The full range of properties

To the largely formal properties discussed in Chapters 5 and 6 we can now add semantic (or semantic–pragmatic) properties of the modal group. One, their occurrence as sentential modifiers, is a property already identified in the narrower context of impersonal constructions. But since ‘subject-raising’ constructions with other verbs become more prominent in late Middle English, this is a less distinctive property at this period. More significant is occurrence as subjectives, both deontic and epistemic; this parameter is incorporated into Table 8.1 (p. 186) along with others already established. Here ‘prediction’ uses are included as subjective epistemics. Among other verbs, subjective deontic and epistemic uses in early English are associated with verbal mood rather than with individual lexemes, though ‘hortative’ let may be a partial exception to this statement (MED leten v. 10a) and particular context-bound instances (e.g. of seme) may be subjective; there are also (as today) adjectives (e.g. lik ‘likely’) and adverbs (e.g. certain, douteles) which may presumably be subjective epistemic. This additional criterion, along with the further evidence that these words are sentential modifiers, reinforces the picture of the mutual predictiveness of properties, and of the prototypical status of the verbs of group A: this seems very clear for late Middle English, less so for Old English. Indeed, for late Middle English there is virtually a further criterion.

In this book I have tried to offer a relatively nonabstract account of major aspects of the grammar and history of English auxiliaries, not simply relying on reinterpretations of established data, but undertaking some further investigations as appropriate. I do not (of course) pretend to have provided a complete account, but have focused on specific topics which I am confident will need to be integrated within a more complete statement. My conclusions can be best presented under four headings.

Synchrony. I have argued for a new analysis of English auxiliaries, claiming that a series of their properties has a natural interpretation if they are analysed as a word class which is distinct from verbs particularly in that they are not subject to the regularities of verbal morphosyntax. What look like inflected forms are not, but (with the partial exception of being and having) are holistic or unitary items incorporating tense, mood or nonfinite categories lexically. This gives a rational account of the ordering of auxiliaries, of the absence of particular categories, of some curious facts about ellipsis, and of a range of other particular properties. It also accounts for the learnability of such properties. A formal analysis within the account of the lexicon given in HPSG was outlined, and a small set of lexical redundancy rules was shown to give a rather complete account of auxiliary constructions. It was more speculatively suggested that the distinct categoriality of auxiliaries in the case of modals (whose properties are central to the class) partly but crucially involves the interaction of subjectivity with their formal distinctness from verbs.

The ancestors of our modern auxiliaries before the Modern English period certainly had more verblike properties than their modern descendants, as we saw in Chapter 4. And most of the formal differences which distinguish so sharply between modern auxiliaries and full verbs had not then developed. But it is too simple to conclude that they were therefore verbs four square in all essential syntactic properties with other verbs at this period, with (say) Lightfoot (1979), for they already share striking characteristics which apparently isolate them. This chapter will be taken up with the establishment and discussion of two of the most striking of these characteristics, their behaviour in ellipsis and with impersonals. These are important because they hold at the level of sentential syntax and they look on the face of it like potential wordclass properties. The conclusion which will seem most plausible for Old (and Middle) English is that there was already a formal subcategorial distinction within the class of verbs. In the next chapter this position will be reinforced, and Chapter 7 will discuss the notional identity of this grouping.

At the end of Chapter 4 three questions were isolated as methodologically central in determining word-class status. The first two were:

1. (i) Do these words have formal properties which distinguish them from other verbs?

2. (ii) If so, do such formal properties typically cooccur with each other, so that there is a measure of predictability between properties which might characterize a distinct group?

This chapter will focus on the period of particularly rapid change in early Modern English in which the status of modals and auxiliaries was substantially clarified. Lightfoot (1979) interpreted this as showing the development of a class ‘modal’. My views differ from his in many respects. But we agree that the beginning of early Modern English saw a particularly significant set of changes, as (with less enthusiasm) does Plank (1984, see 348). In discussing this, I will need to recap some work familiar particularly from these authors, but my interpretation will differ and I will cover a wider range of data than Lightfoot.

We can conceptualize early Modern English changes affecting auxiliary group verbs under three headings. First comes a series of changes to earlier properties which further differentiate the modal group from full verbs: these are (A)–(D) of §9.1 below. Second is a striking further series of changes which look like new developments. These tend to make the auxiliary group more coherent and distinct. I will argue that both these sets of changes are illuminatingly interpreted as the development of a ‘basic-level’ category within Rosch's approach to categorization. Finally there is the general adoption of periphrastic DO. This must surely be interinvolved with these other changes if only because of the striking coincidence of date, but the nature of the connection is less clear. I will consider what other factors might be involved, and will sketch a speculative account relating changes in the status of finiteness to changes in word order, the obligatoriness of subjects, and the rise of periphrastic do.

So far I have suggested in Chapter 1 that auxiliaries show the prototypicality structuring typical of word classes, and in Chapter 2 I have argued that their grammatical behaviour reflects their lack of verbal morphosyntax. The argument was informal, and the claim is presumably compatible with a range of linguistic theories, at least in the sense that they do not preclude such an analysis. In this chapter I will underpin my argument by showing that there is a simple and coherent account of the type of lexical structuring of auxiliaries which I have posited, in which relevant generalizations are captured, within the formal framework of Head-driven Phrase Structure Grammar (HPSG). This includes an account of the distribution of periphrastic do, which I shall suggest is found when tense cannot be realized as an affix because it has some characteristic which is properly that of a word, thus providing an answer to question 7 of the list at the beginning of my §1.5. The discussion will necessarily be somewhat technical, and some readers may prefer to take the demonstration of this chapter for granted, and proceed directly to Chapter 4; I have recapped the small amount of discussion that is relevant to succeeding chapters so that the book coheres if it is read in this way.

In adopting HPSG I do not, of course, wish to suggest that it provides the only possible formalization of the account of Chapter 2.

Introduction

Formal criteria are primary to the establishment of word classes, but such classes typically have a notional core. Moreover, the diachronic relationship between formal and semantic characteristics needs investigation, particularly in the light of Brinton's recent claim that in a closely related area semantic change precedes syntactic change and that it is abrupt (1988: 161f.). So in this chapter I turn to two topics central to the history of modals: the development of recognizably ‘modal’ uses within a small group of verbs, the ancestors of today's group of modals, and the growing semantic coherency and isolation of that group.

The existence and status of distinctively modal uses in Old English especially needs clarification. BT gives headings which correspond to some of the distinctive areas of present-day modality in Old English, but on the whole cites less than fully convincing examples in support (see Goossens 1982 for discussion: I am rather more sceptical than he is about some of these instances). The most important semantic account for Old English is Standop (1957). He identifies the central members of the group as modalen Hilfsverben, and isolates the parameters which structure their semantic space. But he is concerned to classify the majority of uses within a historically oriented framework of sense development. Thus the synchronic significance of potentially modal uses is never focussed on as a distinct topic. On the other hand, Traugott's (1972) discussion of the sense changes undergone by these verbs from Old English is focussed on the development of grammaticalized realizations of performatives.

Introduction

The first part of this book has established the essential grammatical and lexical characteristics of the modem auxiliaries. Armed with an understanding of the most recent point of their development, I want in the second part of the book to ask how the modem auxiliaries developed, and what kind of grammatical status they had in earlier English. Their Old English ancestors already appeared in constructions which can sometimes be translated using modem auxiliaries, so they had at least some ‘notional’ points of contact with their modem congeners, if arguably often a contextual one, but their grammar was clearly much closer to that of nonauxiliary verbs. Most strikingly, periphrastic do did not appear in Old English, so that as in present-day French and German it was finite verbs generally which appeared in inverted interrogatives and were involved in the placement of sentential negation. The ancestors of the modem modals were also not as sharply distinctive in morphology and subcategorization as today. One view of these differences is that ‘pre-modals’ (and presumably other ‘pre-auxiliaries’ too) were simply verbs in Old English (see Allen 1975, Lightfoot 1979, Roberts 1985). In Lightfoot (1979) this leads to an account of the emergence of modals as a sudden, cataclysmic development early in the Modem English period, and to an account of syntactic change in terms of sharp discontinuities. But this account leaves the nature of earlier developments obscure; it claims that the definitional properties of ‘premodals’ developed essentially by chance, in some quite mysterious way.

As observed before, the present aspectual theory focusses on the interaction between temporal and atemporal structure in sentences like (1)–(4). This interaction can be (crudely and provisionally) captured by an aspectual ‘feature algebra’ based on two parameters discussed above and collected in Figure 8. However appealing this compositional machinery may be, it fails to give a satisfactory account of why it is that an NP is [+sqa] or [-sqa]. For example, to assign [+sqa] to both a sandwich and three sandwiches is correct but it does not explain anything unless we know what these NPs share. This extends, of course, to NPs like many sandwiches, at least three sandwiches, most sandwiches, etc. which also may express [+sqa]-information. A detailed enquiry is also necessary because certain NPs look as if they are [+sqa], whereas they are to be interpreted as [-sqa]. One needs to know more about the NPs themselves in order to make sure that the process of compositional interpretation proceeds as it should. Therefore, Part II will focus on the atemporal side of Figure 8. Two results are aimed at. The first one is to indicate exactly under which conditions NPs are [+sqa] or [-sqa]. This means that the representation of NPs should contain or lack certain information which is interpreted as contributing to terminative or durative aspect. The second one is to indicate exactly where atemporal structure ‘meets’ temporal structure.

Introduction

In this chapter, attention will be given to cases adduced as counterexamples against the Plus-principle. The most famous one is undoubtedly anachronistic. Vendler's John pushed the cart looks like a counterexample to the Plus-principle: John is [+sqa], the cart is [+sqa], and it would be playing ostrich simply to call the verb push [-add to]. Yet the sentence can be used duratively. How come? And how serious a counterexample is this against the Plus-principle, which is the real foundation of a compositional approach to terminative aspect? For an answer to this question, I will discuss a set of verbs like push and stroke which are a sort of hybrid between [+add to] and [-add to]. When generalizing over the behaviour of the verbs push and related verbs in English and duwen and related verbs in Dutch, I shall use the term ‘push-verbs’. Their properties bring their analysis into the realm of a controversy between adherents to the so-called Small Clause Analysis and the so-called Complex Predicate Analysis for (terminative) sentences like John pushed the cart away. This is explained in section 14.2 and I shall take sides in favour of the (syntactic) Complex Predicate Analysis. This makes it possible to devise a semantics for verb stems and their argument frames in section 14.3: I will distinguish between a transitive scheme, an unergative scheme and an unaccusative scheme in which verb stems are (non-compositionally) inserted to form the lowest syntactic verbal category, the verb.

INTRODUCTION TO PART I

The aim of this book is to present a theory about aspectual properties of sentences in natural language. These are properties allowing sentences to express temporal structure with respect to which they are interpreted. Sentences may pertain to states or processes or events, they may express boundedness, duration, repetition, semelfactivity, frequency, habituality, and many other forms of temporality. The question put in its simplest and crudest form is: how do they do this? The answer to this question will be given in terms of the opposition between terminative aspect (roughly, the expression of boundedness) and durative aspect (roughly, the expression of unboundedness). More particularly, it will be given by presenting a theory about the way terminative aspect is compositionally formed on the basis of semantic information expressed by different syntactic elements, in particular the verb and its arguments. It will be argued that an aspectual theory is explanatorily adequate only if it treats the opposition between terminative and durative aspect structurally. The focus of the theory proposed in the present study will be on the interaction between the temporal and atemporal information contributed by constituents involved in aspect construal.

Part I is introductory in the sense that it will try to clarify crucial issues which are involved. In chapter 1, some preliminary notions will be explained, identifying step by step the scope of this study. The provisional picture emerging is that of a simple ‘aspectual feature algebra’.