2.1 The future is uncertain / will is not an inflection

If the three arguments against treating -ed as a tense look weak, why do most grammarians take them seriously when they apply to will? Before looking at these arguments in detail, we should consider the supposed differences between the two cases, which are these:

1. D. The future, unlike the past, is inherently uncertain and therefore has something in common with modality.

2. E. The -ed morpheme is an inflection, whereas will is an auxiliary verb; furthermore, will shares the morphosyntactic properties of the English modal auxiliaries.

In his discussion of futurity, Comrie (Reference Comrie, Abraham and Janssen1989: 53) calls argument (D) ‘the conceptual non-argument’, since it may be a fact about the world but that does not make it a fact about language. Comrie points out that if I say It will rain tomorrow and it does not in fact rain the next day, then my statement was simply false; this is not the case for It may rain tomorrow, which is genuinely modal. When will is used for future time reference it simply does not have the properties of modals. We return to this point in section 4.3.4 below.

It is argument (E) which is the reason most commonly given for saying that English only has two tenses, past and present. Jespersen was quite clear on this point:

The same position was adopted by all the grammars co-authored by Geoffrey Leech, starting with Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1972) and continuing up to Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999).

This position rests on a remarkably narrow conception of tense as a purely morphological category: many languages use auxiliaries and particles to express time relations, and studies of tense in a range of languages, such as Comrie (Reference Comrie1985) and Dahl (Reference Dahl1985), do not assume that tense is only inflectional. It is generally agreed that morphological tense is more highly grammaticalised than other means of expressing time, but this does not mean that periphrastic tenses are not grammaticalised at all. Vetters & Skibinska (Reference Vetters, Skibinska, Borillo, Vetters and Vuillaume1998) comment that periphrastic constructions tend to evolve into morphological ones, for example in the development of the future tense in French, so that the distinction is not stable (Reference Vetters, Skibinska, Borillo, Vetters and Vuillaume1998: 250). Similarly, Declerck (Reference Declerck1991: 10) points out that ‘whether or not a tense is morphologically marked is the result of a historical development determined by arbitrary factors that have nothing to do with the essence of the phenomenon of tense itself. I therefore think it is unwarranted to claim that tenses can only be marked by an inflectional morpheme and not by a free morpheme such as an auxiliary.’

In a quantitative study of tense and aspect categories across languages, Bybee & Dahl (Reference Bybee and Dahl1989: 56) note that 73 per cent of the languages in their sample express the past tense using a bound expression, whereas the figure for future tense is 46 per cent (cf. also Dahl & Velupillai Reference Dahl, Velupillai, Haspelmath, Dryer, Gil and Comrie2008). This difference may be significant, but this still leaves 54 per cent of languages using a periphrastic expression for future tense.

In non-literary French, the default way to refer to past time is by using the auxiliary avoir ‘to have’ + past participle, whereas future time is expressed using an inflection:

1. (1) Nous avons mangé hier dans un restaurant.

   ‘We ate yesterday in a restaurant.’   (literally ‘We have eaten’)

2. (2) Nous mangerons demain dans un restaurant.

   ‘We will eat tomorrow in a restaurant.’

If we only accept inflectional tenses, then French would have a future tense but not a past tense, a conclusion that few would accept.

Comrie dismisses argument (E) as ‘the formal non-argument’ against a future tense (Reference Comrie, Abraham and Janssen1989: 54–5), and Huddleston (Reference Huddleston1995: 414) concedes that this argument ‘has no force’.

It seems fair to conclude that arguments (D) and (E) do not provide a good reason to distinguish between -ed and will. We return now to the first three arguments as they apply to will: we shall see that they are equally unconvincing.

At this point it will be useful to list informally the different uses of will and would (see table 1).

Table 1.

Uses of will and would

We shall mostly ignore constructions with would such as would like (as in We would like to see Mr Brown now, please), wouldn't mind (cf. I wouldn't mind a coffee right now) and would rather (e.g. I would rather go tomorrow).

2.2 Argument (A): will has modal uses

Concentrating for now on the left-hand column of table 1, we can pick out uses (iii–viii) as those where will appears to express something other than simple future time reference. The first argument (A) against treating will as a future tense is the existence of these uses.

Faced with data of this kind, some obvious questions need to be asked. Firstly, what is the nature of these non-temporal uses of will? Are they all of the same kind, and should they be grouped together as ‘modal’? Secondly, are they fundamentally different from future-time will, or are they best analysed as pragmatic inferences from a basic future time meaning, given the right contextual conditions? A comparison with other languages is likely to be revealing here: if we find similar phenomena in languages with an uncontroversial future tense, that would tend to support a pragmatic analysis. Thirdly, how common are they? If they are rare, marginal, or restricted in particular ways, that would make them less convincing as evidence against a tense analysis of will.

In response to the first question, it is surely only use (viii), often called the epistemic use of will, which has a strong claim to be called ‘modal’. The core modal auxiliaries may and must have similar epistemic uses,Footnote ³ and although there are some interesting differences between epistemic will and epistemic must (see Gilbert Reference Gilbert, Dendale and van der Auwera2001: 126–8 and section 4.3.4 below), the two are certainly analogous and can often be substituted one for another with little or no change of meaning: for example, in this exchange from the British National Corpus (BNC):

1. A: And that's all, there's nothing

2. B: No there's nothing down that way as far as I know.

3. A: Okay, it must be one of these then. Okay.   [cf. It'll be one of these then.]

4. B: It's one of these, yes.   (BNC KE2)

The other uses are less clearly modal. Intention (iii) and volition (iv) are not expressed by other modals in English. Characteristic properties (v) and persistent habits (vii) are likewise not usually taken as coming under the heading of modality, though there is a parallel with can (Oil can float on water / He can sit in his room all day staring at the TV): we are in the domain of genericity here, not modality. There is a similar parallel with can/could in polite requests (vi): Can/could you help me look for my purse? It is worth pointing out that apart from the epistemic use of will, none of the other uses has parallels with the core modals may and must. Thus apart from the epistemic use, to which we return directly, there is little here to convince us that will is fundamentally modal.

In response to the second question, even some of the strongest advocates of a non-temporal analysis of will concede that several of these uses can be accounted for by the interaction of a basic futurity semantics with elements of the context. Thus Haegeman argues that the volition and intention uses (iii, iv and vi in table 1) are determined by a ‘volitional’ proposition in the right context (typically with first- and second-person subjects), and are not separate from the future use (Haegeman Reference Haegeman1983: 97). There seems to be general agreement that the persistent habit use (vii in table 1) depends on stress on will, and is not separate from characteristic properties (v). The latter is more interesting: it also arises with the simple present and past tenses in the right context (a non-specific subject, or an expression like always to indicate genericity):

1. (3) Oil floats on water.

2. (4) He sits in his room all day staring at the TV.

3. (5) He sat in his room all day staring at the TV.

However, with the past and present tenses the time reference does not change when the sentence is generic (a past generic like (5) is still past), but a generic with will does not refer to future time. There is perhaps a sense of ‘If you take any instance of oil, it floats on water’, with the idea that any such instance will have to be in the future; but this is different from past and present generics. It also remains to be explained why the French future tense is not always used for characteristic properties in the same way:

1. (6) Weakened batteries will result in lower volume or poor tonal quality.

2. (7) Des piles insuffisamment chargées entraînent un volume plus faible ou une qualité sonore médiocre.   [from the INTERSECT translation corpus (Salkie Reference Salkie2008)]

It does not seem implausible, however, to derive these generic uses of will from a future time sense: they do not appear to have an intrinsic connection with modality.

Turning now to the third question, the frequency of these ‘modal’ uses of will, the evidence is that they are very rare. The first corpus-based study of this question was Wekker (Reference Wekker1976). Using a corpus of 500,000 words of written English (novels and newspapers), and 100,000 words of spoken (radio and TV programmes), Wekker found 1,547 instances of will. He examined the first-person, second-person and third-person examples in turn, and his main results were as follows.

• • I/we will: 209 instances, of which 46 are non-volitional like (8), most of the remainder ‘ambiguous between a volitional and a simple future interpretation’ like (9) (Reference Wekker1976: 41).

1. (8) Next week we will learn for the first time the nature of the Cabinet minutes.

2. (9) If he should decide to instruct us further in the matter we will let you know.

• • You will: 70 instances, of which 67 ‘express simple futurity and . . . show no trace whatever of either the speaker's or the subject's willingness or intentions’ (Reference Wekker1976: 54), e.g.:

1. (10) I can assure you, sir, that when you come back to Hue in a year, you'll find a Gideon Bible in your hotel room.

• • Third-person subject: 1,268 instances, of which 1,241 (98%) express simple futurity, ‘without saying anything about the wishes or intentions of the subject of the sentence’ (Reference Wekker1976: 60), e.g.:

1. (11) The new season will run until March 21, and the theatre will be announced shortly.

The next substantial corpus-based study was Larreya (Reference Larreya1984), although a direct comparison with Wekker is not straightforward because Larreya's figures cover will and would. Larreya found a total of 2,241 instances in a corpus of 276,500 words of written English and 28,400 of spoken. The breakdown of this total is:

Larreya makes it clear that the ‘conditional’ category are all instances of would (Reference Larreya1984: 304); if these are left out, the total is 1,746 and the percentages are:

Some of these examples also involve would rather than will. Nonetheless, the non-temporal uses still appear to be fairly marginal, though it would have been useful to know how many of Larreya's 1,571 actually include volition.

Mindt (Reference Mindt1995) used a much larger corpus of 80 million words. No details about the corpus are given, but see Mindt (Reference Mindt2000: 596) for the corpora of 240 million words used during his ten-year project. Mindt's categories are not easy to match with the ones we are concerned with here, but they are nonetheless of interest:

• • Certainty/prediction: 71 per cent

1. (12) I am sure it will be published.

2. (13) Technique, rather than tactics, will be needed from the jockey.

• • Volition/intention: 16 per cent

1. (14) I will be perfectly frank with you.

2. (15) One thing I'll ask of you, keep away from the . . . Cadwells.

• • Possibility/high probability: 10 per cent

1. (16) You hope I'll miss you.

2. (17) Indeed he will probably improve for a season or two.

• • Others: 3 per cent

Others include inference/deduction (. . . when steps sounded on the veranda, she said ‘That'll be Vetch's boy’) and habit (you will never be lonely because your family will come along and pay you visits) (Reference Mindt1995: 60).

Mindt adds that 94 per cent of his instances of will have future time orientation.

Two other corpus-based studies suggest that the non-temporal uses of will are less rare. Coates (Reference Coates1983) used the Survey of English Usage corpus, containing 725,000 words. Her figures only relate to 200 instances of will, described as ‘a representative sample’ (Reference Coates1983: 2). Coates gives the following breakdown:

• • Willingness: 26 instances

1. (18) Give them the name of someone who will sign for it and take it in.

• • Predictability: 19 instances

1. (19) A commotion in the hall. ‘That will be Celia,’ said Janet.

• • Intention: 46 instances

1. (20) But I will bring you more today I promise

• • Prediction: 96 instances

1. (21) It will be lovely to see you.

All but one of the ‘intention’ examples have a first-person subject.

Palmer (Reference Palmer1990) also uses the Survey corpus. He gives no specific figures, saying: ‘This is not a textual study, and is not concerned with . . . giving statistical information about the use of modals’ (Reference Palmer1990: 29). He goes on to comment, however: ‘there are very few examples in the Survey of futurity will and shall that can, without any doubt at all, be regarded as simply predicting a future event’ (Reference Palmer1990: 137–8).

It is noticeable that Wekker, who believes that will is a future tense marker, has statistics supporting this view; while the figures given by Coates and the informal remarks of Palmer support their view that will is a modal. Larreya is a strong supporter of the modal position, however, and Mindt is agnostic – and their figures support Wekker. The figures in Leech et al. (Reference Leech, Hunt, Mair and Smith2009) show a small decrease in the frequency of will over the past 50 years, but a far smaller decrease than for the core modals (see note 3 on which modals are core). All in all, the corpus findings seem to support the tense analysis.

We return now to the epistemic use of will (viii in table 1). This use is found in languages which have an uncontroversial future tense, such as French or Italian (Rocci Reference Rocci2002):

1. (22) Pourquoi donc a-t-on sonné la cloche des morts? Ah! Mon Dieu, ce sera pour Madame Rousseau.   (Proust)

   ‘For whom did they toll the passing-bell just now? Oh dear, of course, it would be for Mme. Rousseau.’   (translation by Scott Moncrieff, 1922)

   ‘Now who were they ringing the passing-bell for? Oh, dear God, it must have been for Mme. Rousseau.’   (translation by Lydia Davis, 2003)

2. (23) – A qui est-ce?

   – C'est lui qui l'aura oublié ici . . . Il l'avait dans les mains hier au soir . . . (Simenon)

   ‘Whose is it?’

   ‘It's he who will have left it here . . . He had it in his hands yesterday evening . . .’

3. (24) Bella stoffa, bella . . . l'avrà pagata un occhio, immagino, laggiù in città non scherzano.   (Buzzati)

   ‘Nice material . . . you will have paid a lot of money, I imagine; there in the city they do not joke.’

4. (25) – Non c'è Marco?

   – Sarà sarà in ospedale, è diventato papa oggi.   (spoken conversation)

   ‘Marco is not there?’

   ‘He will be in the hospital, he became a father today.’

This epistemic use of the future tense is restricted and rare in French and Italian, but that is also true of English (cf. Celle Reference Celle2005 for discussion of the restrictions in French and English. In her sample of 1,000 instances of the French future, Celle did not find a single instance of the epistemic use (Reference Celle2005: 208)). This indicates that the use of future tenses to express an inference or a deduction is available, though not necessarily common, across languages; it does not support the view that will is fundamentally modal. A similar conclusion is reached by Bybee, Perkins & Pagliuca (Reference Bybee, Perkins and Pagliuca1994) and Tsangalidis (Reference Tsangalidis1999).

Notice also that be going to can be used to express an inference, as in these examples from the BNC:

1. (26) It seems unlikely then, that ‘standard procedures’ of recording space and time are going to be relevant to the unique identification of utterance acts. (F9V)

2. (27) So next year's output is to be a hundred and fifty percent of this year's. Because you add fifty percent to the hundred percent and a hundred plus fifty is a hundred and fifty. So the output is going to be a hundred and fifty percent of what it was last year. . . (KE2)

See Fleischman (Reference Fleischman1982: 91) and Tsangalidis (Reference Tsangalidis1999: 160) for similar remarks. The evidence thus suggests that the ‘modal’ uses of will are rare, and are either not modal or mostly explainable on pragmatic grounds.

2.3 Argument (B): the other modal auxiliaries have modal meanings

We do not need to linger long on this argument. It is true that will has the same grammar as other modals, but that does not necessarily mean that it has the same meaning. Huddleston & Pullum et al. (Reference Huddleston and Pullum2002: 109) list eight morphosyntactic properties shared by be and auxiliary have – but there is no suggestion that they therefore share semantic properties. Huddleston & Pullum et al. concede: ‘This argument is not in itself decisive . . . but it does place the onus of proof on defenders of the future tense analysis to demonstrate why will (and shall) should be differentiated from the others as tense auxiliaries vs. mood auxiliaries’ (Reference Huddleston and Pullum2002: 209). Given the weakness of the arguments against the tense position, the question of where the onus of proof lies is debatable.

2.4 Argument (C): English has many ways of referring to future time

This argument starts with data like these:

1. (28) I'm going to visit Germany next March.

2. (29) The train leaves in 2 hours.

3. (30) Helen is taking her exam tomorrow.

4. (31) Frank may arrive next week.

5. (32) Julie will visit Germany next April.

Is there any reason to pick out one of these as ‘the future tense’? In fact, there are several. The other ways of referring to the future are restricted in various ways. For example, The train leaves in 2 hours can only be used for a ‘pre-determined and certain’ situation (Declerck et al. Reference Declerck2006: 182): you can't say, *Leeds Utd win against Everton tomorrow, though you can say Leeds Utd play Everton tomorrow. The present tense is typically used in this way for statements about the calendar, future situations that result from a definite plan or a regular pattern which already exists at the time of speaking (Declerck et al. Reference Declerck2006: 182–3). The will + infinitive construction has no such restrictions. In addition, the will + infinitive construction can express future time by itself. The present tense normally needs a time adverbial for a future time interpretation (Huddleston & Pullum et al. Reference Huddleston and Pullum2002:132). Furthermore, French also has aller + infinitive, the present tense, and modals with future time reference, and this is true of many other languages (see in particular Bybee, Perkins & Pagliuca (Reference Bybee, Perkins and Pagliuca1994) for other examples).

We conclude that the five major arguments against the tense analysis are far from convincing.

3.1 Frequency

As we have seen, the future time use of will is overwhelmingly the most frequent in naturally occurring speech and writing. Most corpus studies agree that the proportion of pure future time uses is over 90 per cent. Usually statistics of this kind would convince most linguists that the other uses were marginal; it is unlikely that they would be held up as representing the basic meaning of a word. Whether the different uses of a word correspond to distinct senses, and whether distinct senses can be derived from a more abstract underlying meaning, is a controversial issue, particularly with English modals (cf. Papafragou Reference Papafragou2000): the frequency of different uses is, of course, not the only pertinent factor here, but a very high frequency of one use is surely relevant.

It is also worth noting that will is far more frequent than the core English modals. The figures from the large corpus used in Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999), presented in table 2, are revealing (compare also similar figures in Coates Reference Coates1983: 23–5).

Tabel 2.

Frequency of modals in the Longman corpus (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 486)

Since over 90 per cent of these occurrences of will probably express future time, this table tells us that there is a large difference in frequency between these occurrences and uses of the core modals – unsurprising if will is basically a tense.

3.2 Comparison with languages which do or do not have a future tense

It is useful to compare English with a language like French, which definitely has a future tense, and Finnish, which definitely does not. Compare these extracts from weather forecasts:

The question is whether English looks more like French or Finnish. Leaving aside the fact that the French future is inflectional whereas will is not, it appears that English and French both have a morpheme which is routinely used to refer to the future in weather forecasts, whereas Finnish has no option but to use the present tense. Put this way, it appears that English is more similar to French than to Finnish. It is true that the simple present tense is also found in English weather forecasts, along with other expressions (Tomorrow looks like being cloudy / Thursday is expected to be sunny, etc.), but that does not affect the argument: meteorologists speaking French and English have a choice which is denied to their Finnish counterparts. (A similar comparison of the three languages is made by Dahl & Velupillai Reference Dahl, Velupillai, Haspelmath, Dryer, Gil and Comrie2008.)

3.3 Typological comparison

Dahl (Reference Dahl1985) includes English in his list of languages which have a future tense, ranking English twentieth out of 49 languages in its closeness to the ideal future tense category, higher than French or Spanish which are normally taken to have uncontroversial future tenses. He comments that ‘the traditional view of the Future as a tense can . . . be defended’ (Reference Dahl1985: 107). In more recent work, Dahl (Reference Dahl and Dahl2000) looks at a range of European languages and argues that a number are genuinely ‘futureless’, including ‘at least all Finno-Ugrian and Germanic languages except English’ (Reference Dahl and Dahl2000: 326). With respect to English, Dahl argues that ‘will . . . has (in combination with shall) become what is probably the most highly grammaticalised future marker in the Germanic languages' (Reference Dahl and Dahl2000: 322). However, Dahl & Velupillai (Reference Dahl, Velupillai, Haspelmath, Dryer, Gil and Comrie2008) are more cautious, noting only that ‘one result of the progressive grammaticalization of futures is that the temporal component of their semantics becomes more dominant relative to the modal component’.Footnote ⁴

3.4 Explanations for non-temporal uses of will

We saw in section 2.2 that it is relatively easy to account for the uses in table 1 under the headings of (iii) intention, (iv) volition, (v) characteristic properties, (vi) polite requests, (vii) persistent habits and (viii) inference and deduction. We return to the volition uses in section 5. That leaves only uses (ix) and (x) which involve conditional sentences: we consider these in our discussion of Huddleston (Reference Huddleston1995) below.

4.1 Argument F: Would is the past time counterpart of will

Like many of the arguments that we shall consider in this section, this one is put forward in Huddleston (Reference Huddleston1995). A third of his paper tries to demonstrate that would is the past tense counterpart of will (compare also Huddleston & Pullum et al. (Reference Huddleston and Pullum2002: 209), where this is described as a ‘major argument’ against the tense analysis). Huddleston claims, accurately, that this relationship between will and would is fatal to any analysis in which will is ‘the marker of a future tense contrasting with preterite and present in a three-term tense system: will is not in paradigmatic contrast with preterite and present tense but contrasts syntagmatically with either’ (Reference Huddleston1995: 415).

Although this argument is correct, it is basically irrelevant, since no serious study of English tense has argued for a three-term system. The actual contenders are:

• • Two tenses (the mainstream view)

• • Four tenses (Huddleston's position, and that of Huddleston & Pullum et al. Reference Huddleston and Pullum2002)

• • Eight tenses (e.g. Declerck Reference Declerck1991)

It is arguable that the eight tense analysis is the obvious, default one. Suppose we distinguish between the tense morphemes ‘past’ (represented schematically as -ed), have and will, and the tense configurations which result from using one or more of these morphemes in a clause. The three competing analyses can now be reformulated as claims about tense configurations:

• • Two tenses: no configurations are permitted, since only -ed (vs its absence) counts as a tense morpheme.

• • Four tenses: -ed and have can combine, yielding the four configurations in table 3.

• • Eight tenses: -ed, have and will can combine, yielding the eight configurations in table 4.

Table 3.

The four-tense analysis

Table 4.

The eight-tense analysis

These eight tense configurations are perfectly compositional: the first three contain only one morpheme, the next four contain two (as their two-part names suggest), and the last one contains three. They are the only possible configurations owing to the morphology of the three tense morphemes (will only has finite forms, so it cannot follow have, etc). Whether the eight-tense analysis can account for the semantics correctly remains to be seen, but it is fully compatible with the observation that would is the past counterpart of will (i.e. that would can be analysed as will + -ed). Huddleston's claims about will and would are thus irrelevant to choosing between the different analyses.

We should add that in French the future tense inflection can be combined with a past tense inflection to yield the ‘future in the past’ tense configuration (Il mangera [future)] vs Il mangerait [future in the past]):Footnote ⁵ this is also true of many other languages, as noted by Bybee (1995) and others. Thus there is nothing ad hoc about combining the future and past tense morphemes. Furthermore, the analysis in table 3 is not subject to Huddleston's objection that it treats the meaning of Pamina will live as ‘future in present’, which is implausible (1995: 418). The future tense configuration in table 3 simply lacks the -ed morpheme: that is, it is not specifically marked as past.

4.2 Argument G: Generalisations about English modals

4.3 Argument G: semantic and pragmatic similarities between will and the other modals

Before examining Huddleston's arguments under this heading, we need to consider the nature of the issues here. The central claim of the modal analysis is that will fits seamlessly into a framework for representing the meanings of the modal auxiliaries of English. The strength of any proposed modal analysis of will thus depends crucially on whether the overall framework for modality is plausible. Modality is a complex area, and scholars disagree radically about whether it should be broadly or narrowly defined, how to distinguish it from related areas such as mood and evidentiality, where to draw the line between semantic and pragmatic aspects of modality, whether it is in principle desirable to seek a single theoretical framework for English modals, and many other issues.

Huddleston does not propose a theoretical framework for modality, and his discussion adopts the descriptive classification of Palmer (Reference Palmer1990), with one modification – the notion ‘degree of modality’ – to which we shall return. His arguments claim that will can be accommodated within Palmer's system. Arguably, though, almost anything loosely related to modality will have similarities with Palmer's system, since as a descriptive system it does not attempt to be exclusive. The most that we could concede to Huddleston is that he may develop some semantic generalisations about English modals which extend to will. In reality, we do not need to concede even this.