Jump to ContentJump to Main Navigation
The Oxford Reference Guide to English Morphology$

Laurie Bauer, Rochelle Lieber, and Ingo Plag

Print publication date: 2013

Print ISBN-13: 9780198747062

Published to Oxford Scholarship Online: December 2015

DOI: 10.1093/acprof:oso/9780198747062.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy).date: 23 March 2019

Verb inflection

Verb inflection

Chapter:
(p.60) (p.61) Chapter 5 Verb inflection
Source:
The Oxford Reference Guide to English Morphology
Author(s):

Laurie Bauer

Rochelle Lieber

Ingo Plag

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780198747062.003.0005

Abstract and Keywords

This chapter describes the morphosyntactic categories of the verb in English, drawing on evidence from different subdisciplines of linguistics. It starts out with the morphological differences between lexical and auxiliary verbs. Then the distinction between regular and irregular verbs is examined, paying special attention to the various ways in which irregular verbs may be classified. This is followed by a discussion of the inflection of the modal auxiliary verbs. Additional topics dealt with include auxiliary clitics and weak forms, the encoding of negation, and the incorporation of infinitival to into certain verbs.

Keywords:   lexical verb, auxiliary verb, regular verb, irregular verb, modal, clitic, negation

5.1 Prospectus

In this chapter we will be concerned with how the morphosyntactic categories of the verb are encoded in English. After a look at the distinction between lexical and auxiliary verbs we will deal with lexical verbs. Finally, we will turn to the inflection of the modal auxiliary verbs and to the incorporation of infinitival to into certain verbs.

5.2 Lexical versus auxiliary verbs

English distinguishes between two kinds of verbs, lexical and auxiliary. The distinction is based on a variety of diagnostic criteria, most of them syntactic in nature. Thus, auxiliary verbs precede lexical verbs in the same clause, have certain types of complement, and behave in a peculiar way in a number of syntactic constructions, such as negation, inversion, ellipsis, and emphasis. The particulars of the syntactic behaviour of this class of words are well described in the standard grammars and will not concern us here in great detail since we focus on the morphological side of the matter. It should be noted, however, that with regard to their syntactic behaviour, the auxiliaries do not show a completely uniform behaviour across all syntactic constructions, and we will see that their morphological behaviour is also not necessarily identical across different lexemes.

We will start our discussion of the morphological differences between the two major classes of verb with the lexical verbs as a reference point. The number of different forms in the inflectional paradigms of lexical verbs in standard varieties of English ranges from eight for a single verb (BE, with the forms be, am, are, is, was, were, being, been) to only three for some 40 verbs (e.g. PUT with put, puts, putting). The different paradigms of individual verbs show different kinds and numbers of syncretism, to be discussed more extensively in Section 5.3.5.1. If we implement an analysis that recognizes morphosyntactic contrasts in the system as soon as at least two verbs show this contrast (that is, we ignore some of the categories that are found only with BE), we can distinguish the six morphosyntactic categories given in Table 5.1, illustrated with three different verbs, with, five, four, and three (p.62)

Table 5.1 Verbal inflectional paradigms

Verb inflection

different forms, respectively. We deal with BE more fully in Section 5.3.4.1: it has a formal person–number distinction within the past tense category (was versus were), one within the non-3rd singular non-past category (am versus are), and one between the past tense form and the counterfactual form for the 1st and 3rd person singular (I was, she was versus I were, she were).

The details of this analysis will be discussed in Section 5.3. What is important for the discussion of the dichotomy between auxiliary verbs and lexical verbs is the fact that we can safely posit six morphosyntactic categories for the English verb, with varying degrees of syncretism across different verbs. We have a distinction between past forms and non-past forms, there is a form that is used in infinitival, subjunctive, and imperative clauses, a present participle form and the past participle form. The present participle occurs in progressive constructions, the past participle in passive and perfect constructions. Both kinds of participle can also be used as adjectives (as in a disturbing story, one of the invented alphabets, see Chapter 14 for discussion), and the present participle can serve as an event nominal (as in the teaching of the arts, see Chapter 10).

There is variation in non-standard varieties of English as to the kinds of morphosyntactic category, the number of such categories and the ways they are encoded. The present-day standard system described here is historically derived from a much richer system of categories and realizations, and some contemporary conservative dialects of English have more different verb forms than the standard, encoding more different categories. Some dialects of English have different or fewer morphosyntactic category distinctions in their system. Another parameter of variation is the distribution of the different forms across the different categories, which can differ across dialects. For example, northern dialects of Britain are known for their peculiar subject–verb agreement pattern (the ‘northern subject rule’, for example, Pietsch 2005; Wales 1996: 45) where what is called the 3rd singular non-past form in Table 5.1 is used variably in certain plural and non-3rd singular environments. Another well-known example is the use of habitual be in African-American English, some varieties of Caribbean English and IrE (e.g. Rickford 1986; Green 1998). A description of such dialectal systems is not attempted here. However, we will document variation with forms that express (p.63) the same morphosyntactic category, such as preterite spelledspelt, or participle beatbeatenbet (ScE, NZE, IrE).

Let us now turn to the auxiliaries. The class of auxiliary verbs that emerges from syntactic analyses of verbal behaviour potentially has the members shown in (1).

(1)Verb inflection

The items in (1a) are the traditionally acknowledged set of so-called modal auxiliary verbs. Perhaps controversially, we list COULD, MIGHT, SHOULD, WOULD as separate lexemes rather than as past tenses of CAN, MAY, SHALL, or WILL, since their semantics does not correspond simply to that of CAN, MAY, SHALL, and WILL with different time reference. Morphologically, the modals can be distinguished from lexical verbs in important respects. Need and dare, however, behave differently from the other modal verbs in (1a), and we will discuss their behaviour in detail in Section 5.4. The remaining modals, which we will call ‘core modals’ for ease of reference, have impoverished inflectional paradigms, lacking participles (*canning, *maying, *shoulding) and 3rd person singular forms (*oughts, *shalls), and for some of them even past tense forms (*musted, *oughted, *shalled, etc.). Core modals also share another morphological property: they generally do not serve as bases for further word-formation. The core modal WILL is also used as a lexical verb, with interesting morphological properties. This will be discussed in Section 5.3.6.

The auxiliaries in (1b) are non-modal auxiliaries, and they share much of their syntactic behaviour with the modal auxiliaries. They are, however, also used as lexical verbs, and their special status has earned them the name of ‘primary verbs’ (e.g. Quirk et al. 1985). In terms of morphology, they are characterized by the full paradigms that are typical of lexical verbs, and will therefore be treated together with the lexical verbs, but in a distinct section (Section 5.3.4).

The items in (1c) are controversial in their status. The aspect-marking verb USE, as in She used to follow him to work every day, partly behaves like an auxiliary in some varieties of English, but in others it does not (see Section 5.4.2 for further discussion). Finally, the infinitival marker to is of unclear grammatical status, with analyses ranging from conjunction to VP-subordinator (e.g. Huddleston et al. 2002: 1183–7) to non-finite modal auxiliary verb (e.g. Pullum 1982). Given its (almost) invariable morphological behaviour we can afford to remain agnostic as to its status and will consider it only in connection with its behaviour regarding certain verbs preceding it, as in, for example, gonna, wanna (see Section 5.5).

(p.64) The lexical verb GET also occurs in passive constructions in the same syntactic position as BE, which, together with its lexical morphology, can be taken as an argument for its belonging to the set in (1b). Similarly, the verb GO, in its present participle form plus a preceding form of BE plus a following TO is in competition with will, and could therefore also be included in the auxiliary verbs. There are, however, strong syntactic arguments not to include them in the class of auxiliaries, as they behave like lexical verbs in crucial syntactic tests like question formation and negation. Morphologically, nothing unexpected happens with these verbs either. The future-oriented construction with GO can be seen as an idiomatic use of the present participle of this verb, and passive GET enjoys the same full paradigm as in other syntactic constructions. The two candidates GET and GO also lack a specific morphological property that all finite auxiliary verb forms share and that distinguishes auxiliaries from lexical verbs, namely that they can act as a morphological host for the suffixed negator n’t. Future-marking GO and passive GET cannot. The two verbs are thus syntactically and morphologically not distinct from (other) lexical verbs and will not be treated as auxiliaries here. Similar arguments would hold for some apparent auxiliaries as used in certain dialects of English, such as ScE want (as in This car wants washed).

We are slightly controversial in adding the marginal example BETTER, which, in some varieties, occurs in sentences like I better do it, bettern’t I? The fact that it can take reduced n’t shows its status, but it is rare in writing or formal speech where the older and fuller form had better is preferred. We have little to say about this form beyond noting its existence.

Based on this brief overview of the two classes of verbs we will now turn to the details of the inflection of lexical verbs. Further intricacies of the morphology of auxiliary verbs will be discussed in Section 5.4.

5.3 Lexical verbs

5.3.1 The verbal paradigm

In Table 5.1 we briefly introduced the paradigm of English lexical verbs with only minimal justification and discussion. In this section we will take a closer look at the morphosyntactic categories, their justification and realization across verbs. Table 5.2 recasts the paradigm, from a different perspective, focusing on the forms and their names instead of the morphosyntactic categories. We will treat the verb BE separately in Section 5.3.4.

Table 5.2 has only five rows in the first column, as it recognizes the fact that the traditional categories of infinitive, subjunctive, and imperative collapse formally with the non-3rd singular non-past tense form. We use the label plain form to express the fact that this form is identical in form with the base form, that is the form to which affixes are attached. The difference between infinitive, subjunctive, and imperative is no longer morphologically marked in (p.65)

Table 5.2 Verbal paradigms: terminology, forms and morphosyntactic categories

Verb inflection

English, but is syntactically realized. That is, in certain syntactic constructions or sentence types the plain form is used. There is thus only one morphosyntactic category involved, which we could label ‘infinitive-subjunctive-imperative’ (cf. Huddleston 2002b: 83). We will use the less clumsy, and well-established term ‘plain form’ instead. The reasons for recognizing the morphosyntactic category of non-3rd singular non-past tense in spite of its being also realized by a plain form is that this form contrasts with the 3rd singular form, given the same sentence type (I want some cake versus She wants some cake). We will use the shorthand plain non-past form for this use of the base form. The base form, printed in small capital letters, will also be used as the citation form of the respective lexeme.

The proposed analysis implies that the finite versus non-finite distinction is a syntactic rather than a morphological one, as the non-finite infinitival form is identical to the finite subjunctive, imperative, and plain non-past forms. This turns the distinction between these identical forms into one of clause types.

As already pointed out in Section 5.2, the number and kinds of syncretisms that different verbs show is variable and deserves further study. For example, from a comparison of the two sample verbs TALK and GIVE we see that TALK, like all regular verbs, has four different forms, while GIVE has five. However, if we also take into account the different vowel in the preterite we might as well state that there is a base form syncretism with five of the six forms, plus an additional base form (with a different vowel) as the preterite. The different patterns of syncretisms will be discussed in more detail in the discussion of irregular verbs in Section 5.3.5.1.

The semantics of the non-past and the past tense form require some comment. The non-past is used not only to refer to events which habitually happen at a time including the present (Mary leaves for work at eight o’clock) and for states which are currently in force (Wellington is the capital of New Zealand, This coffee tastes wonderful), but also for eternal truths (The sun rises in the east) and for ongoing events in the here-and-now, as in a running commentary (Rooney scores for England!) or stage directions (Mrs Smith picks up her knitting) and (p.66) performatives (I apologize for what I said). In academic discourse, it may be used to discuss the content of any written text, whether modern or historical (Austen depicts her heroines with great honesty). It is also used for scheduled future events (The plane arrives at noon tomorrow), and in subordinate clauses more generally for future events (He will tell you when he arrives). Despite our label of ‘non-past’ it may be used to narrate a presumed past event, typically in colloquial language or in jokes (A horse walks into a bar, and the barman says, ‘Why the long face?’).

The past tense is used as a narrative tense, to relate completed events in the past, but also to mark attitudinal distance from an event, for instance for reasons of politeness (I wondered whether you had time to see me) or to express a hypothesis or doubt (If you left now, you could be in Edinburgh before dinner). For further details see, for example, Quirk et al. (1985) or Huddleston (2002b).

We will now turn to the discussion of the distinction between regular and irregular verbs, a distinction that we have already used in our discussion without proper justification.

5.3.2 Regular versus irregular verbs

As shown in Tables 5.1 and 5.2, we can distinguish between six morphosyntactic categories that are morphologically encoded. Depending on how these categories are realized, grammarians have introduced the distinction between regular and irregular verbs. In spite of what the terminology suggests, this distinction is not unproblematic, as will become clear as we go along.

According to most analyses, regular verbs are those that form their past tense forms with the predictable allmorphs of the so-called regular suffix ‑ed (as in showed, talked, ended), while all other verbs are irregular. There is also some irregularity found in the marking of 3rd person singular non-past, but this irregularity is restricted to the three verbs in standard varieties: BE, DO, and SAY. The respective forms are is, does /dʌz/ and says /sez/ (~ regular /seɪz/ in some varieties), and do not provide a sensible basis for a general category distinction. The recent verb text has, in some varieties, irregular third persons, either as texes or as textes, with the /ɪz/ allomorph occurring in an unpredictable environment.

The irregular verbs as defined above may form their preterite and past participles by various other means, for example by ablaut (singsangsung), by suppletion (gowentgone), by partial suppletion and addition of an unexpected ‑t (bringbroughtbrought), by having the suffix ‑en in the participle form (showshowedshown), or some combination of these (givegavegiven). There is even a group where all three forms concerned are homophonous (putputput). Thus, there seems to be a clear dichotomy of a nicely predictable set of forms and a mixed bag of largely unpredictable miscellaneous forms.

One problem with such a view and the above definition of what is regular and irregular, is, however, that it is not always clear which verbs should be treated as irregular. Sometimes the right suffix allomorph is accompanied by some additional change in the base, as for example (p.67) with havehad, dodid, teachtaught, which makes the forms appear regular in terms of suffixation, but irregular in terms of base allomorphy. Another problem is that some verbs have variably regular and irregular past tense forms, for example bettedbet, knowedknew, learnedlearnt, which would make them belong to both classes. Traditional wisdom also holds that irregular verb forms need to be stored in the mental lexicon, while the regular suffix is generally applied to new verbs, and the regular verb forms need not be stored (and mostly are not). However, as we will see in some of the discussions to follow, this is a rather simplistic view of the matter that is unable to account for a number of empirical facts. Finally, the terminology may suggest that the irregular verbs form their preterites and past participles in rather arbitrary ways. As we will see, however, there are a number of clear sub-patterns within the set of irregular verbs.

In spite of these problems we will use the established distinction here and treat all those forms as regular whose past tense is fully predictable as being marked exclusively by the expected allomorph of the ‑ed suffix, with no accompanying change of the base. All other verbs are treated as irregular. Those verbs that have both regular and irregular variants of the past tense will be treated with their irregular variants in the section on irregular verbs. The theoretical and psycholinguistic status of the distinction between regular and irregular will be discussed in more detail in Section 5.3.5.2.

One could of course ask the question whether regular verbs and irregular verbs might also differ in other respects, that is, whether the difference in inflectional patterning is accompanied by other, non-inflectional properties that distinguish the two kinds of verb. And indeed, such differences can be found.

For example, it has long been observed that, on average, irregular verbs tend to be of higher frequency than regular verbs. Empirical studies using corpus frequencies (e.g. Baayen and Moscoso del Prado Martín 2005) have shown that the probability of a verb being irregular dramatically increases with higher token frequencies. For illustration, we may look at the frequency of verbs in the BNC lemmatized word list (Killgariff 2006), which contains the 1,281 most frequent verbs as found in this corpus. The top 13 verbs in this list are all irregular, with the first regular verb, LOOK, occupying rank 14. In contrast, there are only two irregular verbs among the bottom 100 verbs, BET and BLEED. High frequency protects word forms from falling out of memory, and it therefore does not come as a surprise that the most irregular verb is also the most frequent one, BE, which is in fact the second most frequent word of English overall (after the), according to the BNC frequency list.

Apart from a frequency difference, there is also a difference between regulars and irregulars that concerns their morphological families. A morphological family of a given word consists of all morphologically complex words (i.e. types) in which that word occurs as a constituent. For example, the morphological family of accept would contain accept, acceptable, acceptably, acceptability, acceptance, accepter/acceptor, accepting, acceptive (based on the attested forms in COCA). Verbs with irregular preterites tend to have morphological family sizes that are significantly smaller than those of regular verbs, although the irregulars tend to (p.68) have a higher token frequency and would therefore be expected to enter further derivation, all other things being equal (Baayen and Moscoso del Prado Martín 2005).

Finally, there is some debate about the role of semantics in verbal inflectional patterning. Contrary to standardly expressed views that the distinction between regular and irregular verbs is one of form only (e.g. Kim et al. 1991), there is robust empirical evidence that regular and irregular verbs may differ in their semantic properties. Evidence for semantic effects on verb inflection has emerged from production experiments, as documented in Bybee and Slobin (1982) and Ramscar (2002). Baayen and Moscoso del Prado Martín (2005) provided large-scale distributional evidence on the role of semantics in inflection. Controlling for frequency, the authors show that the number of synonyms is larger for irregular verbs. Furthermore, sets of synonyms for irregular verbs are much more likely to contain other irregular verbs than sets of synonyms of regular verbs are to contain other regular verbs. This means that irregulars tend to live in a semantically more densely populated space, having more semantic relations to other words and being more similar to each other than regulars. Incidentally, this may have contributed to the survival of the many irregular forms in the history of the language, as forms with stronger semantic ties to other words tend to have stronger lexical representations in memory.

In sum, it seems that the difference in the inflectional behaviour of verbs goes together with differences in other areas. We will return to some of these facts in the theoretical discussion in Chapter 23.

5.3.3 Regular verbs

Regular verbs have only four different forms, one of them being the plain form, already dealt with above. In the following we will discuss the three suffixed forms and some issues relating to the orthography of the suffixed forms.

5.3.3.1 Preterite and past participle

Regular verbs are characterized by the fact that their preterite and past participle are fully predictable and involve the suffixation of one of three allomorphs to the base form. The allomorph /ɪd/ or /əd/ occurs after base-final /d/ and /t/. All other bases take /d/ or /t/, depending on the voicing of the base-final segment. Bases with voiced final segments take /d/, bases with voiceless final segments take /t/. This is illustrated in (2). We transcribe the syllabic allomorph as /ɪd/, ignoring the considerable variation between /ɪd/ and /əd/.

(2) Verb inflection

(p.69) 5.3.3.2 3rd singular non-past

The 3rd singular non-past form is derived from the base form by adding the pertinent allomorph of the suffix ‑s. This allomorphy is similar to the allomorphy of the plural ‑s (from which it differs only by not being subject to haplology), and also similar to that of the past tense suffix (in the general pattern of variation). Verbs ending in a sibilant (i.e. one of the sounds /s, z, ʃ, ʒ, tʃ, dʒ/) take the allomorph /ɪz/ or /əz/, all other bases take either /z/ or /s/, depending on the final segment of the base. If the base ends in a voiced segment the voiced allomorph /z/ is chosen, if not, the unvoiced allomorph /s/ is chosen. Examples are presented in (3). As with the past tense suffix, the variation between the two epenthetic forms is ignored.

(3) Verb inflection

5.3.3.3 Present participle

Apart from the plain forms, the present participle is the only form that is of the same structural make-up for all lexical verbs, no matter whether regular or irregular. It is derived by adding the suffix ‑ing to the base form. The suffix has two allomorphs, /ɪŋ/ and /ɪn/, the latter being prescriptively viewed as a non-standard form, though it may be heard from otherwise standard speakers. Variation between these two has been frequently subject to sociolinguistic investigation (e.g. Trudgill 1974).

5.3.3.4 Orthographic issues

The default spelling of the suffixes is <ed>, <s>, and <ing>, but the attachment of these strings to the orthographic representations of the verbal bases may bring with it some adjustment in the orthographic representations of certain verb forms. These adjustments concern the doubling of base-final consonant letters before <ed> and <ing>, the deletion of final mute <e> before the same suffixes, and the change of <y> to or <ie> before <ed> and <s>, respectively. These changes follow the general principles of morphology–orthography interaction in morphology that are laid out and illustrated in Chapter 4.

In addition to what follows from these general principles, the following peculiarities should be noted with verbal inflected forms. First, the 3rd singular suffix ‑s has two spelling variants, <s> and <es>, with the latter occurring after bases ending in <s> (<kisses>), <z> (<buzzes>), <x> (<boxes>), <sh> (<pushes>), and <ch> (<matches>) and the relevant words ending in <o> (e.g. <vetoes>, see Chapter 7). The suffix ‑ed has also two spelling variants, <ed> and <d>. The latter is used with words that end in <e> (baked, created, dyed, hoed), (p.70) and after specific base allomorphs of certain verbs (e.g. did). The spellings <paid>, <laid>, and <said> (and the spellings of pertinent derivatives, such as <mislaid>) are lexical exceptions to the rule that base-final <y> is preserved after vowel letters if it is part of a vowel digraph (as in <employed> or <stays>). With regard to ‑ing, bases ending in <ie>, such as <die, lie, tie, vie>, have <y> instead of <ie> before ‑ing.

5.3.4 Irregular verbs I: The primary verbs BE, DO, and HAVE

5.3.4.1 BE

Apart from its being a lexical verb, specifically, a copula, BE is used as an auxiliary verb in progressive aspect constructions and in passive sentences. The full paradigm of BE is given in (4). This verb is the only one in standard English that distinguishes the three person categories in the singular. Instead of a distinction between 3rd singular and non-3rd singular, BE encodes all three persons in the singular differently in the non-past (amareis). In the past tense paradigm there is a distinction between 2nd singular and non-2nd singular (werewas). Furthermore, (4) has a column for the counterfactual. As already mentioned, this category is not formally marked with any other verb, and with all other verbs the preterite is used in the respective clauses, which means that the counterfactual in English is a syntactic, and not a morphosyntactic category. With BE, however, we find a morphological distinction between the counterfactual and the past tense encoded for the 1st and 3rd singular.

(4) Verb inflection

What is also striking with this verb is the degree of suppletion. One can differentiate between six different bases that are phonologically quite distinct (am, is, are, was, were, be).

Although BE is the most irregular verb of English with the highest number of different forms and morphosyntactic distinctions, the distribution of its forms has things in common with other irregular verbs and indeed with regular verbs. The non-finite forms are all derived from a single base, be, as is the case in regular verbs and many irregular ones. Furthermore, in standard varieties there is no person distinction in the plural forms, and the levelling of forms (p.71) in the past tense goes beyond the levelling found in the non-past. Finally, the suffixes used for the participles are also used by other verbs (‑ing by all verbs, ‑en by many irregular verbs).

The combination of negation and BE brings about some interesting forms and issues, which we will discuss in Section 5.4.4. There is also some variation in the forms of BE used in different dialects of English, as for example the number levelling of the preterite forms (I waswe was), the use of been as a preterite in many non-standard dialects, and the use of invariant be in African-American English.

5.3.4.2 DO

Like BE and HAVE, DO can be a lexical verb or an auxiliary. The auxiliary verb DO is used in all constructions requiring DO-support, namely in negated sentences, interrogatives, negative inversion, emphatic, and elliptical constructions. The paradigms of the lexical DO and the auxiliary verb DO differ, in that the auxiliary does not have participle forms and the auxiliary behaves differently in negative contraction. Negative contraction will be discussed with that of the other auxiliaries in Section 5.4.4. In some varieties of English, the auxiliary and the lexical verb display further morphological differences. For instance, the regional English speakers described in Cheshire (1982: 34–6) have a distinction between lexical dos /duːz/ and auxiliary do, both used for all persons in the non-past. Furthermore, there are varieties that have a paradigm dodonedone for the lexical verb, and dodid for the auxiliary (Anderwald 2009: 126). The form done is also used as a perfective aspect marking auxiliary in Southern U.S. English and African-American English (see Green 1998).

The paradigm for the lexical verb in standard English is given in (5):

(5) Verb inflection

Like regular verbs, the verb DO has six forms in its paradigm, and it also shows the predictable allomorphs of the 3rd singular and past tense suffixes, the same syncretism of the two plain forms as regular verbs and the same suffix for the present participle. However, we find base allomorphy with the 3rd singular (/duː/ ◆ /dʌ‑z/), with the preterite (/duː/ ◆ /dɪ‑d/), and with the past participle (/duː/ ◆ /dʌ‑n/), and the past participle suffix is one that occurs with irregular verbs, not with regular verbs.

(p.72) 5.3.4.3 Have

The primary verb HAVE is similar to DO in that its paradigm also closely resembles that of regular verbs, with only some base allomorphy disturbing the otherwise regular picture. Consider (6).

(6) Verb inflection

The 3rd singular, the preterite, and the past participle all show the predictable regular suffixes, but do not have the base-final /v/. Disregarding this base allomorphy, the formal distinctions and syncretisms are exactly the same as those of regular verbs.

5.3.5 Irregular verbs II: all others

In this section we take a closer look at the large number of remaining irregular verbs, all of which show some deviation from the regular paradigm discussed above. Their exact number is hard to determine, as it at least partially depends on how lexemes are treated that are putatively derived from irregular verbs (e.g. misspell). In many cases, these are no longer morphologically complex (e.g. become), but others certainly are (deepfreeze, outswim). Another problem is of course the sampling procedure itself. Some lists we find in the literature contain verbs that other lists do not, irrespective of morphological complexity. Due to these problems, the number of irregular verb lexemes given in the literature ranges, for example, between 167 (Anderwald 2009: 3) and 470 (listed by englishpage.com). Palmer et al. (2002: 1600–10) give 176, Quirk et al. (1985: 104) 261. Figures are of course lower in those sources that define the term ‘irregular verb’ in a more restrictive way, for example by reserving it for verbs that use vowel change to mark the past tense, or for verbs that do not use a suffix.

What is clear is that the class of irregular verbs seems much less open than the class of regular verbs. Claims that it is a closed class are certainly too strong, as new verbs can in principle enter the irregular class by compounding, derivation, or back-formation. For example, we found 126 such irregularly inflected forms at englishpage.com that were not listed by Quirk et al. (e.g. HAND-FEED, INTERWEAVE, OUTSWIM, OVERHANG, PRESHRINK, REGRIND, TELECAST, TEST-DRIVE, TYPEWRITE, UNDERBID, UNHANG). There are extremely rare cases where verbs move from the regular class into the irregular one. A well-known historical case (p.73) involved TEACH, as a consequence of various sound changes that differentially affected different forms.

5.3.5.1 The classification of irregular verbs

The kinds of deviations from the regular pattern which we find with irregular verbs are manifold and have been the source of diverse attempts at categorizing the verbs into different classes. All classifications refer exclusively to properties of and relations between three forms—the plain form, the preterite, and the past participle—as the other two forms, the 3rd singular and the present participle, are nicely predictable for all lexical verbs, with the exception of BE, HAVE, DO, and SAY, as discussed above. Many irregular verbs have the suffix ‑en, which has the allomorph /n/ after base-final vowels (and base-final /r/ in rhotic varieties), and /ən/ elsewhere.

Turning now to existing descriptions and classifications of irregular verbs we can state that the number and nature of the proposed classes differ from researcher to researcher, depending on which particular feature or features are chosen as the basis for the classification. Most available classifications are based on the forms as spoken, which is an obvious road to take as changes in vowel quality across the different morphosyntactic categories are very frequent and important, and phonological vowel quality is not mapped straightforwardly onto the spelling (consider plain form <read> /riːd/ ◆ preterite <read> /red/).

Quirk et al. (1985), to begin with, distinguish seven classes (sometimes with subclasses) on the basis of the combinations in the distribution of three properties: whether there is a suffix used, whether preterite and past participle is identical, and whether all three forms have the same vowel. Palmer et al. (2002: 1600–9) set up four classes (with various subclasses) based on four central criteria, that is whether there is some allomorph of the regular suffix discernible, whether there is only a vowel alternation, and whether the past participle is formed with the suffix ‑en. The fourth class is residual and contains ‘other formations’. The theoretical basis of these two classifications is not very clear. Both taxonomies rely on formal similarities between different sets of forms, with some of the criteria being syntagmatic in nature (i.e. suffixation), some of them paradigmatic in nature (identity of vowels, or identity of certain forms, across the paradigm). Why it is exactly these similarities and not others that form the basis of the taxonomy seems to be partly arbitrary, partly a matter of frequency of particular similarities, and partly even a reflection of the historical strong verb classes as they survive in the modern system.

For further illustration of the problem of classification let us first look at a classification based on the presence and quality of vowel change from plain form to the preterite. On the basis of this feature, Baayen and Moscoso del Prado Martín (2005) end up with 32 classes (one of them containing all verbs with no vowel change). In their sample, which is based on CELEX, the number of verbs in each class ranges from only 1 to 38. Another classification based on vowel alternations is also possible, focusing on whether forms have the same or different vowels. We have classified the verbs from the Quirk et al. list accordingly and (p.74)

Table 5.3 Kinds of vowel alternations

Class

Vowel pattern

Example

A (N=39)

a–b–c

sing ◆ sang ◆ sung

B (N=134)

a–b–b

keep ◆ kept ◆ kept

C (N=42)

a–b–a

give ◆ gave ◆ given

D (N=3)

a–a–b

shear ◆ sheared ◆ shorn

E (N=75)

a–a–a

hit ◆ hit ◆ hit

ended up with the classification in Table 5.3, listing the kinds of vowel patterns using the letters a, b, and c to denote different vowels, and using capital letters A to E to label the resulting verb classes. Note that there are many lexemes among the 261 irregular verbs given in Quirk et al. (1985) that have more than one irregular paradigm, which increases the number of paradigms discussed from 261 to 293. We also give the frequency of the pattern per class. We can see that the most frequent patterns are a–b–b and a–a–a, covering more than two thirds of the verbs.

Anderwald (2009) sets up five classes on the basis of whole form syncretism. Given three forms in the paradigm, and given that of these, none, two, or all three could be identical, one arrives at five logically possible syncretisms, which in fact are all attested and which constitute Anderwald's five irregular verb classes (with various subclasses). This is schematically represented in Table 5.4. All combinations occur, although with varying frequency. The type frequency of each pattern, based on the Quirk et al. (1985) list, is given in the first column, including more than one irregular pattern for a given verb if listed and pertinent.

This is an insightful classification as it allows us to relate the irregular patterns to the regular pattern. Thus all regular verbs and a large number of irregular verbs distinguish between the plain form and that for the preterite and the past participle (class 2, cf. regular TALK and irregular HOLD). One could therefore claim that this difference is a typical characteristic of the English verbal system, as was done by early authors such as Lowth (1762: 85–6, cited in Anderwald 2009: 98). The plain form and the preterite are the same for 46 verbs. Four of

Table 5.4 Syncretism of whole forms

Anderwald's class

Plain form

Preterite

Past participle

Example

1 (N=108)

a

b

c

give ◆ gave ◆ given

2 (N=129)

a

b

b

hold ◆ held ◆ held

3 (N=10)

a

b

a

come ◆ came ◆ come

4 (N=4)

a

a

b

beat ◆ beat ◆ beaten

5 (N=42)

a

a

a

hit ◆ hit ◆ hit

(p.75) these verbs have different past participles (e.g. BET with its past participle variant betted, and BEAT with its past participle variant beaten), which makes them class 4 verbs, and 42 of them also have a past participle of the same form (class 5, e.g. COST, CUT, HIT, HURT, PUT). There are only ten verbs that have the same past participle form as the plain form but a different past tense form (class 3).

Another classificational approach based on syncretism could focus on different kinds of bases for one verb if there is a suffix present in at least one of the three forms. Thus, we could posit that some verbs have more than one base form and that these different bases have different combinatorial properties. For example, GIVE may be analysed as having one base form give, which is used as a plain form, as the base for the suffix ‑s and, crucially, also for the past participle suffix ‑en. This type of approach gives us another kind of syncretism, base syncretism. Table 5.5 summarizes base syncretism patterns for suffixed irregular verbs, assigning capital letters to the classes.

All of the five possible syncretism combinations are attested in the Quirk et al. list. Class A shows that even suffixed verbs can have three different base forms (cf. drivedrovedriven). Class B verbs take the same base for the participle as for the past tense (cf. also awakeawokeawoken); in class C the plain form and participle bases are syncretic. The constellation of bases labeled class D is rare and only two verbs show it: SHEAR and SWELL (latter with the forms swellswelledswollen). Class E shows an interesting implicational pattern. All verbs in this class have an irregular participle suffix, but they differ in the marking of the preterite, which can go either unmarked (beatbeatbeaten), or have a regular suffix (show ◆ showed ◆ shown), or an irregular suffix (spellspeltspelt).

Our overview has shown that there are very many conceivable properties and relations between forms that could be employed as a basis for classification, some of which have been used in the literature, some not. In (7) we list a number of features and their respective potential values. Most of these, sometimes in combination, have been used in the literature to set up classes or subclasses. Different authors have used different kinds of codings for the same kind of structural phenomenon, and we list the different codings separately. We also indicate some of the sources that have used the respective criterion for their classification.

Table 5.5 Syncretism of bases, verbs with suffixed forms

Class

Plain form base

Preterite base

Past participle base

Example

A (N=27)

a

b

c

ride ◆ rode ◆ ridden

B (N=55)

a

b

b

keep ◆ kept ◆ kept

C (N=33)

a

b

a

give ◆ gave ◆ given

D (N=2)

a

a

b

shear ◆ sheared ◆ shorn

E (N=34)

a

a

a

show ◆ showed ◆ shown

(p.76)

(7) Verb inflection

Let us summarize our discussion of possible classifications of the similarities between different sets of verbs. There is a long (and potentially open-ended) list of criteria and their combinations that can be employed to group forms into classes that share certain properties, and the general absence of insightful pertinent methodological principles leads to classifications that may be intuitively plausible, but mostly seem to lack any theoretical or practical relevance. In particular, the whole question of productivity appears to have been ignored. This will be discussed in the next section.

5.3.5.2 Productivity, variability, and irregularity

There is a widespread opinion that only the regular past tense formation is productive, while irregular verb inflection is not. In order to see whether this is really the case we will first look at the variation that many irregular verbs show and then revisit the potential evidence that irregular past tense inflection is not totally unproductive.

Of the 261 verbs listed in Quirk et al. (1985), 72, that is, almost three out of ten, are listed with variant forms (some of them regular, some irregular). If one included the great many non-standard forms that can be found in colloquial usage (e.g. knowed with hundreds of attestations in COCA), this figure is even higher. This means that the system is in fact much (p.77) more variable than is allowed for in theoretical accounts which claim that irregular forms block the formation of regular forms. Of the 72 verbs that have variable forms, 66 have only two competing patterns, but four verbs have three competing patterns and two verbs have four competitors in some of the cells of the paradigm. If we had widened the data set from that provided by Quirk et al., we would have found even greater variability. The different sets of verbs are illustrated in (8), (9), and (10).

(8) Verb inflection

(9) Verb inflection

(10) Verb inflection

(p.78) What is remarkable with the verbs that have more than two competing patterns is that all competing patterns are irregular ones, which raises the question of whether in the set of verbs with only two competing forms it is always a regular pattern competing with an irregular one, as suggested by the example in (8). This is not the case, however. Of the 66 verbs that have two competing patterns, 19, that is almost one-third, have two irregular patterns competing with each other.

A closer inspection of this variability reveals that all forms with an a–b–c pattern (in the vowel or in the whole form) have a variant with an a–b–b pattern. Furthermore, those forms whose bases end in /t/ or /d/ have a tendency to develop a–a–a variants, either instead of a–a–b suffixed variants (e.g. bitbitbit ~ bitten), or instead of forms with different vowels (shitshat ~ shit, or spitspitspit ~ spitspatspat). The recent usage of text as an invariable verb in some varieties of English illustrates this point perfectly. In sum, there are three patterns that seem to be able to attract new verbs, the regular verb pattern, the two-stage a–b–b ablaut pattern (in various instantiations) and the complete homophony pattern a–a–a.

If we go beyond the set of words from Quirk et al., and also include, for example, the variable irregular verbs that Anderwald (2009) investigates as ‘non-standard’ verb forms, this picture is independently corroborated. Of these verbs, we find 16 attested in COCA with a frequency of more than five regular preterite and past participle forms per lexeme (BLOW, BURST, BUST, CATCH, COME, DIG, DRAW, GIVE, GROW, HEAR, KNOW, RUN, SEE, SHINE, SING, SINK, TEACH, TELL, THROW). In addition, three verbs show up with irregular competing past tenses, levelling the respective paradigms to an a–b–b ablaut pattern (seeseenseen, singsungsung, sinksunksunk—and, indeed, all of the verbs which standardly have /ɪ/‑/æ/‑/ʌ/ ablaut). There are even places where verbs borrow forms from other lexemes, as with the confusion of bought and brought in several varieties of English.

Further evidence concerning the potential productivity of some of the irregular patterns can be gleaned from studies involving past tense formation with nonce verbs, such as Bybee and Moder (1983), Prasada and Pinker (1993), Marcus (1995, 1999), Xu and Pinker (1995), Albright and Hayes (2003) or Wagner (2010). These experiments have also shown that speakers create irregular past tense forms, and they do so especially if the nonce base form has certain phonological properties. For example, generalizing across different studies, ablaut forms (instead of regular preterites) seem most likely with nonce forms that are monosyllabic, have a complex, /s/-initial onset, have /ɪ/ as their base vowel, and end in a velar nasal (e.g. Bybee and Moder 1983; Albright and Hayes 2003; Wagner 2010: ch. 3.2.2). Thus, the data in these studies show that there are rather clear patterns, but these patterns cannot be generated by very general deterministic rules that produce a correct output from any given input form. Instead, one needs either highly constrained rules that are specified to apply to very small subsets of the data (as in Albright and Hayes 2003), or analogical models that can operate on the basis of similarity of forms to existing (cohorts of) forms in the lexicon (e.g. Keuleers 2008).

(p.79) 5.3.5.3 Orthographic issues

The regular spelling alternations, as discussed in Chapter 4 and in Section 5.3.3.4, apply, and there are not many orthographic idiosyncrasies to be noted in addition. The irregular suffix ‑en has three spelling variants, <en>, <n>, and <ne>. The variant <n> occurs after bases that end in <e>, <y>, <w>, <r> (<given>, <lain>, <shown>, <sworn>), the variant <ne> is restricted to two verbs (<borne>, <done>), and <en> occurs elsewhere.

5.3.6 Defective paradigms

As mentioned in Section 5.2, the core modals are characterized by the fact that they have defective paradigms, that is they lack certain forms. There are also, however, a few lexical verbs that are, at least to some extent, defective. In most instances, this means that speakers are unsure about what the relevant part of the verb might be, and avoid the issue whenever possible. This gives rise to a gradient notion of defectiveness: some forms of some verbs are less frequently used than might be expected, and in the limiting case are not used at all. Instances are the verbs STRIDE and BEWARE, the former of which, for some apparently arbitrary reason, does not have a generally agreed past participle (strode, strid, stridden, strided). The verb BEWARE is even more impoverished and has only the plain form, with the consequence that we only find it in infinitival, subjunctive, and imperative constructions.

There are also, however, less clear cases. REPUTE and RUMOUR, for example, have been said to be defective (Ward et al. 2002: 1435), as they are assumed not to occur outside passive constructions and therefore not to have the pertinent other forms. However, the example in (11), from COCA, implies that this is really a preference rather than a strict absence of forms, though it is less clear whether this is the case for rumour.

(11) Verb inflection

Ward et al. (2002: 1435–6) also mention that the verb SAY cannot occur in an active verb + object + infinitive construction (*They say Kim to be a manic depressive). This looks like a matter of verbal diathesis rather than defectiveness.

Another potential candidate for defectiveness is the lexical verb WILL ‘wish, want’, which behaves syntactically like the modal verb and could therefore be classified as another primary verb, together with BE, DO, and HAVE. Lexical WILL has the full set of forms we know from regular verbs (willwillswilledwilling), and even an alternate preterite form would, but far-reaching restrictions on the usage of some of these forms make this verb appear to live on the verge of defectiveness. The first peculiarity concerns the 3rd singular non-past form. While willed is quite frequent as a preterite or past participle (with more than a hundred attestations in the BNC) with no apparent restriction as to its subject (apart from sentience), (p.80) the 3rd singular form wills seems highly restricted, especially in BrE. In the BNC we find only 12 attestations of wills, nine of which have God as a subject, and such usage appears to be very formal. The same is apparently not true in NAmE. The form wills occurs in COCA more than 300 times with a wide variety of subjects. Some subjects denote power figures (God, Allah, the president) but others are attested (proper names, a young Polish boy, pronouns that do not refer to power figures). We frequently find wills followed by a reflexive (she wills herself ). There is no particular formal flavour. Another peculiarity is the alternate preterite form, would. Historically, this form originated as a preterite and subjunctive form, but is now largely restricted to collocations with rather, as in (12).

(12) Verb inflection

The analysis of potentially defective verbs suggests that defectiveness, outside the domain of the core modal auxiliaries, is best seen as a very rare, and perhaps gradient, phenomenon.

5.4 Auxiliary verbs

In this section we will look more closely at the morphological behaviour of auxiliary verbs. The auxiliary verbs BE, DO, and HAVE were treated in Section 5.3.4, and we will include them here only with regard to those aspects of auxiliary behaviour that were not discussed in that section, namely weak forms, clitics, and negation.

Potential candidates for auxiliary status are the verbs in (13), where (13a) gives the core modal auxiliaries and (13b) three auxiliaries whose morphological behaviour is, however, quite different from that of the core modals.

(13) Verb inflection

We will deal with each group in turn in the following two subsections, and subsections 5.4.3 and 5.4.4 will then be devoted to clitics and weak forms, and to the negative forms of auxiliaries.

5.4.1 Core modal auxiliaries

5.4.1.1 The semantics of modals

The semantics of modals have been dealt with in many publications including Coates (1983), Quirk et al. (1985), Palmer (1974, 1990, 2001), Huddleston (2002b). These studies indicate that the whole question of the semantics of the English modals is far from simple and not (p.81) easily summarized. We have no information to improve on the descriptions in such texts, and the complex semantics associated with the modals has little direct relevance to the forms. We should also note that there are many periphrastic expressions which carry modal meanings, but which are not usually listed as ‘the modals’: forms such as be able to, be going to, be possible that, be willing to, better (not), have (got) to, maybe, perhaps. We should also note that there is a great deal of dialectal variability in the form and meaning of modals in English. This ranges from the existence of ‘double modals’, such as might could in some varieties of English, to the fact that haven’t got to may mean ‘there is no obligation to’ in some dialects and ‘must not’ in others, to the fact that in some varieties shall is never or very seldom used at all. Overall statements of precisely what the modals mean ‘in English’ are thus difficult. Rather than spend a great deal of space on this topic, we present Figure 5.1 (adapted from Coates 1983: 26) to illustrate the complexity of the meanings associated with the various modal verbs. While many readings are presented in Figure 5.1, and the mapping between forms and meanings is clearly complex (as Figure 5.1 illustrates), the meanings illustrated are closely related to one another, to the extent that a given modal in context may be ambiguous between readings.

Verb inflection

Figure 5.1 Modal verbs and their meanings/readings. Bold lines represent primary use, thin lines represent secondary or infrequent uses.

Adapted from Coates (1983: 26).

(p.82) 5.4.1.2 The form of modals

As mentioned in Section 5.2, the core modals can be morphologically characterized by their defective paradigms. They do not have non-finite forms that would allow them to occur in the respective non-finite syntactic constructions, and they also lack 3rd person singular non-past forms.

Since we have listed could, might, should, and would as separate modals on the basis of their meaning, we might argue that the core modals do not even have a preterite form. However, there are places where could (and the other forms listed here) can act as the preterite of CAN, etc. under sequence of tense rules in indirect speech. This is illustrated in (14).

(14) Verb inflection

It thus seems that could and would have a dual function: one as a preterite form, and one as a separate modal verb with their own meanings. Might and should are awkward as past tense forms of may and shall, partly because of the restricted usage of may and shall, and partly because the meanings of the erstwhile preterites have diverged semantically from the original base to such a great extent.

(15) Verb inflection

While (15b) is clearly possible, it sounds pedantic or old-fashioned, while (15d) does not sound as though it is equivalent to (15c).

We might thus argue that we have modals like must and ought which have no preterite, modals like can and will which have a preterite form that is homophonous with another modal, and modals like may and shall which are transitional between the two.

5.4.2 Other auxiliaries

The three verbs given in (13b) are rather intricate cases, with NEED and DARE sharing some peculiarities. These two modals behave syntactically both as lexical verbs and as auxiliary verbs. For example, they can take both bare infinitives (a sign of being an auxiliary) and (p.83) to-infinitives (a sign of being a lexical verb). Interestingly, this distinction has other correlates, syntactic and morphological. Huddleston (2002b: 111) implies for DARE that the auxiliary use goes together with a lack of 3rd person marking in the non-past. This is, however, not the case, as the many attestations to the contrary show, two of which, from COCA, are shown in (16).

(16) Verb inflection

Given that the morphological and syntactic facts do not clearly map onto each other, it is not clear whether a lexical distinction between auxiliary DARE and lexical DARE is warranted at all. Rather, we find variant forms expressing the same morphosyntactic property (e.g. 3rd singular non-past).

With NEED, however, we do find a neat complementary distribution of structural properties. When NEED is used with a bare infinitive, that is as an auxiliary, its paradigm is defective and has only one form, the non-past need, and no preterite, thus aligning with MUST and OUGHT. This is illustrated in (17a–c), from COCA.

(17) Verb inflection

The verb USE as an aspectual marker is also ambiguous in its status. For many speakers it always behaves syntactically like a lexical verb, for others it has some of the syntactic properties of auxiliaries. The construction is only used with anterior temporal reference, which has the morphological consequence that, independent of syntactic behaviour as a lexical or auxiliary verb, only the preterite and past participle forms occur. This has led to some variation in the spelling of the form too. Thus one can easily find attestations of <use to>, where standard orthography would have <used to>, as in (18), from COCA.

(18) Verb inflection

Furthermore, the construction is incompatible with the present participle (*She was using to come by every day) and the verb thus ends up with a highly defective paradigm.

(p.84) 5.4.3 Auxiliary clitics and weak forms

Being function words, auxiliaries occur most often unstressed, but must receive at least some degree of stress in certain syntactic positions, for example before the gap in stranded position (e.g. We’ll help you if we can, see, for example, Palmer et al. (2002: 1613) for further discussion). Traditionally, we speak of ‘strong’ and ‘weak’ forms. In (19), we give these allomorphs for all pertinent auxiliaries. In general, the weak forms involve a more centralized vowel and sometimes the loss of a consonant in addition. The non-strong realizations of are are treated further below.

(19) Verb inflection

Some of the verbs listed in (19) have yet other forms, namely clitic forms that attach to the word immediately preceding them. There is also one verb, DO, that has a proclitic, that is /d/, for the forms do and did, which attaches to the following pronoun you, as in non-past D’you need a hand? and past D’you have a good time? (20) gives the clitic forms.

(p.85) (20) Verb inflection

The clitic forms are not simply to be seen as further reduced variants of weak forms for at least three reasons. First, not all verbs that have weak forms have enclitic forms, second, the syntactic distribution of the clitics is more restricted than that of the weak forms, and, third, some pronoun–clitic combinations have weak and strong forms themselves (e.g. we’re and did you illustrated in (20)). Note however that, depending on the kind of environment, the distinction between some weak forms and the corresponding clitic is not always very clear. For example, the orthographic representation <’ll> after a consonant letter, as in that’ll do it, may be realized with only the clitic /l/ or with the weak form /əl/. Furthermore, some sources list /ə ǁ ər/ as weak forms of are, and not, as we do, as clitics. We opt for a clitic analysis of ’re on the grounds that a weak form analysis cannot account for the facts that the pronoun and the auxiliary form an indivisible phonological unit in /jɔː, jə, wə, ðeə, ðer, ðər/. Furthermore, there are weak and strong forms of the pronoun clitic combinations themselves, as shown in the last column of (20).

The distributional restrictions of the clitics are quite intricate and vary for different sets of forms. In general, the clitics attach to preceding NPs, but there are some systematic exceptions to this. The least restricted forms are the clitic forms of has and is. They attach to the widest range of preceding constituents, including NP subjects (Jane's not at home), some subordinators (the book that's on the shelf ), interrogative words and phrases (How's your mother?; How bad's your ear?) and a few more constructions (see Palmer et al. 2002: 1614–16 for more discussion).

More restricted is the encliticized realization of did, had, and would, /d/. It only attaches to vowels and it occurs in fewer syntactic environments than the has/is clitic 's. Which environments are possible and which ones are not, is not entirely clear, however. Not all grammars also list ’d as a clitic form of did, but the existing evidence strongly suggests its inclusion (p.86) as such. Some examples from COCA are given in (21) (we have eliminated the spaces round clitics used in the corpora for tagging purposes)

(21) Verb inflection

As with the clitic forms of has and is, the use of the ’d clitic may lead to ambiguities in specific contexts. For example, I’d cut it can mean ‘I would cut it’ or ‘I had cut it’, or how’d they find us? can mean ‘how did they find us’ or ‘how would they find us’.

The most restricted clitics are ’ll, ’m, ’re and ’ve, for which it has been claimed that they attach only to NPs that are their own and sole subjects (e.g. Kaisse 1983: 98). This seems to be true for ’ve and ’re (and trivially for ’m), as shown in (22) for ’ve, taken from Kaisse, but not for ’ll, as many attested examples like those in (23) show (taken from COCA).

(22) Verb inflection

(23) Verb inflection

One can also find spellings with two clitics attaching to one another, but in these cases the second clitic, which is usually ’ve, is a weak form and realized with a schwa. Consider (24), from COCA.

(24) Verb inflection

5.4.4 Negation

All of the auxiliaries, in contrast to all other verbs of English, have negative forms, in which the negative element /nt/, spelled <n’t>, is suffixed to the auxiliary. In these cases, the negation has sentential scope. The negative form of the auxiliary is in free variation with constructions that have the affirmative auxiliary form and the free sentential negator not, as shown in (25) and (26), from the BNC. Examples like that in (26b) may be dialectal, being more common in northern British English.

(25) Verb inflection

(p.87) (26) Verb inflection

In contrast to the free form negation, negative auxiliary forms can never be used to express VP-negation, that is a type of negation in which the negation has scope only over the VP, and not over the whole clause. This leads to a potential meaning contrast between the two constructions, at least in writing, as shown in (27).

(27) Verb inflection

The negative suffix can be attached to each of the (non-clitic) forms of the respective verbs’ paradigms. The suffix integrates phonologically with the base, which, depending on its phonological structure, may or may not lead to the addition of a syllable (cf. monosyllabic aren’t with disyllabic couldn’t). If the addition of n’t leads to an increased number of syllables, the nasal of the suffix is syllabic (e.g. /dɪdn̩t/). Bases with final /n/ do not have geminate nasals. Six of the negative auxiliary forms show deviant allomorphies or spellings, or both. They are listed in (28). Note that shan’t listed in (28) is not NAmE, and is rare or non-existent in some other varieties as well.

(28) Verb inflection

A number of additional peculiarities of auxiliary negation are noteworthy. First, the aspectual verb USE cannot take the negative suffix in all dialects or with all speakers. This is expected, given its variable status as auxiliary or lexical verb discussed above. If the verb is lexical in status, it does not have a negative form. A possible negative for used to is didn’t use(d) to, as illustrated in (29) from the BNC. (p.88)

(29) Verb inflection

Second, the verb CAN has an additional negative form, in which the free form negator is suffixed to the form cannot. This form has initial stress and, just like the negative form can’t, cannot be used to express VP-negation. The form cannot is more formal in character than the form can’t, but the syntactic distributions of can’t and cannot seem to be equivalent (pace Palmer et al. 2002: 1611). This is illustrated in (30), with examples from the BNC (although Palmer et al. are right to the extent that the construction in (30a) is rare though found in both BNC and in COCA, usually spelled <can not>).

(30) Verb inflection

Third, not all of the negative auxiliary forms that can be found in the language are uniformly considered part of the standard English system. Thus, ain’t occurs as an additional suppletive negative form of am, are, is, have, and has in many varieties, and is described as being informal or non-standard. Another such form is amn’t, which is especially frequent in IrE and ScE, but seems not to be restricted to these areas. Another variant form for the 1st singular non-past BE is aren’t, which, again contrary to what Palmer et al. (2002: 1612) suggest, can be found in both interrogative and affirmative clauses, especially in spoken registers, and more especially in BrE, as shown by the examples from the BNC given in (31).

(31) Verb inflection

5.5 Incorporation of infinitival to

The infinitive marker TO combines with seven verbs to form morphologically complex units that deserve special treatment. The verbs are WANT, GO, HAVE, USE, GOT, OUGHT, and SUPPOSE. The spellings for the respective forms of WANT TO and GOING TO, <wanna> and <gonna>, are established, for the other verbs we use <hafta>, <useta> , <gotta>, <oughta> and <supposeta> /s(ə)pəʊstə ‖ s(ə)poʊstə/. The phenomenon is also known as to-contraction, a label that suggests that it is a phonological process of fusion. However, it can be argued that we are not dealing with on-line phonological reduction due to fast speech but are faced (p.89) with complex forms in their own right, which all show some peculiarities in phonological shape and behaviour that are not readily described under the assumption of a syntactic and phonological concatenation process of the two words involved. For ease of reference, we will use the traditional term ‘contraction’ to refer to the phenomenon, but do so without commitment to the underlying analytical assumptions. We will discuss the facts in detail for WANT TO, but we find the same patterns of usage with the other forms. We therefore discuss the properties of the other verbs in more detail only if they differ from those of WANT TO.

The form wanna can be used in all contexts in which the verb form want and the infinitival marker TO, introducing the complement of WANT, are syntactically adjacent. Notably, the noun want adjacent to infinitival to does not contract (it virtually never occurs), nor do we find contraction if the to following want introduces an adjunct, or a complement of another verb. The examples of adjuncts in (32) are from Hudson (2006: 605).

(32) Verb inflection

The form wanna is only a possible realization of WANT TO in those contexts where the plain form or the plain non-past form of WANT, that is want, is possible, making incorporation with wants, wanted, wanting impossible. The form is phonetically realized as /wɒnə ǁ wɑːnə ~ wɔːnə ~ wʌnə/, with an additional alternation in the final vowel in some varieties, where schwa alternates with /ʊ/ if the following word starts with a vowel (see Hudson 2006: 606–8 for discussion). This variety-dependent alternation can also be found with the other verbs of this group.

The future-oriented progressive form going to has a variant with incorporated infinitival marker, gonna, which can be realized as /ɡənə, ɡɒnə, ɡʌnə ǁ ɡɔːnə, ɡɑːnə/. Contraction affects go only in the pertinent idiomatic usage; the homophonous verb GO cannot contract if followed by to (*He's gonna jail meaning ‘He is going to jail’).

The verb HAVE in its idiomatic obligation-expressing usage with an immediately adjacent to-infinitive can realize the two non-past forms have to and has to as /hæftə/ and /hæstə/. Notably, these forms show regressive devoicing that is not found across the board (cf. li[v]e to, do[v]etail, li[v]etrapping, fi[v]e times).

The aspect-marking construction with used to has already been described in Section 5.4.2. The contracted form of this construction useta, pronounced /juːstə/, is phonologically different from the non-aspect-marking form used to, as it does not normally alternate in stranded positions. In contrast, non-aspectual used to, as in this is not what she is normally used to cannot end in a schwa. In some dialects useta is used as a base form (as in She didn’t useta).

The idiomatic construction (HAVE) GOT TO expresses an obligation or necessity and, in its contracted form is pronounced /ɡɒtə ǁ ɡɑːɾə/. In many varieties of English, including NAmE, gotta does not rhyme with not to, which is an indication of the lexical status this form enjoys. In some varieties of NAmE, and possible others as well, gotta can be a 3rd singular form. The (p.90) form gotsta is found in NAmE and is used with all persons (contrary to what is suggested in Pullum 1997: 89), as in (33) from COCA.

(33) Verb inflection

The contracted variant for ought to is /ɔːtə ǁ ɔːɾə ~ ɑːɾə/. As pointed out by Pullum (1997), oughta does not ryhme with thought to (as in you oughta be happy versus they are thought to be happy), as the latter, at least in NAmE, cannot have a schwa. Thus, oughta, like the other verbs discussed, has its own lexically specified phonological peculiarities.

Finally, the contracted form of SUPPOSED TO, supposeta, shows the same kind of non-alternation of schwa and /ʊ/ as the previously discussed verbs. Pullum (1997: 89) provides the wonderful minimal pair of supposed to versus host to given in (34).

(34) Verb inflection

At the syntactic level, supposeta co-occurs only with the clitic form of the copula, if there is a choice (?You are supposeta take out the garbage).

Appendix Irregular verbs in English

We provide below a list of 387 irregular verb forms in English derived from the various sources at our disposal. We make no claim about the status of the forms we list beyond that we believe that they are used. Specifically, some may be standard, others non-standard or dialectal, and we make no attempt to mark such matters; we simply list the forms alphabetically in each category. The lack of a form cannot definitely show that it does not exist, merely that we have not found it listed or used. Under each fundamental verb, we list verbs created from it by word-formation, provided that they too have irregular forms: such verbs are listed in italics. Grandstanded is not listed because it is regular. Where there are alternative forms existing side-by-side there may be, but need not be, a semantic difference between them. Forms which were originally past participles but which are now only adjectives (forms such as drunken, sunken) are ignored. Forms which are phonologically regular but orthographically irregular (laid, paid) are also ignored.

Base form

Preterite

Past participle

Also regular

Notes

abide

abode

abode

Yes

be

was were

been

bear

bore

borne

forbear

forbore

forborne

overbear

overbore

overborne

beat

beat bet

beat beaten bet

browbeat

browbeat

browbeat browbeaten

begin

began begun

begun

bend

bent

bent

unbend

unbent

unbent

bereave

bereft

bereft

Yes

beseech

besought

besought

Yes

bet

bet

bet

Yes

bid

bad bade bid

bade bid bidden

<bade> may be pronounced /bæd/ or /beɪd/

forbid

forbad forbade

forbid forbidden

outbid

outbid

outbid outbidden

overbid

overbid

overbid

rebid

rebid

rebid

underbid

underbid

underbid underbidden

bind

bound

bound

rebind

rebound

rebound

unbind

unbound

unbound

bite

bit

bit bitten

frostbite

frostbit

frostbitten

bleed

bled

bled

blow

blew

blown

Yes

break

broke

broken

breed

bred

bred

crossbreed

crossbred

crossbred

inbreed

inbred

inbred

interbreed

interbred

interbred

outbreed

outbred

outbred

overbreed

overbred

overbred

bring

brought

brought

build

built

built

overbuild

overbuilt

overbuilt

prebuild

prebuilt

prebuilt

rebuild

rebuilt

rebuilt

burn

burnt

burnt

Yes

sunburn

sunburnt

sunburnt

Yes

burst

burst

burst

Yes

bust

bust

bust

Yes

buy

bought

bought

overbuy

overbought

overbought

cast

cast

cast

broadcast

broadcast

broadcast

Yes

forecast

forecast

forecast

miscast

miscast

miscast

overcast

overcast

overcast

rebroadcast

rebroadcast

rebroadcast

Yes

recast

recast

recast

roughcast

roughcast

roughcast

sand-cast

sand-cast

sand-cast

telecast

telecast

telecast

typecast

typecast

typecast

catch

caught

caught

Yes

chide

chid

chid chidden

Yes

choose

chose

chosen

cleave

cleft clove

cleft cloven

Yes

cling

clung

clung

clothe

clad

clad

Yes

come

came come

come

Yes

become

became

become

overcome

overcame

overcome

cost

cost

cost

creep

crept

crept

Yes

(regular in combination creep out)

crow

crew

crew

Yes

cut

cut

cut

recut

recut

recut

undercut

undercut

undercut

deal

dealt

dealt

misdeal

misdealt

misdealt

redeal

redealt

redealt

dig

dug

dug

Yes

dive

dove

dived dove

Yes

do

did done

done

misdo

misdid

misdone

outdo

outdid

outdone

overdo

overdid

overdone

predo

predid

predone

redo

redid

redone

undo

undid

undone

drag

drug

drug

Yes

draw

drew

drawn

Yes

outdraw

outdrew

outdrawn

overdraw

overdrew

overdrawn

redraw

redrew

redrawn

withdraw

withdrew

withdrawn

dream

dreamt

dreamt

Yes

daydream

daydreamt

daydreamt

Yes

drink

drank drunk

drunk

Yes

outdrink

outdrank

outdrunk

overdrink

overdrank

overdrunk

drive

drove

driven

outdrive

outdrove

outdriven

test-drive

test-drove

test-driven

dwell

dwelt

dwelt

Yes

earn

earnt

earnt

Yes

rarely in writing

eat

ate eat

eaten

overeat

overate

overeaten

fall

fell

fallen

befall

befell

befallen

feed

fed

fed

hand-feed

hand-fed

hand-fed

overfeed

overfed

overfed

spoon-feed

spoon-fed

spoon-fed

underfeed

underfed

underfed

feel

felt

felt

fight

fought

fought

outfight

outfought

outfought

find

found

found

fit

fit

fit

Yes

refit

refit

refit

Yes

retrofit

retrofit

retrofit

Yes

flee

fled

fled

fling

flung

flung

fly

flew

flown

outfly

outflew

outflown

test-fly

test-flew

test-flown

forsake

forsook

forsaken

freeze

froze

frozen

deepfreeze

deepfroze

deepfrozen

Yes

quick-freeze

quick-froze

quick-frozen

unfreeze

unfroze

unfrozen

get

got

got gotten

beget

begot

begotten

forget

forgot

forgot forgotten

give

gave give

given

Yes

forgive

forgave

forgiven

misgive

misgave

misgiven

go

went gone

gone went

Yes

forgo

forwent

forgone

undergo

underwent

undergone

grind

ground

ground

regrind

reground

reground

grow

grew

grown

Yes

regrow

regrew

regrown

outgrow

outgrew

outgrown

hang

hung

hung

Yes

The regular verb hang is normatively taken to be a hyponym of execute, though colloquially the irregular form is used here as well.

overhang

overhung

overhung

unhang

unhung

unhung

rehang

rehung

rehung

have

had

had

hear

heard

heard

Yes

mishear

misheard

misheard

overhear

overheard

overheard

rehear

reheard

reheard

heave

hove

hove

Yes

hew

hewed

hewn

Yes

hide

hid

hid hidden

unhide

unhid

unhidden

hit

hit

hit

hold

held

held

behold

beheld

beheld

uphold

upheld

upheld

withhold

withheld

withheld

hurt

hurt

hurt

keep

kept

kept

Yes

ken

kent

kent

Yes

esp. ScE

kneel

knelt

knelt

Yes

knit

knit

knit

Yes

reknit

reknit

reknit

Yes

unknit

unknit

unknit

Yes

know

knew

known

Yes

lead

led

led

mislead

misled

misled

lean

leant

leant

Yes

leap

leapt

leapt

Yes

outleap

outleapt

outleapt

Yes

learn

learnt

learnt

Yes

mislearn

mislearnt

mislearnt

Yes

relearn

relearnt

relearnt

Yes

unlearn

unlearnt

unlearnt

Yes

leave

left

left

lend

lent

lent

let

let

let

sublet

sublet

sublet

lie

lay

lain

underlie

underlay

underlain

light

lit

lit

Yes

moonlight

moonlit

moonlit

Yes

relight

relit

relit

Yes

lose

lost

lost

make

made

made

Yes

premake

premade

premade

remake

remade

remade

unmake

unmade

unmade

mean

meant

meant

meet

met

met

mow

mowed

mown

Yes

plead

pled

pled

Yes

prove

proved

proven

Yes

disprove

disproved

disproven

Yes

put

put

put

input

input

input

Yes

output

output

output

Yes

quit

quit

quit

Yes

read

read

read

misread

misread

misread

proofread

proofread

proofread

reread

reread

reread

sight-read

sight-read

sight-read

rend

rent

rent

rid

rid

rid

Yes

ride

rode

ridden rode

outride

outrode

outridden

override

overrode

overriden

ring

rang rung

rung

Yes

The regular verb is denominal

rise

rose

risen

arise

arose

arisen

run

ran run

run

outrun

outran

outrun

overrun

overran

overrun

rerun

reran

rerun

saw

sawed

sawn

Yes

say

said

said

gainsay

gainsaid

gainsaid

see

saw seen

seen

Yes

oversee

oversaw

overseen

seek

sought

sought

sell

sellt sold

sellt sold

Yes

outsell

outsold

outsold

oversell

oversold

oversold

presell

presold

presold

resell

resold

resold

undersell

undersold

undersold

send

sent

sent

resend

resent

resent

set

set

set

beset

beset

beset

inset

inset

inset

misset

misset

misset

offset

offset

offset

preset

preset

preset

reset

reset

reset

typeset

typeset

typeset

upset

upset

upset

sew

sewed

sewn

Yes

oversew

oversewed

oversewn

Yes

resew

resewed

resewn

Yes

unsew

unsewed

unsewn

Yes

shake

shook

shaken shook

shave

shaved

shaven

Yes

shear

shore

shorn

Yes

shed

shed

shed

shew

shewed

shewn

Now old-fashioned. See show.

shine

shone

shone

Yes

The transitive form is more likely to be regular

outshine

outshone

outshone

shit

shat shit

shat shit

Yes

shoe

shod

shod

Yes

shoot

shot

shot

outshoot

outshot

outshot

overshoot

overshot

overshot

show

showed

shown

Yes

See also shew.

shred

shred

shred

Yes

shrink

shrank shrunk

shrunk

preshrink

preshrank

preshrunk

shrive

shrove

shriven

Yes

shut

shut

shut

sing

sang sung

sung

outsing

outsang

outsung

sink

sank sunk

sunk

sit

sat

sat

outsit

outsat

outsat

resit

resat

resat

slay

slew

slain

Yes

sleep

slept

slept

outsleep

outslept

outslept

oversleep

overslept

overslept

slide

slid

slid

backslide

backslid

backslid backslidden

sling

slung

slung

unsling

unslung

unslung

slink

slunk

slunk

Yes

slit

slit

slit

slitted is found, but is denominal

smell

smelt

smelt

Yes

outsmell

outsmelt

outsmelt

Yes

smite

smote

smitten

sneak

sneak snuck

snuck

Yes

sow

sowed

sown

Yes

speak

spoke

spoken

misspeak

misspoke

misspoken

outspeak

outspoke

outspoken

overspeak

overspoke

overspoken

speed

sped

sped

Yes

outspeed

outsped

outsped

spell

spelt

spelt

Yes

misspell

misspelt

misspelt

Yes

spend

spent

spent

misspend

misspent

misspent

outspend

outspent

outspent

overspend

overspent

overspent

underspend

underspent

underspent

spill

spilt

spilt

Yes

overspill

overspilt

overspilt

Yes

spin

span spun

spun

unspin

unspun

unspun

spit

spat spit

spat spit

spitted is denominal

spoil

spoilt

spoilt

Yes

spread

spread

spread

spring

sprang sprung

sprung

stand

stood

stood

misunderstand

misunderstood

misunderstood

understand

understood

understood

withstand

withstood

withstood

steal

stole

stolen

stick

stuck

stuck

unstick

unstuck

unstuck

sting

stung

stung

stink

stank stunk

stunk

stove, stave

stove

stove

Yes

strew

strewed

strewn

Yes

stride

strode

strid stridden strode

bestride

bestrode

bestrid bestridden bestrode

strike

struck

stricken struck

string

strung

strung

hamstring

hamstrung

hamstrung

restring

restrung

restrung

unstring

unstrung

unstrung

strive

strove

striven

Yes

swear

swore

sworn

forswear

forswore

forsworn

outswear

outswore

outsworn

sweat

sweat

sweat

Yes

sweep

swept

swept

swell

swelled swole

swollen

Yes

swim

swam swum

swam swum

outswim

outswam

outswum

swing

swung

swung

take

took

taken

betake

betook

betaken

mistake

mistook

mistaken

overtake

overtook

overtaken

partake

partook

partaken

retake

retook

retaken

undertake

undertook

undertaken

teach

taught

taught

Yes

misteach

mistaught

mistaught

reteach

retaught

retaught

tear

tore

torn

retear

retore

retorn

tell

tellt told

tellt told

Yes

foretell

foretold

foretold

retell

retold

retold

text

text

text

yes

think

thought

thought

Thunk is sometimes used jocularly as a past participle.

outthink

outthought

outthought

overthink

overthought

overthought

rethink

rethought

rethought

thrive

throve

thriven

Yes

throw

threw

thrown

Yes

outthrow

outthrew

outthrown

overthrow

overthrew

overthrown

thrust

thrust

thrust

tread

tread trod

tread trod trodden

Yes

retread

retread retrod

retread retrodden

wake

woke

woken

Yes

awake

awoke

awoken

Yes

reawake

reawoke

reawaken

rewake

rewoke

rewaken

Yes

wear

wore

worn

rewear

rewore

reworn

weave

wove

woven

Yes

interweave

interwove

interwoven

Yes

reweave

rewove

rewoven

Yes

unweave

unwove

unwoven

Yes

wed

wed

wed

Yes

rewed

rewed

rewed

Yes

weep

wept

wept

wet

wet

wet

Yes

rewet

rewet

rewet

Yes

win

won

won

rewin

rewon

rewon

wind

wound

wound

interwind

interwound

interwound

overwind

overwound

overwound

rewind

rewound

rewound

unwind

unwound

unwound

wring

wrung

wrung

write

wrote

written

handwrite

handwrote

handwritten

miswrite

miswrote

miswritten

outwrite

outwrote

outwritten

overwrite

overwrote

overwritten

rewrite

rewrote

rewritten

typewrite

typewrote

typewritten

underwrite

underwrote

underwritten

(p.91) (p.92) (p.93) (p.94) (p.95) (p.96) (p.97) (p.98) (p.99) (p.100) (p.101) (p.102)