There is a certain malaise in the polarity literature concerning the compatibility between factive predicates and Polarity Sensitive Items, with seemingly contradictory proposals being found even in studies published by the very same author. Thus Giannakidou (2009: 1889–1890) affirms that “polarity items are excluded from veridical sentences” and that “affective verbs are veridical and indeed strongly: p is true also in the speaker’s model (factive complements are presupposed to be true).” Giannakidou (2006: 595), however, had argued that any can be “rescued” after an affective predicate if the latter makes a non-veridical inference available in the global context of the sentence with which the polarity item can be associated, as is the case in (1):
Larry regrets that he said anything. → Larry would prefer it if he had not said anything.
In contrast to negative emotive predicates like regret, “a positive emotive verb (…) is not affective and does not admit PIs” (Giannakidou 2006: 577), as shown in (2):
*Larry is glad that he said anything.
Giannakidou observes in addition that “factivity in general is not a sufficient condition for PIs: factive verbs that are not emotive, such as know, do not allow any” (Giannakidou 2006: 577).
*John knows that Bill said anything.
Giannakidou (2006: 586) goes on however to mention the fact that PIs such as any can occasionally be found with positive emotive factives, as illustrated by (4):
Bill is glad that we got any tickets at all.
This is said to occur only if “context inferencing makes salient somehow a quasi-negative proposition,” in this case the implication that it was almost impossible to get tickets and so Bill had not expected that he would be able to purchase any at all.
For a paradigmatic Polarity Sensitive Item (PSI) such as any to be licensed by contextual inferences is not a straightforward matter, as not all negative inferences validate any in factive contexts (for other PSIs, see Crnič 2011: 67–70). Atlas (1996) points out for instance that if PSIs could be licensed by negative inferences indiscriminately, the negative implication associated with the determiner some, to the effect that not everyone is covered by the predication, should make any felicitous in a context such as (5) below:
Some students have submitted an essay.
However, this turns out not to be the case:
*Some students have submitted any essay.
Likewise, the negative inferences associated with focus environments such as those given in (7) below should also make a PSI acceptable:
- It was John who talked to someone, nobody else.
- Johntalked to someone, not me.
Nonetheless they do not, as Giannakidou (2006: 582) herself notes:
- *It was John who talked to anybody.
- *Johntalked to anybody.
Horn (2016) argues that this is due to the fact that the negative inference is not at-issue in focus constructions, what is under discussion being rather “the assignment of the value (x=John) for a given variable (λxFx).”
Another illustration of the complex relations between polarity items and negative inferences is the distinct behavior of the otherwise synonymous expressions almost and barely. As discussed by Horn (1996), although both of these adverbs can be paraphrased by ‘not quite’, only barely licenses any:
- He has barely finished any of the essays.
- *He has almost finished any of the essays.
According to Horn (2016), this is due to fact that the at-issue meaning profiled by almost is a positive one, whereas the at-issue content conveyed by barely is negative: He has almost finished the essays is an argument in favor of considering the essays practically completed, even though in fact they are not, while (9a) is an objection to considering the essays as validly completed, although of course they are in fact finished.
These claims raise certain empirical questions that we wish to address in this paper. The first of these is whether there really is a difference between negative emotive factives such as regret and positive ones such as be glad with regard to their compatibility with a Polarity Sensitive Item such as any. The second empirical point is whether any really is incompatible with non-emotive epistemic factives such as be aware and realize, as Giannakidou (2006: 577, 595) claims. These empirical issues have led us to undertake a systematic large-scale corpus investigation of PSI any licensed by factive predicates – to our knowledge the first such investigation to be carried out. Given the considerable theoretical energy expended on the subject, one would have thought the data on the subject to be well-established. There is however significant variability in the judgements on any’s actual usage in factive contexts in studies based on author-produced examples. For instance, Kadmon and Landman (1993) give I’m sorry I said anything as felicitous and contrast it with the allegedly anomalous ?I’m glad I said anything, while a comparable sentence is deemed acceptable by Giannakidou in (4) above. Examining the occurrences of such sequences in a corpus allows one to establish the usage profile of the item under study, to pin down the exact contribution of any and of its environment to the message conveyed, and to determine the relative frequency of the various types of contexts. The theoretical issues can then be addressed concerning the purported distinction between the “licensing” and “rescuing” of NPIs proposed by Giannakidou (2006: 592, 2011: 1684–1687) and claimed to be supported by experimental results from Xiang et al. (2013). Is it really conceivable that an NPI can appear “in the syntactic scope of an expression that does not have the semantic potential to license it” (Giannakidou 2011: 1687)? And if so, how is one to characterize the environments in which this can occur? We will be arguing that this may be achieved by the type of proposal put forward by Horn (2016), according to which any in a veridical environment is licensed by an at-issue negative meaning. However, the adequacy of this account will be shown to benefit from a better knowledge of the actual behavior of the phenomenon in real usage.
The article is organized in three parts. First we present the usage profile of any in the scope of negative (e.g. regret) and positive (e.g. be glad) emotive factive expressions. This is compared with a newly-identified category of emotive factives that do not prejudge the positive or negative character of the fact-evoked sentiment, but relate instead to the feeling caused by a contradiction of expectations (e.g. surprised, amazed, etc.). After this, the possibility of any occurring with epistemic factives is examined. A final assessment is offered that focusses on the conditions of distribution imposed by the indiscriminate-coverage lexical value of any (cf. Duffley and Larrivée 2015).
2 Negative versus positive emotive factives
This section is concerned with the distribution of any under negative and positive emotive factives. We first describe the method used for identifying these contexts as well as those with epistemic factives, which are explored in the following section. The relevant factive expressions are taken from the exhaustive list provided by Wyse (2009) 1. The corpora used for the study were the 4.2-billion-word News on the Web (henceforth NOW) corpus, the 450-million-word Corpus of Contemporary American (henceforth COCA) and the 100-million-word British National Corpus (henceforth BNC). Searches were performed for the factive predicates from Wyse’s list followed by any within the maximum range available on the Brigham Young University website of nine words to the right of the polarity item 2, since relevant sequences of any were not found in searches to the left of the factive licenser. The occurrences were then examined individually to retain only those in which any is licensed by the factive expression, exclusive of other potential licensers such as negation, modality and genericity. 3 If the factive licenses the Polarity Sensitive Item, the sequence containing the PSI should be infelicitous outside of the commanding factive. Thus in the following sequences:
- … natives who have trod the boards of New York, were surprised to discover that any number of important new plays have fallen through the cracks …
- Awarethat any serious tightening of short-term interest rates added billions to federal debt-service costs, even the feistiest central bank might hesitate to tighten the monetary screw when inflation threatened,
since the subordinate clauses can stand on their own due to their generic dimension:
- Any number of important new plays have fallen through the cracks.
- Any serious tightening of short-term interest rates added billions to federal debt-service costs,
it cannot be said that any is licensed by the factive here and such sequences were therefore left aside. Another criterion of exclusion was whether an expression classified as belonging to the factive category actually conveyed a factive value in a specific attestation. If a given sequence can be followed by the negation of the suspected factive clause, then that clause is not being used factively after all. The attestation below will serve to illustrate this criterion:
Yet they call for smaller, less intrusive government and are indignant over any limits on the Second Amendment right to bear arms.
Doubts about the factive character of “limits on the Second Amendment right to bear arms” are confirmed by the fact that this sequence can be followed by a rejection of the existence of such limits:
They are indignant over any limits on the Second Amendment right to bear arms, but fortunately there aren’t any such limits.
Such a rejection is incompatible with a factive interpretation of the complement.
Having defined the remit of our investigation, we can now move on to describe the results found.
3 The corpus data
This section presents the usage profile of any under the command of emotive factives. One of our objectives here was to assess whether negative emotives are the only actually attested environment in which this polarity item is found. The 22 negative and 22 positive emotive factives listed in Table 1 below were therefore examined in the corpus:
Negative and positive emotive factive predicates.
|Negative emotive||Positive emotive|
|be sorry||be pleased|
|be disappointed||be overjoyed|
|be appalled||be grateful|
|be sad||thank God|
|be angry||thank one’s lucky stars|
|be pained||thank heavens|
|be shocked||be honored|
|be indignant||be moved|
|be disgusted||be satisfied|
|be horrified||be delighted|
|rue||be tickled pink|
|be ashamed||be exhilarated|
|be upset||be ecstatic|
|be mad||be glad|
|be pissed (off)||be gratified|
|be insulted||be euphoric|
|be furious||be jubilant|
A certain group of factives from the list assembled by Wyse could not be classified as either positive or negative, being related rather to the general notion of counterexpectation. The eight counterexpectative factives that we examined were the following: be surprised, be amazed, be astonished, be taken aback, be baffled, be puzzled, be nonplussed and be perplexed. These do not presume whether the feeling experienced is positive or negative, as one can be surprised in either a pleasant or an unpleasant way. Table 2 shows the results found for this group of predicates:
Relative frequency of any licensed by counterexpectative emotive factive predicates.
|COUNTEREXPECTATIVE EMOTIVE||any in scope||total of predicate in corpus||rel. freq. any|
|be taken aback||4||5135||0,00078|
The results of the search for the negative emotives are given in Table 3 below:
Relative frequency of any licensed by negative emotive factive predicates.
|NEGATIVE EMOTIVE||any in scope||total of predicate in corpus||rel. freq. any|
|be pissed (off)||0||3984||0|
Negative emotive factives are thus 1.8 times more frequent with any than are counterexpectative factives.
The biggest gap however lies between negatives and counterexpectatives, on the one hand, and positive emotive factives, on the other. Negative emotives are 46.4 times more frequent with any than positive emotives, counterexpectatives 25.8 times more frequent.
Relative frequency of any licensed by positive emotive factive predicates.
|POSITIVE EMOTIVE||any in scope||total of predicate in corpus||rel. freq. any|
|thank one’s lucky stars||0||433||0|
|be tickled pink||0||172||0|
This raises the question of what negative emotives and counterexpectatives might have in common as a factor favoring the occurrence of any. That common denominator is not difficult to find: it would appear to involve a negative bias against the occurrence of the event denoted by the subordinate clause. With negative emotives, this takes the form of a wish that the event had not occurred; with counterexpectatives, of an expectation that it would not take place. This is confirmed by the contrast between positive emotives, on the one hand, shown in Table 4, and negatives and counterexpectatives, on the other, with respect to their occurrence with some versus any. Table 5 below shows that some is 22.8 times more frequent than any with positive emotives, whereas it is only two times more frequent in the company of any with negative emotives and only 4.6 times more frequent than any with counterexpectatives 4.
Relative frequency of some and any in the scope of positive, negative and counterexpectative emotive factive predicates.
|POSITIVE EMOTIVE||some in scope||total of predicate in corpus||rel. freq. some||rel. freq. any|
|thank one’s lucky stars||0||539||0||0|
|be tickled pink||0||379||0||0|
|NEGATIVE EMOTIVE||some in scope||total of predicate in corpus||rel. freq. some||rel. freq. any|
|be pissed (off)||11||4225||0,0026||0|
|COUNTEREXPECTATIVE EMOTIVE||some in scope||total of predicate in corpus||rel. freq. some||rel. freq. any|
|be taken aback||119||5769||0,021||0,00078|
Since the notion of “unidentified particular referent” expressed by some tends to presuppose the existence of the referent (cf. Larrivée and Duffley 2014), this item shows a certain repulsion both for negative emotive contexts, where the speaker wishes the event had not occurred, and for counterexpectative contexts in which no referent was expected to materialize. Positive reactions, on the other hand, are more concerned with the reality that provoked them than with a wish for non-existence or an estimation of low chances of occurrence of an eventuality.
These observations suggest a connection with Schlenker (2005)’s application of the notion of counterfactual reasoning to the use of the French subjunctive after factives such as être heureux que, illustrated in (15):
Jean est heureux qu’il pleuve.
‘Jean is happy that it is raining-SUBJUNCTIVE’
Schlenker argues that (15) implies that John holds the counterfactual belief that if it were not raining, he would not be as happy as he is. The relation between this account and the use of any with factives can be made through Farkas (1992)’s proposal that the Romance subjunctive is linked to intensional predicates that introduce sets of worlds, i.e. values that vary across a number of different possible scenarios. Farkas (1992: 101) suggests that the occurrence of the subjunctive after factive emotives in French is due to the fact that they “classify situations according to the reaction/emotion they produce or according to some implicit set of criteria (what one considers good/just, etc.);” this makes them like desideratives, directives and modals “in that the attitude they denote does not anchor the complement proposition to any particular world.” The same sort of analysis is also applied to the subjunctive by Giannakidou and Mari (2015), who argue that emotive factives create an emotive modal base in which an individual has the sentiment denoted by the factive only when the proposition expressed by the complement is true, i.e. the individual does not have the sentiment in question if the proposition is not verified. It is this nonveridical component that allows the subjunctive to be licensed after such expressions. Crosslinguistically, the relation of Polarity Sensitive Items to counterfactuality is further exemplified by the ability of the Korean PSI amwu to occur in concessive clauses (cf. Lee et al. 2000) and by the fact that in this language concessives also give rise to expletive negation, which is sometimes treated as a type of polarity-sensitive phenomenon (cf. Choi and Lee 2017). PSI any itself occasionally occurs in concessives in English as well, as demonstrated by the following examples gleaned from the Internet:
- The meeting and the public show of Mugabe speaking authoritatively and, for the most part, cordially with his Generals was a sign that he remains firmly in charge, even if there was any disgruntlement in some quarters.
- Even if there was indeed any physical contact, I would like to raise the fact that the boys had just ran over 1km at an intense pace.
- It’s still frustrating how many people I talk to who just can’t see Paul as “electable”. I guess that’s mainly due to media saying that over and over, to the point I even start to believe it sometimes. I think, despite anyone having any disagreements with Paul’s position, that’s the biggest hump to get voters over.
Bias against an event’s occurrence, either in the form of a preference for or an expectation of its non-occurrence, could therefore be identified as the reason why negative emotives and counterexpectatives allow a Polarity Sensitive Item such as any despite their factive value (cf. Linebarger 1987). This impression is strongest with negative emotives, since if someone is sorry that a certain event occurred they wish it had not and would presumably have liked to prevent its occurrence. Among the counterexpectatives, it is strongest with the predicate amazed: if someone is amazed that anyone survived a plane crash, there was a very strong expectation that no one could survive such a violent accident. What about positive emotive factives, one might ask. They too can be interpreted counterexpectatively: if I am pleased that anorexic Maya has eaten anything at all, her eating was a desirable thing, and may well have been unexpected. Thus all of the attestations of the predicate happy with any in our corpus suggest that the subordinate event might not have been the case, and it is clear from a case such as Juliet was happy to be earning any money at all that the contentment derived by Juliet from earning some money is based partially on the alternative prospect of her perhaps earning none whatsoever. In this respect, Kadmon and Landman (1993: 384–385) observe that the occurrence of any under the predicate glad as in (17) below produces what they call a ‘settle for less’ type of message, i.e. ‘I am willing to settle for what I have, and be glad even about that’ (see also Schwarz 2000):
|A:||But these tickets are terrible!|
|B:||Be glad we got any tickets!|
the interpretation of glad in these examples is closer to surprised than to happy. The sentence voices some tacit negative expectation, along the lines of I didn’t expect to get any tickets. Note that the embedded NPI becomes decidedly worse (perhaps even ungrammatical) when the context rules out a surprise/negative expectation reading of the factive predicate: ??I’m so glad I bought any car!
The common factor in the occurrence of factives with any is therefore the negative bias/counterexpectative dimension. Thus in (17) above any indicates that the speaker is glad that they got whatever tickets they were able to get, indifferently as to the quality of the seats. This naturally tends to implicate that they were not expecting to get any tickets at all, and that the seats they did get were not the best possible ones. Similarly, in (18) below any construes Natalie Wood’s mother as being thrilled at whatever connection she could get to a movie star, indifferently as to the moral unsavoriness of the relationship:
Among the book’s claims: At 16, Natalie was raped by a “powerful, married movie star”; her mother, thrilled at any connection to the screen idol, hushed up the crime.
The essential import of any in such usage concerns the quality of complete indistinguishability of one referent from any other member of its category (cf. Duffley and Larrivée 2015). This value is exploited to convey the message of utter indifference as to what kind of connection was involved with the actor in (18), or that of being glad no matter what kind of ticket the person was able to purchase in (17). With counterexpectatives, the negative expectation covers all possible forms of occurrence of the thing in question, however small or piddling they might be. With negative emotives, what lies behind the occurrence of any is the wish that one had been able to preclude some event whatever shape or form it may have taken. With positive emotives, on the other hand, for any to make sense there has to be some circumstance under which an object’s materialization is desired or welcomed no matter what form it might take.
The use of any in non-emotive factive environments is described in the next section.
4 Emotive versus epistemic factives
Epistemic factives can be divided into two subtypes. Some express “the subject’s state of knowledge” (Shankland 1981), as in (19):
I knew that the tire was flat.
Others denote “the manner in which the subject came to know the truth” (Hooper and Thompson 1973), as in (20):
I noticed that the tire was flat.
Field (1997: 806) notes that epistemic factives index “a great degree of certainty” with respect to the truth of the proposition expressed by their complement. This should make them a particularly hostile environment for any according to Giannakidou (2009)’s hypothesis, as it seems inconceivable that predicates such as those in (19) or (20) could be anything other than veridical. Moreover, it seems difficult to imagine that such predicates could readily involve a counterfactual dimension that would allow the rescuing of any via an inference in the global context of the utterance.
This was confirmed by the fact that all of the occurrences gathered with any to the right of find, know and understand involved a modal or a generic licenser. The examination of the complete list of epistemic factives in the corpus brought to light only 48 uses with epistemic factives that did not contain any such element. A good number (19/48) involved a minimization of the existence of the referent of the noun phrase containing any. This sometimes took the form of a depreciation of the quantity of existing cases (8 examples), as in:
But a senior member of the royal household told the Mirror that any inaccuracies in the diagrams were tiny.
The adjective tiny is used here to downplay of the importance of the inaccuracies. This is similar to the type of impression found in concessive contexts: one could paraphrase (21) by ‘if there were any inaccuracies in the diagrams, they were tiny,’ which shows that the speaker is fictitiously presenting the inaccuracies as hypothetical entities that may not actually exist, even though they are in fact real. Another form of minimization involved qualitative belittlement (6 examples), illustratable by:
However, the company noted that any statistics on explicit security spending are inherently “soft”.
A third form of deflation, attested in four contexts, consisted in conceding the prior existence of the referent but pointing to its current non-existence:
And I remember any concern over my first boyfriend vanished when Dad discovered the new beau was a promising centre. Rugby opened doors.
Closely related to these minimizing contexts were a number of cases (13/48) in which the legitimacy of the existence of the referent of the noun following any was called into question, as in (24) and (25):
A spokesman for the law firm told the ICIJ any allegations the firm helped shield the proceeds of the Brink’s-Mat robbery “are entirely false”.
A representative of the band told Esquire Magazine that any association between the group and the alt-right was “pretty ridiculous”.
These all occurred under the verb tell and contained nouns denoting speech acts, such as allegation, claim, insinuation, suggestion and report.
Finally, a third group of uses (16/48) came to light in which the existence of the noun’s referent was conceded, but not under the mode claimed by some other source. Thus in (26) below the existence of interactions between Weinstein’s lawyers and investment customers is acknowledged, but only inasmuch as they were based on the belief that the money would be properly invested and not on the belief that it would be kept by Weinstein:
The judges also found that any interactions Weinstein’s lawyer had with investment customers was based on his belief that the money would be properly invested, not kept by Weinstein.
This notion is accompanied by an impression of the minimization of the importance of the interactions, a feeling which is even stronger in some contexts:
She told Ghomeshi’s lawyer that any inconsistencies in her statements are due to memory lapses or being nervous.
The common thread running through all of these non-modal, non-generic occurrences of any under epistemic factives is the placing of certain limits on the existence of the referent that is the object of knowledge or perception. This is linked to a concessive impression whereby this referent’s existence is acknowledged as a fact, but not as significant a fact as is thought, or not in the way claimed by someone else. The contribution of any to such contexts is to provide a maximally indefinite referent whose hold on existence is represented as minimal and whose magnitude or importance is therefore amenable to depreciation. Thus in (21) above the import of any allows the application of the adjective tiny indiscriminately to whatever inaccuracies might be present in the diagrams, no matter how insignificant, thereby suggesting that the inaccuracies are practically nonexistent.
The bottom line concerning any’s occurrence with epistemic factives as attested in our corpus however is that no contexts were found in which an epistemic factive actually licensed the occurrence of the PSI. In all cases, the sequence containing any was felicitous outside of the commanding factive. Thus in (27) above, it is possible to say that any inconsistencies in her statements are due to memory lapses or being nervous. This only makes sense: if the epistemic factive signifies knowledge or perception of a fact, it should be possible to assert this fact as existing independently of the knowledge or perception of it. The examples discussed in (21)-(27) above suggest however that the PSI can be associated with emotive dimensions absent from the epistemic predicates themselves. Emotive dimensions have been shown to be associated with counterfactual preferences and expectations that make available a negative reading that allows any under emotive and counterexpectative factives; here they are linked to concessive minimization implying that the materialization of something is so slight as to be practically nonexistent. Whether the occurrence of any with emotive and counterexpectative factives should be treated as licensing proper or as a special form of rescuing is discussed in the next section.
5 Licensing versus rescuing
Larry didn’t say anything.
Larry regrets that he said anything.
We must note first of all that this distinction looks somewhat like a stipulation with a view to saving the nonveridicality theory of PSI licensing: all PSIs are claimed to be licensed by nonveridicality, so uses in veridical contexts such as those under the command of a factive expression are said to be legitimized by something else, here a “rescuing” process.
This distinction is grounded moreover on the assumption that the properly semantic level in language is that of Logical Form, which corresponds to “those aspects of semantic representation that are strictly determined by grammar, abstracted from other cognitive systems” (Chomsky 1977: 5). Logical Form is taken to be the level which fully determines the meaning of a sentence and on which the licensing of linguistic items takes place. In our view however, Logical Form is not a product of the grammar alone, but rather requires the intervention of pragmatic factors in order to be determined (other authors who share this view are Atlas and Levinson 1981; Carston 2002). It does form part of what is communicated, but it is situated on the level of an effect produced by the interaction of linguistically-signified semantic content with contextual and situational factors, and not of a cause. Consequently, we do not subscribe to the distinction made by Giannakidou and Quer (2013: 122, 130) between “agrammaticality,” postulated to be determined by Logical Form regulated exclusively by the grammar, and “infelicity,” which is taken to be mere pragmatic oddness. Based on this distinction, they claim that a sequence such as (30) below is ungrammatical because it violates a grammatical licensing condition on negative polarity items, while (31), uttered in a context in which there is only one cookie, is merely infelicitous and “doesn’t damage the sentence as badly as the unbound polarity variable does”:
*Bill bought any presents.
?Eat any cookie!
However the average English speaker would be far more mystified upon hearing (31) in a situation in which there was only one cookie in sight than by (30), for which one could imagine a context in which it might make sense; cf. (32):
He bought any ticket to any NBA game in the league.
Of course, the context makes it clear in (32) that the ticket-buying is a repeated occurrence, but this just confirms our point about Logical Form being conditioned not only by the grammar but by pragmatic factors as well.
Giannakidou and Quer seem moreover to have unwittingly begged the question of the existence of a real distinction between agrammaticality and infelicity by only giving a context for (31), which biases the discussion towards a pragmatic treatment of this sequence, while (30) is treated as if people could make sense of it outside of any context. Moreover, regarding the purported agrammaticality of (30), a careful look at the data shows that in some situations episodic past perfective predicates are indeed compatible with any, as attested by (33) below:
A Columbia University psychiatrist reports that students come to him to find out what is wrong with them if they are not having intercourse. “My virginity was such a burden to me that I just went out to get rid of it,” a junior at the University of Vermont revealed to a Boston sex counselor. “On a trip to Greece, I found any old Greek and did it so it wouldn’t be an issue any more.”
(TIME Magazine Corpus)
This suggests that the supposedly grammatical licensing condition on NPI any in (30) is in fact partly pragmatic, i.e. it depends on the speech situation and the communicative intentions of the speaker, just like the oddity of (31) in a situation with a lone cookie. We would thus concur with Israel (2011) that polarity is essentially a pragmatic phenomenon. 5
The implications of this for the distinction between licensing and rescuing are clear: there is no motivated difference between them, as both are partly pragmatic and partly semantic. The fact of any occurring under the negative operator in the Logical Form of (28) above is a consequence of the logical interaction between the meaning of the negation and that of any; proof of this is provided by the fact that, due to its meaning, the determiner some is not interpreted as falling within the scope of negation in the logical formula (at least not in default cases – see Larrivée 2012 for discussion):
Larry didn’t say something.
This is because some denotes a particular but unidentified thing that Larry is represented as not having said (something the speaker knows about and was expecting Larry to mention, for instance). This idea must be paraphrased by a logical formula with the existential operator outside of the negation, as in ∃x (NEG (Larry say x)). Any, on the other hand, evokes a qualitatively indefinite referent as not having been said; this implies an indiscriminate negation of all the things that Larry might possibly have said at the moment referred to. Since this conveys the idea that no entity escapes the negation, it is paraphrasable by a logical formula with the existential inside the negation, as in NEG (∃x (Larry say x)). As for the alleged rescuing of any in (29), the fact that regret logically entails the truth of the action regretted precludes paraphrasing this utterance by a logical formula in which ‘Larry said anything’ is under the scope of negation. This does not prevent however the indiscriminate application of the counterfactual ‘wished-he-had-not-said’ side of the regret to all the possible things that could have been said in this situation, including what actually was said.
To conclude, we would argue that the divide between linguistic semantics and pragmatics is not to be drawn on the level of decontextualized Logical Form vs. subsequent contextually-based inferences. To obtain a Logical Form, one must already invoke context (cf. Carston 2002). We would advocate rather an approach to linguistic semantics that builds up from stable sign-meaning units in which a stable linguistic form is paired with a stable meaning. Since there is generally no one-to-one correlation between meaning and form on the level of the sentence, this entails that, properly speaking, Logical Form is generally not a linguistic-semantic level. A semiological approach to semantics (cf. Langacker 2000) draws the line between linguistic semantics and pragmatics in terms of the divide between what is linguistically expressed and what is not linguistically expressed but still communicated in the message conveyed by the speaker to the hearer. From this perspective, the occurrence of any in both (28) and (29) above is a complex semantico-pragmatic phenomenon that depends both on the semantic content of any and on its interaction with the meanings and implicatures associated with it in these contexts.
The purpose of this paper has been to investigate the actual usage of any in veridical factive contexts. A clear answer emerges from the systematic corpus investigation conducted here: any is found with emotive factives that have an expressive dimension yielding a negative or counterexpectative reading. This confirms an intuition that goes all the way back to Baker (1970: 182), who remarked that
speaking intuitively, we can say that each of these predicates expresses a relation of contrariness between a certain fact and some mental or emotional state. For example, we say that we are surprised when a certain fact does not conform to our expectations (…).
This empirical confirmation has been achieved by investigating whether epistemic factives license any to the same degree as emotive factives do, and whether any is equally at home in positive and negative emotive environments.
Our study also provides further support for Horn’s (2016) claim that licensing is based on at-issue content. Thus, the felicity of a sentence such as Larry is glad I said anything relies on the availability of the counterexpectation of the undesirable possibility that I might have remained silent; it is because the factive predicate amazed readily implies such counterexpectations that a sentence like Larry is amazed I said anything is universally accepted as felicitous in studies based on decontextualized examples. The meaning of any converges with this counterexpectative dimension through the exploitation of the referential indiscriminacy that it signifies: the amazement applies to whatever the person said, no matter how trivial it might have been. With negative emotives (e.g. regret), the wish that some object or action, whatever shape or form it may have taken, not have been realized lies behind the occurrence of any; since the very notion of regret involves wishing that something was not the case, the counterfactual idea is at-issue here. With positive emotives such as be happy, on the other hand, in order for any to make sense there has to be a circumstance under which some object’s materialization is desired or welcomed no matter what form it might take. Thus in (35) below the context makes it clear that reference is made to pessimists who are wary even in the face of record corn crops and refuse to take for granted that there will be any increase in harvests or profits at all:
… crops dropped about a third from record high prices this summer, according to the United States Department of Agriculture. But even those prices mean record incomes for many farmers in the region, and continued strong demand from overseas has kept prices from plummeting. From1993 to1995, the value of agricultural and livestock exports from the northern Plains grew 51.7 percent, to $9.5 billion from $6.3 billion, according to Ernest Goss, a professor of economics at Creighton University in Omaha, who does a regular survey of the region. Some people in Geneva seem happy that there is any growth at all. “It’s been a really good year, no question,” 24-year-old Brian Lewis shouted above the roar of a combine in a cornfield at the edge of town. He was working a 14-hour day, and the machine he was operating was spitting out a steady stream of corn into grain trucks that came and went almost continuously. “Where I’m from, this is big,” he said, referring to the generous harvest.
(‘Corn Crop Brings Smiles to Nebraska, But Some Are Wary’ in COCA)
This can be contrasted with (36):
The report talked about nine out of 10 homes not having adequate staffing, and that’s something that ombudsmen as resident advocate and problem solvers have been talking about for many years, and we’re happy that it’s finally getting some attention so that maybe some things can be done to address the problem.
Here the speaker simply rejoices at the attention which is finally being given to the problem, and there is no suggestion of the counterexpectative attitude that he was not expecting any attention to be devoted to the issue at all and so is happy at whatever heed is being paid to it, no matter how little.
The counterexpectativity account proposed here might seem to be discounted by the fact that any can occur with counterexpectatives even when the latter are negated:
We don’t resent any development, but we do want to see development in keeping with the locality.
I don’t regret saying anything.
I’m not amazed that any segment of the population has a negative bias towards woman.
It should be noted however that these are all cases of focus on the indiscriminacy notion conveyed by any: (37) can be paraphrased by ‘just any’; (38) and (39) have stress on any and imply ‘even the extreme value on the scale’. As argued in Duffley and Larrivée (2010, 2012), any is used in this type of context for its meaning of utter indiscriminacy, which is the focus of the negative here. It is merely fortuitous that the predicates which are negated in these contexts should be counterexpectatives.
The study reported here supports the view that there is a meaning-based regularity underlying the use of any, such that setting usage in factives apart from other uses is not warranted. There is little empirical reason to identify these as “rescued” rather than “licensed”, apart from the fact that some theories may find it difficult to integrate such attestations. After all, any has the same reading of qualitative indefiniteness with factives as in other similar contexts: in I’m glad the movie was any good it evokes the same indiscriminate coverage of all degrees of quality as in Any rust would ruin the mechanism, where its indiscriminate exhaustification of quantity covers all possible quantities of rust, even the tiniest ones. Clearly, licensing by at-issue meaning seems a more generally encompassing and unified approach than one based on a distinction between licensing and rescuing.
Indeed, in more recent work Giannakidou (2015) has moved away from a rescuing account, arguing that emotive factives show “weak licensing,” which is attributed to the existence of a conflict between their veridical presupposition (I regret I caused you any inconvenience → I caused you some inconvenience) and their nonveridical assertion (I may not have caused you any inconvenience, in which case I don’t have any regrets). Giannakidou and Mari (2015) argue that the weak licensing account is supported by the observation of variation across different languages in the use of the subjunctive or the indicative after emotive factives, some requiring this mood to be used exclusively (e.g. French), some allowing both moods (e.g. Catalan), and some selecting the indicative (e.g. Romanian). Giannakidou (2015) adduces the low frequency of any in the scope of emotive factives as another piece of evidence in favor of this account. However, some term of comparison is needed in order to evaluate whether this frequency is really significant or not. The comparison of the frequency of occurrence of any and some in the scope of emotive factives which has been made here seems to lend support to Giannakidou’s position: the tables given above show that any is 4 times less frequent than some with positive emotives but only 2 times less frequent with negative emotives. Interestingly, pure counterexpectatives even favor any a bit more than negative emotives, any being only 1.4 times less frequent than some in the scope of these predicates. This points to the need to examine the behavior of counterexpectatives more closely, in particular their use with subjunctive or indicative complements in languages that utilize these moods in subordinate clauses. Such an investigation falls beyond the scope of the present study, but we do hope to have opened up some new lines of enquiry with respect to factive predicates.
We wish to thank Larry Horn, Anastasia Giannakidou, Louis de Saussure, and the audience at the workshop “The pragmatics of grammar: negation and polarity”, held May 19–20, 2015 at the Université de Caen, for their input. What we have done, or not done, with it is our sole responsibility.
Atlas, Jay David. 1996. “Only” noun phrases, pseudo-negative generalized quantifiers, negative polarity items, and monotonicity. Journal of Semantics 13. 265–328.
Atlas, Jay David & Stephen C. Levinson. 1981. It-clefts, informativeness and logical form: Radical pragmatics. In Peter Cole (ed.), Radical pragmatics, 1–62. New York: Academic Press.
Baker, C. Lee. 1970. Double negatives. Linguistic Inquiry 1. 169–186.
Brandtler, Johan. 2012. The evaluability hypothesis: The syntax, semantics and pragmatics of polarity item licensing. Amsterdam & Philadelphia: John Benjamins.
Carlson, Gregory N. 1981. Distribution of free-choice any. Papers from the Seventeenth Regional Meeting of the CLS, 8–23. Chicago: Chicago Linguistic Society.
Carston, Robyn. 2002. Thoughts and utterances. The pragmatics of explicit communication. Oxford: Blackwell.
Chierchia, Gennaro. 2006. Broaden your views. Implications of domain widening and the “logicality” of language. Linguistic Inquiry 37. 535–590.
Choi, Yoonhee & Chungmin Lee. 2017. Expletive negation and polarity alternatives. In Chungmin Lee, Ferenc Kiefer & Manfred Krifka (eds.), Contrastiveness in information structure, alternatives and scalar implicatures, 175–201. Dordrecht: Springer.
Chomsky, Noam. 1977. Essays on form and interpretation. New York: North-Holland.
Crnič, Luka. 2011. Getting even. Cambridge, MA: Massachusetts Institute of Technology dissertation.
Dayal, Veneeta. 2013. A Viability Constraint on alternatives for Free Choice. In Anamaria Falaus (ed.), Alternatives in Semantics, Basingstoke: Palgrave Macmillan. version on V. Dayal’s website, http://www.rci.rutgers.edu/~dayal/viability-2012.pdf, accessed on May 13, 2013.
Duffley, Patrick J. & Pierre Larrivée. 2012. Collocation, interpretation and explanation: the case of just any. Lingua 122. 24–40.
Duffley, Patrick J. & Pierre Larrivée. 2015. A fresh look at the compatibility between any and veridical contexts: The quality of indefiniteness is not strained. Lingua 158. 35–53.
Farkas, Donka. 1992. On the semantics of subjunctive complements. In Paul Hirschbühler (ed.), Romance languages and modern linguistic theory, 69–105. Amsterdam & Philadelphia: John Benjamins.
Field, Margaret. 1997. The role of factive predicates in the indexicalization of stance: A discourse perspective. Journal of Pragmatics 27. 799–814.
Giannakidou, Anastasia. 2009. The dependency of the subjunctive revisited: Temporal semantics and polarity. Lingua 120. 1883–1908.
Giannakidou, Anastasia. 2011. Positive polarity items and negative polarity items: Variation, licensing, and compositionality. In Claudia Maienborn, Klaus von Heusinger & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, 2nd edn 1660–1712. Berlin & New York: De Gruyter Mouton.
Giannakidou, Anastasia. 2015. Veridicality conflict produces weaker licensers: A fresh look at emotive verbs. Paper presented at CRISCO workshop the Pragmatics of Grammar: Negation and Polarity, University of Caen, 19–20 May.
Giannakidou, Anastasia & Alda Mari. 2015. Emotive factives and the puzzle of the subjunctive. Paper presented at the Chicago Linguistic Society, University of Chicago, 23–25 April.
Giannakidou, Anastasia & Josep Quer. 2013. Exhaustive and non-exhaustive variation with free choice and referential vagueness: Evidence from Greek, Catalan and Spanish. Lingua 126. 120–149.
Hoeksema, Jack. 2013. Polarity items in Strawsonian contexts – A comparison. In Eva Csipak, Regine Eckardt, Mingya Liu & Manfred Sailer (eds.), Beyond ‘any’ and ‘ever’, 47–78. Berlin & Boston: De Gruyter Mouton.
Hooper, Joan. 1975. On assertive predicates. In John P. Kimball (ed.), Syntax and semantics, 91–124. New York, NY: Academic Press.
Hooper, Joan & Sandra Thompson. 1973. On the applicability of root transformations. Linguistic Inquiry 4. 465–498.
Horn, Laurence. 2016. Licensing NPIs: Some negative (and positive) results. In Pierre Larrivée & Chungmin Lee (eds.), Negation and polarity. Experimental and cognitive perspectives, 281–305. Dordrecht: Springer.
Israel, Michael. 2011. The grammar of polarity: Pragmatics, sensitivity, and the logic of scales. Cambridge: Cambridge University Press.
Langacker, Ronald W. 2000. Grammar and conceptualization. Berlin & New York: Mouton de Gruyter.
Larrivée, Pierre. 2012. Positive polarity items, negation, activated propositions. Linguistics 50(4). 869–900.
Larrivée, Pierre & Patrick J. Duffley. 2014. The emergence of implicit meaning: Scalar implicatures with some. International Journal of Corpus Linguistics 19. 526–544.
Lee, Chungmin, Daeho Chung & Seungho Nam. 2000. The Semantics of amwu-N-to/-irato/-ina in Korean – Arbitrary choice and concession. In Akira Ikeya & Masahito Kawamori (eds.), PACLIC 14: 14th Pacific Asia conference on language, information and computation proceedings, 413–424. Tokyo: Waseda University International Conference Center.
Schlenker, Philippe. 2005. The lazy Frenchman’s approach to the subjunctive. In Twan Geerts, Ivo van Ginneken & Haike Jacobs (eds.), Romance languages and linguistic theory. Selected papers from ‘Going Romance’ 2003, 269–309. Amsterdam & Philadelphia: John Benjamins.
Schwarz, Bernhard. 2000. Notes on ‘even’. Stuttgart: Stuttgart University unpublished manuscript.
Shankland, Mary Ellen. 1981. Factivity from a discourse perspective. Linguistic Notes from La Jolla 10. 20–32.
Wyse, Brendan. 2009. Factive/non-factive predicate recognition within question generation systems. Milton Keynes: The Open University MS thesis. http://computing-reports.open.ac.uk/2009/TR2009-09.pdf (accessed 15 August 2014).
Xiang, Ming, Julian Grove & Anastasia Giannakidou. 2013.Dependency-dependent interference: NPI interference, agreement, and global pragmatic inferences. Frontiers in Psychology 4. (accessed 6 October 2015).
Wyse bases his list in large part on that given by Hooper (1975). We adopt Hooper and Thompson’s (1973: 479) definition of emotive factives as verbs that “express some emotion or subjective attitude towards a presupposed complement.”
The NOW corpus only allows searches up to five words to the right of the search item.
Note that we are not concerned here with whether any belongs to the category of Negative Polarity or Free Choice; for discussion of this topic, see inter alii Aloni (2007), Carlson (1981), Chierchia (2006), Dayal (2013), Giannakidou (2001), Giannakidou and Quer (2013). The purpose of our study is to pin down the actual distribution of any with veridical licensers, not to engage in the theoretical debate around the NPI-FCI question.
Overall, some is 1.4 times more frequent than any across all contexts in the three corpora used for the study.