PAUL BERCKMANS
THE QUANTIFIER
T H E O R Y OF
EVEN
I. INTRODUCTION
It has long been accepted that the use of e v e n typically expresses unexpectedness or surprise. There is also some agreement that e v e n implies the existence of a contrast class whose members share a relevant property with the object in focus of e v e n . For example, the use of e v e n in such a sentence as (1): (1)
Even Ted was sober,
is thought to express that Ted's sobriety was unexpected, or at least less expected than the sobriety of other individuals who are members of a relevant class. There is much disagreement about the precise nature of this class, its size, and how it is ordered. On Bennett's (1982) account, for example, e v e n requires only the existence of at least one other individual who shares a relevant property. Fauconnier (1975) has offered the now popular theory that e v e n sentences imply not just a related less surprising item but a scalarly ordered contrast class, with the object in focus of e v e n at the end of such a scale. So, in example (1), Ted is presumably the surprising item at the end of a scale of individuals who can be rank-ordered according to their degree of expected drunkenness. Very recently, W. G. Lycan (1991) and Stephen Barker (1991) have argued (separately) that sentences with e v e n make an implicit reference to a universal class. My aim in this paper is to examine a number of problems with respect to their versions of what I will call the universal quantifier theory of e v e n . I will accommodate Lycan and Barker, retaining the idea that e v e n implies quantification over a wider class, but I will produce evidence suggesting that e v e n is ambiguous as between universal quantification and existential quantification. II. BARKER'S
UNIVERSAL
INSTANTIATION
THESIS
Before looking at the universal quantifier theory of e v e n , we must pay some attention to Bennett's treatment of e v e n , which has been a point of departure for Barker and Lycan. Bennett argues that, although e v e n does Linguistics and Philosophy 16: 589-611, 1993. © 1993 Kluwer Academic Publishers. Printed in the Netherlands.
590
PAUL BERCKMANS
not contribute to the truth-value of a sentence, it does contribute in an important way to the assertibility of such a sentence. To appreciate the assertibility conditions Bennett offers, we must be aware of some notational conventions. (I will use these throughout the paper.) Let S be any sentence with e v e n , let S* be S from which e v e n is deleted, and let Sj be any grammatical sentence identical with S* except for the focus of e v e n . (Bennett calls Sj a neighbor.) To give an example, if S is "Even Ted was sober," then S* is "Ted was sober," and possible neighbors are "Tom was sober," "Harry was sober," "Everybody was sober," etc. Bennett stipulates the following conditions for the assertibility of e v e n : (i) (ii) (iii)
Sj is true and mutually believed by the speaker and hearer, and salient for them (e.g. it has just been authoritatively asserted); the truth of S* and that of Sj can naturally be seen as part of a single more general truth; it is more surprising that S* is true than that Sj is true.
A felicitous utterance of "Even Ted was sober," for example requires at least that Ted was sober, that both Ted and, say, Frank were known to be sober, and that Ted's sobriety was more surprising than Frank's. More generally, an e v e n sentence is assertible only if "there is a neighbour sentence which is known, related and less surprising." 2 Before advancing their own theories, Barker and Lycan give some examples showing that the conditions of assertibility are not sufficient, i.e. some assertions of e v e n sentences may be infelicitous although the conditions (i)-(iii) are satisfied. They also show that this set of conditions is not necessary, attacking especially condition (i). Their criticisms are plausible enough and I need not add anything to that discussion here. Without abandoning Bennett's idea that e v e n contributes to the felicity conditions of the statement, Barker suggests the following assertibility conditions for e v e n sentences: (i)
(ii)
S* and Sj are asserted as universal instantiation cases of an implied or stated Su (where S, is a universal generalization of S*); S* is an extreme case of S,.
The assertion of such an e v e n sentence as "Even Ted was sober" thus implies some universally quantified sentence like "Everybody was sober,"
1 Bennett (1982, pp. 405-406). 2 Bennett (1982, p. 406).
THE
QUANTIFIER
THEORY
OF EVEN
591
with the understanding that " T e d was sober" is an extreme case of that sentence. To forestall a number of obvious counterexamples, we should immediately add two qualifications about the implied contrast class. Although Barker doesn't say much about the precise nature of this implied class, it will be appropriate to presume that this class is contextually determined. No one would seriously take (1) to imply that everybody in the entire universe was sober, but rather that the contrast class in that example consists of, say, all individuals present at a party, or present at the estate of the individual. This adjustment will take care of an otherwise obvious sort of counterexample. Secondly, we should probably tolerate that the generated S, sometimes be taken nonliterally, metaphorically or in some hyperbolic or exaggerated sense. Suppose, for example, that one hears someone say: (2)
Foreign cars are enormously popular. Even Pat Buchanan drives one.
On Barker's account, the use of e v e n in (2) would have to entail the statement " E v e r y o n e drives foreign cars." This statement is patently false, not only if we take the class of all car owners as the relevant class; the context doesn't appear to suggest a determinable class for which the statement comes out true. But one could understand the implied statement as an exaggeration in the same way that " n o b o d y buys American cars anymore" will typically be understood in a nonliteral hyperbolic way. Since such uses of universal quantifiers are quite common, there is little harm in accepting that universal generalizations of e v e n sentences, too, sometimes involve such a nonliteral universal quantifier. With these qualifications in mind we can look at some examples for which universal quantification will be difficult to maintain, even if we make liberal use of contextually determined contrast classes and metaphorical uses of quantifiers. Consider (3a) and (3b): (3a) (3b)
The denazification was organized so sloppily initially that even Eichmann and Mengele escaped. His quiz average was so low that even a perfect final could not save him.
A universal generalization of (3a) cannot readily be rendered nonliteral. For (3a), the universal instantiation hypothesis postulates the existence of a less extreme neighbor, perhaps some lesser known or less infamous Nazi criminal. But we must also take (3a) and its neighbor as an instantiation of such an implied Su as " T h e denazification was organized so sloppily
592
PAUL BERCKMANS
initially that everybody escaped." Even if we allow a very generous nonliteral reading of the quantifier ranging over a narrow, contextually determined contrast class containing only high-ranking Nazis, universal quantification is difficult to sustain. Such a generalization is contradicted by history, 3 The " s o . . . that even" construction, however, doesn't always give rise to such failure to imply a universal generalization, as shown in (3b), which implies that nothing (including a perfect final) could have saved the student. But I see no immediate resolution for (3a) that could supply a satisfactory universal generalization. A second problem for Barker's account that I will discuss briefly relates to his discussion of what Bennett calls the comparative-intensifying sense of e v e n . Bennett had pointed out that his account does not capture the use of e v e n in such comparisons as (4): (4)
Their house is even larger than ours.
Bennett argues that this comparative-intensifying use of e v e n is a lexically different use, finding some justification in the fact that in French different words are used for these separate senses of e v e n . 4 Barker responds that this comparative-intensifying use of e v e n can be "easily" subsumed under his universal quantification account. One of the examples he discusses is
(5): (5)
Mr. Big is even richer than Kerry Packer.
After first stressing that we shouldn't take a paraphrase of what would be implied by (5) as literal, i.e., as stating that Mr. Big is richer than everybody else in the entire world, Barker suggests that the relevant comparison class is that of all Australians (I assume that Kerry Packer is a rich Australian), offering this paraphrase:
(5')
All millionaires in Australia are such that Mr. Big is richer than they are.
I am not convinced that (5') represents a satisfactory paraphrase of (5). For one thing, if the universal instantiation thesis is argued for, it would
3 Notice that although Eichmann and Mengele are very close to the end of a scale (that of Nazi war criminals) they are not quite at the end of a scale, as we would expect if Fauconnier were right about even. 4 Bennett (1982) points to the difference in French between encore and mOme. In D u t c h a similar distinction exists between zelfs, for the original use of even, and nog, for the comparative-intensifying use. But there are too m a n y circumstances in which nog and zelfs can be used interchangeably to make this sort of evidence compelling. I look at the linguistic data from French and D u t c h with extreme caution, but I accept the distinction.
THE QUANTIFIER
T H E O R Y OF E V E N
593
have been reasonable from the beginning to say that e v e n sentences imply universal quantification over a c o n t e x t u a l l y d e t e r m i n e d relevant comparison class, not the entire universe of discourse. Barker's example (4) and paraphrase (4') appear to address that point. The question remains whether uses of e v e n exist in comparative sentences for which the very idea of universal generalization is not plausible - even with a narrow contrast class. A different reading of (5) suggests the incorrectness of Barker's treatment of the comparative-intensifying use of e v e n . 5 Suppose that during a conversation about wealth Kerry Packer is mentioned to show that wealth is always acquired harmfully. The following exchange takes place after someone suggests that this isn't always so: A: B: A:
Take the case of Mr. Big. Hold on, though; we are talking about really rich people. Does Mr. Big qualify? Mr. Big is even richer than Kerry Packer.
In the preceding conversation A's last statement does not imply a universal generalization. One further example that presents a difficulty is this. Suppose that police officers discover the bodies of murder victims as they enter the apartment to which they were called by a neighbor. After quickly checking the site for a possible suspect, they rush to their car to call in the discovery. One could describe their return to the apartment and what they found there as follows: (6)
After they returned for a second inspection they found even more bodies in the closet.
This sentence could be taken as ambiguous in at least two ways. In one sense, (6) could mean that the number of bodies found on the second trip exceeded that of all bodies found initially. This reading could be paraphrased as "After they returned for a second inspection they found in the closet more bodies than the number of all bodies found earlier." Universal quantification in this example ranges over a comparison class, the class of bodies found first. Barker's universal paraphrase looks correct if we have such a comparison in mind. The other reading of (6) however, the one presumably intended by the utterer of (6), cannot be accounted for by Barker's hypothesis. One could assert (6) if during the first inspection three bodies turned up and during the second visit only two, thus violating the universal generalization and the requirement that there exist some less extreme Sj, since the existence of three dead bodies (I presume) is more 5 A referee of Linguistics and Philosophy offered this example.
594
PAUL
BERCKMANS
extreme than the existence of two. The use of even in this second sense points to a use of even as an intensifier of a comparison between two individuals, objects or sets of objects instead of a comparison between a member of a universal class and the rest of that universal class. While no explicit " m o r e . . . than" construction is present here, a comparison is invited between the number of bodies found during the first visit and the total body count after the discovery of the additional bodies. Comparisons of states of affairs when no larger class is entailed or presumed are generally counterexamples. Consider the following exchange between A and B in which B describes to A a situation at a convention hotel: (7)A: B: A:
I heard the elevators were really slow. Well, only two were working! That's even worse.
A comparison is made between two states of affairs: the state of affairs that consists of having two working elevators, and the state of affairs that consists of having all (or most?) elevators work slowly. Does A's statement entail the statement "Every state x is such that having only two working elevators is worse than x ? " Hardly. Presumably, having only one working elevator or none at all, is far worse. A universal class for (7) or a relevant St would be difficult to argue for without resorting to ad hoc restrictions. Barker discusses some other examples which appeared not to entail a contrast class. One involved a report of an encounter with a yeti: (8)
It was very tall, eleven, even twelve feet!
Barker argues that this even statement should not be taken literally to mean that the yeti was eleven feet and even twelve feet. He suggests that eleven f e e t and twelve feet are extremes of a range within which certain relevant heights fall. He paraphrases (8) as (8'): (8')
The heights eleven feet, even twelve feet accord with my impression of the yeti,
which in turn is thought to be an implicit conjunction of statements whose conjuncts (8S~) and (8S) are instantiation of (8S,): (8Sj) (8S) (8S,)
The height eleven feet accords with my impression; Even the height twelve feet accords with my impression; Any height falling roughly between eleven and twelve feet accords with my impression of the yeti's height.
The resulting paraphrases and the universal generalization probably reflect what a speaker would have meant by (8). But the analysis leaves a few important aspects of the structure of the preceding even sentence unex-
THE QUANTIFIER
THEORY
OF EVEN
595
unexplained. Let us note first that, in such sentences as (8), e v e n appears acceptable only at the end of an enumeration, not at the start or midsentence. It would be rather odd to say "It was very tall, even eleven, twelve feet!" Notice also that it is infelicitous to say "It was even twelve feet!" It is apparent that whatever appears in the focus of e v e n requires another element for comparison in an enumeration. Such an enumeration often shows up as an explicit d i s j u n c t i v e enumeration, or can be rendered explicitly as a disjunctive enumeration, as in (9): (9)
It was very tall, eleven, or even twelve feet!
We might add (10) as an example, and its more explicit form (10'): (10) (10')
By year's end I expect the Capital Growth Fund to go as high as $17.50, $18.00, even $19.00; By year's end I expect the Capital Growth Fund to go as high as $17.50,.$18.00, or even $19.00.
The progression of the items in such disjunctive enumerations reflects a certain rank order which allows an explicit comparison. 6 But the explicit disjunction in such enumerations as (9) and (10') cannot be explained away that easily. To find a universal reading for a disjunctive enumeration like (10'), we first have to ignore the disjunctive surface structure of the sentence. Then, to avoid reading the statement as a conjunction we must take the disjuncts as extremes or parameters of a range within which universal quantification exists. Finding a general justification for Barker's solution will be challenging, especially since it is not at all obvious which e v e n statements are to be taken as literal or nonliteral disjunctions, or, for that matter, when a n y disjunctive enumeration is to be taken literally or nonliterally. Consider (11): (11)
He must be Harvard, or even Yale. 7
As in the previous examples, the items in this enumeration can be rankordered, for example, in virtue of the property of desirability, a Yale student scoring higher than a Harvard student as a marriage prospect. But what is absent here - and this makes it different from the previous enumerations - is a range. It is highly implausible that " H a r v a r d " or " Y a l e " are the extremities between which a number of other schools could fall. Quite the contrary, it appears that the speaker is suggesting for the
6 T h e existence of such enumerations lends support to Fauconnier's view that the object in focus of even is located at the end of scale, and it should thus be no great surprise that even in such sentences modifies an end-of-scale object. 7 Verbatim from Where the Boys Are, said by Yvette Mimieux.
596
PAUL BERCKMANS
individual only two possibilities, Harvard and Yale, not a universal class including Harvard and Yale or a range from Harvard to Yale. Barker would suggest that a paraphrase of (11) would look like this: (llSj) • (11S) (11S,)
Harvard accords with my impression of his university; Even Yale accords with my impression of his university; Either establishment accords with my impression of his university.
The suggestion is that either functions as a universal quantifier whose domain has only two members, Harvard and Yale. 8 I doubt if such a solution is very promising. I am troubled by the suggestion that either is a universal quantifier. Surely, either is not always a universal quantifier. If it were, all disjunctions would somehow be associated with universal quantifiers. That is clearly not true. So perhaps only some either sentences are universally quantified sentences. If that is so, Barker will need to explain under which circumstances such either sentences are universally quantified sentences and under which they are not, i.e., when they are just disjunctions. This brings us back to the earlier criticism that there is no hard and fast rule that tells us when to take an even disjunction literally as a disjunction or nonliterally as something else, for example, as a conjunction, or as a universally quantified sentence, as proposed in this example. But even when we grant that such an either sentence as (11S,) is a universally quantified sentence, I reject the idea that it is a universal generalization of (11). Barker's account of even understands an even sentence as implying a universally quantified sentence. A simple way to generate such universal generalizations would add a great deal of plausibility to the theory. 9 1 am not certain if Barker's account can provide such a method. Let me illustrate the point I am making. Imagine that Evans is so desperate to become a doctor that he is willing to attend the worst medical school that would accept him if he shouldn't get in anywhere else. In a conversation during which Colson Medical College (a very bad school) and the Robertson Institute of Medicine (arguably the worst school) are mentioned, someone asks Evans whether he would really attend such poorly regarded institutions. To this he replies: (12)
I'd go to Colson, or even Robertson,
8 Barker (1991, p. 23). Barker's own example is "Either w o m a n is right for the job," said after two suitable w o m e n are interviewed for a position. 9 Lycan's thesis is straightforward in this respect.
THE QUANTIFIER
THEORY
OF EVEN
597
indicating that he would go anywhere, no matter the quality of the school. Here, the relevant Su is something like "I would go anywhere, including Colson or Robertson," which I think is entailed by the corresponding even sentence. But I doubt strongly that "I would attend either Colson or Robertson" is a universal generalization of (12), as we would expect if Barker is right about these sorts of disjunctions. Along this line of reasoning, I see no reason why we should accept "Either establishment accords with my impression of his university" as the universal generalization of (11) instead of the more evident " H e must be any school including Harvard or Yale." Barker's account does not seem capable of explaining why such similar even sentences would entail such different universal generalizations. It further rests on an unelaborated view of either as a universal quantifier that would require much more explanation. As a result, if we want to maintain some form of a universal quantifier account of even, we need to find a suitable modification of Barker's universal instantiation thesis. III.
LYCAN'S
MODIFIED
ACCOUNT
Lycan wilt argue that even does contribute to the truth value of the statement, and, motivated by his own theory of conditionals, he advances a general account of even, only, even i f and only if. On Lycan's initial account of even, a sentence like (1): (1)
Even Ted was sober,
implies that all members of a relevant wider class of individuals were sober, paraphrased as: (la)
Everybody including Ted was sober.
E v e n is thus a universal quantifier. After looking at a number of difficulties
for this account Lycan considers two modified versions of the thesis, the "within reason" thesis, and the "plus" thesis. The first analyzes even as " e v e r y . . . within reason," the second as " e v e r y . . . plus." To illustrate, (1) gets paraphrased as follows:
(lb) 0c)
Everybody within reason, including Ted was sober, Everybody who was expected to be sober was sober, plus Ted (who wasn't expected to be sober).
Lycan rejects the "within reason" thesis as ad hoc, adopting the "plus" thesis as his final version of his universal quantifier theory of even if it should turn out that the defects of the original analysis cannot be reme-
598
PAUL
BERCKMANS
died. L y c a n concludes that, despite the shortcomings of any of the different versions of the universal quantifier thesis, "it is hard to d o u b t that e v e n involves universal quantification in some way. ''1° L y c a n ' s " p l u s " thesis not only avoids the sorts of objections we a d v a n c e d earlier against B a r k e r ' s account; it explains these counterexamples if we give a " p l u s " rendering to the universal generalization Su in B a r k e r ' s explication of e v e n . 11 F u r t h e r m o r e , L y c a n adopts B e n n e t t ' s distinction b e t w e e n the intensifying use of e v e n and the c u s t o m a r y use of e v e n . H e treats e v e n in disjunctive e n u m e r a t i o n s as comparative-intensifying uses, thus eliminating comparisons and disjunctive e n u m e r a t i o n s as counterexamples. 12 But L y c a n ' s " p l u s " version, too, is not resistant to difficulties. Let us look at some. A first group of c o u n t e r e x a m p l e s involves the use of e v e n with b e f o r e and after. T h e idea of universal quantification is hard to resist in such examples as (13) and (14): (13) (14)
E v e n after he was diagnosed with cancer he kept smoking; H e fooled a r o u n d even before they were married.
T h e implication in (13) is that the m a n s m o k e d before as well as after the diagnosis. In (14) the subject m a y be t h o u g h t to have b e e n fooling a r o u n d b e f o r e and after the wedding. E v e n in such examples - we might argue with L y c a n - entails universal quantification over a relevant range of m o m e n t s in time. Such a reading, in contrast, is hardly appropriate for (15) and (16): (15) (16)
E v a n s kissed M a r y even before he k n e w her n a m e ; 13 H e was executed even after the q u e e n herself m a d e a plea to spare his life. 14
Unlike the actions described in (13) and (14), the actions referred to in 10 Lycan (1991, p. 29.) 11 I am still not convinced that (3a) is adequately captured by the "plus" account. For (3a) "The denazification was organized so sloppily initially that even Eichmann and Mengele escaped", the "plus" account would require the implication that all those who were expected to get away did in fact get away. Sure enough, Eichmann and Mengele got away, but it is questionable that their escape was less expected than the escape of, say, Klaus Barbie, or Joseph Stangle. It is also not certain that those who were expected to get away, did, in fact, get away. One could sincerely and truthfully utter (3a) without having to presume or commit oneself to the truth of what would be implied under the "plus" paraphrase. lz Lycan suggests that the comparative-intensifying use can be subsumed under his analysis without spelling out the details of such an analysis. In light of our earlier examples I doubt if such a reduction will be successful. 13 Consider also "Evans was well-published even before he got his Ph.D." 14 From an NPR report on drug-related executions in Malaysia. ,
THE
QUANTIFIER
THEORY
OF EVEN
599
(15) and (16) do not occur over a range of time but rather take place at a particular moment in time. As result, it makes no sense to think that even entails quantification over a time interval, even though the focus of even in these examples is a range of time. We should find it rather unnatural to think that anyone who asserted (15) would intend the implication that Evans kissed Mary continually, i.e. at each moment during a particular interval, commencing at some time before he learned her name, ending at some moment after he got to know her name. Quite analogously, universal quantification appears impossible for (16) altogether. Executions, we know, take place at a particular moment in time. These counterexamples might be less damaging than they seem at first blush. The ordering of events that takes place as a result of the use of before and after allows a certain comparison. If we think of before as roughly equivalent in meaning with at an earlier p o i n t in time, it follows naturally that even before means something like even at an earlier p o i n t in time or at an even earlier p o i n t in time. This sort of counterexample, Lycan could now argue, will disappear once we understand what even earlier than or even before does for a sentence. H e had discussed the following apparent counterexample: (17)
Bill is even taller than John.
Following Bennett, Lycan points out that (17) is ambiguous. It may be taken to imply that, among other things that are true of Bill - e.g., that Bill is smarter, more handsome, richer, etc. - Bill is also taller than John. This sense of even is perfectly consistent with the universal quantifier thesis: "Bill is everything that John is not." The other reading of even in (17), however, suggests a comparison among tall individuals which we can explicate as "While Harry is tall and John is very tall, Bill is taller yet." The second reading implies that both individuals are tall, while the first one does not. On the second reading it is not obvious how universal quantification can enter into the picture. But Lycan avoids having sentences like (17) imply universal quantification when he accepts even as a comparative intensifier. A proponent of the universal quantifier thesis could now argue that even with before and after functions as a comparative-intensifier in expressing a contrast with "Mike kissed Mary as soon as he knew her name", or " H a r r y kissed Mary on their second date", and so on. In the other example even intensifies a comparison with, e.g., "John was executed after his mother made a plea," or " H a r r y was executed after the ambassador made a plea." As a result, we needn't worry about finding an acceptable paraphrase with a universal quantifier if that is the sort of reading we have in mind. But
600
PAUL BERCKMANS
the other reading, in fact, the one intended, requiring the other sense of e v e n , is still possible for (15) and (16). We are left no choice but to postulate the original sense of e v e n , and hence, universal quantification. 15 A different way to attack my proposal that (15) and (16) are inconsistent with Lycan's thesis is to paraphrase them as (15') and (16'): (15')
(16')
Evans kissed Mary before any of the personal-relation-establishing things had happened between them, even his learning her name; He was executed after all the expectable things were done to save him, even the queen's pleading for his life to be spared. 16
In (15') "even before he knew" is understood as "before he even knew," whereas in (16') "even after the queen" is understood as "after even the queen." The idea is thus that the object of e v e n is not time. These readings of (15) and (16), I submit, seem quite plausible. But I don't think they're quite adequate as paraphrases of (15) and (16). In (15), no implication is present that a n y of these things ever took place. It appears that (15) is assertible when only a few of these personal things had happened. It doesn't imply that n o t o n e of those things happened, as expressed by (15'). Similarly, (16) leaves open the possibility that some things did not happen, e.g. a plea from the Pope. This possibility is excluded by (16'). 17 1 see no plausible resolution, thus these examples are left unaccounted for.iS One could perhaps show that even in these examples resembles the archaic emphatic use as that of even in such locutions as "He, even the k i n g , . . . . " I will not attempt to explore this avenue (which, I think, will fail in any case), since I will offer my own account below. Notice also that a restricted nonliteral reading of a paraphrase with a universal quantifier will not do here. Such paraphrases appeared acceptable for some of the examples I advanced against Barker, but here, with before and after, that is not a likely solution. 16 These examples were suggested by a referee of Linguistics and Philosophy. 17 This proposed move doesn't always seem to work. Consider: 15
(a) (b)
Even after the ground attack was in full swing, many Iraqis didn't believe the land war had actually started; They kept fighting even after a cease-fire was called.
If we were to take (a) and (b) to mean (a') and (b') respectively, an incorrect result follows: (a') (b')
After even the ground attack was in full swing many Iraqis didn't believe the land war had actually started; They kept fighting after even a cease-fire was called.
In (a') the implication appears that other attacks bad taken place prior to the ground attack. This implication is not present in (a). In (b') (which appears anomalous in any case) one could read the implication that things other than a cease fire were called. These examples suggest that even is linked to before and after. 18 These before and after cases may indicate a more general failure of the universal quantifier thesis in the context of prepositional phrases referring to times or places. Compare, for
THE QUANTIFIER
THEORY OF EVEN
601
Some of the difficulties posed by enumerations did not go unnoticed by Lycan, who offers an interesting discussion of the following example: (18)
You have to be eleven or twelve or even thirteen to get your ears pierced.
According to his own universal quantifier thesis, we would have to paraphrase (18) as something as unlikely as (18'): (18')
Every age at which any of the relevant children is allowed by their parents to get their ears pierced, including age thirteen, is at least eleven and is required by some parent.
Lycan avoids this absurd entailment of (18) by defending the comparison intensifier view for that use of e v e n . 19 Indeed, a comparison is made between the relative ages at which children can have their ears pierced. Lycan adds that "the comparative-intensifying use of e v e n is of necessity comparative: at least two items must be mentioned and compared, one of which has a positive scalar property and the other has the same property to a greater degree. ''2° In contrast with the preceding enumerations where some scalar ordering is present, we may point to the following examples where such an ordering does not exist: (1%) (19b)
A vouvray, an Australian chardonnay or even a good vintage champagne would go well with scallops in creamy leek sauce; Carol Nelson, a basket weaver, teaches others to dye fabric and dried grass using the juices of beets, berries and even onion skins.
example:
(a) (b) (e) (d)
Roaches are a problem in my Florida vacation home. They're even in the refrigerator; I never felt the earthquake, and I was even in San Francisco when it happened; We found spaghetti sauce everywhere, even on top of the refrigerator; I didn't notice anything peculiar about her singing, and I was even on stage with her;
For (a) and (c) universal quantification is very plausible, but for (b) and (d) it seems to fail. There is also a reading of "We met even twice," i.e. not just once, which doesn't accept universal quantification easily. This may be the case too in "Even a year ago they were still enemies." I am not certain if generalizations can be made for a particular set of propositions, so I won't press the question here. ~9 Barker's solution, of course, would involve quantification over a range, made explicit in such a paraphrase as "Every age, ranging from eleven, to twelve, up to thirteen, is the age that is required by some parents to get their children's ears pierced." But I have suggested (and I will argue below) that such enumerations should be treated logically as disjunctions. 2o Lycan (1991, p. 24)i
602
PAUL
BERCKMANS
It had appeared that what made the comparative-intensifying use work in enumerations was the scalar ordering of the items in the enumeration, w i t h t h e last item in the enumeration, i.e., the object in focus of e v e n , being highest on a scale. With respect to (19a) (which was overheard in a posh wine store) however, it appears difficult to rank-order on a scale any of the wines or any of their relevant properties, nor can we place the item in focus of e v e n at the end of a relevant scale. In fact, there seemed to be a certain randomness to this statement, made by a salesman, as he glossed over racks filled with white wines, responding to a query concerning wines compatible with creamed scallops. In (19b), a newspaper caption commenting on a picture of an artist, the enumeration reflects no scalar ordering of substances but rather presents a loosely organized collection of items. The association between an element at the end of a scale and the object of e v e n , which had appeared to be an essential element of the comparative-intensifying e v e n , is missing. I want to conclude cautiously that since no scalar comparison is present - as in (18) - a comparativeintensifying use will be difficult to argue for. If that conclusion is correct, we are forced to take e v e n as implying universal quantification. The original version, however, is highly undesirable in light of certain well-known facts about food and wine and dyes. Neither a red bordeaux wine nor a red burgundy, for example, would go well with creamy scallop dishes, and not all substances are suitable for the purpose of dyeing, although these possibilities would be implied by the thesis. Notice that it is also difficult to think of the items in the disjunction as extremes of a range, thus rendering Barker's solution unfruitful. Lycan's "plus" account is helpful here, supplying this satisfactory paraphrase: (19a')
In addition to anything you would expect to go well with scallops in creamy leek sauce such as a vouvray, or an Australian chardonnay, a good vintage champagne would go well, too.
Although these disjunctions are thus not immediate counterexamples against Lycan's "plus" thesis, we are left with a rather unattractive division: e v e n in a scalable disjunctive enumeration should be taken as a comparative-intensifier to avoid an undesirable universal quantification reading, but, in an otherwise very similar sort of disjunctive enumeration, the absence of a scalable property makes the intensifier reading improbable, thus requiring the "plus" reading. I am not certain if we should be willing to accept such different accounts of e v e n for such cases. A final sort of counterexample against the "plus" account involves the use of e v e n in such examples as (20):
THE QUANTIFIER
(20)
T H E O R Y OF EVEN
603
Some of the most loyal White House officials weren't with the president on this one, even Sununu and Baker weren't. Scowcroft, of course, stood beside his leader.
" S u n u n u " and " B a k e r " are obvious instantiations of an explicit existential quantifier. It would be rather odd, however, to suggest that "Sununu" and " B a k e r " are, somehow, instantiations of a universal quantifier which on Lycan's " e v e r y . . . including" or " e v e r y . . . plus" thesis should be implicitly present. Such a view would only work if we thought that a universal quantifier would range over a subset of the class associated with the existential quantifier: (20')
Some of the most loyal White House officials weren't with the president on this one, and of all those that one expects to be with the president, none were with the president, p l u s Sununu and Baker. Scowcroft of course, stood beside his leader.
Such a move would require a great deal of logical maneuvering and seems rather implausible.
IV.
APPOSITION,
ENUMERATION
AND EXISTENTIAL
QUANTIFICATION
Thus far, I have discussed a number of difficulties for Barker's universal instantiation hypothesis which can be resolved if we accept the existence of a comparative-intensifying use of e v e n , along with a modification like Lycan's "plus" account. Of the three types of counterexamples I advanced against Lycan's thesis, the cases of e v e n b e f o r e ~ e v e n a f t e r , and the case in which e v e n indicates an instantiation of an explicit existential quantifier cannot promptly be remedied. My objection tO the intensifier use of e v e n for disjunctive enumerations was based on the observation that some of these enumerations involve no scale. I am willing to accept that one can perhaps hear these enumerations in a scalar way, although my own intuitions tell me otherwise. If I am right about these cases, however, I will need to explain what e v e n does in such disjunctions. One case of interest in which the object of e v e n indicates an instantiation of an explicit existential quantifier suggests an interesting relationship between existential quantifiers and e v e n . In our attempt to understand the meaning of e v e n , we have thus far looked at the relationship between the object in focus of e v e n and a possible universal contrast class. But it might be a good idea to think about the sorts of beliefs and intentions a speaker is thought to possess when he uses an e v e n sentence. It will be important
604
PAUL BERCKMANS
to take into account the speaker's state of mind and other information about the speaker to cast doubt on the hypothesis that universal quantification is consistent with these psychological facts. Let us look in some detail at the following examples. Imagine a conversation between Mr. Evans and his fifteen-year-old daughter who is about to go out wearing an outrageously skimpy outfit. Mr. Evans forbids his daughter to go out dressed that way, adding (21): (21)
Even Cher wouldn't wear that!
What prompts Mr. Evans to utter (21)? He must have some beliefs or thoughts about the kind of person that would wear such outfits, and he presumably has some relevant beliefs about Cher to assert (21), e.g., that Cher is the sort of person one ordinarily expects to wear extravagant outfits. It would be entirely appropriate for Mr. Evans to utter (21) having no knowledge or making no presumption at all about an entire class of individuals that can be expected to wear such outfits. It might have been the case, for example, that Mr. Evans, who rarely watches T V and knows little about what goes on in pop culture, happened to see a TV show featuring Cher. But had he seen a show featuring say, Madonna, a comparison with Madonna, not Cher, would likely have been made. In such circumstances the use of e v e n does allow the speaker to make a relevant comparison with a contrasting individual without implying the ascription of any further properties to that individual and without implying his membership in a comparison class, or any knowledge thereof. 21 In the preceding example, the object in focus of e v e n is intended to bring to mind a certain kind of person, some of whose characteristics become the subject of comparison, and such e v e n sentences are intended to bring to the attention of the hearer an instance or token of a stereotype. How this stereotype gets instantiated will depend on varying psychological or social facts about the speaker. This relationship between a stereotype and one of its tokens can be paraphrased explicitly as "somebody like . . . . " In such cases, e v e n announces an instantiation of an existential quantifier, its use implying some form of existential quantification. If this is correct, a theory of e v e n that aims at stating truth-conditions must postulate an ambiguity as between universal quantification and existential quantification. Before looking at the details of this account, however, we should gloss 21 Paul Kay (1990), criticizingFauconnier's scalar view, argues that even requires the existence of only one less extreme item, not a scalarly ordered class or a universal class. This view followsBennett's. That this view can be reconciledwith my own will become apparent.
THE
QUANTIFIER
THEORY
OF EVEN
605
over a second example to verify for ourselves the initial plausibility of the hypothesis• Suppose Mr• Robards is charged six dollars for an Alabama Slammer which he ordered at Joe's Bar and Grill• In disgust, he says to the bartender: (22)
Even at Max's they're only $4.50!
In this sort of situation we need not presume on the part of the speaker any knowledge of all taverns in town or even all expensive ones in town. No more is required for the use of even than that the speaker think that Max's is ordinarily very expensive but not that expensive, and that he think that the hearer knows some relevant facts about Max's, suitable for the purpose of contrast or comparison. The speaker can assert (22) without implying that Max's is the most expensive place, thus serving as an endof-scale item, and without implying that drinks are cheaper everywhere else, including at Max's. Notice also that the context does not seem to determine any particular kind of contrasting universal class, which, for example, would be limited to the class of all bars known to both speaker and hearer, nor would we be warranted in presuming that any such class is tacitly assumed to be present by the speaker or hearer. What is striking about the use of even in these sorts of cases is its function as an indicator of a certain kind of thing, a stereotype or model. This use of even is sometimes made more explicit with the phrase "some • . . like . . . . " or "some individuals, for example . . . . " I paraphrase (21) and (22) as follows: (21') (22')
Somebody like Cher wouldn't wear that; At a place like Max's an Alabama Slammer is only $4.50;
or, more colloquially, (21") (22")
Cher, say, wouldn't wear that; At, say, Max's, an Alabama Slammer is only $4•50.
I want to make the case that "even x is P , " at least in these cases, does not mean " e v e r y o n e . . . plus (surprisingly) x is P " but "someone (surprisingly) like x is P . " To compare what (21) implies on Lycan's account versus on mine, consider paraphrase (21') and (21"): (21') (21")
Somebody like Cher, wouldn't wear that; Nobody you would ordinarily expect to wear that would wear it, plus Cher.
While the truth-conditions of these two paraphrases are radically different, sentences with explicit existential quantifiers and sentences with explicit
606
PAUL BERCKMANS
universal quantifiers share interesting similarities. In both paraphrases (21') and (21") the object in focus of even appears in apposition; the apposition indicator and the quantifier in the first appositive differ. It will be important to appreciate the logical, syntactic and semantic relationship between even and the items in apposition. It is obvious that a universal quantifier as well as an existential quantifier can be the first appositive in a sentence (and also such quasi-quantifiers as many, most, etc.) followed by a second appositive indicated by even or especially, for example, etc. This second appositive may refer to a member included in the class denoted by the first appositive. Lycan drew attention to the similarity between even and especially in an attempt to turn this similarity into evidence for his thesis. He remarked that operators other than even "indicate contrasts within the membership of a contextually indicated reference class, such as especially," adding these examples: 22 (23a) (23b)
Everyone was embarrassed, especially Ned; Ned especially was embarrassed.
The suggestion is that especially implies universal quantification over the reference class, just as even does. But this similarity between especially and even can be turned into evidence against his thesis. I do not doubt Lycan is right to think that these uses of especially and even imply a wider reference class, but I don't think it is conclusive that a universal quantifier should range over this class. We have already discussed example (20) - which exhibits this appositional structure - as a counterexample against Lycan's thesis: (20)
Some of the most loyal White House officials weren't with the president on this one, even Sununu and Baker weren't. Scowcroft, of course, stood beside his leader.
Here the apposition indicator announces prominent members of a wider class without universal quantification. To generalize, contrast: (24a)
The honesty of some of our leaders - especially/even Prichard's was called into doubt. The honesty of all of our leaders - especially/even Prichard's was called into doubt. The honesty of many/quite a f e w / m o s t . . , of our leaders especially/even Prichard's - was called into doubt. -
(24b)
-
(24c)
These examples illustrate the capacity of even and especially to announce 22 Lycan (1991, p. 22).
THE QUANTIFIER
T H E O R Y OF
EVEN
607
a member of a wider class, no matter which quantifier ranges over that class. 23 It is clear that the appositive indicator announces only those members that are captured by the quantifier. In other words, in (24) even or especially announce only those members of the class whose honesty was called into doubt. These observations, I think, are so obvious that no controversy should arise over them. But what makes even and especially different from such apposition indicators as namely and f o r example and what makes them troublesome - is their ability to announce members of a contrast class when no reference to such class is explicitly made, as in (25): (25)
Even Prichard's honesty was called into doubt.
Such even sentences should thus be thought of as logically dependent on a sentence whose terms in apposition include a quantifier and a singular term. One aspect of the relationship between (24) and (25) may add an interesting perspective to the discussion. Any universally quantified scheme, we know, is equivalent to a conjunction of all of its universal instantiations; an existentially quantified scheme is equivalent to a disjunction of its existential instantiations. We could thus reformulate Lycan's thesis as saying that even sentences express surprise about one or more of the conjuncts in a truth-functional expansion of a universal quantifier. So, "Even Ted was sober" can be paraphrased as (26a)
Tom was sober, Dick was sober, Harry was sober, as expected, plus Ted (a surprise).
E v e n would thus flag an unexpected element in a truth-functional expan-
sion of a universal quantifier. This truth-functional expansion may sometimes show up explicitly, with or without a quantifier in opposition: (26b) (26c)
Everybody you would expect to be sober, i.e. Tom, Dick, Harry, they were all sober, even Ted was. Tom, Dick, Harry, even Ted was sober.
23 Barker explicitly rejects the proposition that e v e n can be used in combination with an existential quantifier. H e gives as an example a situation in which someone who is asked whether anyone passed a difficult test says "Well, some passed but they were an odd collection. The class brains Smith and Brown passed and even last year's dunce, George." Someone else replies "Even George passed." Barker (1991, p. 9). Barker thinks the last e v e n sentence is odd. I don't think it is odd at all, but even if it were, I doubt if that fact would invalidate the apparent assertibility of such sentences as (20) and (24)(a), and thus the general proposition that e v e n can be associated with existential quantification.
608
PAUL BERCKMANS
The realization that e v e n announces an item in a truth-functional expansion of a quantifier will help us explain the odd surface structure of some examples discussed earlier, especially those which involve disjunctive enumerations. Having excluded universal quantification for these examples, it seems natural now to treat them as truth-functional expansions (partial or complete) of an existential quantifier. Recall Lycan's original example of a disjunctive enumeration, which made him postulate the comparative-intensifying use of e v e n : (18)
You have to be eleven or twelve or even thirteen to get your ears pierced.
The disjunction in (18) follows the natural pattern of a truth-functional expansion of an existentially quantified scheme that can easily be made explicit: (18")
Some advanced ages exist before which children are not allowed by their parents to have their ears pierced: eleven, twelve, or (and this is quite outrageous) thirteen.
The existential quantifier in apposition does seem to explain why the enumeration is a disjunction, and it explains why e v e n turns up at the end of the sentence when we think of the enumeration as representing a scalarly ordered expansion of the quantifier. The assumption that e v e n flags an item in an expansion of an existential quantifier in these examples is more attractive than the suggestion of e v e n as a comparison intensifier, especially since it also accounts for my own counterexamples with disjunctive enumerations: (19£')
A variety of really different wines would go well with scallops in creamy leek sauce: a vouvray, an Australian chardonnay, a good vintage Champagne (and that may come as a surprise).
It is thus not necessary to invoke two different lexical uses of e v e n for these sentences, one for disjunctive enumerations where scalar comparison - and thus the intensifying use - is possible, the other for uses requiring the "plus" theory. Both types of examples can be explained by the hypothesis that an existential quantifier is entailed. And one more similarity between e v e n and other appositive indicators becomes apparent, Insofar as e v e n is an appositive indicator, it often functions as a device to announce the end of an enumeration, just as f i n a l l y may indicate the last term in an apposition. E v e n thus fulfills a dual role when used in such enumerations: in addition to carrying an expectation-contravening connotation, it signals the final term in an apposition. To put it in Fauconnier's terms, when no
THE
QUANTIFIER
THEORY
OF EVEN
609
scale is present even announces only an end-of-enumeration item, but when a scalar ordering exists even also signals an item comparatively highest on a scale. Let us attempt the following set of conditions for any even sentence S (where Se is an existential generalization of S*): (i) (ii) (iii)
S* and Sj are elements of a truth-functional expansion of an explicit or implied S, or Se; the truth of S* would be more surprising than the truth of Sj; There is a contextually determined universe in which at least S* and Sj are satisfied.
This explication is meant to be general enough to account for the ambiguity between the existential reading and the universal reading. Aside from the obvious benefit of resolving the counterexamples which I advanced, the main attraction of this account lies in its ability to unify a number of different theories which initially enjoyed enormous plausibility. We can adopt the Barker/Lycan idea of a contrasting class and a quantifier while also accommodating examples advanced by Bennett and Kay that required the existence of only one contrasting item. While universal quantification cannot be sustained as a single rule, we can still formulate a general principle for the assertibility of even sentences and their truth-conditions: "even announces an unexpected element in a truth-functional expansion of an existential quantifier or a universal quantifier." We are also in a position to drop Lycan's "plus" account, which seemed unaesthetic from the start. The result is a favorable similarity between the existential paraphrase of an even sentence and a universal paraphrase of an even sentence, both of which will contain minimally an apposition whose first term is a quantifier and whose second term (or nth term) will contain a member of the class over which the quantifier ranges, announced by including (for the universal reading) or namely, like, etc. (for the existential reading). My hypothesis takes care of the other counterexamples as well. First, consider the examples with even before and even after that demanded universal quantification on Lycan's account. That situation did not look promising. But if we take even in such sentences as an indicator of an instantiation of an existential quantifier, everything falls into place. In these examples even before and even after are then paraphrased as "at some surprising time, namely":
(15') and
John kissed Mary at some surprisingly early time, namely the time before he knew her name,
610 (16')
PAUL BERCKMANS H e was executed at a time when we would have expected leniency, that is, after the queen herself made a plea to spare his life.
It is natural to use a relative clause-indicator like n a m e l y , or t h a t is, in such e v e n sentences to bring to mind a particularly m e m o r a b l e property of a certain kind of event or state of affairs, or, in these examples, a m o m e n t in time. To indicate the unusualness of an event, the speaker brings to the attention of his audience a characteristic sample of the sorts of circumstances under which the event takes place. 24 The examples for which Lycan suggested the comparative intensifying use, i.e., disjunctive enumerations, can be given a reading with an existential quantifier too. But not all apparent intensifying uses of e v e n can be reduced to a quantifier. E v e n in my own example (7) " T h a t ' s even worse," cannot readily be paraphrased with a quantifier, universal or existential. This suggests that Bennett is right when he identifies these two different senses of e v e n .
V. CONCLUSION My claim that the original use of e v e n is ambiguous as between a universal quantifier and an existential quantifier leaves a n u m b e r of questions unresolved. An account of e v e n which postulates such ambiguity should offer some insight into the factors which determine an appropriate reading. Let us note, however, that questions of quantificational ambiguity are not unique to such particles as e v e n . They often arise with the use of possessive noun phrases, mass nouns, and singular and plural indefinite descriptions, to name only a few instances. Consider, for example, how the noun phrase "second-and-third-year law students" gives rise to an existential/universal ambiguity: (27) (28)
Second-and-third-year law students are employed each year by the College of Law faculty as research assistants, Second-and-third-year law students are required to take at last three credits per t e r m in other graduate programs.
24 I expressed my reservations about the suitability of Lycan's "plus" account to deal with such examples as (3a) '°The denazification was organized so sloppily initially that even Eichmann and Mengele escaped." Once we think of Eichmann and Mengele as examples of the sorts of people one would have expected not to get away, my own paraphrase, i.e., '*The denazification was organized so sloppily initially that - contrary to what one would expect - some notorious war criminals like Eichmann and Mengele escaped" seems to describe appropriately what a speaker has in mind when he utters (3a).
THE Q U A N T I F I E R T H E O R Y OF E V E N
611
In (27), I think, the intended reading is one which supports an existential quantifier, paraphrased as "a number of second-and-third-year law students." This can be contrasted with (28), which appears to require a universal quantifier. Important determinants of an interpretation wilt undoubtedly include contextual features. Semantic properties of the components of the even sentence, too, may force a particular reading, as indicated by the apparent difference in modality. In (28) the functor is required (indicating necessity) appears to demand a universal quantifier. Similar modal differences may appear to determine how an even sentence should be understood. While I argued that universal quantification in "He must be Harvard, or even Yale" didn't appear appropriate, it does seem right if we change the example to "He could be Harvard, or even Yale." If we take this ambiguity of even seriously, it is best studied not just in light of the intensional properties of even sentences but especially against the background of a discussion of questions of ambiguity in quantificationY
REFERENCES Barker, Stephen: 1991, 'Even, Still, and Counterfactuals', Linguistics and Philosophy 14, 138. Bennett, Jonathan: 1982, 'Even if', Linguistics and Philosophy 5, 403-18. Fauconnier, Gilles: 1975, 'Pragmatic Scales and Logical Structure', Linguistic Inquiry 6, 35375. Kay, Paul: 1990, 'Even', Linguistics and Philosophy 13, 59-111. Lycan, William G.: 1991, 'Even and even if', Linguistics and Philosophy 14, 1-36.
School of Law University of North Carolina Chapel Hill, NC 27514 U.S.A.
22 Much of this paper has benefitted from discussions with Bill Lycan and my wife Sheree Berckmans. I am indebted to two referees of Linguistics and Philosophy and especially to Bill Lycan for their invaluable comments on an earlier draft of this paper.