or any useful selectional restrictions on the combinatory possibilities of sets of items such as Hakutani & Hargis list under Q (or are listed above in (2.25)(c) and (d)). Even the authors' unwillingness to reject (2.55), with its adjacent repetition of the same Q item be going to, must be supported in principle: there seems to be no absolute block on the adjacent repetition of forms of Q items, as in the following: (2.58)

?It's an awful nuisance having to have to be here at 7.

In this connection it is significant that Quirk et al (1985:147) claim that "chain-like structures" with not only catenatives but also semi-auxiliaries and full verbs, and obviously lacking any absolute constraints, such as: (2.59)

We are going to have to enjoy seeming to like listening to his music.

are "improbable, but structurally possible". All of this points to something like the following: there are purely pragmatic constraints, not to mention processing problems, on sentences like (2.54) and (2.55), while (2.56) and (2.57), for example, can simply be rejected as ungrammatical - (2.56) since it has might fulfilling a non-finite function, which

25 modals cannot do, and (2.57) since it has the verb is to in a form standardly treated as non-existent. The new Aux rule. The final problem with the Hakutani & Hargis account concerns the basic rule set out in (2.50). It was suggested above that the only valid constraints on sentences like (2.54)-(2.55) can be pragmatic; but this would imply that at least the rule was correct. Let us now consider the following invented sentence, uttered or written in a situation where membership of a club increasingly requires football expertise: (2.60)

More and more people are having to have been able to play football when they were young before being accepted as members.

The first auxiliary structure here is 'be+ing, Q, have+en, Q', which includes 'be+ing' and 'have+en' in the reverse position of that stipulated by (2.50); yet the sentence seems better than (2.54)-(2.55), which follow the proposed rule. Schachter (1983) discusses the problems of auxiliary order, in terms of the original Aux rule, and, more specifically, the question whether the order normally stipulated is absolute and arbitrary, or independently motivated. His conclusion is that the standard order of elements is determined by the interaction of a basic principle of English word order (head before dependants) with the semantic properties of the individual auxiliaries involved. He offers an example of possible progressive + perfective order in the auxiliary (1983:161): (2.61)

?Whenever I see you, you're always just having returned from a vacation.

(Schachter 1983:161). What is important here is not whether examples such as (2.60) and (2.61) are 'acceptable' so much as Schachter's general argument, which suggests that auxiliary word order is ultimately a pragmatic matter, and that rules like (2.50), which possibly represent a quantitatively valid observation, have no absolute value. (2) Chapin (1973), similarly, attempts to provide syntactic grounds for distinguishing a group of quasi-modals (this term is used again, though not for quite the same phenomena). He argues that the set have to, need, be able to, be about to, be going to and be supposed to, with, in some dialects, used to and ought to, form a well-defined closed system, functioning, unlike true predicates, as clausemates in underlying structure and thus, when combined, not separated by clause boundaries. Two supporting arguments are particularly relevant to our discussion, and will be investigated here. These are that (a) only one relative order for any adjacent pair is possible, and (b) repetition of a quasi-modal is impossible. Fixed relative orders. For the first claim, data such as the following axe used: (2.62)

Joe has to be able to think.

(2.63)

*Joe is able to have to think.

(2.64)

Al is going to have to say grace.

(2.65)

*A1 has to be going to say grace.

(2.66)

He used to be supposed to be about to need to have to be able to jump.

26 In all these cases the deontic sense is specified for have to; obviously, (2.65) is normal when has to is understood epistemically. The basic question here is whether there is an absolute order of elements, reflected positively in (2.62), (2.64) and (2.66), and negatively in (2.63) and (2.65), or whether any apparent ordering restrictions are determined pragmatically. For (2.65), one of the supposedly ungrammatical examples, a completion such as . . . when the guests arrive, if the trick is to work seems to make the sentence perfectly natural in the required deontic sense; if it looks odd as it stands, this is simply because we rarely state requirements for ail action that is to follow a specified future event after a further interval, as opposed to one that is seen as immediate. There is a further problem in the stipulation that has to in these examples be deontic since (a) this would seem to mean allowing alternative ordering rules for the quasi-modals according to their senses, which would considerably complicate the system proposed, and (b) such a procedure would require that the deontic and epistemic senses of items 1 8 correspond to semantically distinct entities, whereas often a unitary, general sense, to be interpreted pragmatically, appears more appropriate (cf 3.5.2). (2.66), which is designed to present decisive proof of the basic order of quasi-modal items, and does, in fact, include all the items Chapin lists as 'regular' quasi-modals, together with used to, which is treated as a dialectal variable, but excluding be going to (which, according to Chapin's examples, would have to go somewhere between used to and have to), presents a more complex problem. Used to certainly has to stand first; but there is no indication that the order of the other items is immutable: for example, the permutation of need and be about to produces a sentence no worse than (2.66), while a reshuffle of the items, as in: (2.67)

He used to need to have to be supposed to be able to be about to jump.

is about as, but certainly not more, odd, pragmatically, than (2.66). Moreover, there is no evident restriction on mixing the 'quasi-modals' with other items, specifically catenatives in the sense of (2.25) (e), as in: (2.68)

He used to want to need to have to seem to be supposed to be about to . . .

Quasi-modal repetition. The second argument concerns a block on the repetition of a quasi-modal. Chapin offers: (2.69)

*Mort is able to be able to jump.

This is said to be grammatically blocked in a way that is different from any block on: (2.70)

*Jake wants to want to spit.

on the grounds that with true predicates, which lack ordering restrictions, sentences such the following axe possible: (2.71) 18

Al tried to begin to try (to begin) to work.

Chapin specifically refers to a three-way contrast between a deontic sense of have to, as 'obligation of the subject', an epistemic sense and a sense of pragmatic necessity (as in It has to rain today if the plants are to live) (1973:4); cf 3.2.

27 Further, the following is shown to be possible: (2.72)

Hank has to expect to have to jump.

because a full verb (expect) intervenes between the two occurrences of have to, and thus the occurrences are in different clauses. This means that, grammatically, the alternating repetitions of try and begin in (2.71) can continue without restriction, while (2.66), assuming that this includes the complete set of quasi-modals in its unique order, cannot be extended with any of those items. To examine this, let us consider the following: (2.73)

He always has to be supposed to be about to have to be able to do a serious job in order to appear to be doing his best.

Here the quasi-modals have to and be about to appear both in the reverse order of the required order illustrated in (2.66) and with have to repeated within the same clause, but the sentence, again, does not appeax worse than (2.66). Conclusion. It is reasonable to compare these two accounts, since Chapin's set of items, apart from be supposed to, is included in Hakutani ¿z Hargis's; but they are essentially contradictory in their findings. Thus, Hakutani & Hargis's PS rule, in (2.50), allows any order of elements under Q, and allows juxtaposed repetition of a Q item, as was shown in (2.55) with its double occurrence of be going to, while in both these cases, Chapín axgues for the precise opposite. There seem, then, to be such problems with these accounts, individually and in combination, that we are forced back to the position taken by Quirk et al (1985:147), who reject any grammatical restrictions on the occurrences of quasi-modals, alongside other 'intermediate' items, and accept "improbable, but structurally possible sequences" like that in (2.59) above.

2.3.2

Semantic criteria

I shall now discuss five semantic criteria that are of potential relevance to the establishment of a set of periphrastic items. The first two derive from observations made by Quirk et al (1985:126-127) as evidence of the tendency of auxiliaries to be semantically independent of their subjects (cf (2.22)(h)(i) and (iii)); here they will be applied to the modal idioms and semi-auxiliaries as set out in (2.26).

2.3.2.1

Lack of subject-auxiliary restrictions

Quirk et al (1985:127) give the following examples to illustrate the difference between auxiliaries and main verbs in their dependence on their subjects:

28 (2.74)

The man ought to be here at five.

(2.75)

The bus ought to be here at five.

(2.76)

The man hopes to be here at five.

(2.77)

*The bus hopes to be here at five.

What is significant here is the normality of (2.75), showing that the auxiliary ought to is not dependent on the bus in the way that the main verb hopes is in (2.77). If the modal idioms and semi-auxiliaries are tested in the frame The bus - be here at five, the following emerges: (2.78)

subject-auxiliary dependence: unrestricted had better is to have got to have to be about to be apt to

be be be be be be

bound to due to going to likely to meant to supposed to

doubtful

restricted

be able to be obliged to be unable to

would rather would sooner be willing to be unwilling to

The 'unrestricted' and 'restricted' classes here axe clearly distinct, but the classification of 'doubtful' items is difficult, as will now be shown. One dictionary definition of be able to starts "If a person or thing is able to do something . . . " (Sinclair 1987), but strongly suggests personal, or animate, ability ('thing' is in fact only illustrated by a citation involving a frog). By contrast, Coates (1983:124-125) states, with illustrations, that ube able to covers the entire range of meaning covered by can", which clearly rules out any such restriction. Palmer, more cautiously, suggests that be able to is subject oriented, without being restricted to 'ability' senses; the following is an illustrative text (1990:88): (2.79)

In the past, we've had small seminars in our rooms and these were obviously not able to contain them.

There seems, in fact, to be a cline of availability for the use of be able to in senses possible for can: at one end axe examples like Lions can be dangerous (showing existential modality, cf Palmer 1990:107), where be able to is excluded; between are cases such as The bus can do the journey in 20 minutes, where be able to is possible though not normal; and at the other end are cases of explicit ability, as in He can/is able to touch the ceiling now. The restriction involved here seems to lie in the ease with which subjects like the bus or our rooms can be treated as agent-like entities. Be unable to requires no special comment. In the case of be obliged to, also treated here as a doubtful case, Jespersen (1940:323) states that inanimate subjects axe possible because of the semantic equivalence of this item and must. This comment suggests that the lack of total restriction on subject-auxiliary dependence shown in the 'unrestricted' and 'doubtful' lists in (2.78) reflects semantic identification of an item with a modal, which was stipulated at the beginning of 2.1 as one of the three features periphrastics would ideally be expected to display.

29 2.3.2.2

Voice neutrality

Quirk et al (1985:127) use the following examples to illustrate semantic equivalence or non-equivalence between active and passive verb forms (cf (2.22)(h)(iii)): (2.80)

Thousands of people will meet the president.

(2.81)

The president will be met by thousands of people.

(2.82)

Thousands of people hope to meet the president.

(2.83)

The president hopes to be met by thousands of people.

Here the point is the semantic equivalence of the first pair (which applies for the will of futurity, but not of volition, cf Palmer 1990:158) but not of the second. Testing of the items in the frames Thousands of people - meet the president and The president - be met by thousands of people produces exactly the same result as for the first criterion, as set out in (2.78). The 'doubtful' category may be illustrated by the following pair: (2.84)

Norbert is able to beat Martyn at chess.

(2.85)

Martyn is able to be beaten at chess by Norbert.

Whereas the ability in (2.84) is clearly that of the subject, in (2-85) the situation is unclear, since interpretations identifying both the agent and the subject (if losing a game is seen as an ability) are possible. 2.3.2.3

Idiomatic status

This criterion concerns the relative semantic transparency of the items under discussion. The status of the five items classified as modal idioms in (2.26) obviously needs no comment in this connection, so only the semi-auxiliaries will be considered here. Of these, have to is an anomaly in the list, since it is not structured be - to\ but, synchronically, to have to do something is semantically distinct from to have something to do, and thus have to can be treated as idiomatic, together with have got to. For the other semi-auxiliaries the test of idiomatic status is whether the item between be and to can be treated as semantically equivalent to its use in other contexts. The following sets out the findings: (2.86)

idiomatic had better would rather would sooner is to have got to have to be about to

be apt to be bound to be due to be going to be meant to be supposed to

doubtful

non-idiomatic

be able to be likely to be obliged to be unable to

be willing to be unwilling to

30 The problem of determining whether particular items are idiomatic or not may be illustrated by the cases of be able to and be willing to: able and willing collocate readily (eg, a person may be able but not willing), but it seems doubtful whether be able to shows able in its full 'ability' sense in the way that be willing to shows willing. On the other hand, it could hardly be claimed that be able to diverges widely from the basic sense of able, so it cannot easily be considered fully idiomatic. 2.3.2.4

Modal meaning

'Modal meaning' is understood in a broad sense so as to cover concepts such as 'necessity', 'obligation', 'possibility', 'permission', 'ability' and 'volition'; 'futurity' is also often treated as a modal notion, so that will be taken into account here too (cf discussion in 3.2). Since we are concerned with possible equivalents to modal verbs, it should be revealing to see which quasi-modal items do in fact have 'modal meanings'. Of the items in question here some can readily be assigned to particulax categories, eg, had better expresses 'obligation', and would rather 'volition'. Others are perhaps not so readily accounted for: be apt to, be likely to, be meant to, be supposed to. However, these can be related to the concepts 'necessity' or 'obligation' in that they express modified versions of these values (be apt to and be likely to concern probability, which can be considered a qualified form of 'necessity', and be meant to and be supposed to concern what is expected, a qualified form of 'necessity' or 'obligation'), so it seems that all these items can be associated with modal concepts as here understood, as follows (these are not, of course, exclusive categorizations) : (2.87)

2.3.2.5

necessity/obligation

ability

volition

futurity

had better is to have got to have to be apt to

be able to be unable to

would rather would sooner be willing to be unwilling to

be about to be due to be going to

be be be be be

bound to likely to meant to obliged to supposed to

Epistemic-non-epistemic alternation

This meaning alternation is a striking feature of some, or (depending on the definitions of the terms) all, modal verbs, but the phenomenon is not exclusive to 'modal meaning' as manifested by modal verbs, since a potential for two related, but distinct, meanings of such kinds has wide incidence in language (cf, especially, Tregidgo 1982, Sweetser 1982 and 1990, and Traugott 1989). Among the items under consideration here, have to and have got to clearly show ambiguity of this kind, as in the following example: (2.88)

He's got to/He has to be in the room when we go in. = (a) I'm confident he will be; (b) he's required to be.

Here the (a) gloss indicates an epistemic sense, and the (b) gloss a non-epistemic one.

31 Such systematic ambiguity is also possible with a few other items, as will now be shown; in each illustrative example, the epistemic and non-epistemic senses, respectively, are indicated in partial glosses. (2.89)

The government is bound to announce the election date soon. = (a) this can confidently be expected; (b) it is under an obligation.

(2.90)

The tunnel is supposed to be closed. = (a) it is said to be; (b) this has been required.

(2.91)

They are meant to offer excellent service. = (a) they are said to; (b) they are intended to. This is going to be the news. = (a) I'm certain it is; (b) it's coming shortly.

(2.92)

The last two items, at least, axe not normally treated as showing epistemic-non-epistemic alternation 19 , but in their potential for systematic ambiguity they are clearly parallel to should and will, respectively. Whether any others of these items show such an alternation is doubtful. Thus, if the two senses of the following: (2.93)

He can't finish the work by three. = (a) it isn't possible that he will; (b) he isn't capable of it.

can each be carried by: (2.94)

He isn't able to finish the work by three.

then be able to should join our list; but sense (a) seems impossible for (2.94). We may briefly consider here why it is that periphrastic items tend to manifest epistemic senses only to a relatively restricted degree. Derived forms of words commonly have only a subset of the readings available for the original form, and thus, for example, in the case of must the nominal use (eg, It's an absolute must) only has a deontic sense (Smith 1989:91); if, as is well documented (eg, Bybee & Pagliuca 1985, and Traugott 1989), epistemic senses axe, normally, secondary developments, and the basic, default, sense of the modal must is deontic, then the derived nominal simply has not extended its readings. This principle could be extended to semantically related forms, so that periphrastics typically have a narrower range of readings than their modal counterparts; this can be illustrated by the epistemic reading for be going to in (2.92), which is much less well established than it is for the close synonym will, as in this will be the news.

19

I know no reference to this phenomenon for be meant to, and for be going to only Haegeman (1981:402404) and Fleischman (1982:91); cf (5.79) and (5.80) and related comments in 5.3.1.1.

32 Two final questions remain here. The first concerns the possible addition of want to to our list of items showing epistemic-non-epistemic alternation. Bolinger (1980:297) argues that potential for "epistemic, aspectual or modal meaning" is a semantic characteristic of modal and modal-like items. 'Epistemic' meaning is illustrated by, alongside other wellestablished items, want to, especially in the wanna form. Bolinger gives the following examples: (2.95) (2.96)

Pick it up with both hands. You wanna be careful it doesn't fall. ?There wants to be a little more circumspection in all this. 20

(1980:295). This claim is strange: a paraphrase of wanna in (2.95) would be something like 'axe advised to', and of wants to in (2.96) 'needs to', and neither of these looks epistemic. Whether wanna can in fact be epistemic is unclear. The following, however, seems perfectly possible in a colloquial context: (2.97)

They want to be pretty stupid if they believe everything he says. = Ί assume they are'.

Want to is not included in our list of 'intermediate items', but the example is revealing in that it provides further evidence that the phenomenon being discussed here is more widespread than is generally assumed. The second issue is rather more speculative. The potential alternations of meaning discussed so fax in this section have contrasted epistemic and non-epistemic uses. However, if this is seen as merely one prominent kind of alternation typical of quasi-modals, then we should perhaps also consider a case like be about to. The temporal use ('to be just ready to') is well established, while a further volitional use has recently gained in usage in BE; the entry in LDOCE (1987:3) runs "not about to infml, esp. AmE very unwilling to", illustrated by: (2.98)

I'm not about to lend you any more money.

Another use of be about to, apparently different from the volitional meaning, and treated, some 30 years ago, as a recent innovation in standard AE, is glossed "the actor is not the sort of person from whom such a deed can be expected" (Joos 1968:24; cf Haynes 1967:174-176). Whether there are two or three distinct meanings of be about to here, the difference between them is surely as great as that between typical epistemic-deontic pairs of meanings.

20

Example (2.96) is termed "not too unusual" - hence the '?' in the text.

33 2.3.2.6

Conclusion

Of these five semantic criteria of possible relevance to the determination of a set of periphrastic items, the fourth is essentially redundant, since it was shown that it is not difficult to predicate 'modal meaning' of all the items in question. However, if we take the remaining four, the following list shows which items count positively (ie, show idiomatic features) on at least three criteria: (2.99)

modal idioms

semi-auxiliaries

had better is to

have to

be due to

be about to be apt to be bound to

be going to be meant to be supposed to

have got to

The case of be able to should be noted: this item was found to count as 'doubtful' on all four criteria.

2.3.3

Phonological criteria

In this section I shall explore the potential of the items being discussed here for abnormal phonological reduction. By this is meant reduction phenomena that are normally only found in 'rapid familiar' style, but with certain items regularly occur even in 'careful colloquial' speech (to use the terminology of Gimson 1980:297-298). These phenomena do not appear to have been investigated in detail for BE; comparable phenomena in AE, however, have been widely discussed, and they will be mentioned briefly at the end of this section. The main type of phenomenon in question here concerns the reduction to a single / t / of a pair of homorganic stops at the juncture between the verbal element and to in items like have got to. In each case I shall contrast a specific item with a phonologically similar 'normal' item, always in the context of a phrase or sentence. After noting each pair to be contrasted, unreduced forms and then reduced forms are shown. The first, unreduced, forms are possible but exceptional for the quasi-modal items; the reduced forms are possible and, in fact, normal for the quasi-modal items but not for the other items. The following items axe considered here: need, ought to, used to, have got to, be bound to, be supposed to. The first three count for Quirk et al (1985:137) as marginal modals and are not in the lists being discussed here in 2.3; they are included because they illustrate the kind of phenomenon in question. 21

21

For transcriptions of BE, I follow Jones (1977), and for AE I retain the systems used in the sources cited.

34 (2.100)

(2.101)

need to - cf lead to we n e e d t o h a v e a b r e a k

t h i s will lead t o disaster

... /ni:dta/...

... /li:dta/...

... /ni:ta/...

... */li:ta/...

ought to - cf taught

to

h e o u g h t t o sing

(2.102)

h e was t a u g h t t o sing

... /o:tta/...

.. /to:tta/...

... /o:ta/...

.. */to:ta/...

used to - cf used to ( f r o m to use) he used t o s m o k e

(2.103)

(2.104)

(2.105)

it was used t o clean t h e air

. . . / j u : z d t a / or / j u r s t t a / . . .

.. /ju:zdta/...

... /ju:sta/...

.. * / j u : s t a / . . .

have got to - cf got i o ( ' m a n a g e d t o ' ) I've got t o see t h e play

I got t o see t h e play

... /gotta/...

... /gDtta/...

... /gota/...

...•/gota/...

be bound to ( ' c e r t a i n ' ) - cf be found

to

h e ' s b o u n d t o c o m e soon ... /baondta/... ... /baunta/...

t h i s was f o u n d t o b e false ... /faondta/... ... */faunta/...

be supposed

to - cf be exposed

to

he's s u p p o s e d t o d r i n k less ... /s(a)paozdta/... ... /spaosta/...

he's exposed t o d r i n k i n g ... /ikspaozdta/... ... */ikspaosta/...

In c o n t r a s t t o t h e i t e m s have got to, etc, t h e i t e m s be about to, be apt to, be meant

to

a n d be obliged to d o n o t s e e m t o allow t h i s kind of r e d u c t i o n ; t h u s , they were meant to go does n o t r e d u c e t o . . . / m e n t a / . . . , which would suggest they were men to go. J u n c t u r a l r e d u c t i o n a n d assimilation also occur in specific, again a b n o r m a l , environm e n t s in t h e t w o i t e m s be going to a n d have to: (2.106)

(2.107)

be going to - cf be going to ( m o v e m e n t ) I ' m going t o r e a d t h e m

I ' m going t o R e e d h a m

/aimgaun)tari:dam/

/ a i m g a w r ) tari: d a m /

/aii)ganari:dam/

*/air)ganari:dam/

have to - cf love to I ' d h a v e t o go ... /haevta/... ... /haefta/...22

22

I ' d love t o go ... /Livta/... ... */kfta/...

Cowper (1992:122-123), surprisingly, treats this contraction of have to as obligatory, in contrast to the optionality of the want to contraction (/wana/), which involves contrasting uncontracted */haev tuw/ with uncontracted /want tuw/. There is no reason for assuming BE-North American English differences here.

35 Aside from these cases, the only other phenomena that could be singled out concern had better and have got to, where the first, auxiliary elements, in addition to reducing to / d / and / v / , respectively, regularly disappear altogether, leaving /beta/ 2 3 and /gota/, with a phonological shape similar to that of /o:ta/ (ought to). The data reported above have limited value in that not all potential periphrastics are likely to undergo abnormal reduction: for example, be able to /bieiblta/ does not provide suitable loci for marked phonological reduction (beyond the regular weakening in /bi/ and /ta/). This account has been concerned with BE so fax. In AE there is a comparable set of phenomena, which have been extensively discussed, with particular reference to the case of wanna. A list of verb + to items showing junctural reduction, apparently complete and thus pointing to a closed set, is given by Postal L· Pullum (1978:2), as follows: want to /wäna/, be going to /gana/, have to /haefta/, ought to /o:ra/, used to /yu:sta/, have got to /gara/ and be supposed to /s(a)posta/. As can be seen, there is some degree of correspondence between this listing and the findings above for BE; the crucial differences between them are simply that want to is missing in the BE list, and be bound to in the AE one. Regarding want to, complete loss of the /1/ only occurs in BE in very rapid or casual speech, and the non-standard status of the /wona/ type of pronunciation is reflected in the (perhaps rather exaggerated) claim that the wanna spelling regularly implies a negative judgement by the writer of the speaker being represented (Kress & Hodge 1979:66-67). In any case, want to is absent from the lists of potential periphrastics in the present study. As to be bound to, Bolinger in fact notes that this item functions in the way being discussed "for some speakers" (1980:293), so variable usage in AE must be assumed. The relevant items that have been shown to manifest abnormal reduction here are the following: (2.108)

2.4

had better, have got to, have to, be bound to, be going to, be supposed to.

Conclusion

The findings of the preceding sections 2.3.1-3 are summarized in (2.109). In addition to this, a column is added to show when an item is regularly treated as equivalent or related to a particular central modal. The inclusion of an item in column d is based on general assumptions regarding semantic equivalence, and constitutes the basis of the comparisons in the following chapters. 24 23

24

In this connection, it is noteworthy that Palmer attests the occurrence of beitern't he? (/betnt hi/) in children's speech (1974:165). Reference works and grammars do not normally deal with this issue to any extent; exceptions are the lists in Todd & Hancock (1986:395), with can-be able to/get io, may (deontic)-Zei me, may (epistemic)-seem to, must-have (got) to, should-had better and will-be about to/be going to/is to; and Rutherford (1975:65, 1977:140), with can-be able to, may-be allowed to, must-have to, should-be supposed to and will-be going to.

36 (2.109)

summary of findings in 2.3 verb

criteria a

b

e

d

modal idioms had better would rather would sooner is to have got to

+ +

+

+

-

-

should -

+ -

-

+ +

+ +

— · must/will + must

-

+

+

+

must

+

+

+

will

semi-auxiliaries have to be able to be about to be apt to be bound to be due to be going to be likely to be meant to be obliged to be supposed to be willing to be unable to be unmlling to

+ — — can + + - + + - + + + must + + - + + — — + + — — — — — —

-

— must + should — will — can't — won't

explanation of criteria: a - syntactic criteria, as summarized in (2.49) b - semantic criteria, as summarized in (2.99) c - phonological criteria, as in (2.108) d - full modal commonly considered to be equivalent or related: this reflects both the description in Palmer (1990) and the implications of the conceptual representations of the modals and 'marginal auxiliaries' in Quirk et al (1985:221,236); cf also Footnote 24

Examination of (2.109) shows that the following items qualify on all three criteria, and in addition axe associated with a central modal: had better, have got to, have to, be bound to, be going to, be supposed to. These clearly have a primary claim to be considered periphrastics, in the sense set out at the beginning of this chapter. Of the other items, be able to and be willing to are included in the description on the grounds that they are regularly treated not only as being semantically close or equivalent to central modals, but also as having suppletive functions. It was mentioned earlier in this chapter that need

37 (to) and ought to are to be included because of their especially close relationships to must and should, respectively. In addition, is to will, somewhat arbitrarily, be included since it is fairly idiomatic, and relates to both must and will. The items be unable to and be unwilling tò are not discussed separately. On the basis of the above discussion, the following groups of semantically related items will be discussed in the descriptive chapters: (2.110)

must, have to, have got to, need (to), be bound to (4) should, ought to, be supposed to, had better (5.2) will, be going to, is to, be willing to (5.3) can, be able to (5.4)

The discussion from now on will focus on these items and groupings.

3.

Approaches to this area

3.1

Introduction

In this chapter the major approaches that have been adopted in treatments of the general area of M-P alternation will be reviewed critically, and the approach to be followed in the present work explained. The ultimate aim is to establish a framework for the subsequent description which is as well motivated and soundly based theoretically as possible. However, it is quite clear that there are a number of major problems facing such an endeavour. Firstly, a very considerable vaxiety of descriptive notions have been applied in this area, and, as will be seen, among these there are a number of cases where unrelatable or conflicting findings are involved. Further, while some proposals seem to relate only to one pair or set of items, others claim, or appear, to have wider application - obviously, the latter axe more relevant to our discussion. Finally, an adequate account of this area must be maximally responsive to variation in use according to style. Before going further, I shall try to clarify precisely what kinds of claims could in principle be made in this area. Three aspects of semantic contrast between potential M-P pairs or sets can be distinguished here: (3.1)

(a) (b) (c)

the degree of equivalence, the nature of the relationship, and the extent of application.

This can be illustrated briefly by what seems to be the only comment on M-P alternation of general relevance to be found in the traditional scholarly grammarians. Jespersen (1931:4) gives the following list of modals together with the "substitutes used for the missing perfects" : (3.2)

he he he he he he

can: may: must: ought to: shall: will:

he has been able to he has been allowed to he has had to, has been bound to, has been obliged to it has been his duty to (no substitute) he has been willing to

As to possible distinctions in meaning between these modals and their periphrases (ie, between can and be able to, etc), Jespersen states (ibid) "these substitutes do n o t . . . represent exactly the same shades of meaning as the verbs themselves". In terms of (3.1), Jespersen is claiming that (a) the degree of equivalence is less than complete (cf "shades of meaning" ); (b) there is no indication of the nature of the difference; and (c) five groups of items are involved (treating it is one's duty to as a, periphrastic). I shall now consider these three dimensions of possible M-P contrast in more detail. The degree of equivalence between pairs or sets of items - aspect (a) - may not be absolute in two respects. Syntactically, to take can and be able ¿o as an example, can,

39 being a modal, lacks non-finite forms, so that it cannot be fully interchangeable with all the forms of be able to. Semantically, in comparing these items we may be focusing on the 'ability' sense of can, while the use of be able to is not being restricted in this way, so that alternation between the items may involve some of the uses of can, and all the uses of be able to. This can be illustrated as follows: (3.3)

can

be able to

Here, X shows the syntactic-cum-semantic availability of 'ability' can, Y that of be able to, and Ζ the area of overlap, where both items are possible. Relevant questions here are (i) how far Ζ represents total semantic equivalence between the items, and (ii) how far Y can be considered to function suppletively - ie, how far that area can be considered to cover ground missing in X . Examples (1.13) and (1.14), here repeated, are sufficient evidence that there is not always full semantic equivalence (1.13) (1.14)

*I ran fast, and could catch the bus. I ran fast, and was able to catch the bus.

Further, consideration of the following two pairs: (3.4) (3.5) (3.6) (3.7)

I can go there any time I like at the moment. I am able to go there any time I like at the moment. ?I can go there when I'm older. 1 I'll be able to go there when I'm older.

suggests that while can and be able to may be interchangeable in present/general time reference situations, be able to functions suppletively for can in a future-related context. In contrast to such a case where equivalence may, in an appropriate context, be absolute, one can also envisage a minimal degree of equivalence (perhaps in terms of cognitive synonymy, cf 1.5) which would involve be able to simply being treated as a rough paraphrase of can for some of its uses; in this case there is no special interest in M - P pairs (or groups) since the degree of equivalence or interchangeability would be similar to that which applies between lexical items and their glosses in dictionaries. The nature of the relationship between items - aspect (b) - is the crucial question of this study, and will be the subject of the rest of this chapter. It is clear that a wide variety of types of relationship may be involved, and a major question will then be whether, and how, these can all be related together or subsumed within some general framework. The extent of application of any distinction identified - aspect (c) - would ideally be across all the M - P pairs and sets. Many proposals or comments have only concerned a

It has to be specified that the ability sense is intended here, since (3.6) is natural in a general 'possibility' sense, as in Well, if it's impossible now, I can always go there when I'm older.

40 single set of items (most commonly must and have (got) to), and wide application has rarely been claimed; and this matter is, naturally, complicated by the problems involved in determining which modals are to be related to which periphrastics. The present study will devote primary attention to the must-have

(got) to group, because these items have

always attracted much the most interest, their interrelations are rather complex, and they suggest possible parallels in other pairs and groups. Generalizations across sets of M - P items can, clearly, only be made with great caution; but such conclusions as seem justified will be set out in 6. The rest of this chapter will be organized as follows. Firstly, two basic sets of concepts that are widely used in treatments of this area of the grammar will be examined in detail: in 3 . 2 I shall discuss the main types of modal meaning posited in accounts of modality, together with their related terminology, with particular attention to the deontic/rootepistemic dichotomy, and in 3 . 3 turn to the notion of a pervasive subjective-objective dichotomy in the use of modal verbs that is reflected in M - P alternation. In 3 . 4 some generative approaches to the area of M - P alternation will be explored: these should have the advantages of both setting the question within the dominant theoretical paradigm, and combining the syntax and semantics in a principled manner. In 3 . 5 , in marked contrast, some pragmatic approaches will be explored, which will involve concentrating on issues such as the interpretation of polysemy and ambiguity but will take little account of syntax as such. I shall try to justify the adoption of a framework of such a kind for the description in the present study, and show how it is to be applied. In the following three sections I shall then discuss various specific aspects of this area that have been found relevant to M - P alternation - in 3 . 6 , the notion that implied/implicated actuality is a potential distinguisher in certain cases such as that illustrated above in the contrast between (1.13) and (1.14), together with related proposals; in 3 . 7 , the possibility that style and variety are significant factors; and finally, in 3 . 8 , the general notion of markedness, with modals seen as maximally unmarked items, as a central organizing concept. These discussions will then be summarized, and conclusions drawn, in 3 . 9 .

3.2

Types of modal meaning and related terminology

An epistemic vs non-epistemic/deontic/root distinction in the use of modal verbs is widely assumed, and was involved in the discussion in 2 . 3 . 2 . A distinction of this kind is, in fact, one of two oppositions that are prominent in discussions of modality, the other being that between 'subjectivity' and 'objectivity' (to be discussed in 3 . 3 ) . Various other terms, in particular 'dynamic,' are also used to denote types of modal meaning; and since there is considerable variety of practice in the use of such terms in the literature, I shall here briefly set out how they are generally understood, indicate some of the problems inherent in such types of classification, and show how they are to be used in the present study. There is general recognition that a basic distinction can be made between epistemic and non-epistemic modal meaning in languages (cf Palmer 1986:18-21, and the literature cited there; also, in relation to English, Huddleston 1984:164-176).

Epistemic

modality

41 (hereafter 'EM') appears to be more easily defined than other types: in Lyons' words (1977:797), it involves (3.8)

Any utterance in which the speaker explicitly qualifies his commitment to the truth of the proposition expressed by the sentence he utters, whether this qualification is made explicit in the verbal component . . . or in the prosodie or paralinguistic component.

(This, appropriately, makes it clear that the contrast is not exclusively related to modal verb use; cf comments in 2.3.2.5.) Alternatively, from another perspective, EM (3.9)

characterizes the actuality of an event in terms of alternative possible situations, or worlds. (Chung & Timberlake 1985:242)

EM is illustrated in the (a) glosses for (2.88)-(2.93) above, and there axe two clear-cut epistemic occurrences of may in the following: (3.10)

he will send it by the end of June, that's what he says, now, he may not send it quite as soon as that and it may take a hell of a long time to come if he puts it into the diplomatic bag (CEC:35/79-84)

The use of may here concerns what is possible in alternative, conceivable circumstances, and contrasts with what is actually known, and, in this case, what a person claims he will do; it is distinguished from the use of may for the granting and requesting of permission, as in: (3.11)

a

yes, now, this is very interesting, because we haven't thought of this sort of A I've gots lots and lots of them a intermediate category - may we borrow these? (CEC:413/160-162)

The following use of may illustrates the possibility of ambiguity between these two readings: (3.12)

Part 3 is postgraduate as far as the academic level is concerned - you may read it as an undergraduate, but it's beyond the degree course (CEC:168/979-982)

It is not clear whether the speaker means 'you are allowed to read it' or 'you possibly do/will read it', and it is doubtful whether the difference matters in this context. The important thing here is that two separate interpretations are possible, even if in this particular instance the distinction between them is lost or obscure. Certain aspects of EM in English are unclear or axe disputed in the literature. Firstly, it is not agreed how widely the category extends, and hence what its modal exponents are. A definition like that in (3.8) seems to allow a range between, and excluding, total commitment and total non-commitment to the truth of a proposition. Total commitment

42 is standardly expressed by a declarative sentence, and an item like epistemic must expresses a high but less t h a n total degree of commitment; similarly, while the item may expresses acceptance of the possibility of a proposition being true, non-commitment might be considered to go below t h a t , and to be expressed by, eg, I don't know whether ... . Most accounts of E M allow a range of modal exponents apart from must and may, such as will, should and can't. I shall not discuss questions concerning the exact semantic relations of these items to one another: I assume that they signify distinct, not necessarily relatable, values or types of commitment. Although a single epistemic system is usually assumed, it is possible to talk in terms of partially distinct systems: for example, Coates posits two parallel scales, involving 'assumption' (inferential) and 'assessment' (non-inferential) (1983:18-19), while Palmer argues that the English system is partly evidential (1986:Ch.2). However, such accounts assume an overriding epistemic system, contrasting to a large extent with a partially parallel deontic or root system. Many accounts of this area find epistemic meanings for most, if not all, of t h e central modals. A typical example is Coates (1983), who has distinct epistemic meanings for t h e following, which include two marginal items: can't, could, may, might, shall, should, will, would, must, needn't and ought to. Shall, will and would are here treated as 'marginally epistemic' (1983:244), and can't is given in t h a t form, as the affirmative can is not normally capable of EM. Such analyses may use concepts such as 'possibility' and 'necessity' informally as extreme or focal points of reference. An alternative model takes these two concepts as logical, polar values relatable to the existential and universal quantifiers, respectively, and, as such, exclusive definers for EM. This approach, which will be illustrated in 3 . 4 , raises m a j o r problems, such as the non-correspondence of logical values such as 'necessity' and t h e value such terms have in epistemic metalanguage, and t h e question how other supposedly epistemic exponents (eg, should, will and can't) ¿ire to be treated. A f u r t h e r area of disagreement in the definition of EM centres on the status and expression of 'prediction' and 'futurity'. In particular, it is unclear which uses of will should be treated as epistemic, and whether EM can, in principle, relate to f u t u r e time (this is partly why Coates assigns marginal epistemic status to the items shall, will and would). This issue involves the question whether epistemic f u t u r e meaning is, in fact, possible, b o t h for must (eg, Palmer 1990:54), and for may (McDowell 1987:passim); it will recur in 3.4 and, for must, 4.6.1. As with t h e previous problem area, if such uses are t o be excluded, the question arises how they are to be categorized otherwise. Non-epistemic types of modal meaning in English are sometimes grouped together as root uses (eg, Coates 1983, and generally in the American tradition); sometimes t h e t e r m deontic is used in this broad sense. This sense is hard to define precisely, in t h e way t h a t E M was defined above, but Palmer, in a cross-linguistic discussion of mood and modality (1986:96), suggests t h a t the relevant modal uses are those "containing an element of will"; here he adopts a phrase from Jespersen (1924:320-321), who contrasts such modal categories with those "containing no element of will", which include EM and, in general, categories t h a t deal with alternative possible worlds. Rather similarly, Bybee & Pagliuca (1985:63) refer to "the modalities that predicate conditions of either an internal or external n a t u r e on a wilful agent: these are the notions of ability, obligation, desire and intention";

43 they term such modalities 'agent-oriented'. These contrast with non-agent-oriented, nonepistemic, modalities, including 'root possibility', with root modality as a superordinate term for all agent- or non-agent-oriented, non-epistemic modalities (1985:77). The term 'root' suggests a kind a basic or prior status (Palmer 1986:103), and this may or may not be found desirable2; but it is a convenient cover term for all non-epistemic modal meanings, and will be used as such here. The term 'deontic' is then available for its narrower, and commoner use, according to which it is (3.13)

concerned with the necessity or possibility of acts performed by morally responsible agents. 3 (Lyons 1977:823)

This sense is illustrated both in the (b) glosses for (2.88)-(2.91)4, and in the uses of may in (3.11) and, if interpreted non-epistemically, (3.12), and will normally be followed here. However, the precise domain of deontic modality ('DM'), like that of EM, is variously understood. Thus, if the notion of personal will is taken as criterial for DM, this may be taken to include 'insistence' (as expressed by shall and will) (Brown 1991:117).5 Alternatively, DM may be defined more narrowly so as to cover only cases where the speaker initiates the action, otherwise sometimes termed 'performative' (Palmer 1990:Ch.4). DM in this sense would be illustrated in (3.11), with a request by the speaker for permission from the addressee seen as the converse of the granting of such permission, and, in the appropriate reading, for (3.12) - if the speaker were personally giving the permission 6 ; such senses will on occasion be referred to as typifying performative modality ('PM'). The most obvious problem in the classification of types of modality in English concerns the ability and volition uses of the modals. Palmer treats such uses as instances of dynamic modality, which also includes non-epistemic possibility and necessity, but without performative uses (1990:35-38); this is also termed 'subject-oriented' modality (Palmer 1974:100-102). The term 'dynamic' is often used in roughly this way, but again, there are variations; for example Perkins (1983) excludes volition; and the term 'pragmatic' is also used in roughly this sense (cf Footnote 18 in 2). Palmer (1990) focuses on the three categories of EM, DM and dynamic modality, but allows for the possibility of sub-divisions of dynamic modality, as well as other, more marginal, kinds. Other scholars extend the obviously unsatisfactory epistemic vs nonepistemic/root dichotomy in various ways: for example, Perkins (1983) includes boulomaic

3

4

5

6

The common tendency for items expressing root-like modalities to develop epistemic meanings has already been mentioned (2.3.2.5). This definition is not entirely satisfactory, in that it requires appropriate senses for 'necessity' and 'possibility'. The (b) glosses for (2.92) and (2.93) are left out of consideration here, since the first refers (primarily, at least) to futurity, and the second to ability. In fact, Brown includes the notion 'insistence' under both 'root deontic' and 'root dynamic' modality, without explanatory comment. It is sometimes objected that 'deontic' is, etymologically, inappropriate for the broader, root meaning, since it should denote 'duty' or 'obligation', which only constitutes a small area of its use in that sense. However, objection could also could be made to its use in its narrower, performative meaning, since it then applies to permission as well as obligation.

44 modality (for wishes and desires), while Melrose (1983) adds affective modality (as an emotive 'overtone' that can itself qualify other kinds of modality). The term 'root' is given a very different sense by Quirk et al (1985:219-221), who use it for a type of non-epistemic possibility and necessity, which would roughly correspond to a sub-division of Palmer's dynamic uses, while they introduce the terms 'intrinsic' and 'extrinsic' for, broadly speaking, the areas covered by root and epistemic modality, respectively, in their standard, wider senses. The more salient aspects of this rather confusing situation can be summaxized as follows: (3.14)

modal category (a) epistemic

(b) non-epistemic (c) root (d) agent-oriented (e) deontic

(f) performative (g) dynamic

some uses *(1) (2) (3) *(1) *(1) (2) (1) (1) *(2) (3) *(1) (1) (2)

as in (3.8) as in (1), including all reference to futurity as in (1), restricted to possibility and necessity all modal uses excluded from (a)(1) all non-epistemic non-epistemic, excluding (e)(2) as in (c)(1), without root possibility as in (c)(1) permission, etc, without ability and volition as in (f)(1) permission, etc, initiated by the speaker non-epistemic, excluding (f)(1) ability and volition

Stars in the 'some uses' column in (3.14) show the ways in which the terms EM, nonepistemic, root, DM and PM will be used here. Fortunately, however, the use of these terms is not as such of central importance in this study, and it will be argued later that the epistemic-root dichotomy is, firstly, not always, or equally, clear-cut, and secondly, not always of obvious significance. However, the terms require some clarification since, as will be seen in this chapter, some scholars claim that M-P alternation is directly relatable to the expression of specific types of modality. In the following section, questions of subjectivity and objectivity will involve further examination of various aspects of these different types of modality.

3.3 3.3.1

Subjectivity and objectivity Introduction

The terms 'subjective' and 'objective' often occur in accounts of modality in English, with reference primarily to DM, but also sometimes to EM (for example, in the accounts by Coates 1983, Perkins 1983, and Palmer 1990), and axe sometimes related specifically to

45 M - P contrasts (eg, Gramley & Pätzold 1992). In this section I shall explore some ways in which these terms have been used that may be relevant both specifically to the discussion in 3.4 and in general to the present investigation. We can first distinguish between (a) the widespread informal descriptive use of 'subjective' to signify speaker involvement or orientation of some kind and 'objective' either for its lack, or for some kind of external orientation, the terms often being relative to one another and complementary, and (b) uses of these terms to denote some specific theoretical claims. To a large extent, however, essentially the same descriptive claims are often involved. The first sense can be illustrated by reference to a pair discussed in 1.1: (1.4)

My girl must be home by ten.

(1.5)

My girl has to be home by ten.

According to the claimed interpretation, the speaker's will or interest is involved in the first case, with the modal, while the requirement has some kind of external orientation in the second, with the periphrastic; in such senses the first can be termed subjective, and the second objective, whether independently or in relation to each other. Something similar is argued for by Gramley & Pätzold (1992:158-159), in relation to the following pair: (3.15)

We said they could help out.

(3.16)

We said they were able to help out.

In (3.15) (assuming an ability sense), the ability attributed to they, the immediate subject of could, is an 'assessment' and hence subjective, while in (3.16), were able to conveys 'an objective possibility'. The semantic distinction assumed here is reflected in the force of the main verb said - in (3.15) as a claim, without certainty, and in (3.16) as a statement of a fact. An implication is that ability can and be able to systematically contrast in terms of subjectivity vs objectivity, but it is not clear exactly how this is to be understood whether, for example, I/she can play the piano pretty well has to be, in some way, a personal, subjective claim, and I am/she is able to play the piano well is a true, objective observation. Apart from can-be able to, Gramley & Pätzold note must-have to, needn'tdon't need to/don't have to and can-be allowed to as parallel pairs. Whatever problems of such kinds may arise with these concepts, the general sense of potential alternation between speaker-based authority or perspective (which may be associated with modals) and externally based, and/or objectively valid orientation (which may be associated with periphrastics) will be assumed here, and when the terms are used in the present study it will normally be in such an informal sense. In the second type of use of the terms, where specific theoretical claims axe involved, three roughly distinguishable approaches can be singled out, and will be discussed in turn. These are characterizable as involving contrasts located in (a) the semantic structure of sentences (this can be related to the standard, informal use outlined above; 3.3.2.1), (b) the speaker's point of view (3.3.2.2-3), and (c) the relative status of the 'observer' and the 'observed' in cognitive semantic theory (3.3.2.4). The first of these approaches, which

46 are by no means compatible, has been very influential and will receive much the greatest attention. Since it is widely claimed that there is at least some correlation between the use of modals and subjectivity, and that of periphrastics and objectivity, I shall then examine some arguments that have been proposed within the first of these frameworks to account for the different functioning of modals and periphrastics ( 3 . 3 . 3 ) . I shall then go on to discuss the notions of 'deontic source' and 'deontic target', since these relate to ideas about the speaker's role, and hence subjectivity and objectivity ( 3 . 3 . 4 ) . Finally, I shall review a number of descriptive proposals within this general area ( 3 . 3 . 5 ) .

3.3.2 3.3.2.1

Various approaches Semantic structure: modal and propositional elements

Lyons' views on subjectivity and objectivity in language and his emphasis on the importance of subjectivity within semantics 7 are set out in various places (on modality, especially 1977:Ch.l7 and 1981:235-242; on general issues, 1982 and 1987:176-177). Lyons' basic claim is shown in the following: (3.17)

there is a distinction to be drawn, in the structure and use of language, between a subjective component in which the speaker (or, more generally, the locutionary agent 8 ) expresses himself and an objective component comprising a set of communicable propositions. (1982:105)

This is applied to modality in the following ways: in uttering a subjectively modalized utterance (3.18)

a locutionary agent can be expressing his own beliefs and attitudes, rather than reporting, as a neutral observer, the existence of this or that state of affairs. (1981:237)

By contrast, in objective modality (3.19)

the modality is represented as something that holds, as a matter of fact, in some epistemic or deontic world which is external to whoever utters the sentence on particular occasions of utterance. (ibid)

Lyons rejects any semantic-pragmatic dichotomy (eg, 1987:154-160). In the present discussion I am using the term 'semantic' in Lyons' sense, to include material which many people would term pragmatic. References to 'the speaker' are always to be understood in this broader sense to refer to a speaker or writer; similarly, he, etc, for non-specific reference, is to be understood gender-neutrally.

47 Lyons explains his treatment of subjective and objective modality in terms of a general underlying semantic framework for sentence structure, consisting of the following three elements (1977:749-750): (3.20)

(a)

(b)

(c)

The neustic - "the speaker's commitment to the factuality, desirability, etc, of the propositional content": this represents the speaker's personal involvement, and as such is the locus of the inherent 'subjectivity of utterance' (cf Lyons 1981:240). The tropic - "that part of the sentence which correlates with the kind of speech-act that the sentence is characteristically used to perform" : this represents the conventional illocutionary force associated with the sentence-type. The phrastic - this denotes the propositional content.

Lyons argues that any modalized sentence can be subjective or objective, depending on whether the neustic or the tropic, respectively, is qualified. His discussion of this topic is primarily in terms of EM (1977:749-753, 798-809), but it is made clear, eg, in (3.19), that both EM and DM are involved. There is an obvious difference between this three-part model of semantic structure, where the subjective-objective contrast is located outside the proposition itself, and the simpler two-part structure, where the contrast is between the modal and propositional elements, so that, in effect, (b) and (c) in (3.20) are collapsed; the latter seems normally to be envisaged, both by Lyons himself (eg, in (3.17)) and by other scholars, but the non-total correspondence between these two models will have occasional relevance to the ensuing discussion. This case for a basic subjective-objective opposition in the use of modals can be illustrated by an example of four-way ambiguity provided by Lyons (1981:237-238): (3.21)

He may not come.

This is given the following glosses for the four meanings in question: (3.22) (3.23) (3.24) (3.25)

I-think-it-possible that he will not come, (subjective epistemic) Relative to what is known, it is possible that he will not come, (objective epistemic) I forbid him to come, (subjective deontic) It is not permitted that he come, (objective deontic)

Combined subjective-cum-objective modality is also allowed, and illustrated for epistemic possibility by Perhaps it may be raining, where perhaps is said to express the speaker's assessment and it may the objectively observed possibility (Lyons 1977:804); there is no mention of such an option for DM. I shall now discuss in turn each of the four types of modality involved in the glosses in (3.22)-(3.25), first the types of EM, and then the types of DM.

48 The gloss given in (3.22) shows what is generally taken to be the normal subjectivity of EM - the expression of a judgement of the speaker. By contrast, (3.23) shows a judgement that arises externally and is simply reported by the speaker. This 'objective' type of EM appears in various types of expression, as the following extensions of (3.21) show: (3.26)

They reckon he may not come.

(3.27)

It makes no difference what I think about it - he simply may not come.

In (3.26) the judgement is naturally interpreted as being associated with they, and is thus 'objective' in relation to the speaker. In a broader sense, according to which every reallife utterance is marked by the inherent 'subjectivity of utterance', this is still subjective, although the subjectivity applies in the first instance to the higher clause They reckon. This ambiguity can perhaps be dealt with roughly in terms of the more complex, threeelement model of semantic structure set out above in (3.20): this allows for the ubiquitous subjectivity of utterance to be located in the neustic, while the objectivity claimed is relatable to the tropic, so that a gloss in Lyons's terminology (¿is in 1977:Ch.l7) would be along the lines of '(I say) they reckon - (it is so) - he may not come'. (3.27) is somewhat different: here, the speaker explicitly dissociates his own opinion from the possibility referred to, which is presented as independently existent. This second type may seem close to logical or alethic/aletheutic modality, where the possibility in question is purely a matter of the content of possible worlds. A naturally occurring example of this type of EM is quoted by Huddleston (1984:167): (3.28)

Senator Withers may have misled Parliament.

This was uttered in a situation where the speaker "was acknowledging that the possibility existed" and not expressing his own assessment of a possibility; in other words the speaker could have qualified his words by adding I accept . . . or The fact is that ... , for example. The precise status of this second type of objective EM is doubtful, at least in normal language use; after all, even though it may be conveyed by (3.27) or (3.28), it can hardly be conveyed by (3.21) per se, without any contextual information. Lyons elsewhere, at some length, illustrates objective EM - which he takes to be essentially the same as alethic modality - with the case of a community where the marital status of a member is established quasi-mathematically and as if independently of human involvement (1977:797-798), but this illustration is hardly satisfactory, since the calculations involved all require human reasoning, which is by definition fallible and leads to conclusions of various degrees of strength rather than to certainties. Palmer (1990:6-7) suggests the following example of alethic EM: (3.29)

John is a bachelor, so he must be unmarried.

However, he comments: (3.30)

Yet it would be no less natural not to use must and to say so he is unmarried, and it seems likely that no clear distinction is recognized by native speakers between this alethic use and the epistemic use of must.

49 It is often observed that, logically, epistemic must should entail is, while in normal language use the reverse is the case (cf, in particular, Karttunen 1972:11-15). The following text illustrates the relationship required in logic: (3.31)

The question is not whether they [sc the angles of a triangle] do add up to two right angles but whether they must. If they must, then the conclusions of Euclid would hold for all possible universes, but if it only happens that they do, and our conviction that they must has no better foundation than our habit of thought, then it is quite worth while to find out what are the consequences of the opposite assumption. (R.H. Thouless, Straight and crooked thinking. London: Pan, 1973:119)

A contrast here is being made here between normal epistemic must (which rests on "our habit of thought"; = subjective) and a mathematical use of must which relates to all possible worlds (in a literal sense; = objective/alethic), and thus entails do. The fact that the argument in (3.31) involves a special use of language (which reverses the normal entailment relations between items) is made explicit, and there is no need to stipulate a distinct type of EM to explain it. In this connection, Huddleston claims (1984:167): (3.32)

Because It must be Kim's, but it isn't is contradictory we can say that It must be Kim's (interpreted epistemically) entails It is Kim's.

However valid this may be in logical terms, it does not represent normal language use, and Huddleston subsequently admits that It is Kim's is 'stronger' than It must be Kim's. The treatment of this subject by Lyons, in particular, has generated a considerable amount of discussion in the literature about objective EM in the second, alethic-like sense, but it should be emphasized that Lyons' own discussion is essentially speculative (1977:797): (3.33)

This is not a distinction that can be drawn sharply in the everyday use of language; and its epistemological justification is, to say the least, uncertain.

In the present work, 'objective EM' will normally be used only in the first sense, where there may be, for example, overt subordination, as in (3.26), or contextually determined distance from the speaker, as in (3.27) and (3.28). 'Alethic' will be reserved for the second sense. The status of the other two types of modality under discussion - subjective and objective DM, as illustrated in the two glosses (3.24) and (3.25) - is less problematic, and the two pairs are not appropriately seen as parallel. It is widely recognized that DM in principle includes both speaker-initiated and externally-initiated/reported uses, and that the first type has the illocutionary force of a directive, and the second that of an assertive. Since the directive type of DM is often treated as typified in performative uses of the modals, I shall briefly discuss PM at this point.

50

Two distinct approaches to PM, in relation to the use of modal verbs, can be found in the literature. Firstly, there is the strict performative hypothesis, as in Ross (1970), according to which the illocutionary force of a sentence is represented syntactically in the highest clause in underlying structure: this approach is adopted in relation to modal verbs by Boyd L· Thome (1969) and Lakoff (1972). The strict performative hypothesis has plenty of general problems in its application (cf Levinson 1983:246-251, and the discussion of Lakoff 1972 below in 3.3.4). Secondly, there is the informal, descriptive use of the term 'performative' to denote a subset of deontic uses (cf 3.2), without any specific claim being made regarding underlying structure, as in, for example, Coates (1983) and Palmer (1990). This second approach to performatives is illustrated in the following: (3.34)

Deontic modality is essentially performative. By using a deontic modal, a speaker may actually give permission (may, can), lay an obligation (must) or malie a promise or threat (shall). (Palmer 1990:69)

Palmer justifies distinguishing PM from dynamic modality on the grounds that such deontic uses axe both quite distinct from general dynamic uses, and are associated with a specific set of items. Thus, to mention some exponents of DM (in its usual broader sense), shall is always, may normally, must sometimes, and have (got) to, the only periphrastic involved, never performative (1990:69-70) - periphrastics being excluded from directive/performative use according to this account. Palmer does, in fact, apply the term 'performative' to EM as well as to a subset of DM on the grounds that speakers of epistemically modalized utterances "actually make a judgement about the truth of the propositions", which can be seen as parallel to issuing orders and the like (1990:10). It is revealing to compare the account cited in (3.34) with the approaches to performatives in two other accounts. Coates refers to performative uses of must alone among the modals; such occurrences axe treated not as a separate category, but as typifying the extreme subjective core of the cline of meaning covered by must (1983:32-33). As to the musi-related periphrastics, have got to, but not have to, is seen as subjective, but never actually performative. By contrast, Matthews (1989:201-202) treats only may and shall as having clear-cut 'performative deontic' uses, with the status of must as a true performative left in doubt; however, in a later discussion (1991:124-126), he treats must, along with a number of other items, as "typically performative", and only will, among the true modals, as "exclusively performative". Clearly, if the kind of pattern of use Palmer claims could be shown to have wider application, there would be stronger support for (a) positing a distinct directive/performative category, (b) associating it either exclusively or centrally with pure modals, and (c) associating external orientation with periphrastics. However, consideration of the rather different, and to some extent contradictory, findings of Palmer, Coates and Matthews on this issue suggests that extreme caution is needed with such a category. The issue is, in fact, more complex than this, since, for example, it will be seen below (4.4.2.3) that there is reason to talk of performative-like uses of the periphrastic have got to. Moreover, if such uses can, in fact, usefully be distinguished, a further question is whether they should be treated as central/basic/prototypical or peripheral.

51 Coates (1983:33) sees the performative use of must as its stereotypical or core use, corresponding to its psychological stereotype, while Bybee, Pagliuca L· Perkins (1991:23) see directive/performative uses as indirect speech acts that are secondary developments of exponents of agent-oriented modality (cf 3.2). It is not clear whether these views are compatible. The general question how far, and how exclusively, subjectivity is related to modals, and objectivity to periphrastics, will be central in the descriptive chapters below. 3.3.2.2

Point of view

One approach to 'points of view' concerns the omnipresent 'subjectivity of utterance' (cf (3.20)), which entails that "whenever speakers (or writers) say anything, they encode their point of view towards it" (Stubbs 1986:1). In a somewhat different sense, Mitchell argues that utterances often "betray an implicit viewpoint or perspective on the world" (1986:1) which is crucial to correct interpretation. Expressions with this quality are termed 'perspectival', and in the light of their typical functioning Mitchell offers an informal classification of points of view ('POV's). Establishing the POV of a particular expression is a matter of the interaction of contextual data and pragmatic principles, and successful communication depends on speaker and hearer agreeing on the 'most natural' or 'favoured' POV. Mitchell's taxonomy is as follows (1986:7, here slightly modified): (3.35)

POV description

term used

1 2 3 4 5 6 7

Τ 'you'

speaker hearer shared speaker + hearer contextually relevant individual intrasentential antecedent general absolute, nonperspectival

'other' 'bound' 'generic' 'absolute'

POVs 6 and 7 need some clarification. To take one of Mitchell's examples, if something is found to be boring, in POV 6 it is generally found to be so - that is, the quality is relative to people's general reactions; while in sense 7 the quality is treated as an inherent property, which, strictly, means that there is no POV. POV 7 thus has an objectivity comparable to that of alethic modality. Mitchell shows the effect of modalization on POV in relation to the following set, which include an instance of EM: (3.36)

That point was clear.

(3.37)

Was that point clear?

(3.38)

That point must be clear.

52 (3.36) can have any POV except ' y ° u \ (3.37) any except Ί ' or 'us', while in (3.38) "we seem to be shifted from Ί ' to 'you' " (26) - presumably, the first for must and the second for clear. We could speculate further about how this might work with DM, with particular reference to M - P alternation. They must go, lacking context, favours POVs 1 and 4, while in the context (He says) they must go it favours POV 4, but also allows 6; if a periphrastic is used, They have (got) to go favours POVs 4 and, perhaps, 1, while (He says) they have (got) to go, as with must, favours 4, but allows 6. Can she leave? favours 2 and 4, while Is she allowed to leave? favours 4 and perhaps 6. These 'favoured' POVs are those which are found to be pragmatically more/most 'natural', and thus constitute default interpretations (cf 3.5.2); and they clearly suggest one approach to significant M - P distinctions. 3.3.2.3

Empathy perspective

Another approach to the relation between the speaker and the proposition is put forward in Kuno's theory of empathy perspective (1987:Ch.5). Empathy is here defined as: (3.39)

the speaker's identification, which may vary in degree, with a person/thing that participates in the event or state that he describes in a sentence. (Kuno 1987:206).

What is significant here is, firstly, the possibility for the speaker to identify/empathize with any speech event participant, and, secondly, the claim that this 'empathy' may vary in degree between total and minimal. 9 Kuno's basic examples concern various formulations of the proposition 'John hit his brother Bill', and the correlations between these and possible types of speaker identification. The discussion concerns describing or reporting language, as opposed to, for example, issuing orders or requests, but seems applicable to various modal uses. In the following corpus text with must: (3.40)

and then, you, you must do that, I suppose, with standardization (CEC:55/1174-1175)

the function of must is nothing like that of a directive: the context and the co-occurrence of I suppose and standardization show that the speaker is sympathizing, or identifying, \frith the addressee. If have (got) to is substituted here, as in: (3.41)

and then, you, you have (got) to do that, I suppose, with standardization

the speaker's expression of sympathy is markedly reduced. This example suggests that the choice of a modal or a related periphrastic may reflect the speaker's empathy, in Kuno's sense.

This is certainly the sense of (3.39); but elsewhere, Kuno assumes a categorical distinction between partial and total speaker identification with a participant (1987:204).

53 The following pair of examples show a comparable M - P contrast: 1 0 (3.42)

May/can I leave now?

(3.43)

Am I allowed to leave now?

The speaker in this pair might be seen as appealing for permission in two slightly different ways: with may/can in (3.42) he assumes that the addressee has average identification, or empathy, with him (the roles here are reversed, since the utterance is a question), while with be allowed to in (3.43) reduced identification is assumed. In the subsequent discussion, I shall assume the possibility of an empathy factor, in the general sense of sympathy felt by a speaker (or hearer) for a speech participant, or, more generally, with a state, requirement, etc, referred to by the proposition. 3.3.2.4

The observer and the observed

Langacker (1985; 1987:128-132; 1990) discusses the role of the speaker/observer/self (S) in relation to the predication/observed/other ( 0 ) in terms of subjectivity and objectivity. The subjective/objective issue is seen here as deriving from "the dual role of ground elements in a deictic expression" (these elements include the speech event, its participants and its setting): "they serve both as the source of the predication, and as participants within the predication" (1987:128). The essential variable is the role of S in relation to 0 . When S and 0 have the "optimal viewing arrangement" (which means that S is in the best possible position to focus on 0 , and 0 is as clearly defined from its background or surroundings as possible), S and 0 are, respectively, maximally subjective and objective. The opposite situation has S at the focal point of 0 , rendering S maximally objective: this is typified by self-reference, when the speaker becomes at the same time conceptualizer and primary object of conceptualization. Although Langacker's data may be relatable to some other claims in this area, this approach to subjectivity and objectivity can be distinguished from Lyons' proposals (3.3.2.1) as follows. 11 Firstly, the speaker is assumed always to be involved in the predication: "the speaker (or hearer) participates in a construal relationship with respect to every linguistic expression" (1987:128). This suggests that language is to be seen as inherently subjective, and that there is no subjective-objective dichotomy, since the crucial variable here is the role of the observer (S), ranging between extreme subjectivity and egocentricity, which "attenuates the subjective/objective distinction" (1985:122). Further, objectivity is treated as a property of what the speaker 'observes', which may, naturally, include himself; in other approaches the term has been used to denote distance from the speaker. Langacker relates 'subjectification' (viewed primarily as a diachronic process) closely to grammaticalization, and treats this as present in the use of be going to as a future form, and, at a further stage of the process, in the use of the English modals (1990). 10

11

May-be allowed io are not included a m o n g the M - P pairs described in t h e present study, b u t some observations m a d e a b o u t t h e m seem relevant t o t h e other pairs and sets. Pace T r a u g o t t (1989:36), who implicitly equates t h e relevant concepts in Lyons (1977) a n d Langacker (1985).

54 The fact that be going to and the modals represent different degrees of grammaticalized subjectification is reflected in the fact that be going to functions as a main rather than auxiliary verb. The inherent subjectivity of the modals (viewed as the extreme stage of a developmental process) is characterized as follows: (3.44)

The crucial development is that the conception of directed potency loses its profiled status, the source of that potency (either the speaker or some other facet of present reality) being offstage and subjectively construed. The resulting expressions are grounding predications: . . . they no longer profile the grounding relationship (directed potency), but only the grounded process serving as its target. (1990:27)

What is interesting here for our purposes is the notion that the modals have a general, or vague, semantic value that simply qualifies the process denoted by the main verb, without, for example, specifying where the source may be located. Periphrastics, like full lexical verbs, are therefore to be seen as inherently less subjective. Although Langacker's discussion is in very general terms, it supports an approach to M - P contrasts in which the modal is seen as maximally unspecified, or unmarked, while the choice of a periphrastic allows a modalized predication to be more precisely specified both syntactically and semantically.

3.3.3

Subjectivity, objectivity and modal-periphrastic alternation

In this section I shall examine certain correlations that have been observed between (a) subjective and objective modality, and (b) the use of modals and periphrastics, respectively. Before going on to these, it should be mentioned that there is no evidence that alethic modality is in any way reflected in M - P alternation: thus, the objective meanings of must in (3.29) and (3.31) are equally well (but not better) carried by the use of have to. The main source here is Palmer (1983; cf 1990:10-11), who adopts the model in Lyons (1977) (in its simpler two-part form), as discussed in 3.3.2.1, and deals with the pairs can-be able to, must-have (got) to, will-be going to and will-be willing to. Palmer's argument is that the English modals are devices for indicating subjectivity, in the sense of indicating the speaker's attitude or commitment, and that periphrastics lack this function. His evidence concerns restrictions on the use of both modals and periphrastics, and can be summarized as follows: (1) where modals and periphrastics 12 co-occur, it has to be in the form Ί modal + 1(+) periphrastic'; (2) modals are associated with the present tense;

12

Palmer refers to 'semi-modals', but I shall use the term 'periphrastics' for consistency with our general discussion.

55 (3) periphrastics are not used for PM, or, normally, for EM (since EM is inherently subjective); (4) î/-clauses axe associated with periphrastics. A further argument that will be discussed in this connection and that is relatable to the Lyons model of sentence structure is the claim by Perkins (1983) that (5) the item be, an element in many periphrastics (and other compound verbal units), is associated with objectivity. These five arguments will now be discussed briefly in turn; while many of the points involved are clearly well founded, I shall suggest that there are some complicating factors that may make the whole case less straightforward than it appears. (1) The occurrence and co-occurrence restrictions on modals and periphrastics (cf 2.3.1.2) are claimed to directly reflect the underlying semantic structure of sentences: 'modal (subjective) + propositional (objective)', with the subjective element corresponding to the neustic and the objective element to the rest of the structure in (3.20). This is a cleax-cut and perfectly plausible claim as far as it goes and will not be discussed further. (2) It is observed of modals, specifically in their PM and EM uses, that either they only have a present tense form (eg, must) or their past tense forms are normally not used for past time reference. This is related to the notion that performative functions, whether deontic or epistemic, can only apply in present (including future) time. In this connection, Palmer discusses the specific case of can/could, and the notorious restriction shown in: (1.13)

*I ran fast, and could catch the bus.

Palmer argues that could, and similarly would, are unavailable for such a typically propositional use, "precisely because they are modals" (Palmer 1983:213). This argument is complex and depends on a sharp distinction being drawn between performative and other deontic uses, which is not necessarily easy to malie (cf 3.3.2.1). Further, it has to be shown that periphrastics do not, in fact, have relevant PM and EM uses, and do not have past tense forms for past time epistemic use. It will be seen below, in 4, that have (got) to is questionable in these respects. The evident incompatibility between modals and propositional use will be returned to in 3.6. (3) It is claimed that periphrastics are not available for clear-cut subjective uses, either in PM or in EM (which is inherently subjective) (Palmer 1983:208-209). The claim relating to PM is often made, and is illustrated in the following (ibid): (3.45)

He must come and see me tomorrow.

This is said to show the speaker imposing an obligation, and not to be replaceable by have (got) to. The corresponding sentence: (3.46)

He has (got) to come and see me tomorrow.

56 thus has to be a report rather than a command. This is typical of claims often made in this area, and will be assumed here to reflect at least a tendency in interpretation; it will be explored in 4. The claim concerning EM involves the non-availability of periphrastics for EM, the only exception being the AE use of have to. This is related to the argument that EM is not normally objective (that is, that epistemic judgements are intrinsically speakerbased), and that periphrastics are associated with the objective, propositional element in semantic structure. This would have to mean that the epistemic use of have (got) to (which certainly has currency in BE), must always be objective. A corpus occurrence of epistemic have (got) to follows: (3.47)

if you've seen the - if you've seen all the old Frankensteins you've got to, you've got to know the jokes (CEC:632/740-741)

Even though it might be argued that this is 'objective', in that the speaker's conclusion is based on explicit reasoning (if ... , then . . . ), it is hard to see that this is objective in the categorical sense intended here. Moreover, there are other cases of epistemic have (got) to, as in the familiax You've got to be joking, which appears to alternate freely with You must be joking, where it is hard to find evidence of objectivity of any kind; and in both these cases the use of must seems to make no important difference in this respect. As to the general claim for the non-availability of periphrastics for EM, if we consider the whole range of M - P pairs/sets (cf 2.3.2.5), it could just as well be argued that be able to and be willing to axe exceptional in not having epistemic uses; but then their modal counterparts, can (ability) and will (volition), do not either. The second part of this claim, then, is very problematic. (4) The close association of modals and subjectivity, specifically in EM, is given as the reason for the possible oddity of modals in ¿/-clauses 13 , as in the following (Lyons 1977:805806): (3.48)

If it may be raining, you should take your umbrella.

It is claimed that the following is, normally, more natural: (3.49)

If it is possible that it will rain, you should take your umbrella. 1 4

Lyons merely says that sentences like (3.48) may seem odd; however, they are sometimes excluded as ungrammatical (eg, Jenkins 1972:26,38). The first of these two sentences, it is argued, is odd because may naturally has subjective interpretation, and this is not natural for tin «/-clause, where the subordination removes the verb may be raining from the speaker's immediate, subjective focus (cf (3.26)-(3.28) and related discussion above). This would suggest that may is possible in (3.48) to the extent that it can be used 13

14

Although this discussion is entirely in terms of ¿/-clauses, the phenomena in question also appear to occur in other types of subordinate clause. To be exact, (3.49) corresponds to If it may rain . . . , and (3.48) to If it is possible that it is/will be raining ...

57 objectively; this is illustrated in the following, where the may clause is subordinated to a higher, non-modalized clause: (3.50) (3.51)

If it really is true that it may be raining, you should take your umbrella. If you really believe it may be raining, you should take your umbrella.

These could be considered explicitly objective, hence allowing may in unambiguously objective senses. Another example of the oddity of epistemic modals in ¿/-clauses is (Palmer 1983:213): (3.52)

?*If John must be in his office, I'll go there.

Clearly, a paraphrase like (3.53)

If it's certain that John's in his office, I'll go there.

would be normal 15 , making this pair seem parallel to (3.48) and (3.49). However, the use of the corresponding, supposedly objective, periphrastic, as in: (3.54)

?If John has (got) to be in his office, I'll go there.

is, in the required epistemic sense, scarcely better than (3.52). Testing with some other verbal periphrases in the saine context gives the following: (3.55) (3.56) (3.57)

?If John's bound to be in his office, I'll go there. If John's sure to be in his office, I'll go there. If John's certain to be in his office, I'll go there.

(3.55) is about as good as (3.52) and (3.54), but (3.56) and (3.57) axe fine. This evidence suggests that only items with no claim to modal or periphrastic status, as in (3.53), (3.56) and (3.57) are fully natural in epistemic «/-clauses (to ignore cases like (3.50) and (3.51), where the modal is further subordinated within the «/-clause). The question remains why periphrastics, which might be expected to be either typically objective or at least capable of objective meaning, should be unnatural in the 'objective' context of epistemic ¿/-clauses. If we compare the effect when must and have (got) to are used in «/-clauses with deontic senses, as in the relevant readings of (3.52) and (3.54), both sentences are possible, although the use of must is less natural than that of have (got) to. On the other hand, both the following pair are equally good: (3.58) (3.59)

If you must be late, at least try to be quiet. If you have to be late, at least try to be quiet.

This suggests that there may be no serious restriction on the use of modals in deontic senses in «/-clauses, unlike the situation with epistemic senses. The real problem, then, lies in some incompatibility between the 'objectivity' required in an epistemic «/-clause and the epistemic use of both modal and periphrastic items. 15

I am not claiming that (3.52) and (3.53) are synonymous.

58 Epistemic meaning, unlike deontic meaning, is characteristically subjective; so if it is assumed that both modals and periphrastics axe capable of relatively subjective meaning, incompatibility may arise between this 'natural' subjectivity and the kind of objectivity required in epistemic ¿/-clauses. (Verbal phrases like be possible that in (3.49) and be certain that in (3.53) are, of course, not subjective in this way.) This discussion suggests that the restriction illustrated by (3.48) and (3.52) may not in fact bear significantly on the basic issue in this section. (5) Regarding the objectivity of be, Perkins states (1983:68): (3.60)

all complex verbal expressions which incorporate the verb be express objective modality, the objectivity being a function of the fact that the modality itself is actually asserted.

The verb be is taken to assert the existence of a state of affairs independently of the speaker, and this argument is applied to the contrast evident in the following pair: (3.61)

Ali could defeat anyone he fought with when he was 25.

(3.62)

Ali was able to defeat anyone he fought with when he was 25.

It is stated that in (3.62), but not in (3.61), the ability was actually realized, and that this is due not to the meaning of able but to that of was. This claim is supported by the fact that there is no implication of realization in: seemed "1

(

looked > able to defeat anyone he fought with when he was 25. felt This argument is difficult toJaccept, since, for example, in: (3.64) Ali was willing to talee on anyone he met when he was 25. where the verb was willing to includes be (and Perkins also treats it as a periphrastic), there is no implication of fulfilment, while in: (3.65)

Ali proved able to defeat anyone he fought with when he was 25.

which lacks the verb be, clearly implies fulfilment. Verbs like be able to and be willing to may well be more 'objective' than the corresponding modals, but the question still remains unanswered why able to or willing to, whether with be or with prove, should imply fulfilment. This issue will be returned to from another perspective in 3 . 6 . This claim for the 'objectivizing' function of be affects, of the items selected in (2.110), the following: be able to, be bound to, be going to, be supposed to, be willing to and is to. This leaves out only the following periphrastics: have to, have got to, ought to and had better. Of these, ought to is a special case (because of, in particular, its close relation to should), and the other items all have a form of have as their first element. All these items - except ought to - are also treated by Perkins as inherently (ie, in their core meanings) objective, though without any specific argumentation such as is applied to be able to.

59 Conclusion. Despite a certain amount of obvious plausibility for the general association of modals with 'subjectivity', and of periphrastics with 'objectivity' (whether these concepts are understood as pragmatic tendencies or as elements in underlying semantic structure), it has been shown that there are a number of difficulties attaching to any strong claims of the kinds seen in this section.

3.3.4

Deontic source and deontic target

The roles of source (ie, imposer of a deontic value) and, to a lesser extent, taxget (ie, recipient of this value) are often treated as significant variables in the use of modalized expressions. In this section the nature of these concepts and their potential application to the analysis of M - P alternations will be explored. The study of the universals of mood and modality by Chung L· Timberlake (1985:241250) treats sources and targets, with specific, but not exclusive, reference to DM, in the following way: (3.66)

A mode 1 6 may include a source from whose point of view the event is characterized as actual or non-actual. For primary events 17 the source is typically the speaker; it is the speaker who identifies the event as actual, or imposes it on the addressee . . . A mode may include a participant target who is responsible for the actuality of the event. For primary events the target is typically the subject. . . . the deontic mode often includes a source of obligation/permission (for example, the speaker). But an event can also be imposed without any explicitly named or understood source . . . the deontic mode typically includes a participant target who can be held responsible for the event (for example, the subject and/or semantic agent). But an event can also be imposed without being imposed on any particular target, as in "let there be light" or "there must be no smoking". (1985:241-242,247)

The terms 'source' and 'target' require brief comment. The source of a modalized sentence is standardly seen, in the case of DM, as the person imposing or authorizing the requirement, permission, etc, expressed, and, in the case of EM, uttering the epistemic judgement. The target is the recipient of the deontic value imposed, "responsible for the actuality of the event". It will be seen in this section that these axe not simple or unproblematic notions. The following are the crucial features of (3.66) for our purposes: (3.67)

16 17

(a) (b) (c)

the general relevance claimed for source and taxget; the optional assignability of these roles; and the notion of default or prototypical identification (ie, speaker as source, subject as target).

The term 'mode' here means, roughly, type of modality. These are events expressed by matrix clauses.

60 In talking here of the participant roles of source and target and their optionality, we are cleaxly dealing with pragmatic interpretation. The notion of default interpretation as presented in (3.66) may be assumed to work in ways such as the following: in the interpretation of a modalized utterance, the speaker is taken to be 'source' if there is, for example, no suggestion of another source, or of any reason why the speaker should not be understood to be the source. Further, if there is a relevant distinction between the use of a modal and of a periphrastic, such inferences may apply in the case of one item but not the other. Default interpretation of such a kind, if commonly applied, may result in a prototypical or stereotypical interpretation for an item. This will be discussed in more detail in 3.5.2. For further investigation of these notions, I shall turn to the account of the speech event participants in modalized utterances by Lakoff (1972:236-239), which was developed from arguments in Newmeyer (1975:69 18 ). It only deals with deontic must, but this is in the context of a general discussion of modality, including the status of periphrastic equivalents, so it must be assumed to be generally applicable to M - P pairs. Lakoff proposes an underlying three-place predicate structure for deontically modalized sentences, to represent the following elements: (3.68)

(a) (b) (c)

the logical subject, or imposer of the obligation - here the source; the bearer or receiver of the obligation - here the target; and the complement sentence.

This is illustrated by a consideration of the possible interpretations of such sentences as: (3.69)

The witch must be kissed by every man in this room.

(3.70)

There must be peace and quiet.

The deontic source may be filled by the speaker or another person, or it may be blank. (To be more exact, if we follow Lakoff's remarks elsewhere in the same paper, quoted in (1.2), it may also be the speaker together with another person.) In the cases of (3.69) and (3.70), this means that the speaker's attitude to the obligation could be glossed, variously, Ί say', 'x says', 'it is said' or Ί pass on what χ says with my support'. It is questionable whether such a finite listing of possible sources can really be justified. We can consider some corpus data: (3.71)

18

Β

you've heard, probably, we're very often bedevilled more by what the candidates, more by difficulties of marking than by what we ought to set the candidates, you know A yes, that's a devil, I agree Β and then, you, you must do that, I suppose, with standardization (CEC:55/1167-1175)

Lakoff in fact cites Newmeyer (1970), but the relevant contents of the two references are compatible.

61

(3.72)

Β you could study literature in a foreign language, couldn't you? A yes, I could Β you feel it, it must be English A yes (CEC:776/1107-1112)

In the first text, which was cited in part earlier as (3.40), the speaker is sympathizing with the subject/addressee. The use of I suppose makes it clear he can in no way be the source himself; but he is hardly 'reporting' the requirement either (which would require an introducer like you know . . . ). The source is not apparent, though it evidently has something to do with the consequences of standardization. In the second text A agrees with Β that his chosen language must be English, but neither A nor Β can be said to be 'imposing an obligation'. Here, even if A is seen as the 'source', there is no question of anything like 'imposition'. Cases such as (3.71) and (3.72) suggest that the concept of deontic source as an identifiable person - or even factor - in deontically modalized utterances, including the option that there is no source, is in practice simply too strong. This discussion of deontic sources has only dealt with must. Our real aim, however, is to see how this factor may bear on the alternation between M-P pairs. For this purpose we shall see what effect the substitution of have (got) to would have in (3.69)-(3.72). Appropriate contexts have to be considered. In the case of (3.69), assuming someone is stating the requirement (these are the rules of the game), regardless of who invented them or is 'imposing' them, all three items must, have to and have got to are possible, and no necessary distinction between them can confidently be made. The same applies to (3.70), assuming, for example, that someone is saying this to a crowd of people. However, two factors that appear potentially relevant to M-P choice in these examples are the following: have (got) to would (a) suggest motivation distinct from the interlocutors, and (b) be more appropriate in informell language. The first of these could be related to Lakoff's claims in that it points to an inferred, default interpretation, but the second is unrelated to them. In the cases of (3.71) and (3.72), where the speaker can in no way be imposing an obligation, but, rather, is showing sympathy with the addressee, have (got) to would simply reduce the speaker's interpersonal involvement. To move to different items, I shall next consider possible contrast in 'source' reference between may/can and be allowed to.19 To return to a pair of examples discussed earlier in 3.3.2.3, the use of may/can in: (3.42)

May/can I leave now?

suggests that the speaker is looking to the addressee as the locus (but not necessarily the actual source) of authority, while replacement of may/can by be allowed to, as in: (3.43)

Am I allowed to leave now?

suggests that an authority is being appealed to which is independent of - but may still be associated with - the addressee; and this is reflected in the fact that please is more naturally added in (3.42) than in (3.43). This supports the notion that some speaker vs external source contrast in the use of M-P pairs may indeed have more general validity. 19

On the use of these items for illustration, cf Footnote 10 above.

62 In the case of the deontic target, Lakoff argues on the basis of extensions to (3.69) that provide more specific context, as follows: (3.69a) (3.69b)

. . . , or they will all be turned into star-nosed moles. . . . , or the leader of the coven will demote her to leprechaun.

(3.69c)

. . . - that's the law.

In these cases, it is claimed, the target in (3.69a) becomes every man, in (3.69b) the witch, and in (3.69c) zero. These are reasonable default interpretations, but can hardly be taken as more. To illustrate this further, we may consider what happens if the first clause of (3.69b) is changed into the corresponding active form, as in: (3.69d)

Every man in this room must kiss the witch, or the leader of the coven will demote her to leprechaun.

Who is now to be the target? Of the three personal agents present (every man, the witch and the leader of the coven), the grammatical subject, every man, might seem to be the obvious choice; alternatively, since (3.69d), like (3.69b), points to her as em interested party, it might be the witch·, the leader of the coven, the third possibility, is not in the same clause as must but might also be considered, in the words of (3.66), as in an extended sense "responsible for the actuality of the event". There is no easy way of fixing on an obligatee among the available agents, without taking into account overt and non-overt contextual data. Part of the problem here lies in the vague use of terms like 'target' or 'bearer of the obligation' (Lakoff). The identifications related to (3.69a-c) associate the obligatee with an agent who has an interest in the performance of the action denoted in the complement. If, however, the verb kiss (exceptional in being a 'reciprocal and symmetrical' verb, Quirk et al 1985:364) were replaced in these examples by, for example, stab, it would be easy to show that one person might be given the task of seeing that the stabbing was done, and another might 'have to' do it: the context of utterance could specify this, leave it to be inferred, or simply leave it vague. This can be illustrated further with Lakoff's second example, (3.70), which is offered as a one-place structure without a target. Again, a flexible approach seems called for. There must be peace and quiet, as it stands, names no agent, but if it were uttered in a noisy room it would fairly clearly denote the occupants of that room as something like targets; while if said by someone in authority to subordinates deputed to execute this requirement, they themselves would be the obligatees in the first instance, and their charges would then be, in turn; and so on. I shall finally consider the two corpus texts quoted above. (3.71) is overtly unproblematic, since there is an explicit addressee who could be considered to be under an obligation. However in (3.72) the must predicate indicates a requirement that A is presumed by Β to feel; and it would be difficult to identify a specific target - it could be A or any person who wants to study English. To sum up, there axe obvious problems involved in attempts to establish procedures for the identification of deontic sources and targets, and while the source is clearly a

63 potentially significant variable in the comparison of M - P pairs, there is no reason to see identification of the target as relevant to our investigation. It will be assumed here that source and target are pragmatically inferred, and that the notion of prototypical or default identification, as suggested in (3.66)-(3.67), is directly relevant to this. Here we may note Sweetser's comment on (3.69) that "any pragmatically reasonable interpretation of the modal imposer and imposee is possible" (1990:67). In fact, to return to Lakoff's 3-place predicate structure outlined above in (3.68), I shall assume that the modality of an item like must is best seen as applying in the first instance to the complement sentence as a whole (the third element in the proposed structure), and only secondarily to such sources or targets as may be inferred. This would suggest greater semantic similarity to the structure of epistemic sentences, where the modality is generally recognized as applying to the whole proposition.

3.3.5

Some relevant descriptions

In this section I shall illustrate some descriptive approaches to this area that in various ways involve associating modals with subjectivity and periphrastics with objectivity. I am here concerned with general approaches to the area rather than details of descriptive accounts, which will be dealt with in 4 and 5. In principle, there are three ways in which this connection between M - P alternation and subjectivity-objectivity might work: item type

subjective/objective?

(a)

modal periphrastic

subjective objective

(b)

modal periphrastic

subjective subjective objective

(c)

modal

subjective objective objective

(3.73)

periphrastic

These will be demonstrated in turn. In model (a) there is a straight correlation between modals and subjectivity, and periphrastics and objectivity, respectively. This is illustrated in the following statement of this contrast between must and have to by Leech (1969:227-228): (3.74)

Have to indicates the general idea of obligation, but must more precisely identifies the speaker as the person who gives the orders. Thus, in a military context, You must be back in camp ... would probably be spoken by an officer giving the orders, while You have to be back in camp ... could well

be spoken by an ordinary soldier informing his comrades of orders issued by someone else. Must implicates the speaker as authority irrespective of the person of the subject of the clause. This relates to the following pair: (3.75)

You must be back in camp by ten.

(3.76)

You have to be back in camp by ten.

Similarly, Coates (1990:56) distinguishes between the following pair: (3.77) (3.78)

You must fill in this form You have to fill in this form.

in these terms: (3.79)

In the first sentence the speaker imposes an obligation on the addressee, while in the second, the speaker informs the addressee of an obligation but distances him/herself from the authority source.

This first model is the commonest in the literature. Model (b) in (3.73) can be illustrated by the following quotation, which was given earlier in 1: (1.2)

when the speaker agrees with, or takes upon himself, the atomic meaning of the modal, he can use the simple modal form. Otherwise he must use the periphrastic variant. (Lakoff 1972:239)

This system, which is applied to a set of M - P pairs, can be represented as follows: (3.80)

exponent modal periphrastic

meaning atomic + speaker involvement ( atomic + speaker involvement atomic

{

Since the modal is here restricted to the ' + speaker' option, it can be considered the marked item. Model (c) in (3.73) can be illustrated by the following quotations from Kruisinga (1925), dealing with the pairs may-be allowed to20 and must-have to, respectively: (3.81)

20

You (he) may go expresses that the permission is given by the speaker, or by someone else. You are (he is) allowed to go can only be used when the permission is given by another than the speaker. (1925:320)

On the use of this pair, see Footnote 10 above.

65 (3.82)

Both must and to have21 may express necessity. But necessity, if due to a personal will, can only be expressed by must (subjective necessity). If it is due to circumstances it may be expressed by must or by to have (objective necessity). (1925:346)

These two statements are not quite comparable, as can be shown in diagram form, with (3.83) for (3.81) and (3.84) for (3.82); this will also enable some degree of comparison with (3.80): (3.83)

exponent may be allowed to

(3.84)

exponent must have to

source of authority r speaker 1- another person another person source of authority f person 1· circumstances circumstances

There is parallelism here in that in each case the periphrastic is externally oriented in relation to the speaker, and has the narrower range of use. In these cases, the periphrastic could be considered the marked item. The three models in (3.73) differ in the relative status of modals and periphrastics in M-P pairs, but share the association of modals with the speaker and subjectivity, and of periphrastics with objectivity. Numerous descriptions reflect or imply this: for example, Palmer (1990:Ch.4) treats a set of modals (may, can, shall, must, and, doubtfully, should) as available for the expression of PM, and thus used for the granting of permission, etc, specifically by the speaker (cf 3.3.2.1). He accepts cautiously the idea that these modals contrast with periphrastics, specifically in respect of their speaker involvement, in the cases of must-have to, may-be allowed to and should-be supposed to. Certain basic, prima facie problems with such accounts were noted in the discussion in 1.2; and the argument in 3.3 has generally suggested that, whatever tendencies may be evident, any kind of obligatory connection between the use of modals and an initiating role for the speaker cannot be justified. Some other perspectives on the speaker's role will now be discussed. The following proposals have in common that they are concerned with identifying the role of modals in relation to the speaker's viewpoint (3.3.2.2), rather than his actions; in this sense they represent a modification of the kind of accounts just discussed. First, to quote an EFL practice book, Allen presents a contrast between the use of must for the expression of "obligation or compulsion from the speaker's viewpoint" and have to for "external obligation" (1959:38-39), illustrated by pairs such as:

21

To have (got) in Kruisinga (1925) denotes the verb (got) to, as here.

have (got)

in all its uses, including those of

have

66 (3.85)

You must go now. (I want to go to bed.)

(3.86)

What a pity you have to go now. (It's time for you to catch your train.)

(3.87)

We must begin before five. (Or we shan't finish in time for our supper.)

(3.88)

We have to begin before five. (That's the time arranged.)

This suggests the following pattern: (3.89)

exponent

nature of obligation

must have to

speaker's viewpoint external

The speaker's viewpoint obviously offers a wider perspective than his authority, and thus could, perhaps, subsume the concept of deontic source as discussed in 3.3.4. This can be further illustrated with another M-P pair: (3.90)

You know, you really should talee better care of yourself.

(3.91)

You know, you really are supposed to take better care of yourself.

(3.90), with should, naturally expresses the speaker's personal opinion or viewpoint, while in (3.91) there is an assumption of an external source for the advice. Significantly, these interpretations are reflected in the likely senses of really: in (3.90) it means something like Ί strongly feel', and in (3.91) 'this is certainly the case'. A relevant argument is put forward by Bouma (1975:324-325), in a critical discussion of the general Larkin/Lakoff approach discussed in 1.1 and 3.3.4. Bouma claims that such an approach does not stand up to consideration of data such as the following: (3.92)

I must go with you, but I don't want to.

(3.93)

I have to go with you, but I don't want to.

(3.94)

You should vote for Agnew, but I hope you don't.

(3.95)

You are supposed to vote for Agnew, but I hope you don't.

Bouma points out that sentences like (3.92) and (3.94) should be "ungrammatical or at least questionable" in relation to sentences like (3.93) and (3.95), respectively, if the use of the modal is as argued in (1.2), for example, since (3.92) and (3.94) would show the speaker both associating with the requirement (by the use of the modals) and rejecting it (in the conjoined clauses). He argues that these two sentences are, in fact, as good as their periphrastic counterparts, and if (3.92) is at all odd this is a stylistic matter - thus introducing a new factor (cf 3.7). Bouma goes on to claim that the M-P distinction is actually rather different from what Larkin and Lakoff propose: (3.96)

the choice of one form over another has to do with the speaker's view of the type of commitment of the actor. That is, in . . . (3.94) the speaker feels that the actor is somehow morally committed (perhaps because of some prior agreement) whereas in . . . (3.95) the speaker is focusing on some kind of directive (perhaps on election rules in this particular case).

67 In other words, the speaker's role remains crucial, but he is not seen as the originator of or sympathizer with the obligation, but rather as the observer of the actor's situation. This allows the speaker to adopt the actor's point of view, and it associates the use of modals with a moral perspective, and that of periphrastics with a factual/objective one, as in the following diagram: (3.97)

exponent

speaker's focus

perspective

modal periphrastic

subject's commitment external

moral objective

Bouma's argument allows us to reconsider some of the data discussed above in 3.3.4. Lakoff's problem cases (3.69) and (3.70) can now be seen as involving some kind of 'commitment' for the actors (whoever exactly they may be); the counterparts of these sentences with have to would then present the relevant requirement objectively as 'some kind of directive'. In both (3.71) and (3.72) it was noted that one aspect of the speaker's role is the expression of sympathy for the subject, and this could perhaps be seen as the speaker's view of the subject's potential moral commitment. A somewhat parallel distinction, in relation to will and be going to, is observed by Larkin in the following pair: (3.98) (3.99)

?Hang on! We'll crash! Hang on! We're going to crash!

Larkin states (1976:395) that (3.100)

a speaker of a will sentence is less talking about an already determined state of affairs than he is about one that still involves the decisions and attitudes of individual actors.

This claim, that the use of the modal indicates reference to attitudes as opposed to already settled events, is related by Larkin to the distinction between must and have to (cf (1.4)(1.6)): the parallel seems to lie in the idea that the use of a modal suggests human initiative, whether exercised or observed by the speaker, whereas the use of a periphrastic refers to external forces. The will-be going to contrast is a complex issue and will not be discussed further here, but will be returned to later in this chapter (3.5.3) and then in the descriptive section (5.3.1). The accounts discussed in 3.3.5 display a variety of approaches to the role of the speaker and other participants in sentences that extend the scope of subjectivity and objectivity beyond questions of identifying sources and obligatees.

3.4

The generative framework

Generative theory has had relatively little to say in detail on the area of modality. In its early stages, the Auxiliary rule:

68 (3.101)

Aux —ν Tense (M)(have+en)(be+ing)

(based on Chomsky 1957:39) was generally found to provide an adequate account of the structure of the auxiliary complex; some attempts to accommodate sets of quasi-modal items within extensions of this framework were discussed in 2.3.1.2. However, an obvious challenge was presented by the semantics of the modals, as illustrated by the ambiguity of: (3.102)

Rioters may break windows, (adapted from Ross 1969:88)

which can be paraphrased in either of the following ways: (3.103)

Rioters are allowed to break windows, (deontic)

(3.104)

It is possible that rioters break windows. (epistemic) 22

Evidence of this kind was taken to show the existence of two rather different syntactic/semantic structures, suggesting that epistemic and root readings in effect constituted a pair of homonyms; but while epistemic uses of modals were widely analysed as underlying intransitive, raising verbs, root uses presented more problems, as an analysis in terms of underlying transitive, control verbs was found to be unsatisfactory (for this discussion, cf Ross 1969, Newmeyer 1970 and 1975, and Pullum k Wilson 1977). The treatment of M - P alternation proposed by Larkin and Lakoff, which was discussed in 1 and further in 3.3.4, also belongs within this general tradition. The possibility of establishing some more satisfactory approach to the analysis of modality and the modals has been given some attention within the richer, and more abstract, framework of Government-Binding theory. In the following discussion, the major sources are the semantic analysis of some aspects of the English modals by McDowell (1987; also 1991), together with suggestions by Drubig (1993). McDowell does not deal with periphrastics as such, with the exception of comments on differences in use between can and be able to, and must and have to, which will be referred to later, but the absolute distinctness of the modals from normal lexical items, including periphrastics, is implied in the argument. She also assumes a basic epistemic-root distinction (radically different from standard analyses), but there is no subjective-objective dichotomy. 23 By contrast, Drubig treats periphrastics systematically alongside the modals, and takes the subjectiveobjective contrast as a basic and pervasive semantic distinction like that between epistemic and root senses. McDowell's analysis of the modals produces three distinct (1987:Ch.4): 22

23

groups of

items

There are, of course, further ambiguities in (3.102), in that both deontic and epistemic senses can apply to either present/general or future time. McDowell is essentially concerned with EM, which she, like most scholars, takes to be intrinsically subjective (1987:73), while the subjectivity-objectivity question does not arise in connection with non-epistemic modality.

69 (3.105)

Group 1

Group 2

Group 3

may - epistemic must - epistemic

may - future possibility must - deontic should will

can may - deontic

Syntactically, the items in the three groups in (3.105) are treated as follows (assuming an analysis of S as IP and S' as CP). Those in Groups 1 and 2 axe generated in I, and differ in that the first group (the 'true epistemics') are operators that move to COMP at LF, while the second group are lexical heads; those in Group 3 are generated in AUX, a category introduced by McDowell under VP, and subsequently move to I. This system reduces the epistemic and deontic systems to the contrasting pair must and may: this is to be related to the author's restriction of epistemicity to the expression of strong and weak quasi-assertions, directly relatable to the universal and existential quantifiers, respectively.24 However, this is complicated by the addition of a separate non-epistemic reading for 'future possibility' may, which is consequent on the categorical exclusion of future reference in EM; and if non-deontic 'future necessity' must were not ignored in this account, that would presumably similarly be given a separate reading; cf comments on this issue in 3.2. 25 Should, will and can are assigned unitary readings, and the remaining modals are treated as past tense forms. Drubig (1993) adopts the McDowell analysis, though with certain modifications, but suggests relating the subjective-objective dichotomy directly to M-P alternation. A basic dichotomy between two sets of modal uses, with periphrastics as alternatives for one of these sets, results: (3.106) exponents modalities expressed subjective or objective scope restrictions

" b

24

25

lexical items

functional operators

modals periphrastics root"

modals

objective narrow: over VP

subjective epistemic ?subjective deontic6 subjective wide: over S no questions not in conditions no stressing

includes objective epistemic, objective deontic, dynamic, and (doubtfully) subjective deontic this presents a crucial problem, since there are reasons both for treating the subjective uses of deontic and epistemic modals as parallel, and for isolating subjective epistemic sentence operators

Elsewhere, McDowell does not insist on this restriction, but allows for "at least two values" of quasiassertion, which leaves open the possibility of there being other exponents of EM (1991:318). Elsewhere, McDowell allows future reference quasi-assertions such as The bus must leave/be leaving tomorrow, which are clearly epistemic (1991:323-324).

70 Such a framework makes very specific predictions about the syntax and semantics of the items in question. For example, periphrastics are treated in (3.106) as synonymous alternative forms, sometimes required suppletively, for the root meanings of the modals, these being understood to cover the relevant epistemic, deontic and dynamic uses (this reflects model (3.73)(c) in 3.3.5). Also, this approach makes the modals systematically ambiguous, not only between root and epistemic senses but also, within these, between subjective and objective senses, resulting in four-way ambiguity (cf the four distinct interpretations for (3.21) in (3.22)-(3.25) above). Such claims can only be investigated in appropriate detail in the descriptive chapters below. However, some of the specific arguments supporting (3.105) and (3.106) can be investigated individually, and this will be done in the following sections 3.4.1-2.

3.4.1

Scope phenomena

Root meaning has often been associated with scope over the VP (narrow) and epistemic meaning with scope over S (wide). This is reflected in the contrast observed earlier between (3.103) and (3.104), and is a feature of (3.106), with the qualification that there the root-epistemic contrast is closely equated with the objective-subjective opposition. I shall follow the widespread practice of using negation as a convenient way of demonstrating scope relations: with wide scope, the negative comes within the scope of the modal and the main verb is negated, and with narrow scope, the modal itself comes within the scope of the negative. The discussion will draw on the analyses by McDowell (1987), Drubig (1993), and also Coates (1983:especially 237-239), who investigates this issue in some detail, and Palmer (1990:passim). (Perkins also reviews the whole problem of scope and modal verb negation, 1983:47-49.) The first three analyses are summarized in (3.107); Palmer's findings are too complex to set out in a simple form. (3.107)

wide scope (a)

(b)

(c)

26 27

McDowell (1987) 26 may - epistemic may - future possibility must - epistemic must - deontic should will Drubig (1993) subjective uses (primarily epistemic) Coates (1983) epistemic uses, except must27 must - root

narrow scope can may - deontic

objective uses (primarily root) root uses, except must

The wide scope items here are the items in Groups 1 and 2 in (3.105). Coates treats epistemic musi as non-negatable, the relevant sense being supplied by can't (1983:238).

71 As can be seen, there are some similarities - eg, the contrast between wide scope for EM and narrow scope for root modality in the case of may is implied in all three accounts (assuming, in the second account, that EM is only subjective and that root modality is only objective). However, there are some striking differences here. In what follows, I shall discuss the question of determining the application of scope in uses of should and will, since they highlight the problems here well. Should. To take root/deontic occurrences first, Coates presents the following as a typical instance of negated root should (1983:63): (3.108)

They [beggars] shouldn't be allowed to go about like that.

This is assigned narrow scope, to be paraphrased by 'it is not advisable . . . ' (1983:238). Palmer gives dynamic shouldn't the wide scope paraphrase 'it is necessary . . . not . . . ' (slightly modified) (1990:125-126). McDowell has the following example (1987:207): (3.109)

John should not miss class again.

This is similarly paraphrased for wide scope: (3.110)

It is (desirable (not (John miss class again))).

Finally, Drubig's claim, in (3.106), that items have wide scope when subjective and narrow scope when objective would suggest that (3.108) could be paraphrased 'it is advisable that . . . not . . . ' i f the speaker is giving some kind of instruction (subjective), and 'it is not advisable that . . . ' if a demand is being reported (objective). The general problem with these paraphrases is that there is no obvious difficulty if the alternative paraphrase, pointing to the alternative scope relation, is used, as in: (3.111)

It is (not (desirable (John miss class again))).

for (3.109), as Coates notes elsewhere (1983:239). To move to epistemic uses, Coates assigns epistemic shouldn't wide scope, but has no relevant instances. A possible case from the CEC follows: (3.112)

I should think you'd allow two and a half 28 , because it's an imponderable with traffic, isn't it? c shouldn't be that bad on a Friday morning A I wouldn't have said so, no (CEC:292-293/836-841) Β

Coates' paraphrase for shouldn't here would be Ί assume that . . . not . . . '/'It is probable that . . . not . . . ' (1983:238). Palmer paraphrases epistemic shouldn't 'it would be reasonable to conclude that . . . not . . . ' (1990:62,125-126), similarly entailing wide scope. 29 28 29

Sc hours. I am treating the use of should in (3.112) as epistemic, but it might be claimed that it is a case of merger or indeterminacy, since there is no ambiguity of meaning involved between root and epistemic interpretations (cf remarks by Coates 1983:14-17 and Palmer 1990:197-200).

72

McDowell only allows should a root reading, since her definition of EM is restricted to 'strong' (must) and 'weak' (may) statements of likelihood. Yet the example she uses, (3.109), is capable of a second epistemic-like sense, as in 'it is reasonable to expect that John won't miss class again' (or, perhaps, 'it is not reasonable to expect that John will miss class again'). This and the root reading shown in (3.110) are obviously distinct senses, since one can utter the sentence with a commitment to its truth in one sense but not in the other. 30 Finally, Drubig, as for root readings, assigns scope according to the subjectivity/objectivity parameter. This brief overview suggests a certain obscurity, or even arbitrariness, in the choice of a gloss for the senses in question, and hence in the assignment of scope relations for should. A further complication here lies in the fact that glosses for narrow scope can often be claimed to show negative raising. Will. Coates (1983:176) gives examples of root uses of will not/won't such as: (3.113)

Heather . . . was still repeating in sing-song: "I will not give in". (= Ί am not willing to give in'/'I refuse to give in')

as showing narrow yield a paraphrase the other hand, as with intend, could

scope. Wide scope here, which McDowell claims for will, would here like Ί am willing not to give in', which is obviously inappropriate. On Coates herself remarks (1983:239), an alternative type of paraphrase, suggest wide scope, as in "I intend not to give in", a possible sense for

(3.113). McDowell's sole example for will not (1987:207): (3.114)

John will not miss class again,

is paraphrased (3.115)

It is (strongly predicted (not (John miss class again))).

However, without 'strongly' in the paraphrase 31 , the following, with narrow scope, would also seem possible: (3.116)

It is (not (predicted (John miss class again))).

Similarly, a volition sense for (3.114) would seem to require a narrow scope paraphrase, just like (3.113); but will is only given one reading in this account. In conclusion. It looks as if plausible paraphrases for negative should and will sentences can be used to justify either narrow or wide scope, and this makes it difficult to justify absolutely any of the options illustrated in (3.107) for these items - (a) wide scope for all uses of most items, (b) wide scope for subjective and naxrow for objective, or (c) wide scope for epistemic and narrow for root. While these comments only apply to two 30

31

Thus, John should not miss class again, but I expect he will makes sense in the root, but not the epistemic, sense. This paraphrase with strongly only seems appropriate when will is stressed. It seems, in fact, to be chosen by McDowell to help create a semantic pair, or 'dual', with 'future possibility' may ('it is weakly predicted . . . ').

73 items, they suggest a considerable problem here. More importantly, it can, in any case, be argued that a clear-cut scope contrast between two readings of a modal only applies in the case of may (wide scope for epistemic and narrow scope for root uses), and even then the negated root use, as in You may not touch that ('you are not allowed . . . ') is rare and marked for formality: no such cases occur in the CEC, and Palmer records finding none in the SEU (1990:75).

3.4.2

Restrictions associated with subjectivity

I shall now turn to some major restrictions associated with subjective modality, and specifically associated with EM (cf (3.106)) - non-compatibility with (a) questions, (b) subordination in conditions and (c) the incidence of stress. I shall take these topics in turn. An incompatibility between questions and EM has frequently been claimed or commented on, and is explained within GB in terms of the impossibility of an epistemic operator moving to COMP when that position is occupied by a question operator, a block which becomes more evident if epistemic modals are treated as making a kind of assertion, and thus conflicting directly with an expression of interrogation (McDowell 1987:246). A block of this kind is also discussed by Coates (1983:242-243), where the absence in her corpus of epistemic questions with the items must, should, ought to, may, might and could is reported. However, an absolute block of such a kind cannot be claimed. Thus, Palmer (1990:62-63) reports some relevant occurrences: (3.117)

Can they be on holiday?

(3.118)

Would you ring me? Might that be the best?

He refers specifically to the rarity of must in epistemic questions, but treats the relevant use as grammatical, and elsewhere cites the following SEU text: (3.119)

I like to think about those days and how tough it was for the average Englishman what a hard life they must have had and mustn't there be endless stories about this mansion (1979b:193)

The two occurrences of must here aie unambiguously epistemic. Palmer treats the rarity of epistemic must? as reflecting the rarity of epistemic mustn't - in other words, epistemic must is rarely used non-assertively (Palmer 1979b). The occurrence of interrogative must in the following text is. also unambiguously epistemic: (3.120)

Cautiously he drew his downward gaze, riveted to the parapet surface, closer to him. He was keeping going, but how? Must he not suddenly tire? He felt tired, his movements less steadily mechanical. (I. Murdoch, The green knight. London: Penguin, 1994:27-28)

74 The ungrammatically of epistemic modal questions is apparently not universally claimed within the GB framework: Radford (1988:567,573) cites sentences such as Might John suspect that he will resign?, which is also unambiguously epistemic, as being grammatical. 32 The restriction claimed for epistemic questions is extended to include relevant tagquestions: McDowell (1987:266-267), treats the following as ungrammatical: (3.121)

*John must be a doctor, mustn't he?

Yet such tag forms, specifically mustn't, are common in corpus data, eg: (3.122)

but he must have thought it an eventuality, mustn't he? - he must have predicted (CEC:460/1083-1084)

The claim for a block on epistemic interrogatives and tag-questions is quite cleaxly unfounded. In contrast to this, McDowell cites examples such as: (3.123) (3.124)

Does someone have to be lying? Does John have to know the truth?

as evidence that the block does not apply to periphrastic items (1987:238). These sentences are attributed to Leech (1971)33, but they seem little better than their counterparts, such as Must John know the truth?. This block on epistemic questions with modals is absolute for McDowell, who does not consider 'objective' EM, but if the subjective-objective dichotomy is accepted as pervasive, it might be possible to argue that any apparent occurrences of epistemic questions, as in (3.117)-(3.124), are objective. It must suffice to say that these examples of EM all look standardly speaker-oriented, ie subjective. However, Drubig (1993) suggests that the restriction of questions to objective modality in EM may also apply to DM (this is a further illustration of the uncertain status of subjective DM in relation to subjective EM and objective DM, cf (3.106)). This would mean that the following typical text: (3.125)

May I leave my telephone number? (Palmer 1990:78)

would now only be interpretable objectively, which would mean that it would be glossed 'am I permitted/allowed (by some other authority) to leave my telephone number with you?', and the periphrastic form Am I allowed to ... ? would be equally appropriate. These consequences are also highly problematic. 32

33

T h e possibility of indirect epistemic questions, as in the following text with could have, provides another kind of evidence here: he eventually withdraws and in in all modesty makes a, some confession of his virtues - I was wondering if it could have been fear (CEC:869/1132-1135) This corresponds to a direct 'could it have been fear?'. This is misleading, since Leech merely says t h a t have (got) to has two ways of forming questions, without reference t o any availability for epistemic uses (1971:72-73).

75 The incompatibility of epistemic modals with syntactic subordination, specifically in conditional clauses, was discussed in 3.3.3, where it was suggested that the relevant data, if both EM and DM are taken into account, are rather complex and, while subjectivity and objectivity appear to be involved, no absolute restriction of subordination to objective modal use can be justified. The final syntactic restriction to be discussed here concerns the incidence of stress, specifically with EM, which is claimed only to have objective interpretation. Aijmer (1980:129-130) axgues that the Grundbedeutung of must is "true without exception (in all possible worlds)" (which points to an objective, even alethic sense), and that pragmatic weakening produces meanings that are closer to 'uncertain' than 'certain'. However, (3.126)

it always seems to be the case that the 'Grundbedeutung' can be preserved when must is stressed.

This claim regarding must cannot be accepted. 34 In the CEC, only one of the first ten occurrences of epistemic must does not carry stress, and the following is a typical text with two occurrences (stresses are marked): (3.127)

you 'must have 'had a 'copy of it - it's - 'must be 'somewhere (CEC: 125/1145-1147)

Here the speaker is anything but certain, and the stress on somewhere starts as a fallrise, characteristic of uncertainty or an appeal for information. Stress can emphasize positively, that is, indicate the high likelihood of the claim, but it also, apparently more often, serves to stress the uncertainty the item suggests. (But this is irrelevant to Aijmer's claim, which is for the certainty rather than the likelihood of the proposition.) In fact, epistemic must, which is generally agreed to be normally subjective, regularly carries stress; Householder, for example (1971:87), states that epistemic must always has its full vowel (which is naturally associated with stress), while Coates records the occurrence of roughly two-thirds of stressed epistemic occurrences in a corpus sample (1983:48-49). There is no reason for attempting to apply the claim for stressed must to root uses; stress or emphasis has no obvious correlation with 'objectivity', and one use of root must that is particularly strongly associated with stress is the type illustrated in I must say/admit/warn, which has no objective basis (Coates 1983:49). There are, then, serious difficulties with all these claimed restrictions on the use of subjective EM; and more generally, there is an impression that the application of categorical dichotomies such as we have seen in this section is simply not an adequate way of dealing with "the most complex and unsystematic aspect of English syntax" (Newmeyer 1970:196).

34

Similar claims are also made in relation to German (Öhlschläger 1989), and Hungarian (Kiefer 1986). It is possibly significant that none of these three accounts cites naturally occurring data.

76

3.5

Pragmatic approaches

In this section some explicitly pragmatic treatments of this area will be investigated, and the approach to be followed in the present study will be set out. This will involve concentrating on questions of polysemy, ambiguity and, in general, methods of contextual interpretation, in relation both to modality and, specifically, to M-P alternation. A pragmatic approach will entail maximally general semantic interpretations, which may tend to treat modals and periphrastics either as completely unitary items or as unitary within a basic root-epistemic dichotomy (to ignore more difficult cases such as can and will where such a dichotomy may not seem adequate to the polysemy involved). The justification for adopting such an approach to the meaning and use of modals and periphrastics rests on consideration of various factors such as the regularity of rootepistemic alternation and the semantic closeness of the various uses of individual items. This can be illustrated by Halliday's use of the following examples in illustration of the nature of the root-epistemic 35 dichotomy (1970:326-328): (3.128) (3.129) (3.130)

You must be very careful. You must be very careless. You must be very sympathetic.

Halliday assigns the first a deontic, and the second an epistemic interpretation, while the third is treated as ambiguous between these. In such cases, the dichotomy between the two readings is clear: (3.128) favours deontic unique future reference, and (3.129) epistemic present, or general, time reference, while (3.130) readily allows either. These interpretations are, of course, culturally determined: in principle, both readings are always available. This dichotomy is widely evident in the language, and not peculiar to modals and periphrastics (cf 2.3.2.5). On the other hand, an example such as: (3.131)

This weather must break soon.

shows an occurrences of must where the dichotomy between deontic and epistemic readings is not strongly, if at all, evident. Since we are concerned with M-P alternation, it is worth considering the effect if have (got) to replaces must36 in (3.128)-(3.131): now the same interpretations are possible, but (3.129) and (3.130) seem to favour deontic interpretation. This points to one kind of distinction between these items, since they have the same inherent polysemy but may favour different interpretations. In the rest of this section, I shall first discuss various approaches to word meaning, and then go on to the nature of pragmatic interpretation, and explain the approach to be adopted here. Subsequently I shall discuss some specific issues that have arisen in this connection. Several themes that have been discussed earlier in this chapter are, whether or not explicitly seen as such, interpretable as pragmatic in nature, particularly in relation to the possible identification of source and target (3.3.4) and some related descriptions (3.3.5). 35 36

Halliday does not use the terms 'epistemic', 'root' or 'deontic', but their appropriacy is obvious. I a m not suggesting t h a t have to and have got to are fully interchangeable here.

77 3.5.1

Various approaches to word meaning

The general approach to word meaning assumed here was discussed briefly in 1.5, and illustrated in (1.16). According to that discussion, lexical items have a 'semantic normality profilé' made up of their meanings in all possible contexts of use; and full synonymy between items involves the absolute equivalence of such profiles. Total synonymy in this sense will be difficult to establish and is disproved as soon as a single context can be shown where one item is more natural than another. In contrast to this kind of synonymy there is a lower limit of 'cognitive synonymy', which can roughly be equated with truthconditional equivalence. However, it was shown in 2 . 1 that equivalence of this kind may in practice be very difficult to demonstrate. Despite such problems, the informal notions of lower-level equivalence and an upper level of virtually unattainable synonymy suggest a useful methodology for the demonstration of the relation between semantically close items, and allow distinctions between items above the level of cognitive synonymy to be treated as pragmatic in nature. It is central to any approach to word meaning that it should be able to deal with issues of polysemy and semantic closeness between items, in this case primarily between M - P pairs. A number of approaches to this - specifically to the meanings of the modals - are possible and have been adopted, involving, for example, (a) extensive polysemy, (b) 'basic' or 'core' meanings ('core' in the sense of a specific semantic element - a lowest common denominator - present in every use of an item), (c) semantic matrices, and (d) componential features, but these all run into well-documented difficulties which need not be recapitulated here (cf Palmer 1990:14-20). It is significant that, in view of such problems, Palmer himself does not adopt any formal approach to word meaning, but emphasizes the considerable indeterminacy in the use of modal items (1990:20-22 and 197200) - a feature which is reflected in the various kinds of overlap in use between modals and periphrastic equivalents, illustrated below most clearly in 4 . 2 . 1 . Palmer suggests that modals are likely to be characterized by "a conglomeration of vaguely related meanings, each linked in some way to at least one of the others in the set, but not necessarily sharing any common feature with, or directly linked to, all of them" (1990:15). Related to this is the problem of establishing a principled way of determining distinctions between individual items, whether between modals or between M - P pairs. Matthews (1993:57-59) discusses various possible methods for eliciting, specifically, the meanings of modals in epistemic uses. He criticizes the use of sentential paraphrase, as in Hermerén (1978), constituent analysis, the 'lexical approach', which treats the modals as if they were freely occurring lexical items, as in Strang (1968), and corpus-based analysis, as in Coates (1983) and Palmer (1990), and argues for the 'ceteris paribus' method, which uses the fundamental linguistic principle of minimal distinctivity. The crucial question here is how fax this method alone can provide the insights required. Matthews applies it in exploring the distinctions between modal items as they are used epistemically in the frame He be at home as possible answers to a request for information. As an example, this shows the distinction between should and ought to to lie in the difference between what cannot be said with certainty but 'seems a reasonable prediction' - should, and what cannot be

78

said with certainty but 'seemed a reasonable prediction' - ought to (1993:61). But this supposedly rigorous method is obviously problematic, since what it produces here is two minimally distinct glosses which are only supported by the author's intuition; and it is quite unclear how these should be applied in explicating uses of the items. It seems, in fact, doubtful whether any one intuition-based approach should have privileged status, and it could, rather, be argued that the status of intuition itself requires justification. Nevertheless, minimal pairs and larger groups in this sense will be used on occasion in the present work, but only as one technique to help clarify potential distinctions between items. In general, if modals are treated as polysemous rather than monosemantic, differences, both in syntax and in semantics, can if necessary always be associated with distinct items that have the same form. If, on the other hand, they are treated as monosemantic, variations or alternations in meaning have to be explained pragmatically, and distinctions between uses of items are likely to be characterized by vagueness. In practice, scholars vary very widely over this issue. Of those who adopt a unitary approach, many assume a basic or primary dichotomy between root and epistemic uses, and may derive one from the other (eg, Antinucci & Parisi 1971, Kratzer 1977, Tregidgo 1982, and Sweetser 1990). Sometimes such a monosemantic system is favoured in principle, but proves to be not quite workable: for example, Ehrman (1966) finds unitary basic meanings for all the modals except may; Huddleston assumes extensive monosemy but also accepts that "at least some polysemy needs to be recognized" (1984:165), which appears from his discussion to apply in particular to will·, while McDowell (1987) claims monosemy as the general rule, and only allows ambiguity, and hence polysemy, for must (two-way) and may (three-way). None of these accounts deals explicitly with periphrastics as such. Of the three major accounts of the area of modality in English - Palmer (1990), Coates (1983) and Perkins (1983), the last two make fairly specific proposals regarding word meaning (they all take some account of periphrastic equivalents), and these will now be summarized. Coates (1983), motivated like Palmer by the overwhelming evidence of indeterminacy, uses a fuzzy set model which assigns items meanings characterized by 'cores', 'skirts' and 'peripheries'. The core in this sense (which differs from that referred to above) is the semantic focus of a meaning, corresponding to the cultural stereotype (ie, what is popularly assumed to be the essential meaning), and first acquired by children, although it may be relatively rare in occurrence. Most occurrences of an item belong in the skirt and periphery: the skirt covers a range or continuum of meaning stretching between the core and the periphery, while the periphery may represent an extreme value (eg, objectivity) and itself tend towards semantic independence. Coates argues that it should be possible in such a way "to be precise about indeterminacy" (1983:10), but, while it may suggest a good way of dealing with some problems of both ambiguity and merger (which here involve different types of interaction between two distinct semantic cores), this system can involve up to seven distinct meanings and five fuzzy sets for individual items, and can too easily serve to reinforce the normal impression of extensive, unmanageable polysemy.

79 Unfortunately for our purposes, Coates does not treat such periphrastics as come into her account in terms of the fuzzy set model. 37 Perkins (1983) is motivated by the desirability of avoiding the extensive polysemy so typical of accounts of modality, and modifies the notion of 'core' meaning - in the first sense used above - by positing basic meanings which function with sets of variables, and constitute 'semantic frames'. In this model, relevant laws (eg, rational or social; these indicate what kind of domain is in question, and correspond to EM and DM) determine that relevant circumstances (evidence or a deontic source) entail/preclude, etc, a consequence (the truth or the occurrence of the proposition). Clearly, this provides the framework for a considerable amount of vaxiation, and aspects of the frames can be further specified by, for example, constraints on the details of the 'circumstances'. Periphrastics axe also treated within this account, and as a general point modal verbs axe characterized as being less specified within these frames than any other modal exponents, and thus maximally unmarked. Perkins' model also suggests the possibility of a clear distinction between semantics and pragmatics in sentence interpretation. 38

3.5.2

Pragmatic interpretation

It is assumed here that, with the possible exception of a few problem cases, all the items under discussion are monosemantic. This means that the wide vaxiation in use must be explained in terms of pragmatic interpretation. It is possible, very roughly, to distinguish between different types or levels of interpretation in relation to these items, as follows (no claim for psycholinguistic validity is being made, and this may not be equally relevant to all items): (3.132)

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

type of modality; contextual inferences; conventionalized inferences; negotiation of ongoing discourse.

Firstly, there is the identification of the type of modality in question: since our account often concerns must, this can be illustrated by the discussion of this item by Smith (1989:Ch.8), who argues that the distinction between deontic, epistemicand alethic meanings is pragmatically inferred in context, with the 'basic', or default, meaning being deontic. This approach will be followed here, with the following important qualifications: (a) the priority of deontic interpretation is not assumed to be necessary true for all items, (b) the possibility is allowed that merging or indeterminacy may sometimes crucially blur the root-epistemic (or, indeed, any other) distinction, and (c) the notion of a distinct alethic meaning remains questionable (cf 3.3.2.1). 37

38

Coates' concept of a semantic core entails that intuitions about the meaning of an item may not represent quantitatively dominant use. A somewhat different conception of semantic prototypes involves meanings that are quantitatively dominant but not necessarily most typical within the grammar (cf Leech 1994:28-29). For further discussion of Coates (1983) and Perkins (1983), cf Westney (1984).

80 Secondly, assuming a clear distinction, where appropriate, between, for example, deontic and epistemic interpretations, there are a variety of inferences - in principle, infinite - that may be drawn in the further interpretation of modalized utterances. In the case of DM, these may prominently concern, for example, the source of the requirement, etc, its illocutionary force, its degree of urgency, the identification of the 'target', and the temporal reference. In the case of EM or other types of modality, somewhat different sets of inferences might be identified. Commonly occurring inferences may give rise to default interpretations - for example, in a deontic utterance with must, that the speaker, in the contextual absence of other agents, is responsible for the instruction, while with have to the speaker merely reports this; or that the temporal reference with DM is the immediate future, but with EM it is the present. It is specifically in the area of default interpretations that differences between semantically close items, such as an M - P pair, may arise. Common inferences, and related default interpretations, can become conventionalized in the language, and result in 'semanticization' (cf Hopper & Traugott 1993:75-86 on the role of pragmatic inferencing on grammaticalization). This is reflected in the tendency of modals, together with certain other items, to appear highly specialized in use (Quirk et al 1985:147): (3.133)

From the semantic point of view, modal auxiliaries are often specialized towards the expression of certain speech acts, eg, giving advice (ought to, should) . . . These characteristics of modals are shared in various degrees by modal idioms, semi-auxiliaries, etc 3 9 For example, had better and have got to are more closely associated with speech acts such as giving advice and orders than is the semi-auxiliary have to, which is capable of undergoing variation of tense 40 and aspect.

Quirk et al also refer to the tendency of modals, etc, to develop into 'pragmatic particles', as in the use of may in May you be happy! (wish), and also of informal want to in You want to be careful with that saw (warning/advice), in which uses the structure is fixed as to, for example, the position of may and the person and tense of want to (1985:147148). Such highly conventionalized, perhaps fossilized, uses can be treated as involving a distinct, third type of pragmatic interpretation. Other examples include the use, in appropriate contexts, of you must as a polite, friendly offer, as in You must have a drink, and of you can as a direct instruction, as in You can leave now, in each case uses that may appear to be far removed from their 'normal' meanings (for an review of the relation of modality to politeness in English, see Palmer 1990:191-194). These comments might suggest a correlation between restrictions on the morphology of an item and the speech acts it typically expresses, such that the modal-full verb gradient shown in (2.25) is characterized by a gradient between semantic/pragmatic specialization and polysemy. However, if this specialization is seen in terms of the process of conventionalization of commonly occurring inferences, it remains possible to characterize the modals as basically vague, semantically, and, more generally, maximally unmarked in 39 40

For these categories, cf 2.2, especially (2.25). This implies, misleadingly, that have got to has no past tense form.

81

relation to other items. Thus, the example of should/ought to referred to in (3.133) could be explained in the following way: the modal simply refers to the desirability, etc, of an action; in a second person use, the speaker is interpreted as the authority or source if one is sought and there is no other contextual evidence of one; and this type of use becomes conventionalized as a marker of personal advice. The point here is the fact that the specialization is facilitated by the items being basically vague or semantically unmarked. It may also be noted here that two items, such as the should/ought to pair referred to in (3.133), can apparently become conventionalized in similar or identical ways without their basic meaning being identical.41 Inferences in this sense have much in common with conversational implicatures, in that, for example, they are computed within contexts of use according to conversational principles (such as making sense, finding relevance or supplying missing information) that resemble Gricean maxims, and are defeasible. On the other hand, these inferences fall between generalized and particularized implicatures (cf Levinson 1983:126-127): they are attached to specific items, but also require appropriate contexts for them to apply. Finally, modality, or more specifically modal items, may be used in the negotiation of ongoing discourse between speakers. Thus, in relation to the following text (abbreviated): (3.134)

but it's basically saying you must watch your child twenty-four hours a day - you shouldn't go to bingo . . . and so you must watch your child all the time

Coates claims that, whereas semantically shouldn't expresses a weaker sense of obligation than must, in this context it is interpreted as a strong prohibition, contrasted with what the addressee must do (1990:58-59). In the present study, I shall normally be dealing with interpretation types (b) and (c) in (3.132). A final issue will now be discussed briefly. Since the contrast between epistemic and deontic/root modality is so prominent in the polysemy of modals and periphrastics (cf 2.3.2.5), we may consider what kinds of regularities constrain the development of this feature. If epistemic senses are seen as extensions from root senses - and there is every reason to see the standard (but not necessarily universal) line of development in this way - it is natural to look for some motivation for such a uniform semantic change. There have been various approaches to this; recently, the notion of systematic metaphorical extension from a real world or content domain to another, epistemic domain of reasoning has suggested a principled basis for the phenomenon (Sweetser 1990:Ch.3). In this approach, root-epistemic polysemy is seen as the conventionalization of a regular mapping between two distinct domains (op cit:64). Such an approach clearly suggests that root-epistemic alternation has a special status within pragmatic interpretation.

41

The comment by Taylor (1989:56) that two words may be distributionally and referentially equivalent but have different prototypes is relevant here.

82 3.5.3

Processing constraints

Haegeman (1987 and 1989) proposes locating the distinction between will and be going to in different contextual processing constraints. The two items in question are claimed to be truth-conditionally equivalent (cf 2.1), so that any distinction between them in use is pragmatic and must be relatable to demonstrable contextual factors:

(3.135)

at the level of sentence meaning be going to and shall/will axe equivalent, and . . . the difference between them is to be found in the constraints they impose on the processing in context of the utterance in which they occur. The two forms are thus interpreted as ways of grammatically encoding processing constraints on the utterance in which they appear. (1989:291-292)

The explicit claim here for truth-conditional equivalence between two related items (an issue not usually discussed), and for the consequent distinction between semantic (sentenceinternal) and pragmatic (discoursal or contextual) questions, is obviously attractive in that it suggests the possibility of a principled framework for the whole investigation. This initial claim is, however, problematic. If the items would and was going to are accepted as past tense forms of will and be going to, respectively, then truth-equivalence should also be expected of them. However, in relation to the following pair: (3.136)

The Queen would arrive three hours later.

(3.137)

The Queen was going to arrive three hours later.

Haegeman admits that "In the event that the Queen did not arrive, the first sentence appears to be false, but the second true" (1989:313), with resulting non-equivalence. She suggests as a solution that in such a situation the use of (3.136) rather than (3.137) would not be false so much as non-cooperative or counter to the pragmatic principle of relevance in that it would increase the processing cost for the interlocutor; however, Palmer (1990:149-150) sees this non-correspondence as fatal to the truth-conditional claim. Haegeman's aim is to present the widely accepted basic distinction between the items will and be going to (more strictly, certain uses of these items), in terms of present vs future 'orientation', within a theoretically satisfactory framework. The approach adopted is that of Relevance Theory, as set out by Sperber & Wilson (1986), and, more specifically, as applied to tense distinctions by Smith (1981; cf 1989: Ch. 10). This theory is concerned with the ways in which utterances axe interpreted or processed against a broader contextual background: linguistically, context is seen as a set of propositions that are of potential relevance to interpretation; and the use of a specific item or grammatical form may impose a constraint on the choice of such propositions. The way in which the interpretation of will and be going to works is set out as follows (Haegeman 1989:305):

83 (3.138)

Be going to ... imposes a constraint on the processing of the proposition with which it is associated. It signals that the proposition is relevant in a context including at least some present tense propositions, or, in other words, it guarantees a contextual effect if the utterance is processed against a present context. Will, on the other hand, signals that the hearer should extend the immediately accessible (present) context for the processing of the proposition and should process the utterance against future propositions.

Linguistic contextual material is not necessarily overt, but, presumably, any potentially relevant contextual data must be realizable as propositions. A present context exists by default, in the time and place of the production of an utterance (op cit:306). Some of the evidence presented will now be reviewed. First, the oddity of a will sentence such as: (3.139)

The rock will fall,

in contrast to its counterpart: (3.140)

The rock is going to fall.

when they are uttered in isolation (cf a similar pair in (3.98)-(3.99) and relevant discussion in 3.3.5) is explained not as a sentence-internal issue but as due to inadequate contextualization: in order to process a will sentence some future propositions are required, and in their absence there is only a default present context. The sort of proposition that would be relevant for (3.139) would be, for example, an «/-clause such as if you push it. Further data with conditionals are given (op cit:307): (3.141)

You're going to be fired if you ever go near his computer.

(—)

(3.142)

You'll be fired if you ever go near his computer.

(+) 4 2

(3.143)

You're going to be fired if you go on like this.

It is argued that (3.141) and (3.142) have a future time protasis (the reference is to any possible future event - ever), and in (3.141) this creates a context which clashes with the use of be going to, which requires a present context. On the other hand, the protasis in (3.143) is set in present time (like this), and this is appropriate for the use and interpretation of be going to. These are fully justified observations as far as they go, but Haegeman fails to add the fourth sentence to her set in (3.141)-(3.143): (3.144)

You'll be fired if you go on like this.

This is at least as good as (3.143), although according to the argument there should be a contradiction between will (future oriented) and like this, which is termed a "present context setter" (1987:16). Perhaps, therefore, the use of go on is to be treated as including future propositions. In this case, if you go on like this must be understood to include both present and future propositions, in order to provide suitable contexts for both will and be going to. But if a sentence can so easily contain overt, coexisting present and future 42

I have added these 'relative normality' signs.

84 propositions, the determination of linguistic 'relevant context', in the sense intended, is going to be a very difficult matter. Further discussion of this model of interpretation will be left for the relevant part of the description (5.3.1.3). Despite obvious difficulties here, the methodology of focusing on the actual or possible linguistic contexts of specific utterances, in terms of existing or 'recoverable' propositions, suggests the possibility of a systematic approach to the kinds of question being dealt with in this study (cf the justification of a contextual approach to synonymy and 'relative normality' in 1.5).

3.5.4

Tense, time and modality

A further topic that arises in connection with the previous section is the treatment of tense and time. Haegeman, like Smith (1981; 1989:Ch.lO), adopts the Reichenbach (1947) model as a way of showing, first, a distinction between will and be going to future reference forms and, second, the parallelism found between the will-be going to contrast and that between simple past and present perfect tenses, respectively. It is argued that tenses that differ only in the position of R (the reference point), but share the same 5 (time of speaking) and E (event time) are truth-conditionally equivalent (Smith 1989:111-112), so that he'll leave and he's going to leave, with, respectively, the representations S - R,E and S,R - E, only differ pragmatically, and this appears most simply and obviously in restrictions on possible adverbials, and more complexly in appropriate processing contexts (cf the claimed future and present contexts in (3.142) and (3.143), respectively). A parallel framework of time points would appear, similarly, to be potentially relevant to modal expressions, specifically in root modality, along the following lines. 43 A distinction can be drawn between the time of speaking (5), the event time, or time of due fulfilment (E), alnd the time when the modality becomes relevant, valid, is accepted, etc (to be termed M). The distinctions involved here can be shown by considering the following pair: (3.145) (3.146)

I can go there tomorrow. I'll be able to go there tomorrow.

In (3.145), understood in an 'ability' sense, can go is naturally related to present time, while in (3.146) will be able to explicitly belongs in future time. This distinction could be represented by treating M as either belonging with S (for present time) (ie, S,M - E), as in (3.145), or distinct from S and more closely associated with E (for future time) (ie, S - Μ,E or S - M - E), as in (3.146). Examples (3.4)-(3.7) and the related discussion in 3.1 axe also relevant here. I have omitted R here, on the grounds that it is superfluous for our purposes here. However, there seems to be no problem with its inclusion as a fourth factor. For example, 43

No special excuse is offered for adapting the Reichenbach system, since it has already been adapted in a number of ways - for example, by Comrie (1985), similarly for tense, and by Matthews (1989), for modality, while Lyons (1977:Ch.l7) uses a rather similar system, with the three variables t0 (time of utterance), t\ (time at which the proposition is true) and tj (time of which the proposition is true).

85 if I had seen the picture is represented E - R - S, and I had to see the picture is M,E - S or M - E - S, it would be reasonable to represent I had had to see the picture as M,E - R - S or M - E - R - S. This system will be applied specifically to must-have (got) to (4.5) and can-be able to (5.4.3).

3.6

The 'actuality hypothesis' and related proposals

In his treatment of contrasts in use between modals and periphrastics, which draws on observations in Kruisinga (1931) 44 , Anderson (1971:91-94) deals with various paradigms such as the following: la

I am able to swim

lb

I can swim

2a

I am able to come tomorrow

2b

I can come tomorrow

3a

I was able to swim in those days

3b

I could swim in those days

4a

I was able to catch the bus (this morning)

4b

-

5a

shall be able to run for 5 miles when I grow up

5b

-

6a

I (shall) have been able to swim for years

6b

-

Apart from can-be able to, the following sets of items are also included in Anderson's account: can/may-be allowed/permitted to, will-be willing to and must-have to/be obliged/required/compelled to. It is argued that in their non-epistemic senses each of the modals can be related to one or more be + adjective sequences; that these periphrastic counterparts are systematically ambiguous; and that thus the modal can only partially match its periphrastic syntactically and semantically. The gaps for 5b and 6b in (3.147) have no connection with the semantic restriction in question here, but simply show the sort of suppletive functions that periphrastics have to fulfil. The gap for 4b is what particularly concerns us, since this illustrates a well-known problem. For Kruisinga and Anderson, the crucial point here is that 3a denotes a state, and 4a an occurrence, and while the form was able to is ambiguous between these uses, could is only available for states, and not for occurrences; a state-occurrence contrast of this kind is not claimed to be relevant elsewhere in the paradigm in (3.147). 44

The terminology derives from Kruisinga, and is adopted by Anderson, but I only refer to the latter, since the claims made there are broader and more explicit.

86 There have been other approaches to this area. Firstly, Palmer deals with the relevant phenomena in terms of what can be termed the 'actuality hypothesis' (1977 and 1980; cf 1990:passim), which states that periphrastics, in contrast to true modals, are, in certain contexts, closely associated with the expression, or at least suggestion, of actuality. The basic argument here is based on data like the following, derived from Palmer (1980:92; 1990:1954S): (1.13)

*I ran fast, and could catch the bus.

(3.148)

I ran fast, but couldn't catch the bus.

(1.14)

I ran fast, and was able to catch the bus.

(3.149)

I ran fast, but wasn't able to catch the bus.

It is generally agreed that in a situation where the speaker did catch the bus, could is not possible, and was able to (or some other locution such as managed to) must be used. In a negative context, however, this distinction is neutralized, so that both could and was able to are possible. Palmer argues that in (1.14) was able to is available since it has an implication of actuality, absent in could, which conveys "that the event did, does, or will take place" (1990:194). Palmer uses the term 'implication', but this seems to denote the same as 'implicature' (Palmer 1977:1), and could perhaps be interpreted as a Gricean generalized conversational implicature, since, for example, it is cancellable, as is shown in the following extension of (1.14): (3.150)

I ran fast, and was able to catch the bus, but then I changed my mind and walked instead.

(but see the related discussion in 3.5.2). The 'actuality hypothesis' claims (3.151)

that in certain circumstances a true modal will have no implication of actuality and so may not be used if actuality is implied, whereas a closely associated semi-modal has no such restrictions or, at least, that if there is an implication of actuality the semi-modal is preferred. (Palmer 1990:194)

Later in the same discussion (195-196) Palmer explains the determining circumstances in terms of the interaction of modality, factuality, time reference and tense, along the following lines: (3.152)

45

The future is the period of time that has the least factual status; there is a sense in which we can never know the future. . . . By contrast, the past is the most factual; we can, and often do, know whether events took place. . . . Present-tense forms refer to events both before and after the immediate present, and so include events whose factual status is known and events whose factual status is not known . . .

The first three examples are from these sources; the last is added here for completeness.

87 Since modality is "concerned with events and propositions whose factual status is in doubt", it follows that modals, as the primary exponents of modality, will be most appropriate in reference to future events, and least appropriate in reference to past events whose factual status is established. However, periphrastics are not so restricted, and may be available, or required suppletively, for uses where modals are inappropriate. The essence of Palmer's argument can be shown as follows: (3.153)

time reference

possibility of establishing factual status

appropriacy of use of modal

future present past

least uncertain greatest

highest varied least

In relation to the use of pair can-be able to, this results in the following. Could is excluded when past actuality is implied, and can is inappropriate for present actuality, but often suggests future actuality; be able to must be used when the reference is to a specific action in the past, and is more likely to be used when there is an implication of actuality in present time. The contrast between the non-availability of could for past actions and the availability of can for future actions is to be seen in relation to the non-factuality of future time. An immediate problem with this hypothesis is presented by data such as the following: (3.154) (3.155)

I could just reach the branch. (Palmer 1990:95) But he got the envelope open eventually . . . Several things fell or could eventually be wrested out. (Amis 1991:183)

In these cases, it is cleax that the events referred by reach and be wrested out occurred. Some modification is therefore required; however, since both this specific problem with could and the more general application of the hypothesis to other forms of can-be able to, as well as its extension to other pairs of items (Palmer also deals with will-be willing to, must-have (got) to and will-be going to), are highly complex issues, further discussion will be left for the relevant descriptive sections. In the final approach to the specific problem of could and was able to to be mentioned here, McDowell (1987:185), following Ehrman (1966) in treating the basic meaning of can/could as simply "nihil obstat", and thus involving the potentiality rather than the realization of a proposition, argues that the use of (1.13) has the implication 46 Ί did catch the bus', and is thus inappropriate in that it is less informative than ... I caught the bus (McDowell here refers to an implication that could attach, however unnaturally, to the use of an utterance in a specific situation - eg, if, in this instance, it is known by the speech participants that the bus was caught, or that the ability or permission to catch 46

As with Palmer's theory, this seems to be equivalent to Gricean implicature, particularly in view of the subsequent reference to informativeness.

88 the bus is irrelevant). McDowell explains the appropriacy of (1.14), with was able to, as follows: (3.156)

Be able to ... focuses on one of the permitting conditions, the presence of learned or native ability.

Can/could, by contrast, says nothing about such permitting conditions. This approach allows the two items to be semantically equivalent, but with different conditions of use. McDowell's argument, then, claims that saying X and implying Y (Y being in some sense stronger than X) is underinformative, and hence could, which has the meaning of 'potentiality', is inappropriate in (1.13). However, this does not appear to explain why be able to, which focuses on "the presence of learned or native ability", should be more appropriate and adequately informative. Palmer's actuality hypothesis seems to offer more explanatory potential, and is applied to a set of M - P pairs, so this will be returned to in the descriptive section, particularly in 5.4.3, where the problem of data such as (3.154) and (3.155) will also be dealt with.

3.7

Style and variety

It has sometimes been argued that periphrastics are markedly informal or colloquial forms in relation to their modal counterparts. This would suggest that the choice of a periphrastic may be at least partly determined by the stylistic variety of English being used; it would, of course, be possible to distinguish between a formality parameter and a medium parameter (primarily speech vs writing), but for the present purposes it is assumed that there is a general correlation between informal and spoken language, and between formal and written language, and reference here will normally be simply to formality and informality. A further distinction can also be drawn between varieties, primarily between BE and AE (cf 1.4). Claims that particular items are stylistically marked for formality or informality, or are peculiar to, or typical of, BE or AE are clearly quite different in kind from the arguments we have been discussing so far in this chapter. Must-have (got) to and, in particular, will-be going to have attracted most attention among M - P pairs/sets in this connection 47 , and the issue will be gone into in detail for these items in the description below; would-used to have also often been treated in such a way. 48 Such a formality parameter has not, however, normally been observed in relation to the other items. As far as formality is concerned, Celce-Murcia L· Larsen-Freeman (1983) present what seems to be the only account that deals with a wide range of M - P pairs. Their basic claim is as follows (op cit:87): 47

48

For example, Wekker notes the strikingly low occurrence of be going to in written texts and, generally, in formal style (1976:124). Corpus research, however, has shown t h a t the differences in use between these two items are radically different from what - on the basis of intuition and 'conventional EFL wisdom' - is usually stated, including such a claim relating to relative formality (Willis 1990:49-51,55-56).

89 (3.157)

there is generally a difference in the degree of formality between use of a modal and the corresponding periphrastic modal, the latter being more informal, especially when phonologically reduced.

The initial line-up of matching pairs is as follows (ibid): (3.158)

modal - formal

periphrastic - informal

can will must should, ought to would (past habit) may

be able to be going to, be about to have to, have got to be to, be supposed to used to be allowed to

A subsequent listing, specifically for a deontic 'command-suggestion' scale, gives (op cit: 89): (3.159)

formal

informal

will must had better should might

hafta (have to) better (had better) oughta (ought to) could

These listings relate to AE; and while reduced forms like those included in (3.159), eg, hafta, are typical of informal or rapid speech in general, they are usually seen as rather more typically so in AE than in BE (cf discussion of phonological criteria for periphrastics in 2.3.3). There are some problems with the lists in (3.158) and (3.159). On a detailed level, the fact that ought to appears both as a formal and as an informal item looks rather odd 4 9 , while be to/is to is a surprising inclusion alongside be supposed to; and, more generally, the questions arise (a) how such distinctions can be satisfactorily demonstrated, and (b) if they can, whether they can be shown to relate solely to a formality parameter. It would, for example, be necessary to be able to show that be supposed to differs crucially from should (and, perhaps, ought to), or used to from would, in its informality. Such evidence does not seem to be currently available, though with the rapid growth of computer-based, corpus-related studies elucidation may be expected here. Regarding B E / A E variety distinctions in the use of these items, systematic comment has been rare, a notable exception being the account by Coates & Leech (1980), which will be referred to in 4.3. A basic, more general, question concerns possible relations between a stylistic parameter and other factors that may differentiate between the items in question. In principle, 49

Ought to in (3.158) is termed by the authors an 'intermediate form', classifiable as either a modal or a periphrastic, since the criterion used for periphrastic status is t h a t a multiword item end in to, and ought to regularly loses its to in negation in AE, a process also evident in BE (cf the findings in Svartvik L· Wright 1977, and evidence in 5 . 2 . 1 . 2 below).

90 a stylistic factor could be independent of other factors (so that it could determine the choice of item in cases where no other factor was crucial, or perhaps override the effect of other factors); or it could subsume or be subsumed by other factors.

3.8

Markedness

One general approach to the whole question of M-P alternation that has been referred to at various points above is in terms of markedness. This is a concept that can apply to any aspects of linguistic description, but here I shall be primarily concerned with semantic/pragmatic issues. I suggested in 3.3.5 that, if taken literally, some accounts suggest that modals are marked in relation to periphrastics, while others suggest the reverse; however, all the detailed arguments seem to point in the latter direction, and that is the approach I shall follow here. In discussing the speaker's perspective in relation to M-P contrasts, Bouma argues (1975:325): (3.160)

The reason that the speaker seems to 'go along with the obligation' . . . is that the use of the modal, as opposed to its periphrastic counterpart, allows the speaker to conceal his interpretation of the potential realization of the event. This in effect would mean that the modals are unmarked in regard to their periphrastic counterparts.

Bouma is here contrasting his approach to the nature of M-P relationships with that of Lakoff (1972), for whom "the modal forms are marked and their periphrastic counterparts unmarked for commitment of the speaker" (cf 3.3.4). In Bouma's account, markedness appears to relate solely to the speaker's attitude to the potential realization of the proposition, and he highlights the strategic advantage that the use of modals may entail for a speaker, since he is sheltered by their relative underspecification (which can, of course, result in ambiguity, whether intended or not 50 ). Perkins presents the question of markedness in much more general terms, and his account constitutes the major discussion of the issue (1983:102): (3.161)

The modal auxiliaries are, in fact, the least formally explicit of all modal expressions.

Formal explicitness is illustrated in partly semantic and partly syntactic terms. Perkins characterizes the meaning of modals as follows, using the terminology of his model (1983: 104): (3.162)

50

All they specify is the nature of the relationship between C and X (where X represents either the truth of ρ or the occurrence of e).

Cf Weiser (1974) on the deliberate use of such ambiguity.

91 ('C' here refers to a relevant, determining set of circumstances - evidence in epistemic uses, a source in deontic uses, 'empirical circumstances' in dynamic uses; 'p' to a proposition; and 'e' to an event.) By contrast (ibid), periphrastics (3.163)

(a) (b) (c)

axe often more specific as to C; are less restricted with regard to tense; in the case of forms beginning with be, such as be able to, are explicitly objective.

These three characteristics all constitute some form of restriction: in (a) there are pragmatic constraints on the determining 'circumstances' - eg, the nature of the deontic source; in (b), whereas periphrastics have more, or rather normal, tenses available, individual tense forms are more restricted in their use; and in (c) 'objectivity' (cf (3.60)) constitutes a semantic restriction on what is actually asserted in a proposition. Perkins further relates the modals' unmarkedness to their maximal degree of grammaticalization and of integration within clause structure, all of which gives them a more 'central' status in the language than periphrastics and other fully lexical forms of modal expression (op cit:104-105); these points have been touched on above in connection with the identification of criteria for modals and periphrastics (2.3). To illustrate this claim, I shall now show how the three restrictive characteristics of periphrastic items listed in (3.163) may operate in distinctions between can and be able to. To start with (3.163)(a), it was shown in 2.3.2.1 that there is some degree of restriction on possible subjects for be able to, in contrast to can, so that the sentence: (3.164)

Accidents are easily able to occur on that hill when it gets icy.

is odd in relation to its counterpart with can. This phenomenon could count as a restriction on C - perhaps a requirement either that the grammatical, or semantic, subject be animate, or that ability, a more specific form of possibility, be involved. For a further example, we can return to (3.4): (3.4)

I can go there any time I like at the moment,

which can correspond to either of the following: (3.5)

I am able to go there any time I like at the moment.

(3.165)

I am allowed to go there any time I like at the moment.

This shows that can is semantically vaguer than the periphrastics be able to and be allowed to, which can thus be seen as more highly specified in relation to Perkins' C. To take (3.163)(b) next, examples (3.4)-(3.7) in 3.1, and (3.145)-(3.146) with related discussion in 3.5.4, provided some evidence on the availability of can (of ability) in relation to that of be able to in present and future time reference. Those examples together with a further two follow: (3.4)

I can go there any time I like at the moment.

(3.5)

I am able to go there any time I like at the moment.

92 (3.166) , . ^ '

I can go there later. J I am able to go there later. 11'll be able to go there later.

(3.6) (3.7)

?I can go there when I'm older. I'll be able to go there when I'm older.

Such data show that the forms of be able to are more specialized than can, which can only alternate with could. In this sense too, can is vaguer, since it can, variously, correspond to is able to or to will be able to, or simply not be available, in which case be able to may be used s u p p l e m e n t a r y . In the following further real example, only a future form of be able to is possible, and can is excluded (Palmer 1990:98): (3.168)

I mean, there'll be his mother and grandad so we won't be able to do anything exciting

(cf 5.4.2 for further discussion). (3.163)(c) can be illustrated with the contrast between could and was able to claimed in the following pair from Perkins (1983:68), cited earlier: (3.61)

Ali could defeat anyone he fought with when he was 25.

(3.62)

Ali was able to defeat anyone he fought with when he was 25.

The use of be able to in (3.62) is said to be objective (due to the presence of be), entailing that the ability was actually realized (this argument was discussed in 3.3.3). Finally, in this section, we come to the claims by Haynes and Joos, briefly referred to in 1.1, that point to periphrastics being more highly specified semantically than modals. Haynes argues that periphrastics express "autonomic responses to external circumstances" (1967:212; cf (1.1)), and that their validity is contingent upon independently existent factors (op cit:25). In the case of must and have (got) to, for example, the distinction is presented as primarily being between duty or morality (not specified as externally imposed) and external determining factors, respectively. The following statement by Joos can also be quoted in this connection (1968:21): (3.169)

(1) (2)

modals and quasi-auxiliaries have totally different grammars; they have essentially different kinds of meanings: modals have privative meanings, quasi-auxiliaries are additive too. 5 1

(1) here can be understood as concerning the syntactic unmarkedness of modals, while (2) implies that modals have a basic, simple meaning, whereas quasi-auxiliaries have some extra semantic elements. Joos rarely compares M - P pairs directly; but in the case of can and be able to the latter is said to have as its principal value "time-focused freedom, timelimited freedom" (1968:27), which can be seen as a circumstantial constraint not present with can. Arguments of these kinds all suggest that periphrastic items are more highly specified semantically or pragmatically, while modals, forming a sub-class of function words in the vocabulary, have a reduced or underspecified semantic content. As was noted at the outset, a markedness hypothesis of this kind may subsume other claims for M - P distinctions such as have been discussed above. 51

The senses of 'privative' and 'additive' meaning are complex and peculiar to Joos' account, and need not concern us here.

93

3.9

Conclusion

A full account of the phenomena we have been discussing in this chapter would include the syntax, semantics and pragmatics of the items in question. The aims of the present study are, however, restricted to examining factors that may be relevant to differentiating between the use of modals and that of related periphrastics. In 3.5 I sought to justify the choice of a primarily monosemantic approach which leaves the bulk of 'explanation' to the pragmatics, and the approach outlined in 3.5.2 will provide the basic framework for the description in the following two chapters. In this chapter, a number of specific themes were dealt with which axe of potentially general relevance here - in 3.2, the distinction between various types of modal meaning and in 3.3 various notions relating to subjectivity and objectivity. Aside from such general, pervasive issues, we have also seen a number of specific proposals, whether or not explicitly pragmatic; these will all be taken account of in the description where they seem relevant. In addition, other, quite distinct issues were discussed - specifically the operation of something like an 'actuality' factor (3.6), variation in use according to style and variety (3.7) and the notion of markedness (3.8). It remains to be seen how these various factors can be related together, but the assumption will be made that it is likely to be worth considering M-P relations in general terms of markedness, and more specifically in terms of some kind of subjectivity-objectivity contrast. To return to the three major types of potential contrast between items discussed in 3.1, questions concerning the actual degree of equivalence are central here, and ultimately relate to assumptions made about the nature of word meaning (cf 3.5.1). The possible nature of the relationship between items was, of course, the primary subject of this chapter. Finally, the extent of specific M-P relations is yet to be demonstrated: many of the relevant discussions have been in terms of one or perhaps two pairs or sets, and a major concern of the present study is the question whether there are useful generalizations across pairs/sets of M-P items.

4.

Must, have (got) to, need (to) and be bound to

4.1

Introduction

As was said in 3.1, the items must, have to and have got to will be discussed in much more detail than the other sets of modals and periphrastics, for reasons such as the following. Firstly, must and have to, and, to a much lesser degree, have got to, have received far more discussion in the literature than other items, and, as was shown in 3, a wide variety of descriptive claims have been made for semantic distinctions between them. Secondly, must and have to have been treated as the paradigm case for M - P pairs (cf, especially, Lakoff 1972). Thirdly, have (got) to1 is a quantitatively dominant item, being one of the most frequently used of the periphrastics discussed here 2 - sometimes, possibly, even more than must (cf data in 4.3.1). Fourthly, have got to has received very little individual treatment: among accounts of these items, only Coates (1983) and Melrose (1983) deal with have got to on its own terms, rather than as an alternative form - perhaps a stylistic variant - for have to, while the monograph on have to by Tottie (1971) deals with syntactic, but not semantic, distinctions between that item and have got to. Finally, the very variety of accounts of differences among these items constitutes a challenge, since they can hardly all be equally valid in their observations. Need (to)3 and be bound to are also included in the discussion here, for reasons summarized in 2.4; but they will only be given brief treatment. The introduction of need (to) involves blurring the contrast between modal and periphrastics, since it functions both as a marginal modal, alongside ought to, etc, and as a full lexical verb, and is thus, within our terms, typical neither of modals nor of periphrastics. While need (to) is discussed in close relation to must and have (got) to, be bound to is treated separately, since it is semantically much less close to must than the other items are. This chapter will be organized as follows. First, some basic descriptive problems will be reviewed, and the general approach to be followed will be set out (4.2). Questions concerning possible distinctions in the use of the items according to both variety and style or formality will be discussed next, together with supporting quantitative data (4.3). The use of the items must, have (got) to and need (to) will then be compared in detail, first in deontic meanings and in affirmative, present tense forms, with, in turn, first, second and third person subjects, in order to allow a maximum variety of linguistic contexts (4.4). Then other aspects of the deontic use of the items will be described (4.5). Epistemic uses will be treated separately (4.6), and be bound to will then be described on its own (4.7). The whole discussion will be summarized in 4.8.

'Have (got) to' will be used throughout to denote have to a n d / o r have got to in contexts where the items « e not being considered separately or in contrast. Such d a t a as are available suggest t h a t only be going to may be more frequent. 'Need (to)' will used here to refer to either the marginal modal or the lexical verb; if these are to be distinguished, the former will be termed 'need', and the latter 'need to'.

95

4.2 4.2.1

Framework of this discussion Some basic problems

The first subject to consider is the possibility sometimes argued for that must and have (got) to are, aside from formal restrictions and suppletive functions, partially or completely synonymous. The following texts all, in various ways, suggest a high degree of equivalence between the items: (4.1)

As a your feet, (The

teacher, you must learn to recognize them and do your best to help all pupils realize their potential - you have to be able to think on your continually adapting your lesson to the feedback you get from them. Observer, 20.5.1990:13)

(4.2)

While graduates of other English universities must spend another year in study after their final BA exams to change from being bachelors to masters, all Oxford graduates have to do is wait a few years and make sure they have paid their college bills. (The Observer, 21.4.1991:69)

(4.3)

"I can live with the arrangement, I can live with the general idea, and the details will sort themselves out perfectly well. They will because they must, they've got to. We're all under necessity." (Murdoch 1990:187)

(4.4)

But teachers live in the present. They have to teach now rather than wait for a whole L2 learning framework to emerge. They must get on with meeting the needs of the students, even if they still don't know enough about L2 learning. (V. Cook, Second language learning and language teaching. London: Arnold, 1991:152)

(4.5)

They have been understanding with me since Lockerbie and they have put up with a lot but sometimes people just have to do what they must. (The Observer, 16.2.1992:2)

(4.6)

"I'm afraid I've got to be off now. My parents have come to see me unexpectedly." He paused, to give room for any cries of protest and regret. When none came, he hurried on: "Thank you very much for putting me up, Mrs. Welsh; I've enjoyed myself very much. And now I'm afraid I really must be off. Good-bye, all!" (K. Amis, Lucky Jim. Harmondsworth: Penguin, 1961:80; cited by Nehls 1986:36-37)

(4.7)

"You've got to go, young man. We need your bed. This is a New Zealand hospital and we must put our own lads first. We've had a signal warning us to prepare for casualties. . . . So, there's nothing for it. We have to accommodate them." (Manning 1982:441)

96 (4.8)

and I and I and I think that's all you've got to do - you don't have to be elaborate with them in any way (CEC:328/1395-1397)

(4.9)

It these photographs were genuine, and if these things were all found together, they had to be, they must be, what I hardly dared put in words. (P. Levi, The head in the soup. London: Constable, 1979:14)

(4.1) and (4.2) show must followed by have to, in parallel functions ('this is what being a teacher involves'/'this is what is needed for a BA to become an MA'), while in (4.3) must is immediately followed by have got to as if it is an alternative or preferred formulation. If the order must ... have (got) to illustrated in (4.1)-(4.3) suggests that have (got) to is in some sense 'stronger', in that the later position tends to carry more important information (according to the principle of end-weight, cf Quirk et al 1985:1361-1362), examples (4.4)-(4.6) show the reverse order, with have (got) to preceding must.4 There is possibly, however, a subtle distinction here: whereas in (4.1), for example, must and have to are roughly parallel in function, in (4.4) have to reports a general requirement, and the succeeding must functions to provide illustrative detail (for further related discussion, see 4.8.1). (4.9) also shows the order have to - must, and this is a more telling case, since the items are juxtaposed as if in repetition. (4.7) has all three items in one speech: have got to is used for a second person obligation - the closest in this text to giving an order, while must and have to are used for parallel reported first person obligations. The latter two might possibly be felt to differ in that the have to obligation is imposed from outside (the speaker has been warned to prepare) while the must obligation expresses a more general principle; but the two items are easily interchangeable. (4.8) shows have to and have got to in juxtaposition: there seems to be no difference in effect if the uses are reversed, as in ... that's all you have to do - you haven't got to be elaborate . . . , so, again, equivalence between the items is strongly suggested. Finally, (4.9) shows had to and must together, for past time reference, in free indirect speech. I shall not speculate further here about the effect of the uses of these items in the texts discussed, but it is abundantly clear that at least in some contexts there is minimal distinction between them. Some of these texts will, however, be returned to in the description below. It should be noted that eight of the nine texts discussed are from writing (journalism, fiction and popular academic - ie, relatively informal), and only one from free conversation, and all are in BE: possible distinctions in use according to both variety and style will be discussed below in 4.3. It is more difficult to show need (to) in close contiguity with the other items, but the next text may suffice to suggest a close semantic relatedness between need to and must. Normally need to is considered to indicate a weaker requirement than must, but in the following no absolute difference is evident: (4.10)

A country's success needs to be judged not by the spectacular wealth of those at the top, nor even by its prosperity in some average sense, important

It is also worth noting that (4.1) and (4.4) are similar both in content and in stylistic level.

97 though that is. It must also be judged by the well-being of its poorest and most vulnerable. (The Economist, 18.5.1991:19) Here both modal verbs refer to how wealth is to be judged in terms of various sections of society, and they could be interchanged without creating any anomaly.

4.2.2

General approach

The general aim of the present investigation is, as set out in 1.3, to explore the possibility that, despite evidence such as that just reviewed, there is a systematic distinction between the use of modals and that of related periphrastics, and, more specifically, as illustrated in detail in 3, that this distinction may centre on a contrast between subjectivity and objectivity. In seeking some basis for exploring the distinctions between these items, it seems relevant to consider the fact that have to and have got to are relative newcomers in English in comparison with must. Have to entered English late and was apparently relatively rare till very recently: Jespersen documents some aspects of its development, and reports, for example, only eight occurrences of this item in 300 pages of H.G. Wells (early 20th century fiction) (1940:204-205).5 This suggests that have to may be both narrower or more specialized in its use than must, and felt in some ways to be less widely established. The relevant meaning of have in the OED under which have to appears (no 7) is: (4.11)

To possess as a duty or thing to be done. With object and dative infinitive expressing what is to be done by the s u b j e c t . . . With infinitive: to be under obligation, to be obliged; to be necessitated to do something. It forms a kind of Future of obligation or duty. (1989:Vol.VII,17)

It has been claimed that have to can be used to convey no more than an intention to perform a future activity, with no element of obligation, and it has been suggested, on the basis of uses such as the following: (4.12)

What are you doing tonight? Oh, I have to go to a party.

that the item could in time weaken into a marker of pure futurity (Fleischman 1982:59). This would parallel developments in Romance, since the origin of have to can be related to the type of Romance formation illustrated in the French future -ai ending (OED:ibid). The association of have to with have in the OED definition is supported by Cowper, who argues that there is only one item have, and when this has a to-infinitive complement the interpretation is "some individual (the subject of have) bears an unspecified but pragmatically predictable relation to an obligation or desideratum" (1992:119). The notion in (4.11) that with have to an obligation is something that a subject may "possess as a duty or thing to be done" clearly suggests an orientation in the use of 5

The significance of this figure is explained in 4.3.1, especially in Footnote 8.

98 the item towards an external, definable, existing obligation, and - particularly since this accords with many proposals discussed in 3 - this will be adopted as a hypothesis in examining its use in contrast to that of must here. As to the status of have got to, which seems only to have been in regular use since the middle of the 19th century (Jespersen 1931:51-52), and is much more restricted in its applications, the notion will be adopted here that its use is associated with emphasis (cf Jespersen 1931:53); in a similar sense, Melrose (1983) proposes that it expresses 'affective modality', a kind of emotional loading that standard types of modality may receive. These observations are in principle independent of any question of stylistic differentiation between it and have to (cf 4.3.3). In contrast to these items, must is relatively generally or vaguely characterized in the OED in terms of necessity, obligation, "a fixed or certain futurity", etc (1989:Vol.X,138). Need (to) has been discussed rather little so far. Modal need diifers from the other items in normally being restricted to non-assertive uses. However, need to, specifically in the form needed to, is often seen as comparable in meaning to have to/had to. As a simple demonstration of possible dilference here one could compare the following pair: (1.12)

I have to get up at 7 am every day.

(4.13)

I need to get up at 7 am every day.

An evident distinction here is that need to focuses on the existence of personally motivating reasons for an action rather than its necessity; and this observation will be applied in the ensuing discussion. In view of these considerations, the following minimal hypothesis is proposed as a basis for the description of must/have to/have got to/need (to) in this chapter, with primary reference to deontic uses but potential extension to epistemic ones: (4.14)

Assuming an 'obligation/necessity 6 ' that some proposition be fulfilled, (a) (b) (c) (d)

must makes a maximally unqualified statement as to the requirement; have to focuses on the obligation/necessity as an independent, existing entity; have got to diifers from have to in adding some degree of urgency or emphasis; and need (to) focuses on the motivation for the requirement.

At the same time, the first three items are treated as being, at least potentially, cognitively synonymous. (4.14) takes no account of the possible relevance of varietal or stylistic factors to distinctions in the use of the items; this issue will be taken up in the following section (4.3) before we start on the descriptive sections (4.4-6).

6 These terms will be used here, as they often are, to characterize the essential meaning of the items under discussion: in some contexts one or the other may be preferred when it seems more appropriate, but no special significance is to be attached to their use.

99

4.3

Variety and style

Before going into the general examination of data, I shall discuss the possible relevance of variety and style to the use and choice of these items. As was stated above (3.7), we are concerned here with two broad fields of potential variation in use: variety, primarily in terms of systematic B E / A E differences 7 , and, within a variety, style. Strictly, 'style' should involve at least two distinct parameters: medium (written, written to be spoken, free spoken, etc), and formality (from formal to informal). However, these two factors are generally (eg, in the relevant account by Coates & Leech 1980) treated as largely coinciding (ie, written = formal; spoken = informal), and so they will be here, but the potential distinction of medium and formality parameters will not be ignored, and stylistic contrasts will be seen as diverging tendencies rather than dichotomies. Three particular areas will be examined: the possible general relevance of variety and style factors to the use of must vs have (got) to, the relevance of these factors to the distribution of deontic vs epistemic uses, and the widely observed informality of have got to.

4.3.1

Must vs have (got) to

I shall start with some figures designed to show the use of the three items in written and spoken language, in both BE and AE. Coates (1983:23; cf also Coates h Leech 1980:26) reports, from the million-word Lancaster corpus of written BE, 1131 occurrences of must and 695 of have to, which makes have to much the commonest of the periphrastics listed there (the next commonest, be able to, having 275). The parallel Brown corpus, for written AE, has 1011 for must and 647 for have to (Coates & Leech 1980:26). 8 It is not stated in either case whether the have to figures include have got to, but it seems that they must do, since at one point (Coates & Leech 1980:31), epistemic have to is illustrated by the example: (4.15)

This has got to be some kind of local phenomenon.

Another count for written AE gives the following figures: 35 for must, 4 for have to and 1 for have got to (Melrose 1983:52; corpus of 96,915 words). For that study, only present tense forms were considered 9 , on the grounds that only here are genuine choices possible, The use of modals (including some periphrastics) in Australian English has been investigated (Collins 1988), and to some extent contrasted with the situations in BE and AE (Collins 1991). However, it is not clear to me that there are any significant differences between Australian English and BE in the area being investigated here (on which, cf, also, Trudgill h Hannah 1982:19-20), and I do not find that any of the data in Collins (1991) would be odd in BE. It is instructive to compare these figures for have to with the figure for a count from H.G. Wells (early 20th century fiction) quoted in 4.2.2. If an average of 350 words per page is assumed, that figure of 8 occurrences in 300 pages is equivalent to 8 in 105,000 words, or c 80 in a million. This represents a frequency rate for have to in Wells of c 12% o f t h a t in the present-day written language counts quoted in the text, and supports the notion that have io has become much commoner relatively recently. The use of 'present tense forms' includes the must/have (got) to have + -ed participle forms.

100 as in other cases the option is grammatically determined. 10 If all the forms found were included, the total for have (got) to would apparently have been 30 (op cit:51). For spoken language, my figures for B E from the CEC (total: c 220,000 words) 11 are 253 for must, 301 for have to and 94 for have got to, while for AE, Melrose (with a corpus of 100,522 words) has 7 for must, 66 for have to and 35 for have got to for cases where there is a choice (cf above); the total count, including 'grammatically determined' occurrences, would apparently give 159 for have (got) to (1983:39-40). These figures can be set out as follows: (4.16)

occurrences of must and have (got) to in all uses must written BE

AE(1) AE (2) spoken BE AE

have to have (got) to

have got to

? ? ?

?

4

1

301

94

?

?

66

35

1131 1011 ; 35 i 35

695 647 30 5

253

395 159 101

f

7

i

7

?

?

total

1826 1658 65 ί 40a J 648 166 ι 108" /

source6

(a) (b)

(c)

(d)

(e)

a

these figures are only for present tense forms

b

sources: (a) Coates (1983:23); (b) Coates & Leech (1980:26); (c) Melrose (1983:51-52); (d) my counts from the CEC; (e) Melrose (1983:39-40)

The following are the major conclusions to be drawn from these figures: (4.17)

(a) (b) (c)

must is commoner in written, and have (got) to in spoken language; within spoken language, must is markedly commoner in B E ; must is markedly uncommon in AE speech.

Contrasts of the kind set out in (4.17) are not normally reported, in particular the applicability of (a) to B E as well as AE. Thus, in their investigation in this field, Coates L· Leech (1980:32) suggest a distinction between AE and B E along the following lines: "musí = formal, have to = informal" in AE, while the items are "closer to a state of stylistic free variation" in B E . The first claim is clearly supported by the data, but the second merits comment. Coates & Leech rely for the formal/informal contrast on figures from non-fiction and fiction texts, respectively; but the figures set out above suggest that there 10

11

This decision makes the assumption that an occurrence of must alternates with one of have/has (got) to, and does not allow for must and shall/will have to being alternative options in certain environments (this issue will be discussed in 4 . 5 . 1 ) . This also means excluding data for the past tense forms had to/had got to. The nominal total of 5,000 words for each of the 34 texts gives a total of 170,000 words, but this excludes contributions by 'non-surreptitious' speakers: if these are included, as they are in my counts, a real total of roughly 220,000 words is arrived at.

101 may be a big difference between taking data from fiction (which is written language) and using naturally occurring conversation as evidence for 'informal' language use. A formality distinction relevant to the use of must and have to is also argued for by Bouma, in an AE context (1975:324): (4.18)

. . . must is rarely used except in writing and formal speech. In casual conversation and in non-formal writing have to is dominant, so that the two are essentially in complementary distribution.

This is seen as an overriding distinction in use between the two items (cf conclusion (c) above). The distinction proposed by Bouma is not accepted by Palmer as valid for BE (1990:116), on the grounds that must is common in colloquial BE; the data summarized in (4.16) are hardly decisive here, since they suggest both that must is common in colloquial BE, and that it is overall more typical of written language. The crucial point here, therefore, is the predominance of have (got) to in spoken texts. If distinctions of the kind set out in (4.17) are generally valid, the consequences for description in this area must be considered. If the notion of a kind of 'complementary distribution' of the items (as in Bouma) is accepted, it would follow that other kinds of contrast claimed for them - eg, involving subjectivity vs objectivity - must be correspondingly restricted in their relevance, since, in principle, the actual uses or meanings of these items could vary according to variety and style. This could suggest a radical reassessment of the data: thus, it might be claimed that an example like: (1.4)

My girl must be home by ten.

is rather odd in some spoken contexts, so that the claimed contrast between it and: (1.5)

My girl has to be home by ten.

becomes questionable. It might also suggest that, for example: (1.9)

We must also sit idly by as Serbia and Croatia slip into economic anarchy, with all the dangers that this implies.12

would be normal in writing but not in speech. Similarly, a disputed case like: (1.11)

*I must get up at 7 am every day.

might turn out to be odd in casual speech, but normal in formal use, so that assessing its grammaticality would require setting it in an appropriate context (which might, of course, be difficult). This could make much of the data in the literature hard to discuss, since they are most often invented and presented neutrally as to medium and variety. In the ensuing discussion I shall assume that variety and style must be taken into account, and may be relevant.

12

An extended version of this text was given in 1.2.

102

4.3.2

Epistemic vs deontic uses

The data in (4.16) simply deal with the total occurrences of the items. We can also consider the individual uses of the items. Coates & Leech (1980:31-32) compare the figures for epistemic and deontic uses in BE and AE in written language, and show, in particular, that epistemic must is markedly commoner in BE, and epistemic have (got) to in AE. Melrose (1983) provides figures relating to the three items in spoken and written AE in relation to four types of modal meaning (these are simplified here as epistemic vs deontic). Finally, figures for spoken BE show a marked contrast between must and have (got) to, confirming the general assumption that epistemic have (got) to is rare in BE, but also showing that in spoken BE epistemic uses of must are at least as common as deontic uses. The figures referred to here for epistemic vs deontic uses of the items can be summarized as follows: (4.19)

occurrences of epistemic and deontic uses of must and have (got) to (figures are shown for epistemic and deontic uses in turn - eg, 74:153 means 74 epistemic and 153 deontic cases) must

have (got) to

written BE AE(1) AE(2)

74:153 47:150 3:32

2:226 9:209 0:5

0:4

0:1

(b)

spoken BE AE

126:125 5:2

2:393 6:95

0:301 4:62

2:92 2:33

(c) (b)

a

have to

have got to

source"

(a) (a)

sources: (a) Coates k Leech (1980:31); (b) Melrose (1983:40,52); (c) m y counts from the C E C

Revealing as these figures are, the data involved are not sufficient to warrant strong conclusions such as were set out in (4.17) on the basis of (4.16). As to the widely observed recent increase in the use of epistemic have (got) to in BE, it should be noted that the BE data used derive primarily from the period between the mid 1950s and the mid 1970s, and thus may not reflect current practice.

4.3.3

Have got to as an informal item

Palmer observes that have got to - in contrast to have to - "belongs to a more colloquial style and generally appears only in the spoken texts" (ie, of the SEU) (1990:114). The colloquial nature of have got to, a relatively new form in English, has often been commented on. For example, Poutsma writes (1904:37) "in colloquial language have13 is 13

Have to and have got to are being referred to here.

103 often followed by a redundant got"·, in a prescriptive context, have got to has even been characterized as "colloquial, verging on the low colloquial" (Krapp 1927, cited by Mittins et al 1970:34). This general point is substantiated in numerous ways in the monograph on have to by Tottie (1971). Moreover, the relative status of the two items parallels that of have and have got, which would provide further support for analyses that subsume have (got) to under have (got). The data summarized in (4.16) are, unfortunately, not extensive enough to provide crucial support here, but they do show that have got to is widely used in both BE and AE; and Melrose (1983:21,44) justifiably objects to the claim sometimes made (eg, Leech 1971:73, and Leech & Svartvik 1975:130) that have got to is 'typically British', or rare in AE. It will be assumed here that have got to is generally typical of informal language, and is overall markedly less common than have to in both BE and AE.

4.4

Deontic uses of must, have (got) to and need (to): present tenses, affirmative

I shall now first discuss occurrences of these items in deontic uses (this term being used to subsume some idiomatic uses of items that are only deontic in a broad sense, as in sense (e)(1) in (3.14)). The account draws primarily on corpus texts, with, in turn, first, second, and third person subjects. The three persons are treated separately since typical contexts of use with each may be assumed to vary widely, resulting in different persons showing different prototypical uses, as is suggested by the following comments on must in the OED: (4.20)

In the 1st person, must often expresses an insistent demand or a firm resolve on the part of the speaker. Hence also in the 2nd and 3rd persons, rendering sentiments imputed to others. (1989:Vol.X,138)

Examples supplied include: I must, and will go; let us leave this room, if you must laugh; and he must have rings in his ears. In view of the lack of speaker involvement ('subjectivity') in the second and third person cases, such examples, on their own, could well suggest that the use of must indicates the involvement of the grammatical subject rather than that of the speaker. 14 In each case, must, have to, have got to and need (to) will be discussed in turn. The first three items will be regularly contrasted with one another, while need (to) will only be considered intermittently; (4.14) will provide the basis for the description. General questions of the use and interpretation of present tense affirmative forms will be dealt with in the rest of 4.4; further, specific aspects of deontic uses, including future and habitual reference, will be dealt with in 4.5, and epistemic uses in 4.6. 14

Treating I musi distinctly from you must, etc, also receives support from current corpus research (Hunston 1993).

104 4.4.1 4.4.1.1

First person subjects First person subjects with must

Just as second person must has often been associated with the imposing of orders on others (cf 3.3.4-5), first person must has often been seen as marking the coincidence of source and target, thus expressing, for example, 'self-compulsion' or 'self-obligation' on the part of the speaker (Leech 1987:77). First person have (got) to can then be shown to contrast with must in expressing obligation imposed externally. The following text illustrates this use of must: (4.21)

and then, one night, I suddenly thought, oh hell, you know, I must see somebody (CEC:223/184-189)

This reports a spontaneous decision (suddenly, oh hell). Replacement of must by have to here would suggest some motivation external to the speaker, and hence would not suggest spontaneity; the use of have got to would make little difference from that of must, as this item would, above all, suggest some kind of urgency, whether or not the motivation were external to the speaker. The next text is similar, but lacks the spontaneity of (4.21): (4.22)

. . . she thought: I must have patience. I must wait for her to speak first. She has to be the one to choose the time and place. (James 1986:320)

This seems to make a contrast between first person must and third person have to. As in (4.21), the musts here are vague as to source, though a natural interpretation points to the speaker, while the contrasting has to points to some externality (since the speaker feels that she cannot herself initiate the hoped-for discussion). In spite of examples like these, where a speaker-based obligation is a natural interpretation, a large number of uses of must do not fit into this pattern. I shall now discuss three roughly distinguishable types, followed by a particularly problematic case. The formulaic 7 must say' type. Firstly, there is a common use "in a rather weaker sense with a limited set of verbs all related to the act of conversation" (Palmer 1990:73-74), such as say (most common), admit and warn; out of the 125 tokens of non-epistemic must with all persons in the CEC (cf data in (4.19)), 27 of the 62 first person cases are clearly of this kind. The following is typical: (4.23)

I must admit that the book-club offered to buy us a special pre-Christmas gift (CEC:686/l-3)

There is no sense of 'obligation' here, and it could be considered a fossilized or formulaic use; the use of have to here would also be possible (cf 4.4.1.2 and (4.41)), but that of have

105

got to would not - it would suggest an obligation with some special urgency or pressure. To express a normal obligational sense, a different locution such as I'm obliged to or I'm forced to would have to be used. This use of must has attracted some attention. Fraser (1975) deals with it as a type of 'hedged performative', and explains the use in terms of the interaction of pragmatic principles with various classes of performative verbs, while Hunston (1993) demonstrates the various discourse functions of such expressions as I must admit and I must say on the basis of extensive spoken data. The use of must in the following, though not with one of the verbs commonly used in this sense and less obviously idiomatic, seems to be parallel: (4.24)

there's a very good book about all this part of Germany which has come out fairly recently actually - I must lend it to you, it might be interesting, because it's certainly a period I knew nothing about whatever (CEC:46/319-328)

Here the speaker is expressing what he wants or intends to do, and have (got) to would differ in suggesting a specific external requirement. Vague sources. Secondly, there are uses where the nature and source of the obligation axe essentially vague: (4.25)

a (offers sherry) A well, a very little, because I really must be getting home in a moment (CEC:430/829-831)

(4.26)

A anyway, we must consider seriously the Prom programme b oh yes, I have considered it seriously (CEC:651/134-139)

The first text here illustrates what might be called the I (really) must be going type, which, like the I must say type illustrated in (4.23), has a quasi-formulaic nature (it will be discussed further in 4.5.1). The reason why the speaker wants to leave is vague here: it could be, for example, his own wishes or his recognition of determining circumstances such as the weather or the late hour. This type of locution seems to function as a conventional politeness formula, allowing the speaker to imply a pressing obligation he accepts, perhaps reluctantly (cf also (4.45) and relevant discussion below). It is similarly unclear in (4.26) where the necessity comes from: it could be the speaker or it could be externally motivated, and the force of the must requirement here is similar to that of let's consider and is in no sense an 'imposition'. The use of have to and have got to in these cases would differ in suggesting, respectively, an external requirement and some motivated urgency. In both these texts, must malees a weak statement of obligation. The following texts show occurrences of must where the speaker recognizes clearly what his/her obligations or duties are, but their source is vague:

106 (4.27)

It is impossible for me to make an exclusive choice among the various activities. What seems right for me at any given moment is what I must do. (L. Bernstein, quoted in Newsweek, 5.9.1988:50)

(4.28)

This is the road I am resolved to follow. This is the path I must go. (M. Thatcher, quoted in The Sunday Times, 29.3.1981:14)

In the first text, what I must do, although equated with what seems right for me, suggests undefined artistic motivation, while in the second, the path I must go suggests a mission rather than mere ambition. In these cases the use of must seems singularly appropriate; at the same time, it is significant that the must clauses are relatives, since subordination readily removes a clause from the subjective scope of the speaker. The next text also illustrates this use of must, and allows explicit contrast with have to and have got to: (4.29)

"I must come with you. Please let me drive you. I won't bother you. I just want to look after you, we must comfort each other." "No. Just try to see. I have to make this journey alone, I've got to deal with these things by myself. I've got to think by myself . . . " (Murdoch 1990:481)

The musts in the first speaker's words simply express what the speaker feels strongly should be done, while the use of other items by the second speaker suggests some identifiable source or reason. There seems to be some contrast here between intuition (must) and an appeal to reason (have (got) to). External sources. Thirdly, there axe examples where the speaker refers to an externally imposed requirement, without supporting it, as in the following: (4.30)

I must go home now; father told me to be quick. (Kruisinga 1925:346)

(3.92)

I must go with you, but I don't want to.

(1.9)

We must also sit idly by as Serbia and Croatia slip into economic anarchy, with all the dangers that this implies, (quoted above in 4.3.1)

In (4.30) the speaker's own viewpoint is not expressed, but there is no reason to assume he supports the requirement, while in (3.92) the speaker is explicitly opposed to it. Cases like (3.92), which is invented, seem rare, and are not always accepted as possible; cf (4.235) with related discussion below. In (1.9) the context makes it clear that the requirement is not supported by the writer; the fact that this use of must is unproblematic may be because the text is written. It seems significant here that the problematic (3.92) is naturally assumed to represent speech, while (1.9) is a written text, and would itself be rather odd in informal speech. Here, therefore, it may be assumed that style is a factor relevant to the choice of item. Have to, and similarly have got to in informal style, would be equally possible in all thesé texts.

107 Comments. Discussion of these various uses has shown clearly that there is no necessary association between the use of must and identification of the speaker as the source. We can now consider how it is that the source is, nevertheless, sometimes associated with the speaker. In 3.5.2, it was suggested that must has a default interpretation whereby the speaker is identified as the source in cases where one is sought and there is no evidence to the contrary. To review the texts discussed above, in (4.21)-(4.22) a motivating source is naturally assumed, and in the contextual absence of one, the speaker can be identified as this. (4.23)-(4.24) illustrate an idiomatic or formulaic use, and require no further comment. (4.25) also illustrates a quasi-formulaic use; in (4.26)-(4.29) the source is vague, and in (4.27)-(4.28) it is particularly general and unspecified. Finally, (4.30) has an explicit source, but the speaker's attitude to the requirement is left open, while in (3.92) and (1.9) the speaker's negative attitude to the obligation is made apparent, either contextually or explicitly. At the same time, the possible relevance of stylistic variation in relation to (3.92), which is presumably intended to represent speech, and (1.9), a written text, was proposed, suggesting that there are contexts where must may be odd in informal/spoken style. A special case. To conclude this section, there is one particular use of must that requires special discussion, since it involves conflicting data. In the following text the speaker is explaining to his interlocutor the need to interrupt a walk they are having (the second occurrence of must is in question): (4.31)

"You must excuse me, Mr Edwards," he said, "I am an old man. I must make water." ( The Observer, 21.2.1988:21)

In the next text, the speaker has been crying, and breaks off a speech for the purpose indicated: (4.32)

" . . . Now I must bathe my eyes." (Murdoch 1990:367)

The use of must in these cases is unproblematic within the terms of the present discussion: the speakers report necessities they are aware of, which they self-evidently 'accept'. Have to and have got to would also be possible here, since a speaker can naturally feel 'under a compulsion' and with have got to can indicate urgency. However, Larkin has specifically ruled out this use of musi in connection with bodily functions (1976:393): (4.33)

Must is excluded from sentences that indicate the necessity of engaging in some function of the human body or some act that satisfies some bodily need when the body itself is seen as the compelling force.

This claim is supported by examples like: (4.34) (4.35)

I've got to vomit. I have to yawn.

in which must is said to be impossible. Further, Larkin claims that in:

108 (4.36)

I have to take a rest,

the compulsion is physical, while in: (4.37)

I must take a rest.

it lies in, for example, "social convention, propriety, a tyrannical parent", ie, some definable 'source' with which the speaker can associate. Larkin's argument here must be seen in the light of his general theory, referred to above in 1.1 (cf, especially, (1.7)), that the use of must entails the speaker identifying with, or supporting, the 'source'. Larkin relates this to (4.33) and its supporting examples with the claim that must is excluded because "presumably, we do not identify our 'selves' with our baser physical needs". I am, in any case, arguing here that the general claim, as set out in (1.7), is too strong to support, but a problem remains with the data, since the judgments on the examples are certainly not universal. For example, it is not clear that the choice of must in (4.37) has to be dictated by social convention or the like, and is thus a less direct way of referring to one's physical state: I must take a rest (like I've got to take a rest), is, at least in BE, a strong and direct form of expression, while with I have to take a rest the speaker presents a requirement that is naturally understood as externally imposed, eg, by a doctor. Further, the data conflict with the evidence of (4.31)-(4.32). It is not clear that these problems are to be attributed to general AE/BE or stylistic differences, so the matter must be left here; but another related case will come up briefly in 4.4.3.1. 4.4.1.2

First person-subjects with have to

Just as must is often said to be associated with the self-imposition of first person obligations, the use of have to is often associated with obligations arising for the speaker externally. Whereas the discussion in 4.4.1.1 suggested that the claim for must is not supportable, the claim for have to seems to be substantially justified. The following texts illustrate what the external orientation associated with have to may entail: (4.38)

(4.39)

(4.40)

I then went on to say, well, if you say this is not so, then I have to accept that it isn't so, this is just what I think at the moment, and I kept stressing, you know, I'm not offering you knowledge (CEC:89/357-365) we have to get our calendar together by June for the following March, which means we have to make up our minds what books we want by about May (CEC:266/851-856) My dear, my very dear, I have to tell you the truth. Alison has presented me with an ultimatum. (Murdoch 1990:404)

109 (4.41)

It was a classic piece of investigative reporting . . . I have to say I didn't understand what all the fuss was about. Ever since I came to Fleet Street I have been called slitty-eyed and names like Kowloon Lucy . . . (The Observer, 14.6.1992:59)

An external condition for have to in (4.38) is provided by the «/-clause, and there is an explicit contrast between what the speaker has to accept and what he thinks. The use of must or have got to here would be possible, though rather awkward, and would reduce this contrast. A further comment can be made here: this text has a timeless relation between the protasis and the apodosis (if you say ... I have to accept, as opposed to the future-oriented if you say ... I'll have to accept), and this potential distinction cannot be expressed by the other items since they lack a present-future contrast, so that, for example, I must accept would be ambiguous (or vague) between the meanings of I have to accept and III have to accept (cf 4.5.1). It is striking, in this connection, that the great majority of first person have tos in the CEC have general, or non-specific reference, whereas must is regularly specific. (4.39) also clearly refers to externally determined requirements. The use of must and have got to here would strongly suggest an individual requirement (to be done now, by next June) as opposed to the general (or, perhaps, general-including-particular) interpretation suggested by have to (cf 4.5.2). In (4.40) the speaker feels 'under an obligation'; must here would make the confession self-evident and perhaps spontaneous, while have got to would make the requirement specially emphatic. A similar contrast was evident in (4.29) between the first speaker's unconditional musts and the second speaker's reasoned have (got) tos. In (4.41), finally, we have a quasi-idiomatic use of have to, which seems to be a less common alternative to the I must say type illustrated above in (4.23). Fraser (1975:198199) claims a likely weakening of the performative force of a sentence like (4.23) if a periphrasis, such as I have to, rather than a modal, is used. This is not obviously the case in relation to (4.41), but if it true it could be related to the suggestion with have to of a source external to, and hence less personal to, the speaker. The texts discussed here clearly suggest that the use of have to implies the existence of some externally identifiable source of obligation. The focus is on the existence of the necessity, and, perhaps in consequence (since the necessity is not so closely allied to performance), the reference is primarily general - or habitual - rather than specific; a further consequence, since the focus is on something needing to be done rather than on the nature of the obligation itself, may be the fact that have to can be used for minimally weak 'obligations', as was suggested in 4.2.2 in relation to (4.12). 4.4.1.3

First person subjects with have got to

Have got to has most commonly been treated as essentially equivalent to have to, perhaps as a colloquial variant. The following text would suggest such a semantic closeness:

110 (4.42)

A

we've got to decide what the structure of faculty boards and boards

of studies are going to be in the future Β is that the actual terms of reference? (CEC:80/1318-1323) This can be compared with (4.39), where specific requirements are referred to in a rather similar context, with have to\ in these texts, the items appear interchangeable, one difference being that have to allows general reference, while have got to is strongly oriented to specific reference. Must would differ crucially here in removing the implication of a specific external requirement. In (4.29), discussed in 4.4.1.1, the second speaker's words include have to followed by two have got tos: no distinction between the items is evident in the text, although it would be natural to expect the 'stronger' item to come later, and have got to, which commonly has a sense of urgency or immediacy, easily appears stronger. Here too, must would remove the implication of an external requirement. The prosodie marking on the original of (4.42) does not suggest strong insistence on the part of A; nevertheless, we've got to is often used for very strong speaker-based exhortation, for example at public meetings, so that (4.43)

We've got to malee more provision for regular training in industry, (from a political speech)

can represent the strongest type of urging, more directly than with must of have to. Similarly, "we've got to be prepared", in official announcements maxie to the public by the British government, is singled out for stylistic approval (Ballard 1939, cited by Mittins et al 1970:34) - in marked contrast to the disapproval that have got (to) has often attracted. The following text, addressed to a friend, reports obligations felt similarly strongly, with a succession of four have got tos, followed by a have got, incidental indication that have got and have got to belong in the same style: (4.44)

" . . . But we've got to live, and we've got to live properly. I'm not going backwards. We've all got to be clothed and fed and the twins have got to be educated. It's as simple as that. Hugh's got no pension and no capital outside this house. It's up to me." (J. Trollope, The men and the girls. London: Black Swan, 1992:38)

The following illustrates a further possible distinction between these items: (4.45)

A you know, he, he rang before, and I said, well, look here Β oh! A I've got to run now (CEC:200/25-27)

Ill The extended context shows that the speaker here is reporting an attempt to terminate a troublesome phone call. Have got to always has some sense of urgency, and this is very clearly the case here; and it is worth considering the difference that must, which also often suggests urgency, would make here. Must would not suggest ail existing obligation, and would thus point to a spontaneous decision (I must run now)· by contrast, the use of have got to could suggest an existing commitment, which would obscure the speaker's real intention, which is to avoid the caller. In this sense, the use of have got to may produce a politer formulation. The contrast evident here can be considered in relation to the distinction observed by Ward (1957:2) between unexpected necessity (must) and known necessity (have (got) to), as in the following: (4.46)

. . . I must ring my wife and tell her I'm on my way. We have (got) to be at Hampstead for dinner at eight.

Must is here spontaneous, representing a newly recognized necessity, in contrast to, and partly consequent on, the existing necessity referred to by have (got) to. The following comment by Edmondson et al (1977:279) is to very much the same effect: (4.47)

must conveys the imposition of an obligation, while have (got) to states that an obligation exists, ie, must is a Performance Auxiliary 15 , while have (got) to is the equivalent Statement Auxiliary.

The contrast here is glossed as follows: must = "someone/something requires of you", and have (got) to — "someone/something has required of you". The claim here that deontic must necessarily imposes an obligation cannot be accepted; the notion that have to refers to an already existing obligation seems well founded; while the case of have (got) to is unclear, since an example like (4.43) does not necessarily refer to an existing obligation. 4.4.1.4

First person subjects with need (to)

This item has not been systematically compared with the others in the discussion so far. To take just one example, need to in (4.21) would focus on some requirement or motivation, and remove the personally related immediacy (modal need is not normally used assertively). Two CEC texts with need (to) follow:

15

Sic. "When a speaker uses a performance verb of obligation, he either imposes the obligation himself at the time of speaking, or recreates it, accepting its appropriateness, even though he himself is not the original source". 'Performance' verbs (the term is only used in relation to obligation) are contrasted with 'statement' verbs; the term is "not to be confused with the notion of a performative verb" (Edmondson et al 1977:262-263).

112

she tried to throw out 'establishment' today 16 good heavens on which I wrote a short essay that's absurd - because that's absolutely international - that's absolutely absurd b yes, I need to ask an American I think about that, whether they use it A we certainly use it in Australia (CEC:254/272-279)

(4.48)

b A b A

(4.49)

but you see the trouble is that, for example, what I need to know is if we're going to go away in August some time (CEC:652/151-154)

In each case need to points primarily to the internal logic of the discussion rather than to the necessity for the action, which the use of must or have got to would suggest; it follows that need to can be seen as making a much weaker demand. Have to would differ slightly in its effect, as it would focus on an existing requirement rather than on one arising within a context. 4.4.1.5

Conclusion on first person subjects

The evidence examined so far suggests that the use of must signals an unqualified requirement, specifically as it is perceived by human agents. Matters such as the identification of a deontic source and the attitude of the speaker towards the requirement axe inferred within contexts of use; and it was shown how default interpretation may point to the speaker as the source or supporter of the required action. From the basic meaning, together with stereotypical inferences, specific notions such as irresistibility, immediacy, unexpectedness, urgency and uniqueness easily arise (since, for example, a strong requirement naturally suggests a specific action that is to be done as soon as possible). In addition to this, particular idiomatic, or fossilized, uses were noted, where the sense of obligation/necessity is markedly weakened. By contrast, have to refers to or implies the existence of a requirement that can be perceived and described independently of the immediate situation of the speaker, and thus, it seems, readily denotes general or repeated requirements. This is, perhaps, also relatable to the fact that have/has to is specifically a present form, and can contrast with will/shall have to, while the items must and have got to, which lack distinct future forms, more naturally refer to a unique future requirement. Have got to can be very similar in meaning to have to, but is marked for emphasis or urgency of some kind, and thus can appear closer in its effect to must, and in fact seems capable of having an initiating function. Finally, need to (modal need not being involved so far) focuses on the source or motivation for the requirement rather than on its execution, and thus appears to make a relatively weak statement of necessity. 16

The reference is to a dictionary or glossary being compiled.

113 As to the possible significance of stylistic variation for the use of these items, some evidence for the association of must with formal/writ ten style and of have (got) to with informal/spoken style was shown.

4.4.2 4.4.2.1

Second person subjects Second person subjects with must

It is often claimed that with the use of second person must the speaker typically initiates, or is at least strongly associated with, the 'imposition' of an obligation on the addressee, while have (got) to is neutral in this respect but is naturally used to report or pass on instructions. This issue was discussed in 3.3.5 (cf, especially, (3.74)), and illustrative cases from the literature were quoted in (3.75)-(3.78); one pair is given here again, for convenience: (3.75)

You must be back in camp by ten.

(3.76)

You have to be back in camp by ten.

In (3.74) it was stated that must "more precisely identifies the speaker as the person who gives the orders", in the case of (3.75) typically a military officer. However, such an interpretation is clearly by no means obligatory, but is inferred pragmatically, along lines such as the following. In the assumed context, where an officer is addressing an ordinary soldier, the social relationship is that of unequals, the tone that of an order, and the situation one where orders can be given on the assumptions that they "both are realizable and will be executed. The actual 'source' of the order is unclear: it could be understood as, for example, the speaker (because I say so) or regulations that are being reported (it's in the orders), or both, or it might be left vague. By contrast, (3.76), with have to, clearly suggests an independent or external source for the order; but it would be perfectly possible, here too, for the speaker to be an officer in authority. It should also be noted that a major distinction between the interpretations of (3.75) and (3.76) is the likely habitual reference of the latter, but not of the former (cf 4.5.2): a natural response to (3.76), but not to (3.75), would be What, every night? The extent to which context determines interpretation can be illustrated by comparison of (3.75) and (3.76), in their intended interpretations, with another text involving military orders, this time from fiction: (4.50)

. . . a military police car blocked the route and a policeman directed them to take a barrel track that ran eastwards, away from the battle. Terry put up an angry protest: "That's no good to us. We're not going that way. The 11th Hussars are down in Himeimat and we've got to join them." "No, sir, you must get on the track and stay on it. Himeimat's under heavy fire . . . " (Manning 1982:238)

114 In this situation a military policeman is addressing an army officer about a route: he is passing on an order, which he did not issue (and it is immaterial whether he 'supports' it), to a superior to whom, in the military context, he could in no way give orders. According to (3.74), however, this would be an obvious case for the use of have to, as opposed to must. The substitution of have to for must in (4.50) would, in fact, focus on the existence, or validity, of the relevant obligation, and would thus be more appropriate if a justifying reason or a possible alternative were involved, as in the following extensions: (4.51)

You must get on the track and stay on it - that's the only way you'll get through. (—)

(4.52)

You have to get on the track and stay on it - that's the only way you'll get through. (+)

There is also a contrast to be noted between the uses of have got to and must in (4.50). Terry uses the former to express an urgent, existing necessity, but this is effectively overridden by the new requirement which the policeman initiates - but does not 'impose' for his interlocutors with must (cf (4.46) and related discussion in 4.4.1.3). Performatives. Since the interpretation for (3.75), quoted earlier in (3.74), could be considered performative, I shall first return to consider the question of performative uses before going on to other data on the use of must. This general issue was discussed in 3.3.2.1, and it was there concluded that there were no descriptive advantages in singling out subsets of uses of certain modals as having such a status. Nevertheless, it is worth investigating the relevant uses of must, since this item alone is allowed such uses in two of the general accounts of the modals - Coates (1983) and Palmer (1990). Among other relevant claims in the literature, Lyons (1977:832-833) argues for the existence of two distinct interpretations of deontic must in the following sentence, corresponding to two distinct illocutionary forces: (4.53)

You must open the door.

These are: (4.54)

I hereby impose upon you the obligation to open the door, (directive, performative)

(4.55)

I hereby assert that you are obliged (by some unspecified authority) to open the door. (statement)

The crucial question here is how far performative interpretations such as that in (4.54) really do constitute a distinct set, thus contributing to the claimed ambiguity. In a strict performative interpretation you must open the door should mean the same as the imperative open the door. But this is hardly the case: for one thing, the imperative, unlike the must sentence, demands immediate execution, as is shown by possible addressee responses. Thus, why? and when? are reasonable responses to a must sentence, but not to an imperative, where they could be impolite or even aggressive. The use of must, in contrast to that of the imperative, adds some modification to the obligation, for example

115 by distancing the speaker from immediate and sole responsibility. Furthermore, with a strict dichotomy of interpretations, it should be possible to respond to (4.53) in ways like Are you telling me to or merely passing it on?, but this seems very unnatural. If we consider various possible circumstances under which (4.53) might be used, the following seem plausible extensions, and typical cases where must might be preferred to have (got) to: (4.56)

You must open the door - my hands are wet.

(4.57)

You must open the door - it says so in the instructions.

(4.58)

You must open the door - it was your idea to come down here.

(I ignore likely prosodie distinctions between the three sentences in speech.) In such cases, the speaker justifies the obligation, but it is not obvious that he either imposes or simply reports it: in each case, the addressee is given reasons for the requirement, and it seems irrelevant to ask who imposes or originates it. This suggests that the proposed dichotomy in interpretation may apply to two, perhaps extreme, situations, but not obviously relate to many typical uses; and the alternative to positing a distinct performative category, opposed to a descriptive/reporting use, is to assume an essentially vague meaning for must, and deal with all other aspects of its use pragmatically. For further illustration of this question, I shall now turn to the text that is the closest in the whole CEC to a true performative, and is cited by Palmer (1990:73) as a case where the speaker "clearly takes responsibility for the imposing of the necessity": (4.59)

I told Peter, I've been telling Peter, as I've been telling several people, you know, you must get into permanent jobs, and I've been urging Peter, well, Peter to go back into school teaching or something (CEC:794/672-678)

It is, in fact, doubtful whether even in this text must is really used to impose an obligation. The speaker's role here is suggested contextually: I've been telling Peter and I've been urging Peter, and the fact that the speaker and Peter are friends, point to something like a supportive, encouraging role. Further, the advice is clearly not immediately realizable, but, rather, ein indication of a desirable direction of aim. The source of the injunction, in the absence of other indications, would be inferred as the speaker, presumably drawing on relevant information such as employment forecasts or his own concern for his interlocutor. The use of have to here would make the reference external, in that it would focus on some outside motivation (such as explicit consideration of the addressee's future), while have got to would malte the speaker's words less personally concerned, though not less urgent. It may appear surprising that performative must is so little in evidence in a spoken corpus, but it can be argued that such uses are typical of unequal power situations, where the speaker is in a recognized position to issue commands in the expectation that they will be carried out (cf comments on (3.75) earlier in this section), and that these simply do not arise in normal interaction. Two further corpus-based works are highly revealing in this connection. Firstly, Coates (1983:38) states that true performative uses are naturally rare in a world where "people are either seen as equals or are treated as such, since to do otherwise would be impolite, if not counterproductive". The two relevant texts Coates

116

cites where the speaker is claimed to have authority over the subject (one criterion on her 'strong/subjective/performative'-'weak/objective' cline) appear to relate to (a) private piano tuition and (b) a Magistrate's Court, both of which involve institutionalized power structures; significantly, too, both are written and do not claim to report direct speech. Secondly, Sinclair (1990), in a closely corpus-related pedagogical grammar, does not treat deontic must as an exponent of the function 'giving instructions' at all ( will and, formally, shall are the strongest items given for this), but merely documents its use for 'making suggestions' and 'indicating importance' (228,232,236). It should finally be mentioned here that there is evidence that have got to can come closest among these items to performative uses (cf 4.4.2.3). Various kinds of 'obligation'. So far, the examples discussed have involved some degree of speaker initiation or involvement. We shall now see a series of corpus examples of second person must that show a gradual diminution or alteration of the speaker's role: (4.60)

I can only say that you must take, you must talee Joe Power's advice on this (CEC:386/643)

(4.61)

a you've got the whole flat to yourself now, have you C hm, oh you must come round and see it (CEC:553/450-452) and then, you, you must do that, I suppose, with standardization (CEC:55/1174-1175)17 Β I can't do with an aura of failure, I have enough failure coming from my own depths, you know A yes, you must consider that, that's fair enough (CEC:201-202/113-120)

(4.62) (4.63)

(4.64)

. . . well, if you apply that criterion, then surely you must start to rearrange your estimates of Lawrence's novels, surely (CEC:778-779/1260-1262)

(4.65)

A

well, at the moment I am preparing myself for it to the extent of trying to take Ordinary Level Latin Β ah yes, yes A which a you haven't got A and that in itself Β you must do (CEC:766/637-646)

In (4.60), the speaker explicitly supports another source, a function sometimes proposed as one of the uses of must: here, however, the speaker himself is clearly giving advice rather than an instruction (I can only say). (4.61) illustrates a quasi-idiomatic use for invitations, and, since the speaker's will is involved, treated by Palmer as an extension of 17

An extended version of this text appeared earlier as (3.71).

117 deontic must reflecting social convention (1990:73). In (4.62) and (4.63), I suppose and that's fair enough, respectively, are incompatible with anything like the imposition of an obligation, and indicate the speaker's sympathy or identification with the target, rather than with the source; in (4.62) the source can perhaps be located in the consequences of standardization, while in (4.63) it is circumstantial, related to the interlocutor's sense of failure. (4.64) and (4.65), finally, show uses of must where the speaker recognizes a necessity from the addressee's perspective: in (4.64) the speaker draws the conclusion surely you must start . . . from the addressee's previous comments (if you apply that criterion), while in (4.65), with you must do [sc take Latin], the speaker expresses identification with the addressee in his situation - having to qualify himself in Latin. It should be mentioned that all these examples are markedly typical of BE as opposed to AE. In all these cases, the use of have to and have got to would primarily tend to reduce the speaker's personal involvement or concern; however, for the quasi-idiomatic use in (4.61), see a similar use with have to in (4.72) below. The condition in (4.64) (if you apply ... surely you must start... ) can be compared with the condition in (4.38) (if you say ... I have to accept . . . ), with related discussion above. Need to would be possible in all these cases with, in general, the effect both of suggesting desirability rather than necessity and of pointing to some motivation. The use of must to express the addressee's perspective can presumably be related to Bouma's discussion of the following pair (1975:324-325): (4.66)

You must love your enemies.

(4.67)

You have to love your enemies.

Bouma claims that the use of must in (4.66) shows the speaker observing the subject as a morally committed agent, while have to in (4.67) refers neutrally to the obligation itself (cf (3.94) and (3.95) in 3.3.5). 4.4.2.2

Second person subjects with have to

I shall start by returning briefly to the the have to member of the original contrasting pair (3.75)-(3.76): (3.76)

You have to be back in camp by ten.

The related claim here is that here, in contrast to must, "'have (got) to expresses obligation or requirement generally, without specifying the person exercising power or influence" (Leech 1987:79). It was suggested in (4.14) that have to basically focuses on an existing obligation; in the required interpretation of the present sentence, the addressee infers that such an obligation has an existence, and perhaps source, that is independent of the speaker: thus, (3.76) could naturally be interpreted in its specific context of military life as denoting some existing order. This does not, however, rule out the speaker himself as source: he could use have to to indicate the existence of an authority, or to imply some

118

objective validity for an arbitrary command. The latter use seems to be manipulated by children: You have (got) to come home now can be used to lend non-existent higher authority to an order. Illustrations of the use of have to follow: (4.68)

you have to find a good man and build on him, rather than pick a place on the map (CEC:68/660-662)

(4.69)

Β

(4.70)

what he doesn't realize is that not everybody else can work quite as hard as he can, and you just have to say A yea Β well, I'm, I'm singing tonight, or something A yes Β if he says you've got a seminar (CEC:130-131/207-215) how many hours do you have to do? (CEC:238/1015-1017)

All these texts suggest some circumstantial necessity independent of the speaker: for example, in (4.70) an employer, or terms of employment, are ultimately involved. In each of these cases the obligation is more naturally understood as generalized - concerned with what is in general or in principle appropriate as opposed to what is simply required in a particular situation, even though the reference may be specific. For example, in (4.70) the text as it stands could equally well refer to a specific or a general situation; in fact, the broader context makes it clear that a general situation - terms of regular employment - is involved. The use of must in these three texts would both remove the focus on an existing, external obligation, and naturally make the reference specific or unique, while have got to would, as usual, add urgency. No performative-like uses of have to occur in the CEC, or are cited in the major accounts. However, Bybee et al (1991:23) state that this item can equally well "report a fact . . . or issue a command", the latter illustrated in: (4.71)

Todd, you have to go home now.

This derives from an AE source, but even so, if it represents standard usage, it contradicts a very large number of accounts of the use of these items. It seems perfectly reasonable to argue that (4.71) can be used either as a directive - to give an order - or to report a requirement, as in Todd, Jane says you have to go home now. A final text follows: (4.72)

"You have to admit," said Jack, "that he was the most remarkable person we ever met." (Murdoch 1990:4)

This quasi-idiomatic use of have to is comparable to that of must in (4.61), the sense being Ί want/expect you to admit . . . '. Have got to does not seem to be used in this way, presumably because urgency would not normally be appropriate in such a context.

119 4.4.2.3

Second person subjects with have got to

In the discussion of second person have to it was stated that the replacement of have to by have got to would tend to suggest some special urgency and, possibly, specific as opposed to general reference. This will be illustrated in relation to the following text: (4.73)

and Hart you've got to stand up to, haven't you? (CEC:129/117-8)

The similarity between the items have to and have got to is underlined by the fact that (4.73) occurs in the corpus in close proximity to (4.69), which has have to (the he in the latter is 'Hart' in the former, who in both cases has to be resisted). To consider any possible distinction here it will help to take the incidence of stress and the nature of qualification by just into account. The relevant parts of (4.69) and (4.73), respectively, are: (4.74)

. . . and you 'just have to say . . .

(4.75)

. . . and 'Hart you've 'got to 'stand 'up to . . .

In (4.74) just is stressed, but not have to, and the sense of just is 'only, no more than' (LDOCE 1987:570, meaning 2) - that is, the meaning is 'it's simple: the only thing you have to do is . . . '. In (4.75), by contrast, have got to is stressed, and the message is being emphasized; if (4.75) had qualification by just, as in . . . you've just 'got to ... , the sense of that item would be 'in a way that offers no other choice or possibility' (ibid, meaning 5). 18 The use of must in these cases would be possible, and simply tend to suggest the speaker's sympathy, as was seen in texts (4.62) and (4.63). The difference between must and have (got) to can be illustrated by normality testing with variations on (4.73): (4.76)

You

^

I to stand up to Hart to protect your sanity, haven't you?

(4.77)

You must stand up to Hart to protect your sanity, mustn't you?

(—)

(4.78)

You have to stand up to Hart, I can see that, that's fair enough.

(—)

(4.79) v '

You < f 0 1 1 ^ , have got to

> stand up to Hart, I can see that, that's fair enough.

(+)

Rave (got) to is more appropriate in the first pair, where an external reason is adduced, while in the second pair, where the speaker shows his sympathy with the addressee, both must and have got to seem more appropriate. Another instance of second person have got to follows: (4.80)

18

yes, you've got to be very careful about withdrawing licences (CEC:229/490-492)

The contrast being made is based on (4.69) and (4.73) as they stand. An alternative stress pattern for (4.69) would be . . . and you just 'have to say, and then the sense would be very much as in (4.75); ... you've 'just got to say, with the sense as for (4.74), also seems possible, but less likely.

120 Here the use of have to would make little difference (as the reference is clearly general), but must would be rather odd, presumably because a demonstration of speaker sympathy for the addressee, which would be a natural non-epistemic interpretation, is not expected in this context. I shall now turn to the possible use of have got to for performative-like uses. This option for have to was discussed briefly above, but there is no corpus evidence to support it. Have got to, which is most commonly simply treated as a colloquial alternative to have to, is allowed by Coates (1983:Ch.4) considerable semantic overlap with must, though without the extreme subjectivity that performative status involves. However, in one text presented in 4.2.1, (4.7), which featured all three items, it was noted that have got to was the item whose sense was closest to giving an order. Another such case is reported by Collins (1991:157-158): (4.81)

Well this little old lady, they'd tuck her into bed you know and they'd have her all nice and comfortable you know and 'You stop there' the nurse'd say. 'You gotta stop there, love, don't you get out because you worry the other patients.'

Collins claims that this is performative, and typical of Australian English. Clearly, if anything is to count as performative, this has a strong claim, with its proximity to imperatives (you stop there ... you gotta stop there ... don't you get out). Collins claims that Coates' analysis, which does not allow performative have got to, reflects the rather older BE data used 19 ; certainly, this text (but with got to rather than gotta) seems perfectly natural in BE in a colloquial context. 4.4.2.4

Second person subjects with need (to)

The following text will suffice to illustrate need (to); it also appears to be the only genuine assertive example in the corpus, as opposed to examples like all you need to do ... , which strongly implies you don't need to do much: (4.82)

A Β A c A

no, wait a minute yeah no, I mean, you'd get to well, you need to get to Pu 2 0 , Hammersmith then you met Ham, you you hit Hammersmith, and you've got to get on from there, that's all right (CEC:294/894-904)

This is in the middle of a discussion of recommended road routes, and c's use of need to simply states where a driver would be aiming to get to. Have to (or, perhaps, would have to) would be possible here, with more immediacy of purpose; crucially, the effect of A's following you've got to get on ... is markedly stronger than c's need to. 19 20

Coates' data include the CEC material analysed for the present study. The word cut off and replaced must be 'Putney'.

121 4.4.2.5

Conclusion on second person subjects

The following few remarks can be added to the comments made in 4.4.1.5. With must, the standard commitment of the speaker to his utterances can result in the speaker expressing sympathy with the obligated addressee, or adopting his viewpoint, further evidence that the use of must has no specific relation to the identification of the deontic source - in cases where a source can be identified. While have to is often interpreted as having general reference, have got to contrasts in, like must, normally having specific reference. It was seen that have got to, always a relatively emphatic item, can convey the strongest kind of statement of obligation, and thus have a claim to performative status.

4.4.3 4.4.3.1

Third person subjects Third person subjects with must

As with first and second person subjects, the must-have (got) to contrast with third persons has often been held to rest in a distinction between speaker and external .orientation/origination. A typical example of this is provided by Larkin (1976) in the following pair, cited in 1, and repeated here for convenience: (1.4)

My girl must be home by ten.

(1.5)

My girl has to be home by ten.

Of these, Larkin states (1976:392) that (1.6)

(1.4) implies that the speaker "goes along with" the prohibition that the sentence states, while (1.5) is neutral in this regard.

This claim is parallel to similar claims seen above for second person subjects, but it receives little support in relation to the use of must from data to be examined below, and in any case the question is complicated by stylistic factors. Various sources. The following examples display a wide variety of sources: (4.83)

we, we make it perfectly clear that papers must be in on the first of May (CEC:36/134)

(4.84)

A oh have you? I háven't even seen it really, to speak of b where it's got to A and it must be, that must be one thing we do (CEC:651-652/139-143)

122

(4.85)

Β you could study literature in a foreign language, couldn't you? A yes, I could Β you feel it, it must be English A yes (CEC:776/1107-1112)

(4.86)

ah, Vienna is a Congress stadt 21 , and it must become, a, known as a place friendly to the handicapped (CEC :738/326-330)

In (4.83) the requirement clearly either originates with or is associated with the subject we, and is typical of official announcements; have (got) to and need would be possible with no difference in effect with respect to the source, in view of the explicit we make it perfectly clear, but would appear less official - a stylistic factor. (4.84) follows directly on a text quoted earlier as (4.26); here, also, it is also unclear where the source is (although in principle it could be associated with the speaker). By contrast, in (4.85), cited earlier in (3.72), the source clearly has no connection with the speaker: B, in using must, is essentially eliciting A's wish or requirement. In (4.86), the source is vaguer still, as the necessity arises out of the status Vienna enjoys, and this use appears to be close to a merger, since the sense is partly epistemic (what is sure to happen). In (4.84) and (4.85), the use of both have to and have got to would be possible, the only obvious difference being that have got to would, as usual, add urgency. In (4.86), on the other hand, the other items would suggest a specific, distinct reason outside the general situation referred to. Distinction according to style. The following text requires particular attention: (4.87)

Passengers must cross the railway line by the footbridge, (public notice)

This is typical of official notices or announcements (written or broadcast), and could be seen as representing the will of the initiator of the text; in this case, have (got) to could not be used. However, the information such a notice contains would be referred to in informal speech (whether by a railway official or by a private individual) in ways such as the following: / (4.88) v '

Y°u have to I . .. ,, , , , > use the bridge to get across the line. You ve got to J

(Have to would be more appropriate here for a general comment, have got to for a specific occasion.) Must would not be used here in informal style. (The fact that you mustn't walk across the line would also be used in informal speech is quite distinct, cf 4.5.4.) Here we very clearly encounter the question discussed in 4.3.1, and already encountered a couple of times above - how the parameter of formality, more specifically that of written vs spoken language, may be relevant to the use of must and have (got) to. The following texts, all from written sources, provide further illustration: 21

Sic.

123

(4.89)

" . . . still, it will do to learn verbs and all those tedious things that must be learnt." (B. Pym, A glass of blessings. Harmondsworth: Penguin, 1980:65)

(4.90)

The ceiling is so dark that 30,000 watts of electric light must be used for it to be photographed. ( The Economist, 26.12.1987:117)

(4.91)

He still must spend most of his time opening new factories. {Newsweek, 7.3.1988:13) Euro Disney must attract 11 million visitors a year to the Magic Kingdom to break even. {The Observer, 1.3.1992:32) Since the fall of communism, two Jewish schools have opened in Budapest and are attracting students so quickly that they must turn away applicants for lack of classroom space. {The Guardian Weekly, 1.3.1992:19)

(4.92)

(4.93)

In none of these texts is the writer's will being expressed: the comments are presented neutrally, and in all cases the source of the necessity referred to is left unspecified. This use of must seems typical of public written style, and of journalism in particular, and no distinction between BE and AE is evident. There is little doubt that in informal speech if, for example, a speaker was commenting on the contents of a newspaper report - have to or have got to would be used in all these cases. Thus, in relation to (4.91) an informal comment like The poor fellow still has to spend an awful lot of time opening factories (the reference is to Prince Charles) would be expected, and the use of must would be extremely odd, or rather would naturally be interpreted as epistemic, which would be inappropriate here. Unqualified requirements. The next examples, also from writing (although that does not appear criterial here), illustrate the use of must for maximally general, unqualified requirements of the type seen earlier in (4.27)-(4.28) with first person subjects: (4.94)

It was as though the conceptual space in which Marcus's mind moved did not anywhere connect with that of Ludens - and yet it must connect, there must be a way through. (Murdoch 1990:14)

(4.95)

" . . . What is sought is not one thing among others, but the foundation of things. As I said, something necessary, something which must be so . . . " (Murdoch 1990:163)

The emphasis on the absolute quality of the requirements is, typically, indicated by printed italics in these cases, and while have to or, less suitably, had got to could be used here, must seems singularly appropriate. This sort of use of must is sometimes treated as 'objective', and for Coates (1983) would presumably typify the objective periphery of meaning, as opposed to the subjective core (cf 3.5.1). At the same time, these examples might also be seen as cases where there is a merging with epistemic meaning.

124 Bodily needs. The final point to be discussed here concerns the use of must for bodily needs. In a case parallel to that with first person uses, which were discussed in 4 . 4 . 1 . 1 , Larkin (1976) argues that of the pair (4.96)

Adam must go to the john.

(4.97)

Adam has to go to the john.

only the second is available for the direct expression of a bodily need, while the first would only be used if the 'source of the compulsion' were, for example, social convention. As with the earlier examples, there is no evident justification for the claimed restriction, since (4.97) can easily be used to make, or report, an instruction to, or for, a child, and this has nothing directly to do with bodily needs. In any case, must seems to be readily available for the expression of bodily needs, as in the following: (4.98)

Charles felt that he must certainly vomit within the next ninety seconds. ( J . Wain, Hurry on down. Harmondsworth: Penguin, 1960:105; cited by Nehls 1986:166)

This shows the choice of must, and it makes no difference whether it is seen as a third person use or as a reported form of a first person use (in the latter case it would parallel (4.31)). As in the earlier discussion in 4 . 4 . 1 . 1 , the question here does not seem capable of satisfactory resolution. Comments. The discussion in this section has further illustrated the claim that must is used for the expression of unqualified obligation or necessity, but it has also introduced further strong evidence of a stylistic factor: must is specifically associated with written style and used in cases where the corresponding spoken expression would use have (got) to. 4.4.3.2

Third person subjects with have to

A small number of examples will suffice for the other items. First, have to: (4.99)

some of our people who are doing LEs have to consider which paper to do (CEC:34/39-41)

(4.100)

well, we all understand the circumstances, why it has to be surreptitious, don't we? (CEC:506/697-698)

These two examples show typical uses of have to in speech: the first case shows an external obligation resting on human agents, and in the second has to be is explicitly related to determining circumstances. An explicit contrast between the use of first person must and that of third person have'to for obligations the speaker himself feels and those he identifies for another agent, respectively, is illustrated in the following text, which was discussed in 4 . 4 . 1 . 1 :

125 (4.22)

. . . she thought: I must have patience. I must wait for her to speak first. She has to be the one to choose the time and place. (James 1986:320)

The next text, by contrast, suggests that the use of must and have to can be very close: (4.101)

Disney's own research shows that, of the world's top 10 attractions in terms of attendance (see table), Euro Disney has to be the world's third most popular attraction from the word 'go' if it is to malee money. (The Observer, 1.3.1992:32)

This comes from the same article as (4.92) did, and the use of has to here seems to be parallel to that of must in the earlier case: (4.92) has the structure 'x must y to do z\ and (4.101) lx has to y if z\ that is, both state a necessity required in order to achieve a further aim and are in no evident way determined by the speaker or another definable agent. It was claimed above that the use of must in (4.92) is typical of writing as opposed to speech, so the occurrence of have to in (4.101) requires comment. One evident distinction between these texts concerns the main verbs used: in (4.92) the verb is dynamic (attract), and in (4.101) stative (òe); and dynamic and stative uses of must are strongly associated with, respectively, deontic and epistemic readings (cf data in Coates 1983:Ch.4). This could mean that have to in (4.101) more naturally carries the required deontic meaning than must would. This would also means that have to has a maxked status in relation to must in such a sentence. 4.4.3.3

Third person subjects with have got to

Such stylistic distinction as there may be between must and have to will clearly apply differently to have got to in that this item is markedly typical of informal, spoken language. The following examples are all from the corpus: (4.102)

(4.103) (4.104)

I think she thought I was a career woman who would be only too glad or would say "oh well, he's got to go into a hospital", you know, so she made the decision for me (CEC:218/1048-1052) if you're going to confer a picture into your room it's got to be well painted (CEC:201/110-112) I'm in no way inclined to say that this blanket scheme must go through, or anything of this - a lot of work has got to be done on it (CEC:839/676-680)

The first text shows a quasi-performative use (a decision is involved), comparable to the cases of have got to seen above with second person subjects ((4.7) and (4.81)). (4.103) also reports a strong requirement, which can be seen as externally oriented ( i f . . . it's got to . . . ). In (4.104) both must and have got to are used. It might be possible to claim a clear distinction here between the use of must for a speaker-based requirement, and

126

that of have got to for descriptive comment. At the same time, however, must denotes a more general requirement, while have got to refers to more specific work needed; and it is significant that while must here could be replaced by have got to, the reverse is hardly the case. This suggests that the urgency conveyed by have got to is more naturally related to specific detail than is the case with must. 4.4.3.4

Third person subjects with need (to)

Two corpus texts follow: (4.105)

Marina needs to use a ladder in order to get on to the sun-roof in their house (CEC:281/256-258)

(4.106)

is there a view that the 18th century needs to be strengthened in the teaching? (CEC:519/86-87)

These two examples illustrate slightly different uses of need to. The first case refers to a contingent requirement of general reference (χ is needed in order to achieve y), and could be alternatively expressed by have to, while the second is concerned with a desirable change, and the other items would not be appropriate here (indeed, should would be more suitable here). 4.4.3.5

Conclusion on third person subjects

The section on third person subjects shows further evidence of distinctions in use according to medium: must is used in contexts in written language where have (got) to would be normal in speech. There is also evidence that, in colloquial use, must and have got to may differ in expressing generality vs urgent specificity.

4.4.4

Conclusion

The findings so far will now be summarized, and used as the basis for the further investigation of these items in 4.5-4.6. The framework set out in (4.14) has been followed for the description, and specific, or typical, meanings that may appear to attach to individual items have been attributed to pragmatic inferences. Particular findings include the following. There is no necessary connection between the use of these items and the identification of the source or nature of a requirement, or of the speaker's authority or attitudes, and there is no single, simple distinguishing feature, such as 'subjectivity' vs 'objectivity', always to be observed between them. Further, they may contrast in different contexts in rather different ways and to different degrees, and on occasion seem more or less interchangeable; and while have to and have got to are often close in meaning, there

127

are contexts in which must and have got to may be closer to one another and in contrast with have to. The notion of default interpretation is crucial here and can be related to the operation of stereotypical implications, for example to the effect that with the use of must the speaker sympathizes with the requirement or initiates it, or that with have to the requirement exists independently of the speaker. Certain uses of these items have been treated here as 'idiomatic': in such cases it may be assumed that a particular stereotypical use has fossilized, and acquired a sense not so readily relatable to the normal processes of pragmatic interpretation. Must, as has been emphasized, is used to present a maximally unqualified requirement. The speaker's normal commitment to the validity of his utterances may suggest that must expresses the speaker's requirements, but this is not necessarily the case and there is no indication that it is even the norm; in fact, the opposite can be the case. Typical uses of must may, variously, suggest that the requirement is irresistible, immediate, urgent, moral or unique, for example, but must can also indicate the speaker's understanding, whether or not sympathetic, of an externally defined situation. Have to focuses on some independently existing requirement, and typical interpretations may suggest the lack of speaker concern, external pressure, or - related to the explicitly present tense form of this item - general, and hence habitual, reference. Have got to is often very similar in use, but is marked by urgency, and thus, in relation to the other items, naturally suggests specificity of reference or emphasis. Need to, finally, expresses a desirable requirement and focuses on the source of or motivation for this, and thus often appears simply to be rather weaker than the other items. The fact that individual items may themselves have stereotypical uses is perhaps reflected in the fairly strong association of must with deontic modality when with dynamic uses of verbs and with epistemic modality when with stative uses. Questions of variety and style constitute further complicating factors, as has been seen at various points. Thus, have got to is only used in informal language: all the real examples above were from either real conversation or simulated speech in fiction. It seems reasonable to assume that the meaning expressed by have got to - the suggestion of an existing obligation, marked by some urgency - is itself typical of informal speech as opposed to more formal, especially written, language. Further, there is the more complex situation with must and have to: must in relatively formal style seems to correspond to a certain extent to have to in informal style. Since, however, have to still appears in formal writing, one hypothesis would be that it is used for marked or emphatic functions there.

4.5

Other aspects of the deontic uses of must, have (got) to and need (to)

4.4 dealt with present tense affirmative uses of the items must, have to, have got to and need (to) in deontic meanings. In this section other aspects of the deontic uses of these items will be investigated. The following themes will be taken in turn: first, vaxious types of explicit temporal reference - to future time (4.5.1), habitual or general activity (4.5.2),

128 and past time (4.5.3); then finite non-affirmative forms (4.5.4); and, finally, non-finite forms (4.5.5). This will cover the major aspects of the deontic uses of the items. The comments will be related as far as possible to the general findings summarized above in 4.4.4.

4.5.1

Future reference

In this section the potential of the items under discussion for future reference will be examined. This primarily concerns the functions of the future forms will/shall have to22 in relation to the present forms must, have to and have got to.23 Will need to is also available but rare, and will not be discussed here; but the claims that will be made here for the distinction between the 'present' forms and will have to apply equally to need (to) in relation to will need to. As to possible suppletive functions, it is often assumed that will have to, like had to, has such a function in relation to must. The following text suggests that will have to can also serve as a suppletive form for have got to: (4.107)

cos I've got, I think I shall probably have to come up to town (CEC:270/1041)

Here I've got . . . (sc to come up to town) is corrected to I shall ... have to ... Palmer (1990:121) observes that deontic must "seems more often to refer to future than present events" ; however, in a strict sense, deontic modality is always concerned with events subsequent to the time of speaking (apart from reported habitual requirements). The explicit future form will have to, Palmer states (ibid), "will usually occur if there is any suggestion at all that the necessity is future or conditionally future". The real question here, then, is what the distinction is between such cases and normal uses of future-oriented deontic must, etc The fact that there may be no categorical distinction here is illustrated by two corpus cases cited by Palmer (1990:122), where the immediate contexts are clearly parallel (the speaker refers in each case to an obligation to be fulfilled at a specific time in the future): (4.108)

Yes, I've got to be at London airport at fourish.

(4.109)

Yes, I'll have to be back at the airport at six.

For further illustration, two corpus texts with will have to follow: (4.110)

22 23

. . . we're also going to get rid of those, the questions 3 and 4, that came out, it'll, it'll have to go through the council of course (CEC:51/1001-1003)

Reference to future will forms is to be understood to subsume shall forms when relevant. Have got to is very restricted in its non-present and non-finite uses, and there is no evidence of the use of a future form.

129 (4.111)

well, I think the finalists will have to stop, but the others will carry on (CEC: 118/730-732)

In both these a future environment is established contextually by the other tense forms we're ... going to get rid of those and the others will carry on, and alternative present tense forms for the obligations - it must/has (got) to ... for it'll have to ... in (4.110) and the finalists must/have (got) to ... for the finalists will have to ... in (4.111) - would have clearly different meanings: present tense forms make the obligations more real at the time of speaking and also make the temporal relation between the modalized verb and the other future reference verb obscure. If no such difference is evident between (4.108), with I've got to, and (4.109), with I'll have to, this may be because the requirements belong to the near future (at fourish, at six), so that there is little room for a distinction between a 'present-based' and a 'future-based' requirement. This potential distinction will now be explored further. First, with adverbials of 'near' future reference, such as now (in the sense of 'in the immediate future'), tomorrow or next Tuesday, all four forms in question seem possible without any significant distinction of meaning: must have to 've got to '11 have to

(4.112)

now.

(

tomorrow, next Tuesday. However, when the reference becomes 'distant' rather than 'near', clear differences emerge. To illustrate this, let us consider the case of someone facing the need to buy a new car. In situation (a) he is looking at his papers and considering the expiry date of his car's current roadworthiness certificate: here, he could equally well say any of the following: must

1

!

(4.114)

have to > buy a new car by the end of May. got I'll 've have to to buyJ a new car by the end of May.

In situation (b), on the other hand, he is examining signs of age on his car; here, (4.116) would be much more natural than the three sentences in (4.115): must "j

{

(4.116)

have to > buy a new car pretty soon. 've got to J I'll have to buy a new CM pretty soon.

(—) (+)

The decisive difference between these two situations rests in the temporal location of the recognition or acceptance of the obligation (as opposed to the actual time of its anticipated fulfilment). In situation (a), the obligation can be seen either as existing at the moment of speaking, in that the expiry date and its consequences are fixed, as in (4.113), or as arising at a specific time in the future, as in (4.114); however, in (b) the obligation, although already anticipated, is not yet specific, as it would be in (4.115), and is more appropriately seen as arising at some as yet undetermined time in the future, as in (4.116).

130 The possibility of using three time points in dealing with modal expressions was outlined in 3.5.4: this involves the time of speaking (5), the time of fulfilment of - in the present case - the obligation (E), and the time when the modality becomes relevant or is accepted as valid or binding (M). With all the items must, have to and have got to, M can normally be assumed to be close to 5, while E by definition follows Μ, so that a typical use for a single action could be represented S,M - E.24 However, in the case of will have to, which explicitly contrasts with have to, M is distinct from S, and the typical temporal framework here is S - M E. This claim can be illustrated by the following invented pair: (4.117) (4.118)

I have to paint the bathroom, but I'll avoid facing up to it as long as possible. (—) I'll have to paint the bathroom, but I'll avoid facing up to it as long as possible. (+)

(4.117) is a little odd because I have to suggests that the speaker's acceptance of the obligation ( M ) is closely associated with S (ie, S,M - E), and this conflicts with the speaker's second statement I'll avoid ... , which suggests that the related obligation would have the structure S - M - E. By contrast, (4.118) sets the acceptance of the obligation in the future (ie, S - M - E) and thus removes the incongruity in question. The way in which have to and will have to can contrast with one another in a single text is shown in the following: (4.119)

A do we have to get nails, and stick them in our wall, or Β I think we'll have to - it's a bit big, you see (CEC:113/455-459)

It is natural here to interpret j4's question as immediate in reference (ie, S,M - E), but also potentially general (concerning what one normally does - in this case when hanging a picture), and ß ' s answer as relating more specifically to future requirements that can be foreseen (ie, S - Μ - E). It would be very odd for the forms to appear in reverse order here, with the question shall we have to ... ? followed by the response I think we have to. The functioning of these items in regard to future reference reflects the observations made about them in 4.4.4. Must makes a maximally unqualified statement of necessity, and this entails that unless otherwise indicated an obligation is presented as requiring prompt fulfilment; in other words, S, M and E are assumed to be close together. Have to presents an existing obligation externally; S and M will also naturally be temporally close, but E can easily be separated in time - hence the notion that have to, but not must, is resistible, and hence also the fact that have to readily refers to repeated or habitual action (cf 4.5.2). The future form of have to can then separate S and M. Have got to will resemble must in having S, M and E in close propinquity. Two further aspects of future time reference will be mentioned briefly here. Firstly, other complex future forms axe available apart from will have to, such as be going to have to, eg: 24

With habitual or repeated actions E may be considered to extend in either direction from 5, but still always to follow ili. I ignore this complication.

131 (4.120)

because they're going to have to do it in their jobs, anyway, aren't they, whether they're going for being economists or teachers or geographers, or whatever they do - they're going to have to know about how to get their stuff across (CEC:262/658-665)

Such forms require no special comment here. Secondly, there is what was termed in 4.4.1.1 the I (really) must be going structure, a specific quasi-formulaic type of expression illustrated in text (4.25). This use has an important temporal implication; Jespersen comments (1931:200): (4.121)

while I must go says nothing about the time of going, I must be going denotes the going as immediate.

This allows the be + -ing form to be treated as a substitute missing future infinitive. This also illustrates how futurity can be expressed not in the modal or periphrastic but in the infinitive. In terms of our S - Μ - E structures, I must go has the relation I have termed S,M - E, while I must be going explicitly locates the three time points in maximally close proximity (as if S,M,E).

4.5.2

Habitual reference

The contrast in question here, which has come up at various places in the discussion so far, is that between specific or individual reference and habitual or general reference. Specific reference to an obligation is understood to refer to one performance, while habitual or general reference either relates to a series of performances or fails to specify whether a unique or repeated event is in question. I shall start here by considering the following set, from Coates (1983:54), already referred to in 1.2: (1.11)

*I must get up at 7 am every day. 25

(1.12)

I have to get up at 7 am every day.

(1.15)

*I've got to get up at 7 am every day.

Coates comments: (4.122)

have got to like must cannot be habitual in meaning, unlike have to.

This reflects a fairly common view on the availability of the three forms for habitual reference; however, it is reported that according to informant testing the judgements on the status of (1.15) and (1.11) shown are not universal (Nehls 1986:37). I shall now investigate the use of the three items for habitual meaning, with particular reference to the claim in (4.122).

25

The example with must is not actually given by Coates, but is implied in (4.122).

132

Must. The general opinion that this item excludes habitual interpretation is supported by the absence of any such occurrences in the CEC. This apparent incompatibility between the use of must and reference to a series of actions can be related to the fact that must is typically used for unqualified requirements with immediate effect, and explicit reference to a series of discrete events would markedly qualify both the nature and the immediacy of the obligation. However, the exclusion of habitual reference for must is not absolute: (4.123)

. . . each day for the rest of my life I must take two immunosuppressant drugs . . . and a steroid. (Newsweek, 12.9.1988:45)

Each day has explicit reference to a discrete series of occasions, so we must consider whether there any special circumstances involved here that might 'justify' the use of must. The text is written (though it purports to represent speech), and it would certainly be rather odd in informal BE speech; here we may compare other examples in 4.4.3.1, like (4.92), where must attract 11 million visitors a year clearly has habitual reference. Also, it might be significant that (4.123) represents AE, though the evidence presented in 4.3.1 suggests that must is less typical of AE than of BE. It may thus be assumed that the judgement on (1.11) represents normal usage in an informal, spoken BE context, but may not apply in more formal style. Have to mon:

This item requires little comment, since texts such as the following are com-

(4.124)

I've got most of the material now, but I have to refurbish my lectures the whole time (CEC:116/654-656)

(4.125)

you have to order books anything up to ten months ahead (CEC:265/844-845)

(4.126)

you have to make the article agree with an adjective, an adjective with a noun (CEC:769/805)

The time reference in these cases is either explicitly habitual (the whole time, in (4.124)), or general/unrestricted, since the requirements to order books or malee grammatical items a,gree, in (4.125) and (4.126), relate to any relevant occasion. Such uses can be related to the relative objectivity or distinctness from the here-and-now that typically characterize requirements expressed by have to. Have got to. This item, like must, is clearly restricted in its use for habitual reference, though the judgement on (1.15), here repeated, is much more questionable than that on (1.11), with must: (1.15)

*I've got to get up at 7 am every day.

The following are typical relevant uses from the CEC:

133 (4.127) (4.80)

in a small primary school, you know, you've got to be working all the time (CEC:185/621-623) yes, you've got to be very careful about withdrawing licences (CEC:229/490-492)

In (4.127), the adverbial all the time suggests habituality, while in (4.80) the reference is general rather than particular, but in neither case is there a sense of actual discrete performances. However, despite the lack of more explicit corpus cases, the use of have got to for repeated actions is recognized by some scholars. Thus, Chalker (1984:85) reports the following contrast between have got to and have to: (4.128)

British English uses do have to where the obligation is for habitual/repeated actions already taking place, but often uses have got to where there is an obligation now for a future action or series of actions.

Similarly, Quirk et al (1985:145), in referring to the pair: (4.129)

Jim's got to check the temperature every 12 hours.

(4.130)

Jim has to check the temperature every 12 hours.

comment: (4.131)

whereas (4.129) is likely to have the force of a directive, stipulating what Jim's duties will be in the future, (4.130) is more likely to indicate a habitual action ('This is what Jim's present duties consist of').

Thus, a blanket assignment of ungrammatically to (1.15) cannot be accepted. However, the common claim that, of the three items, only have to can be used habitually is supported here to the extent that with that item the repeated actions are assumed to extend before and after the present, while with have got to, as with must, the repeated actions are restricted to the future. 26 The interaction of future and habitual reference. Since future and habitual/general reference with the items under discussion are often closely related, I shall now show how typical or default interpretations of sentences with must, have to and have got to can involve both future and habitual/general reference. For this purpose we will compare the following invented examples: (4.132)

You have to hurry if you want to catch the 9.29.

(4.133) (4.134)

You've got to hurry if you want to catch the 9.29. You must hurry if you want to catch the 9.29.

(4.135)

You'll have to hurry if you want to catch the 9.29.

26

T h e comment by Matthews-Bresky (1977:8) t h a t "as a tendency, not as a rule, have got to suggests a more limited or circumscribed obligation" (ie, than have to) is also relevant here.

134 The basic difference between (a) (4.132)-(4.134) and (b) (4.135), according to 4.5.1, rests in the time when the obligation is felt to apply; this can be represented as S,M - E and S - M - E, respectively. However, a further crucial distinction arises in terms of the specific-habitual contrast: the default interpretation of (4.132) has general or habitual reference, but those of (4.133), (4.134) and (4.135) have individual reference. This claim can be further illustrated by taking a text discussed earlier: (4.68)

you have to find a good man and build on him, rather than pick a place on the map (CEC:68/660-662)

and considering the effects of possible variations and extensions to it: (4.136)

You have to find a good man whenever these jobs come up.

(4.137)

You must i find a good man whenever these jobs come up. 6 J You've got to J r

(—) \ /

(4.138)

You have to find a good man by the end of the month.

(—)

(4.139)

You must j find You've got to

v

;

a

(+)

good man by the end of the end of the month. ( + )

The original text, with have to, is naturally given a general/habitual interpretation, locating the requirement at any time that is relevant, while if have to is replaced by must or have got to, the reference naturally becomes specific; hence of the examples with have to, (4.136), with general, present-based reference, is better than (4.137), but (4.138), with specific, future reference, is not as natural as (4.139). Need to is clearly capable of either specific or general interpretations without the kinds of restrictions we have seen above: thus, to quote from texts already discussed, (4.49) ( what I need to know) is specific, (4.105) (Marina needs to use a ladder) is general in reference, while (4.106) (the 18th century needs to be strengthened) is vague in this respect, though likely to be specific.

4.5.3

Past reference

Must has no past form, being in origin one itself, but has limited availability for past time reference; this issue will be investigated first. Then I shall discuss in turn had to and had got to, with particular reference to the suppletive functions they possibly perform and to restrictions on the use of had got to, needed to, and finally various other aspects of past time reference. Must. The precise limits to the use of must for past time reference are hard to define. Coates (1983:40), on the basis of her corpus, states that it can function as a past tense form both in normal indirect speech, and to some extent in other contexts where an internal monologue can be hypothesized - ie, in free indirect speech. The use of must for

135 past time reported speech, instead of the available and presumably suppletive had (got) to, is illustrated in the following: (4.140)

I felt I really must get round the Department (CEC:827/2)

The 'original' for this would presumably be I must . . . , and be similar to (4.21): this is a typical use of must, and in such cases must may well be the preferred form when reported. (4.98) is another case where reported past reference must is presumably retained from an 'original' must. Similarly, it seems likely that must, when used for 'absolute' requirements as in (4.27) and (4.28), would remain if reported in past time. It may be hypothesized that where an original must is a marked choice it is likely to be retained in past time reporting. Free indirect speech is illustrated in the following: (4.141)

For a while it was Teresa they must cope with. She was completely frantic. (Coates 1983:40)

Presumably this corresponds to an assumed 'we must cope with Teresa'. This kind of use does not occur in the CEC, and is not referred to at all by Palmer (1990), but it appears to be typical of literary texts: Jacobsson (1979:303-306) provides numerous illustrations from written texts, both in BE and AE. Had to. The following texts illustrate regular past time reference with had to: (4.142)

(4.143)

(4.144)

well, he's away for a couple of days - Alec had to ring me about examination business this morning, and he said, he happened to mention that Oscar was away for a couple of days (CEC:63/363-366) and funnily, this year I got an examination paper he'd set which I had to see was all right (CEC:247/1413-1417) the ghastly thing was the Senior Common-room conversation that one had to conduct (CEC:92/533-535)

In all these have to would naturally be used in corresponding present time contexts (eg, for (4.143), I have to see this is all right). The occurrences of had to in the following display an idiomatic use: (4.145)

(4.146)

when we ate cherries, we didn't just try and suck them in by the stalk - oh no, we had to go one better (CEC:645/1352-1355) so he said, right, well, you'd better ring me then, so I said, OK, well, I'll ring you tonight some time, so I told the others about it and we all had to have hysterics again (CEC :701 /906-911 )

136 The speakers here present the subjects' behaviour as motivated by some inner, irresistible compulsion, and this is clearly the use illustrated by Quirk et al (1985:226) in: (4.147)

The rain was bad enough - but then it had to snow.

Of this, the authors comment "the author . . . jokingly blames the weather's perverse behaviour". This use is similar to the colloquial Do you have to? (also Must you?) used in friendly criticism, where an unreasonable internal compulsion in the addressee is implied by the speaker; it seems to be a metaphorical extension of the normal sense of have to. Since this use can perhaps also be seen as determined by the expectations of the speaker, it might be considered epistemic, and as such will be referred to again in 4.6. The next text shows a typical use of had to for reported speech: (4.148)

I mean, I had said that in order to understand Chaucer's visual imagination we had to understand the visual world in which he lived (CEC:88/318-320).

In this case, an original must or have got to seems possible, as well as have to: this can therefore be considered a case of possible suppletion for must. Had got to. It was stated above that had got to has restricted use; it does not occur in the CEC, and Palmer records no instances in the SEU. Coates (1983:54) cites one occurrence from fiction, as follows: (4.149)

She had got to think of some way out.

and suggests that it is only available for indirect speech (normal or free). Palmer (1990:120) claims that had to, but not had got to, implies actuality for the proposition, the crucial evidence being provided by the following (cf 3.6): (4.150) (4.151)

We had to make a special trip down to Epsom to collect the bloody thing, (from the SEU) We'd got to make a special trip down to Epsom anyway, so it didn't matter very much. (invented)

Palmer argues that in (4.150) it is assumed that the trip took place, while in (4.151) there is no such assumption. To investigate this proposal, we shall reconsider how such a distinction might be reflected in relation to the other examples cited above with had to. Firstly, in the group (4.142)-(4.144), where there is a strong indication that the relevant actions did occur, the use of had got to would in each case be slightly strange, but if it were used, it would be easier to say (to expand parts of the content of (4.142) and (4.143)), Alec had got to ring me this morning but he must have forgotten or I'd got to see it was all right but I couldn't be bothered than to say the corresponding sentences with had to. This suggests that an assumption of the fulfilment of a past obligation is more closely associated with had to than with had got to. Further, the fact that we had got to join in the conversation but kept quiet (cf (4.144)) seems quite as odd as the corresponding sentence with had to

137 suggests that a statement which directly denies the fulfilment of the had got to obligation is odder than one where non-fulfilment is indirectly implied. Secondly, in the case of (4.145)-(4.147), which have the idiomatic use of had to, had got to seems possible in the first text but not in the others, and it may be assumed that this is because in (4.145) there is less direct implication of fulfilment than in the other cases: there is no immediate indication in the text of what the 'one better' was, or whether it occurred, while it seems clear that in (4.146) the subjects did have hysterics, and in (4.147) it snowed. In the reported speech example in (4.148) had got to is possible without any oddity, but then here there is no question of actuality. These findings clearly support Palmer's claim that had to, but not had got to, carries an implication of realization. This particular difference between the use of had to and that of had got to to can be related to the general use of these items in the following way: have to focuses on an existing obligation and in past time reference it is natural to assume that this was fulfilled, particularly if the reference is general or habitual, as is often the case, while have got to focuses on the urgency of a requirement and in past time reference there is not the same assumption of its fulfilment. In other words, had to is concerned with both 'now' and 'then' points of view, while had got to is primarily concerned with 'then'. In terms of our S - Μ - E structures (4.5.1-2), we could represent had to as M - E - S or Μ,E - S, and had got to as M - (E) - S. Needed to. The following two texts illustrate the use of needed to: (4.152)

here are clear evidences of your having carefully picked out the minimum amounts that you needed to read in order to get up to standard on your Α-level texts (CEC:863/815)

(4.153)

I would have thought that he needed to divide his time - one third this external stuff, one third admin (CEC:479/878-881)

As with present need to, needed to relates a requirement to some motivation and thus appears weaker than had to (which would, however, be possible in these texts). The use of needed to in the following is worth noting: (4.154)

The prosecutors, Marcia Clark and Christopher Darden from the district attorney's office, described the defendant as a violent wife-beater who needed to control Nicole Simpson in every way. (The Economist, 28.1.1995:55)

This refers to an internally motivated need, and suggests that need (to) is in fact vague as to the source of the motivation it signals. Thus, to return to an earlier example, the need to use a ladder referred to in (4.105) is clearly motivated by physical problems, whereas in (4.154) an external source for the need to control a person would only be a reasonable interpretation in a society where individual freedom was at a low premium. There is some similarity between the use of need (to) for an internally motivated need seen here and the use of have to (and, to some extent, must) illustrated in (4.145)-(4.147).

138 Other aspects. The discussion so far has been of requirements set in the past, including such requirements when reported. It is, however, also possible to refer to a present requirement for past fulfilment of an action. Deontic meaning with must have + -ed and have (got) to have + -ed, with a present (or future) stipulation that some action have been carried out, is absolutely natural, although rare. Two relevant texts follow: (4.155)

cos if you're going to do honours, you've got to have done it well, you know, and I hadn't done this well, in fact I'd done it badly (CEC:599/311-315)

(4.156)

There is no argument for saying that in a particular locality nobody must have lived here who earns more than twenty pounds a week. (Palmer 1990:117)

In the first text the externality of the requirement in the i/-clause suggests why have (got) to have is more appropriate than must have. The second text, which Palmer quotes, actually as an instance of negation with must and not for the question under discussion here, shows a requirement for a negative proposition in past time. It is important to quote such texts since many descriptions disallow such uses: eg, Coates excludes modal + have -ed forms for root meaning apart from the case of should/ought to have -ed (required, it is stated, "as no past tense forms are available") (1983:240-241); cf, also, (4.204) and related discussion in 4.6.2. To summarize this section, it seems that must is available as the preferred form for use in reported speech where it would be a natural use in the putative original. The past tense uses of the other items essentially reflect their behaviour in present and future time reference.

4.5.4

Non-affirmative forms

This will deal with the use of the items in negative contexts. Interrogative uses of these items are not discussed since there is no indication that any distinctions in use between them arise that do not arise elsewhere; such forms are also rare in occurrence and thus hard to document. The following issues will be discussed: the negation of all the present tense forms of the items under discussion, with special attention to the alternative negative forms needn't and don't need to, and negation in past time. Negation of present tense forms. Mustn't/must not (together with structures like no one/nothing must . . . ) is unique among the modals in that the negation applies not to the modality but to the proposition27; by a suppletive process, don't have to, etc, serve to negate the modality. One question that naturally arises here is whether similar suppletion processes apply semantically - that is, what happens to the normal distinctions in use between must and have (got) to under the restricted possibilities for negation, since it 27

This restriction on the use of mustn't is much more evident in root than epistemic uses; thus, all the cases of must not/mustn't with modality negation quoted by Jacobsson (1979:298-300) are unambiguously epistemic (cf 4 . 6 ) .

139 should be expected that the contexts mustn't is used in do not necessarily correspond to typical positive contexts for must. The following texts illustrate negated must: (4.157)

well, I mustn't go on boringly talking about me (CEC:237/963)

(4.158)

Β well, one just mustn't mind A I don't mind being recorded surreptitiously (CEC:507/757-759) and and and Lev Mastine would be the last person to say nothing must be altered in the college until this grand plan has developed - I mean, of course, we've got to spend a little money (CEC:840/750-753)

(4.159)

(4.160)

his tail carries too low, or whatever Alsatians' tails mustn't do (CEC:727/1036-1037)

The assumption that mustn't is not necessarily restricted to contexts typical of must seems fully justified, as can be shown by considering possible alternative formulations of the must expressions in these texts. Thus, (4.157) could readily be related to I must stop talking ... , and not so naturally to expressions with have (got) to, (4.158) to the form in one has (got) to put up with it, (4.159) to that in everything must/has (got) to be left as it is - it is significant here that have got to occurs later in the same text in a parallel use, and (4.160) to that in Alsations' tails have to hang differently. The following texts illustrate negated have to and have got to (stresses shown in the original transcriptions are marked in texts (4.161)-(4.164), for have (got) to, and subsequently in (4.167)-(4.168), for need (to), for reasons that will become apparent; (4.165)-(4.166) have no prosodie markings in the original): (4.161)

your 'literature 'person will 'go for question '2, and 'they 'don't have to 'read the two 'passages and de'eide (CEC:50/945-947)

(4.162)

you 'don't have to 'wear 'any sort of 'glasses or 'anything, 'do you? (CEC:243/1246)

(4.163)

but you 'see, you 'haven't 'got to either 'like them or dis'like them or 'something with a 'portrait (CEC:215/876-877)

The first two texts, with negated have to, both have general reference, which is typical of that item. In (4.163), the emphasis regularly associated with have got to is made obvious under negation, since got receives stress; a similar effect could be achieved with ''' don't 'Aaue to. Finally, the following texts illustrate negated forms of need (to): (4.164)

the 'fact that a 'third of the 'whole of the de'partment disap'pear in the mi, the be'ginning of 'May 'does affect the 'other 'students - it 'doesn't 'need to, but it 'does, in 'fact (CEC:119/827-832)

140 (4.165)

Christ, I mean, when you're actually diagnosing a patient, you don't really need to know that, do you? (CEC:599/304)

(4.166)

and that kind of thing you'll not just, you you needn't worry with it (CEC:420/361)

(4.167)

and I 'don't 'think we need 'worry about 'that any 'more 'now (CEC:526/403)

(4.168)

although she's 'obviously 'highly 'qualified, her 'field isn't 'one that Ί think we need 'go for, and 'if she is 'also 'rather 'dubious on 'age 'grounds - I 'think that's how 'you should get a'round it (CEC:541/1152-1156)

The first two texts have need to and the last three need. The two questions to be considered here are, first, whether any distinction in use is evident between needn't and don't need to, and, second, how these forms relate to negated have (got) to. As to the first question, one possible point of distinction lies in the likelihood of realization of the proposition. (4.164) juxtaposes it doesn't need to and but it does, while in (4.165) it is not an issue whether the proposition is realized, and thus it might seem that don't need to leaves the realization open; by contrast, needn't/not ... need in (4.166)-(4.168) seems to imply non-realization. Thus, you needn't worry in (4.166) and its variant in (4.167) imply weak imperatives, as in don't worry, while the speaker in (4.168) is clearly negatively oriented towards the applicant under discussion. These interpretations are simply implications, since both the following are perfectly natural, and indistinguishable in effect: (4.169)

You needn't stay up unless you want to.

(4.170)

You don't need to stay up unless you want to.

Perkins (1983:63) has argued for a distinction between needn't and don't need to on the basis of the following: (4.171)

You needn't go to the toilet if you don't need to.

He claims that this sentence is possible because need is compatible with subjective or objective interpretations while need to only accepts an objective interpretation; further, the relationship of need and need to is seen as parallel to that of must and have to, respectively. This is supported by the fact that interchanging the verbs in (4.171), as in: (4.172)

You don't need to go to the toilet if you needn't.

is markedly odd. The proposal that in (4.171) needn't receives a 'subjective' interpretation is compatible with the suggestion above that needn't can function as a weak imperative, as (4.171) seems to be interpreted as 'don't go to the toilet if you don't need/have to'. If the proposed parallels between need and must, and need to and have to, respectively, are tried out, as in:

141 (4.173)

You must go to the toilet if you have to.

(4.174)

You have to go to the toilet if you must.

where (4.173) may be expected to have some correspondence to (4.171) and (4.174) to (4.172), the first is certainly less odd than the second, and can be interpreted as 'go to the toilet if you have to'. This supports the claimed parallel between the pairs of items. At the same time, these interpretations are highly dependent on their contexts and should not, perhaps, be argued for too strongly, as (4.173) and (4.171) are in effect saying, respectively, 'do/don't do χ if there are/are not reasons for this', and the 'subjectivity' and 'objectivity' evident may to some extent arise from the specific juxtapositions of the modals and periphrastics. The second question relating to (4.164)-(4.168) is how interchangeable needn't/don't need to are with don't have to/haven't got to. In general these items can be interchanged with no significant difference, and even the 'idiomatic' use of needn't in (4.166) can, in colloquial style, be expressed by phrases like you don't have to/haven't got to worry, which here are roughly equivalent to you mustn't worry. It is striking that while all these items normally make no suggestion that the proposition will be realized, the two cases where the final auxiliary verb element carries stress - you 1haven't *got to in (4.163) and it doesn't 1need to ... in (4.164) - each imply that the non-required act may well occur. Negation in past time. I shall finally turn to the ways these items are used in expressing the lack of necessity in past time. A marked contrast is often observed between (a) / needn't have -ed and (b) I didn't need to + inf// didn't have to + inf, such that (a) adopts a present time point of view and states that an action, now seen as unnecessary, was performed, while (b) adopt a past time point of view and state that the lack of need for an action was recognized, so that, by normal implication, it was not carried out. Needn't have, in fact, functions as a conditional form, parallel to wouldn't/couldn't have, etc, as can be seen from its use in sentences like: (4.175)

I l^g^J^

I have stayed up if I'd known you weren't coming.

Such cases axe simply counterfactual conditionals, and as such merit no special comment. This distinction between needn't have and didn't need/have to is, however, modified if have or need is stressed ( I didn't ^have/^nted to), in which case it is implied that the action was performed. The stress in these cases functions in a way similar to that in you 1haven't {got to, etc, discussed above.

4.5.5

Non-finite forms

The forms in question here are to have to, having to, had to, to need to, needing to and needed to. Need, being a modal like must, has no non-finite forms. Have got to is classified as a modal idiom (cf 2 . 2 ) and as such has no non-finite forms; nevertheless, non-finite

142 forms with have got to do not seem so bad as they do with the other idioms - a sentence like: (4.176)

To be able to - in fact, to have got to - take a regular bréale at work can seem quite a luxury to some people.

can at first sight/hearing be unproblematic. However, there are no such non-finite forms in the CEC, nor have I ever seen such forms cited. A few examples of non-finite forms of have to follow: (4.177)

I thought he was going to talk us into having to do another complete set of books (CEC:47/786)

(4.178)

and you hear the voice at the other end having to say 'yes' (CEC:192/914-915) because they're going to have to do it in their job (CEC-.262/658)

(4.179)

These illustrate the -ing form as gerund and participle and the infinitive. In some cases one can imagine an 'original' form could be must or have got to, and it may be assumed that, whatever the formal restrictions on must and have got to, these forms automatically become more 'objectified' in ways appropriate for have to in the process of subordination.

4.6

Epistemic uses of must, have (got) to and need (to)

The identification of EM was discussed in 3.2, and here I shall simply refer to epistemic 'uses' and, on occasion, to the possibility of ambiguity or merging between epistemic and deontic senses, which has already been encountered a few times above. This section will be concerned with the availability of the various items for epistemic use and the relation of patterns of use here to those shown in previous sections. There is a marked imbalance in the incidence of deontic as opposed to epistemic uses, and this also varies both according to variety (BE vs AE) and style (formal/written vs informal/spoken). Data summarized in (4.19) show that, for example, must is much more often deontic than epistemic in writing, whereas in speech this is not the case, and in fact the opposite may even apply; on the other hand, despite the widespread observation that epistemic have (got) to is typical of AE but rare in BE, epistemic have (got) to appears generally to be a markedly minority use. In the following, I shall deal with the relative status of must, have (got) to and need (to) as epistemic exponents in present /future time reference, which will involve some discussion of the complex issue of future reference in EM (4.6.1), and then I shall turn to past time reference (4.6.2). The discussion here will be complicated, firstly, by the marked lack of clear-cut data, and, secondly, by problems of definition and interpretation, since scholars in this field

143 seem to work with rather varied concepts of EM, and, in any case, there is abundant evidence that it is by no means always easy to distinguish it from DM. On the other hand, there is no indication that the epistemic uses of the items being discussed axe not relatable to, or derivable from, their deontic uses.

4.6.1

Present and future reference

A clear-cut contrast between present and future reference can often be associated with epistemic and deontic meaning, respectively, as is evident in the salient interpretations of Halliday's well-known pair: (4.180) (4.181)

You must be very careless. You must be very careful. (1970:326-328).

(The temporal reference of (4.180) in its epistemic sense is, incidentally, better seen as general or timeless rather than specifically present.) This temporal contrast can be further illustrated from a corpus text: (4.182)

A experimental psychology is Β psychologists A a very blinkered profession d they must do something - what do they do? (CEC:476/692-696)

The comment by d here is clearly epistemic, and must do is paralleled by what do they do?, which would make no sense with a deontic interpretation. The question arises whether EM can have future reference. As far as mxist is concerned, there are no clear-cut cases in the CEC, and Palmer reports none in the SEU (1990:54), but this must be seen as fortuitous, since, as Palmer notes, there is nothing odd about an example like: (4.183)

Something must happen next week.

Moreover, Anderson (1971:96) gives: (4.184)

They must arrive soon.

as a standard way of expressing epistemic future reference. Coates reports five occurrences of future-reference epistemic must in the SEU (1983:42-43), three of which are quoted below. Since such cases are so rare, they merit examination: (4.185)

. . . this sort of conversation, "have you noticed, President, that the boiled eggs at Sunday breakfast always hard?", and President said, "ah well, the simple truth is that if you're going to boil eggs communally they must be hard . . . " (CEC:96/789-799)

144 (4.186)

Certainly if there is endeavour to x, there must be attention to x. (written text, cited by Coates 1983:42)

(4.187)

McKenzie in, bowls to Edrich, and that pops up, and he must be caught, no, it's over Burge's head . . . (spoken text, cited by Coates 1983:43)

The categorization of these occurrences of must as having future reference can be questioned. (4.185) is paraphrased ' . . . it is necessarily the case that they will be hard', with the temporal reference of must be the same as that of the future boiling of eggs; but general (ie present-based) reference also seems plausible. The same applies to (4.186), which seems as likely to be a conditional sentence of the kind if χ happens, then y happens as one of the type if χ happens, then y will happen. Finally, the must clause in (4.187) can readily be seen as slotting into a sequence of 'instantaneous' simple tenses ( - in, bowls, pops up, must be caught, 's) such as are typically used in sports commentaries. In other words, it is doubtful whether any of these three examples is necessarily future in reference. To return to clearly future cases, texts like (4.183) and (4.184) seem naturally ambiguous between epistemic and deontic interpretations, and readings in such cases may merge: as Lyons notes of the use of will in: (4.188)

The successful candidate will be a woman in her mid-thirties of demonstrated ability.

"the two kinds of modality are hardly distinguishable" (1977:846). However, the following is an unambiguous case of future reference EM: (4.189)

The lorry must surely go tomorrow if it keeps fine. (R. Byron, The road to Oxiana. London: Penguin, 1992:112)

It is thus surprising to find the following, which seems perfectly normal, rejected as ungrammatical by Rivière in a discussion of this and related issues (1981:181) 28 : (4.190)

The plane must land in a few minutes.

The impossibility of future reference epistemic must is also claimed by McDowell (1987); cf discussion in 3.4. 2 9 I shall now move to the use of the other items for present and future reference. The following shows the only unambiguous cases of epistemic have (got) to in the CEC (there are several ambiguous cases): (4.191)

28 29

if you've seen the - if you've seen all the old Frankensteins you've got to, you've got to know the jokes (CEC:632/740-741)

The impossibility of future reference with epistemic must is central to Riviere's argument. It is suggested that the poor compatibility of must and future reference can be attributed to the origin of must in a past tense form (Nehls 1986:78). Yet such an origin does not prevent the use of must for future-oriented deontic use, while the same argument could surely be applied to other originally past tense forms, such as should or ought to.

145 A further epistemic case from the SEU is cited by both Palmer (1990:56) and Coates (1983:53): (4.192)

something has got to give in this second half, I think

This example must have future reference, though it is not treated as such by either Palmer or Coates. It is also striking that has got to here is potentially ambiguous between epistemic and deontic senses, just as must was in cases like (4.183), which has future reference. There seems no reason why will have to should not be available for epistemic uses - it would be perfectly natural in (4.192); however, I know of no such occurrences. Texts (4.191) and (4.192) were from spoken language; the following is an occurrence of epistemic have (got) to in written BE: (4.193)

"Higgins saw them arriving at the Black Swan. Unless one or both are lying Lampart has to be in the clear." (James 1986:334)

Such uses in writing seem to be quite common. I shall now go on consider the status of epistemic have (got) to in relation to epistemic must. Since EM is, by definition, normally 'subjective', and there is some justification for associating deontic have (got) to, as opposed to must, with 'objectivity', in a general sense of minimizing the speaker's role or focusing on external determining factors, it is natural to look for something like circumstantial evidence in epistemic uses of have (got) to. There is, in fact, some support for this here: while typical uses of epistemic must are speaker-based inferences, whether or not 'evidence' is explicit, the present cases with have (got) to seem to show deductions made within the discourse. In (4.191) the argument is 'if you've seen . . . , then you'll know the jokes'; in (4.192) the speaker appears to be referring to something internal to the progress of the sports match he is commenting on; while in (4.193) a detective is concluding a line of reasoning. In these three texts, have to and have got to are roughly interchangeable, just as has often been seen with deontic uses, though the restriction on the use of have got to to informal contexts applies equally to in EM. However, the use of must in these texts could have a somewhat different effect: in (4.191) it could have a meaning like 'surely with all your experience you know all the jokes!', in (4.192) Ί very much hope . . . ', and in (4.193) a hunch might be suggested rather than an explicit deduction. There is, then, evidence for a degree of 'objectivity' in the use of epistemic have (got) to; but this concept needs further clarification, in view of the fact that: (4.194)

She has got to be out.

- while appropriate if the speaker were making the assumption on the basis of past experience of the subject's habits - would be inappropriate in a situation when the speaker rang a doorbell and got no answer (this problem is posed by Swan 1993:355). This restriction, which suggests that the 'evidence' of an unanswered doorbell is somehow too strong for the use of have (got) to, but not for the use of must, can be further illustrated with the following:

146 (4.195)

The plane must land soon.

(4.196)

The plane has (got) to land soon.

Two scenarios can be considered for these. In the first, the speaker is waiting by the arrivals board at an airport, arrival time has passed, and the expected flight is due down next: here both (4.195) and (4.196) seem natural. In the second, the speaker is actually watching the plane descending, and touchdown is expected in a matter of seconds: now (4.195) is fine, but (4.196) seems less natural. The reason for the restriction evident in (4.194) and (4.196) appears to be as follows: have (got) to, epistemically, involves a strong deduction based on some evidence, but the necessary gap between evidence and deduction is reduced too far when the evidence is immediately available and prominent. Must, by contrast, is not so closely related to the nature of the evidence for a claim. The distinction involved here between must and have (got) to in epistemic use is dealt with by Matthews (1991:235), in the following terms: (4.197)

. . . whereas must reflects the spontaneous performance of an inference with respect to some new evidence, have to and have got to are assertions of objectively or logically necessary inferences.

This accords to a considerable extent with the proposals in the discussion above. The question of a BE-AE distinction in the epistemic use of these items, which is very widely reported, remains to be discussed. Palmer (1990:56) states of: (4.198)

You've got to be joking.

that "the more likely British form" is You must be joking. It has also been argued that epistemic have (got) to is an AE-motivated innovation in BE typical of younger speakers (Hughes & Trudgill 1979:23) or of the teenage sub-culture (Coates 1983:57). While there are plenty of indications that epistemic have (got) to is well established in BE today (cf (4.193), for example, from a highly respected author), there is clearly a varietal contrast here that is valid for some BE speakers, and is perhaps salient in cases like (4.197), which is distinct from any demonstrable contrasts in meaning. I shall now move on to the epistemic use of need (to). This is very hard to document, but a typical case would be: (4.199)

This may be the right solution, but it needn't be/doesn't need to be.

In such a case there would be effective equivalence both between needn't and doesn't need to and these two and doesn't have to/hasn't (got) to. The following are corpus occurrences for consideration: (4.200)

Β c

oh gosh, getting married is an awfully complicated business yeah, I know, so I believe, I think, I, one ought to hand it all over to an agency or something - I'm sure this would, would simplify matters Β actually, it needn't be - it can be very straightforward (CEC:657/369-376)

147 (4.201)

I agree with him that it could well happen - I don't see that it necessarily need happen - I think it depends very much on how it's handled (CEC:583/510-512)

The first case is treated as epistemic by Coates (1983:50), with needn't glossed 'it isn't necessarily the case that . . . '. This seems questionable: a more appropriate gloss would surely be 'it isn't necessary for . . . ', and this would firmly point to a deontic sense; it is also relevant that needn't be is followed by positive can be, which is never epistemic. The distinction between the two interpretations is, however, not great, and is perhaps neutralized under negation. (4.201) shows the collocation of necessarily with need, common in non-assertive contexts. Again, it is doubtful whether this should be seen as epistemic rather than deontic, but this is presumably because it has future reference and the modality appears ambiguous. If need(n't) and must are considered semantically related (cf 4.5.4), such an example might be seen as parallel to, for example, (4.183), which similarly has future reference and is ambiguous in its modality.

4.6.2

Past reference

Two types of expression are involved here: epistemic judgements made in the present about the past, and those made in the past about the then present or past; the normal exponent of the first is must have (this will not be illustrated here), and of the second had to (in each case, these forms can also have root readings, so that there is potential ambiguity). As to the first case, the form have (got) to have, as an alternative to must have, is a rare form, and the only occurrence in the CEC, quoted above as (4.155), is clearly a root use. The form is, however, possible in an epistemic sense, as the following vaxiation on (4.191) shows: (4.202)

You've got to have seen some of the old Frankenstein films when you were a kid.

The relevant forms of need (to) are need to have, with needn't have, don't need to have, and the corresponding interrogative forms. There are no epistemic instances of these in the CEC, and the only claimed cases I have found in the literature cannot be accepted. The first follows: (4.203)

The basic questions for the new American administration are two: need the quarrel with Cuba ever have happened, and, can it be put into reverse? (Coates 1983:51)

Need ... have here means 'was it necessary for it to happen', or, perhaps more exactly, 'is there a reason why it happened?' An epistemic reading would be paraphrasable 'is it necessarily the case that it happened?', which would be inappropriate here. Secondly, it is argued that need have in: (4.204)

Ella need not have gone.

148 must be epistemic since it displays a characteristic of epistemic modals (ie, MV + perfect infinitive) (Newmeyer 1975:80). This claim cannot be supported: have got to have and must have, for example, were illustrated in deontic uses in (4.155) and (4.156), respectively, and the salient sense of (4.204) is deontic. 30 Yet, although there are no good corpus examples to be found, the following: (4.205)

"He must have finished by now!" "No, I don't think he need have/Surely he needn't have."

shows that epistemic need have is possible. There are no unambiguous cases of the second type of epistemic past reference in the CEC, but the following texts axe cited in the literature: (4.206)

It had to be there - there wasn't anywhere else it could have been. (Palmer 1990:65)

(4.207)

"Let's sit here for a few minutes, shall we? This boulder looks very inviting." It had to be the same boulder that he and Betty had sat upon so often. Bill gave an inward groan. (Coates 1983:57)

Palmer glosses (4.206) as 'It was always necessary that . . . ' or 'The only possible conclusion was that'; and it is clear that the difference between this epistemic sense and a deontic sense, meaning that there was something inherent in the subject that made it be there, is narrow, just as it is with future reference uses, as in (4.183), for example. Coates, similarly, treats the use of had to in (4.207) as epistemic, but the text also allows, especially in view of inward groan, an interpretation similar to that proposed for (4.145)(4.147) above; it was noted in 4 . 5 . 3 that an epistemic sense might also be considered possible for these cases. With (4.207) a deontic sense would express a resigned or ironic attitude to the speaker's bad luck. The fact that there is no evidence of had got to being used epistemically parallels the situation in root uses: in contrast to had to, it lacks any suggestion of actuality, and this would be particularly odd for an epistemic use, which in past reference would naturally suggest actuality. Nevertheless, its use in a case like I was absolutely convinced it had got to be true, with urgency attaching to the subordinated assumption, seems perfectly natural. Finally, we come to the availability of must for past time epistemic reference. The deontic use of past reference must was dealt with in 4.5.3; a parallel epistemic use is rarely noted in the major literature, but is amply documented for written texts in BE and AE by Jacobsson (1979:306). Just as with deontic uses, many cases can be treated as normal or free indirect speech, and it seems that a hypothetical original comment could always have must. An example from Jacobsson follows: (4.208)

30

He was a detective, and he must be at least six feet tall, big through the shoulder and with the beginning of a gut.

Newmeyer seems to use an idiosyncratic sense of 'epistemic', since, among other examples of EM he gives How can you be so dumb? and I should have done that yesterday (1975:72-73). An epistemic interpretation, at least with a standard definition of EM, would only seem appropriate for the latter if the speaker were an amnesiac.

149 The following is similar: (4.209)

Alison might well have changed her mind. If she had not, some terribly agonizing bloody scene must now be going on at the house in Chelsea upon which none of the victims was likely yet to report. (Murdoch 1990:370)

In these cases, had (got) to would be less natural, perhaps because it has a salient deontic interpretation; and, in relation to (4.209), complex locutions, such as . . . had now got to be going on ... , ... now had to be going on . . . , or would now have to be going on, are perhaps simply avoided.

4.7

Be bound to

This item is discussed separately, since it is not obviously semantically close to the other items, and its use is relatively rare and restricted (eg, there are just six occurrences in the CEC). However, it has been related by scholars to must in ways that merit examination. The following five texts, from the CEC and elsewhere, illustrate various uses of the item: (4.210)

well, I think if one believes this, one, one, you know, is bound to act upon it and stir other people (CEC:469/353-356)

(4.211)

a C

but they've had a film on Hockney yes, they had it on the last day - this was the last night I was there, actually, but α Ά Bigger Splash' C it's bound to come out, though, I think a do you think so? C well, yes, it, you know, it's received such rave notices everywhere that somebody's bound to put it on a what do you mean, come out in the circuits? C well, not, I don't know about the circuits, but it'll probably go on at the Academy or the Curzon or somewhere, I should think (CEC:548/155-178)

(4.212)

I mean, the way the British Army have decided to play it there is one that's bound not to have the most spectacular results when it comes to the suppression of the IRA (CEC:583/531-533)

(4.213)

They never spoke of it in words . . . Perhaps they would not, were somehow bound never to, speak of it again. (Murdoch 1990:231)

150 (4.214)

"But what's happening?" "Can't you understand? What was bound to happen. Listen, Ludens, if I say it to you I can do it . . . " (Murdoch 1990:365)

In (4.210) is bound to has a deontic sense: the meanings is that beliefs should, morally, lead to actions, and this is stronger than, or rather different from, must (and the other items). An alternative epistemic sense would be possible, meaning is certain to, and again quite distinct from epistemic must, etc Both occurrences in (4.211) are clearly epistemic, and the second contrasts with probably ... I should think, which can also be considered exponents of EM in a broad sense; a gloss for is bound to is 'definitely, beyond question'. In (4.212) is bound to is similarly epistemic. In (4.213) the sense seems to be deontic, but an epistemic sense is also possible: the two interpretations are distinct, in that a root sense refers to an obligation on the subject, and an epistemic sense to an action, or rather its lack, that was fated. (4.214) has a comparable effect; was bound to denotes what was fated, or had to be, but in this case the distinction between the readings for (4.213) no longer applies, apparently because of the impersonal subject (what was bound to happen), and here there seems to be merging between epistemic and deontic senses. Such evidence does not suggest any very close semantic similaxity between be bound to and the other items: epistemic be bound to makes a strong claim that seems to rule out the possibility that the proposition may not occur, and is closely paraphrasable by be sure/certain to, while the deontic use, in cases where it is clearly distinct from the epistemic, makes ail explicit statement of a moral requirement. 31 It is striking that, like other periphrastics, it shows alternation between an epistemic and a non-epistemic sense (see 2 . 3 . 2 ) : sometimes the two senses are clearly distinct, but they can merge. A specific claim involving the use of be bound to will now be discussed. Palmer (1990:55-56) argues that while epistemic must normally has present reference (future reference with must being interpreted deontically), be bound to in its epistemic sense functions as a kind of suppletive item for future reference. He locates the crucial distinction between the items in the sense of conclusivity expressed by must and associated with present time and that of inevitability expressed by be bound to and associated with future time. Among other scholars, Coates supports Palmer's claim but emphasizes both the rarity of be bound to and the availability of must for future reference (1983:42-43), while Nehls treats be bound to as only having a very limited suppletive function in relation to must (1986:80). Two problems arise here. Firstly, it is not clear that the items are so clearly tied to particular tense uses: epistemic must can refer to present or future time, as was shown in 4 . 6 . 1 , while be bound to is not restricted to future reference (eg, they're bound to be there by now refers unambiguously to present time). Secondly, it is not obvious that 'inevitability' and 'conclusivity' are reconcilable notions, or can equally well be seen as epistemic: that's bound to be the answer is different from that must be the answer, and is, in fact, closer to that will be the answer. It is, thus, hard to see substantial 31

Downing L· Locke (1992:386) see be bound to as an alternative to must, equally subjective. This is apparently because the epistemic sense, which is the only one considered, is typically subjective; the presence of be in a complex form has elsewhere, as we have seen (3.3.3, especially (3.60)), been taken as a sign of objectivity.

151 justification for assuming any significant relation between be bound to and must of the kind demonstrable between other M-P pairs and sets. 32

4.8

General conclusions

The findings will first be summarized here, together with brief comment on some implications, and then various major proposals made in relation to these items in the literature will be reviewed.

4.8.1

General findings

The proposal set out in (4.14) assumed that the set must, have to and have got to are, in semantic terms, cognitively equivalent but can be considered as separate items. There is certainly ample evidence that these three items axe close in their uses to the extent that they are often more or less interchangeable. However, since the uses of have to and have got to axe relatively recent developments, it is reasonable to assume that they will show meanings that both are more specific than that of must and reflect their origin. The following comments apply in the first instance to deontic uses of the items in present tense forms. The evidence suggests that must is a very general marker of 'necessity'/'obligation', just as, for example, can is of 'possibility', and that any appearance of more specific meaning is to be attributed to pragmatic interpretation. This was shown to explain the use of must for, variously, absolute, unqualified, unconditional, urgent, immediate, irresistible, spontaneous or unique requirements, and the expression of the speaker's wishes or his sympathy with the subject or the obligation. This issue is highlighted in the stark contrast between the often claimed characteristic - or even obligatory - association of must with the expression or imposition of the speaker's will and the evidence produced here which suggests that such a use, far from being typical, is in fact hard to document. Have to focuses on an external, existent obligation that can be perceived or described independently of the speaker. In specific contexts, an obligation may appear, for example, habitual/general, timeless or distinct from the speaker. Have got to is a variant of this form found in colloquial contexts, but is emphatic and presents an obligation as relatively urgent, and may in particular contexts appear, for example, to require immediate fulfilment, be specific in reference or be stronger than have to, and may also make the strongest possible statement of a speaker-based requirement. Because of this, have got to sometimes appears semantically closer to must than to have to. All this results in certain typical contexts of use relating to specific interpretations; and thus we have default or favoured interpretations - eg, that the speaker of a must sentence supports the requirement, or that a have to requirement arises outside the speaker. There 32

An assumption of semantic closeness between be bound to and must presumably lies behind the choice of be bound to as a suitable paraphrase for interrogative need by Coates (1983:51).

152 also are a number of idiomatic or fossilized uses, where the 'obligation' sense has lost its force - eg, the I really must be going, You must come and see us and I have to go to a party (for an intention) types discussed above. Aside from these issues, the importance of style and variety in the choice of these items has been demonstrated at several points. The colloquial nature of have got to is well known, but an association of some uses of have to to informal and of must to formal usage was also shown to be a significant variable. This entails that stylistically neutral comments on the use of these items can only have limited validity, and makes the determination of core uses of the items that much more difficult. It also results in complex interaction between stylistic and other contextual factors, as will be shown below. Distinctions in use between varieties, primarily BE and AE, have on occasion been noted, but, firstly, this has not been done systematically since that would itself be a major undertaking, and, secondly, the evidence here suggests that significant distinctions may not in fact be very great. A particular type of possible contrast between must and have to which relates strictly to discourse functioning was noted tentatively in 4.2.1 in connection with text (4.4). It was suggested that whereas the semantically more highly specified have to was used initially, repeated reference to the requirement might be adequately expressed by the unmarked must. The following text - assuming that more than purely stylistically motivated variation is involved - appears to provide further illustration of this phenomenon: (4.215)

In his study of Native American languages Sapir noted that speakers of different languages have to pay attention to different aspects of reality simply to put words together into grammatical sentences. For example, when English speakers decide whether or not to put -ed onto the end of a verb, they must pay attention to tense, the relative time of occurrence of the event they are referring to and the moment of speaking. Wintu speakers need not bother with tense, but when they decide which suffix to put on their verbs, they must pay attention to whether the knowledge they are conveying was learned through direct observation or by hearsay. (S. Pinker, The language instinct. London: Penguin, 1995:59)

Here the general comment referring to speakers of different languages has have to, but the two ensuing illustrative cases referring to English and Wintu speakers have must. An apparently parallel case of the use of a periphrastic followed by a related modal is discussed in 5.3.1.2. The findings summarized above were also related to the use of non-present tense forms, and, separately, to epistemic uses. The same general observations seem to be applicable across all these uses, allowing for, for example, the suppletive functions of have to and, to a limited extent, have got to, but there are striking imbalances between the two major types of modality, in particular in the very restricted epistemic use of have (got) to, which may partly be attributed to the inherently strong speaker-centredness of EM, which thus favours must. Some evidence was shown that the epistemic use of have (got) to is not especially typical of AE, and may be related to 'objective' deductions rather than evidence leading directly to a conclusion.

153 Need (to) is distinguished from the other items in focusing on the motivation for an action as opposed to its necessity, and as such is usually weaker and not interchangeable. However, in non-assertive uses, and particularly under negation, the distinctions may not be so evident, and thus there may appear to be correspondences or relations, for example between must and needn't or between have to and don't need to. Evidence was shown that need (to), like have to, may indicate motivation arising internally as well as externally to the speaker. Numerous instances of these various phenomena were given above; a group of invented examples will now be discussed, as a way of further illustrating the findings regarding must, have to and have got to. Typical speech/writing distinctions will be indicated where they are relevant. This will also illustrate, again, the difficulties of discussing these items in relative isolation. (4.216)

You must come and see our garden.

(4.217)

You have to come and see our garden.

(4.218)

You have got to come and see our garden.

(4.219)

She must go home at the weekend.

(4.220)

She has to go home at the weekend.

(4.221)

She has got to go home at the weekend.

(4.222)

There must be some sort of authority in charge.

(4.223)

There has to be some sort of authority in charge.

(4.224)

There has got to be some sort of authority in charge.

The purpose here is to see how such sentences are 'neutrally' interpreted, and how this reflects the claims made here for the potentially contrasting meanings of the three items being discussed. BE is normally assumed to be the variety involved. I shall assume that the first set represent informal speech between equals.. (4.216) is a standard type of friendly invitation: the circumstances of utterance rule out any kind of obligation or requirement, the source, in the lack of any other evident motivation, will be taken to be the speaker, fulfilment is hoped for, and the utterance is interpreted as a maximally strong invitation. To the extent that the context is seen as a regularlyoccurring set of factors, and the form as recognizably standard (you must . . . ), this is a conventionalized structure. (4.217) .and (4.218), in the same context, would have the same effect; but they are not similarly conventionalized, or natural, expressions, since have (got) to points to some outside obligating factor, and no such factor would here be involved. In a neutral context - between strangers or non-close acquaintances - (4.217) and (4.218) would suggest some identifiable external reason for the statement. The second set allow more variety of interpretation, arising, specifically, from (a) the ambiguity between deontic and epistemic meaning, (b) the ambiguity between specific (this (coming) weekend) and general (every weekend) reference, and (c) possible speech vs writing distinctions. In speech/informal style, the neutral interpretation for (4.219), specifically when must is stressed, would be epistemic ( . . . otherwise we'd see her around here more), and then general in reference rather than specific (cf the raxity of explicit future epistemic reference with must, which, as was suggested, might merge with deontic

154 meaning); an alternative interpretation would be deontic with specific reference, with a degree of speaker initiation. For (4.220) the interpretation would naturally be deontic, and equally well either specific or general in reference ( . . . her family can't manage on their own, for example, in the context would allow either), while for (4.221) it would be deontic and specific ( . . . she's got a big family party on). In both sentences, an epistemic interpretation would be less expected but possible, but then it could only be general ( ... because she's always up-to-date on all the local gossip). In writing/formal style, (4.219) would equally well allow deontic meaning, (4.220) would seem to be restricted to deontic meaning (whether specific or general), while (4.221) would not occur. Finally, the third set equally well allow deontic and epistemic interpretations, with the formality factor having no obvious significance in (4.222) and (4.223) ((4.224), with have got to, is necessarily informal). The ambiguity here rests partly in the vagueness of be, which could in a deontic sense refer to what is required ( . . . it's laid down in the regulations) and in an epistemic sense to what is evident ( . . . it's so orderly). Have to, whether deontic or epistemic, will be appropriate if an external factor is focused on strongly. As to the use of need to, the following are worth considering: (4.225)

She needs to go home at the weekend.

(4.226)

Does she need to go home at the weekend?

The first sentence points to motivation for the required action: the default interpretation seems to involve an external source, but the motivation could also arise from the subject herself. The time reference is vague between specific and general. The second sentence strongly suggests that the motivation is external. Be bound to was found to be distinct enough in use from must (and the other items) to justify not being taken account of in close combination with them. Some more general considerations that also arise here will be discussed in 6, for example the hazards both of dealing with items in context-free isolation and of depending on intuition for descriptive data, the difficulty of determining 'core' meanings for the items, and the inherent problems in some of the metalanguage used (eg, subjectivity and objectivity).

4.8.2

Review of proposals in the literature

This section will briefly review, in the light of the present findings, first some major proposals for the description of these items (4.8.2.1), and then some individual notions of possible relevance (4.8.2.2). I shall concentrate on proposals that have not been involved in the description above. 4.8.2.1

Major proposals

Here I shall first discuss the findings relating to these items of three major full-length accounts of modals and modality, all British though the first two on occasion deal with

155 BE-AE contrasts. The fourth account is a brief corpus-based account of must, have to and have got to in AE, and will be discussed on its own. Palmer's account of must and have (got) to (1990; originally 1979) distinguishes between three types of non-epistemic necessity - deontic (= performative), external (for obligations arising outside the speaker), and neutral (for uses not specifically relatable to either of the other types) (cf 3.2). The distribution of these 'meanings' across the exponents can be depicted as follows (based on Palmer 1990:116):

(4.227) performative

neutral

external

must have (got) to

The accounts by Coates (1983:Ch.4) and Perkins (1983:Chs 3 and 4) can be, very roughly, equated in this respect, and depicted as follows:

(4.228) core

skirt subjective

periphery objective

must have to have got to

In all these cases, maxked overlap between the uses of the three items is allowed, with must based at the subjective, speaker-oriented end of the scale, and have to at the objective, external end. These accounts differ considerably both from one another and from the findings in the present work, particularly in their treatment of have got to, which they never allow extreme subjective (performative) uses, and in their focusing on one factor as determiner of variation; and they do not attempt to deal systematically with the problem of stylistic variation in usage. The account by Melrose (1983) is mentioned separately, since it (a) is concerned with AE, (b) is very strictly determined by a corpus, supplemented by informant testing, and (c) arrives at narrower and rather different findings from the others. Her findings can be summarized as follows:

(4.229)

(a)

must: in speech - only normally found in the perfect must have + -ed form for epistemic use; otherwise a rare, marked form (this second finding was not supported by informant testing); in writing - the generally preferred form

156 (b)

(c)

have to: in speech - used for deontic and dynamic modality (for the latter term, cf 3.2) 3 3 ; in writing - used for dynamic modality in less formal contexts have got to: in speech - used for 'affective modality', which may refer to all forms of modality when emotionally loaded (eg, with urgency or insistence); in writing - no function

Some of the claims embodied in Melrose's account have already been discussed in this chapter, and receive substantial support; it is not clear that its findings are markedly peculiar to AE as opposed to BE, and the distinct treatment accorded to have got to, as well as the consistent stylistic differentiation, should be noted. The specific point that where intuition was used, in the form of informant testing, the results were at odds with the corpus findings is significant, since the present account has suggested that must is in some ways more narrowly used than is often (intuitively) claimed. 4.8.2.2

Some individual proposals

Individual proposals that may be relevant to the description of these items can be divided into (a) those that associate the use of must either with the speaker, specifically, or a personal will, more generally, as source, or with the speaker's viewpoint, (b) those that focus on various types of necessity involved, (c) the assumption of specific implications of actuality attaching to specific items, (d) claims for the relative inherent strength of the items, and (e) arguments that one or other items should count as marked. The first type of proposal was discussed in some detail in 3.3.4-5 and has been examined extensively in this chapter, especially in 4.4.2.1, and requires no further comment here. Various types of necessity. The second type of proposal has come up at various points - for example, in the association of must, but not have (got) to, with new obligations (4.4.1.3), and similarly with a moral perspective (4.4.2.1). Also under this heading comes the distinction made by McCallum-Bayliss (1985:here 131) between the kinds of background evidence available to a speaker in choosing between must and have to. It is argued that (4.230)

33

By using have to the speaker is making a better substantiated claim than with must; the character of the evidence that has gone into the assessment of the proposition is different - for have to it serves as a testimony to the validity of the statement; with must it conversationally justifies the claim. That is the difference between grounds and evidence.

According to the corpus figures provided, it is also used for epistemic modality, but Melrose does not refer to this.

157 This suggests that the 'evidence' behind the use of have to is stronger than and different from that behind must, in that it can objectively justify the obligation or claim. 34 This argument is illustrated most explicitly by the following example: (4.231)

Did you know that Archie's mother is going to have an operation? He asked his brothers if they wouldn't help pay for it and they absolutely refused . . . a So he must sell his house to pay for her operation. (—)3S b So he has to sell his house to pay for her operation. (+)

where it is claimed that the relevant background information supplied makes response a odd in comparison with response b (ibid:132). The argument is claimed to be equally applicable to EM, so that, to extend the material in (4.231), Archie's mother has to be poor implies that the speaker has solid grounds for his claim, while Archie's mother must be poor is more a reasoned guess. This argument receives plenty of support in the material examined above, and has the rare value of highlighting the close parallelism between the uses of the items in DM and EM. Implied actuality. We next come to the theory that an implication of actuality may attach to specific items under certain contextual circumstances (cf general discussion in 3.6). The possible application of this to the forms had to and had got to was discussed in 4.5.3 with particular reference to (4.150) and (4.151), and was shown to be reflected in other data presented there and to be relatable to the general use of the items. Two other phenomena have also been found relevant here, but it is not clear that the notion of implied actuality is relevant in these cases. Firstly, a parallel claim relating to present tense forms is made on the basis of the following example from the SEU (Palmer 1990:116,195): (4.232)

It's a slow walk down. He's got to fight his way through the crowds.

It is argued that has (got) to (both forms seem to be in question) implicates actuality - that is, the sense is 'has to and does'. By contrast, the use of must would be odd, and could only refer to future time (1990:116; cf 1980:98). This text is, however, rather problematic. The context, apparently a sports commentary, certainly makes it clear that the subject, a boxer, not only has got to, but does (at the time of speaking) fight his way through the crowd. Yet this could surely just as well be treated as an example of the instantaneous use of the present simple, untypically modalized: in this case there would be no anomaly in the 'actualization' of the obligation. (Must would, in any case, be rather odd in this context.) As a further, but distinct, issue, it could be suggested that the availability of the distinct form will have to for future reference may well make have (got) to a more explicitly 'present' form than must (cf 4.5.1). Secondly, Haegeman argues that the use of have to in progressive aspect "gives even more prominence to the actuality" of the proposition (1980:4). She points out that: (4.233) 34

35

Thousands of people «ire having to sleep in the open.

There is a confusing ambiguity in this account in the use of 'evidence', which variously subsumes, and contrasts with, 'grounds'. These 'relative normality' signs have been added.

158 "indicates . . . that people are actually sleeping in the open". In this case, of course, no contrast with the other items is possible, since neither must nor have got to has a progressive form available. However, it is not obvious that the effect of the progressive in (4.233) is totally different from what it would be with a lexical verb: are having to sleep suggests current activity just as are sleeping does, while have to sleep, like sleep, is vague as to exact temporal reference. In other words, if there is an 'actuality' issue here, it is not peculiar to have to. Relative strength of the items. The next type of claim to be discussed concerns the relative strengths of the necessity expressed by the individual items. It is often argued that one or other of these items makes per se a stronger claim, whether in deontic or in epistemic uses. In the following section, the cases for must and have (got) to will first be discussed, and then the question whether such a difference can in fact be demonstrated will be raised. Must as the stronger item. This may be typified by Whitaker's 'The mighty must1 (1966), where must is said to make a uniquely strong, 'irresistible' requirement. Similar claims are made widely; to give a few examples, Kruisinga states that in cases where both must and have to are possible "the difference, if any, is that must is more emphatic" (1925:347); likewise, Alexander (1988:228) states "must conveys more strongly than have to the idea of inescapable obligation". This claim might find support in a text such as the following: (4.234)

Sir David: Major: Sir David: Major: ( The Observer,

'He's got to go.' 'He's not going.' 'But he must go.' 'Sir David, may I repeat myself - he's not going.' 27.9.1992:16)

Reversing the order of have got to and must here would look rather odder, though hardly be impossible. Also relevant to this claim are the following data:

(4.235)

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

??I I I I

must have to need to ought to

get this paper in, but I guess I'll go to the movies instead.

on which Sweetser comments (1990:54): (4.236)

Must has connotations of a directly applied and irresistible force, while have to, ought and need axe resistible forces . . .

The findings in (4.236) are, however, not self-evident. While (4.235)(c) and (4.235)(d) axe natural, (4.235)(b) is hardly better than (4.235)(a), and elsewhere (cf (3.92)-(3.93) and related discussion in 3.3.5) it is argued that both must and have (got) to allow the speaker a degree of non-acceptance of the obligation. However, indirect evidence for the inherent strength of must in comparison with have to can be derived from the following observation by Fleischman (1982:59):

159

(4.237)

For certain speakers of American English, have to currently conveys little more than an intent to engage in a future activity . . .

This is illustrated by examples like the following, already cited above and attested in a situation where no 'obligation' was in question: (4.12)

What are you doing tonight? Oh, I have to go to a party.

It is hard to imagine must being used in such a weak, non-obligational sense. Have (got) to as the stronger item. Some investigators have, by contrast, found have (got) to to be stronger (eg, Dirven & Radden 1977:94; Suzuki 1978:103-104; Smith 1982:82-83; McCallum-Bayliss 1985:131). The claims made will now be illustrated; it is striking that no distinction seems to have been made here between the items have to and have got to. Smith (1982) states that of the pair: (4.238)

You must take that garbage out, damn it!

(—)

(4.239)

You've got to take that garbage out, damn it!

(+)

the latter is more appropriate to the context, which is defined as informal (damn it!), and as such makes a stronger demand. This argument seems to equate 'strength' with informality, and accords with observations earlier in this chapter about the nature of have got to (it is significant that have got to, and not have to, is the item contrasted with must here). Smith's claim is apparently applicable to both deontic and epistemic uses. Other authors focus on the external grounds or inherent logic associated with the use of have (got) to. Thus, McCallum-Bayliss reports a general feeling among informants that sentences with have to seem "stronger, more certain, or absolute" than their counterparts with must; the feeling of strength is then related to the presence of 'grounds' associated with the choice of have to (cf (4.230) above). Finally, Quirk et al (1985:145-146) treat have (got) to as capable of a stronger 'logical necessity' (ie, epistemic) meaning "which cannot be matched by the use of must", as in: (4.240)

There has (got) to be a first time for everything.

(This is reflected in the discussion of (4.194)-(4.196) in 4.6.1 above.) No corresponding claim is made by Quirk et al for deontic meaning. Comparability of strength. Evidence that must and have got to, at least, have comparable strength comes in a text cited by Nehls (1986:36-37) as evidence of interchangeability between the two items, discussed above in 4.2.1: (4.6)

"I'm afraid I've got to be off now. My parents have come to see me unexpectedly." He paused, to give room for any cries of protest and regret. When none came, he hurried on: "Thank you very much for putting me up, Mrs Welsh; I've enjoyed myself very much. And now I'm afraid I really must be off. Good-Bye, all!" (K. Amis, Lucky Jim. Harmondsworth: Penguin, 1961:80)

160 If this text did point to any difference in strength, it might suggest must as stronger, since that occurs second, where the stronger item would be expected (assuming end-focus as a general principle), and collocates with the emphatic really; on the other hand, reversal of the items does not create any oddity. It seems, in fact, impossible to show any absolute difference in strength among these items by means of the standard tests (cf discussion in 2.1), as variations on (4.235) show: (4.241)

It's not just that I ought to get this paper in - I have to/have got t o / m u s t .

(4.242)

*It's not just that I have to/have got to/must get this paper in - I ought to.

(4.243)

?It's not just that I must get this paper in - I have to/have got to.

(4.244)

?It's not just that I have to/have got to get this paper in - I must.

Here, (4.241) and (4.242) are clear cases of normality and abnormality, respectively; but (4.243) and (4.244), as they stand, both simply seem a bit odd. However, with heavy stress on the second, contrasted item they are both natural, and then, as would be expected, (4.243) suggests that there are overriding external reasons for performing the action, while (4.244) suggests that personal feelings are more important than any external requirement. Conclusion on relative strength. The fact that most statements that argue for the greater strength of one (or two) of these items are combined with other claims (regarding, eg, speaker's grounds, or relative formality), together with the impossibility of demonstrating such values absolutely, suggests that 'strength' alone may not be a useful concept for our purposes; and the whole issue is made more complex by data such as: (4.5)

They have been understanding with me since Lockerbie and they have put up with a lot but sometimes people just have to do what they must. (The Observer, 16.2.1992:2)

This text suggests that the items must and have to are not absolutely equivalent, since the effect here is quite different from that with tautologous repetition of the same item, as in I know what I know. Rather, two close but distinct meanings are suggested, as in I know what I can understand/what I can understand I know. The positions of the items in (4.5) cannot easily be reversed (?People just must do what they have to), and the must clause looks similar to the 'absolute' use of must - also in relative clauses - illustrated in (4.27) and (4.28). Such a use could certainly be considered 'stronger', for what that may be worth. The following text is also worth considering in this connection: (4.245)

" . . . Then I saw - it was then I saw - that I must get rid of Franca. I know she's an innocent person, she's put up with so much, I think she'd put up with anything. But I've got to destroy her . . . he's turned her into an animal - a sweet little animal - which I've got to kill." (Murdoch 1990:366-367)

161 Here, too, the uses of must and have got to, although they are parallel within the text, have slightly different uses, and are not interchangeable: must denotes an absolute, overriding necessity the speaker feels, while have got to more specifically indicates a need the speaker is objectively aware of. Markedness. Finally, we come to the notion of markedness, discussed in some detail in 3 . 8 . As was seen there, Perkins (1983:102-104) follows Bouma (1975:325) in treating the modals as a class as unmarked in relation to their periphrastic counterparts - and, in fact, all semantically equivalent forms. Have to can be seen as marked in relation to must in a number of ways: for example, it has a more specific meaning than must (in that it focuses on an independently existing requirement), and is temporally more specific (in that it has future and past tense forms). Coates' approach, in allowing have to only a narrowly limited 'objective' area of meaning (cf (4.228)), points to the same solution. The case of have got to complicates the picture, as this item cannot simply be seen as an alternative form for have to, but, rather, has to be seen as a further marked form in relation to must. Such an assignment of maxkedness is, however, not reflected unambiguously in details of use, including frequency, since, as was seen in 4 . 3 . 1 (cf (4.16)), must has a much narrower range of use than have (got) to in speech, and a wider range of use in writing.

5.

Other modal-periphrastic sets

5.1

Introduction

In this chapter the remaining groups of items listed in (2.110) will be discussed. The main conclusion in the last chapter - that the most appropriate way of distinguishing between a modal and a periphrastic is by treating the modal as maximally unmarked, or general in application, and determining the specializing factors that characterize the periphrastic - will be followed as a primary approach; at the same time, the importance that questions of style and variety may have in the use of these items, which was shown in many ways in 4, will be allowed for. Other distinguishing factors that have been proposed or that appear relevant will be considered where appropriate.

5.2

Should, ought to, be supposed to and had better

These four items will be dealt with in the following way. Since should -.in some of its uses - and ought to are standardly treated as semantically close if not synonymous, this pair will first be compared (5.2.1), and then the other two items will in turn be contrasted with should, generally ignoring possible distinctions in use between should and ought to (5.2.2-3). Conclusions for the set of items will follow (5.2.4).

5.2.1

Should and ought to

In comparing this pair, I shall first discuss some general issues involved in such a comparison (5.2.1.1); then I shall deal with what appears to be the essential distinction between the items, and, in the light of this, review various other points of difference that have been or can be observed (5.2.1.2). 5.2.1.1

General issues

In contrasting should and ought to, a basic descriptive problem arises: while it is generally agreed that these items have closely related uses (and, as such, are the nearest to a synonymous pair among all the items being discussed in this study), this applies to all the uses of ought to but only to certain uses of should. It must, therefore, be clarified exactly which uses of should are involved. Generally, deontic and epistemic uses of should are together singled out as being expressed equivalently by ought to, while other uses of should are treated quite separately,

163 as instances of polysemy, if not homonymy (eg, Quirk et al 1985:227,231,234-235). This approach involves making a sharp distinction between 'modal' and 'non-modal' uses of should, although the uses involved may appear rather close, with the possibility of semantic merging. For example, this kind of analysis will classify the uses of should in the following as distinct: (5.1)

You should try this ouzo.

(5.2)

He insists we should try this ouzo.

Should try in (5.1) is treated as deontic, with ought to as an alternative; in (5.2) it is treated as 'putative' (a use found with volitional verbs and adjectives), with plain try as an alternative to should try, but not ought to try. (In an alternative interpretation, should in (5.2) could be seen as deontic, just as in (5.1), in which case insist would either have a deontic sense, redundantly, or an epistemic sense, with the truth of the required action denoted in the proposition rather than its performance as its complement; in these cases, ought to could also be used.) So long as potential interchangeability between should and ought to is our sole concern, there need be no problem here for the purposes of this study. However, it has been suggested that a number of the 'non-modal' uses, as in the following: (5.3)

How should I know?

(5.4)

I don't see why he should think I'm joking.

are relatable to epistemic uses through the concepts of 'logicality' and 'expectation' (Gailor 1983) - possibly only leaving, among the non-modal uses, the use of should (alternatively would) in reported speech for an original shall. If the domain of 'modal' should is extended in this way, the proportion of the uses of modal - specifically epistemic - should that are potentially synonymous with ought to is substantially increased, and, from another perspective, the availability of ought to as an alternative to 'modal' should is correspondingly reduced. However, the implications of this line of argument will not be followed here. Regarding the deontic-epistemic distinction, although alternation between the two modal senses is here being taken as a typical feature of 'modal' items (cf 2 . 3 . 2 . 5 ) , it has sometimes been observed that epistemic uses are rare for both items, particularly so for ought to. The following attempts to show the figures in the CEC: (5.5) total occurrences deontic epistemic" subordinate 6 conditional 0 ° ' c

should

ought to

211

71

100 18 13 80

66 5

all doubtful cases which allow an epistemic sense, including apparent mergers in subordinate structures, as in it is reasonable (that) you should say that first person variants for would, as in / should go if I had time

164 Uses of these items are often difficult to classify, in marked contrast to those of must, etc, where non-clear-cut cases are rare. Coates discusses this problem in detail, distinguishing carefully between cases of ambiguity (allowing alternative readings) and merger (where two readings can or do coexist) (1983:77-80). A particularly good illustration of this issue, with ought to, occurs in the CEC: (5.6)

A

there was a beer called Newcastle Brown, and do you know, it costs a penny more than Guinness - it's a jolly good beer a is it? A well, it ought to be, at that price (CEC:179/286-291)

This could be seen as either deontic or epistemic without any difference in eifect. 1 It is also worth noting that Halliday, in his account of the two parallel systems of modality, gives the following, with ought to, as an extreme instance of the blending, or merging, of deontic and epistemic meaning: (5.7)

The home team ought to have won, if they'd had reasonable luck. (1970:344)

In the following discussion, deontic and epistemic uses of the items will on occasion be discussed separately, but no great importance is attached to the distinction between them. 5.2.1.2

Comparing the items

In comparing should and ought to, we must start with the possibility that the items are full synonyms. In addition to Quirk et al (1985), referred to above, many descriptive and pedagogical treatments suggest this, whether explicitly or implicitly (eg, Leech 1987:100; Sinclair 1990:passim). There are plenty of indications of interchangeability; in particular, the use of the alternate items in tag-questions (with should following ought to, if not vice versa), seems to be fairly common, eg: (5.8) (5.9)

He ought to come tomorrow, shouldn't he? (Palmer 1990:122) There ought to be a little window by there, shouldn't there? (Perkins 1983:55)

(Cf also Swan 1995:§496.) It is certainly striking that should and ought to are the only pair of semantically related items where this phenomenon occurs, since He's going to come tomorrow, won't he?, for example, is unnatural. The following texts point in the same direction: (5.10)

1

Β

it ought to be, I suppose, in those two, those double grey filing cabinets - do you remember? C yes, it should be in there (cited by Coates 1983:74; here reduced)

Significantly, this text is discussed by Coates - as a merger (1983:78), and is also referred to by Palmer three times - twice as an epistemic-dynamic merger and once as purely dynamic (1990:21,60,124).

165 (5.11)

. . . the view should be most glorious, as one ought to see the whole of the gulf and all the range of Olympus; but, alas! beyond the silvery minarets relieving the monotonous surface of roofs below and the delicately indented shore and the blue gulf, all else was blotted out . . . (E. Lear, Journal of a landscape painter in Greece and Albania. London: Century Hutchinson, 1988:23; originally 1851)

(5.12)

We ought to end this rapid survey of the ideas from which Byzantine art drew its inspiration by recalling a few secondary sources that should not be neglected. (A. Grabar, Byzantium: Byzantine art in the middle ages. London: Methuen, 1965:85)

(5.13)

She thought that she ought to go back to London, knew that she should, but was somehow incapable of summoning the energy to buy her ticket. (A. Brookner, A closed eye. Harmondsworth: Penguin, 1992:244)

(5.14)

Negro children ought to be able to take pride in their own background, and should not be burdened with shame . . . (R. Burling, Man's many voices: language in its cultural context. New York: Holt, Reinhart & Winston, 1970:130)

In all these cases, the close contiguity of the occurrences of the items and their apparent interchangeability suggest a very high degree of equivalence. In (5.11) there is in addition a parallel occurrence of the whole of (the gulf) and all (the range of Olympus), and in each case the alternation of the items (should-ought to, all-the whole o f ) seems to be determined purely stylistically. In (5.12) and (5.13) there is, similarly, no evident syntactic or semantic reason for the alternation of the items within the sentence; in (5.13) vowel quality ( t h o u g h t . . . ought, knew ... should) seems to determine the choice. Finally, (5.14) illustrates a point often claimed, and seemingly also supported by cases like (5.8) and (5.9), that negative forms (and, in general, complex forms) of ought to are avoided, so that should not has to serve as the negative form for both items, leading to the proposal that ought to is redundant - and hence, perhaps, moribund (for a variety of references, cf Nehls 1986:63, and the relevant discussion later in this section). Such claims, however, are not well justified for BE (though AE seems to differ here), since negative forms of ought to can readily be cited, as in the following: (5.15)

yes, because that ought to strengthen your hand, I would have thought, oughtn't it? (CEC:40/341-343)

(5.16)

what else was there? they raised very seriously the question of whether this oughtn't to be 128 or 192 pages in extent (CEC:801/1090-1096)

The reported regular loss of to with negative ought to in AE is mentioned in Footnote 48 in 3. This process is also apparent in BE: (5.17)

Writers of the discarded novels ought not feel aggrieved. ( The Economist, 4.2.1995:93).

166

There is, then, plenty of evidence suggesting semantic equivalence between the two items. However, there is also plenty of evidence of non-equivalence, as careful examination of the corpus-derived data in Sinclair (1990:here 236) strikingly shows. The following examples illustrate the use of the items for the function 'trying to help someone by advising them to do something'; should and ought to are presented as equivalent alternatives: (5.18)

Carbon knives should be wiped clean after use.

(5.19)

You should claim your pension 3-4 months before you retire.

(5.20)

You ought to try a different approach.

While should is clearly a possible alternative in (5.20), ought to could certainly appear odd in (5.18) and (5.19). Possible reasons for this will be discussed later; but first the general claim to be made here for the basic distinction between the items will be set out. It will be argued here that the essential difference between should and ought to lies in the fact that should is a maximally vague or unmarked item, as is typical of the pure modals, while ought to has a more specific meaning, as follows: should simply indicates that the fulfilment of some proposition can reasonably be expected, whereas ought to suggests that there is some good reason for this expectation. The proposal that ought to has a more specialized meaning, together with the fact that it is much less frequent in occurrence 2 , suggests that it can be treated as a marked item in relation to should. I shall now, in the light of this claim, review in turn the major other points of contrast between the items that are evident or have been proposed. The following topics will be discussed in turn: (a) subjectivity and objectivity, (b) the association of ought to with the expression of duties, (c) relative strength, (d) stress, (e) implication of non-fulfilment, (f) style and variety distinctions, and (g) other relations within the modal system. In each case it will be shown how the claims or observations can be related to the operation of pragmatic inference. Finally, a conclusion will summarize these points and touch on a few other aspects of the use of these items. As can be seen, this discussion will return to a number of themes that came up in 4. Subjectivity and objectivity. I shall start by considering the implications of the account of these items by Swan (1980:§550). Swan presents these examples: (5.21)

You should go and see Mary some time.

(+)3

(5.22)

You ought to go and see Mary some time.

(+)

(5.23)

We ought to go and see Mary tomorrow, but I don't think we will.

(+)

(5.24)

We should go and see Mary tomorrow, but I don't think we will.

(—)

Swan comments: (5.25)

When we use should, we give our own subjective opinion; ought to has a rather more objective force, and is used when we are talking about laws,

It is sufficient here to quote the figures for the occurrences of, respectively, should and ought to in Coates (1983:23): 6 1 2 - 1 4 0 (spoken material) and 1285-105 (written material); and in the C E C (cf (5.5) above): 2 1 1 - 7 1 . 3

These 'relative normality' signs have, as usual, been added.

167 duties and regulations (or when we want to make our opinion sound as strong as a duty or law). If we follow the discussion of must and have (got) to in 4, it may be claimed that should has no specific 'subjective' value, but rather conveys the normal speaker's commitment to the content of a proposition, and as a default interpretation can imply that the speaker is expressing his own wish, advice, etc This means that, in the first pair, (5.21), as it stands, is interpreted simply as the speaker's advice, while (5.22) implies that there is some reason for the action - for example, the speaker sees the visit as a moral obligation or feels that it may avert some undesirable consequence; but in a specific context, just as with must/have (got) to, there may be little or no difference between the use of the two items. In the case of the second pair, Swan comments that (5.24) sounds odd as "it would be strange to give oneself advice and say that one was not going to follow it". This may be seen as arising directly from the default interpretation for should: if (5.24) was introduced by Peter told us ... the sentence would not be odd, as the advice now clearly arises outside the speaker. Further, if should was stressed - particularly with a fall-rise nucleus (which is standardly associated with contrast of some kind) - it would function much like ought to, since it would now imply a possible contrast between an obligation and its fulfilment (this reflects the fact that ought to is normally stressed, but should is not; see comments below). This suggests, then, that the comments on (5.23) and (5.24) simply reflect standard pragmatic interpretations, and that the appearance of 'subjectivity' and 'objectivity' arises from this. (It should be mentioned that the discussion of should and ought to quoted here is missing in Swan 1995.) The question of implied non-fulfilment in relation to ought to will be returned to below. Ought to and 'duty'. It is sometimes stated that ought to is closely associated with the expression of duty in a way not typical of should (eg, Aarts L· Wekker 1987:193; Chalker 1990:291); similarly, Gailor (1983:348) states "we tend to avoid ought to when we want to be careful not to bring in an idea of duty". This feature of ought to can then be related to its strong association with deontic meaning (cf figures in (5.5)). Evidence for this might be found in the following pair: (5.26)

The weather should change later in the day.

(5.27)

The weather ought to change later in the day.

Here the first would normally be interpreted epistemically, while the second suggests something like a moral obligation, and hence a deontic interpretation - although if this seemed odd an epistemic sense would also be possible. This distinction becomes more evident if The weather forecast says . . . is preposed, when an epistemic sense is clearly appropriate and (5.26) becomes much more natural than (5.27). A slightly different, but equally important, contrast between the use of the two items in relation to duties is provided by the following: (5.28)

If you've got the flu, you should stay in bed.

(5.29)

If you've got the flu, you ought to stay in bed.

168 Here it is claimed that "the first sentence suggests what would be best for the addressee to do, while the second sentence indicates what the addressee's duty is, and what would be best for anyone" (Aarts & Wekker 1987:193) - this also suggests a degree of objectivity for ought to. Nevertheless, despite this evidence for the association of deontic ought to with duty, such a sense is by no means obligatory, as was shown in (5.20), where helpful advice seemed to be involved, while epistemic interpretations without any suggestion of duty are perfectly natural, as the following corpus text shows: (5.30)

a A

t h a t ' d be the, would it be the viceroy then - yeah, yes well, it ought to be the viceroy, but I'm sure that's not what they

called him (CEC:693/383-385) The assumption regarding the title of the official here has no moral overtones, and simply concerns reasonable expectation. This sense of 'duty', then, may be seen simply as a typical interpretation of ought to, relating naturally to deontic use and the fact that this item suggests an external requirement rather than a vaguely characterized expectation, as with should. Relative strength. Another claim sometimes made - primarily, though not exclusively, in relation to deontic meaning - is that ought to is per se 'stronger' than should (eg, Zandvoort 1969:70; Perkins' treatment of ought to as the conditional counterpart of must, to be discussed below, entails this; cf, also, Matthews 1993:70-72). This claim can be illustrated from some of the examples quoted above: in each of the pairs (5.21)-(5.22) and (5.23)-(5.24) the sentence with ought to can be seen as stronger. Similarly, in the following text should would seem too weak, suggesting that ought to is inherently stronger: (5.31)

Whenever he got a chance, Malek broke into a canter, and one of those bursts turned into a long twilight gallop; he may have thought we were far from home and ought to get a move on . . . (P. Leigh Fermor, Between the woods and the water. London: Murray, 1986:46)

This impression of strength is clearly relatable to the fact, to be discussed next, that ought to, but not should, standardly receives stress; and it also relates to the other themes being discussed in this section. It might, therefore, be assumed that questions of relative strength between the items can safely be subsumed under other considerations, just as was done in the case of the various proposals for relative strength attaching to must and related items in 4.8.2.2. There is, however, the complication that, when stressed, should tends to have some special sense, so that if used in (5.31), for example, it could well imply expected nonfulfilment, which is not the case with ought to, as the text stands. Stress. It was mentioned above in connection with the interpretation of (5.23) and (5.24) that ought to typically carries stress, but should does not. Coates' data from her spoken material show this very clearly: with ought to, 80% of all occurrences, and 11 of the

169

12 epistemic cases, have stress of some kind, while only 28% of a sample total of 200 occurrences of should do so (1983:75-76). In fact, it seems that ought to is standardly associated with some kind of stress. Thus, if we examine a rare instance of unstressed ought to in the CEC (with simplified prosodie marking): (5.32)

'yes, because 'that ought to 'strengthen your 'hand, I would have 'thought, 'oughtn't it? (CEC:40/341-343)

it could be argued that the actual modality is stressed by the phrase I would have Hhought, and reinforced by the tag-question with nuclear stress on ought. It seems that it is only the heavy stress on that that allows the following item not to be stressed; an unemphasized pronoun subject like it or he would necessarily have the following verb stressed. EM is in any case generally associated with stress, and this is reflected in Coates' finding that there is a high correlation between modals that normally carry stress - must, ought to, may and might - and those that have "quantitatively significant epistemic usage" (1983:243-244). Should is an exception for Coates, in having high epistemic frequency but not normally carrying stress. 4 If we examine the group of four items listed, another factor appears to be of relevance: ought to, like may and might, contains a vowel that does not normally reduce. Ought to can reduce to /o:ta/ (cf (2.101) and discussion in 2.3.3), but /d:/ does not standardly reduce in English. In a similar way, may cannot reduce its /el/: the only two items in English that allow reduction of this vowel are unstressed a and Saint in proper names. Might, finally, cannot reduce its / a i / : the only possible exception to this non-reducibility seems to be my, which has very restricted possibilities for reducing its vowel to / i / . 5 Must, the only item among the four that has a regular weak form ( / m a s t / ) , differs from the other three in following a regular phonological reduction pattern ( / λ / to / a / : cf but and some). It should also be noted that a reduction of the vowel in ought to could produce / a t a / , which would strongly suggest unstressed are to. These phonological constraints suggest the following semantic consequences for the items ought to and should: (a) ought to, by virtue of its non-reducible first vowel, naturally carries some stress, and this is relatable to semantic weight; (b) should, since it has regular strong form-weak form alternation, is specially emphasized when it is stressed, and then is close in effect to ought to. This can be illustrated by consideration of some of the examples seen earlier. Firstly, the close equivalence of stressed should and 'normal' ought to was suggested in connec4

This claim concerning should can be questioned, for the following reasons: (a) the figure for its frequency in epistemic usage is the lowest among the items listed; (b) it includes the 'hypothetical' use of should as a first person alternative to would, itself including the hedge I should think (Coates 1983:221-223) - uses which have no 'epistemic' parallels in the other items discussed by Coates, and thus might be assumed to increase the epistemic figure artificially; and (c) the actual incidence of stress on epistemic uses of should seems to be high (at least 13 of the 18 occurrences in the CEC carry stress). This would suggest that should may not, in fact, be an exception to the claimed correlation of stress and high frequency of epistemic usage.

5

This is not included by Gimson among standard unaccented weak forms (1980:255-264); but cf entries in Jones (1977:334) and Wells (1990:467).

170 tion with (5.23) and (5.24). Secondly, the incidence of stress and its consequences can be shown in relation to the first parts of (5.26) and (5.27): The weather should change normally has two sentence stresses, and only exceptionally three, while The weather ought to change virtually must have three. The effect of the stress on the modal items is to add emphasis to the general sense of what is advisable or expected, and this may manifest itself in various ways - for example, by emphasising a moral or well-founded expectation, where fulfilment is assumed, or by focusing on the possibility or actuality of non-fulfilment (the next topic). Implication of non-fulfilment. It is sometimes stated that ought to indicates that the fulfilment or realization of the proposition is less likely than is the case with should. Thus, Close refers to ought to suggesting an obligation "the fulfilment of which is overdue or may be delayed" (1981:121); rather similarly, Gailor claims that should implies confidence about achieving a result, while ought to reduces that confidence and tends to imply that an obligation will not be carried out (1983:348-349). The use of ought to for an obligation that the speaker feels but is prepared not to fulfil was illustrated in (5.23), and the same effect is seen in the following: (5.33)

"What about popping down to the George for a couple in a few minutes? »

"I ought to unpack and so on," Patrick Standish said. "You can do that tomorrow." "All right then." (K. Amis, Take a girl like you. Harmondsworth: Penguin, 1962:18; cited by Nehls 1986:60) In both texts the use of should would only be natural if it was stressed. This can be further illustrated by considering (5.18): the use of ought to here could readily suggest that there was some problem about the proposition being fulfilled (that is, the knives being cleaned), an implication absent with should (this text will be returned to from another perspective in the next section). However, this suggestion of non-fulfilment with ought to is not necessary, or, on occasion, relevant, as texts (5.6), (5.27) and (5.31) show. Should does not normally imply non-fulfilment, and a typical contrast between the items in this respect can be shown by considering the difference in interpersonal terms between As you should know ... and As you ought to know . . . as openers when a speaker is giving information or advice. The use of ought to here, unlike that of should, is likely to seem almost aggressive, and this is because with ought to, but not with should, the speaker naturally suggests that his interlocutor may not, in fact, know. Nevertheless, with stress, should can readily be associated with non-fulfilment or counterfactuality, as in the following corpus text: (5.34)

α Β a Β

what was the climate like there at Christmas? oh, it was fantastic, it was gloriously sunny, and about 68 degrees was it really? hm, it shouldn't be, it should be kind of

171 a yeah Β but it was quite exceptional (CEC:402/1334-1343) Here shouldn't and should are both stressed in the original text, a typical condition for ought ίο-like uses encountered several times above. All the examples discussed have involved present or future time reference. In the case of past time reference, the use of both items naturally - but not necessarily - suggests nonfulfilled actions. This is very clearly shown in the four occurrences of ought to have . . . , together with one ought to with present / f u t u r e reference, in the following (the suggestion of moral obligation is also involved): (5.35)

"That is very satisfying for you," said Rachel. "We are often at variance with people who tell us what we ought to have been. Percy cannot agree that he ought to have been a bachelor, and ought to have had a son. Percy would never agree to both. And Polly may not agree that she ought not to have been born." "Well, now, I think we ought to be getting on with our work," said Agatha, spacing her words. (I. Compton-Burnett, Men and wives. London: Allison and Busby, 1984:269-270; originally 1931).

In view of such data, it may be assumed that the suggestion of likely non-fulfilment is pragmatically inferred in cases where non-fulfilment is a reasonable possibility, and this phenomenon may appear to relate closely to ought to since that item, unlike should, is normally stressed and thus suggests some kind of emphasis, which may be positive or, as here, negative in orientation. 6 Style and variety. The likelihood that style and variety may be important in the choice between modal and periphrastic was discussed in general terms in 3.7, and two specific claims relevant to the present discussion were set out there (Celce-Murcia & LarsenFreeman 1983:87,89), concerning, first, a general distinction between should/ought to and be to/be supposed to in terms of formality vs informality (to be taken up in 5.2.2), and second, a specific contrast between should and ought to, in their deontic uses, also in terms of formality vs informality. This second claim presented ought to (in the colloquial form oughta) as an informal variant of should, alongside other pairs such as must-hafia. The tendency to develop such forms (cf wanna, gotta) is much more marked in AE (to which the relevant account refers) than BE; nevertheless, corpus counts show that ought to is much more typical of spoken, especially conversational, BE than of written BE, and also commoner in BE than in AE, at least in writing. The relevant figures are as follows: Intonation is clearly an important factor in determining positive or negative orientation. Halliday's account of modality in English (1970) is highly suggestive here: in his two parallel systems, corresponding to DM and EM, the relevant category has only a tentative 'undertone' and an assertive 'overtone', but lacks a neutral value. The overtone is explicitly assigned fall-rise intonation, which is associated with uncertainty or unresolved situations; the undertone may be expected to be similarly expressed. In fact, fall-rise intonation is predominant with both should, when stressed, and ought to, generally, in the CEC, and clear-cut falls Me rare.

172 (5.36)

occurrences of ought to per million words (real or extrapolated figures) corpus Brown corpus (AE, written) Lancaster (BE, written) SEU (BE, spoken7) CEC (BE, conversation)

number 68-70 105 257 323

source Kucera & Francis (1967) Coates (1988:23) Coates (1983:23) my count8

Coates suggests that the relatively high figures for spoken language may be connected with the potentiality of the item for carrying stress, which can hardly be represented in writing (1983:70). These data clearly support the claim that ought to is, or may be, informal in comparison with should, though the overall picture regarding the use of these items, including possible A E - B E contrasts, may be rather complex (since, for example, oughtn't (to) is sometimes said not to occur in AE). We can now return to the question why (5.18), and perhaps (5.19), might be odd with ought to replacing should. The first text appears to come from written instructions concerning the use of the tools in question (a speaker would say something like You should/ought to wipe carbon steel knives clean after you've used them). The second could come from user-friendly official documents or, less likely, from informal speech; if the former, ought to would again be inappropriate, but if the latter, it would be possible. These observations can clearly be related to the claim that ought to is less typical of writing than of speech. A restriction of this kind has rarely been noted, apart from the general claims by Celce-Murcia & Larsen-Freeman (1983) referred to at the beginning of this section; however, Close (1981:120-122) observes that in an official letter such as: (5.37)

Applications should be submitted by March 31st at the latest.

ought to cannot replace should. In this connection, Gailor treats the following: (5.38)

Entries should be clearly typed.

as showing that the writer is confident he can require clear typing, and relates it to the common implication of fulfilment attaching to should but not ought to (1983:348); but this could equally well be seen as illustrating the phenomenon being discussed here. Further illustration can be found in printed public notices, such as: (5.39)

This coffee-machine should be turned off after use.

Here, too, ought to would be impossible. Other relations to these items within the modal system. The fact that, at least in reported speech, should is relatable to shall is a complication here. Perkins has extended this relation between the two items to cover all the uses of shall and should, so that should is given a unitary semantic structure as the conditional form of shall. In this way, should is presented as parallel to could, might and would; ought to, the one remaining 'conditional' item (assuming that it is counted among the modals), is then aligned with must (1983:5056). This analysis is justified as follows: 7 8

26 of the 109 'spoken' texts - c 24% - are in fact written material (cf data set out in Coates 1983:2). Cf Footnote 11 in 4 on the basis for this figure.

173 (5.40)

Should... is fairly obviously the conditional counterpart of shall, and ought to ... , although etymologically distinct, is to all intents and purposes a conditional counterpart of must... (1983:54)

This claim is certainly not self-evident, and is not supported by evidence. A text like the following might possibly be found to offer support: (5.41)

. . . he and Nick had travelled rather too fax too quickly. A radical ought to believe, must believe it possible to convert a nation within a generation (M. Drabble, The needle's eye. Harmondsworth: Penguin, 1974:31)

If ought to here more naturally precedes must than should would, this might constitute a kind of evidence that ought to and must axe semantically related; however, it is not clearly the case that should would be unnatural here before must. On the other hand, since Perkins subsequently admits that the claimed difference between the items that underlies (5.40) is "more apparent than real" (whatever that may mean) (1983:56), perhaps the matter should not be taken too seriously, and the items simply treated independently without regard for such claimed relations. Matthews (1993:70-72) rather similarly relates ought to to must/have got to and should to will (of futurity); but convincing evidence is, again, lacking. Conclusion and further remarks. A number of overtly separate factors involved in distinguishing between should and ought to have been discussed here, but it has become clear that these are Very closely related. It was suggested at the outset that ought to is distinguished from should in rather strongly suggesting the existence of some obligation or motivation - this having some parallelism to the basic distinction between must and have (got) to, and this factor can be associated with the degree of emphasis typically attached to it and is reflected in the regular incidence of phonological stress. It was further assumed that the particular uses of ought to for moral duties and likely nonfulfilment are natural contextually determined senses, derivable by pragmatic inferencing. Should is a relatively 'neutral' item, with a wide, rather vague semantic spread; and the subjectivity claimed for it can be seen as a default interpretation. The question of differences between the items in terms of style and variety, which was shown to be of considerable significance, is not so easily associated with these other factors; but here, again, there is a certain parallel with must-have (got) to. The interaction of these various factors is well illustrated in a case like (5.39), where the neutral item should is required, and the use of ought to would be effectively ruled out equally in terms of style (a formal, written style is involved), and of possible suggestions of moral duty or of non-fulfilment. It has sometimes been proposed that ought to is a moribund member, whether central or peripheral (depending on definitions), of the English modal or auxiliary system, and, since it is semantically redundant, can thus be safely ignored. Data of various kinds have been produced, including the markedly low occurrence of ought to, and particularly of its

174 more complex forms, including oughtn't to, and the sheer difficulty of showing any clearcut semantic distinction between it and should. The dilemma is well expressed in Visser's comment that the continued co-existence of the two items "would seem to justify the conclusion that they never had exactly the same meaning, however closely synonymous they always have been" (1969:para.l530). 9 The argumentation typically used to demonstrate the claimed moribundity (cf reference to ought to ss Ά 'decaying modal' in 2, Footnote 9) is reviewed and criticized by Westney (1980:43-45); and the highly revealing comment by Ehrman (1966:66) there cited, that the 'moribundity' of ought to suggested by her data was not reflected by her and her associates' informal speech (in AE), has subsequently been given further illustration for BE in the evidence of the CEC (Svartvik & Quirk 1980) and the corpus-based account by Coates (1983). However, the claim for moribundity persists: for example, Harris (1986:356), after reviewing the situation, concludes, primarily on the basis of informell observation and intuition, "ought is indeed on its last legs". In view of the figures for the occurrence of ought to given in (5.5) and (5.36), which give no support for such a claim, one can only assume that investigations of these items have often been unreasonably dominated by written language, and perhaps by wishful thinking. Both these items are in origin past tense forms. Should is morphologically a past form of shall, though this relation is only fully evident in reported speech. Ought to is rather like must in being an isolated former past tense form which, since it can have no past tense counterpart, may have to serve for past time reference. In 4.5.3, evidence was given of the extent to which must can serve as a past time form. Indication that ought to is a similar case is rare. Jespersen provides some possible lingering evidence (1931:122,128); while a text such as the following deserves consideration: (5.42)

I was sixteen and I ought to be looking for a job. (A. Burgess, Little Wilson and big God. London: Heinemann, 1987:125)

A past tense is clearly required here (a paraphrase of ought to would be, eg, I was supposed to or I was obliged to). Should seems unnatural in its place, no doubt because it tends to carry a weak, rather vague sense, and a stronger sense of 'duty' is required. A final text will illustrate the use of the two items in a written context; this will show how they can be used in ongoing discourse (specifically in writing, as speech adds a prosodie dimension), and how in each case the use of the other item may have a quite different effect (occurrences of the items are marked in italics): (5.43)

Part 4 ought to be read in its entirety by all postgraduate and many other EFL teachers in training. Almost any experienced language-teacher, guided by the section headings and the index, should find the book, if read with some care, both eye-opening and thought-provoking. Historians ought also to be interested. (W.R. Lee, review of A.P.R. Howatt, A History of English language teaching. World Language English 4, 1984:199; my italics)

This assumption - that two distinct forms in a language cannot 'mean the same'- receives its most extended demonstration in Bolinger (1977).

175 The first occurrence (ought to) offers a specific piece of advice, unambiguously deontic (what would be good for these teachers) - should would simply be much weaker or vaguer in effect. The second (should) is just as clearly epistemic - this is a statement of what the author expects will happen, and there is no suggestion of advice, as there would be with ought to. The third (ought to) is ambiguous - or rather, perhaps, a merger between deontic and epistemic senses; the effect of this use is stronger than that of the preceding should, but, at the same time, this statement seems to allow non-fulfilment ie, it is reasonable to expect language teachers to be interested in the book in question, but historians may be too, and this difference would clearly be lost if should was used.

5.2.2

Be supposed to

Be supposed to will be compared with should, and, by extension, with ought to. It was shown in 5.2.1 that ought to can in a certain sense be considered 'objective' in relation to should·, be supposed to has similarly been found to be 'objective' in relation to should. However, it will be shown that the distinctions in use between these items axe of very different kinds. The justification for treating be supposed to as a, periphrastic was set out in 2, and in some respects the closeness between it and should is striking. Thus, in the following three corpus texts, be supposed to could be replaced by should/ought to without producing any serious anomalies: (5.44)

d how long do you anticipate staying on this particular thing, Frank? A well, I'm supposed to be doing a PhD, you see Β we heard that ten years ago (CEC:483/1053-1056)

(5.45)

α it's supposed to be good for you, isn't it? A yeah, Sail's mother always had a bottle every night (CEC:179/318-322)

(5.46)

Β this is all very sad - I feel bad about that A axe we, axe we supposed to know, in general? Β well, I don't know, I mean I suppose so (CEC: 123/1045-1054)

In (5.44) be supposed to allows deontic or epistemic interpretations, in (5.45) epistemic, and (5.46) seems to be a merger: deontic/epistemic classification is often difficult with this item, just as it is with should/ought to. Replacement of be supposed to in these texts by should (or ought to) would have the following effect: in (5.44) and (5.45) the salient sense becomes deontic, while in (5.46) it remains a merger. Be supposed to is thus, in some respects, semantically close to should. In the following discussion, I shall first consider how be supposed to can be given a unitaxy characterization, with particular reference to the discussion of should in 5.2.1 (5.2.2.1). Then I shall examine the nature of specific differences in use between the

176 items; in the course of this discussion, the various proposals that have been made in the literature will be taken into account (5.2.2.2). 5.2.2.1

Be supposed to compared with should/ought

to

A good example of a dictionary definition of be supposed to singles out the following uses (Sinclair 1987; simplified): (5.47)

(a) (b) (c)

something should be done, because of a law, rule or custom; something is planned or intended, especially when it does not occur; something is thought to be so.

Of these meanings, (a) and (b) are deontic, and (c) is epistemic. Meaning (a) can be illustrated by the following example from Swan (1980:§586), offered as an instance of "what people have to do according to the rules or the law": (5.48)

Catholics are supposed to go to church on Sundays.

(There is, of course, a less salient epistemic interpretation 'are believed to . . . '. Here, therefore, we may assume ambiguity.) Meaning (b) is illustrated by examples (5.44) and (5.46); non-fulfilment of the requirement, referred to in the definition, is explicit in the first and likely in the second. Finally, meaning (c) is shown in (5.45). It is proposed here that be supposed to basically denotes what is believed or expected in relation to the subject: the belief/expectation may relate to upcoming actions (deontic) or states of affairs (epistemic), but these uses can be treated as conventionalized specializations. The fact that this item has a passive form is directly reflected in its uses - it is significant that all three meanings given in (5.47) are in passive forms. It differs from most modals and periphrastics in that the epistemic sense seems to have been the primary meaning, with the deontic sense a derivative; this is implied in the relevant entry in the OED (1989:Vol.XVII, 263), where the expression be supposed to is given as a form of suppose that (5.49)

expresses the fact of the subject being credited with some action or quality; now especially = to be expected, intended, meant.

The following example from the OED (19th century) shows be supposed to being used in a way similar to that in (5.44), thus suggesting that merging between deontic and epistemic senses is well established: (5.50)

Still less does it seem consistent with that impartiality which every reviewer is supposed to have.

The problem of establishing a basic meaning for this item is well illustrated in Perkins' analysis. Be supposed to, alongside be expected to, is given a core meaning with a semantic frame that equates it with epistemic may, together with a 'conventional' deontic use that is close to the frame for must. The resulting ambiguity is illustrated in the following:

177 (5.51)

Every soldier is supposed to fight to the death.

Here the epistemic sense could be followed by at least, that's what they say happens, and the deontic sense by in order to defend his country (Perkins 1983:85). However, it could be objected that this analysis of the epistemic sense as a kind of 'possibility' is too weak, while that of the deontic sense, if treated as a norm (it is, after all, a semantic frame) is too strong, suggesting a certain inflexibility in this general approach (cf 3 . 5 . 1 ) , and illustrating clearly the difficulty of dealing with an item such as be supposed to where pragmatic extensions result in an apparently very wide range of meanings. This issue will be further illustrated below in connection with the 'not allowed' meaning. 5.2.2.2

Specific issues

In the following sections I shall discuss some major consequences of the characterization of be supposed to outlined here for the distinction between the use of that item and that of should. Subjectivity and objectivity. Be supposed to can clearly be considered objective in that the expectation arises outside the subject; and this is apparent in the use of the item in (5.44) and (5.46), where there are first person subjects. I'm supposed to be doing and are we supposed to know refer to things expected of the subjects by other, external agents; in the same contexts, the use of should would readily allow first person initiation as the default interpretation, while that of ought to would suggest some kind of external requirement, but not exclude the speaker's own point of view. This distinction can be illustrated further by variations on a text quoted earlier: (5.39)

This coffee-machine should be turned off after use.

As this stands, the source for the requirement is simply left vague. By contrast, the use of be supposed to, as in: (5.52)

This coffee-machine is supposed to be turned off after use.

would create some distance between the speaker/writer and the regulation, and could naturally be followed by . . . but I don't mind what you do (for a deontic reading) or ... but I don't think it usually is, actually (for an epistemic reading). In this sense, be supposed to is markedly more 'objective' than ought to. For the more standard claim for a subjective/objective opposition in the use of these items, we must return to Lakoff's general argument as set out in (1.2). According to this, while speaker involvement can be expressed by either the modal or its related periphrastic, external orientation in the form of lack of speaker involvement can only be expressed by the periphrastic item. The following examples are used for the should-be supposed to pair: (5.53)

We all should vote for Agnew.

(5.54)

We all are supposed to vote for Agnew.

178 In (5.53), the speaker is said to go along with the idea ("he thinks it's a good idea himself"); while in (5.54) he is reporting the suggestion and may or may not approve of it (1972:240). Palmer (1990:81) accepts this argument (but he does not deal with be supposed to in his account). However, this claim seems to be too strong. It was shown in 5.2.1 that speaker support for a should proposal is a common but not necessary implication: it is appropriately seen as a default interpretation of normal speaker commitment to the validity of one's utterances. By contrast, be supposed to is clearly used to report what is believed or expected by an external agent, and here the speaker's role is not important in the same way. Lakoff's distinction between should and be supposed to is, in fact, rejected by Bouma (cf discussion in 3.3.5), who argues that a sentence like (5.53), contrary to Lakoff's claims, does not require the speaker's support, since the following is possible: (3.94)

You should vote for Agnew, but I hope you don't.

Bouma proposes that the use of should in such a case focuses on the actor's moral commitment (regardless of its 'source'), while the corresponding use of be supposed to, as in (3.95), would focus on an external directive; the parallel situation with must and have to was discussed in 4.4.2.1. These comments seem fully justified, and the present discussion is substantially in accord with them, as far as they go. The final proposal to be discussed in this section - although it does not make a direct subjectivity-objectivity claim - concerns the distinction drawn by Edmondson et al (1977: 289-290), who argue that while should and ought to denote a weak obligation that arises from some unspecified source, with the speaker either imposing or recreating it at the time of speaking (1977:262-263), the use of be supposed to denotes a weak obligation that was previously imposed, by either the speaker or a third party. In this way should/ought to count as 'performance' auxiliaries, and be supposed to as a. 'statement' auxiliary, which makes the group exactly parallel to must-have (got) to (cf discussion of the proposals for those items in 4.4.1.3). In the present case, the discussion in this section has suggested that the claim that should/ought to imposes/recreates an obligation at the time of speaking, regardless of its source, is simply too strong. On the other hand, the notion that be supposed to reports an existing obligation that may have arisen from the speaker or an outside source is justified, but it seems that an outside source is strongly suggested, so that, for example, the use of the item for an obligation the speaker himself has made is possible but conversationally misleading. Implication of non-fulfilment. The likelihood that the proposition will not be realized is often commented on (eg, (5.47)(b)). This phenomenon is clearer than was the case with ought to, discussed in this connection in 5.2.1.2, and may be seen as a conventional assumption based on conversational principles, perhaps reflecting a tendency to view human activity negatively. 10 It can be illustrated by comparing two texts seen above. Firstly, the statement in (5.48), in its salient deontic sense, is, in the absence of further contextual 10

Poutsma's comment, in another connection, that a tendency towards 'negative colouring' "has nothing to do with the essential meaning of the idiom", but is "simply due to the fact that people, as a rule, are more given to dispraise than to praise" (1921:83), is pertinent here.

179 information, neutral as to realization. However, in a context where there are contrary indications, or where fall-rise intonation is used, non-realization would readily be implied. Secondly, in (5.44) A's own words and ß ' s response make it clear that be supposed to refers to a non-realized or counterfactual statement. The frequency of cases such as these, where the likely non-fulfilment if not the counterfactuality of the proposition is involved, lies behind the kind of statement seen in (5.47)(b). Negative judgements. A significant distinction in use between be supposed to and should is shown in the following: (5.55)

That book should be quite cheap.

(5.56)

That book is supposed to be quite cheap.

(5.57) (5.58)

?That book should be quite expensive. That book is supposed to be quite expensive.

Should would normally be odd in (5.57), since expensive prices are viewed negatively, and it inherently denotes what one can reasonably expect, as opposed to what is simply expected or believed: hence the comments in 5.2.1.2 about a sense of duty attaching to, in particular, ought to. Of course, in a context where expense is viewed positively, there is nothing odd about (5.59)

A decent St Emilion should/ought to be pretty expensive.

(although the use of should/ought to here may be seen as a merger between deontic and epistemic senses, as in text (5.6)). By contrast, be supposed to merely denotes what is expected to be the case. It may be noted that while (5.55) and (5.57), with should, and (5.56), with be supposed to, can be treated as deontic-epistemic mergers, (5.58) is more naturally interpreted epistemically, suggesting that a deontic sense naturally attracts notions of appropriateness, including the universal desirability of costs being low. It thus seems that while be supposed to remains neutral in this respect in its epistemic sense (which was its earlier sense), should carries over a sense of desirability to its epistemic sense (which was presumably a later sense). The 'not allowed' meaning. A particular problem arises in accounts of be supposed to that introduce the meaning, specific to negative contexts, of 'not allowed/permitted'. For example, the OALD (Hornby 1989) gives two meanings for the item: (i) "be expected or required . . . " (cf (5.47)(a) and (b)), and (ii) "(used in negative sentences) be allowed . . . ", as in (5.60)

You're not supposed to play football in the classroom.

The use of be supposed to here might seem similar to that of should, and replacement by it seems perfectly natural; on the other hand, (5.60) is very naturally interpreted as a prohibition, which is what the definition indicates, a pragmatic extension not normal with should. The fact that be supposed to can - at least in negative contexts - be interpreted as close in sense either to should or must is clear from a consideration of the following texts provided by Chalker (1990:360):

180 (5.61)

We were supposed to do homework every night (but we didn't).

(5.62)

You're not supposed to smoke on the London underground, but some people still do.

Both these uses of· be supposed to are similar to should in allowing non-fulfilment of the requirement. However, if a paraphrase of the first clause in (5.62) were uttered by a railway official, as in: (5.63)

You're not supposed to smoke here.

this could well be interpreted as a prohibition (an equivalent formulation in formal writing could be Smoking is not permitted). This extension of meaning is not, in fact, peculiar to negative contexts, since, in the same context: (5.64)

You're supposed to keep your ticket with you until the end of your journey.

could be a statement of a legally based requirement (and here an equivalent in formal writing could be Passengers are required to retain their tickets ... ). It seems clear that the use of negative be supposed to, usually with negative raising, in a 'not permitted' sense is a particular conventionally established use, which is derived from the basic sense of 'what is expected'; and it suggests that the 'expectation' that this items signals is rather stronger than with should. Past tense uses. The restriction to be discussed in this section arises not from pragmatic interpretation but from the defective morphology of the modals. One instance of a past tense use of be supposed to, for a past expectation (as opposed to forms like is supposed to have done, which refer to present expectations regarding past time, and are close in sense to should/ought to have done) occurred in (5.61); there is another in the following from the corpus: (5.65)

what was one supposed to do, if one, a, does want one? (CEC:106/80-81)

In both cases, explicit past time reference would be impossible with should/ought to, although the sense would justify their use in a present time context, and one may assume that be supposed to is available for such suppletive functions, most obviously in reported speech. In the following text, the form of be supposed to is past in the time both of the expectation and of the action expected: (5.66)

d

is, isn't it because that's where the type was supposed to have originated? Β I think yes, I think probably yes, with the emphasis on supposed (CEC:666/817-819)

Here, again, the relevant sense could not be expressed by should/ought to in a past time context; in addition, these items might introduce a 'subjective' element into the discussion. It is difficult to prove a suppletive connection between should and be supposed to, but there is some evidence that eliciting a past time or reported form of, for example,

181 What should I do? can produce What were you supposed to do? or He asked what you were supposed to do. Style and variety. The final proposal to be discussed here was set out in 3.7 (cf (3.157)(3.159)), and concerns a general distinction between M - P pairs or sets, according to which periphrastics are more informal in use, particularly when phonologically reduced. This claim actually refers to should/ought to in contrast to is to/be supposed to, and it seems very doubtful whether these four items can be treated as semantically close enough to be pairs of formal-informal variants. Aside from this question, as far as should/ought to and be supposed to are concerned I know of no evidence of this alternation: thus, be supposed to certainly occurs in informal conversational contexts in the CEC, but there is no indication that these might correspond to should/ought to in more formal contexts. There may be some distinction at work here between AE (which the description cited in 3 . 7 refers to) and BE; thus, reduced forms such as /spausta/ are rather more typical of AE than of BE (cf overview in 2.3.3); but in the lack of further evidence, this issue cannot be pursued here.

5.2.3 5.2.3.1

Had better Had better compared with should/ought

to

Whereas be supposed to was shown to be in many respects semantically similar to should, this is not so obviously the case with had better. In this section, I shall first examine claims for similarity between the items, and then set out the case for a fairly radical difference. Then I shall discuss specific issues that arise in the description of had better (5.2.3.2). Had better is quite commonly compared to, or at least glossed by, deontic should/ought to, in the sense of 'advising' or offering 'strong recommendation' (eg, Quirk et al 1985:142; Leech 1987:104; Swan 1980:§275). (There seems to be no question of an epistemic use; but cf Footnote 12 below.) The first question will therefore be how far the items are interchangeable. In fact, contextualized examples of should and had better cannot normally be interchanged; to cite one of the few sources'of real data on this item 1 1 , none of the seven instances of had better in Sinclair (1987), presumably all from spoken sources, would be naturally replaced by should (two of these are given below in (5.74) and (5.75)). A rare exception occurs in the following made-up exchange (Altman 1986:85):

11

Had better does not occur at all in a sample of six texts from the CEC ( = c 39,000 words): this can only be put down to specialization of use - many CEC texts are discursive, and occasions for strong advice, close in effect to orders, appropriate for had better, are rare in such contexts, as was also seen in the case of second person must (cf 4.4.2.1). A small, valuable collection of naturally occurring texts with various forms of had better, from fiction and journalism, is provided by Jacobsson (1980).

182 (5.67)

Two students who are friends are talking. Student 1: Student 2: Student 1: Student 2: Student 1:

I'm really nervous. What's the matter? I'm not doing too well in my English class and if I don't get at least a 'B' on this exam, I'm gonna fail the course. Well then, I guess you study. I guess you're right. See ya later.

This is one of a set of three fill-in-the-blank items designed to test second language learners' command of had better. Non-native speakers in the investigation by Altman gave a variety of solutions; what concerns us here is the native speaker group: of the total of 17, eight gave should, and eight gave forms of had better (op cit:86), suggesting that in the given context the items are equally acceptable. I shall discuss below why both items may seem appropriate here, and show that this is, in fact, an exceptional case. First, however, I shall consider the notion that had better is fairly radically distinguishable from should. Had better, like have got to, is a markedly informal item, and the way it is typically used is, for our purposes, well described in the following: (5.68)

Had better is not quite the same as ought to or should. Ought to and should can be used to make general comments; had better gives advice or a warning about future action. It also has a strong suggestion that the action will be done. (Chalker 1990:67)

This comment points to distinctions in use between the items involving the nature of the advice being given, including the contrast between general and specific reference, and the expectation of fulfilment suggested by had better. These topics, which interrelate, will now be discussed in turn. Cumulatively this discussion will suggest that had better has rather little in common with should/ought to. 5.2.3.2

Specific issues

The nature of the advice. It was suggested in 5.2.1.2 that should basically indicates that the fulfilment of a proposition can reasonably be expected. Had better has a rather different function, as can be seen by comparing the following: (5.69)

You'd better ask him when he comes in. (Palmer 1990:82)

(5.70)

You should ask him when he comes in. (invented)

In (5.69) the speaker is making a recommendation for a specific occasion; it is implied that the advice is well founded, and thus that, for example, the speaker has good grounds for giving it, or the authority to do so. Should in (5.70), by contrast, makes a relatively vague recommendation, more naturally understood to have general than particular reference ( = 'any time he comes in' rather than 'next time').

183

Had better is normally only used with a plain infinitive, for future reference. Other past reference forms seem to have been commoner formerly, as in the following text published in 1907 and cited by Jespersen: (5.71)

"How long have you been working here? Perhaps I had better have said 'idling'." (1931:146)

Comparable uses are still found, as in: (5.72)

"Lied!," shouted a relative, "you bloody well had better lied!" (The Observer, 3.5.1981:17)

This is presumably a non-standard form of you had better have lied. The distinction between had better have and should have seems to reflect that between had better and should referred to above, in that the former makes a very specific, threatening remark. One common implication often commented on is indicated in Chalker's reference in (5.68) to "advice or a warning" : in Palmer's words, the speaker "is fairly firm about his advice with the implication that unpleasant consequences may follow if it is not taken" (1990:82). The following shows this aspect of the use of had better explicitly: (5.73)

It's as though I know I'd better believe them - that, when I'm with him, it's safer to believe them. (M. Drabble, The radiant way. London: Penguin, 1988:255)

The chosen action is seen as safer. This is certainly a common, but is not a necessary, implication. For example, in examples from Sinclair (1987): (5.74)

I'd better introduce myself . . .

(5.75)

Better put it over there . . .

although further context is not provided, there is no reason to assume that some kind of threat or warning is involved. Thus, in (5.74) a speaker could well be trying to clarify a social situation, while (5.75) could be equivalent to Put it over there, would you?, etc (Of course, other interpretations are possible: for example, (5.74) could imply ... if you don't want me to call the police.) Accounts, therefore, that treat the . . . 'or else!' aspect of the use of had better as central to the meaning of the item do not seem to be on sure ground: for example, Perkins treats the determining circumstances for the fulfilment of the modality as such that non-fulfilment "entails adverse consequences" (1983:64), while Edmondson et al (1977:297) treat the 'basic meaning' as involving a threat for the subject. The common inference that had better issues a warning as well as a piece of advice may be assumed to lie in the interpretation of better, which simply states that some action would be a better choice, but can very naturally suggest the existence of some worse alternative, with the result that when a context of use includes the possibility of some adverse consequence, the speaker is assumed to be warning against it. 12 Thus, if a speaker says: 12

It is claimed that had better originally had the epistemic sense "of something being better to do than something else", and that the deontic sense arose later (Nagle 1986:111). The gloss given seems perfectly reasonable, but it does not look epistemic at all.

184 (5.76)

You'd better take an umbrella with you.

as opposed to, for example, (5.77)

You should take an umbrella with you.

the existence of adverse weather indications, whether physically visible or reported, would naturally lead to the inference 'if I don't, I risk getting soaked'. A further question arises here: it is not clear what modal value had better has that is, whether it is 'intermediate', like should, or 'strong', like must. The general assumption, which has determined the present discussion, is that it is an intermediate value item; however, there is good reason for claiming that the usual distinction between intermediate and strong values is simply inadequate here. For example, while the use of should is typically concerned with general advice, had better is functionally close to an order, perhaps closer even than must: hence, while the use of should can be politely declined, that of had better cannot (see discussion below). Perkins' analysis of had better can serve as a further illustration of this problem (1983:63-64). The core meaning assigned is a conditional form of the analysis for have (got) to, apart from the condition stipulating 'adverse consequences' mentioned above. Since Perkins presents ought to (but not should) as the conditional form of must (cf 5.2.1.2), and distinguishes have (got) to from must in terms of its objectivity /externality, this suggests that had better has the value of an 'objective' form of ought to. This seems implausible, but well reflects the anomalous nature of the item. General vs specific reference. Examples such as (5.69), (5.74), (5.75) and (5.76) strongly suggest that had better is associated with specific reference; should is vague in this respect, but typically has general reference. Here the elicitation test quoted in (5.67) is relevant. This showed that either should or had better would, exceptionally, fit equally well in a context that seems to have general/non-specific reference (I guess you should/had better study). It is significant that the comment I guess you study lacks adverbial qualification such as regularly, three hours a day or as of now, and is, partly in consequence, vague in temporal reference, which would suggest the appropriacy of should·, on the other hand, the advice being given is urgent, and this justifies the possible use of had better in that urgent application to one's studies is similar to a unique effort. The natural association of had better (as opposed to should) with unique performance and notions of immediacy closely parallels the case of must (as opposed to have to) (cf, especially, 4.5.2). Expectation of fulfilment. A clear distinction between had better and should can be observed in connection with (5.69) and (5.70): the second, but not the first, can reasonably be followed by . . . but I dare say/I'm sure you won't. In this respect, had better is closer to must than should. This discussion has indicated a number of ways in which had better is, despite some claims to the contrary, radically different in use from should. In the following section all the items discussed in 5.2 will be compared.

185 5.2.4

Conclusion

These four items form a very varied set. Should (in some of its uses) and ought to have as good a claim to cognitive synonymy as any pair of items; be supposed to has a rather narrower area of interchangeability with should; while had better, despite a certain similarity, has minimal interchangeability, and is semantically as distinct from should as be bound to is from must. Should was found to be typical of the pure modals in having a vague basic meaning as a general marker of 'what can be expected'; any further details of its meaning are to be seen as pragmatic interpretations. Ought to differs from this in focusing on ail external requirement or reason, and this leads to typical interpretations in terms of, for example, duty and compatibility with anticipated non-fulfilment, in all of which cases should may show some contrast. In some ways the relation between the two items parallels that between must and have to, and a 'subjective-objective' contrast, suitably qualified, might be claimed. However, the relation between the items is considerably complicated by considerations of phonology, specifically stress; while there seem to be some restrictions related to formal, written style, where, again, there is some parallel between should-ought to and must-have to. Claims for the superfluity and moribundity of ought to are not well based. Be supposed to differs markedly from should in stating what 'is expected', etc, of the subject (directly reflecting its passive form), while should simply states what it is reasonable to expect. Nevertheless, there are close similarities in use, and be supposed to appears to fulfil some suppletive functions for should, just as have to does for must. This item could be considered objective in relation to both should and ought to, but here the term 'objective' is being used rather differently from the way it might be used to distinguish ought to from should. Had better, finally, proves to be very different from should, and to have markedly specific uses which distinguish it from the other items; however, some parallels between the uses of had better and must were observed. It is particularly difficult to determine what could be considered the basic meaning of this item.

5.3

Will, be going to, is to and be willing to

This section is concerned with the relation of will to, in turn, be going to, is to and be willing to. Since these other items are far from synonymous (cf the brief discussion of will in relation to be going to and be willing to in 2.1), this will involve comparing each of the other items with specific uses of will, which is itself the most polysemous pure modal in English. The first section will deal with will and be going to ( 5 . 3 . 1 ) . While must and have (got) to have generated the largest amount of literature on an M - P set, will and be going to have proved to be the most baffling pair, and this section will necessarily be rather extended, while the discussions concerning the other two items and the conclusions

186 will be rather brief ( 5 . 3 . 2 - 4 ) . I do not attempt to deal with every aspect of the items under discussion, but simply to focus on some that may differentiate significantly between their uses and be relevant to the general theme of M - P contrasts.

5.3.1

Will and be going to

I shall first try to clear the ground by considering the general findings of accounts of the modals that deal in some detail with the will-be going to contrast ( 5 . 3 . 1 . 1 ) , and then I shall take, in turn, the following specific themes, which have been prominent in discussions of these items: (a) the relative informality of be going to ( 5 . 3 . 1 . 2 ) , (b) distinctions in temporal orientation ( 5 . 3 . 1 . 3 ) , (c) the conditionality or contingency of will ( 5 . 3 . 1 . 4 ) , and (d) subjectivity and objectivity ( 5 . 3 . 1 . 5 ) ; conclusions will follow ( 5 . 3 . 1 . 6 ) .

The

functioning of these items within extended discourse is a particularly important theme; this will be illustrated at various points in the discussion. 5.3.1.1

Major analyses and major issues

I shall first summarize the relevant findings of four accounts of these items. Each has a different focus, but they are very broadly compatible. I shall start with Coates' analysis of these items (1983:Ch.7), since it provides a clear-cut framework. Coates distinguishes the meanings of will as follows: (5.78)

meanings of will, based on Coates (1983:Ch.7) overlap between

also expressed

meanings

by be going tot

1

willingness

2

intention

yes

3

prediction

yes

4

predictability

root

epistemic

Here Coates' four meanings are shown, in groups of two root and two epistemic types each, together with the overlaps that Coates reports. Meanings 1 and 2 are self-evident, though the precise limits of meanings of these kinds are very difficult to define and vary considerably within the literature; meaning 3 is often simply called something like 'future reference' (eg, They'll be there later or I expect they'll be there later), and is not normally treated as epistemic, this term being conventionally reserved for meaning 4, which is typically concerned with 'predictions' for present time (eg, of a noise, That will be the cat trying to get in). Meanings 2 and 3 are those that are here, as indeed generally, seen as

187 also expressible by be going to, so that any comparison between the items concentrates on these. Palmer's account (1990:Chs 3.3 and 7) has the following meanings for will: epistemic (ie, predictability), subject oriented (volition, power - as in Oil will float on water, and habit), futurity, conditionality (with overlap with futurity), and 'modal' futures (intention, planning, etc). Be going to is discussed in relation to the futurity meaning, but this clearly subsumes 'intention' (op cit:144-145). Palmer's and Coates' accounts are in effect fairly similar, and both emphasize the association of be going to with informal style. Perkins deals with the polysemy problem with will by proposing as the core meaning 'K(C is disposed towards X)', indicating that some circumstances are 'positively disposed' towards the realization of the relevant proposition (1983:42). Be going to is then presented as having the same core meaning as will, but differing in being specified as follows: (a) the modality is typically dynamic (ie, it is not deontic/performative or epistemic), and (b) the determining circumstances are objective, and involve "a current ongoing process" (1983:71). Be going to is thus presented as covering a subset of the meanings of will but being more specialized or restricted in its use, making it, effectively, the marked member of the pair. Haegeman (1981, 1983) also proposes a unitary semantic core for will along the following lines: the item has the components (a) non-factuality or current objective uncertainty, to be resolved at a subsequent time, (b) speaker-based subjective certainty, and (c) focus on the time of occurrence as opposed to its current relevance. Be going to differs from will in the following ways: it (a) does not necessarily attribute actuality to a proposition, (b) can express immediacy without an explicit time marker, (c) is oriented to present time, and (d) is incompatible with hypothetical conditions; these differences are subsumed under the general claim that be going to focuses on the present state of affairs, while will focuses on the future (1981:287-308,402-404). Other accounts that have contributed to the discussion of the will-be going to contrast include, notably, Mcintosh (1966) and Wekker (1976); cf, also, Binnick (1971, 1972), Close (1977, 1988), Haegeman (1989) and Nehls (1986:124-136). All the discussions mentioned so far assume a markedly close similarity between the items will and be going to in at least some uses, and hence the need to distinguish between them explicitly. An alternative approach is, of course, to define the items in different terms, thus implying that any apparent semantic overlap or synonymy is coincidental. This can be illustrated from Chalker (1990), who lists the following six meanings for will: (a) willingness, (b) intention, (c) strong insistence, (d) theory (roughly, meaning 3 in (5.78)), (e) habit, and (f) deduction about the present (meaning 4 in (5.78)); while she assigns two meanings to be going to that are not obviously relatable to those provided for will - future actions for which there already exists (a) a decision, or (b) a cause. No explicit comparison between the items is made. However, it is often assumed that the abundant evidence of near-interchangeability between the items (cf, especially, Palmer 1990:Ch. 7) justifies treating them as a typical M - P pair, and that this closeness should be highlighted in descriptions. 13 In the following discussion, I shall concentrate on two specific issues. Firstly, I shall investigate the association of be going to with informal or colloquial style (these terms 13

The decision to contrast these items in descriptions explicitly is often motivated pedagogically.

188 will on occasion be distinguished), since (a) the significance of this has generally been underestimated, and (b) such stylistic issues are a major, recurrent theme of the present study. Secondly, I shall explore the distinction between will and be going to in terms of future and present orientation, respectively: this aspect of the items has long been prominent in accounts of the items, as was clear from the brief overview above, but it has recently been set out within a specific theoretical framework which merits close examination. A note is required on the actual nature of the comparison and contrast being made between these two items (cf 3.1). In much of the literature, the precise semantic area of contrast remains vague, particularly if monosemantic accounts of the items are aimed at: 'reference to futurity' or 'expressions of futurity' are often cited here, but this is unsatisfactory, if exactness is sought, in view of the notorious overlap between various future-oriented 'prediction' and 'modal' meanings. A notable attempt to be precise in this area is made by Mcintosh (1966), who restricts his comparison of the items to cases of 'pure prediction' that avoid modal meanings of volition, intention, etc However, this is only achieved by the careful construction of appropriate data; many scholars would doubt whether futurity can ultimately be distinguished from these other meanings, and a corpus does not necessarily produce clear-cut data here. 14 A further issue involved here is the fact that various other forms are also available for future reference (eg, the present progressive), and these have their own ambiguity or vagueness of reference: for example, we may choose to compare He ΊΙ see Jane this evening with He's going to see Jane this evening, but this ignores the possible use of He's seeing Jane this evening or He'll be seeing Jane this evening, etc, not to mention the inherent ambiguity of He's seeing Jane this evening (ie, between present or future reference). From this perspective the restriction of the discussion to will (in certain uses) and be going to (also, perhaps, in certain uses) may seem rather artificial. The fact that be going to can be used epistemically (cf 2.3.2) would also seem important for any adequate comparison between it and will. The following texts are worth considering in this connection: (5.79)

14

A

I was going to say that nothing that tastes nice is poisonous, but of course the things that are poisonous we don't eat, so we don't know if they taste nice or not, do we? Β no d like the kingfisher, you mean A like the kingfisher, except, we, a kingfisher is probably not going to taste very good - I mean, it's small and bony and highly athletic Β well, slugs don't taste much better, darling, and they're not small and bony and highly athletic, they're slow and sort of cadaverous (CEC :647-648/1445-1462)

Distinguishing between vagueness and ambiguity in relation to the various 'meanings' of will is also very tricky: Robinson & Haegeman (1979) attempt to do this by applying appropriate tests, but the the results are not very convincing.

189

(5.80)

A

when I've got you I know that if medicine will save him he'll be safe, and that's just all, so of course one doesn't want to give this up C well, if one makes that sort of compliment to a person too, they're going to fall for you, so to speak (CEC:218/1094-1102)

The first text is concerned with speculation about how things taste, in a general, timeless way, and be going to here is epistemic with present rather than future reference; the use of will would make no evident difference.15 The second text is less obviously epistemic - in fact, it reflects the marked difficulty of precisely characterizing uses of will, but it involves 'predictability' (cf (5.78), meaning 4), which is essentially epistemic, and here too be going to could equally well be replaced by will. On such limited evidence it is impossible to point to any distinction between epistemic uses of will and be going to, and it is tempting to see the periphrastic as simply a raje, markedly colloquial alternative to the modal (cf 5.3.1.2). This matter will not be discussed further here, but it strongly suggests that be going to is a potential alternative to will for meaning 4 in (5.78), leaving only meaning 1 of will (the volitional use) without the alternative exponent. 5.3.1.2

Style and formality

In 3.7, in (3.158), will-be going to was listed among a set of formal-informal alternates. 16 Whereas such a listing may constitute an oversimplification in the case of some of the other items listed, there has been widespread agreement that be going to is characteristic of colloquial style, equally in BE and AE. Thus, Wekker (1976:123-124) reports the marked association of the item in his corpus with informal spoken style, while the written material where it occurs (relatively rarely) is restricted to (a) novels, and (b) various types of journalism. 17 The nature of these written sources is significant: novels often simulate colloquial speech, while the newspaper material seems all to be from personal contributions, where an individual, sometimes highly informal, style is often adopted. Wekker (1976:125) and Palmer (1990:142-143) give a striking illustration of the importance of stylistic differentiation in the case of weather forecasts. When broadcast, they may use either will or be going to, but if such texts start with be going to, they often then switch to will (cf comments on (5.83) below); when written they normally use will. However, a person casually reporting a forecast he had heard or read would normally only use be going to. In such cases the 'prediction' meaning of the items is involved, and the choice of item is said to be determined solely by stylistic considerations. I shall now explore this question of stylistic alternation where there appears to be no specific semantic or pragmatic justification for the choice; two journalistic texts will be 15

16

17

Palmer notes that the use of 6e going to in this text requires "some special explanation", but he simply treats it as making "a fairly firm statement" about present time (1990:148-149). Be about to was also listed as an informal periphrastic along with be going to, but this does not concern us here. Similar corpus-based comments are made by Palmer (1990:142-144) and Coates (1983:204), and many descriptive and pedagogical grammars note the association of 6e going to with informal speech (eg, Quirk et al 1985:214). Suh (1992:99-100) provides data showing that the frequency of be going to in relation to that of will can be directly related to the degree of informality of the material.

190 used. As a first illustration I shall take a three-page article on future developments in television in The Economist (12.2.1994:5«ri;ej/,9-ll). In this there are 36 occurrences of will, and just one of be going to. The following shows the single occurrence of be going to (op cit:10): (5.81)

"People don't know how to want a lot of the stuff we're going to offer," says Steve Arnold, a manager at Microsoft.

The fact that this presents direct speech allows the option of be going to as opposed to will, which would be odd here in colloquial style. The following is a typical excerpt from the article (occurrences of will are italicized and numbered) (op cit: 11 ) : (5.82)

The broadcast networks will (1) struggle ever harder to hold on to advertisers, which are shifting their spending towards targeted media, and to audiences. Research suggests that, however many channels they have, people only watch seven or so. But the more channels there are, the less likely it is that ABC, say, will (2) be among the seven. And once episodes of programmes are stored on servers, channels disappear. NBC will (3) still unveil episodes of 'L.A. Law' at a certain time each week, just as magazines have a first day on newsstands.

Of the cases of will here, (1) and (3) are typical of the use for the 'pure' future of prediction, which seems to be uncommon in colloquial style; (2) is in the structure likely (that) ... will, which is standard, regardless of style. The second text (K. Whitehorn, 'Bring the jet set down to earth', The Observer, 20.2.1994:25) is by a columnist singled out by Wekker (1976:124) for high frequency of be going to in a corpus from 1969-1970. This text (c 1150 words) has four occurrences of be going to, and six of will. The following excerpts show all the occurrences of be going to (these, together with one occurrence of will, are shown in italics and numbered): (5.83)

(a) (b)

(c)

"When are you going to (1) settle down?" . . . I don't know any global sanctions as strong as "What would the neighbours think?" but at this rate it sure looks as if we are going to (2) have to find some . . . Since he is not in one place long enough to form serious local attachments, he inevitably identifies with others of his own kind, denizens of the airport lounge. The problems of Surrey or Sunderland, where he actually lives, are never going to (3) be as important to him as his own gypsy lore, the structure of rewards and punishments dished out by the firm, the culture of the company. Such people have The Economist and the Wall Street Journal in common with most other people they meet; they aren't going to (4) care much about the opinions of the Scunthorpe Advertiser. They will (5) not, of course, leave pieces of these papers blowing about the fields.

191 Occurrences (1) and (2) of be going to are clearly in colloquial style: the first excerpt simulates speech, and the second has it sure looks . . . , which is markedly non-standard in BE. In excerpt (c), be going to (3) could, in fact, be replaced in colloquial style by will, but less obviously occurrence (4): this seems to be because the lexical verb be readily takes -will even in colloquial style in contexts where other verbs more naturally take be going to.18 The fact that they aren't going to . . . is directly followed by they will not (5) can perhaps be related to the tendency in spoken style already referred to to establish a future situation with be going to and then switch to will: this is documented by Wekker (1976:125), and, further, by Palmer (1990:146), and is related by Wekker to the contrast in implications between present orientation be going to and future orientation will. (A string of be going tos is, in any case, stylistically undesirable in careful style, as here.) The use of be going to in informal spoken style to provide a setting or frame for the description of future situations, with will then taking over for elaboration within this frame is illustrated extensively by Suh (1992); this is shown as an instance of the necessity for discourse factors to be taken into account in description 19 , and is perhaps parallel to a similar case discussed in 4.8.1, there illustrated in text (4.215). This phenomenon can be illustrated from fiction too: (5.84)

" . . . Well, that's what I'm going to do. I've just thought of it." "Not for long, I hope." Yes, that's what I'll do, thought Ludens, I'll run away . . . (Murdoch 1990:562)

This text is immediately followed by 11 further verbs with will. On the other hand, there is no block on reiterated be going tos in informal speech: (5.85)

so the ones that are going to join in this sort of rat race are the first ten or so, probably - now the first six are going to be the only people who are going to get a look in, and the first four are the only people who are actually going to make it - now the first four are the people who are going to really stick at it, you know, for four or five years (CEC:603/568-579)

Here there are five instances of be going to in close succession; there does not seem to be any special reason for this choice, such as special emphasis. There is, then, abundant evidence of the crucial role of style here: in informal, and particularly in colloquial, style be going to is the chosen, or preferred, form in many uses where will would be required in more formal style. This means that the two items can only be properly contrasted within appropriately informal style. These points are worth emphasizing, since the stylistic parameter is often completely ignored in descriptive and theoretical accounts, and sometimes even in pedagogical descriptions (eg, Sinclair 1990:239,240,256). 18 19

Lexical have is similar to be in this respect. Suh finds parallel cases of this phenomenon in (a) the use of used to and would for habitual past time activity, and (b) t h a t of the present perfect and past simple tenses, for framing and elaborating functions, respectively.

192 5.3.1.3

Temporal orientation

As has already been seen, there is widespread recognition among scholars in this field of a crucial distinction between will and be going to in the temporal orientation of their uses - that is, according to whether the utterances they occur in 'belong' in present time (for be going to) or in future time (for will). This notion is inherent in the structure of the be going to form itself, which suggests movement from a present into a future state (cf Traugott 1978, Fleischman 1982:especially 78-102, and Hopper & Traugott 1993:61,8788), and appears to have a long history: Binnick (1971:49) cites a grammar of 1751 that treats be going to as an 'inceptive present', relatable to inchoative verbs in Latin and expressing a 'nascent future', and will as the 'aorist of the future' (ie, a distinct future reference tense), while more recently Jespersen emphasizes the close association of be going to with present time (1931:217-218). Such an approach to the contrast between the items is also implied by Reichenbach's analysis of tenses in terms of the three time points S (speaking), E (event) and R (reference) (1947:96; cf 3.5.4). Reichenbach analyses the French future reference forms je vais voir and je verrais as, respectively S,R - E and S R,E\ if these forms are treated as parallel to the English forms I'm going to see and I shall see, respectively (and semantic closeness is widely assumed), then the distinction between 'present/current orientation' (be going to) and 'future orientation' (will) is reflected in the position of R (cf Fleischman 1982:97, Haegeman 1989:296-298, and Smith 1989:110-111). Despite general recognition of the point involved, adequate illustration of it is curiously difficult, since invented non-contextualized sentences are unrevealing while natural contextualized data do not come in contrasting pairs. Mcintosh (1966:307) provides pairs such as the following (these are intended to be purely predictive in meaning; contextual implications axe retained from the original text): (5.86)

You're going to beat him easily, (suggestion that the match has already started)

(5.87)

You'll beat him easily, (when the time arrives)

(5.88)

He's going to have an awful time when he marries her. (even now, the writing's on the wall)

(5.89)

He'll have an awful time when he marries her. (though things may look all right at present)

These represent clear contrasts. Palmer supplies the following two spoken texts to illustrate the same distinction (1990:145): (5.90)

They've just said they've now completed the survey at last, and according to the result of that they're not going to give us a mortgage.

(5.91)

You see, we have got money with the Abbey National, but they've told us that they won't lend on conversions.

In (5.90) the decision not to lend is based on present information, which is not obviously the case in (5.91), where Palmer terms the use of will 'conditional'. This last point

193 illustrates a pervasive problem with these items: the contrast involved turns out not to be simple, and here the reference to conditionality introduces another factor (cf 5 . 3 . 1 . 4 ) . Many accounts treat an opposition of this kind as prominent, but not such as to subsume all others; for example, Mcintosh (1966) treats it as one of the two basic oppositions between the items (the other is discussed in the next section). However, Haegeman (1989) argues that such a distinction is sufficient to account for all observed contrasts between the items (presumably restricted to predictive uses). This is presented within a specific pragmatic theory, discussed in 3 . 5 . 3 , according to which the items will and be going to are, at sentence level, truth-conditionally equivalent, and differences between them are due to constraints that they impose on their processing within extended contexts. This means that will and be going to signal that the utterances they occur within are to be processed within future and present time contexts, respectively. Since this account makes a number of very specific claims, and criticizes the atheoretical and inexplicit approaches of other accounts such as those just referred to above 2 0 , it seems worth exploring its implications in some detail. The following aspects of this account of will and be going to will be focused on here: (a) the assumed semantic equivalence between the items, and (b) the notion of 'context'; the theory will then be applied to a group of texts from the CEC. Semantic equivalence. The problem of demonstrating semantic or truth-conditional equivalence was outlined in general terms in 2.1, and specifically in relation to the will-be going to pair in 3 . 5 . 3 . Other scholars have sometimes made similar claims: for example, Binnick states that the items - excluding their modal uses - differ not in meaning, but only in their presuppositions (1972:8). The difficulties involved here can be illustrated briefly. For example, let us compare the following pair: (5.92)

That bottle will take 10 litres.

(5.93)

That bottle is going to talee 10 litres.

(5.92) is ambiguous between (at least) a salient 'power' meaning (eg, 'it's big enough to hold that much') and a prediction meaning (eg, 'I'm going to put that much into it'), while (5.93) only accepts the second meaning. This means that it should be possible to affirm (5.92), in the first, salient interpretation, and at the same time deny (5.93); and in fact there is nothing odd about: (5.94)

That bottle will certainly take 10 litres but it's not going to today.

This clearly shows the need to restrict the range of uses of the items: presumably, the 'power' meaning, which seems to be subsumed under 'predictability' in (5.78), is fully distinct from prediction, so that reference to futurity will, or at least can, exclude that sense. To take a rather different case, Comrie (1976:64-65) claims, in connection with the following pair: 20

As an example of this, the claim by Wekker (1976:128) that sentences with will are 'often elliptical' (cf 5 . 3 . 1 . 4 ) is objected to on the grounds that this is an imprecise observation (Haegeman 1989:299).

194 (5.95) (5.96)

Bill will throw himself off the cliff. Bill is going to throw himself off the cliff.

that if Bill does not carry out his intention, (5.95), understood as a prediction, but not (5.96), understood as an expression of intention, is subsequently shown to be wrong. This may not count as conventional truth-conditional nonequi valence, but partially reflects the use of the items in the past tense illustrated earlier in the pair (3.136) and (3.137), respectively. We can compare this with another pair, (5.86)-(5.87): here, if the addressee subsequently does not win the match, the prediction is proved equally wrong in both cases. This illustrates an important distinction: in (5.96), but not in (5.86), an intention is involved as opposed to a 'pure prediction'. 21 Thus, a claim for truth-conditional equivalence should at least presuppose an explicit selection of the meanings of the items, perhaps of the kind Mcintosh undertakes. Haegeman subsumes the uses of be going to for "intention, cause and indications" - ie, presumably all uses - under present orientation (1989:307-308), but it is not clear which of the various uses of will are intended to be included. Context. The notion of context is central to Haegeman's argument. A context is seen (1989:especially 306) as a set of propositions, which may or may not be linguistically overt. The presence of be going to or will tells the hearer to access present or future time propositions, respectively, in order to interpret an utterance; and there is always a present time context by default. One illustration of this was discussed in 3.5.3, with examples (3.141)-(3.143). Haegeman supplies further examples such as (1989:309): (5.97) (5.98)

Now we'll have no bread left for tomorrow's breakfast. Now we're going to have no bread left for tomorrow's breakfast.

In (5.97) will is said to direct the hearer's attention to "the future implications of the proposition (ie tomorrow's breakfast)", and in (5.98) be going to to "the present implications (ie what to do about the situation now)"; presumably other background propositions of this kind for (5.98) could also include the present shortage of bread or the speed at which the bread is disappearing. It should be possible to make the distinction between the two sentences more explicit by means of modifications to the original sentences, as in: (5.97a)

Now we won't have any bread left when we have breakfast tomorrow. ?(+)

(5.98a)

Now we're not going to have any bread left when we have breakfast tomorrow.

the shops. 21

This assumes that 'intention' and 'pure prediction' meanings can be distinguished satisfactorily in use.

195 According to the argument, (5.97a) should be better than (5.98a), and (5.98b) than (5.97b), but this is not obviously the case, as they all seem fine.22 A claimed concomitant of the difference in interpretation between (5.97) and (5.98) is that now has different meanings: it is compatible with both present and future contexts, and thus in (5.98) has its 'prototypical' present time sense, and in (5.97) its 'consequence' sense (ie, 'this means that . . . ') (Haegeman 1989:295). This too is questionable: the present time sense, as in He's living in Nice now, is hardly appropriate for (5.98), and the 'consequence' sense seems natural for both sentences, without evident distinction. In considering this approach to interpretation in more general terms, two crucial questions need answering: (a) what constraints are there on the accessing of contextual propositions - whether overtly linguistic or covert - in pragmatic interpretation? (b) how does this theory relate to other factors found relevant in distinguishing between the uses of the items? To investigate the first question, naturally occurring discourse data, specifically from colloquial use, should provide valuable evidence. There is a paucity of such material in Haegeman (1989): there appear to be only three natural texts longer than single sentences, one derived from Wekker (1976) and two from written material; further, no account is taken of the potential significance of formality level on the choice of item. To explore these questions in natural conversational discourse, I shall now examine a series of six CEC texts. 2 3 Discourse data. In analysing these texts, I shall try in particular to find what support (in Haegeman's sense) the individual tokens of will and be going to whether the use of the alternative item would in each case be possible, and, what contextual support and contextual effect. All relevant occurrences of the including shall, which is treated as equivalent to will, are shown in italics and for convenience of reference. (5.99)

contextual have, and if so, with two items, numbered

I must go down to the bank, you see, and catch this one-twenty-eight 24 I'm going ίο (1) have a new gearbox in my car, and I'm going to (2) collect it tomorrow - I think it's ready by now (CEC:41/423-429)

A present context, appropriate for be going to, is prominent (eg, I must go ... , I think ... , by now). Replacement of be going to by will would be inappropriate here: in (1) it would need some future context such as in I'll have a new gearbox in my car by the time I see you next, and in (2) I'll collect it as soon as it's ready. However, if be going to {1) stays, will would be possible for (2) (see comments on the possibility of switching to will after an initial 'scene-setting' be going to in 5.3.1.2). As to other overt contextual data, it is striking that the unambiguously future adverb tomorrow can co-exist with be going to (1) yet would not alone provide adequate context for the use of will here. 22 23 24

A possible distinction between (5.97) and (5.98) will be mentioned in 5.3.1.4. Suh (19.92:Ch.5) presents some valuable discourse data on these items. Either the time of a particular train or bus ('1.28'), or a bus route ('128').

196 (5.100)

Β

obviously the time to see it is somewhere between August 15th and September 15th, so I'm very much hoping that I'm going to (1) get an answer before I go - either yes or no a what's he going to (2) say yes to? yes, you may come and take a look at us? (CEC:382/458-463)

Here be going to (1) has a present context in I'm very much hoping, and be going to (2) could be said to have one in the current interest of the interlocutors, or simply in occurrence (1). Will would be possible for (1), 'with a future context in before I go, though such a context is not required for will here; I'm ... hoping (in present time) also provides a typical 'qualifying' context for the use of will. Will would not be appropriate in (2), even if it were used in (1), despite the general future orientation of the text (between August 15th ... , before I go, etc): it would seem to need some further context, such as what do you expect . . . or . . . when you ask him. (5.101)

A

and I don't want to, you know, run ourselves out of an external examiner by your saying, oh, to hell with this for a game - I'm not going to (1) have my summers buggered up in this kind of way Β no, I, no, it'// (2) be perfectly all right, Reynard, I've got a, I've got about a week of fairly hard work after the 4th of July (CEC:36/135-143)

Here this kind of way way provides a present context for be going ίο (1), while the reference to the forthcoming week of hard work provides a future context for will (2). Will would also be possible in (1) (I won't have ... , like I'm not going to have ... , naturally indicates strong negative volition - this seems to lie between meanings 1 and 2 in (5.78)); but it is not clear where a future context for it is to be found, apart the fact that the plans being discussed necessarily belong to the future. Be going to would be possible for will (2) and would provide a more reassuring comment than will does; and a present context for it could be found in I've got . . . It does not seem that any shift of orientation between the immediate present and the future would be entailed by either of these switches between the items. (5.102)

Β A

I'm not sure I haven't got a plug somewhere I think VU (1) make another one - I don't know what I'm going to (2) do about this seminar tomorrow (CEC:121/949-952)

Here a future context for will (1) could be found in the proposed future activity, and a present context for be going to (2) in the speaker's present problem; at the same time, I think is a typical qualifier for the use of will (1). The items could be reversed, but then I'm going to make another one could imply that the speaker is commenting on what he is already planning (this would be a present context), while what I'll do about this seminar would strongly imply that the speaker is currently helpless about his seminar (but an overt future context is available in tomorrow).

197 (5.103)

we're having this meeting of CSC assistants on the 4th of July, which is a Saturday - VU (1) have about half a day's work to look at some odd scripts before then, and then I shan't (2) get any scripts from the assistants before about, let me see, 4, 5, 6, 7, about the 8th, so I shall (3) have roughly from the 29th of June to the 8th of July on which I can, I can spend the whole of that time on those two papers if they happen to come, and then again from the 8th of July until whatever time your council meeting is, again, I can spend the whole time on them - you see, I shall (4) get scripts from ten assistant examiners, which will (5) mean a couple of days' work - I shall (6) get those on about the 8th or the 9th of July - well, they'// (7) be out of the way quickly, you see, and, and then I can get straight on to the papers again (CEC:37/182-214) 25

A future context is clearly established here with the various dates, justifying the several occurrences of will/shall. It is significant that the first future reference form is we're having, which establishes the future setting (the same effect has been seen above with be going to). Each of these cases could, individually, be replaced by be going to (though there could then be a stylistic problem; but cf text (5.85) with its string of be going tos); however, the use of be going to would not have any obvious present context to justify it. Future reference will, in colloquial style, often requires an initial 'setter' like, in this text, we're having. (5.104)

he said, will (1) you be back after the weekend?, so I said, well, no, actually, it'// (2) probably be more like 10 days, so he said, oh, well, that's very sad . . . so I said, well, I'm terribly sorry, but, you know, that's how it's going to (3) be, cos I'm going to (4) be away for 10 days at least, so he said, VU (5) give you a ring after the weekend, so I happily put the phone down, thinking, aha, I'm not going to (6) be here (CEC:702/935-959)

A future context for the will occurrences (1) and (2) is explicit in after the weekend, etc Be going to would also be possible in these cases, with a present context presumably recoverable in the current planning involved. At the same time, it is significant that if the main verb with will (1) were not be, but, say, come, will would not be so natural in the appropriate sense; the common co-occurrence of will and be has been noted above. This does not apply to occurrence (2), since probably provides a typical qualifying context for will. Be going to (3) has, presumably, a default present context; will is possible, but would differ in presenting the future objectively, whereas be going to here strongly suggests the speaker's will. The overt context for be going to (4) is very closely parallel to that for will (2), ie future; however, a present context can always be found in the ongoing planning of the speaker. The effect of the two be going tos in (3) and (4) is certainly stronger than it would be with wills; this can perhaps be related to the association of be going to with immediate personal plans as opposed to future possibilities. Will (5) 25

In the interests of intelligibility, this text and the following have been slightly simplified, and the contributions of an interlocutor omitted.

198 also has an overt future context; here, be going to would not be suitable - it could, for example, have the force of a threat. Finally, in (6) will could easily replace be going to (will in (1) has a very similar context); it is not clear what bearing covert contexts could have on the choice of item here. The discussion of these six texts in relation to the first question raised above suggests two crucial problems with Haegeman's pragmatic theory: firstly, there is no evident way of determining which contextual propositions - overt or covert - are to be accessed, or with what priorities, and secondly, even assuming this were possible, it would not be sufficient for the explanation of the choice of future reference items, since a number of other factors that are not readily subsumed under notions of general present or future time orientation appear to be involved. Despite these problems, informal notions of present vs future orientation are obviously revealing in relation to the choice of will vs be going to. The second question raised earlier in this section concerned other factors relevant to the distinction between uses of these items. These will be discussed to some extent in the following two sections.

5.3.1.4

The conditionality or contingency of will

Under this heading I refer to indications that will is typically used for conditional or qualified statements and thus may make a less definite claim for the future than be going to. Mcintosh (1966:especially 308) treats the contingent-absolute contrast as the second basic opposition involved in the comparison of the items, with will correlating with contingency, and be going to with absoluteness. This contrast, which was referred to various times in the previous section, and reflected in the glosses to (5.86)-(5.89), involves will being used in conditional sentences, and generally in contexts that make the futurity in question contingent on some external factor, resulting in the use of the item often appearing elliptical. By contrast, be going to makes independent, non-contingent predictions. This is evident in the regular use of will in the apodosis of conditionals of the type If χ happens, y will happen: be going to is not normal in such contexts, although its use is perfectly possible (cf discussion of (3.141)-(3.144) in 3.5.3, and examples (5.80) above and (5.118) and (5.120) below). There is further evidence of this in the regular use of will rather than be going to in structures such as Maybe/probably/I think/I hope, etc, he'll be there, illustrated in the previous section. The crucial point here is not that be going to is in any way excluded, but that the use of will would be often very odd in such cases without the qualifying phrase or clause. A distinction of this general kind might also be relevant to sentences (5.97) and (5.98): whereas (5.98), with be going to, appears to make an independent statement, and could in appropriate circumstances denote an intention, (5.97), with will, could suggest that there is some contextually available determining factor, eg, since we 've been so greedy or if all those people really come to supper. These instances involve overt language data; but commonly the contingency of will is related to covert data (hence the detailed contextualizations in many accounts of these items). Some illustrations follow. Palmer argues of the following pair:

199

(5.105)

The paint'll be dry in an hour.

(5.106)

The paint's going to be dry in an hour.

that there is an implied condition in (5.105) such as 'if you leave it', while the second expresses inevitability (1990:147). The following pair are similar: (5.107)

It'll cost me a fortune to get them home.

(5.108)

It's going to cost me a fortune to get them home.

Here, Palmer states that the first sentence implies 'if I so decide', and the second 'that's what I'm going to do' (1989:59). Nehls provides a text where the 'condition' for the use of will is not overt but is clearly required and readily recovered (1986:126, quotation here reduced): (5.109)

Cliff:

It's a lovely, delicious paw you've got. Ummmmm. I'm going

to bite it off. Alison: Don't! I'll burn his shirt. (J. Osborne, Look back in anger. London: Faber, 1957:13) The use of will in I'll burn his shirt implies a condition like if you bite it o f f . Binnick (1972:3) makes essentially the same point in his claim that sentences like (5.110)

The rock'll fall,

as opposed to (5.111)

The rock is going to fall.

are incomplete or elliptical (1972:3), while Larkin's comments quoted in (3.100) on the oddity of: (3.98)

?Hang on! We'll crash!

can be interpreted similarly. It is also claimed that the notion of contingency can be extended to cover causation, in view of the following data (Binnick 1972:4): (5.112)

We're going to get rid of our teacher, and then we'll be happy.

(5.113)

We're going to get rid of our teacher and then we're going to be happy. 26

Binnick states that in (5.112) we'll be happy is the result of the first clause, while in (5.113) we're going to be happy has no such motivation. These may represent default interpretations, but are certainly not obligatory. These arguments suggest that be going to maíces a more definite or strong prediction, as happens in the following text (where will would be perfectly normal in place of 're going to): 26

It is not clear whether the difference in punctuation between these two sentences has any significance; it certainly should not, in t h a t the use of be going to is a feature of spoken style.

200 (5.114)

But I said "Don't have any doubt about this, dear friends, that if you are going ahead with doing that, you're going to be dealing with me!"

Here Palmer claims that the use of be going to makes the event "more immediate or more certain" - "the future is 'extra-real'" (1990:148-149) (but cf the evidence of (5.85)). There is a major potential problem in the apparent contradiction between the conditional or contingent use of will and the use of will for 'pure futurity' - for future events that cannot be otherwise within the world we know, as in He'll be 65 in 1999 or 2000 will be a leap year.27 Such predictions are of the strongest kind possible, and could be termed absolute. Palmer suggests that the use of will in such cases is a default use: be going to would be inappropriate because of a lack of present-future relation, and will is not so much appropriate as less inappropriate than be going to (1990:148). For an alternative approach, we can return to Haegeman's pragmatic theory (1989), discussed in the last section. Haegeman posits a universal present context by default, which entails that a future context needs to be supplied for the processing of will sentences. This is the reason for the 'elliptical' nature of will sentences such as (5.110), assuming that no future context is available; by contrast, (5.111), with be going to, has a ready-made present context by default (Haegeman 1989:306), and, in consequence, is likely to appear more definite, and hence stronger. 28 At the same time, the fact that will properly 'belongs' in distinct future time seems to allow it to refer to fixed and definite events such as birthdays or leap years. In other words, both fixed events and contingencies of various kinds provide equally appropriate future contexts for the processing of will. If this is correct, it provides a framework for combining the topics of both 5.3.1.3 and 5.3.1.4. 5.3.1.5

The objectivity of be going to

I shall finally turn to the notion that be going to is an objective item. Perkins (1983:7172) argues that with be going to the determining circumstances are objective, and these are identified as a present process that will result (with greater or lesser certainty) in the relevant prediction. In this sense, be going to is presented as typical of all complex verbal expressions with be (including, for example, be bound to and be supposed to); the general proposal was discussed in 3 . 3 . 3 (cf (3.60)). In a case like (5.111), which Perkins discusses, the objectivity could lie in incipient visible movements, and this might then be considered a virtually certain prediction; similarly, in (5.106) the prediction might be based on the paint's recognized drying properties. This proposal has some similarity to the argument discussed in 4 (especially 4.8.2.2) that with the use of have to the speaker has objective evidence, while with must he merely has personal grounds. However, it is difficult to see this as applying to all uses of be going to, since, for example, a speaker of (5.106) might simply be making a strong prediction, while in informal style be going to is often simply the obvious choice for future reference that is not explicitly qualified. Further, in some of its uses be going to is closely associated 27 28

Be going to is also possible in such cases in colloquial style. Cf Joos' comment that with be going to the future event is "assumed rather than determined" (as it is with is to), and "taken for granted as a proper part of future reality without any suggestion that there had to be a cause to make it so" (1968:22).

201 with expressing the speaker's point of view - a kind of subjectivity by any definition; while some uses of will - eg, for 'pure futurity', as just discussed in the previous section - may appear markedly objective. 5.3.1.6

Conclusion

This discussion is difficult to summarize satisfactorily, for the following reasons, which have all been highlighted above: (a) the exact range of meanings of each item that can be compared is hard to define; (b) the marked polysemy of will makes it hard to argue for any kind of generally valid core meaning, and hence for any basis for a comparison between the items; (c) the restriction of be going to to informal spoken style means that any relevant contrast can only be properly demonstrated in informal contexts; and (d) the degree of interchangeability between will and be going to appears to be even higher than was observed with must-have to-have got to. Nevertheless, allowing for these factors, temporal orientation seems to constitute the basic, or, alternatively, the most significant, difference in use between the items. Some final illustrations of how these items can function in contrast will now be given. Firstly: (5.115)

What will you do this weekend?

(5.116)

What are you going to do this weekend?

Here, as a casual enquiry, the first question is odd, since future reference with personal intention or planning (and both factors are natural here) is appropriately expressed by be going to, as in (5.116), or the present progressive, as in What are you doing this weekend? (5.115) needs some justifying context, for example a condition ( . . . if it rains) or temporal qualification ( . . . now the others have gone)·, in consequence, lacking such context, it could have other implications, such as the addressee's helplessness. Two further pairs follow: (5.117)

She'll get annoyed if you go on like this.

(5.118)

She's going to get annoyed if you go on like this.

(5.119)

I'll get annoyed if you go on like this.

(5.120)

I'm going to get annoyed if you go on like this.

Will is normally used in the apodosis of such conditional sentences, and (5.118), with be going to, could be seen as the marked member of the first pair; an appropriate overt present time base for the use of be going to could be, for example, I know . . . or / can see ... , and, lacking this, the sentence could, for example, have the force of a warning. Whereas in (5.117), where the two clauses are equipollent, the conditional connection is clearly focused, in (5.118) the main clause has relative independence, and the conditional connection is partly causal. In the second pair, however, (5.119) is the odd member; the reason is that while neutral observations about other people's likely behaviour (such as will suggests) are normal, this is not the case when talking about oneself, when a speaker will naturally

202 express his intentions (with be going to), as in (5.120) - unless he is observing his emotions objectively.

5.3.2

Will and is to

The periphrastic is to is also sometimes compared to will, more so than to must. However, the overlap of meanings between the items is narrow, and the discussion here will be correspondingly brief. In the following, I shall first discuss the various uses of is to, and then compare the relevant ones with the comparable uses of will. Quirk et al treat is to as essentially "expressing futurity, with various connotations of 'compulsion', 'plan', 'destiny', etc, according to context" (1985:143,217-218): this points to futurity as the basic meaning. Alternatively, the deontic use may be seen as basic: thus, Jespersen treats the 'usual meaning' of 'obligation, destiny, etc' as weakened into a simple futurity use (1931:354-355). Rather similarly, Perkins (1983:68-70) has a unitary analysis of the item, which effectively equates it with be going to, with the sole difference that the determining circumstances, instead of being marked for present orientation, are "consciously organized (typically)". 29 The 'weakened' pure future use is seen in this account as an incidental consequence of the fact that any expected/determined action necessarily occurs in subsequent time; this may be compared to the weakened use of have to for intended future actions illustrated in (4.12) in 4 . 2 . 2 . Perkins uses the following example to illustrate this: (5.121)

You are to marry him within six months.

This is ambiguous between a command (strong) and a prediction (weak). We are concerned here with meanings which correspond to uses of will, and here the 'futurity' use and an 'obligation' use "to give or relay commands and instructions" (Palmer 1990:166) can be distinguished. This means ignoring, for example, a 'possibility' use and the idiomatic use seen in John is to blame·, these are all documented by Palmer (1990:164-166). Future reference. The use of is to for future reference is, in fact, semantically closest, among the available future reference forms, to the relevant use of the simple present tense (Quirk et al 1985:217). However, there is some evidence of alternation between is to and will here. Poutsma (1904:33) quotes the following text in illustration: (5.122)

The building is to be seven storeys in height, but a portion of it will rise to ten storeys.

Nehls (1986:130-131) cites another extended text with an initial is to followed by multiple wills. Such examples show that will can be used as a neutral future marker once a future 29

Joos makes similar comments in contrasting is to with be going to: the former means "a practical determining of subsequent events", while the latter deals with events "assumed rather than determined" (1968:22; cf Footnote 28 above).

203 scene has been set with is to, just as happens with be going to (and the present progressive tense with future reference). However, generally the items do not appear close in their uses: is to refers to planned events in a maximally objective way, while future uses of will and be going to are, typically, either purely predictive or involve issues of personal intention and will. Instructions. In the deontic 'obligation' sense the items seem close to being interchangeable. Thus, a typical use of is to in: (5.123)

You're to report to the police station every Friday.

could, roughly, be replaced by will. This use, like the future reference use above, is markedly objective in that an order is uttered relatively impersonally. Leech (1987:102) refers to the 'imperative flavour' of is to here, and treats it as similar in meaning to have (got) to and ought to (is to seems, in fact, stronger than both of these, and ought to is hardly comparable, since it allows non-fulfilment). The corresponding use of will to issue instructions is rather rare and belongs, perhaps even more than is to, in strictly defined power structures. Palmer cites the following text, where the speaker tells the subject, in the third person, what to do "by indicating what has been decided (either by custom or by him)" (1990:142): (5.124)

and the President, with head averted and profile turned to me, said "Mrs Dodgson will walk on my right" (CEC:93/579-583)

It is this sense of is to that Lakoff relates to will, with the following examples (1972:240; cf (1.2) and (3.80)): (5.125)

John will take out the garbage.

(5.126)

John is to take out the garbage.

In (5.125) the speaker is said to issue an order, while (5.126) can also be used to report someone else's order. This association of a speaker source with will, and its lack with is to, reflects a standard M - P tendency; and it receives support from comments by Leech (1987:102), Palmer (1990:166) and Nehls (1986:29), though they all allow is to equally to initiate or to pass on commands. In fact, it seems that here, as with have (got) to, be supposed to, etc, an identifiable source is required, and, by default interpretation, this is assumed to be distinct from the speaker, though the speaker is not excluded (cf, for example, the discussion of may vs be allowed to in examples (3.42) and (3.43) in 3.3.4). Despite the possible closeness of the M - P pair in (5.125)-(5.126), the interpretational options for the sentences are not parallel: (5.126) favours an 'instruction' sense, while (5.125) favours a sense where the speaker undertakes that John will do the job, and a strict 'instruction' sense is hardly available here in ordinary circumstances. Palmer (1990:81), in commenting on the Lakoff proposal, simply remarks that the distinction between the items "seems to be rather different" from the one proposed.

204 Conclusion. It has been shown here that is to could be considered a relatively objective counterpart to will in certain future and obligation uses, but such a relation is very restricted in extent and hardly significant in comparison with those that characterize other M - P pairs and sets examined.

5.3.3

Will and be willing to

Be willing to is a much simpler item to deal with. Its meaning of 'willingness' is transparent, and differs from the volitional uses of will, which are close to intention. Some ways in which it contrasts with will, by virtue of its narrower sense, are illustrated in the following pairs: (5.127) (5.128) (5.129) (5.130)

They aren't going to visit us, but they are willing to. *They aren't going to visit us, but they will. He wasn't willing to open the files, but he did so in the end. *He wouldn't open the files, but he did so in the end.

In the first pair, the distinction between be willing to and be going to, and hence the compatibility between the assertion of one and the denial of the other, is clear, as is the impossibility of specifying an exclusively volitional meaning for will.30 In the second pair, wasn't willing to is not incompatible with explicit subsequent performance, but wouldn't is, apparently because while be willing to narrowly denotes willingness or readiness, will suggests a stronger emotion, and wouldn't in past time reference typically means 'refused to'. Be willing to, then, has a narrower, more specific meaning, than will, as is typical of periphrastics, and it may be assumed to have standard suppletive functions. It has attracted some attention because of the gap in the following partial paradigm 3 1 : (5.131)

He was willing to come any time he was free.

(5.132)

He would come any time he was free.

(5.133)

He was willing to come on that occasion.

(5.134)

*He would come on that occasion.

This parallels the situation with can and be able to, and has been commented on by Anderson (1971:92) and Palmer (1980; 1990:163-164,194-196). Palmer deals with this in terms of the 'actuality hypothesis', discussed in 3.6, and attributes the gap at (5.134), ultimately, to the lack of compatibility between modality (as expressed by fully modal exponents), and factuality; this general issue will be returned to in 5.4 in connection with can and be able to. A further possible problem with past tense forms of will and be willing to arises in negation, as is illustrated by (5.129) and (5.130). Palmer (1990:196) argues that negative volition, unlike negative ability, does not imply negative actuality, so that there should be no problem with (5.130); since, however, wouldn't in (5.130) clearly does 30 31

For further relevant data, see (2.13)-(2.20). Conditional senses for would are not intended here.

205 imply negative actuality, this must be due to analogy with the behaviour of could.

An

alternative explanation for this particular restriction was suggested above - that wouldn't strongly suggests a meaning like 'refused', and is for this reason inappropriate. In present tense uses, a parallel with past tense meanings (cf, also, the question of the present tense meanings of can and be able to, to be discussed in 5 . 4 . 3 ) would suggest that will could mean 'is willing to (and may do in the future)', and be willing to 'is willing to and does'. The evidence here is unclear. For example, if we compare: (5.135)

I'll give you a hand any time.

(5.136)

I'm willing to give you a hand any time.

the first in fact makes the more positive offer. Again, we may assume that the use of will suggests intention and future performance, while be willing to is more restricted in what it implies. This restriction does not apply so clearly in the case of habitual actions: (5.137)

John'll always give a hand if things get a bit tight.

(5.138)

John's always willing to give a hand if things get a bit tough.

This is presumably because it is difficult to be regularly 'willing' and never perform. Be willing to, then, is a pretty clearly defined item, with a restricted potential for close overlap with will·, there are grounds for terming it 'objective' if this means that the item has a well defined sense concerning the subject's feelings, and it has the features of a syntactically and semantically marked item. Comparison of this item with will, just as was the case for be going to, is complicated by the polysemy of will, and resulting default interpretations of that item.

It is reasonable to assume be willing to fulfils suppletive

functions for will in the same way as, for example, be supposed to does for

5.3.4

should.

Conclusion

The discussion in 5 . 3 has concerned will in relation to three other items. The problem of comparing items with a variety of uses is highlighted here, since will is particularly polysemous, while both be going to and is to have ranges of uses which are not readily subsumed under one 'meaning'. Nevertheless, alternation between will and be going to seems to rest primarily on distinctions in temporal orientation and stylistic variation. While all these items behave idiosyncratically, it is possible to single out some common factors that characterize the periphrastic items in contrast to will.

Firstly, they are

semantically narrower, most obviously in the case of be willing to. Secondly, they can be seen as 'objective', in that details of the source or motivation for the action are relatively explicit in contrast to the normal vagueness of the modals. Thirdly, unlike true modals they are not constrained by the general incompatibility of modality and explicit factuality. Stylistic differentiation is crucial in the case of be going to, but not evident for the other items.

206

5.4

Can and be able to

While the true modals are often treated as polysemous, however difficult the definition and distinction of the various 'meanings' may be (will is the best example of this), can is often seen as having three major meanings - 'possibility', 'permission' and 'ability' that are in some way related, perhaps together with other minor meanings that can be subsumed under, or associated with, these. The closeness of these meanings, together with the claimed lack of an epistemic sense (this ignores the use of can't in That can't be true, which is semantically the propositional negation of epistemic That must be true), has suggested that can is, untypically for the modals, appropriately seen as a monosemous item (Leech & Coates 1980). Leech and Coates argue that the uses of can can be treated in terms of the gradients of restriction, which ranges between 'possibility' and 'permission', and of inherency, which ranges between 'possibility' and 'ability'. 32 This range of meanings can be shown in a diagram, derived from Leech & Coates (1980:85): (5.139)

possibility (restriction) permission

(inherency) ability

This diagram suggests that 'possibility' forms a kind of semantic focus and functions as the unmarked meaning (1980:85). Essentially the same approach is followed by Coates (1983:Ch.5), though here 'permission' and 'ability' are seen as two distinct core meanings, with 'possibility' as the overlapping weak peripheral area between the strong cores (this interpretation involves polysemy rather than monosemy). Be able to clearly relates closely to the 'ability' use of can, but, unlike, for example, be willing to, another transparent periphrastic, it has a wider semantic range than its elements suggest. In the rest of this section I shall first investigate the availability of be able to for the meanings expressed by can (5.4.1), and the suppletive functions be able to performs (5.4.2). Then I shall examine the 'actuality hypothesis' in relation to certain restrictions observed in the use of can- be able to (5.4.3), and briefly consider stylistic aspects (5.4.4); a conclusion will follow (5.4.5).

5.4.1

The semantic range of be able to

Firstly I shall see how far be able to can be used along the gradients depicted in (5.139) above, assuming 'ability' to be an appropriate semantic base. This question was raised 32

The 'permission', 'possibility' and 'ability' uses of can can, roughly, be equated with the 'deontic', 'neutral dynamic' and 'subject-oriented dynamic' uses, respectively, in Palmer's account (1990:Chs 4 L· 5).

207 briefly in 2.3.2.1, where it was claimed that be able to, alongside be allowed to among other items, has doubtful subject-auxiliary dependence, with examples showing that rooms and buses, at least, can be able to do things, with a possible constraint lying in the ease with which the grammatical subject can be related to personal activity; however, some uses of can, for example in the existential modality of Lions can be dangerous (Palmer 1990:107), or the EM of You can't be serious!, absolutely resist replacement by be able to. Coates (1983:124-125), in contrast to the majority view, claims that be able to is capable of covering the entire semantic range of can and gives examples showing be able to in all the three meanings in (5.139). However, it will be shown below that this certainly does not mean that any occurrence of can can be replaced by be able to. To test the availability of be able to for the meanings associated with can, I shall consider the examples with can provided by Coates (1983:Ch.5). It might be expected that the availability of be able to in these contexts would increase with explicit and unambiguous 'ability'-like meanings, and decline with 'permission'-like meanings, or that, in terms of (5.139), it would decline as the meaning moved first up the inherency and then down the restriction gradient. Coates (1983:92) lists eight occurrences of can in inherency meanings extending from core (ability) uses to periphery (possibility) ones. The expected result is, broadly, produced: in the extreme core cases, as in (5.140)

It is now getting quite difficult to find choirboys . . . who can still sing treble.

be able to is equally suitable; while in the cases Coates terms more indeterminate, such as (5.141)

Every believer can be a faithful distributor of the gospel.

(5.142)

You can't see him because he's having lunch with a publisher.

the use of be able to introduces some change in the meaning. Thus, in (5.141) be able to would focus on a believer's personal ability as opposed to what he is, in some undefined way, in a position to do; while in (5.142) it would make the salient interpretation not a statement of impossibility or prohibition but rather a statement of the addressee's personal lack of ability. In other words, the use of be able to in such texts would produce an 'ability'-like meaning that would be either inappropriate or overspecific. Consideration of examples of the restriction gradient (Coates 1983:88) produces rather similar results. Thus, in the following case of core permission: (5.143)

You can start the revels now.

be able to would simply have the wrong meaning. However, in the following case, also illustrating core permission: (5.144)

Poppy · · · can't drive (her car) because she hasn't got any insurance on it

be able to seems natural. This is because an 'ability' meaning here is just as reasonable as a 'permission' one, since the legal basis for the prohibition, which determines the inability, is explicit in the text; that is, there are two interpretations for can't: 'isn't allowed to, in that she hasn't got any insurance' (permission), and 'isn't able to, because she hasn't got any insurance' (ability). On the other hand, if the insurance problem was not referred to,

208 the sense of can would remain vague, and the use of be able to would have a narrowing or over-specifying effect. The basic meaning of can, then, can be characterized as 'possibility' or 'lack of restriction' though this in specific contexts may be interpreted as, for example, 'permission' or 'ability', whereas be able to is transparently concerned with 'ability' and is available for use where this sense (which is ultimately relatable to human, or at least animate, faculties) is both appropriate and not overly narrow or specific.

5.4.2

Suppletive uses of be able to

There are evident restrictions on the availability of can (and could) in certain types of non-present time reference and non-finite use, and in such cases it may be assumed that be able to can fulfil a suppletive role. Thus, examples (3.4)-(3.7) showed how in a context where both can and be able to might be assumed to be semantically available, explicit future reference, as in (3.6)-(3.7), is only expressible with will be able to. This situation parallels that for must and have (got) to, discussed in detail in 4.5.1, and needs no further comment here. Other examples of the presumed suppletive role of be able to in relation to can, similarly paralleling the have (got) to-must situation, appear in the following texts from the CEC: (5.145)

nevertheless, to judge from such of the letters as I've been able to, to get hold of and look at closely, they've never committed themselves to anything (CEC:58/133-135)

(5.146)

we had reached the point of thinking that we weren't going to be able to reach a policy decision (CEC:59/167-169)

In these cases, there seems no doubt that if the relevant can form were available it would be used, and be more normal than the be able to option. Thus, in relation to (5.145), the speaker would say, in a present time context, I can't get hold of all the letters, and I'm not able to get hold of all the letters would simply seem an overly explicit formulation. Similarly, in relation to (5.146) a speaker would naturally say We can't reach a policy decision. The situation with past time reference is, of course, not parallel to that with must and have to, since both forms could and was able to are available; but they are no longer interchangeable in the same way. The restriction on the use of could as opposed to was able to was discussed briefly in 3.6, the relevant data appearing in (1.13)-(1.14) and (3.148)-(3.149). The problem here centres on cases like the following: (1.13)

*I ran fast, and could catch the bus.

Various proposals have been made in connection with the obligatory replacement of could in (1.13) by was able to and other locutions, and the issue will now be discussed further.

209 5.4.3

Could and actuality

The next subject focuses on the often observed gap in the modal system that makes (1.13), quoted above, markedly odd. The obligatory use of be able to (or some other verb such ás succeed) for the expression of the intended sense here constitutes a well known learning problem, as is attested by Quirk (1981) and Swan L· Smith (1987). However, the restriction involved here has been found to have wider implications, both for the uses of the items can and be able to and for other M-P pairs. In 3.6 various comparable approaches to this problem were discussed, and it was argued that Palmer's 'actuality hypothesis' offers the most explanatory potential. I shall start here by exploring the precise nature of the restriction on the use of could, in relation to was able to, since this is not a simple matter. The following texts, the first two introduced in 3.6, will be discussed: (3.154)

I could just reach the branch.

(3.155)

But he got the envelope open eventually . . . Several things fell or could eventually be wrested out. (Amis 1991:183)

(5.147)

Mr. Thies, born in 1944, is an historian and journalist who floated into the centre of government power almost by chance and drifted out of it again, evidently without much regret. For nearly two years he could watch the chancellor closely. Some of his findings . . . (The Economist, 11.2.1989:97)

(5.148)

. . . the folk singer, who could persuade President Carter in 1979 to send the Seventh Fleet and the U.S. Air Force to rescue the Vietnamese boat people . . . (The Observer, 21.2.1988:64)

(5.149)

luckily, Juliane's mother could come up at short notice (conversation)

The use of could in (3.154) seems to conflict with the restriction illustrated in (1.13), and Palmer relates this to its use in sentences such as the following: (5.150) (5.151)

I could almost reach the branch. (Palmer 1990:95) He was laughing so much, he could hardly get a word out. (Palmer 1990:94)

All three cases axe seen as essentially negative-oriented, since they occur in 'semi-negative' contexts: "What is relevant is both that there is a negati ve-type word in the context and that that the implication is that, even if the event took place, it almost did not" (op cit:95). Further, the fact that, despite the normality of (3.154), the following variant of (1.13) is markedly odd: (5.152)

?I ran fast and could just catch the bus.

210 is commented on as follows: "It is where could is closest to did that it becomes most unacceptable"; by contrast, "one can be in a position to reach a branch . . . for some time" (op cit:96). All this suggests that the contextual factor crucial for the appropriate use of could when performance of the proposition is involved is an indication that the event (a) 'almost' did not occur or (b) was potential for an extended period. These two conditions have in common that an actual individual performance is not the focus of attention, even if it is clearly implied. Text (3.155) is similar to (3.154) in that the single achievement referred to - getting the contents out of the envelope - proved difficult. If reference to this difficulty is removed, the result is rather odd: (5.153)

? . . . Several things fell or could be taken out.

A modification like the following - with could explicitly in a 'single achievement' sense is even odder: (5.154)

? . . . The letter could be taken out of the envelope.

This is because (5.154) unambiguously involves a single achievement, while (5.153) is vaguer in that several things are involved. In the next case, (5.147), could watch denotes actuality (since findings are involved) extending throughout a specific period (For nearly two years) without indication of any difficulty of achievement. However, there is no reference to an individual action, and the sense is, rather, 'He was in a position to watch', with the implication ' . . . and this is what he did all the time'. That is, the action is generalized, so that the factuality is in one way minimized. (5.148) differs in referring unambiguously to a single achievement; here, it seems that the sense of could persuade is something like 'the folk singer, who was the sort of person who would, and incidentally, did persuade . . . '. This would focus on the potential rather than the actualization. We can compare this with an extension of one part of the text again, specifically in a 'single achievement' sense: (5.155)

? . . . the U.S. Air Force could rescue the Vietnamese boat people.

If this is odd it is because it is not easy to interpret could rescue as denoting generalized potential. Rather similarly, in (5.149) could come up can be seen as meaning 'it was possible for her . . . , and she actually did', and here obstacles that might have to be overcome are implied. In all these cases, (3.154)-(3.155) and (5.147)-(5.149), the use of was able to would make little difference, but would focus rather more on the achievement. For example, if we adapt (5.148) as follows: (5.156)

. . . the folk singer, who was able to persuade President Carter . . .

the sense now is more clearly ' . . . was strong enough to, and did' than ' . . . was the sort of person who would, and did'. The above data show that could can, in certain contexts, coexist with actual occurrences, and even with a unique occurrence; however, it is crucial to its use then that the

211 meaning focus is primarily on something other than the unique performance of an action. By contrast, was able to can strongly imply individual realization. The fact that be able to, at least in its past tense forms, so markedly implies actuality must be seen as an idiosyncratic feature of the verb. Palmer extends the actuality hypothesis to present/future uses as well, and argues that an implication of present time actuality is excluded for can, which refers to future performance. This derives from the basic incompatibility between modality and factuality (cf (3.152)-(3.153)). Palmer cites examples such as the following: (5.157)

By bulk buying in specific items, Lasky's are able to cut prices on packages by as much as 30% or so. (1990:90)

(5.158)

Liverpool can win the cup next year. (1990:99)

(5.157) is said to imply actuality in present time; can "would not be impossible" here, but be able to is preferred. The reference here seems to be to general/habitual activity (which can subsume individual occurrences) rather than to an individual action, and parallels the use illustrated in (3.147) 3a-b (I was able to/could, swim in those days): both verbs are possible in each case. (5.158) is taken to refer to present ability that is expected to be actualized individually in the future, and as such would not be quite equivalent to either is able to (which is an inappropriate tense) or will be able to (which would indicate future ability with possible future actualization). With general/habitual activity (eg, Liverpool can win when they really try), the temporal reference is vague but present-based, and here are able to would be equally possible. Coates provides two instances of is able to where can is actually excluded, and these merit close examination; one follows: (5.159)

(film synopsis) The prosecutor is not concerned with him as an individual and is himself quite convinced of his guilt. But in the end a friend is able to prove the man's innocence to the satisfaction of the court officials. (1983:127)

It is significant that this and the other text quoted, which is from literary comment, axe concerned with discussing works of art, since this constitutes a special type of historic present (Quirk et al 1985:181-182). Reference to an individual occurrence in present time, for example in a simultaneous commentary, similarly requires be able to (eg, He moves fast, and he's able to return the ball, but hardly ...he can return the ball). These cases of is able to - where can seems impossible - are, in fact, exact parallels to the past tense situation illustrated in (1.13) and (1.14). The contrast involved in these examples is between can meaning 'can and (weakly implied) may do in the future/any time', is able to meaning 'can and (strongly implied) does, generally/individually)', and will be able to meaning 'will be able to and (implied) will do'. This can set out as follows:

212 (5.160)

item

time of ability

time of potential fulfilment

can is able to will be able to

present present future

future/general present present future

This situation is explained within the actuality hypothesis as follows (1990:91,195-196): be able to has an implication of simultaneous actuality, while can does not readily refer to events whose factual status is or can be established, and ail implication of future actuality is necessarily non-factual (there is some here parallel with the options for must and have to in future reference (cf 4.5.1)). This can be shown with analyses of the following set of sentences using the notation set out in 3.5.4 and used in 4.5: (5.161)

I can go there any time I like.

S,M,E

(5.162)

I'm able to go there any time I like.

(5.163)

I can go there tomorrow.

S,M,E S,M - E

(5.164) (5.165)

?I'm able to go there tomorrow. I'll be able to go there tomorrow.

S, M - E S-Μ,E

(5.166)

?I can go there when I'm 64.

S-

(5.167)

?I'm able to go there when I'm 64.

S-M,E

(5.168)

I'll be able to go there when I'm 64.

M,E

S-M,E

In view of the restrictions evident in these sentences with present tense forms, we can say that can is inappropriate for S - Μ,E, and am able to for a structure where S and E are temporally distinct (ie, both S,M - E and S - Μ,E), while will be able to specifically requires the distinctness of S and E, as in S - Μ,E or S - M - E.

5.4.4

Style and formality

In 3 . 7 the argument was set out that periphrastics, in general, are less formal than their modal equivalents, particularly when phonologically reduced. This claim has been discussed in connection with, in particular, will and be going to (5.3.1.2). No evidence is offered to support the claim for can and be able, and the situation is made more difficult by the fact that the contrary claim is also made - that be able to is a relatively formal item in relation to can: Palmer (1990:90) cites occurrences of be able to in writing where can would be expected in speech; but it may be suspected that in such cases the use of be able to is simply designed to achieve more specific meaning, in contrast to the marked vagueness of can. There is, in brief, no clear indication that a formality factor is generally significant here.

213 5.4.5

Conclusion

The contrast between can and be able to is rather different from those found for other pairs and sets of items, which justifies LakofF's exclusion of the pair from her original proposal (cf 1.1 and the list in (1.3)). However, in a more general sense, the distinctions in use observed between can and be able to as regards the expression of actuality are to be related to the basic incompatibility between true modals and factuality, discussed in 3.6, and depicted in (3.153). It is not clear that a distinct 'actuality hypothesis' is, in fact, needed to account for the behaviour of these items; on the other hand, it must be seen as an idiosyncratic feature of be able to that it readily relates to actualization, and thus is able to fulfil what could be considered a suppletive role in relation to can/could. There is no evidence that a formality factor is relevant to the choice of item here - in this respect, too, can-be able to seem to differ from some other M - P pairs.

6.

Conclusions

The aim of this study was to investigate the functioning of periphrastic verbal items in relation to the modals they seem to relate to, and, more specifically, to consider the possibility that there are systematic distinctions to be observed between the use of the two sets of items. The focus of the investigation was the proposal, widely expressed in various forms but particularly associated with Larkin and Lakoff, that modals express speaker-related meanings and intentions, while periphrastics express modal values that arise externally or independently of the speaker. At the same time, we were dealing with pairs or sets of items that are demonstrably close to synonymy, and sometimes apparently fully interchangeable. For a number of reasons explained in the text, the must/have (got) to set was used for detailed investigation, and, partly on the basis of the findings for these items, other M - P pairs and sets were treated more briefly. Any comments on this subject assume a well-defined type of M - P relationship. In fact, however, the ideal M - P pair does not exist: the most clear-cut examples are probably can-be. able to and can/may-be allowed to (which was not discussed here), but these are complicated by, among other things, the existence of the can/could pairing. To take a couple of other examples, must-have to-have got to, although the best documented case, form ail idiosyncratic group in a number of ways - eg, the complex relationship between the three items, the ways in which a formality factor is relevant, and the occasional availability of must for past time reference. Should-ought to, next, form a unique pair: ought to has the highest claim among non-central modals to a quasi-modal status, and it and should are widely felt to constitute the closest to a synonymous pair, a fact reflected, for example, in the availability of should for use as a tag-question for ought to; nevertheless, it seems reasonable to treat ought to as a periphrastic-like item in relation to should. In general, the items that might count as periphrastics form a strikingly heterogeneous collection, and for this reason a number of tests were applied in 2 in order to identify potential members of such a class, and these show that, although there are several syntactic, semantic and phonological features that tend to distinguish such items, any categorization or choice among them must ultimately be arbitrary. Despite all this, the phenomenon of alternating modal and 'periphrastic' items is a marked feature of English, and as such merits examination. The most basic general point to be made about the pairs and groups discussed, one which in various ways can subsume some of the other distinctions observed, is the marked status of periphrastics relative to their corresponding modals (cf preliminary discussion in 3.8). In the case of the modals, this is most obviously reflected, morphosyntactically, in the fact that they have just one form (or, in some cases, two, if pairs like will-would are treated as related present and past tense forms), and, semantically, in their essentially vague or minimally specified meanings (eg, must as a general marker of necessity or obligation, and can as signalling 'nihil obstat'); these aspects of the modals, in relation to all other modal exponents, are set out in detail by Perkins (1983:Ch.8). In spoken language, this general indeterminacy of the modals, when prosodically non-salient, as they

215 normally are, can be contrasted with the various more specific interpretations that become possible with intonational prominence (Brazil 1995:117-118). In contrast to the modals, periphrastics have, in the ideal case, all the forms of a full verb, and fulfil suppletive functions - in supplying, most obviously, nonfinite forms, but also finite forms for past and certain types of future reference, while they have, in various ways, more highly specified meanings. Determining when periphrastics really are fulfilling such roles is not a simple matter; evidence on this was discussed at various points in the text. I shall briefly review how this claim for semantic markedness can be seen to apply to the items examined. In 4, it was argued that have to and must both refer to a necessity or obligation, while have to differs, essentially, in implying the existence of some specifiable source, a feature have got to shares to a certain extent. Need (to), which in any case cuts across any clear-cut M - P contrast in having both auxiliary and quasi-modal forms, differs from the other items in focusing on the motivation for a requirement. The final item in this group, be bound to, with a sense like 'unquestionability' or 'inevitability', has a more specific meaning than must, but it does not relate significantly closely to it. In 5, the situation involving should and three other items was found to be basically similar. Whereas should expresses a general notion of what can be expected, ought to suggests, more specifically, that there is some reason for this expectation; with be supposed to the focus is on the fact that an expectation is held - ie, the passive form is reflected in the implication of an agent; while had better, a highly anomalous item, states that a specific, individual action is rather definitely proposed. The case of will in relation to be going to, is to and be willing to is fairly clear in general terms: it is a highly polysemous item, while the other three only correspond to, or overlap with, it in certain semantic areas.

The precise relationship of will and be going to, however, is problematic here,

since the dominant contrast between them - ignoring the stylistic issue for the moment - lies in the nature of the time reference or temporal orientation involved. But this too can perhaps be seen in terms of markedness: will - in its relevant uses - malees plain predictions, or simply refers to future time, while be going to is concerned with a more complex relation between present and future time. The contrast with is to is, similarly, in their expression of futurity - with is to the future event is strongly expected or required but this merges into a sense of obligation; while be willing to contrasts with will solely in its transparent volition sense. Finally, can has a much wider, vaguer semantic range than be able to, and here the periphrastic can be seen as marked in that it specifies 'ability'like qualities of the subject. The restricted power of can to refer to events, as opposed to states, in contrast to the lack of such a restriction in be able to, a feature reflected to a certain degree in other M - P pairs, could be considered to show markedness; however, this feature of can is probably better seen as ultimately reflecting the indeterminacy typical of modals. Some general points of contrast have been axgued for from among these various factors. Firstly, it has ofteïi been proposed that the modals can be categorized as 'subjective' and periphrastics as 'objective'. The term 'subjective' can, minimally, be taken to mean that the speaker's viewpoint or normal commitment to a proposition is not seriously constrained by external considerations, and 'objective' can then be applied to an expression which focuses on or assumes the existence of some motivating force, normally outside the

216 speaker; the senses here may be relative - that is, one item is, standardly, objective in relation to another, which may in turn be seen as less objective or as subjective in comparison. In such senses, the modals can generally be categorized as subjective in relation to the periphrastics. Such uses of the terms can readily be related to notions of markedness, since the objectivity of periphrastics involves more highly specified meanings. However, caution is required with concepts of subjectivity and objectivity. A number of claims have been made in this area which are simply too strong (cf, especially, the discussions in 3 . 3 and 3 . 4 ) . For example, some statements concerning deontic sources, in particular those that treat modals as necessarily expressing speaker origination, are simply untenable, while notions that subjectivity and objectivity constitute a clear-cut dichotomy, relatable to directly contrasting syntactic and semantic structures, do not receive adequate support in the data examined here. Further, in some M - P cases the relationship in this respect is far from simple. For example, there is some justification for regarding should as subjective in relation to ought to, but there is much more for regarding both those items as subjective in relation to be supposed to; while is to is not easy to categorize in this respect in comparison to the other three items. Elsewhere, for example with the will-be going to pair, such a distinction or contrast simply does not seem relevant. Notions of subjectivity and objectivity are also relatable to various theories which involve the speaker's point of view or empathy ( 3 . 3 . 2 . 2 - 4 ) : these broaden the perspective of speaker involvement to include sympathy both for deontically involved participants and, more generally, for the modality of propositions. These ideas were reflected to some extent in various descriptive proposals discussed in 3 . 3 . 5 , and receive some support in the descriptive chapters above. The 'actuality hypothesis', which was discussed at several points above, claims to account for a number of points of semantic contrast observable between various M - P pairs. It was found that this hypothesis can be substantiated, to varying degrees, and seems to deal better than other proposals with the relevant phenomena, but it was suggested that it in fact reflects a more deep-rooted feature of modality and the use of modal verbs in English - that, being essentially concerned with alternative possible worlds, they do not naturally coexist with factuality. This, together with certain conventionalized uses of the relevant periphrastics, seems to lies behind the phenomena that the actuality hypothesis and similar arguments seek to explain. The relevance of questions of language style and variety was outlined in 3 . 7 and has been shown to be crucial to several areas of contrast here. It is quite clear that a formality/informality parameter may be relevant to the selection of an item, and that some periphrastics are more characteristic of informal, conversational style than are their related modals. This can result in situations where, apparently, a true modal would be chosen in a formal situation and a periphrastic in an informal one, and this naturally raises the question whether such a parameter of variation can in any way be related to other distinctions such as have been outlined above. Here there is no simple answer. Two approaches to this seem possible: either a kind of complementary distribution is assumed, with the two items meaning essentially the same in different types of context of use, or the actual meanings of the items are seen as partly relatable to or determined by the level of formality of a context. To take the items must and have (got) to, an approach

217 that assumes complementary distribution would suggest, for example, that must is used in a formal context for meanings that have to or, in particular, have got to expresses in an informal one, while, from another perspective, must in a colloquial context might be restricted to an epistemic interpretation, but in a formal context it would be equally capable of deontic and epistemic senses. In the second approach, which allows a causal connection between levels of formality and the meanings typically expressed by items, it might be argued, to take the will-be going to pair, that the meanings expressed by be going to are typical of colloquial discourse, that is, that the form itself, metaphorically indicating movement into a future state, is appropriate for the expression of intentions, comments and informal observations relating to upcoming time that are close to the speaker's immediate interests, and that such uses are typical of colloquial style, in contrast to more specific and objective ways of referring to intentions, future time, etc. But this must remain pure speculation. Despite cases such as those just discussed, there is no evidence that a formal-informal contrast applies generally to all M - P pairs/sets. Variety differences, primarily between B E and AE, were not intended as a central part of the investigation, but in general relatively little significant difference of this kind was noticed. This account was primarily sentence-based, but· a large number of descriptive details were demonstrated within extended contexts, and this provided ample evidence of the need to take account of a discourse level in interpretation. Discourse considerations involve not only questions of the relevant use and appropriacy of specific items within particular types of text, but also the operation and interaction of items within extended discourse - for example, the way in which a periphrastic may be used to 'set' a text, and then be optionally followed by related modals, a phenomenon that seems to be shown by, at least, have to-must and be going to/is to-will. On a detailed descriptive level, the findings suggest that several assumptions widely made about the use of these items simply lack adequate evidence, or are contradicted by available data - eg, in the case of must, deontically, claims regarding its standard use to initiate demands or orders and its non-availability for habitual reference, and, epistemically, its nonavailability for future reference or in question forms; while, of the must/have to/have got to group, there are indications that have got to can come closest to performative-like status. The common claim that ought to is totally synonymous with should (in relevant senses) and, in any case (perhaps in consequence), moribund cannot be upheld. In the case of will-be going to, failure to note that any contrasts observed between the items are only applicable to relatively colloquial style may lead to misleading assumptions about the semantics of those items. More generally, it was shown that making a sharp-cut dichotomy between epistemic and non-epistemic uses of items is sometimes hard to justify (see below); in any case, there is a lack of consensus as to the precise definition of epistemic modality, particularly in relation to its possible use for future time. Some other concepts used in relation to these items must also be viewed with caution, eg, the notion of a distinct variety of performative modality, or the claim that some items are inherently stronger than others. In general, it is clear that modals, and to a lesser degree, periphrastics, are highly idiosyncratic items

218 in every respect, and the impression is gained that, whether for descriptive simplicity or for theoretical tidiness, some scholars and grammarians work with grossly oversimplified notions about them. Certain patterns emerge that have no obvious significance for the whole field, for example the close association of some items with unique reference, eg, must and had better, and of others with general reference, eg, should and have to. This investigation assumed a predominantly monosemantic approach to word meaning, with the following roughly distinguishable features of interpretation, as outlined in 3.5.2: (a) the identification of the type of modality, (b) certain general contextual inferences, and (c) further conventionalized, fossilized inferences. The first feature involves, for most items, primarily determining whether a specific modal use is deontic or epistemic; this choice has certain typical syntactic and semantic consequences - for example, EM is associated, in terms of co-occurrence probabilities, with stative verbs and perfective aspect. These associations are not in any sense binding, but they may create stereotypical expectations in use. However, there axe reasons for not treating ail epistemic-deontic split as producing two radically different items, but, rather, for emphasizing the semantic unity of relevant items for reasons such as the following, which were all illustrated in the text above. (a) The root/deontic vs epistemic contrast is a widely established phenomenon in language and can be seen in terms of the regular metaphorical extension of meanings - where such meanings are appropriate - from an external, real world to a parallel internal, mental world (Sweetser 1990:Ch.3). (b) Despite cases like may and must, where the deontic and epistemic readings are syntactically and semantically highly differentiated, in other cases such as should and be supposed to it is often difficult to determine whether an individual occurrence should be treated as deontic or epistemic, while such a distinction is commonly blurred in future reference and is sometimes obscure in past reference. (c) There axe problems involved in setting up syntactic and semantic distinguishing features for the alternative readings; for example, perfective aspect is not excluded for deontic uses, but is simply more much commonly found epistemically. (d) Apart from the strict root/deontic-epistemic dichotomy, there is often difficulty in distinguishing between various senses of items, eg, in the cases of will and can and related periphrastics. The contextual inferences referred to above determine dominant uses of items, and here the notion of default interpretation is crucial. At this level of interpretation, numerous features typical of the use of individual items are pragmatically inferred, eg, speaker-centredness, association with habitual activity, uniqueness of reference, urgency of performance, expectation of non-fulfilment, potentiality, sympathy or a moral attitude. Through such processes, modals and periphrastics also tend to develop certain fossilized, idiomatic meanings, such as were discussed at various points above. In this connection it was also noted that modals and modal expressions, though basically very vaguely specified, axe often associated with specific illocutionary forces. With the kind of variation that all this allows, it is sometimes difficult to determine what the basic or core meanings of items are: here will, is to and had better, for example,

219 seem much more problematic than must and be able to. Questions of basic/core meanings relate to other complex issues involving stereotypes, prototypes and acquisition processes; but these lie outside the concerns of this study. A final question related to word meanings concerns the ways in which they are presented: in the area of modality, glosses are notoriously hazardous, since they risk on the one hand being meaninglessly obscure, and on the other being misleadingly specific. Descriptive metalanguage is commonly vague and impressionistic, but this may be unavoidable: there seems to be no reason why glosses should be expected to be easily formulated, particularly for items with highly complex functions such as modals and periphrastics. This also has pedagogical consequences, since the relevant facts of usage are not easily summarized in simple statements and rules (Westney 1994). Questions of data and methodology were discussed in 1.4-5, and will now be returned to briefly. It is widely recognized, though not universally accepted in practice, that a complex area such as modality crucially requires ample natural language data as opposed to the invented products of intuition; the absolute inadequacy of dealing with any items simply in terms of invented, context- and style-free contrasting sentence pairs should be fully apparent. A number of factors bear on this: firstly, several recent treatments of this general area have adopted a corpus-related, if not corpus-based approach, often with striking results (eg, Hermerén 1978, Coates 1983, Palmer 1990); secondly, the wide variety of data and accounts offered in the literature suggests the need for extreme caution; and thirdly, data have during the past two decades become publicly available in ways previously unknown. Moreover - crucially - the hazards of depending on native-speaker intuition as a sole authority for details of usage are well documented - eg, Preston (1989:3) comments "While there are areas of research where intuitions serve linguistics, one place where they serve nothing is in the area if direct, objective language use", while Davies, in a monograph on the native speaker (1991:14), refers to the "chimera of intuition". In the present study, a combination of methods was used for purely practical reasons: some items can easily be documented in quantity, while others, or specific uses of them, are simply rare and hard to attest. For this reason, the conversational corpus was always used as a basic point of reference, but other sources of data, including the relevant literature and, inevitably, intuition, as the ultimate arbiter, were drawn on.

Bibliography Aarts, F.G.A.M. h H.C. Wekker. 1987. A contrastive grammar of English and Dutch. Leiden: Nijhoff. Aijmer, K. 1980. Evidence and the declarative sentence. Stockholm: Almqvist & Wiksell International. Alexander, L.G. 1988. Longman English grammar. London: Longman. Allen, W.S. 1959. Living English structure. 4th ed. London: Longman. Altman, R. 1986. Getting the subtle distinctions: Should vs had better. Studies in Second Language Acquisition 8:80-87. Amis, K. 1991. The folks that live on the hill. London: Penguin. Anderson, J. 1971. Some proposals concerning the modal verb in English. In A.J. Aitken, A. Mcintosh & H. Pálsson, eds. Edinburgh studies in English and Scots. London: Longman, 69-120. Antinucci, F. h D. Parisi. 1971. On English modal verbs. Papers from the 7th Regional Meeting of the Chicago Linguistic Society:2S-39. Ballard, P.B. 1939. Teaching and testing English. London: University of London Press. Binnick, R.I. 1971. Will and be going to. Papers from the 7th Regional Meeting of the Chicago Linguistic SocietyAQ-52. -

1972. Will and be going to II. Papers from the 8th Regional Meeting of the Chicago Linguistic Society Bolinger, D. 1977. Meaning and form. London: Longman. 1980. Wanna and the gradience of auxiliaries. In G. Brettschneider L· C. Lehmann, eds. Wege zur Universalienforschung. Tübingen: Narr, 292-299. Bouma, L. 1975. On contrasting the semantics of the modal auxiliaries of German and English. Lingua 37:313-339. Boyd, J . L· J.P. Thorne. 1969. The semantics of modal verbs. Journal of Linguistic 5:57-74. Brazil, D. 1995. A grammar of speech. Oxford: OUP. Brown, K. 1991. Describing modality in English. In R. Bowers h C. Brumfit, eds. Applied linguistics and English language teaching. London: Modern English Publications and the British Council, 109-129. Bybee, J.L. h W . Pagliuca. 1985. Cross-linguistic comparison and the development of grammatical meaning. In J. Fisiak, ed. Historical semantics. Historical word-formation. Berlin: Mouton, 59-83.

1987. The evolution of future meaning. In A.G. R a m a t , O. Carruba h G. Bernini, eds. Papers from the 7th International Conference on Historical Linguistics. Amsterdam: Benjamins, 109-122. Bybee, J.L., W . Pagliuca & R.D. Perkins. 1991. Back to the future. In E.C. Traugott h Β. Heine, eds. Approaches to grammaticalization. Vol.2: Focus on types of grammatical markers. Amsterdam: Benjamins, 17-58. Carter, R. h M. McCarthy. 1988. Vocabulary and language teaching. London: Longman. Caton, C.E. 1966. On the general structure of the epistemic qualification of things said in English. Foundations of Linguistics 2:37-66. Celce-Murcia, M. h D. Larsen-Freeman. 1983. The grammar book. Rowley: Newbury House. Chalker, S. 1984. Current English grammar. London: Macmillan. 1990. English grammar word by word. Walton-on-Thames: Nelson. Chapin, P.G. 1973. Quasi-modals. Journal of Linguistics 9:1-9. Chomsky, N. 1957. Syntactic structures. The Hague: Mouton. Chung, S. k. A. Timberlake. 1985. Tense, aspect and mood. In T . Shopen, ed. Language typology and syntactic description. Vol.3: Grammatical categories and the lexicon. Cambridge: CUP, 202-258. Close, R.A. 1977. Some observations on the meaning and function of verb phrases having future time reference. In W.-D. Bald L· R. Ilson, eds. Studies in English usage. Frankfurt: Lang, 125-156. 1981. English as a foreign language: its constant grammatical problems. 3rd ed. London: George Allen & Unwin. 1988. The future in English. In W.-D. Bald, ed. Kernprobleme der englischen Grammatik. Sprachliche Fakten und ihre Vermittlung. München: Langenscheidt-Longman, 51-66. Coates, J. 1983. The semantics of the modal auxiliaries. London: Croom Helm.

221 1990. Modal meaning: the semantic-pragmatic interface. Journal

7:53-63.

of Semantics

Coates, J . & G. Leech. 1980. T h e meanings of the modals in modern British and American English. York Papers

8:23-34.

in Linguistics

Collins, P.C. 1988. T h e semantics of some modals in contemporary Australian English. Australian

Journal

8:233-258.

of Linguistics

1991. T h e modals of obligation and necessity in Australian English. In K. Aijmer k B . Altenberg, eds. English

corpus

linguistics:

Comrie, B . 1976. Aspect. -

1985. Tense.

studies

in honour

London: Longman, 145-165.

of Jan Svartvik.

Cambridge: CUP.

Cambridge: CUP.

Cowper, E.A. 1992. Infinitival complements of have. Croft, W . 1990. Typology and universals. Cruse, D.A. 1986. Lexical

Journal

in applied

Edinburgh: Edinburgh UP.

linguistics.

Dirven, E . k G. Radden. 1977. Semantische

Syntax des Englischen.

Downing, A. k P. Locke. 1992. A university

course

Drubig, H.B. 1993. Modalverben.

37:115-135.

of Linguistics

Cambridge: CUP.

semantics.

Davies, A. 1991. The native speaker

Canadian

Cambridge: CUP.

in English

Wiesbaden: Athenaion. New York: Prentice Hall.

syntax.

Unpublished notes.

Edmondson, W . , J . House, G. Kasper k J . McKeown, J . 1977. A pedagogic

grammar

of the English

verb.

Tübingen: Narr. Ehrman, Μ . E . 1966. The meanings

of the modals

Firth, J . R . 1957. Modes of meaning. In Papers

in present-day in linguistics

Fleischman, S. 1982. The future in thought and language:

American

diachronic

T h e Hague: Mouton.

English.

London: OUP.

1934-1951.

evidence

Cambridge:

from Romance.

CUP. Francis, G. 1993. A corpus-driven approach to grammar: principles, methods and examples. In M. Baker, G. Francis k E. Tognini-Bonelli, eds. Text and technology.

In honour

Amsterdam:

of John Sinclair.

Benjamins, 137-156. Fraser, Β . 1975. Hedged performatives. In P. Cole k J . L . Morgan, eds. Syntax Speech

and semantics.

Vol.3:

acts. New York: Academic Press, 187-210.

Gailor, D. 1983. Reflections on should,

ought to, and must. English

Language

Teaching

Journal

37:346-

349. Gimson, A.C. 1980. An introduction

to the pronunciation

Gramley, S. k K.-M. Pätzold. 1992. A survey

of English.

of modern

English.

Greenbaum, S., G. Leech k J . Svartvik, eds. 1980. Studies

3rd ed. London: Arnold.

London: Routledge.

in English

linguistics

of English

Linguistics

for Randolph

Quirk.

London: Longman. Haegeman, L. 1980. Have to and progressive aspect. Journal 1981. The use of WILL

and the expression

of futurity

in present-day

14:1-5.

British

English.

P h D Diss,

University of Ghent. Unpublished. 1983. The semantics

of 'will' in present-day

British

English:

a unified account.

Brussels: Verhandeling

Letteren. 1987. 'Be going

to' and

'will':

a pragmatic

Duisburg:

account.

Linguistic Agency University of

Duisburg. 1989. Be going to and will: a pragmatic account. Journal

of Linguistics

25:291-317.

Hakutani, Y . k C.H. Hargis. 1972. T h e syntax of modal constructions in English. Lingua

30:301-332.

Halliday, M.A.K. 1970. Functional diversity in language as seen from a consideration of modality and mood in English. Foundations

of Linguistics

6:322-361.

Harris, M. 1986. English ought (to). In D. Kastovsky k A. Szwedek, eds. Linguistics and geographical

boundaries.

Vol.1: Linguistic

theory

and historical

linguistics.

across

historical

Berlin: Mouton de

Gruyter, 347-358. Haynes, C.S. 1967. A grammar American

English.

of modal

Heny, F . k B. Richards, eds. 1983. Linguistic Reidel.

and catenative

auxiliaries

in contemporary

informal

spoken

P h D Diss, New York University. Ann Arbor: University Microfilms International. categories:

auxiliaries

and related puzzles.

Vol.2. Dordrecht:

222 Hetmerén, L. 1978. On modality in English: a study of the semantics

of the modals. Lund: CWK Gleerup.

1986. Modalities in spoken and written English. An inventory of forms. In G. Tottie & I. Bäcklund, eds. English in speech and writing: a symposium. Acta Univ. Ups., Studia Anglistica Upsaliensia 60. Uppsala, 57-91. Hopper, P.J. L· E.C. Traugott. 1993. Grammaiicalization. Cambridge: CUP. Hornby, A.S. 1989. Oxford advanced learner's dictionary of current English. 4th ed. Oxford: OUP. Householder, F.W. 1971. Linguistic speculations. Cambridge: CUP. Huddleston, R. 1980. Criteria for auxiliaries and modals. In Greenbaum, Leech & Svartvik, eds. 1980, 65-78. 1984. Introduction to the grammar of English. Cambridge: CUP· Hughes, A. L· P. Trudgill. 1979. English accents and dialects. London: Arnold. Hunston, S. 1993. I must say I'm a little sceptical: an investigation of a discourse use of some modal auxiliaries. Paper presented at the Annual Meeting of the British Association for Applied Linguistics, 1993. Unpublished. Jacobsson, B. 1979. Modality and the modals of necessity must and have to. English Studies 60:296-312. 1980. On the syntax and semantics of the modal auxiliary had better. Studia Neophilologica 52:47-53. James, P.D. 1986. A taste for death. London: Faber. Jenkins, L. 1972. Modality in English syntax. Indiana: Indiana University Linguistics Club. Jespersen, O. 1924. The philosophy of grammar. London: Allen & Unwin. 1931. A modern English grammar. Vol.IV. London: Allen L· Unwin. 1940. A modern English grammar. Vol. V. London: Allen & Unwin. Jones, D. 1977. Everyman's English pronouncing dictionary. 14th ed. London: Dent. Joos, M. 1968. The English verb - form and meanings. 2nd ed. Madison: University of Wisconsin Press. Kajita, M. 1968. A generative-transformational study of semi-auxiliaries in present-day American English. Tokyo: Sanseido. Karttunen, L. 1972. Possible and must. In J.P. Kimball, ed. Syntax and semantics. Vol.1. New York: Academic Press, 1-20. Kiefer, F. 1986. Epistemic modality and focus. In W. Abraham & S. de Meij, eds. Topic, focus, and configurationality. Amsterdam: Benjamins, 161-179. Krapp, G.P. 1927. A comprehensive guide to good English. New York: Rand McNally. Kratzer, A. 1977. What 'must' and 'can' must and can mean. Linguistics and Philosophy 1:337-355. Kress, G. k R. Hodge. 1979. Language as ideology. London: Routledge & Kegan Paul. Kruisinga, E. 1925. A handbook of present-day English. Pt II: English accidence and syntax 1. 4th ed. Utrecht: Kemink en Zoon. 1931. English accidence and syntax 1. Groningen: Noordhoff. Kucera, H. & W.N. Francis. 1967. A computational analysis of present-day American English. Providence: Brown UP. Kuno, S. 1987. Empathy perspective. Ch.5 of Functional syntax. Anaphora, discourse and empathy. Chicago: University of Chicago Press. Lakoff, R. 1972. The pragmatics of modality. Papers from the 8th Regional Meeting of the Chicago Linguistic Society:229-246. 1981. The 'Dear Robin' column. CATESOL News, January 1981 (quoted in The Linguistic Reporter 23/8:2). Langacker, R.W. 1985. Observations and speculations on subjectivity. In J . Haiman, ed. Iconity in syntax. Amsterdam: Benjamins, 109-150. 1987. Foundations of cognitive grammar. Vol.1: Theoretical prerequisites. Stanford: Stanford UP. 1990. Subjectification. Cognitive Linguistics 1:5-38. Larkin, D. 1976. Some notes on English modals. In J.D. McCawley, ed. Syntax and semantics. Vol.7: Notes from the linguistic underground. New York: Academic Press, 387-398. LDOCE. 1987. Longman dictionary of contemporary English. 2nd ed. London: Longman. Leech, G.N. 1969. Towards a semantic description of English. London: Longman.

223 -

1971. Meaning and the English verb. London: Longman. 1987. Meaning and the English verb. 2nd ed. London: Longman. 1994. Students' grammar - teachers' grammar - learners' grammar. In M. Bygate, A. Tonkyn & E. Williams, eds. Grammar and the language teacher. New York: Prentice Hall, 17-30. Leech, G. L· J. Coates. 1980. Semantic indeterminacy and the modals. In Greenbaum, Leech k Svartvik, eds. 1980, 79-90. Leech, G. h J. Svartvik. 1975. A communicative grammar of English. London: Longman. Levinson, S.C. 1983. Pragmatics. Cambridge: CUP. Lightfoot, D. 1979. Principles of diachronic syntax. Cambridge: CUP. Lyons, J. 1977. Semantics. Cambridge: CUP. 1981. Language, meaning and context. London: Fontana. 1982. Deixis and subjectivity: Loquor, ergo sum? In R.J. Jarvella & W. Klein, eds. Speech, place, and action. Chichester: Wiley, 101-124. 1987. Semantics. In J. Lyons, R. Coates, M. Deuchar & G. Gazdar, eds. New horizons in linguistics 2. London: Penguin, 152-178. Manning, O. 1982. The levant trilogy. London: Penguin. Matthews, P.H, 1981. Syntax. Cambridge: CUP. Matthews, R. 1989. Reference time and modality. In A. Schöpf, ed. Essays on tensing in English. 2. Time, text and modality. Tübingen: Niemeyer, 189-246. 1991. Words and worlds. On the linguistic analysis of modality. Frankfurt: Lang. 1993. Predicted possibilities, necessary inferences and the like: 'epistemic modality'. In Papers on semantics and grammar. Frankfurt: Lang, 55-78. Matthews-Bresky, R.J.H. 1977. Modals of strong obligation. Zielsprache Englisch 7: 6-11. McCallum-Bayliss, H. 1985. The modal verbs: universal/univocal1 lexical items. P h D Diss, Georgetown University. Ann Arbor: University Microfilms International. McDowell, J.P. 1987. Assertion and modality. P h D Diss, University of Southern California. Ann Arbor: University Microfilms International. 1991. Quasi-assertion. Journal of Semantics 8:311-331. Mcintosh, A. 1966. Predictive statements. In C.E. Bazell, J.C. Catford, M.A.K. Halliday h R.H. Robins, eds. In memory of J.R. Firth. London: Longman, 303-320. Melrose, S.L. 1983. 'Must' and its periphrastic forms in American English usage. MA thesis, University of California, Los Angeles. Unpublished. Mitchell, J.E. 1986. The formal semantics of point of view. P h D Diss, University of Amherst, Massachusetts. Ann Arbor: University Microfilms International. Mittins, W.H., M. Salu, M. Edminson & S. Coyne. 1970. Attitudes to English usage. London: OUP. Murdoch, I. 1990. The message to the planet. London: Penguin. Nagle, S.J. 1986. Inferential change and syntactic modality in English. P h D Diss, University of Southern California. Ann Arbor: University Microfilms International. Nehls, D. 1986. Semantik und Syntax des englischen Verbs. Teil II. Die Modalverben. Eine kontrastive Analyse der Modalverben im Englischen und Deutschen. Heidelberg: Groos. Newmeyer, F.J. 1970. The 'root' modal: can it be transitive? In J.M. Sadock & A.L. Vanek, eds. Studies presented to Robert Β. Lees by his students. Edmonton, Champaign: Linguistic Research Institute, 189-196. 1975. English aspectual verbs. The Hague: Mouton. OED. 1989. The Oxford English dictionary. 2nd ed. J.A. Simpson h E.S.C. Weiner, eds. Oxford: The Clarendon Press. Ohlschläger, G. 1989. Zur Syntax und Semantik der Modalverben des Deutschen. Tübingen: Niemeyer. Palmer, F.R. 1974. The English verb. London: Longman. 1977. Modals and actuality. Journal of Linguistics 13: 1-23. 1979a. Modality and the English modals. London: Longman. T h e title appears in both forms.

224 1979b. Non-assertion and modality. In D.J. Allerton, E. Carney h D. Holdcroft, eds. Function and context in linguistic analysis. A Festschrift for William Haas. Cambridge: CUP, 185-195. 1980. Can, will, and actuality. In Greenbaum, Leech & Svartvik, eds. 1980, 91-99. 1983. Semantic explanations for the syntax of the English modals. In Heny k Richards, eds. 1983, 205-217. 1986. Mood and modality. Cambridge: CUP. 1987. The English verb. 2nd ed. London: Longman. 1989. Why future tense? English Today 5/4:59. 1990. Modality and the English modals. 2nd ed. London: Longman. Perkins, M.R. 1983. Modal expressions in English. London: Pinter. Plank, F. 1984. The modals story retold. Studies in Language 8:305-364. Postal, P.M. 1974. On raising. Cambridge, Mass.: MIT. Postal, P.M. h G.K. Pullum. 1978. Traces and the description of English complementizer contraction. Linguistic Inquiry 9:1-29. Poutsma, H. 1904. A grammar of late modern English. Part 1 - The sentence. Section 1 - The elements of the sentence. Groningen: Noordhoff. 1921. The characters of the English verb and the expanded form. Groningen: Noordhoff. Preston, D.R. 1989. Sociolinguistics and second language acquisition. Oxford: Blackwell. Pullum, G.K. & D. Wilson, D. 1977. Autonomous syntax and the analysis of auxiliaries. Language 53:741788. Quirk, R. 1981. International communication and the concept of Nuclear English. In L.E. Smith, ed. English for cross-cultural communication. London: Macmillan, 151-165. Quirk, R., S. Greenbaum, G. Leech L· J. Svartvik. 1972. A grammar of contemporary English. London: Longman. 1985. A comprehensive grammar of the English language. London: Longman. Radford, A. 1988. Transformational grammar: a first course. Cambridge: CUP. Reichenbach, H. 1947. Elements of symbolic logic. New York: The Free Press. Rivière, C. 1981. Is should a weaker must? Journal of Linguistics 17:179-195. Robinson, B. L· L. Haegeman. 1979. A note on some aspects of the use of will in affirmative declarative sentences. Journal of Linguistics 15:107-110. Ross, J . R . 1969. Auxiliaries as main verbs. In W. Todd, ed. Studies in philosophical linguistics 1. Evanston, 111.: Great Expectations Press, 77-102. 1970. On declarative sentences. In R.A. Jacobs & P.S. Rosenbaum, eds. Readings in English transformational grammar. Waltham: Ginn, 222-272. Rutherford, W . E . 1975, 1977. Modern English. Vols 1 and 2. New York: Harcourt, Brace, Jovanovich. Schachter, P. 1983. Explaining auxiliary order. In Heny & Richards, eds. 1983, 145-204. Shopen, T . 1971. Caught in the act. Papers from the 7th Regional Meeting of the Chicago Linguistic Societfjr:254-263. Sinclair, J., ed.-in-chief. 1987. Collins COBUILD English language dictionary. London: Collins. 1990. Collins COBUILD English grammar. London: Collins. Smith, M.A. 1982. On the acquisition of must, have to, and havegot to by German speakers. In W. Kühlwein & A. Raasch, Hrsg. Stil: Komponenten - Wirkungen. Bd 2 (Kongreßbericht der 12. Jahrestagung der GAL 1981). Tübingen: Narr, 82-86. Smith, N.V. 1981. Grammaticality, time and tense. Philosophical Transactions of the Royal Society, London Β 295: 253-265. 1989. The twitter machine: reflections on language. Oxford: Blackwell. Sperber, D. & D. Wilson. 1986. Relevance. Oxford: Blackwell. Strang, Β.M.H. 1968. Modem English structure. 2nd ed. London: Arnold. Stubbs, M. 1986. Ά m a t t e r of prolonged fieldwork': notes towards a modal grammar of English. Applied Linguistics 7:1-25.

225 Suh, K.-H. 1992. A discourse analysis of the English tense-aspect-modality system. P h D Diss, University of California, Los Angeles. Ann Arbor: University Microfilms International. Suzuki, N.Y. 1978. A generative semantic analysis of the English modals. P h D Diss, Georgetown University. Ann Arbor: University Microfilms International. Svartvik, J . h R. Quirk, eds. 1980. A corpus of English conversation. Lund: C W K Gleerup. Svartvik, J. h D. Wright. 1977. The use of ought in teenage English. In S. Greenbaum, ed. Acceptability in language. The Hague: Mouton, 179-201. Swan, M. 1980. Practical English usage. Oxford: OUP. 1993. Language conundrums. English Language Teaching Journal 47:354-355. 1995. Practical English usage. 2nd ed. Oxford: OUP. Swan, M. L· B. Smith, eds. 1987. Learner English: a teacher's guide to interference and other problems. Cambridge: CUP. Sweetser, E.E. 1982. Root and epistemic modals: causality in two worlds. Proceedings of the 8th Annual Meeting of the Berkeley Linguistics Society :484-507. 1990. From etymology to pragmatics: metaphorical and cultural aspects of semantic structure. Cambridge: CUP. Taylor, J.R. 1989. Linguistic categorization - prototypes in linguistic theory. Oxford: Clarendon Press. Thomson, A.J. & A.V. Martinet. 1980. A practical English grammar. 3th ed. Oxford: OUP. Todd, L. & I. Hancock. 1986. International English usage. London: Croom Helm. Tottie, G. 1971. Have to: a study of usage and acceptability in present-day British English. Stockholm Theses in English 4. Stockholm University. 1985. The negation of epistemic necessity in present-day British and American English. English World-Wide 6:87-116. Traugott, E.C. 1978. On the expression of spatio-temporal relations in language. In J.H. Greenberg, ed. Universals of human language. IV: Word structure. Stanford: Stanford UP, 369-400. 1989. On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 65:31-55. Tregidgo, P.S. 1982. Must and may: demand and permission. Lingua 56:75-92. Trudgill, P. & J. Hannah. 1982. International English: a guide to varieties of standard English. London: Arnold. Twaddell, W . F . 1963. The English verb auxiliaries. 2nd ed. Providence: Brown UP. Visser, F.T. 1969. An historical syntax of the English language, Pt III. Leiden: Brill. Ward, J . M . 1957. Peculiarities in English grammar. London: Longman. Weiser, A. 1974. Deliberate ambiguity. Papers from the 10th Regional Meeting of the Chicago Linguistic Society:723-731. Wekker, H.C. 1976. The expression of future time in contemporary British English. Amsterdam: NorthHolland. Wells, J.C. 1990. Longman pronouncing dictionary. London: Longman. Westney, P. 1980. W h a t can be done with should and ought to? In G.K. Jember, ed. Tübingen studies in literature and language: language and its manipulation. Denver, CO: T h e Society for New Language Study, 22-51. 1984. Review of Coates (1983) and Perkins (1983). International Review of Applied Linguistics 22:229231. 1986. Notes on scales. Lingua 69:333-354. 1994. Rules and pedagogical grammar. In T . Odlin, ed. Perspectives on pedagogical grammar. Cambridge: CUP, 72-96. Whitaker, S.F. 1966. 'The mighty must': some suggestions concerning the modal auxiliary. English Language Teaching Journal 21:12-15. Willis, D. 1990. The lexical syllabus. A new approach to language teaching. London: Collins ELT. Zandvoort, R.W. 1969. A handbook of English grammar. 5th ed. London: Longman.