Exclusive Company: Only and the Dynamics of Vertical Inference ...

Journal of Semantics 13: 1-40 © Oxford University Press 1996 

Exclusive Company: Only and the Dynamics of Vertical 

Inference 

LAURENCE R. HORN 

Yale University 

Abstract 

The semantics of only says this: it asserts that no proposition from the set of relevant contrasts 

C other than the one expressed by its sister sentence a is true. There is in addition an 

implicature that a is in fact true. There is an industry devoted to the issue of whether the latter 

ingredient is an implicature (conversational or conventional), a presupposition, or part of the 

truth-conditions . . . For our purposes, we don't need to decide. 

(von Fintel 1994: 133) 

The goal of this paper is to find out whether for our purposes, we do. 1 

1 SYMMETRICALISM AND ITS DISCONTENTS 

The lion's market share in this industry has always been controlled by the 

conjunctionalists. For Peter of Spain and his fellow scholastics, (ia) entails the 

conjunction of (ib) and (ic) and thus entails each of them singly. 

(1) a. Only God can make a tree. 

b. God can make a tree. 

c. No one distinct from God can make a tree. 

The 'only' particle, solus or tantum, is an EXPONIBLE to be expounded or 

unpacked into 'an affirmative copulative proposition whose first part is the 

same proposition without only'—the 'PRAEIACENS ' or PREJACENT—'and whose 

second part is a negative proposition denying the predicate of all others apart 

from the subject' (Tractatus Exponibilium 21 ff, in Mullally 1945: 106-7). 

Peter's analysis has been co-sponsored or endorsed by a wide range of 

twelfth-, thirteenth-, fourteenth- and twentieth-century scholars, 2 from 

Burley (cf. Pinborg 198 r; De Ryck 1985), William of Sherwood (cf. O'Donnell 

1941; Kretzmann 1968), Ockham (1980: 133), Billingham and Alnwick (cf. De 

Ryck 1982) to Kuroda (1969: 348), Lakoff (1970), Taglicht (1984), van der 

Auwerea (1985), Keenan & Stavi (1986), Krifka (1992), Moser (1992), Burton- 

Roberts (1993), and von Fintel (1993). A related proposal is elaborated in two 

papers in this journal by Jay Atlas (1991, 1993); on Atlas's account, (2a) unpacks 

Downloaded from http://jos.oxfordjournals.org/ by guest on September 12, 2014

2 Only and the Dynamics of Vertical Inference 

not simply into the Petrine conjunction in (2b), but into the more complex 

representation of (2c). 

(2) a. Only Muriel voted for Hubert. 

b. Fa & ~3y(y # a & Fy) 

(e.g. 'Muriel voted for Hubert and nobody other than Muriel did' 

c. 3xVy[(x-y«Fy)&(Fy-y-a)] 

e.g. 'Exactly one individual, and no one other than Muriel, voted for 

Hubert' 

Since it is a consequent of his account, one he regards as a virtue, that (ia) entails 

(ib) as well as (ic), I'll regard it for now as a variant species of the same 

symmetricalist genus. 

But if only sentences have positive entailments, how do we account for the 

data in (3)-(5), which seem to demand an asymmetric theory on which the 

negative exponent as in (ic) is somehow more equal than its positive 

counterpart (ib) Note for starters that the pattern in (3) (see Horn 1969, 1970, 

1972) shows that the prejacent can be suspended with the help of an epistemic 

rider, while the exclusionary clause—like any true god-fearing entailment— 

cannot be. 3 

(3) a. Only Kim can pass the test, and it's possible even she can't. 

b. #Only Kim can pass the test, (and/but) it's possible that someone else 

can. 

c. #Everyone passed, but (for all I know/it's possible that} someone didn't. 

d. All the world is queer save thee and me, and even thou art a little queer, 

(attributed to Robert Owen on separating from his business partner in 

1828) 

e. All the world is queer save me and thee, and sometimes I think thee is a 

little queer. 

(attributed to 'an unidentified Quaker speaking to his wife') 

f. —Only Kim can pass. [us. #Kim and only Kim can pass.] 

—Yes, that's true. %And (what's more/in fact), maybe even she can't. 

Downloaded from http://jos.oxfordjournals.org/ by guest on September 12, 2014 

Notice the epistemic qualifier that must in general be present when suspending 

the prejacent: we can't normally say that only Kim can pass the test, and perhaps 

she can't. But the contrast between (3a, b) is undeniable, the latter as anomalous 

as any attempt to remove an entailment recognized by the speaker, as in (3c). 

Further, there are attested citations in which an exception is canceled without 

benefit of an epistemic rider, such as that in (3d), a variant of the Bartlett's 

familiar quotation in (3c). The exclusive counterpart of these exceptives would 

be something like 'Only you and I are sane and sometimes I wonder about you'. 

I find the exchange in (31") imaginable as well, although intuitions apparently 

differ here.

Laurence R. Horn 3 

As we shall see in more detail below, only subjects license negative polarity 

items (such as those boldfaced below) in the predicate, as in a classic example 

cited by Klima (1964:311) as evidence for the [+ affective] status of only. 

(4) Only young writers ever accept suggestions with any sincerity. 

Further, fronted only phrases (with an adverbial, PP, or NP object in focus) 

trigger inversion, a property that in modern English is confined to phrases of 

negative character. The examples in ($c-f) are lifted fromjacobsson (1951: 50- 

55), Jacobson (1964: 309-10), and Visser (1969: §1436). 

(5) a. Only in Montreal can die-hard film buffs expect to see not only one but 

three films from Mongolia. 

(New York Times article on the Montreal Festival des filmes du monde, 

13 August 1995) 

b. Only in New York can one seek shelter from a storm under takeout 

mussels in white wine sauce. 

(New York Times Metropolitan Diary, 20 July 1994) 

c. Only in stories does a dropped glass betray agitation. 

(Graham Greene) 

d. Only thus can the corruption of society be checked. 

e. Only one feature did I notice in the landscape, a large, white villa ... 

f. Only a year before he had built himself a mansion in Kensington. 

Smith (1922) notes the minimal pair in (se-f), where narrow-scope only fails to 

license inversion in the latter case, but this is the same point made by Klima 

(1964) and others with respect to the classic pairs in (5'): 

(5') a. Not even two years ago could you swim there. 

(—you couldn't swim there even 2 years ago) 

a'. Not even two years ago, you could swim there, 

(—you could swim there not even 2 years ago) 

(Klima 1964: 300) 

b. In not many years will Christmas fall on Sunday, 

b'. In not many years Christmas will fall on Sunday. 

(Klima 1964: 301) 

c. In no clothes does Robin look attractive, 

c'. In no clothes, Robin looks attractive. 

(sex-neutral adaptation of examples from C. Bird cited byjackendoff 

1972: 364) 

Further, inversion in the wide-scope contexts exemplified in (sa-e) is 

obligatory: 'In the course of fairly extensive reading in more recent literature, I 

have not come across a single example of straight order after only modifying a 

whole clause' (Jacobsson 1951: 54). 



In exhibiting the behavioral symptoms of downward monotonic expressions, 

only a phrases are for most speakers differentiated from the corresponding 

non-monotonic a and only a conjunctions. Thus compare: 4 

(5") a. Tonight and only tonight, the opalescent squid will mate. In a frenzy 

they seize each other, coupling again and again ... 

(Voice-over from 3-D IMAX movie Into the Deep) 

b. %Tonight and only tonight will the opalescent squid mate. 

In the light of these diagnostics, we might consider casting aside the 

symmetric conjunctive theory in favor of one of a range of asymmetric analyses 

which agree that the relation of (ia) to (ib) is not logical entailment while 

differing on just what if anything it is. On the analysis in (6), 

(6) Horn (1969): only (x — a, Fx) 

Presupposes: Fa 

Asserts: ~3y(y ^ a & Fy) 

whose silver anniversary the world recently celebrated, (7a) presupposes rather 

than entailing (7b), thus preserving the distinction between the negativeasserting 

only Muriel sentence and its truly conjunctive Muriel and only Muriel 

counterpart in (7d). 

(7) a. Only Muriel voted for Hubert. 

b. Muriel voted for Hubert. 

c. Nobody distinct from Muriel voted for Hubert. 

d. Muriel and only Muriel voted for Hubert. 

On this view, the suspension, polarity, and inversion facts of (3)-(i) would 

follow from the premise that these properties are associated with the negative 

assertion or entailment of the only sentence. A number of subsequent analyses 

(e.g. Konig 1991; Barker 1993) have essentially followed the same line, which is 

also echoed in the otherwise rather divergent analyses of negative polarity 

phenomena in Ladusaw (1980) and Linebarger (1981, 1987). 5 

Some asymmetricalists have pushed the presuppositional envelope further 

into the pragmatics, deriving (7b) as a non-truth-conditional aspect of conventional 

meaning—a conventional implicature or pragmatic presupposition—or 

even as a conversational implicature. On such accounts, the positive component 

of only is not part of what is said but part of what is implicated in an utterance 

like (7a). Thus in Horn (1979), the positive component of meaning contributed 

by only is taken to be a non-truth-conditional constraint on the appropriate 

assertion of (7a); the notion appealed to (with some reservations) is that of 

conventional implicature, a la Grice (1975) and Karttunen & Peters (1979). 6 The 

same position is endorsed in Rooth (1985), Krifka (1993), and Fretheim (1995). 

More radically, the positive component can be taken to represent not a 



conventional but a conversational implicature. McCawley (1981: 227) is the 

first to argue explicitly that the use of (7a) conversationally implicates (7b): if the 

speaker knew that nobody voted for Hubert, it would be a violation of the 

Quantity maxim (Grice 1975) to assert (7a) (cf. Horn 1992 in defense of this 

view and Atlas 1993 m opposition). 

The most asymmetric theory of all would of course be one on which the 

logical form of (7a) is just that of (7c), with the positive component excised. This 

is in fact Theory G, the line urged by Geach (1962: 187) in Reference and 

Generality on which no relation whatever obtains between the only expression 

and its positive counterpart. Thus 'F (some a)' is not deducible from 'F (only a)' 

either as an entailment or as a non-truth-conditional aspect of conventional 

force; nor is there any pragmatic inference involved. Geach's argument from 

logical convenience—'It is formally much more convenient to treat the 

exclusive proposition as having precisely the exclusive force of its supposed 

second [negative] component'—appears to fly in the face of intuition, entailing 

as it does that '"F (only 0)" will thus be true when "F( )" is true of nothing at 

all', rendering Only Hilary could reform health care vacuously true. 

2 MORE EVIDENCE FOR ASYMMETRY: THE CASE OF 

ONLY IF 

Before inspecting these asymmetricalist approaches for flaws, it's worth noting 

that the distribution and meaning of only if reinforce the conclusion that the 

positive component of only clauses is not simply entailed, or else A only ifB 

would be equivalent to A if and only ifB. As McCawley has observed, A only ifB 

is better paraphrased by 'Not A if not B' than by 'If A then B' (or by 'B if A'); he 

points out that (8a) is equivalent not to (8b) but rather to (8c) (McCawley 1981: 

49-54; cf. also McCawley 1974 and 1993: 81-8 and, independently, Barker 

I993)- 7 

(8) a. You're in danger only if the police start tapping your phone. 

b. If you're in danger the police start tapping your phone. [^(â)] 

c. If the police don't start tapping your phone you're not in danger. 


Crucially, as observed by Geis (197 3), van der Auwera (1985), Lycan (1991), and 

von Fintel (1994) in addition to McCawley and Barker, only if is compositional, 

its meaning a function of the meaning of only plus the meaning of if. Following 

Geis and McCawley, we can read only ifB as 'under no circumstances other than 

B\ in the same way that only on Sundays is read as 'on no days other than 

Sundays', so that (9a) amounts to the negative conditional in (9b). On the related 

account of Lycan (1991: 126-7), (9 a ) i s analyzed into (9c).


(9) a. I'll go only if you go. 

b. I won't go if you don't go. 

c. I'll go only in (the event that/events in which} you go. 

(— I'll go in no event other than one in which you go) 

And so, not surprisingly, we have the parallel suspension and cancellation 

effects via the epistemic riders of the type of (10): 

(10) a. He goes to church only on Sunday, and not even then if there's a football 

game on television. 

b. He'll go to church only if you do, and not even then if... 

c. I'll go only if you do. And maybe not even then. 

In either case, a switch to an overt conjunction yields a contradiction with the 

cancellation clause: 

(10') a. #He goes to church on Sunday and only on Sunday, and not even then 

if there's a football game on television. 

b. #He'll go to church if and only if you do, and not even then if... 

c. # I'll go if and only if you do. And maybe not even then. 

The inversion and polarity facts pattern accordingly: 8 

(11) a. If you build it, they will come. (. . .*will they come) 

b. Only if you build it will they come. (. . .*they will come) 

c. If and only if you build it, they will come. (. . .%will they come) 9 

b'. Only then will they come. 

(. . .*they will come) 

c'. Then and only then, they will come. (. . .%will they come) 

Note the contrast with non-negative adverbials (as in (11')) 

(11') Even (if you build it/then), they will come. 

(.. .*will they come) 

and with overtly conjoined a and only a conditionals (as in (1 ic) and (12)). 

(12) a. Only if George works hard can he ever hope to pass. 

(Barker 1993: 256) 

b. *If George works hard (he can/can he) ever hope to pass. 

c. If and only if George works hard (*he can/%can he) ever hope to pass. 


Curiously, the parallelism between if and only if on the one hand and Muriel 

and only Muriel on the other is supported by the conjunctionalist van der 

Auwera (1985 and personal communication, 14 June 1994), who argues that 

just as 'Horn's old example' (7a) is equivalent to (yd), so too if and only if is 

essentially just a stylistic/pragmatic/discourse variant of the semantically and 

truth-conditionally identical only if. In the light of the contrasting patterns we 

have observed here, that conclusion is ruled out and, given the parallelism with


'ordinary' only sentences, a conjunctionalist account of the latter is equally 

untenable. 10 

By way of summary we can offer a State of the Industry report, with a display 

of the competing theories on only ranked in order of increasing asymmetry: 



c. No one distinct from God can make a tree. 

Theory C/C A , the conjunction analysis: (ia) entails the conjunction of (ib) and 

(ic). 

Theory P, the presupposition analysis: (ia) entails (ic); (ia) semantically/ 

logically presupposes (ib). If (ib) is false, (ia) is neither true nor false. 

Theory PP, the pragmatic presupposition or conventional implicature 

analysis: as in P except (a) pragmatically presupposes/conventionally implicates 

(ib). (ib) is non-truth-conditional but semantic condition on the appropriate 

utterance of (ia). 

Theory M, McCawley's conversational implicature analysis: (ia) entails (ic); a 

speaker uttering (ia) normally (but not necessarily) conversationally implicates 

Theory G, Geach's pure negative analysis: (ia) entails (ic); no representation of 

any relation between (ia) and (ib). 

On the last four theories, the prejacent (ib) does not follow from (ia) by classic 

entailment; on the last three, it does not represent a truth-conditional 

inference, and on the latter two it has no semantic status whatsoever. 

3 THE SYMMETRICALISTS STRIKE BACK: POLARITY 

AND MONOTON1CITY 

In recent work, Jay Atlas has subjected the last three approaches to a withering 

critique on the grounds that we must of course reject any analysis that takes a 

sentence like Only Hillary could reform health care to be true—and merely 

misleading—on the grounds that nobody could reform health care. As (13a) 

shows, the putative prejacent implicature staunchly resists cancellation. 

(13) a. #Only Hillary could reform health care, and even she couldn't, 

b. #1 love only you, but (maybe) I don't love you either. 

Of my acknowledgement (Horn 1992: 182) that Theory M predicts the 

semantic coherence of (13b), Atlas retorts (1993: 314 and fn. on 317): 


[This] consequence strikes me as so outrageously counter-intuitive as to be a reductio of the 

theory. The idea that simultaneously / love only you could be true while / love you is false just


seems crazy to me ... I just have a very hard time with Horn's theory about the truthconditions 

oil love only you. 

Further, Adas denies that only NP subjects are monotone decreasing (or 

downward entailing), given the lack of any valid inference from (14a) to (14b). 

(14) a. Only Socrates entered the race. 

b. Only Socrates entered the race early. 

This observation actually dates back seven centuries to Peter of Spain, who 

notes (in the Tractatus Syncategoremata, Mullally 1945: 33) that no entailment 

goes through either way in (14): 

(14') a. y- Only Aristotle moves.-*\ 

b. ^»-Only Aristotle runs. -^ 

But Peter didn't have polarity or inversion to worry about. How does Atlas 

account for the downward or negative effects associated with only NP The 

suspension facts he disputes, along with the putative compositionality of only if. 

The inversion facts he does not acknowledge. And as for polarity licensing, 

Atlas maintains on the basis of the contrast between (15a) and the more 

marginal or unacceptable (15b-d) 

(15) a. Only John ever suspected David Alexander. 

b. *Only Bill wants Sam to finish the report until Friday. 

c. "Only Phil will give Lucy a red cent. 

d. Only I was all that keen to go to the party. 

that—contrast to popular opinion, including that of virtually everyone who has 

considered the issue, from Klima 11 to, amazingly, both Ladusaw and 

Linebarger—only NP does not in fact license NPIs: 

I do not find the syntactic observation that only triggers Negative Polarity Items sufficiently 

well-grounded, and so I do not believe that there is yet sufficient ground to take only to be a 

negative lexical item. (Atlas 1993: 31 3) 


In fact, though, popular opinion is correct, as the data in (16) show. 

(16) a. (Of all her friends,) Only Phil would lift a finger to help Lucy. 

b. Only your wife gives a hoot about what happens to you. 

(McCawley 1981: 83) 

c. My nose and my lungs are only alive at all because they are part of my 

body and share its common life. 

(C. S. Lewis, Mere Christianity, cited injacobsson 1951) 

Only NP is indeed a polarity licenser, although as a non-overt negator and a 

non-anti-additive quantifier in the sense of Zwarts (1986) and van der Wouden 

(1994), not every polarity item will freely occur within its scope. In this respect,


OM/y-initial phrases are quite parallel to another quantifier whose downward 

monotonic status is widely conceded, NPs of the form few + CN. 

The data in (17) review the non-discrete nature of polarity licensing; note 

that both semantic and morphological considerations come into play. 12 

(17) a. John didn't realize that 

he had swallowed any marbles. 

*Mary loved him anymore. 

*Ralph arrived until midnight. 

b. If 

c. 

you had ever done a thing like that... 

*John were here anymore ... 

*John had arrived until midnight... 

#Only Chris slept a wink last night. 

Nobody but Chris slept a wink last night. 

Nobody slept a wink last night but Chris. 

Whether or not it is a full-fledged downward monotonic operator on its 

own semantic merits, only NP patterns with the downward-entailers and in 

contrast with the truly non-monotonic NP and only NP, which triggers few if 

any polarity, inversion, or cancellation effects. And while the intuitions on (13) 

and (14) of Peter of Spain and Jay of California cannot be simply dismissed, it's 

worth reviewing some evidence, especially an elegant argument due to the 

thirteenth-century logician Walter Burley, suggesting that only phrases may in 

fact be more downward monotonic than they appear. 

For Ladusaw (1980: 165-6), the assumed polarity-triggering effect of only 

phrases follows from their downward monotonicity, as seen in (18): 

(18) a. Only Qohn] F had ever read anything about phrenology, 

b. Only John walks — Only John walks slowly. 

Recall that while upward monotonic quantifiers license inferences from subsets 

to supersets, as in (19a), downward monotonic quantifiers license inferences 

from supersets to subsets, as in (19b). Non-monotonic quantifiers, typically 

analyzable as conjunctions of quantifiers of mixed monotonicity types (e.g. 

exactly n — at least n + at most n) license inferences in neither direction. 


(19) a. Some Greeks voted for Bill. 

b. No Greeks voted for Bill. 

c. Exactly 756 Greeks voted for Bill. 

Some Greeks voted. 

No Greeks voted. 

Exactly 756 Greeks voted. 

(20) mont: some men, many women, Hillary, most Democrats, all linguists, everyone, 

the tallest 30-year-old, (at least) 10 cats. .. 

monl: no men, few women, at most 10 chickens, not every doctor, nobody. . .

io Only and the Dynamics of Vertical Inference 

As recognized by Barwise & Cooper (1981: 193), quantifiers with the same 

monotonicity orientation combine with and, excluding but, as seen in (21), 

(21) a man and three women few violins and no cellos 

some students and every professor no men and very few women 

most men and any FC women no dogs and hardly any NP i cats 

while monotonicity-discordant quantifiers most combine with but: 

(20) John {# and/but] no woman 

few women {# and/but) many men 

most of the dogs {# and/but) few of the cats 

(vs. most of the dogs [and/#but] a few of the cats 

Crucially, only phrases pattern as downward monotonic quantifiers: 

(23) many men {# and/but) only 3 women 

no men {and/#but} only 3 women 

4 CONVERSION AND EXISTENTIAL IMPORT 

But in fact only phrases must be downward monotonic, given the converse 

relation between only and all, recognized since the medievals: 

(24) a. Tantum animal est homo convertitur in istam: omnis homo est animal, per 

istam regulam: Exclusiva affirmativa convertitur in universalem P 

(John of Holland, in Bos 1985: 27) 

b. only (A, B) (only As are Bs) « all Bs are As ~ B c A 

The interdefinability of exclusives and universals appears in a variety of traditional 

sources, e.g. Peter of Spain's Tactatus Exponibilium (Mullally 1945: 106-7), 

and is exploited, defended, or assumed—often in the form of the subset relation 

in (24b))—by a bevy of moderns, including Lobner (1987), Chierchia & 

McConnell-Ginet (1990: 427), de Mey (1991), Johnson-Laird & Byrne (1991: 

128), Higginbotham (1994: 459-63), and von Fintel (1994, 1995), enough of a 

quorum to challenge the assertion that 'these days, those who adopt this analysis 

are few and far between' (von Fintel 1994: 13 5). 

Given this convertibility, to say that only Democrats supported Clinton is to 

say that all Clinton-supporters were Democrats. But, as has been recognized for 

a couple of millennia (cf. Horn 1989: Section 1.1.3 for summary and Moravcsik 

1991 for a new look), there is an existential inference, generally assumed to hold 

in the pragmatics, that is characteristically associated with the assertion of a 

universal. 14 Thus we can infer that (for all the speaker knows) there were indeed 

Clinton-supporters; otherwise the ^//-statement would be too uninformative 

to assert. 15 



(25) Only Democrats supported Clinton -*• All Clinton-supporters were 

/ Democrats 

(II) / ^ 

Democrats supported Clinton 

\^-— There were Clinton-supporters 

..There were Democrats who were Clinton-supporters 

But now we obtain the conclusion that there were indeed Democrats who 

supported Clinton, which is in fact the strongest positive proposition licensed 

by (25). Another way to put the same point is that it's just as true, but just as 

uncooperative, to assert (25) if you know that nobody supported Clinton as it is 

to assert that all John's children are bald on the grounds that John is childless. 16 

For a conjunctionalist, the unpacking of (25) yields the indefinite proposition 

Democrats supported Clinton, which as a stage-level predication (see Carlson 

1977) can be read existenrially. But, as von Fintel (1994: 139 ff.) points out, no 

such reading is possible when we shift to individual-level predicates. Thus (26a) 

does not entail (26b), which amounts essentially to the universal claim in (26c). 

The speaker of (26a) is at most committed to the weaker exitential claim that 

there are professors who are confident, i.e. to (26d). 

(26) a. Only professors are confident. 

b. Professors are confident. 

c. All professors are confident. 

d. Some professors are confident. 

Such cases thus provide a difficulty for the conjunctive analysis, even assuming 

the coherence of an entailment analysis for corresponding stage-level predications, 

as in (25) or in (26'). 

(26') a. Only professors are available, 

b. Professors are available. 

The point carries over, mutatis mutandis, from bare plurals to definite NPs in 

the focus of only. When Thomas Wolfe stipulated sixty years ago in his 

eponymous story that 'only the dead know Brooklyn', he wasn't committing 

himself to the claim in (27b), which seems to suggest that the whole ghostly 

crue—in particular Aristotle—knows Brooklyn. 


(27) a. Only the dead know Brooklyn, 

b. The dead know Brooklyn. 

Nor does (27a) entail the truth of a generic (rather than strictly universal) 

reading of (27b). Rather, the commitment is merely existential, as is clear from 

the opening sentence of the narrative: 'Dere's no guy livin' dat knows Brooklyn 

t'roo and t'roo, because it'd take a lifetime just to find his way aroun' duh f— 

town' (Wolfe 1935: 91).


Thus, a claim about only + CN is simply not equivalent to a claim about all 

and only the members of a given set, a fact rendered explicitly in (28a), 

excerpted from an Internet posting on lactose intolerance: 

(28) a. Only humans and not even all humans are genetically equipped to 

drink milk into adulthood, 

b. Humans are genetically equipped to drink milk into adulthood. 

The corresponding claim in the prejacent, (28b), is not interpreted existentially 

but universally or at least generically, and again represents a stronger commitment 

than is licensed by the assertion of (28a). 

Of course, an individual-level only predication may be consistent with the 

corresponding all and only statement, as in (29), but the point is that (29b, c) are 

not valid consequences of (29a), even though they happen to be true. What 

follows from (29a) is not (29b) or (29c), but (29d). 17 

(29) a. Only birds have feathers. 

b. All birds have feathers. 

c. All and only birds have feathers. 

d. All feathered things are birds. 

In recognizing the mutual convertibility of exclusives and universals, we 

obtain the equivalences of (30), preserving monotonicity orientations for the 

corresponding argument positions as indicated. 18 

(30) a. only As are Bs all BsareAs every (B)(A) 

*— DE environments —' 

non-DE environments —I 

Finessing the issue of existential import, only behaves like an upward monotone 

determiner that combines with a focus argument to form a downward 

monotone quantifier, while every /all is a downward monotone determiner that 

combines with a restrictor to form an upward monotone quantifier. We can 

represent the correspondence in (30) schematically by three-dimensionalizing 

the traditional Aristotelian square of opposition: 



only P is S 

AU 

only non-P is S 

E 

/ 

/ 

isP 

/ 

.'TT 

some S is P 

/ 

1 

P also is S 

not only non-P is S 

no S is P 

UE 

no 

/ 

some S is not P 

not all S is P 

/ 

/ 

O 

not < 

But it is far from clear that only is ever actually a determiner. Note that even 

if it is a determiner when it takes a common noun focus, as in (29a), this analysis 

fails to extend to the cross-categorial only operator that combines with or 

modifies full NPs (including proper names, descriptions, and pronouns), VPs, 

PPs, and adverbs. Note in particular that only combines with the in either order. 

only the N, the only N. Although the conjoinability of only with all in contexts 

like (29c) seems to testify to its Det-hood, all and only may not be a determiner 

here, as von Fintel (1995) has noted: cf. All and only the migratory birds. .. 

The category status ofonly/ CN is of theoretical interest in that if only is 

a determiner in this frame it stands as the unique counter-example to an 

otherwise striking generalization that all natural language determiners are 

conservative (Keenan & Stavi 1986). 

Conservativity is defined as in (31) and illustrated in (31'): 


(31) a determiner D is CONSERVATIVE iff [D(A)(B) ~ D(A)(A n B)] 

(31') a. No ravens are white «• No ravens are white ravens 

b. Most owners are greedy ** Most owners are greedy owners 

c. Only willows weep T 4 Only willows are weeping willows 

Only is not conservative because the biconditional in (31 'c) is only valid left-toright: 

only willows are willows that weep but not only willows weep. But if only 

is never a determiner, always combining with full NPs including those with 

null determiners as in (29a) and (31 'c), the problem for the conservativity 

universal is dissolved; see Thijsse (1983); Chierchia & McConnell-Ginet (1990:


426-8); van Benthem (1991: fn. 4); and von Fintel (1995) for arguments to this 

effect. 

De Mey (1991), on the other hand, argues that only is a determiner in 

sentences like (31 'c). Even if de Mey is essentially correct, however, there is 

another take on the question. 19 As van Benthem (1986: 8) observes, 

conservativity is motivated by 'the privileged role of the first argument in a 

determiner statement: it "sets the stage"'. But in an only sentence like (31 'c), it is 

not only plus its focus argument that 'sets the stage', as the principle of 

conversion shows. If (31) represents traditional conservativity, let us say that 

every natural language determiner is either traditionally conservative or NEO- 

CONSERVATIVE in the sense defined in (32) and illustrated in (32'): 

(32) a determiner D is NEO-CONSERVATIVE iff [D(A)(B) •• D(A B)(B)] 

(32') Only willows weep


In addition to universals, comparatives, and various overt and implicit 

negations (Alnwick and Billingham obviously knew their Fauconnier and 

Ladusaw), the set of terms with 'negative vim' (termini habentes vim negationis) 

crucially includes exclusives. 

Similar inferential mechanisms are found in Ockham and date back to 

anonymous authors of the late twelfth and early thirteenth centuries. The key 

idea for our purposes is that expressions with vis negationis block upward 

inferences and allow downward ones within their scope. Thus a position like 

that of the restrictor of universals licenses inferences from superior to inferior, 

from set to subset, if all humans are rational, then all women (but not 

necessarily all animals) are rational; but a position like that of the predicate or 

nuclear scope of universals licenses inferences in the opposite, upward 

direction: if all humans speak a complex language, it follows that all humans 

speak a language, but not that they all speak a complex Indo-European 

language. 

It seems to have been Walter Burley who first recognized that the inference 

patterns associated with omnis obtain in the corresponding (opposite) positions 

of tantum. The valid ser-subset consequences in the exclusive pair of (33a, b) 

matches that in the restrictor of the universal in (33c, d). 21 

(33) a. Tantum Sor[tes] videt animal. 'Only Socrates sees an animal' 

b. Tantum Sor[tes] videt asinus. 'Only Socrates sees an ass' 

c. Omnis homo est animal. 'Every man is an animal' 

d. Omnis albus homo est animal. 'Every white man is an animal' 

(Burley, 'De exclusivis': Pinborg 1981; De Rijk 1985; emphasis added) 

For Burley, the oddness of the downward inference from (33a) to (33b) in the 

case where the animal Socrates sees is an ox rather than an ass is no more 

problematic than the well-established downward inference from (33c) to (33d), 

which remains valid in the absence of white men. The issue of existential 

import is to be handled similarly in both cases. And while, as we saw, Petrus 

Hispanus and Atlas Pomonensis reject the inference from (34a) to (34d) (see 

(14), (14') above), Burley derives it by double conversion: (34a) converts to (34b), 

which licenses the downward inference of (34c), which converts back to (34d). 

(34) a. Tantum horn movetur. => b. Omne movens est homo. 

'Only a man is moving' 'Every moving [thing] is a man' 

u 

d. Tantum homo currit.


make a tree and the set of tree-makers is non-null, then the truth of the 

prejacent follows. No specific rule—entailment, presupposition, implicature— 

need be stipulated for deriving (35 b) from (35 a). 



c. No one other can God can make a tree. 

d. {x I x can make a tree} / 0 

On this view, we correctly predict that when 1 ask who's coming and you 

answer 'only John', all I must presuppose—all my model of the context must 

accommodate—is that someone is coming, not thatjohn is: (36c) but not (36d) is 

jiven'. 22 

(36) a. —Who's coming tonight 

b. —Only John (is coming). 

c. Someone is coming. 

d. John is coming. 

A neo-Burleyan approach also fares better than its market competitors in 

capturing the appropriate intuitions for the following context. In September 

1994,1 place a wager with you, staking $100 on the proposition that only Seattle 

will win more than 60 games in the upcoming regular season for the National 

Basketball Association. Now consider the following four possible scenarios on 

how season and bet play our, the last column reflects my sense of the proper 

resolution of the wager. 

(37) SCENARIOS Theories Theory P Theories Intuition 

C/C A 

PP/M/G 

Seattle wins 62 games; 

every other team wins < 60 W W W W 

Seattle wins 62 games, 

Orlando wins 61 L L L L 

No team wins more than 

60 games 

San Antonio wins 62, every 

other team wins < 60 

[=ACTUAL RESULTS] 


The first two scenarios are straightforward: when the prejacent holds, all 

theories correctly predict that I win the bet if the exclusion holds and lose if it 

fails. But only the neo-Burleyan theory allows us to distinguish the existential 

failure in the third case, where no accommodation is possible, from the mere 

falsity of the prejacent in the fourth, where accommodation of the existential 

proceeds unimpeded. 23


6 CONVERSION AND NPI LICENSING 

Now imagine a language just like medieval Latin, but with grammatical 

correlates of vertical inference, say the licensing of polarity items and inversion. 

Call such a language English. 

The interdefinability of only and all allows a semantic account of polarity 

licensing that avoids the empirical pitfalls of a strictly configurational account. 

For the construction grammarians Fillmore, Kay & O'Connor (1988: 529), nonsubject 

only is not a polarity trigger. Similarly, for Progovac (1993, 1994), only 

licenses NPIs only from Comp position as in (38a, b). On her analysis, the 

apparent subject only phrase in (38a) is itself raised to Comp at LF, whencethrough 

Spec-Head Agreement in Comp—it will license an abstract polarity 

operator which in turn licenses NPIs. Where no such raising to Comp is 

possible, as in (38c, d) only is not a trigger: 

(38) a. [n> Only Mary showed any respect for the visitors] 

b. [ CP Only to his girlfriend [ c did [jpjohn give any flowers]]] 

c. *John gave only his girlfriend any flowers. 

d. *John told Mary about any books. 

This analysis successfully distinguishes only from overt negation, the direct NPI 

licensor, whence the minimal pair in (39a, b); the latter contrasts with the NPIlicensing 

only phrase that appears in S-structure subject (LF Comp) position in 

(39c) (from Klima (1964: 311)). 

(39) a. His sister will not expect him to write any more novels. 

b. *His sister will only expect him to write any more novels. 

c. Only his sister will expect him to write any more novels. 

But Progovac's approach cannot account for the contrasts presented in (40)- 

(45), which show that non-Comp only can indeed license NPIs, provided that 

they are located outside the semantic focus of only. (The relevant NPIs are in 

boldface, and the focus of only in small caps.) 


(40) a. There was only ever ONE GLOVE there that I saw. 

(Los Angeles police lieutenant Frank Spengler, OJ. case court 

testimony, 16 March 1993) 

a'. I've only ever gone there ONCE. 

b. I only go there {SOMETIMES/*EVER}. 

(41) r.. I only eat any meat WHEN I'M DEPRESSED. 

b. I eat meat only WHEN I'M DEPRESSED ABOUT (SOMETHING/*ANYTHING) 

(42) a. Only ONCE have I ever been there. 

b. *Only EVER {do I go there/have I been there).


(43) a. Only WHEN I'M DEPRESSED do I ever eat any meat. 

b. *Only when I'm EVER DEPRESSED ABOUT ANYTHING do I eat meat. 

(44) a. He only bothered to write WHEN HE NEEDED MONEY. 

a'. *He only BOTHERED TO WRITE, [i.e. he never called] 

a". Only WHEN HE NEEDED MONEY did he bother to write. 

b. You only lift a finger to help WHEN YOU WANT SOMETHING. 

b'. *You only LIFT A FINGER TO HELP. 

b". Only WHEN YOU WANT SOMETHING do you lift a finger to help. 

(45) a. The only one who said a word was CHRIS. 

b. # Chris only SAID A WORD (literal only; no'NPI idiom'reading) 

The contexts in which NPIs are acceptable do indeed license downward 

inferences in the Burley-Ladusaw sense—if we ignore existential import. 24 

Thus if I eat meat only when I'm depressed, then it follows that in particular I 

eat pork (if I eat it at all) only when I'm depressed, but not that I eat meat only 

when I'm depressed about linguistics. 

Crucially, the argument of only that corresponds to the restrictor of a universal 

is a downward, superior-to-inferior-inference licensing, and NPIfriendly 

environment, regardless of its syntactic position. 25 And regardless of its 

morphology as well; thus, in (46b), where A is a singleton set, the relevant B 

position surfaces not as a standard restrictor-of-universal but is instead 

introduced by the determiner the only, which semantically amounts to the 

converse of adverbial only: 

(46) a. Only As are Bs ~ All Bs are As 

b. Only A (Bs/is a B) ~ The only {B/B'er} is A 

As expected, NPIs are licenced in this context, but not just any NPI. 26 

In the context of (47), we find pronominal all and the only thing occurring 

interchangeably as heads of an NPI-hosting relative clause. 27 

(47) a. The only thing he ever does anymore is complain, 

b. All he ever does anymore is complain. 


According to the line pursued here, the relation between an only predication 

and its prejacent—treated variously over the centuries as an entailment, a 

presupposition, a conventional or conversational implicature, and a Geachian 

ghost—is reconstructed as an indirect deduction from an existential premise. 

The linguistic correlates of downward entailment can be attributed to the 

convertibility of only predications and the corresponding universals, each 

downward-inferential in the corresponding (restrictor) argument. 

But while conversion is always possible, the strongest case for only phrases 

defining strict downward inferential patterns are those where only takes 

common noun or adverbial focus. De Mey (1991: 102-4) points to the crucial


role of the semantic type of OM/V'S focus in determining the content of the 

expression. He acknowledges that the non-semantic line he tentatively 

endorses for the existential proposition is most convincing for common noun 

subjects as in (46a), less so for proper names as in (46b), and least of all for 

cardinal foci as in (46c). 

(49) a. Only students read books. 

b. Only John slept. 

c. Only three pilots slept. 

Indeed, McCawley (1993: 587) himself, the progenitor of Theory M, acknowledges 

the acceptability contrasts in the exchanges of (50) and concedes that he 

is unable to account for this variation. 

(50) a. —Did only Southerners vote for Hubert 

—Yes, indeed only Lyndon did. 

b. —Did only Muriel vote for Hubert 

-{Yes/()No), indeed no one did. 

c. —Did only Muriel, Ed, and Lyndon vote for Hubert 

—Yes, indeed only Muriel and Ed did. 

You don't have to be an Atlas to hold up (50c) as anomalous. 28 As in (13) 

above, or the cleft of (51) (adapted from Frege's dictum), 

(51) It is only in the context of a sentence that a word has (any) meaning. 

the inference to the prejacent in such contexts strongly resists cancellation, yet 

here too the point of the only sentence is to assert the exclusion, whence the NPI 

licensing effect. 

7 BARELY HALF RIGHT-OR ALMOST HALF WRONG 


Scalar orientation, as reflected by polarity distribution and related diagnostics, 

is evidently parasitic on—rather than identifiable with—strict entailment-based 

downward monotonicity. This conclusion is supported by the behavior of the 

scalar adverbs almost and barely. The pattern in (52) indicates that almost is in 

some sense the more negative member of the pair. 

(52) a. s- The Celtics almost won the game. 

b. / The Celtics barely won the game, -v 

c. v*-*- The Celtics won the game.


(53) a. #He almost {budged/slept a wink/touched a drop/spoke to anyone), 

b. He barely {budged/slept a wink/touched a drop/spoke to anyone). 

This result is especially problematic for Linebarger's theory of polarity, on 

which an NPI is sanctioned in a given frame in the absence of overt negation if 

it occurs in a proposition p that entails or implicates some other overtly 

negative proposition whose truth in turn 'virtually guarantees' the truth of p: 

(54) The Negative Implicature (NI) account of NPI licensing 

(Linebarger 1991: 166; cf. Linebarger 1981, 1987) 

a. Basic constraint: 

A negative polarity item NPI contributes to a sentence S expressing a 

proposition P the CONVENTIONAL IMPLICATURE that the following 

condition will be satisfied in the context of the utterance: 

•P entails or implicates some proposition NI, which may be identical to 

P; in the LF of some sentence S' expressing this negative implicatum NI, 

NPI occurs in the immediate scope of negation. 

b. Conditions on the saliency of NI: 

i) The AVAILABILITY requirement: The speaker must be actively 

attempting to convey NI by the utterance of S. 

ii) The STRENGTH requirement: The truth of NI itself must virtually 

guarantee the truth of P. 

iii) The FOREGROUND requirement: If NI ^ P, the representation of NPI 

may not occur as background information in either P or NI, nor 

may NI itself serve as background to P. 

For (5 5a) the strength requirement is presumably not met by (5 5b), but it's hard 

to see how the truth of (55c) could fail to guarantee that of (55a), since the two 

are evidently truth-conditionally identical, as Atlas (1984) has observed. 

($5) a. He almost finished. 

b. He didn't finish. 

c. He didn't quite finish. 


But then (56a) should provide as happy a home for an NPI as (56c) does: 

(56) a. He almost finished {some/*any} of the essays. 

b. He didn't finish any of the essays. 

c. He didn't quite finish any of the essays. 

Clearly this is not the case. 

On the other hand, barely passes the tests for what we might call downward 

quasi-entailment, the facts in (52) notwithstanding. If he's barely alive, he's 

barely ambulatory, ignoring the case in which he's not ambulatory at all. But if 

he's almost alive, it doesn't follow that he's almost ambulatory; no downward


inference is licensed. If she's barely studied linguistics, she's barely studied 

syntax, but not vice versa. Again, almost yields an upward rather than downward 

inference. But not an upward entailment, at least not if the relation in (57) is 

semantic. 

(57) They almost won >- They didn't win 

In fact, as has been demonstrated in a number of studies (Horn 1972; Ducrot 

i973;Sadock 1981; Atlas 1984; Fillmore etal. i988;Lundquist&Jarvella 1994), 

it's not clear just what it is. For Sadock (1981), [almost p] is true iff p is true in a 

world not very different from the real world; it conversationally implicates 

(rather than entailing or conventionally implicating) —>p. On the related and 

considerably earlier account of Ducrot (1973), [presque p] has a positive 

assertion (p is 'non loin' from being true) and a negative presupposition (p is 

false), while [a peine p] has the reverse characterization and is in fact virtually 

identical to [presque non-p]. 29 What is clear from Ducrot's account and from 

the patterns in (58) and (59) is the constrast between the upward scalar orientation 

of almost and the downward orientation of barely. 

(58) The tank is 

almost half empty -*•[ 

U— barely y half full 

almost half full—»J~ 1/2 'j-«— barely half empty 

J 1/2 j 

(59) a. Good news—My printer is almost functional! 

b. Bad news—My printer is barely functional! 

If our tank in (58) is almost half empty (or barely half full), we'd better stop to 

fill it, even though it contains more gasoline than if it's almost half full (or 

barely half empty). Similarly, I take comfort from the report that my printer is 

almost functional and bemoan the news that it's barely functional, even though 

it's in the latter case that it actually works. 

In each case the point of the almost sentence is to assert a positive predication 

even at the cost of conceding a negative, while barely has just the reverse effect. 

Following Stalnaker (1974, 1978), I assume that an ASSERTION is a potentially 

controversial proposal to increment the context by adding the content of the 

assertion to the store of what is already presupposed, where a PRESUPPOSITION is 

a proposition whose truth is either already established in the common ground 

or can be ACCOMMODATED to it: 

The RULE OF ACCOMMODATION for presupposition (Lewis 1979: 340): 


If at time 1 something is said that requires presupposition P to be acceptable, and if P is not 

presupposed just before (, then—ceteris paribus and within certain limits—presupposition P 

comes into existence at (.


Note that even if the assertion of the upward-orientated (56a) semantically 

commits the speaker to the negative proposition in (56c), the latter is not part of 

the point of the utterance and thus plays no role in the computation of polarity 

licensing. 

The downward orientation of barely and only expressions follows from the 

dynamic content of these expressions, while their positive inference is at most 

an inert aspect of meaning to be accommodated. Converging evidence for this 

conclusion is provided by a variety of constructions, including the asymmetry 

in the scalar suspenders displayed in (60) (see Horn 1970, 1972, 1989: Ch. 4). 

(60) a. He's {barely/only} 16, if that, 

b. # He's almost 16, if that. 

Since Karttunen & Peters (1979), one standard technique for distinguishing 

what is (conventionally) implicated from what is said focuses on the interaction 

of the diagnosed material with the scope of higher illocutionary, attitudinal, or 

epistemic predicates. In the frames of (61), the prejacent—the proposition that 

Hillary trusts Bill—behaves as though it's outside the scope of what is said, or 

more specifically what is hypothesized, discovered, regretted, or asked. 30 

(61) a. If [only Hillary trusts Bill], all is well. 

b. I just discovered that only Hillary trusts Bill. 

c. It's too bad that only Hillary trusts Bill. 

d. I know Hillary trusts Bill, but does ONLY Hillary trust Bill 

e. #1 know nobody besides Hillary trusts Bill, but does only Hillary trust 

Bill 

In particular, as seen in the contrast between (6id, e), once the discourse model 

contains the exclusion, the only sentence is pragmatically deviant—whence the 

anomaly of (62). 31 

(62) #1 know that nobody besides Kim is coming, but 

I wonder if only Kim is coming. 

I just found out only Kim is coming. 

if [only Kim is coming], we'll be all right. 


The contrasts in (63) speak to the same point, in particular the distribution of 

but vs. and, reflecting contrastive vs. shared orientation of the connected 

clauses. The non-monotonic conjunctions in (63b, d), lacking scalar 

orientation, block contrastive but. (See Horn 1991 for extensive discussion of 

informationally redundant affirmation.) 

(63) a. Only Muriel voted for Hubert, but {she did/#nobody else did). 

b. Muriel and only Muriel voted for Herbert, but {#she did/#nobody else 

did).


c. —Only {Lee/3} passed the test. 

—Yes, but {Lee/3} did pass it. 

d. —{Lee and only Lee/3 and only 3} passed the test. 

—#Yes, but {Lee/3} did pass it. 

The patterns for almost and barely are parallel, as seen in (64): 

(64.) a. almost {but/# and} not quite (cf. Sadock 1981) 

b. The Knicks almost won the title in 1994 {but/#and} they didn't win. 

c. The Knicks barely won the East in 1994 {but/#and} they did win. 

d. The real is only the base. But it is the base. (Wallace Stevens) 

The same opposition emerges within the theory of argumentation 

elaborated by Anscrombre & Ducrot (1983: 20 ff.; cf. also Ducrot 1973; 

Lundquist & Jarvella 1994, and the cautionary comments of Konig 1991: 194, 

n. 5). In the frame of (65), the downward-asserting scalar particles in the family 

of a peine and seulement are rhetorically harmonic with the negative quantifier 

in the subject position and so exclude mais, while a contrastive connective is 

mandated for the rhetorically opposed upward-asserting presque: 

(65) a. Peu d'automobilistes depassent le 120, (a peine/#presque) 20%. 

'Few drivers exceed 120 [kph], (barely/#almost) 20%' 

b. Peu d'automobilistes depassent le 120, (mais (quand meme)) 

{presque/#a peine} 20%. 

'Few drivers exceed 120, (but (even so)) {almost/#barely} 20% did' 

Our analysis oionly bears a kinship with Israel's (1994,199 5) pragmatic scalar 

model for polarity phenomena and there are also some preliminary but 

promising results from psycholinguistic studies by Johnson-Laird & Byrne 

(1991) and by Moxey and Sanford and their colleagues (Moxey & Sanford 1986, 

i993;Sanford etal. 1994) supporting the distinction between what is asserted or 

more generally posed as against what is sup posed as inert material to be accommodated. 

Finally, we come around once more to Theory C A , Atlas's amendment to the 

conjunctionalist doctrine, on which (2a) is assigned the logical form in (2c). 


(2) a. Only Muriel voted for Hubert 

c. 3xVy[(x — y •* Fy) & (Fy — y — a)] 

e.g. 'Exactly one individual, and no one other than Muriel, voted for 

Hubert' 

It will be recalled that on his account (Atlas 1991,1993), the prejacent—here, the 

proposition that Muriel voted for Hubert—is entailed. Atlas insists that the 

prejacent is 'not just presupposed', supporting this position with an argument 

drawn from Taglicht that I see as somewhat equivocal. 

Atlas (1991: 134) endorses Taglicht's observation (1984: 88) that (66a) is 'hard


to reconcile with the claim that the positive proposition in on/y-sentences is not 

part of the assertion'. 

(66) a. Each of them stuck to one drink; Bill drank only whiskey and Harry 

drank only beer. 

b. Bill drank only whiskey. 

c. Bill drank whiskey and Bill drank nothing other than whiskey. 

Rather, to assert (66a), and thus to assert (66b), amounts to asserting the 

conjunction in (66c). 'Surely' Atlas concurs, 'there is something correct in 

Taglicht's observation. His example strongly suggests that Bill drank whiskey is 

not just presupposed.' 

Now it is clear that when (66b) is uttered within the enriched context of 

(66a)—where the common ground contains the proposition that Bill consumed 

exactly one (type of) drink—the conjunction in (66c) is indeed entailed, but this 

does not in itself guarantee that (66c) is semanrically entailed by the logical 

form of (66b). 32 

In any case, As Atlas insightfully observes, the fact that it is only topical NPs 

that induce presuppositions, combined with the fact that the focus of only 

cannot be topical, yields the prediction that the prejacent cannot be a presupposition. 

At the same time, he argues (1991: 139), 'in asserting Only a is F, we do 

not thereby assert a isF.. . What we do entails a is F, but it does not "say" it' 

(emphasis in original). 33 

The position on which Atlas alights is thus not entirely at odds with the one 

defended here, although we differ in many respects, some of which have 

already been touched on here. 34 In particular, while Atlas is certainly correct in 

claiming that (2a) does not 'say' that Muriel voted for Hubert, there is no reason 

to believe that it 'says' that exactly one individual did so either. Rather, we have 

argued, the assertion of (2a) requires an accommodation of the existential 

proposition that the set of voters for Hubert was non-empty, which—combined 

with the asserted exclusion—yields the inference that this set was the singelton 

{Muriel}. 

Further, as a referee notes, the representation proposed by Atlas fails to 

generalize naturally to cases of only with plural subjects. Indeed, it is not 

obvious how the logical form of (2c) extends to any instances of the cross-categorial 

operator that do not conform to the only a is F schema. 35 

Returning to Burley and friends, how do we explain the tension between the 

symmetric two-conjunct unpacking of only and the insight that only, as the 

converse of all, licenses downward inferences in the predicate Thus Ockham 

(Summa Logicae, Part II, Chapter 17) explicitly endorses both claims: 


• Every exclusive proposition has two exponents: one affirmative and the other negative. For 

example, 'Only a man is an animal' has these exponents: 'A man is an animal' and 'Nothing 

other than a man is an animal'. (Ockham 1980: 133)


• There is a valid consequence from an exclusive proposition to a universal proposition with 

the terms transposed, and conversely ... Therefore this consequence is valid: 'Only what is 

necessary is true; therefore everything true is necessary'. (Ockham 1980: 141) 

The scholastics' willingness to overlook their inconsistent insights on only 

reflects the standard analytical practice for a wide range of complex operators. 

In each case, the components of the analysis are in effect laid out on the 

refectory table and treated as conjoined, even where asymmetries would seem 

to emerge. This tradition can be traced back at least to Aristotle's modal logic, 

with his treatment of one sense of possible as a two-sided operator, akin to what 

we would now call contingency: if a thing may be, it may also not be (De Int., 

22b20, Pr. An., 32ai8-28; see Horn 1972, 1989 for discussion). While this 

analysis did not survive into the modal logic of the scholastics, analogous twosided 

treatments emerged for a variety of other operators and predicates. 

Thus, on Peter of Spain's account of exponibles (Mullally 1945: 104-29), 

time-axis predications like that of (67a) and comparatives like that of (68 a) are 

unpacked into the simple conjunctions of (67b) and (68b) respectively. 

(67) a. Socrates ceases to be white. 

b. 'Socrates is not now white and immediately before this was white' 

(Peter of Spain, Treatise on Exponibles, line 167 ff.) 

(68) a. Socrates is more white than the ass. 

b. 'Socrates is white and the ass is white but the ass is not equally as white 

as Socrates' 

(Peter of Spain, Treatise on Exponibles, line 229 ff.) 

For the Stoics, Boethius, Peter of Spain, and a majority of the late medievals 

(as documented in Bochenski 1961: 119-20, 197 and Ashworth 1974: 161-2), 

the principal or only reading of or is as a 'proper', 'complete', or 'strong' disjunction, 

one whose disjuncts cannot both be true: p orq, taken 'properly', amounts 

to the assertion that either p is true and q false, or p is false and q true. This too is 

essentially a conjunctive analysis, as on the standard definition of so-called 

exclusive disjunction: 36 


(69) p@q~(pvq)&~(p&q) 

Particularly interesting in this two-sided vein is Peter of Spain's treatment of 

'reduplicatives', causal connectives like inquantum 'insofar as' or secundum quod 

'according as'. After suggesting a perceptive analysis that incorporates the first 

explicit invocation of presupposition in the Western canon— 

The first rule is that a reduplicative word presupposes \praesupponit] a given predicate to inhere 

in a certain subject and denotes [JeMoJaf] that what it is directly attached to is the cause of that 

inherence. (Peter of Spain, Treatise on Exponibles, line 105, in Mullally 1945: 112-13)


—at crunch-rime, Peter once again retreats to a straightforward conjunctive 

analysis on which (70a) is simply unpacked into a conjunction of the clauses in 

( 7 ob-d). 

(70) a. Man in so far as he is rational is tearful. 

b. Man is rational. 

c. Man is tearful. 

d. Every rational [thing] is tearful. 

It is against these conjunctive analysf that the treatment of what we might 

call 'two-sided only' must be understood—and that of two-sided all as well. As 

we have seen, the traditional account of universals is as bilateral as the 

traditional account of exclusives; both Only A is B (Only As areBs) and EveryBis 

A (All Bs are As) are taken to entail the existence of Bfs) while excluding the 

possibility of any non-A B(s). 37 But even as the scholastics accepted the 

unpacking of Tantum/Solus homo est rationalis into a conjunction of'homo est 

rationalis' + 'nichil aliud ab homine est rationale', there are some suggestions 

that these two components may not be on all fours: 

It is asked why 'alone' [solus] is called an exclusive rather than an inclusive; for when someone 

says 'Socrates alone is running', Socrates is included under running but the others are excluded. 

It must be said that it is because the inclusion occurs not as a result of the force of the word but 

as a result of the statement as it is before the 'alone' is inserted into it. The exclusion, on the 

other hand ... does occur as a result of the force of the word ['alone']. (William of Sherwood, 

Treatise on Syncategorematic Words, XI.6, in Kretzmann 1968: 71-2; cf. also St. Thomas Aquinas 

(Pegis 1945) on only in Sutntna Theologica, la, q. 31, arts. 3, 4) 

As we saw above in our exploration of the whiskey and Zantac examples, the 

context in which only occurs may indeed entail inclusion, but not necessarily 

because of the presence of only. 

If the Williams of Ockham and Sherwood, Walter Burley, and their 

colleagues had only taken the trouble to consult Grice on implicature, Lewis on 

accommodation, and Stalnaker on assertion and presupposition, the reign of 

the conjunctive analysis might have been far shorter-lived and the industry 

devoted to explicating the relation between an only sentence and its prejacent 

counterpart might have emitted fewer pollutants into the scholarly atmosphere. 

Yet it must be conceded that the atmospheric conditions in the conjunctionalist 

heyday had charms of its own, as Ockham's Summa Logicae makes clear. 

After cataloguing various 'improper' senses of only, those which are taken with 

restricted scope ('no more than [within a fixed domain]') as opposed to the 

purely exclusive 'proper' sense, 38 Ockham (1980: 137) remarks that 


These are the senses, then, in which the exclusive expression can be taken improperly. And 

perhaps there are still other senses in which it can be taken improperly. But since they are not 

as widely used as the ones we have dealt with, I will leave them to the specialists.


A glorious picture indeed: monasteries crammed to the spires with specialists 

on only, labouring away on the fine points of the semantics of exclusive 

propositions. Those were the days! 

LAURENCE R. HORN 

Deft, of Linguistics 

Yale University 

P.O. Box 208236 Yale Station 

New Haven, CT 06520-8236 

e-mail: lhorn@yalevm.ycc.yale.edu 

Received: 04.08.95 

Revised version received: 28.11.95 

APPENDIX 

The licensor that shouldn't be 

I have concentrated in the text on trying to 

demonstrate that the non-focus B position in 

exclusive predications of the form Only A is 

B, Only As are B(s), corresponding to the 

restrictor of the corresponding universal in 

All Bs are A, Every B is A, is downwardinferential 

(although not necessarily strictly 

downward entailing). But since the nuclear 

scope of a universal is not downward inferential, 

we would expect the focus of only not to 

be either (see the cube of opposition on p. 13). 

And given that the former position does not 

license NPIs ("All my friends have ever eaten any 

squid), we would expect the latter not to. 

The first of these expectations is satisfied, 

but the second is not. Contra Ladusaw (1980: 

165-6), who takes only as well as only NP to 

define a downward entailing context, Linebarger 

(1981, 1987) shows that the first argument 

or focus position of subject-modifying 

only is not downward entailing; (Aia) does 

not entail (Aib). Inexplicably, this does not 

seem to bother the NPIs that frolic happily 

within this context, as (Aic) shows. 

(Ai) a. Only people who have had a debilitating 

illness themselves can appreciate 

what an ordeal this was. 

b. Only people who have had polio can 

appreciate what an ordeal this was. 

c. Only [people who have ever had a 

debilitating illness themselves] ... 

Indeed, the focus of only—if we are careful to 

make sure that the relevant material is in the 

focus—represents an upward entailing 

context, precisely as does the nuclear scope of 

universals. Thus there is no downward 

entailment from (A2a) to (A2b), but there is 

an upward one in the opposite direction. 39 

(A2) a. Only faculty can vote. 

b. Only F[tenured faculty] can vote. 

One more diagnostic that supports the 

ascription of opposed vertical orientations for 

the two positions in only predications is provided 

by the distribution of assertionnarrowing 

(and implicature-suspending) 

indeed or what's more constructions. 

(A3) Only people who have had a debilitating 

illness themselves can appreciate 

what an ordeal this was—(indeed/what's 

more/in factj, only people who've had 

polio can truly appreciate it. 

Similarly, if only [tenured faculty] can vote 

on a given matter, then a fortiori only faculty 

can vote on ir, correspondingly, we get the 

scalar patterns in (A4): 40 

(A4) a. Only faculty can vote on this matter; 

indeed, only tenured faculty can. 

b. #Only tenured faculty can vote on 

this matter, indeed, only faculty can. 



c. #Only tenured faculty can vote on 

promotions; indeed, only tenured 

faculty can vote on promotions to 

full professor. 

d. Only tenured faculty can vote on 

promotions to full professor, indeed, 

only tenured faculty can vote on 

(any) promotions. 

which on semantic grounds should be DE 

only in their nuclear scope. Thus if not every 

Catholic priest is a bachelor, it follows that 

not every priest is a bachelor, but not vice 

versa (the married and spinster priests may all 

be Episcopalians). But while the restrictor of 

not all, not every is a UE position, it also allows 

NPIs within a relative clause: 

Nor is Linebarger's discovery of the non-DEness 

of the focus of only a new one: the 

fourteenth-century tracts of Alnwick and 

BiUingham both cite Tantum homo currit; igitur 

tantum animal currit as a valid instance of the 

upward 'ab inferiori ad superius' inference 

(DeRijkio82:8, 51). 41 

What, then, explains the acceptability of 

NPIs in the upward-oriented environment of 

(Aic) I would maintain that polarity items 

are possible in this context not because of the 

semantic properties of only but in spite of 

them. 42 Given minimal pairs like those in 

(Aj)smd(As'), 

(A5) a. Only the students who had ever read 

anything about polarity passed, 

b. Only the students who hadn't ever 

read anything about polarity passed. 

(A6) a. Only the guests who had seen any of 

the suspects were questioned, 

b. Only the guests who hadn't seen any 

of the suspects were excused. 

there is no point in seeking to extend DE-ness 

to cover only as has been suggested for determiners 

like most and few (see note 39), for 

adversatives like sorry and glad (cf. Ladusaw 

1980; Linebarger 1987; Krifka 1990), and for 

the only and related superlative triggers 

discussed in note 26. The above pairs illustrate 

the untenability of the assumption, 

made explicitly by Krifka (1990: 174) and 

tacitly by others, that triggering proceeds 

algorithmically, with an NPI licensed just in 

case it falls within the scope of an odd 

number of downward entailing NPI licensors. 

43 

It's worth noting that another site where 

unlicensed NPIs show up is in the restrictor 

of determiners of the form not all, not every. 

(A7) a. Not all the guests who ate any of the 

contaminated squid became ill. 

b. Not everyone who has ever been to 

Groningen works on polarity. 

It appears that the downward or negative 

property of only or not all with respect to one 

argument position is somehow illicitly 

extended to the inappropriate (UE) argument. 

In any case, it must be recognized from the 

data in (4o)-(45) that not just any position 

within the focus of only licenses polarity 

items; the licensing is best when the NPI 

occurs within a restrictive relative within 

subject only phrases. Moreover, not just any 

NPI occurs in these environments. As seen in 

(A8), only the most permissive NPIs, any and 

ever, show up here, and in particular the minimizers 

that are licensed by the only (see note 

26) are blocked within the focus of only: 

(A8) a. Only those travelers who (get any 

sleep/#sleep a wink) on the plane 

will enjoy their layover in Amsterdam. 

b. Only (those) students who have 

ever smoked any hashish . . . 

#been to Europe in years.. . 

# missed so much as one question 

eaten {any/#all that much) squid 

The only student to 

#Only those students who 

move a muscle 

lift a finger 

drink a drop 

say a word 



Our focus-of-on/y NPIs are subject to 

another restriction, one whose exposition 

requires a bit of background. Heim (1984.) has 

observed a subtle contrast in polaritytriggering 

within the downward-entailing 

context defined by the restrictor of universals. 

She notes that the moderately restrictive 

minimizer NPIs in (A9) can only occur when 

there is perceived to be a meaning connection 

between the subject and predicate: 

(A9) a. Every restaurant that charges so 

much as a (single) dime for iceberg 

lettuce ought to be closed down, 

b. £very restaurant that charges so 

much as a (single) dime for iceberg 

lettuce actually has four stars in the 

handbook. 

This requirement is relaxed for sufficiently 

permissive NPIs: 

(A 10) a. Every restaurant that I have ever 

been to happens to have four stars, 

b. Every restaurant that advertises in 

any of these papers happens to have 

four stars. 

Now it appears that while the focus of only 

admits just the liberal NPIs any and ever, as 

seen in (A8), these items are constrained in 

our environment in the same way that the 

stricter minimizers are in Heim's, while the 

items semantically restricted in Heim's 

environment do not occur at all within the 

focus of only. The crucial distinctions are the 

one between (An) and (A 12) on the one hand 

and that between (Ai 1) and (A13) on the 

others. 44 

(Ail) a. Only those students who have any 

siblings need to complete the 

survey. 

b. Only those students who have ever 

been to Europe need to complete 

the survey. 

(A 12) a. #Only those students who have 

even a single sibling need to complete 

the survey. 

b. #Only those students who so much 

as glance at anyone else's sheet need 

to re-take the exam. 

(A 13) a. #Only those students who have 

any siblings happen to have passed 

the exam. 

b. #Only those students who have 

ever been to Europe happen to have 

passed the exam. 

Thus Heim's generalization applies in 

attenuated fashion for the environment 

defined by the focus of only, the declencheur 

malgrelui. 

One possible approach to the licensing 

properties of only (as suggested by a referee) 

would take only sentences like (A 14) to 

represent not the converse of the universal 

A14b) per se, but rather that of its doubly 

complemented counterpart (A 14c). 

(A 14) a. Only B's are A's 

b. All A's are B's. 

c. No non-B's are non-A's. 

This suggestion is consonant with 

McCawley's argument (1974 et seq) cited 

above for taking A only ifB to translate not if 

A then B but rather if not-B then nol-A. In 

addition, Johnson-Laird & Byrne (1991: 128) 

provide experimental evidence to support the 

claim that while Only the bakers are athletes and 

All the athletes are bakers have the same truth 

conditions, the two converses differ in the 

mental representations or models speakers 

assign to the two sentences, with the only 

versions processed as inherently more negative: 

The assertion containing 'only' makes 

explicit right from the start not merely 

that some bakers are athletes but also 

that anyone who is not a baker is not an 

athlete. Hence, the two models are 

equivalent in content, but the equivalence 

is not immediately apparent 

because the initial model for 'all' makes 

explicit just the affirmative information, 

whereas the model of'only' makes 

explicit both affirmative and negative 

information. 

Relating (A 14a) directly to (A 14c), rather than 

to the logically equivalent (A 14b), offers a 


3o Only and the Dynamics of Vertical Inference 

promising tool for exploring why only NP 

statements, beyond sharing downward 

orientation of universal restrictors, appear to 

be somehow more 'negative' than the corresponding 

all statements, while at the same time 

providing a new beachhead for the assault on 

the focus-of-on/y licensing problem. 

NOTES 

1 This paper has had several previous 

incarnations since its origin in Horn 

(1992: 178-83). I would like to acknowledge 

the audiences of these earlier 

presentations—at SALT II (Columbus) in 

1992, at Rutgers and CUNY in 1993, at 

Groningen in 1994, at the LSA (New 

Orleans), MIT, and the Ottawa Conference 

on Negation in 1995—for their 

observations, hints, and objections. I am 

especially grateful to Jay Atlas, Barbara 

Abbott, Jack Hoeksema, Bill Ladusaw, • 

Jim McCawley, Victor Sanchez Valencia, 

Kai von Fintel, and the two Journal of 

Semantics referees for their generous analytic 

and bibliographic suggestions and 

complaints. Given the analysis to be presented 

herein, I can safely assert that only 

the author is responsible for any remaining 

defects. 

2 I am indebted to Victor Sanchez Valencia 

for directing me to many of the medieval 

sources cited here and below. The earliest 

one either of us has been track down in 

the current connection is Abelard (1079- 

1142), whose analysis of (i) into the 

conjunction in (ii) appears in Tractatus III 

(Topica) of his Dialeclica (Abelard 1970: 

332-3)- 

(i) Socrates tantum est Socrates. 

(ii) Socrates est Socrates et nulla alia res 

est Socrates. 

3 Altmann (1977: 311-17) and Reis (1977: 

56-8) note similar suspension effects in 

German: 

(i) Nur Hans mag Bier, und moglicherweise 

nicht einmal er. 

'Only Hans wants beer, and possibly 

not even he [does]' 

(ii) Hans verdient nur 1000—und moglicherweise 

sogar noch weniger. 

'Hans earns only 1000, and possibly 

even less' 

4 The facts are actually somewhat more 

complex. For many speakers a and only a 

sequences do trigger inversion; for some, 

they license negative polarity items as 

well. Perhaps one relevant factor for the 

permissivists is that the syntactically 

closer conjunct is the negative one, so that 

the semantic non-monotonicity of the 

conjoined phrase may be overridden by 

the proximity of its negative component 

to the diagnostics in question. (The 

impossibility of Only XP and XP phrases, 

as discussed in Horn 1969, 1991, makes 

this hypothesis difficult to test.) If this line 

is correct, inversion in contexts like that 

of ($"b) is subject to an effect parallel to 

the role of the Closest Conjunct Principle 

(see Morgan 1972) in determining 

number agreement in English. It should 

be noted that even speakers in the liberal 

inversion and polarity dialects distinguish 

the suspension and cancellation properties 

of the prejacent of only a from that of 

the corresponding a and only a sequences. 

(See note 9 below for related considerations.) 

5 I read Thomas Aquinas (Summa Theologica 

la, q. 31, arts 3,4 in Pegis 1945: 311- 

14) as endorsing an asymmetric position 

on only, although I confess that with Peter 

of Spain, who later became Pope John 

XXI, figuring among the conjuncrionalist 

host, the presence of a saint on my side of 

the ledger would provide moral support. 

Be that as it may, Peter's analysis did 

precede his election to the papacy, thus, 

whether or not the conjuncrionalist 


doctrine is correct, there are no grounds 

to assume its infallibility. 

6 In the midst of a twelve-page exposition 

of the formal properties of their conventional 

implicature analysis for even, loo, 

and also, Karttunen & Peters (1979: 32) 

provide the following characterization of 

only, reproduced here in its entirety: 'The 

case of only is more complicated, 

although it also involves a distinction 

between focus and scope.' It should be 

clear from the current presentation that I 

can sympathize with their diffidence. 

Indeed, I seek here to bite off only a small 

corner of the only nur, for comprehensive 

cross-linguistic expositions of the synchronic 

and diachronic properries of the 

form and meaning of exclusives and other 

focus particles, see Konig 1991 and 

Nevalainen 1991. 

7 Based on the behavior of counterfactual 

conditionals, McCawley (1993: 546) ultimately 

determines that A only ifB is best 

paraphrased by 'Not A unless B\ The difference 

between the two versions of his 

analysis need not concern us here. 

8 Lycan (1991: 133) actually takes inversion 

in cases like (1 ib) (and presumably in 

those like (5)) to constitute a negative 

polarity item, 'caused by the not within 

only', but, given the difference in patterning 

between the diagnostics mentioned 

above, it seems safer to take these two 

phenomena as related but distinct. 

9 Not everyone rejects inversion and NPIs 

in the (1 ic, c') contexts; the caveat of note 

4 applies here as well. A referee contributes 

this attested example from Doonesbury: 

[Text of letter] Dear Guys, 

How has NAFTA affected the strip's 

production—S. B., Washington, D.C. 

[Mike and Zonker's response:] 

Excellent question, S.B. The strip is still 

written in American, but the art is 

completed by Mexican cartoonists who 

dream of a better life ... 

Next, the final inked drawings are 

shipped to Canada, where each piece is 


signed ... And carefully checked for 

spelling! Then—and only then—are the 

strips rushed to your dairy case! (Trudeau 

1994; ellipses and boldface in original) 

Note the contrasrive emphasis on the 

negative (only) conjunct; evidently some 

non-monotonic operators are less nonmonotonic 

than others. It is also worth 

noting that the diverse environments— if 

and only if, then and only then, tonight and 

only tonight—lend to pattern alike for a 

given speaker. 

10 For more on the relarion between only if 

and //conditionals, see von Fintel (1994, 

1995)- 

11 Indeed, only subjects were present at the 

conception: 'As for the grammatical 

similarities of neg, wli, and only, these 

will now be described as resulting from 

the presence of a common grammaticosemantic 

feature to be referred to as 

Affect(ive)' (Klima 1964: 313). 

12 'The constraints on occurrence of yet and 

anymore are stronger than the constraints 

on ... any and ever, while . .. the constraints 

on negative polarity until are 

stronger still' (Horn 1970:318); see related 

work by Borkin, Ladusaw, Lawler, 

Linebarger, Ross, and now especially 

Hoeksema 1992 and van der Wouden 

1994, all demonstrating the variable 

nature of polarity effects. The overtness 

parameter for negative morphology in 

the licensing of polarity items and related 

diagnostics, as well as the parallel between 

only- and few- initial phrases, can be 

further detected in the patterns of'Klima 

diagnostics' illustrated below: 

(i) (#Only Hillary/Nobody but 

Hilary), I don't think, has faith in the 

plan, 

(i) (#Few people/Not many people), I 

don't think, have faith in the plan, 

(ii) (#Only Hillary/Nobody but Hillary) 

has faith in the plan, do they 

(ii') (#Few senators/()Not many senators) 

have faith in the plan, do they 

(iii) (#Only Democrats/Nobody but 



Democrats) opposed NAFTA, not 

even those from strong labor 

districts. 

(iii')(#Few Republicans/Not many 

Republicans) opposed NAFTA, not 

even those . .. 

13 Only animal is man (Only animals are 

humans) is converted into this: Every 

human is (an) animal, by this rule: An 

affirmative exclusive is convened into a universal. 

14 The fact that only some universals—those 

with every or all, rather than any—tend to 

induce the existential inference might 

suggest that import is a conventional 

property of particular determiners. But 

there is growing evidence that freechoice 

any is an end-of-scale indefinite 

and not a true universal, although its 

semantics mirrors that of universals in 

certain contexts: cf. Vendler 1967, 

Kadmon & Landman 1993, Lee & Horn 

1994, Jennings 1994: I9of£, and especially 

Haspelmath 1993. Moravcsik (1991) 

argues that the existential proposition is 

uninferrable just in case the universal 

statement has the force of a law-like 

generalization. I shall remain agnostic 

here on how existential import in universals 

should ultimately be treated; for our 

purposes, the essential point is the correlation 

of existential import in universals 

and in only statements. 

1 5 Aspects of the analysis presented here are 

prefigured in McCawley 1981: 226, de 

Mey 1991, and Hoeksema 1991. As 

pointed out by a referee, (25) would not 

be entirely vacuous even if it were known 

than nobody supported Clinton, but it 

would be a relatively unhelpful and inefficient 

way to report the emptiness of this 

set. 

16 This last proposition is of course the 

classic presupposition-violator of Austin 

1962, and as such it figures in a curious 

claim ofJay Atlas's: 

My analysis also explains the linguistially 

anomalous character of ## # Only a is F 

and a is not F by showing that it is a logical 

contra-diction, not merely a contrapresupposition 

(as in Horn's [1969] analysis). 

Compare the less anomalous, 

contra-presupposition sentences: ##All 

John's children are bald and John is childless, 

# The king of France is bald and there is no king 

ojFrance/he doesn't exist. (Atlas 1991: 139) 

For Atlas's Theory C A , as we noted above, 

the first of these three anomalies involves 

an entailmcnt (the only clause entailing 

the prejacent, which thus cannot be 

negated without contradiction), but it's 

not clear why the third doesn't also 

involve a contradicted entailment. 

Whether or not existence is presupposed 

(or conventionally implicated) in singular 

expressions, it is in any case entailed in 

positive descriptions; a better 'less anomalous, 

contra-presupposition sentence' 

would have been The king of France isn't 

bald—because there is no king of France. And 

what of Atlas's leniency toward the 

second of his three defendants I see no 

reason why the universal case isn't exactly 

as intolerable as the only one preceding it; 

indeed, they are bad to the same extent 

and (1 argue) for the same reason. I just 

have a very hard time with Atlas's theory 

of the truth conditions of All John's children 

are bald. 

17 Note that (29c) is technically false, but 

then so is (29a), given the existence of 

feather-beds and parkas. Here, as elsewhere, 

the universe of discourse from 

which alternatives to the focus of only are 

drawn is tacitly restricted (cf. William of 

Ockham [Kretzmann i968];Rooth 1985). 

18 As a referee observes, the equivalences in 

(30a, b) finesse the very real differences 

between the collective all and the distributive 

every in English, on which see 

Vendler 1967. Depending on the semantic 

properties of the focus of only, either 

every or all provides the more appropriate 

converse. 

19 I am inebted to Kai von Fintel for (his 

attempt at) clarifying my thinking on this 



20 The reference to 'preposed negation' 

alludes to the well-known fact that it is 

only when material falls within the scope 

of the relevant operator that (what would 

now be called) scale reversal occurs: 

compare Non animal currit; ergo non homo 

currit vs. Homo non currit; ergo animal non 

currit (see William of Sherwood's Syncategoremata, 

in O'Donnell 1941: 73). 

21 For Burley, the correspondence between 

universals and exclusives can also be 

expressed within supposition theory (cf. 

Sanchez 1994): the A position in (30) 

(focus of on/y/nuclear scope of all) is said 

to have suppositio confusa et distributiva, 

while the downward-inference-licensing 

B position has suppositio (mere) confusa. 

22 I am indebted to Roger Schwarzchild, 

who called my attention to the exchange 

in (36a, b) and to the fact that the problem 

it poses for the traditional presupposirional 

approach to the prejacenr inference . 

(a problem noted by Anita Mittwoch, 

personal communication to Schwarzchild) 

is dispelled by the current analysis. 

23 Theory P correctly predicts anomalymore 

precisely, a truth-value gap—in the 

former case (while the conjuncrionalist 

account renders the wagered preposition 

simply false and the implicature theories 

render it true) but comes a cropper in the 

latter, while Theories C/C A , PP, and G 

score in the fourth scenario but drop the 

ball in the third. 

24 The A position constituting the focus of 

only is upward entailing and should never 

trigger NPIs. Life is not that simple, 

however, this matter is considered in the 

Appendix. 

25 The unacceptability of (38c)—if indeed it 

is unacceptable (judgments vary)—cannot 

be predicted semantically, but if the 

present analysis is right, it can't be 

attributed to the non-LF-Comphood of 

the only phrase either, pace Progovac. 

Instead, it appears to involve a parochial 

structural or functional constraint on 

licensing from indirect object position, as 

seen from the acceptability of (ii): 

(i) *John gave only HIS GIRLFRIEND any 

flowers. (-(38c)) 

(ii) John gives only FLOWERS to any of his 

girlfriends. 

See Horn & Lee (1995) for a fuller critique 

of Progovac's configurarional theory of 

polarity. 

26 As Hoeksema (1986) has observed (cf. also 

van der Wouden 1994; 76-7,136), the only 

patterns with superlative triggers (the 

X-est. . .). There is etymological support 

for this relationship: Hoeksema notes that 

Dutch de enigste functions as a 'lowbrow 

variant' of standard de enige 'the only 

(one)', and we can add the onliest—attested 

by various non-standard American and 

British dialects—as a parallel. Both the only 

and superlatives license a subset of NPIs, 

essentially any, ever, and what Hoeksema 

calls the 'domain-settlers', i.e. those that 

define a domain of quantification within 

which an extreme scalar value can be 

picked out: 

(i) The only one who could ever reach 

me was the hard-lovin' son of a 

preacher man 

The first rime ever I saw your face... 

Jay is the {only/youngest} linguist to 

have ever eaten any squid souffle, 

the first rime in years you've been 

(*all that) happy to see me ... 

the youngest woman who {has yet 

climbed/'need climb} Mt. Everest 

Hoeksema observes that while the triggers 

in the class of the only/the first/the 

A-est do not define a strict D£ environment—if 

Kim is the youngest American 

to have {ever/yet} climbed Mt. Everest, it 

doesn't follow that Kim is the youngest 

American to have climbed Mt. Everest 

with a 500 pound gorilla strapped to her 

back—they do conform to a weakened 

version of monotonicity or DE-ness, 

necessitating an extra premise: 



(ii) 

X is the (A-est/only) Y 

X is aZ 

all Z are Y 

. .X is the {A-est/only) Z 

27 This interchangeabiliry evidently applies 

cross-linguistically. When the ill-fated 

defense witness Rosa Lopez in the O.J. 

Simpson murder case was interrogated by 

the prosecutor about the time period 

during which she had seen the infamous 

white Bronco parked innocently by the 

curbside, her response—rendered by her 

English interpreter as 'All I said was it was 

after 10:00'—actually began 'Yo solamente 

dije . . .' 

28 And in fact, as we would predict, the 

acceptability of the CN focus case of (50a) 

hinges on the satisfaction of existential 

import. The effect of the scalar indeed is to 

strengthen the exclusionary assertion, up 

to but not reaching the zero set: 

(i) — Did only Democrats support the 

President 

— Yes, indeed [only northern 

Democrats did/only Ted did/ 

#no one did). 

(ii) — Only mammals can understand a 

syntactically complex language. 

— Yes, indeed only {humans/ 

#vertebrates) can. 

(iii) — Only mammals can understand 

the semantics of quantification. 

— #Yes; indeed none of THEM can 

either. 

29 See also in this vein the claim 

Fermatically sealed in the margins of 

Fillmore, Kay & O'Connor (1988: 529): 

'There is independent evidence that barely 

may be analyzed semantically as "almost 

not", but space does not permit reviewing 

it here.' 

30 It must be allowed that in certain contexts 

both Petrine conjuncts do indeed fall 

within the scope of a proposirional 

attitude predicate. Thus the utterance of a 

sentence like (i) may be taken to express 

the corresponding conjunction in (ii), 

(i) Mary will be upset if only Bill makes 

it to her dinner party, 

(ii) Mary will be upset if Bill and only Bill 

makes it to her dinner party. 

i.e. as suggesting that Mary is apprehensive 

about an evening alone with Bill: if 

no one shows up at all she will cheerfully 

pop a pizza into her microwave and a tape 

into her VCR. I have suggested elsewhere 

(Horn 1992: 182-3) tnat îs case requires 

an invocation of the notion of explicature 

or pragmatic enrichment (a la Sperber & 

Wilson 1986; Carston 1988), so that the 

prejacent, while not constituting part of 

the linguistic meaning contributed by 

only, enters into the determination of 

what is said and hence into the (enriched) 

proposirional content. 

31 The behavior of the prejacent in the 

contexts of (62) contrasts crucially with 

that of a meaning component which is 

normally presupposed but may in the 

appropriate context be promoted to a 

dynamic aspect of content. Thus (as first 

pointed out by Karttunen 1973) while a 

change-of-state predication like that of (i) 

normally induces a presupposition about 

that past (as in (ia)) and makes an assertion 

about the present (as in (ib)), 

(i) Kim has stopped smoking. 

a. Kim used to smoke. 

b. Kim does not smoke now. 

this can be reversed in a particular context 

as in (ii), where the normally presupposed 

proposition (ia) is successfully brought 

within the scope of the illocutionary 

operator, a move that does not succeed 

for the prejacent of the only clause in (62). 

(ii) I know that Kim doesn't smoke now, 

but 

I wonder if Kim has STOPPED 

smoking (within the last six 

months). 

I just found out that Kim has 

stopped smoking. 

if [Kim has stopped smoking], I 

want to interview her for my 

study. 



32 Since writing this paragraph, it has come 

to my attention that the same argument is 

made by Delgado Lavin (1993: 6). Our 

conclusion here is prefigured by De 

Morgan, who attributes the tendency to 

infer not all. .. from an assertion of some 

... to the principle that 'the common 

usage of mankind inclines to partition', 

even in the natural, if fallacious, inference 

of an affirmative conclusion from 

negative premises. To make his case, De 

Morgan (1862: 277) underlines the telltale 

deductive adverb in the old Mother 

Goose rhyme: 

Jack Sprat could eat no fat, 

His wife could eat no lean; 

And so betwixt them both, 

They licked the platter clean. 

'Two abstinences', De Morgan remarks, 

'make complete consumption.' But not, 

we need hardly add, by virtue of the 

semantics of the purely negative determiners. 

3 3 Even in a more blatantly enriched context 

than Taglicht's, one in which the 

prejacent is a trivially entailed tautology, 

the exclusive predication does not 

ASSERT this tautology. The sentiment of 

Abelard's Socrates tantum est Socrates (see 

note 2) now appears in the guise of a commercial 

slogan for an anti-ulcerative prescription 

medicine: Only Zantac is Zantac. 

The point of the advertiser is clearly not 

to say or assert that Zantac is identical to 

itself but to assert that nothing else is 

identical to it or shares its relevant 

properties. 

34 Recall that Atlas's assumption of an 

entailed prejacent results in his denial of 

downward inferenriality of only XP, and 

correspondingly in his failure to appreciate 

the significance of the monotonicity 

effects (polarity licensing, inversion triggering, 

suspension, illocurionary scope) 

associated with exclusives. Since Atlas 

evidently accepts the standard nonconjunctive 

line on universals, he is also 

incapable of capturing the Burleyan 

insight recognizing the conversion of only 

AisBOallBisA. 

35 While I have for the most part concentrated 

my own forces on count-nounfocus 

only, the converse analysis extends 

naturally to all those varieties of only 

familiar from the literature (see e.g. Horn 

1969; Rooth 1985). Thus, if I like only 

spicy food, then everything I like that is 

food is spicy (with the implication that I 

like some food). If I like only spicy food, 

then everything I like is food that is spicy 

(with the implication that 1 like something). 

If 1 only like spicy food, then every 

relation that I have with spicy food is (no 

more than) liking. Space precludes a more 

detailed exposition, but I see no impediment 

to a straightforward adaptation of 

Rooth's (1985) cross-categorial alternative 

semantics for capturing our conception 

of what is said by ... only. . . 

36 My use of @ for 'rwo-sided' or exclusive 

disjunction is designed to recall Latin aut. 

This notation is prompted by the claim of 

Quine (1952: 5), Geach (1972: 15), and 

others that aut was unambiguously 

exclusive, as against the inclusive disjunction 

expressed by Lat. vel, the etymon of 

the classical V connective of the prepositional 

calculus. This empirical claim is, 

however, demonstrably incorrect; cf. 

Horn 1989: Chapter 4 and Jennings 1994 

for discussion. 

37 The conjunctive character of both all and 

only is explicit in Peter Tartaret's square 

of opposition (c. 1480) for exclusive propositions, 

reproduced in Bochenski 1961: 

234-36. 

38 I have set aside this issue of scalar (—'no 

greater than') vs. pure exclusive (—'no 

other than') understandings of only (cf. 

Horn 1969, Konig 1991, and Delgado 

Lavin 1993 for discussion) as orthogonal 

to those that have concerned us here, 

although I would argue, as does Konig, 

that the distinction can be treated pragmatically. 

39 In this respect, only differs from another 

unexpected polarity licenser, the deter- 



miner few (Few students who have any 

problems. . .). Few is not strictly DE in its 

restrictor: if few professors are wealthy, it 

doesn't follow that few professors who 

bought stock in Netscape the week it 

opened are wealthy. But few does define a 

limited DE environment in the sense of 

Heim (1984: 100-4), wno makes a similar 

point with respect to most. Crucially, the 

determinersû; and most, unlike only, are 

not UE, as their behavior in frames like 

those of (A2) and (A4a, b) shows. 

40 Note that the UE determiners some and 

many pattern with only in (A4a, b) while 

DE determiners—all, no, not all—yield the 

opposite distribution. 

41 Thus the medievals were able to predict 

that the A argument position in TantumA 

est B allows upward inferences and the B 

argument downward ones (see Burley's 

derivation of (34d)). But in fact the same 

tradition universally took such exclusives 

to represent garden variety exponibles, 

equivalent to the conjunction [A est B] et 

[nichil aliud ab A est B]. There is an obvious 

tension between, on the one hand, the 

ubiquitous definition of only statements 

as static (non-monotonic) conjunctions 

which shouldn't license vertical inferences 

in either position and, on the other, 

their recognition that lantum behaves as 

though it's left upward and right downward 

monotonic, precisely as the converse 

of omnis should be. Thus, the 

inference patterns we have attested here 

were both essentially captured and systematically 

excluded by the scholastics. 

42 On the approach to the acceptability of 

NPIs within the focus of only advocated 

in Israel 1994, it is not what a sentence 

like (Ale) says but what it presupposes that 

accounts for its licensing properties. 

Crucially, Israel takes this presupposition 

to have universal force: All people who have 

(ever) had a debilitating disease themselves . . 

If the arguments I have presented are 

correct, no such derived licensing is tenable, 

since Only As are Bs presupposes not 

All As are Bs, but rather the existential 

proposition that Bs exist, which in turn 

yields the inferences that (some) As are Bs. 

Any strengthening of this inference to a 

generic or a universal, as we have seen, 

depends on epistemically relevant 

properties of the predicate and/or the 

particular context. 

43 Other multiply-negative pairs exhibit the 

same promiscuity: 

(i) None of the guests who had seen any 

of the suspects were excused. 

None of the guests who hadn't seen 

any of the suspects were questioned, 

(ii) There's no one here who hasn't 

already eaten any of the infected 

meat. 

There's no one here who hasn't yet 

eaten some of the infected meat. 

44 The anomalous examples of (Al 3) are salvageable 

if a causal connection can be 

forged between the condition in the first 

clause and the effect in the predicate, a 

task rendered more difficult by the anticausal 

content of happen to. 

Atlas, Jay David (1991), 'Topic/comment, 

presupposition, logical form and focus 

stress implicatures: the case of focal par- 


REFERENCES 

Abelard, Peter (1970), Dialectica 2nd revised 

edition (L. M. de Rijk, ed.), Van Gorcum, 

Assen. 

AJtmann, Hans (1977), Die Gradpartikeln im 

Deutschen, Niemeyer, Tubingen. 

Anscombre, Jean-Claude & Ducrot, Oswald 

(1983), L'Argumentation dans la langue, 

Pierre Mardaga, Bruxelles. 

Ashworth, E J. (1974), Language and Logic in the 

Post-Medieval Period, D. Reidel, Dordrecht. 

Atlas, Jay David (1984), 'Comparative adjectives 

and adverbials of degree', Linguistics & 

Philosophy, 7, 347-77-


tides only and also', Journal of Semantics, 8, 

127-47. 

Atlas, Jay David (1993), 'The importance of 

being "only": testing the neo-Gricean 

versus neo-entailment paradigms', Journal 

of Semantics, 10, 301-18. 

Austin, J. L. (1962), How to Do Things with 

Words, Harvard University Press, Cambridge, 

MA. 

Barker, Stephen J. (1993), 'Conditional 

excluded middle, conditional assertion, 

and "only if", Analysis, 53, 254-61. 

Barwise, Jon & Cooper, Robin (1981), 

De exclusivis: an edition', Vivarium, 23, 23- 

54- 

Ducrot, Oswald (1973), La Preuve et le dire, 

Maison Mame, Paris. 

Fillmore, Charles, Paul Kay & O'Connor 

Mary (1988), 'Regularity and idiomaticity 

in grammatical constructions: the case of 

let alone', Language, 64, 501-38. 

Fretheim, T. (1995), 'Pragmatic implications 

of "not until" in Norwegian', paper presented 

at the 15 th Scandinavian Conference 

of Linguistics. 

Geach, P. T. (1962), Reference and Generality, 

'Generalized quantifiers and natural Cornell University Press, Ithaca. 

language', Linguistics and Philosophy, 4, Geach, P. T. (1972), Logic Matters, University 

159-219. 

Bochenski, 1. M. (1961), A History of Formal 

Logic, trans, by Ivo Thomas from the 1956 

German edition, University of Notre 

Dame Press, Notre Dame. 

Bos, E. P. (198 5), John of Holland: Four Tracts on 

Logic, Nijmegen, Ingenium. 

Burton-Roberts, Noel (1993), 'On preservation 

under negation', Newcastle-Durham 

Working Papers in Linguistics I, 18-42. 

Carlson, Gregory (1977), 'Reference to kinds 

in English', unpublished Ph.D. dissertation, 

University of Massachusetts. 

Carston, Robyn (1988), implicature, explicature, 

and truth-theoretic semantics', in R. 

Kempson (ed.), Menial Representations, 

Cambridge University Press, Cambridge, 

155-81. 

Chierchia, Gennaro & McConnell-Ginet, 

Sally (1990), Meaning and Grammar, MIT 

Press, Cambridge, MA. 

Delgado Lavin, E. (1993), 'Only and only-if 

conditionals', unpublished MS, Universidad 

del Pais Vasco, Victoria, Spain. 

de Mey, Sjaak (1991), '"Only" as a determiner 

and generalized quantifier', Journal of 

Semantics 8, 91-106. 

De Morgan, Augustus (1862), On the Syllogism 

and Other Logical Writings, P. Heath (ed.) 

(1966), Routledge & Kegan Paul, London. 

De Rijk, L. M. (ed.) (1982), Some 14th Century 

Tracts on the Prbationes Terminorum 

(Ingenium, Nijmegen.) 

De Rijk, L. M. (1985), 'Walther Burley's tract 

of California Press, Berkeley. 

Geis, Michael (1973), 'If and unless', in B. J. 

Kachru et al (eds), Issues in Linguistics: Papers 

in Honor of Henry and Renee Kahane, University 

of Illinois Press, Urbana. 

Grice, H. P. (1975), 'Logic and conversation', 

in P. Cole & J. L. Morgan (eds), Syntax and 

Semantics y Speech Acts, Academic Press, 

New York. 

Haspelmath, Martin (1993), A Typological 

Study of Indefinite Pronouns. PhD dissertation, 

Freie Universitat Berlin. 

Heim, Irene (1984), 'A note on negative 

polarity and downward entailingness', Proceedings 

of the North Eastern Linguistic Society, 

14, 98-107. 

Higginbotham, James (1994), 'Mass and 

count quantifiers', Linguistics and Philosophy, 

17, 447-80. 

Hoeksema,Jack (1986), 'Monotonie en superlatieven', 

in C. Hoppenbrouwers etal. (eds) 

Proeven van Taalwelenschap, TABU, Groningen. 

Hoeksema, Jack (1991), 'The semantics of 

exception phrases', in J. van der Does and J. 

van Eijck (eds), Generalized Quantifier 

Theory and Applications, Dutch Network 

for Language, Logic and Information, 

Amsterdam, 245-71. 

Hoeksema, Jacob (1992), 'Negative polarity 

items in discourse', paper presented at the 

First International Conference on Literary 

Semantics, University of Kent at Canterbury. 



Horn, Laurence (1969), 'A presuppositional 

analysis of only and even', CLS, 5, 97-108. 

Horn, Laurence (1970), 'Ain't it hard (anymore)' 

CLS, 6, 318-27. 

Horn, Laurence (1972), 'On che semantic 

properties of logical operators in English', 

UCLA dissertation', distributed by Indiana 

University Linguistics Club (1976). 

Horn, Laurence (1979), 'Only, even, and conventional 

implicature', LSA Paper, Los 

Angeles. 

Horn, Laurence (1989), A Natural History of 

Negation, University of Chicago Press, 

Chicago. 

Horn, Laurence (1991), 'Given as new. when 

redundant affirmation isn't', Journal of 

Pragmatics, 15, 305-28. 

Horn, Laurence, (1992), 'The said and the 

unsaid', SALTII, 163-92. 

Horn, Laurence & Lee, Young-Suk (1995), 

'Progovac on polarity: review article on Ljiljana 

Progovac, Negative and Positive Polarity' 

(Cambridge University Press, Cambridge, 

1993),/ of Linguistics, 31,401 -24. 

Israel, Michael (1994), 'Polarity sensitivity as 

lexical semantics', unpublished MS, 

University of Californa at San Diego, to 

appear in Linguistics and Philosophy. 

Israel, Michael (1995), 'Negative polarity and 

phantom reference', BLS 21. 

Jackendoff, Ray S. (1972), Semantic Interpretation 

in Generative Grammar, MIT Press, 

Cambridge, MA. 

Jacobson, Sven (1964), Adverbial Positions in 

English (Uppsala dissertation), AB Studentbok, 

Stockholm. 

Jacobsson, Bengt (1951), Inversion in English, 

Almqvist & Wiksells, Uppsala. 

Jennings, R. E. (1994), The Genealogy of Disjunction, 

Oxford University Press, New 

York. 

Johnson-Laird, P. N. & Byrne, R. M.J. (1991), 

Deduction, Lawrence Erlbaum, Hove. 

Kadmon, Nirit & Fred Landman (1993), Any. 

Linguistics & Philosophy 16: 353-422. 

Karttunen, Lauri (1973), 'STOP: is there a 

presupposition' Unpublished MS, MSSB 

Workshop on the Semantics of Natural 

Language, Ann Arbor. 

Karttunen, Lauri & Peters, S. (1979), 'Conventional 

implicature', in C.-K. Oh & 

D. A. Dinneen (eds) Syntax and Semantics 

11: Presupposition, Academic Press, New 

York, 1-56. 

Keenan, Edward L. & Stavi, Jonathan (1986), 

'A semantic characterization of natural 

language determiners', Linguistics and 

Philosophy, 9, 253-326. 

Klima, Edward (1964), 'Negation in English', 

In J. A. Fodor & J.J. Katz (eds), Prentice- 

Hall, Englewood Cliffs, 246-323. 

Konig, Ekkehard (1991), The Meaning of Focus 

Particles, Routledge, London. 

Kretzmann, Norman (1968), William of Sherwood's 

Treatise on Syncategorematic Words, 

University of Minnesota Press, Minneapolis. 

Krifka, Manfred (1990), 'Some remarks on 

polarity items' in D. Zaefferer (ed.), Semantic 

Universals and Universals of Semantics, 

Foris, Dordrecht, 150-89. 

Krifka, Manfred (1992), 'A framework for 

focus-sensitive quantification', SALT II, 

215-36. 

Krifka, Manfred (1993), 'Focus and presupposition 

in dynamic intepretation'.yowrna/ 


Kuroda, S.-Y. (1969), 'Attachment transformations', 

in D. Reibel & S. Schane (eds), 

Modern Studies in English, Prentice-Hall, 

Englewood Cliffs, 331-51. 

Ladusaw, William (1980), Polarity Sensitivity as 

Inherent Scope Relations, Garland, New 

York. 

Lakoff, George (1970), 'Repartee', Foundations 

of Language, 6, 389-422. 

Lewis, David (1979), 'Scorekeeping in a 

language game',/ Phil. Logic, 8, 330-59. 

Linebarger, Marcia (1981), The Grammar of 

Negative Polarity, IULC, Bloomington. 

Linebarger, Marcia (1987), 'Negative polarity 

and grammatical representation', Linguistics 

and Philosophy, 10, 325-87. 

Linebarger, Marcia (1991), 'Negative polarity 

as linguistic evidence', Papers from the Parasession 

on Negation, CLS 27, 165-88. 

Lobner, Sebastian (1987), 'Quantification as a 

major module of natural language seman- 



tics', in J. Groenendijk et al. (eds), Studies in 

Discourse Representation Theory and Generalized 

Quantifiers, Dordrecht, Reidel, 53-85. 

Lundquist, Lita & Jarvella, Robert J. (1994), 

'Ups and downs in scalar inferences', Journal 


Lycan, William G. (1991), 'Even and even if, 

Linguistics and Philosophy, 14, 115-50. 

McCawley, James D. (1974), 'If 2nd only if. 

Linguistic Inquiry, 5, 632-35. 

McCawley, James D. (1981), Everything that 

Linguists Have Always Wanted to Know about 

Logic but Were Ashamed to Ask, University 

of Chicago Press, Chicago. 

McCawley, James D. (1993), Everything that 

Linguists Have Always Wanted to Know about 

Logic but Were Ashamed to Ask, 2nd edition, 

University of Chicago Prss, Chicago. 

Moravcsik, Julius (1991), '"All A's are B's": 

form and content'', Journal ojPragmatics, 16, 

427-41. 

Morgan, Jerry (1972), 'Verb agreement as a 

rule of English', CLS 8, 278-86. 

Moser, Megan (1992), 'The negation relation', 

University of Pennsylvania dissertation. 

Moxey, Linda & Sandford, Anthony (1986), 

'Quantifiers and focus', Journal of Semantics, 

5, 189-205. 

Moxey, Linda & Sandford Anthony (1993), 

Communicating Quantities: A Psychological 

Perspective, Lawrence Erlbaum, Hove. 

Mullally, Joseph P. (1945), The Summulae 

Logicales of Peter of Spain, University of 

Notre Dame Press, Notre Dame. 

Nevalainen, Terttu (1991), But, Only, Just: 

Focusing Adverbial Change in Modern English 

t$oo-tgoo, Memoires de la Societe Neophilologique 

de Helsinki, Tome LI. 

Ockham, William of (1980), Ockham's Theory 

of Propositions: Part II of the Summa Logicae, 

trans, by AJ. Freddoso & H. Schuurman, 

University of Notre Dame Press, Notre 

Dame. 

O'Donnell, J . R. (1941), 'The syncategoremata 

of William of Sherwood', Mediaval 

Studies, 3, 46-93. 

Pegis, A. C. (ed.) (1945), Basic Writings of Saint 

Thomas Aquinas, Random House, New 

York. 

Pinborg, Jan (1981), 'Walter Burley on 

exclusives', in H. A. G. Braakhuis et al. 

(eds), English Logic and Semantics, Ingenium, 

Nijmegen, 305-26. 

Progovac, Ljiljana (1993), 'Negative polarity: 

entailment and binding', Linguistics and 

Philosophy, 16, 149-80. 

Progovac, Ljiljana (1994), Negative and Positive 

Polarity: A Binding Approach, Cambridge 

University Press, Cambridge. 

Quine, W. V. O. (1952), Methods of Logic, 

Routledge & Kegan Paul, London. 

Reis, Marga (1977), Prasuppositionen und Syntax, 

Niemeyer, Tubingen. 

Rooth, Mats (1985), 'Association with focus', 

University of Massachusetts dissertation. 

Sadock, Jerrold (1981), 'Almost', in P. Cole 

(ed.), Radical Pragmatics, Academic Press, 

New York, 257-71. 

Sanchez Valencia, Victor (1994), 'Monotoniciry 

in medieval logic', Language and Cognition, 

4, 161-74, Yearbook 1994 of the 

research group for theoretical and experimental 

linguistics of the University of 

Groningen. 

Sanford, Anthony, Moxey, Linda & Paterson, 

Kevin (1994), 'Psychological studies of 

quantifiers', Journal of Semantics, II, 153- 

70. 

Smith, Arvid (1922), 'Strodda anmarkningar 

till engelska syntaxen', Moderna Sprdk, 16, 

80-123. 

Sperber, Dan & Wilson, Deirdre (1986), 

Relevance, Harvard University Press, 

Cambridge, MA. 

Stalnaker, Robert (1974), 'Pragmatic presuppositions', 

in M. K. Munitz & P. Unger 

(eds), Semantics and Philosophy, New York 

University Press, New York 

Stalnaker, Robert (1978), 'Assertion', Syntax 

and Semantics Q: Pragmatics, Academic Press, 

New York. 

Taglicht, Josef (1984), Message and Emphasis: 

On Focus and Scope in English, Longman, 

London. 

Thijsse, Elias (1983), 'On some proposed universals 

of natural language', In A. ter 

Meulen (ed.), Studies in Modeltheoretic 

Semantics, Foris, Dordrecht, 10-36. 



Trudeau, G. B. (1994), Syndicated 'Doonesbury 

cartoon strip, 12 January 1994. tions', Natural Language Semantics, I, 123- 

von Fintel, Kai (1993), 'Exceptive construc- 

van Benthem, Johan (1986), Essays in Logica 48. 

Semantics, Foris, Dordrecht. 

von Fintel, Kai (1994), 'Restrictions on quantifier 

domains', University of Massachu- 

van Benthem, Johan (1991), 'Linguistic 

universals in logical semantics', in Semantic setts dissertation. 

Universals and Universal Semantics, Foris, von Fintel, Kai (1995), 'A note on bare plurals, 

Berlin, 17-36. 

bare conditionals, and only', unpublished 

van der Auwera, Johan (1985), 'Only if', MS, MIT, to appear in MIT Working Papers 

Logique et Analyse, 28, 61-74. 

in Linguistics. 

van der Wouden, Ton (1994), 'Negative contexts', 

University of Groningen disserta- 

Brooklyn', From Death to Morning, Charles 

Wolfe, Thomas (1935), 'Only the dead know 

tion. 

Scribner's Sons, New York, 91-7. 

Vcndler, Zeno (1967), Linguistics in Philosophy, Zwarts, Frans (1986), 'Categoriale gramm- 

Cornell University Press, Ithaca. 

Visser, F. Th. (1969), An Historical Syntax of the 

English Language, E. J. Brill, Leiden. 

atica en Algebraische Semantiek', Ph.D. 

dissertation, University of Gronigen.

Exclusive Company: Only and the Dynamics of Vertical Inference ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?