Not-negation revisited: variation between a and any in verb complements in contemporary spoken American English

Abstract

In not-negated English sentences with indefinite expressions following the verb, there is variation between the indefinite article and any as determiners of nouns. The standard view is that singular count nouns take the indefinite article and singular non-count and plural nouns take any. However, it is possible to encounter examples like it isn't any threat, there isn't any lock or I don't have any problem.

The article studies variation between the indefinite article and any as post-verbal determiners of singular nouns in 21,084 not-negated sentences in the spoken component of The Corpus of Contemporary American English, COCA SPOK. The indefinite article is dominant with 90 per cent of the tokens. Variation is extremely rare in sentences with copular be and much more frequent in sentences with existential be and have. Among the reasons for variation between verb types is the use of do-support with have (but not with be). Expressions such as have a job/car/home or there's not a/an with uncontracted not may also prevent the use of any. Variation occurs mostly with abstract nouns such as problem, choice, way, place, reason. This finding is surprising as abstract nouns have rarely been discussed in the literature on varying countability of nouns.

1 Introduction

In negated English sentences with indefinite expressions following the verb, there is variation between negation by means of not and no, not-negation and no-negation (Tottie 1991b; Biber et al. 1999; Wallage 2017, 2020).² Parallel constructed examples are given in (1)–(4):

1. (1) I don't have a car | I have no car.

2. (2) There was not a ship in sight | there was no ship in sight.

3. (3) I didn't see any blood on the floor | I saw no blood on the floor.

4. (4) There weren't any dogs in the garden | there were no dogs in the garden.

No-negation goes back to Old English, where the negative element ne was fused with an to form nan, which has developed into Modern English no. Not-negation derives from ne-a-wiht and became more and more used in Middle English; for a recent detailed study, see Wallage (2017). The determiner no can be used with all categories of nouns: count or non-count (car, blood), singular or plural (ship, dogs). When not-negation is used, the standard view (as expressed by e.g. Svartvik & Sager 1977: 236–7 and Quirk et al. 1985: 256–7) is that singular count nouns such as car or ship take the indefinite article, and that singular non-count nouns like blood and plural nouns like dogs take any, as demonstrated by (1)–(4). My purpose here is not the study of not-negation vs no-negation; my focus will be on the variation between the indefinite article and any as determiners of complement nouns in sentences with not-negation. I will, however, occasionally make comparisons with parallel cases of no-negation.

1.1 Variation between the indefinite article and any

Exceptions to the standard view – article with singular count nouns and any with non-count nouns and plurals – are easy to come across; see e.g. (5)–(7), where any is a determiner of count nouns, and (8), where both a and any are used as determiners of the same noun, threat.

1. (5) Ma drove six hours [to the prison] to pick Dyer up. “She couldn't fly,” Ma said. “She uses a wheelchair and doesn't have any ID.” (San Francisco Chronicle, 23 August 2020)

2. (6) I'm sure [a hungry prowler] would have come here first, because our icebox is on the back porch and there isn't any lock. (Rice 2018: 75–6)

3. (7) Unlike Ms. Huffman, who has released a lengthy, emotional statement expressing shame about her actions … Ms. Loughlin has not made any public statement. (The New York Times, 7 May 2020)

4. (8) The article pointed out that [coyotes] aren't statistically a threat, but this doesn't mean that they aren't any threat. (Letter to the Editor, San Francisco Chronicle, 28 March 2014)

Most modern standard grammars, e.g. Quirk et al. (1985) or Biber et al. (1999), have had little to say about variation between the indefinite article and any. Quirk et al. (1985: 256) advocate using the article, and Biber et al. (1999: 168) state that not any is the ‘correspondence’ of no, and do not explicitly mention not a/an with count nouns. However, Huddleston & Pullum et al. (2002: 381–2) make a few observations on the topic. They point out that the indefinite article a is the determiner of choice for count singulars and that ‘that preference carries over to non-affirmative contexts’, but they acknowledge that there can also be variation between the article and any as determiners of count nouns, henceforth art and any.³

Theoretical and philosophical linguists have also made some observations on this type of variation. Kadmon & Landman (1993: 357) propose that the use of any induces ‘widening, strengthening’ of the meaning of the NP, viz. the addition of ‘additional semantic/pragmatic characteristics’. They claim that ‘any induces widening … whether it carries prominent stress or not’, quoting examples with completely unstressed any (1993: 362), but Rohrbaugh (1997: 311) adduces evidence that that ‘the widening function cannot be dissociated from emphatic focus as marked by phonological stress’.

1.2 Two problems: countability and the use of any

Two important factors must be taken into account before proceeding further, viz. countability and the meaning and use of any. The divide between count and non-count nouns is not clear-cut. The fuzziness of the countability category is well known and acknowledged by major grammars, e.g. Quirk et al. (1985: 245ff.) and Biber et al. (1999: 242ff.). There is also a considerable literature on the subject; see e.g. Drożdż (2020) and Husic (2020), both with extensive bibliographies. Allan (1980: 548) points out that nouns have ‘countability preferences’, depending on their use in noun phrases. In the same spirit, Huddleston & Pullum et al. (2002: 334) assert that ‘[t]he count vs non-count distinction applies to senses or uses of nouns’; they regard nouns as largely polysemous and give a number of examples of variability, adding that ‘when we speak of count nouns and non-count nouns … we are concerned with nouns as used with a count and non-count interpretation respectively’ (2002: 335). Importantly, Huddleston & Pullum et al. also point out (2002: 382) the possibility of different interpretations of abstract nouns as count and non-count (which then leads to variable use of the indefinite article and any):

[Non-affirmative a]ny is found with various singular abstract nouns where the distinction between count and non-count is somewhat blurred: They didn't make any attempt to justify their decision (cf. They didn't make an attempt/much attempt … with count and non-count interpretations respectively).

Another factor determining variation between art and any is the multiple uses of any. Huddleston & Pullum et al. (2002: 392) specify that ‘non-affirmative [i.e. polarity-sensitive] any is usually but by no means always unstressed’ (italics added) and point out that it ‘can be stressed, for example, when it is the focus of negation’. This applies to both typically count and non-count nouns; thus any can be stressed or unstressed with both types.

There are empirically attested examples of any as a determiner – stressed or unstressed – with count nouns in Sahlin (1979: 89). On the basis of the prosodically transcribed London–Lund Corpus of Spoken English (LLC; Svartvik & Quirk 1980), Sahlin provides examples of the ‘indefinite non-assertive article, lacking in stress’ as in (9) and (10) and of ‘a stressed indefinite non-assertive unlimited quantifier’ as in (11) and (12). Stress is indicated by ' (all other markings of prosody, taken from Svartvik & Quirk, can be ignored for the present purpose).

1. (9) Have you made any 'serious at'tempt to \PREPARE yourself 'for it (S.3.1.635)

2. (10) There was never any /\NEED for a re'public of IRELAND (S.2.8.265)

3. (11) … I don't have 'any /\ANSWER to 'that (S.3.6.659)

4. (12) … it would not be right for me to /\GO into 'any :detail at 'this stage (S.11.4.568)

There is also another type that is always stressed, free-choice any, as in Any policeman will be able to tell you or Just any present will make her happy (see also Kadmon & Landman 1993; Rohrbaugh 1997; Horn 2000: 157ff.; Huddleston & Pullum et al. 2002: 381). Free-choice any can also be used in non-affirmative contexts, where it has a fall–rise intonation and is often preceded by just, as in (13), or where old is inserted before the NP head, as in (14):⁴

1. (13) She won't marry just any man. He has to be tall, dark and handsome.

2. (14) He doesn't want any old car. It has to be a Ferrari.

This article is a quantitative study of not-negated sentences with non-affirmative any, often referred to as negative polarity any, stressed and unstressed, in variation with the indefinite article as determiners of nouns. (Free-choice any will be removed from analysis.) Sahlin does provide some quantification but bases it on educated guesses concerning countability, which makes them less reliable.⁵ Tottie (1994) also made intuitive classifications, with the same disadvantages. In order to prevent such problems, I shall take another approach and look for actual cases of variable usage of art and any with different verb types and nouns as described below.

1.3 Material and method

Tottie (1991b, 1994) were based on small corpora (the London–Lund Corpus and the Helsinki Corpus) available at the time and suffered from scarcity of examples, as variation between art and any in negative sentences is a low-frequency phenomenon. It is therefore of interest to return to the topic in the era of mega-corpora. For this study, I have used The Corpus of Contemporary American English, COCA (Davies 2008–), for the search of cases of variation between art and any. I have chosen to concentrate on American English as there is evidence that any is more common in North American English (Childs et al. 2018) than in British English, where any variants have been deemed ‘not possible’ (Hawkins 1978: 188) or only marginally acceptable (Hogg 1977: 142).

I decided to use the COCA section comprising spoken material, COCA SPOK, consisting of recordings from radio and television programmes from 1990 to 2019, a total of 126.1 million words. The present study concerns only the indefinite article and any as determiners of singular nouns in sentences with not-negation, art and any. Cases with zero determiners are not part of this study. The only negators included are full or contracted forms of not, henceforth fnot and n't, occasionally subsumed as neg. Tokens including other negators, such as never, nor, etc., are not included. Searches were restricted to unpremodified nouns.

Pilot studies indicated that the types of verbal constructions that had been found to determine the choice of no-negation and not-negation (Tottie 1991b; Childs 2017; Wallage 2017; Childs et al. 2018) were also determinative for the choice of art and any. The searches were therefore carried out for three of these types of verbal constructions: copular be (becop), existential be (bex) and main verb have. The fourth category distinguished in Tottie (1991b), lexical verbs, is extremely heterogeneous and is not included in the present study. have is used as a lexical verb in American English and will at least to some extent serve to represent lexical verbs.

For manageability, the search was restricted to sequences of finite verb forms in past and present tense forms of becop and bex, and present tense forms with do-support of have, as shown below. I shall use the term sequence in conscious avoidance of construction with its theoretical implications. Examples are is not * fool, there isn't * reason and I don't have * bicycle, where * denotes the site of the variable. Tokens were searched according to the following schemas:

becop    _is|'s|was_ not|n't_ a|an|any_ NSg

bex     There_ is|'s|was_ not|n't_ a|an|any_ NSg

have     _do|does _not|n't_ have_a|an|any_ NSg

1.4 Using COCA: Advantages, problems and initial results

The great advantage of using COCA is it size and searchability, but there are also problems. A major issue in using COCA for the current purpose is that the corpus transcription does not indicate intonation and stress patterns, and that it is not possible to access the recordings on which it is based.⁶ This is a serious limitation, as it is not possible to determine whether any is stressed or unstressed, and thus the matter needs to be discussed on the basis of context. This study must therefore be limited to ascertaining the variability of the nouns used as complements by comparing the number of art and any tokens of each noun and the ratio between the two variants. As high ratios of art and any will mean little if the total number of tokens is low, I will focus on items with the highest number of hits.

Some practical problems also occur because of inherent characteristics of COCA. One issue is that nouns and adjectives can have the same form but different syntactic functions, e.g. official and individual. COCA searches for nouns therefore occasionally yield hits that are adjectives classified as nouns, as in (15), including nouns as part of genitival premodifiers, as in (16). This is particularly a feature of becop sequences, where such hits will include premodified nouns and not the simple nouns sought for this study. (All COCA tokens used as examples include the source and year of the recording.)

1. (15) … the letter is not any official endorsement of the initiative… (CNN03)

2. (16) … it's not any individual’s responsibility (NPR04)

A similar problem is caused by compound nouns, as the search will retrieve not only singulars, but plurals, as in (17). As singular compound nouns were likely to be infrequent, all compound nouns were excluded.

1. (17) Nobody is being checked. There is not any bag checks or anything like that … (CNN15)

Any causes other types of problems. Free-choice any will not be automatically spotted in the initial searches carried out with the present search method, but tokens like (18) or (19) are rare and will be caught in the detailed survey of the most frequent items (see below).

1. (18) This is not any city. It is Jerusalem, which is the holy city. (CBS96)

2. (19) Few places could the theft of student newspapers raise such ire; but this isn't any place, this is Berkeley … Politics are so intense in Berkeley … (CNN02)

As COCA does not distinguish between count and non-count nouns, it will include all singular nouns among the hits, including those that normally take any. I shall give initial overall search results for introductory overviews of each verb type. As the differences between the use of art and any are large, this should suffice for a start. My method will be to first establish the most frequent sequences in each of the verb categories, and then manually vet all tokens and remove problematic cases of the types flagged above. Core non-count nouns (such as money, evidence, progress, news, etc.) will be automatically excluded if they have no art variants. The ensuing analysis will then show the degree of countability in context of the remaining relevant noun complements.

Initial searches produced 21,084 hits, with very different proportions of any: 0.5 per cent in becop sequences, 15 per cent in bex sequences and 26 per cent in have sequences. In what follows, I shall present results concerning sequences with becop in section 2, bex in section 3 and have in section 4. In these sections I will present examples of high-frequency complements with variation between art and any and discuss instances of semantic, pragmatic and grammatical differences. A summary of results and a discussion will follow in section 5 and a conclusion in section 6.

2 Sequences with be copula

becop sequences make up the most numerous type, with 12,444 hits. They provide a good starting point, as distributions are clear and the number of relevant sequences with any is low and manageable. It would be easy to simply dismiss variation here as marginal, but there are some findings worth mentioning.

2.1 Results

The total number of hits included over a hundred tokens of a lot of and a number of non-count nouns as complements. Even with these shortcomings, the result of the initial search shown in table 1 clearly demonstrates that any sequences make up only a fraction of the total – there are only 56 tokens, not even 0.5 per cent of the total.

Table 1. The distribution of any and art as determiners in becop sequences, with proportions of fnot and n't

The presence of fnot co-occurring with any is also a bit of a surprise. Several researchers have found that this co-occurrence is rare or even unacceptable, e.g. Poldauf (1964), Bolinger (1977), Tottie (1991b: 277, 306ff.). Bolinger (1977: 60ff.) proposes a syntactic explanation, viz. that not is not part of the verb phrase in a sentence like there was not any trouble.⁷ Another possibility could be rhythmicity, i.e. the fact that stressed and unstressed syllables tend to alternate so that rhythmic clashes are avoided, the so-called Principle of Rhythmic Alternation first formulated by Sweet (1970 [1887]) and further developed by Schlüter (2005). Adjacent fnot and any would then produce a clash, but more empirical data is necessary to support this hypothesis.

2.2 Variation between art and any

To check actual variation between art and any, the 56 any tokens shown in table 1 were manually checked, and irrelevant items were removed, i.e. misclassified adjectives, compound nouns, instances of any without art variants, and tokens with free-choice any as exemplified in (18) and (19) discussed in section 1.

Table 2. Complements with art and any variants in becop sequences

The remaining 34 any tokens all had art variants, as shown in table 2, with the numbers of any always lower than of art. The table lists the number of occurrences of each noun, with totals of any + art. For easy reference, proportions of any are indicated as percentages when totals are ≥13.

In a few cases there were meaning differences between art and any tokens of the same noun: kind, sort and business. The most frequent any -complement is kind, which occurs with any 13 times, compared with 23 times with art, and there is a subtle difference between the art and any sequences. The art variants are similar to the hedging pragmatic marker kind of (see e.g. Beeching 2016: 172ff.). As a hedge, kind of/kinda usually modifies adjectives or verbs, as in He's kind of nice and I kind of like him. This hedge use is less frequent with nouns, but that seems to be the function in (20) and a few other examples. The art variant in (20) can be seen as the negation of the affirmative sentence “Mein Kampf” is a kind of garden variety of anti-Semitism, but this is not the case with the any variant in (21). The any examples seem more emphatic and often appear in argumentative or legal contexts, and any may have been stressed in (21). While the hedge a kind of downplays the importance of the referent, any kind of specifies the uniqueness of the referent. The synonym of kind, sort, exhibits the same meaning difference between art and any variants. Example (22) is mildly jocular but (23) is strongly argumentative. Like many other nouns, business has different meanings when used as a count and a non-count noun: ‘commercial company’ in (24) and ‘a matter that one has the right to meddle with’ in (25).

1. (20) “Mein Kampf” … is not a kind of garden variety of anti-Semitism … (NPR 00)

2. (21) … this nutty, bigoted pastor is not any kind of spokesman for America. (NPR12)

3. (22) You know, he's not a sort of raving leftist radical or anything. (NPR03)

4. (23) [on nuclear situation after tsunami] We are sure we are fine. We are sure this is contained … We are positive this is not any sort of Chernobyl. (Fox11)

5. (24) It is not a business, though. I mean it is not like Ford – you know, some company … (CNN15)

6. (25) … it is not a criminal offense and is not any business of ours to investigate. (ABC90)

It is difficult to find any semantic difference between the art and any tokens with way. Both (26) and (27) convey emphasis, and not is stressed in both variants:

1. (26) … it's not helping. So that is not a way to rehabilitate. That's not a way to pay debt. It's not a way to allow people to transform their lives. (CNN18)

2. (27) … it's a fluke that it happens, and it's not any way that we can depend on to make something that's reliable. (NPR01)

A few items with low frequencies of any are worth mentioning. Thus (28) with art before part is a simple statement, whereas (29) appears to be more argumentative, possibly with stressed any. Accident occurs in 99 art tokens, as in (30). The two any tokens, (31) and (32), occur in statements functioning as questions, which may have influenced the choice of any. See also (58).⁸ Note that any can hardly have been stressed in (31) and (32).

1. (28) For example, the “Washington Post” is not a part of Amazon, the company. (CNN18)

2. (29) Forget what you heard. He's not any part of this. (NBC04)

3. (30) … the fire was deliberately set by a human being. It wasn't an accident. (NBC08)

4. (31) This was a murder? This wasn't any accident? (NBC08)

5. (32) So it wasn't any accident on your part, any --- (CNN03)

Joke is a strongly count noun that occurs 38 times in COCA SPOK in the sequence It is not|n't a joke. In (33) any is likely to have been primed by the preceding core non-count fun with any:

1. (33) … prison isn't any fun and it isn't any joke. (ABC18)

Example (34) is interesting, as it contrasts with (18), repeated here for convenience, with stressed free-choice any. In (34) the speaker characterizes her hometown in Southern California and any must have been unstressed.

1. (18) This is not any city. It is Jerusalem, which is the holy city. (CBS96)

2. (34) It's a big, sprawling, nameless grid of mini-malls and … Burger Kings and auto shops. It's not cool. It's not downtown. It's not uptown. It's not any town. It's sand. (CBS99)

3 Existential sequences

Only sequences where there immediately precedes be and where neg directly precedes art/any were included, so tokens like … there was not much chance for coverage, not any chance for questions? (CNN01) were not retrieved.

3.1 Results

The number of hits with bex sequences was considerably lower than for becop, 2,049, but the proportion of any tokens was much higher, 314 or 15 per cent in the initial count, as shown in table 3. The table also shows which form of neg was used, fnot or contracted n't, and with which form of the verb be it occurred, 's, is or was. As in the case of becop sequences, these interconnected factors turned out to be important for the choice of variant.

Table 3. Distribution of any and art as determiners in bex sequences, with proportions of fnot and n't

Among the reasons for the high rate of art, the most important one is certainly simply the fact that (strongly) countable nouns are a majority in English (Biber et al. 1999: 242). Note that, again, few any tokens have full form not, only 60, or 19 per cent of the total number of any sequences, whereas fnot is prevalent with art sequences, 65 per cent.

Some other factors are worth pointing out. As with becop sequences, full-form not is most frequent after the contracted form 's, where it is obligatory after 's in there's not. According to Rupp & Britain (2019: 25), there's has undergone grammaticalization and can be regarded as a single presentational morpheme that can also be used with plural nominal subjects. My data suggest that grammaticalization may now have extended to the whole bundle there's not as a negative existential quantifier; see (35). Furthermore, not a/an is a routinized collocation with the meaning ‘not a single, not even one’ often used with minimizers, such as shred or scintilla, as in (36). Another reason for the dominance of art in bex sequences is the enormous number of tokens of a lot (of), over 400, as in (37).

1. (35) … the border is secure. There's not a problem right now. (Fox17)

2. (36) … there's not a shred of physical evidence linking him to the crime… (ABC13)

3. (37) … when you suffocate someone, there's not a lot of blood. (CNN19)

3.2 Variation between art and any

For the survey of actual variation between art and any, the inventory of the most frequent any tokens was cleared of tokens without art equivalents, compound nouns, misclassified adjectives and tokens of free-choice any. Table 4 lists first the top seven any complements in descending order of frequency. There are no others with five or more tokens – the remainder all have ≤ 3 tokens and mostly occur only once or twice. art variants are listed to the right of any, followed by totals of an +art. Percentages of any are included for quick reference.

Table 4. Complements with art and any variants in bex sequences

Three of the most frequent any complements have higher frequencies of any than of art: question, way and reason; doubt has almost equal numbers. Two additional items have higher proportions of any than of art: kind and hope, but totals are low. The remaining items have low proportions of any, and word and chance were only included because the any/art pairs merit some comment.

A couple of complements show semantic/pragmatic differences between the variants, viz. question and kind. In bex sequences art tokens of question often refer to the speech act and can be glossed ‘sentence worded or expressed so as to elicit information, query’ (OED s.v. question, n., II3a) as in (38). The meaning of question can also be ‘a matter (of concern)’, as in (39) or ‘doubt’, usually with any as in (40).

1. (38) … there was not a question to either Romney nor Obama about gun control. (Fox13)

2. (39) … [this expression] is one of the great distracting phrases, because there's not a question of blame. (PBS96)

3. (40) Look, we need different leadership. There isn't any question about it. (ABC18)

As in becop sequences, both art and any are used with kind, and the same meaning difference appears with bex: art tokens function in a way similar to pragmatic particles. The relative clause in (41) indicates that the sentence is a negation of a positive statement like there is a kind of respect for the profession. ⁹ Example (42) on the other hand does not express the negation of there was a kind of program but a denial of the totality of possibilities, and any may have been stressed.

1. (41) … there isn't a kind of respect for the profession that I think some of us would like … (NPR94)

2. (42) … I started a battered women's group because there -- there wasn't any kind of program … (Ind03)

Other pairs show collocational differences, e.g. doubt and reason. All art tokens of doubt had the routinized bundle not a doubt in my/our mind, as in (43). Of the 15 tokens with any, only one had doubt in my mind. In most cases, doubt was followed by about it, about that or a modifying clause as in (44). Reason had a majority of any tokens, most of them with n't plus a to-infinitive, and there are only three tokens of art, all with for +ing; see (45) and (46).

1. (43) There's not a doubt in my mind that he did not kill my mother. It was an accident … (CBS13)

2. (44) … there isn't any doubt that China is doing a lot of saber-rattling at this point … (ABC95)

3. (45) … I believe now … that there's not a reason for having an abortion. (NPR92)

4. (46) … it is one of the … mysteries of Watergate. There wasn't any reason to do it. (ABC92)

Way, place, sense and problem show great similarity in meaning and collocations between any and art sequences, as shown by (47)–(54):

1. (47) So the difficulty was, there wasn't a way to get that in before the jury. (CNN11)

2. (48) The problem was … we did too good. There wasn't any way to pick up the oil. (NPR10)

3. (49) … these beautiful kids … were made to feel like there wasn't a place for them … (Fox08)

4. (50) … there wasn't any place I could go that they didn't know Bob Hope … (CNN96)

5. (51) … people are taking this seriously, but there's not a sense of alarm. (CNN00)

6. (52) I felt right at home. There wasn't any sense of danger or foreboding evil. (CBS94)

7. (53) … we were able to establish that there wasn't a problem. (NBC14)

8. (54) We got through in a couple minutes so there wasn't any problem. (CNN13)

Chance and word both have nonce tokens of any, compared with 18 with art. Not a chance and not a word are entrenched bundles in COCA SPOK, with 89 and 126 tokens, respectively, often used without a verb phrase. Examples (55) and (57) are typical; as the NPR programme is available online we can know that (56) definitely had unstressed any.¹⁰ There is not a single instance of there isn't a/any word with n't in COCA SPOK, and (57) is typical. Example (58) with any, wasn't any word, is therefore a surprising example that shows that even the most entrenched bundles can have variation between art and any. Note that the interviewer is a reporter born and bred in the US. It is a statement functioning as a question, which may account for the use of any; see (31) and (32) above.

1. (55) There's not a chance you're going to see John Edwards on the stand. (CNN12)

2. (56) … you know, we have to make Obamacare work. There isn't any chance to blow it up if we don't come up with our own system. (NPR17)

3. (57) There's not a word in the federal Constitution about marriage. (Fox15)

4. (58) [An interviewer claims that President Reagan had not had brain surgery]: He didn't have any stitches. RON-REAGAN-JUNIOR- Yes, he did. Yeah, he sure did. [Interviewer …]: There wasn't any word … it just wasn't made public? (ABC11)

4 Sequences with have

The verb have is used as a regular lexical verb in contemporary American English, and anomalous finite forms now mostly occur in fixed collocations like I haven't the faintest (idea) (see Biber et al. 1999: 160ff.). The forms investigated for any and art with have are do-negated sequences in the present tense. Note that because of do-support, the sequence have art NSg is identical in affirmative and negative sentences.

4.1 Results

The initial search of COCA for tokens of have with either art or any and with either fnot or n't produced the distribution of complements shown in table 5, with nearly 6,600 hits. The numbers in the table include 511 tokens with a lot of complements among the art results, and many core non-count nouns among the any results – the top items are money (131), evidence (74) and information (58). Both art and any instances include misclassifications of adjectives as nouns and compounds like bomb damage assessments, motor functions and lab values.

Table 5. Distribution of any and art as determiners in have sequences, with proportions of fnot and n't

Notwithstanding the problems cited above, table 5 provides a good idea of the distribution of art and any. As with becop and bex sequences, the number of any tokens was much lower than that of art tokens, 1,726 vs 4,865, but the proportion of any was higher, 26 per cent. Full not tokens are much less frequent than with be sequences, only 6 per cent with any and 8 per cent with art. As all tokens have do-support, the negator and the complement are not adjacent, and the negator type cannot determine the selection of any or art.

4.2 Variation between art and any

Table 6 lists the 25 top-ranking variable complements after the removal of irrelevant hits. The large number of any tokens with have justifies a somewhat different approach to presentation here than that given for becop and bex sequences. Table 6 thus lists complements with any frequencies over 50 per cent at the top and those with art frequencies at the bottom for a better overview. The table also lists the complements in the order of proportions, rather than numbers, of any and art in columns IV and V. Moreover, there is an overview of the frequencies of corresponding affirmative tokens of have_art_NSg in COCA SPOK for comparison in column VI.¹¹ This has been introduced to test the claim that ‘affirmative art carries over to’ negative sentences (Huddleston & Pullum et al. 2002: 381). Complements discussed in the text are printed in bold. For reasons of space, I will limit most of my discussion to items with high numbers of occurrence and robust proportions of art and any.

Table 6. Complements with any and art variants in have sequences

Note that, as with be sequences, all the complements are abstract nouns. This is in itself an interesting fact, as abstract nouns have been given little attention in research on countability; cf. Husic (2020), Drożdż (2020). However, the variability between count and non-count senses has been pointed out by Huddleston & Pullum et al. (2002: 382) as quoted in section 1.2. above. Moreover, a quick check of the entire COCA corpus shows that the majority of top-ranking complements in both affirmative and negative sentences are abstract nouns.

A comparison between columns V and VI demonstrates that most complements with high proportions of art in sequences with not-negation do indeed have high numbers of occurrence of art in affirmative sentences, and that those with high proportions of any mostly have low numbers of occurrence in affirmative sentences. This overall correspondence thus supports the claim that art in affirmative sequences is carried over to not-negated sentences. However, a closer look at individual sequences shows that there are many exceptions to this principle, and that there are a number of other factors that govern the choice of determiner. First of all, where there is a semantic difference between art and any tokens, the choice of variant depends on the speaker's intentions. Four any/art pairs show the same semantic/pragmatic differences between variants as found for be sequences above: kind, question, business and sense.

Kind has a solid majority of any, with 97 per cent, or possibly even 100 per cent, and kind of is not functioning as a hedge in e.g. (59). The one token with art, (60), is cryptic. It is truncated and ends with cool, which may be either a noun or an adjective. The speaker is a choreographer commenting on a video, and a kind of here may function as a hedge. Question shows the same meaning differences between art and any tokens as found with bex sequences: art tokens usually have the meaning ‘query’, whereas any tokens signify ‘doubt’; cf. (61) and (62). There can also be semantic differences between have a sense and have any sense, but postmodified (63) and (64) appear synonymous. Non-postmodified have any sense is used to mean ‘have your wits about you’ in (65). For business, see section 2, examples (24) and (25).

1. (59) Pete Buttigieg … Indiana, doesn't have any kind of organization. (Fox19)

2. (60) Mr-FELD: It doesn't have a kind of cool … (CBS99)

3. (61) I don't have a question. I just wanted to thank you … (NPR08)

4. (62) I don't have any question that it's in the American tradition [to help] … (NPR05)

5. (63) I don't have a sense of reverence for “Gone With The Wind”. (NPR19)

6. (64) … these people don't have any sense of what the money is being used for … (PBS12)

7. (65) … they're thinking I don't have any brains anymore … I just don't have any sense or don't know what I'm doing. (CBS 92)

Some sequences with a high number of affirmative art sequences still show substantial variation between art and any in negative sentences. Thus way has almost a fifty-fifty distribution: the proportion of any is 45 per cent. And although affirmative sequences with have a problem and have a choice are frequent, negative art sequences reach only 74 per cent, and art and any variants are often used with no perceptible difference of meaning; see (66)–(71):

1. (66) … I don't have a way to print braille easily. (PBS15)

2. (67) We don't have any way right now to treat Alzheimer's. (ABC11)

3. (68) And I don't have a problem with background checks. (NPR13)

4. (69) They don't have any problem with assaulting people. (Fox11)

5. (70) … you've already been told what to eat. You don't have a choice… (NPR13)

6. (71) … the argument is, well, we're weak, so we don't have any choice. (PBS15)

Sequences with high proportions of any can usually be linked to low frequencies of affirmative sequences with art, e.g. those with doubt, desire and reason. Have * doubt thus has 36/38 any tokens, or 94 per cent. The two tokens with art contain a doubt in my mind and could be instances of the type that signifies ‘not a single’; see (36), (43), (55), (57) and (58) above. There are only three tokens of affirmative have a doubt in COCA SPOK, but it is interesting that the plural negative sequence have any doubts occurs 57 times. This may have contributed to the use of any in (73) and others. As shown above in (62) speakers often prefer to express having a doubt by using any question.

1. (72) I don't have a doubt in my mind that he’s guilty … (CNN95)

2. (73) I mean, I don't have any doubt about that. (CNN19)

Desire is also infrequent in affirmative sentences – there are only 16 tokens in COCA SPOK. But note that the object of desire in affirmatives with art is often a specified one, as shown in (74), and in negative sentences it is not necessarily specific, as shown by (75). Have art reason is slightly more frequent in affirmative occurrences, with 60 tokens, but there is still a majority of any with negation, 38/45 (84 per cent). There may be a couple of different explanations for this. As with desire there is usually a specific reason in affirmative sequences, understood or expressed, as in (76), but in negated sentences there can be more than one reason – no motive, no girlfriend and loving one's wife, as in (77).

1. (74) … they have a desire to see their children grow up in peace. (CNN01)

2. (75) … I haven't ever run for anything. I don't have any desire to run for anything. (Fox05)

3. (76) … they've made a really, really good iPad, so now I have a reason to upgrade. (CNN16)

4. (77) … you know, I don't have any reason to kill my wife. I don't have any motive. I don't have a girlfriend. [My wife and I] love each other very much. (CBS13)

Two sequences are exceptional: those with idea and clue. Although have * clue is more colloquial than have * idea, the two sequences are synonymous, but they have opposite determiner patterns.¹² Both sequences also run counter to the correspondence of a high art frequency in affirmative sequences to a high art frequency in not-negated sequences, and vice versa. Idea has a high frequency of art in affirmative sequences, 188, but it is still one of the top any-dominant complements with 127/140 (91 per cent) any tokens. Have * clue has a low frequency of art in affirmative sequences (only 26 tokens) but still has a high proportion of art with 108/117 (92 per cent) and only 8 per cent of any in negative sentences; see (78)–(81):

1. (78) … most people don't have an idea of what Afghanistan is like… (PBS12)

2. (79) … officials say they don't have any idea on the whereabouts of bin Laden. (Fox02)

3. (80) Even his employees don't have a clue. (ABC10)

4. (81) … we don't have any clue about her origins. (CNN13)

In this case it is appropriate to seek a historical explanation of the different patterns. Idea has been used in English with the meaning ‘an item of knowledge or belief; a thought, a theory’ at least since the seventeenth century (OED s.v. idea, n. III, 12a.), and its use in negative sentences has always been dominant; see (82), which has no-negation, the historically older form.¹³ The original meaning of clue was ‘thread, ball of yarn’ and the meaning ‘key [to a solution]’ is also recorded in the seventeenth century, but the use with negation is first recorded in the twentieth century, with not-negation, the more recent type (OED s.v. clue, n. 2b, 2e.); see (83). The preference of art over any still prevails with clue, as shown in table 6, but interestingly, there are also 147 instances of have/has no clue in COCA SPOK.¹⁴

1. (82) … my feet were all blisters. You have no idea how they smarted. (Burney, Cecilia 1782 (OED))

2. (83) That doesn't bring us any nearer to finding out … We haven't a clue to it. (Mason, House of Arrow 1924 (OED))

4.3 Resisting any – household words

Although it is clear that there are many exceptions to the observation that art carries over from affirmative to negative sentences, there is one type of have sequences that shows almost total resistance to any. Thus 13 out of the 30 top-frequency art tokens have no any variants at all, viz. job, gun, home, strategy, name, license, father, date, vote, house, dog, car and life. The most striking case of complete lack of variation is job, with a total of 126 art tokens like (84) and not a single token with any.¹⁵ Examples (85)–(87) are other typical examples of high-frequency instances without any counterparts in COCA SPOK.

1. (84) She doesn't have a job, can't pay her medical bills, and now she's been sued. (NPR98)

2. (85) … people who fear they can't protect their child if they don't have a gun. (CBS18)

3. (86) And what's his name? Dr. BROWN: Oh, he doesn't have a name yet. (CBS96)

4. (87) They don't have a license to drive, but that doesn't stop them … (CBS93)

What all these complements have in common is that they denote something that we usually have just one of, and that the sequences are frequent collocations or bundles in non-negated contexts. Have a gun occurs over 200 times in COCA SPOK without negation, have a home over 100 times, have a license over 50 times, have a name about 100 times.

Going beyond the nouns listed above, it is possible to find many more any-resistant lexical items. Two types stand out: words denoting family (but not the word family; see (88) and (89) below) and words denoting everyday possessions. Thus there are no any variants with any of the following items, listed according to the frequency with art in COCA SPOK:

≥10 art sequences: mother, husband, baby, child, boyfriend, girlfriend; phone, camera

≥ 4 art sequences: kid, daughter, son, mom, parent, nanny; radio, basement

Example (77), repeated from above, is interesting in that it demonstrates the entrenchment of have a girlfriend, following two tokens with any:

1. (77) He spontaneously said, you know, I don't have any reason to kill my wife. I don't have any motive. I don't have a girlfriend. (CBS13)

The sequence have * family shows meaning differences between art and any variants. With art family is normally used to mean ‘nuclear family’ consisting of parents and children, as in (88) where the meaning is clearly ‘hasn't started a family’. any family refers to ‘extended family’ or ‘relatives’, as is clear from (89):

1. (88) He's a young man … He has lost his job. His -- he doesn't have a family yet. (CNN01)

2. (89) [After flooding she] and her husband … live with their two sons … in a trailer … [and she says]: We're going to probably go stay in a motel. We don't have any family … (NPR93)

Seven sequences have only one any token: budget, case, government, record, relationship, vote, case and system. Have a budget occurs 115 times in COCA SPOK, have a record 182 times, have a relationship 295 times and have a system 251 times; clearly all stable sequences. But note that in (90), the preposition is to and the complement is fact, whereas relationship otherwise mostly refers to relationships (personal, sometimes sexual) with individuals:

1. (90) leon panetta: what you heard here in the previous interview doesn't have any relationship to fact. (ABC94)

5 Summary and discussion

COCA SPOK yielded over 21,000 hits with not-negation and either any or art as determiners of indefinite complements. The largest number was 12,444 be copula sequences, followed by 6,591 have sequences and 2,049 with existential be. art was totally dominant overall with 90 per cent of all tokens and only 10 per cent any, but the distribution differed between the verb types: becop sequences had the lowest proportion of any, 0.5 per cent, followed by 15 per cent for bex and 26 per cent for have sequences.

The distribution of fnot and contracted n't also differed between the three types of sequences. The frequency of fnot was high in both becop and bex sequences with art (83 and 65 per cent, respectively), but low in art sequences with have, only 8 per cent. This difference is related to the use of do-support with have but not with becop and bex sequences, where neg and the determiner are adjacent. Especially the collocation 's not makes for a high incidence of art, 38 per cent. Most any tokens (90 per cent) had n't, corroborating earlier observations concerning the avoidance of adjacency of any and not (Poldauf 1964; Bolinger 1977: 60ff.; Tottie 1991b: 277, 1994).

Variation between the determiners was defined as the occurrence of both any and art with the same complement and the same verb. After establishing the most frequent tokens with any, the inventory was manually cleared of non-count nouns by eliminating tokens that had no variants with art, free-choice any tokens, compound nouns and misclassified complements. This made it possible to establish actual variation, which was low in becop sequences, higher in bex sequences and highest in have sequences, mirroring overall results.

Most complements were abstract nouns. Only seven of them occurred with all three verb types, as shown in table 7. The table shows those nouns plus question, which only occurred with bex and have. The complement nouns are listed according to their number of tokens in the have column, which has the largest number of variable tokens; note that some percentages in the other columns are based on low numbers.

Table 7. Complements with any/art variation in all three sequences: have, bex, becop

Only four lexical items appear in all three verb categories, viz. sense, way, kind and doubt. Way and kind are the only ones with high numbers of occurrence in becop sequences. bex shows variation with the same lexical items as have sequences, but proportions are different. In a few cases there are semantic or pragmatic differences between art and any variants, most of them occurring with question or kind. Question ‘query’ usually has art and question ‘doubt’ any. A kind of sometimes functions as a hedge but any kind of does not.

The dominance of art can be explained by the fact that affirmative sentences are more frequent than negative sentences, and affirmative sequences with art are deeply entrenched in the language. art then ‘carries over to non-affirmative contexts’ according to Huddleston & Pullum et al. (2002: 381). This correspondence is demonstrated most clearly in section 4, where it is shown that overall, not-negated art sequences usually are related to high frequencies of art in affirmative sequences. The use of art is supported by other factors such as the already mentioned use of fnot, but there are many collocational and other lexico-grammatical factors that lead to a preference for any. However, there is a group of sequences where the complements refer to everyday items, ‘household words’, that appear to be resistant to the use of any. Thus have a job/car/dog/home/father/name/license do not have variants with any in COCA SPOK. According to Huddleston & Pullum et al. (2002: 382), a ‘special context’ is necessary for the use of any here:

I would normally say, for example, I haven't got a car rather than I haven't got any car. The latter cannot be ruled out, but it needs some special context, as when I make an emphatic riposte to someone who thinks I do have a car.¹⁶

Special context is a loose label for discourse factors, referring here to old or shared information. Wallage (2017: 116ff.) uses the term pragmatic activation to refer to the effect of overtly stated circumstances or shared knowledge that leads to the use of particular linguistic forms. Previous mention is a case in point in (91), where any raccoon refers to an animal that has been mentioned in the previous context:

1. (91) … McCormick … said the old farmer enticed him into the barn, supposedly to get rid of a raccoon … “he wanted me to poke the ’coon out.” McCormick says he was suspicious because he knew there wasn't any raccoon … (CBS94)

In (91) any is used in an emphatic denial of the alleged presence of a raccoon in the barn. It is likely that wasn't was stressed and any unstressed. Pragmatic activation can also explain the use of any in examples (5)–(8), repeated here for convenience.

1. (5) Ma drove six hours … to pick Dyer up. “She couldn't fly,” Ma said. “She uses a wheelchair and doesn't have any ID.”

2. (6) I'm sure [a hungry prowler] would have come here first, because our icebox is on the back porch and there isn't any lock.

3. (7) Unlike Ms. Huffman, who has released a lengthy, emotional statement expressing shame about her actions … Ms. Loughlin has not made any public statement.

4. (8) The article pointed out that [coyotes] aren't statistically a threat, but this doesn't mean that they aren't any threat.

Shared knowledge of flight rules, thieving and outdoor ice boxes will explain the use of any in (5) and (6), and previous mention of statement and threat its use in (7) and (8). Discourse factors have not been considered in this study, which is based on short text extracts. Further work based on longer extracts will be necessary to substantiate the importance of context.

Example (91) is an unusual example in several ways. Raccoon is one of a small number of countable nouns denoting concrete entities in the sequence there wasn't any NSg in COCA SPOK. The fact that the majority of all variable complements in my material are abstract nouns appears not to be a consequence of the nature of COCA SPOK, even though the topics are mostly politics and crime, and mention of concrete items is not often called for. A search of the entire COCA for both affirmative bex and have sequences shows that abstract nouns are the most frequent complements in not-negated sentences; thus the most common concrete nouns, food, man and woman, do not even make it to the top-twenty list. The versatility of abstract nouns is noted by Huddleston & Pullum et al. (2002: 334, 382) but has been given little attention by countability scholars. As pointed out by Husic (2020), most writers have discussed only fluctuation in countability between mass and count nouns, but she shows that abstract nouns can also have countability preferences. The fact that actual variation between art and (unstressed) any takes place in unstressed complements with abstract nouns should therefore not be a surprise: they have both count and non-count interpretations and should be able to take unstressed any like any mass nouns or core non-count nouns in sequences like there isn't any food/wine/meat/news or we don't have any wine/money/evidence.

6 Conclusion

The question remains why unstressed any is used rather than art with core count nouns such as raccoon or lock, and why abstract nouns take on their count or non-count guise. It is clear that if we wish to achieve a more definitive account of the variation between the indefinite article and any as determiners of indefinite noun complements of verbs we need both large corpora with prosodic transcription of stress and long enough text extracts for discourse study. Corpora of British and American English as well as other Englishes will be required as there are indications that there are differences between varieties. More work also needs to be done on the use of any in questions, a type that has attracted little attention by scholars. What still needs much more research is the variation in contemporary English between not-negation and no-negation, an immense area waiting for intrepid scholars, theoreticians as well as empiricists.