236 6 6MB
English Pages [163] Year 1980
I
Outstanding Dissertations •
In
Linguistics edited by Jorge Hankamer • Harvard University
A Garland Se ries
Reference to Kinds in English
Gregory N. Carlson
-:/
Garland Publishing, Ine. • New York & Londoll 1980
Fore~lOrd
Library of Congress Cataloging in Publication Data
Carlson, Gregory N 1948Reference to kinds in EngHsh. (Outstanding dissertations in linguisticsl Rev. version of author's thesis, University of Massachusetts, Amherst, 1977. Bibliography: p. 1. English language-Semantics. 2. English language-Noun phrase. 3. English \anguage-Number. I. Title. 11. Series. PE1585.C34 1980 425 79-6619 ISBN 0-8240-4551-3
In the spring of 1976, Terry Parsons and Barbara Partee taught a course on Montague grammar, ~/hich I attended. On the second to the final day of class, Terry went around the room asking the students if there were any quest ions at all that remained unanswered, and promised to answer them on the last day of class. r asked if he really meant ANY question at all, \~hich he emphatically said that he meant. As I had encountered a few quest ions in my lifetime that remained at least partially unresolved, I decided to ask one of them. What is life? What is the meaning of lire? After all, Barbara and Terry hOld promised to provide answers to any question at all. On the final day of cla'>s Barbara wore her Montague grammar T-shirt, and she and Terry busied themselves answering our questions. At long last, they ca me to my question. I anticipated a protracted and invalved answer, but their reply was crisp and succinct. First Barbara, chalk in hand, shal'/ed me the meaning of I ife.
Terry then stepped up and showed me what life really is.
As we were asked to show on 01 homework assignment earl ier in the year, this is equivalent to: ~t.
CD
1980 Gregory N. Carlson All rights reserved
All volumes in this series are printed on acid-free, 250-year-life paper. Printed in the United States of America
Leaving me astounded that I hOld been I iving in such darkness for all these years, the class then turned to the much stickier problem of pronouns.
ACKNOWLEDGEMENTS
The fact that my name alone appears on the ti tle page as the author of this dissertation should not be taken too seriously, for this could not possibly have cone into being were it not for the concerted efforts of a large number of other people. I have many to thank, but I will be brief. Barbara Partee, had she been present at UMass for the past year would have doubtless served as my dissertation director. Were it not for her work this dissertation simply could not have appeared. In spite of her absence, we managed to correspond with sufficient regularity so as to give rise to a fruitful exchange of ideas, and her many comments and suggestions improved dramatically the contents cf this work. My fi le of correspondence with her reveals that she put in a tremendous amount of time on reading and criticizing the material that I sent her. My debt to her is enormous. While perhaps not every cloud has a silver lining, this one did, and in Barbara's absence Robin Cooper came here for the year, and graciously assented to serve on my committee. His constant guidance and interest was of incalculable value to me, and t was consistently impressed with Robin's grasp of everything from confusing technical detai 15 to the most general aims of the field of I inguistics. I am certain that he taught me mare than he may th i nk. Lisa Selkirk served as my committee chairman, and we spent many hours together talking. Her support was of great value to me, and t thank her heartily for her time and eHort..PYp(d) (x I [~' (w ' 3z s [R(z,x) & find (w,z) 1) (b) 1) =~' (w i 3z s [R(z,d) & find+(w,z) 1) (b) b.
Syntax:
Bi 11 tried to find dogs
\
./
Bi I1
tried to find dogs
/
tried
In none of these sentences, whether the object of find is a quantified NP or a bare plural, does the subject ever stand in relation to any particular stages at all, even though in al I ca ses there is an existential quantifier that ranges over stages present. This is because it is present only as apart of the interpretation of the verb find, which must appear within the scope of try (as we have no mechanism in the grammar for quantifying in verbS). So no matter what the direct object of find is, whether it be a bare plural, a name of aperson, or a quantified NP, the subject of the sentence is never trying to find any particular STAGES of something. So this factor of the interpretation never gives rise to any ambiguity in such cases. Before 90ing on to the next section, a brief observation is in order. The semantics of the bare plural proposed make it so that we interpret (54) as being a relation between Bill and some particular thing, just as when Bi 11 is trying to find John, Bi 1I stands in a relation to some particular thing -- John. However, in a case like (54) the particular thing Bill is trying to find is not any object whatsoever, but rather a particular kind -- dogs. When we look at a sentence like (54) with the view that bare plural NP's refer to objects, it appears to be enigmatic at first sight. But once we thlnk of bare plurals as referring to kinds, the sentence is no longer so puzzling.
3.2 Oifferentiated scope. It was observed that there are cases where the existential use of the bare plural, if thought of as being existential Iy quantified, appeared to be capable of exhibiting narrower scope than any other quantified NP. for instance, there is sarnething that is very strange abaut (S7a) that is not strange about (57b) •
\
,.
{~~;:ed~~~s
b.
Dogs were everywhere.
to find dogs
I
find dogs
/
find
89
(57)
\
dogs
~~~s of dogs
}
were everywhere.
I nterpretat ion:
(In these interpretations, certain reductions have been carried out automatically for the sake of perspicuity, such as the treatment of the subject NP Bill, and the object of find, which is to be treated just like,bit:.e in this respect). One will see that as a result of treating bare plurals I Ike proper names the two derivations above reduce to identically-interpreted structures. Hence, there is no predicted ambiguity for any senten ce like (54).
Wi th the quant i f i ed NP' 5 as subj ects, the genera I i zat ion appears to be that the quantifier of the subject must hold scope over the universal quantifier present ;n the predicate. If we th;nk of the bare plural as being exlstentially quantified, then the bare plural in (42) behaves exceptionally in that It may appear to hold narrower scope than the universal of the predicate. But the treatment proposed here does not have to treat the bare plural as exceptional with respect to this construction in any way. Let us first examine the interpretation of the sentence some We translate the predicate be evergwhere as
dog was everywhere.
90 - Carlson
Carlson - 91
a predicate that denotes a set of individuals that have areal ization in each place: (58)
Translating some dog as: P'"3x[Dog' (x) & "p(x»), and treating some äog as subject, and be everywhere as predicate, we arrive at the following translation. p3x i [Dog' (x) &
&
s "p(x») (A)..xiVyi [place' (y)-+ 3z [R(z,x)
At(Z,y)JJ).
s ",,3)(i [Dog' (xl & Vyi [Place' (y)-+3Z [R(z,x) & At(z,y)JJJ This says that (with respect to a given world and time) there is some dog which, for all (relevant) places, has a stage in every one of those places. Given the nature of objects, such an assertion speaks of a strange state of affairs, where a given object is able to be in more than one place at a given time. Thus the sentence is necessarily false when taken I iterally. Note that we are not claiming this sentence to be syntactically or semantical Iy ill-formed in any way. It is just astrange literal assertion to make. All other quantified NP's will werk in a similar way. When we compare this to a semantic structure accorded the bare plural, we find that a natural interpretation arises if the bare plural is the subject of the sentence. (57b) is given the following translation. (60)
several d09S} some dogs for twenty minutes. { lots of dogs
,.
John ki lIed
b.
John killed dogs for twenty minutes.
>..xiVyi[place'(yi)-+3zs[R(Z.X) & At(z,y») Here, place' is a set of places, and ~ is a 2-place relation meaning one is at the other.)
(59)
(61 )
let us adopt a variant of the analysis of Dowty (1972) with respect to 'for' time adverbials as a means of showing again the manner in which the proposed analysis can account for cases of differentiated scope. I treat kill on the happening reading as being again a relation that holds between stages. Dowty treats that 'for' time adverbial as being a universal quantifier over a restricted range of times, and the times are related to propositions by the predicate At. let us now give a possible semantic interpretation for John killed same dogs for twenty minutes.
(62~ 9xi~,{x)
&
Vt:tE20 minutes
[R(z,x)
&
kill+(y,Z)J]]
IAt{t,~x3z[R(y,j)
&
The claim made he re in this structure is that there is same dog such that for all (relevant) times that are members of some particular 20-minute periodthere was a John-stage that killed a stage of that dog. This requires that the same dog have two or more stages that were ki lIed -- the same dog must be ki lIed repeatedly. Again, in the world we live in, this is a rather strange state of affairs indeed. The sentence John killed dogs for twenty minutes, however, requires no resurrected dogs to make sense of it. We can see why this is so if we examine the following structure which we would associate with the sentence.
PYP(d)(x i "tyi[Place'(y) .... .3z s [R(z.x) & At{z,y)J])" s Vyi [Place' (y)-+3z [R{z,d)
(y,z)])(')J &
At{z,y)J]
This says that at every place, there is some stage that is the stage of the individual denoted by d. Since d is a KIND, and not an object, d can have various stages appearing in different places at a given time, so there is nothing particularly odd about the interpretation given (57b) and it could be true as weIl as false dependlng on how dogs are distributed. The crucial difference is whether or not the stages that appea, in every place have to be the stages of the same object or not. in (57a) they must be; in (57b) there is no need for the stages to be the stages of the same object (a particular dog) , so long as they are stages of the same KIND. A similar sort of solution sheds light on the nature of the 'for' time adverbials. Senten ce (6Ia) speaks of astrange state of affalrs. but there is something perfectly natural about (6Ib), which requires no special assumptions in order to be true.
This makes the claim that at each instant (of relevance) in the twentyminute period, there was a John-stage that killed a stage of the individual d (dogs). Ther'e is no necessity in the stages that were killed to be stages of the same dog, and in fact the natural fnterpretation Is the one where the stages are in fact stages of different individual dogs. But all the stages are stages of the same individual -the kind DOGS. This ean be derived even t~ugh we introduce the expression dogs outside the scope of the universal quantifier. But if we introduce the quantified NP's outside the scope of the universal quantifier (which we must, apparently, in the ca se of 'for' time adverblaIs, though Dowty's analysis does not account for this). the quantified object NP must hold scope over this universal. But in treating bare plurals I ike proper names, we find that we can come to an understanding of why 1t is that the bare plural on the existential interpretation behaves as it does with respect to these time adverbials in
92 - Carlson
Carlson - 93
and .(64b) must be interpreted on the transparent reading, where same partlcular books are being sought.
contra-distinction to all other quantified NP's. It is predicted that cases of differentiated scope will arise whenever the following sort of situation holds. There must be an existential quantifier that ranges over stages that appear as part of a predicate. In the presence of certain material (for any number of reasons), that predicate cannot be appl ied directty to an argument, but must first have this extra material added before applying to an argument. This extra material may contain quantification whlch holds scope over the existential quantifier that ranges over stages, but it may or may not hold scope over the argument that the predicate is ultimately applied to. If it Ilolds scope over the existential quantifier ranging over stages, but not tlle quantlfication present in the argument NP, differentiated scope wil I arise. This is because the stages within the scope of the intervening quantlfier need not be stages of the same OBJECTS If a bare plural is the argument, but they must be Stages of the same objects if a quantified NP is the argument.
These facts can be accounted far in a relatively stralght-forwar? manner, as presented in PTQ. A quantifier can only bind those var~ables that are within its scope, and it wil I leave ~nbound any variables that do not fall within its scape. let us use (64b) as an exam~le. If we bui Id up the sentence in roughly the following way, we find that the scape of the existential quantifier does not include the variable in the secand conjunct (the pronoun). (65)
~~ B;II~ t ry
3·3
Anaphora. For the purposes of this section, we will be making the assumption that all cases of intra-sentential pronominalization arise via semantic bending of variables. While this assumption is perhaps dubious in fact (see chapter 5), it will not seriously affect the form of tlle argument presented here, and it greatly facil itates exposition.
a
,.
Bill is trying to fi nd
,ome} {two
(66)
Bill is trying to find bOOks, and Jill 15 trying to find them, too.
find x
x
unbound
find xl. This, too, is the sort of structure that would be associated with the opa~ue reading of the first conjunct. This is what blocks (64b) from havlng an opaque reading in the first conjunct. In order to bind the variable in the second conjunct, it is necessary to introduce the NP a book outside the scope of the opacity-inducing rry in the following manner.
trying to find them, too.
c.
find
/\
/
[Bill try[3x[Baok(x) & Bill Find x]] & [Jill try Jill
books, and Jill is
Bill is trying to find a book, and Jill is trying to find it, tao.
bOOk~
Since we are assuming that pronouns only arise through semantic binding of variables, this structure cannat give us sentence (64b). In rough form, its translation would be:
~"y
b.
,
!
It was noted in Chapter 2, that unlike quantified NP's. in contexts such as those presented in (64) bare plurals da not require identity of reference of particular individuals. (64)
J i 11
16])
In (64a) and (Mb), Bill and Jill must be trying to find the same particular books. In (64c), however, the books they are looking for may be quite different anes, and I think it is true that there is no reading of (64c) which means they must be looking for the same books. In addition, it is further noted fhat the first conjuncts of (64a)
sod
Bi J 1
J I11
tey
Uy
/
find
.1
3x[Book(x)
&
x
[Bi 11 t ey to find x
finä &
x
J ill uy to find xJJ
Carlson - 95
94 - Carlson In this case both variables are bound by the quantifier, and thus the pronomin~l form in the second conjunct may arise. This entails that the first conjunct have a transparent reading only, as the NP a book is introduced outside the scope of try. When we were thinking of the bare plural form in (64c) as being existentially quantified, the fact that we only find an opaque reading for the first conjunct appeared as an anomaly because there,. too, we can successfully pronominal ize. But given the treatment outllned here, this is the only expected reading we would anticipate finding. In order to obtain the pronoun, we must introduce the NP books outside the scope of try as we had to do with a book above.
(68)
';',\
JiII
This treatment also allows us to account for the cases noted in Chapter 2 where there was pronominal ization occurring between a generic useof the bare plural and an existential use.
tey
tey
/
!n the second formula, the variable x ranges over objects, and the formula makes the claim that there is some object that is a book, and Bill is trying to find one of its stages, and Jill is trying to find ont" of its stages as weIl. They are seeking stages of the same object in both cases, and hence the identity of reference. should stress that in both cases, with bare plurals and with quantified NP's, we are obtaining identity of reference. It is Just that in the former case. the identity of reference is required of the KIND sought, but in the latter the identity of reference is required of the OBJECTS sought.
books
fi nd
thus may be stages of different individual bOOks, they ARE trying to find the same KIND of thing •• books. Just because they are looking for the same kind of thing, it does not mean that they are looking for the same OBJECTS. Thi5 is what gives rise to the paradox, then, the 'lotion somehow that books makes reference to objects. Once we dispense with that not ion, we are able to gain an understanding of the processes that are at werk here.
I
x
find
x
(71) Semantically, the syntactic structure sketched above would give rise to the following formuJa, to which more detai I has been added in order to obtain reference to stages (though this remains schematic in cer· tain respects). (b'" books, the individual) (69)
;\.PVP(b)(xi[Bill try[yi 3zs[R(Z,x) Jill try [;..i 3u s[R(u,x) Bill try [yi 3/[R(z,b) [R(u,b)
&
&
& find+(y,Z)]]] &
find+(w,u)]]). '"
&
find+(y,z)]] & Jill try [.;,i 3u S
find+(w,u)ll
Compare with more detailed analysis of (64b): (70)
Jxi[BoQ-'~'(x)
[Bill try[yi 3zS[R(Z,x)
&
.
Jill try[;..l :3 u s [R(u,x)
&
find+(y,z)]]) &
+
&
find (w,u)]l]
The first formula above makes the claim that there Is some individual b such that Bill is trying ta find same stage af b, and Jill is trying to find some stage of b. Tho.ugh the stages may be different, and
I see eagles every day, even though they are nearing extinction.
Essentially, the sentence says that I see STAGES of a certain kind of thing (eagles) every day, even though that kind of thing is nearing extinction (and not its stages). In botn cases what I see and what is nearing extinction are the same entity -- eagles. It is the fact that we can only 'see' stages of things that gives rise to the existential reading in the first clause, and it has nothing to do with an ambiguity of the bare plural per se. 3.4 Summary. In this section, we have seen how a unified analysis of the bare plural not only Jets us account for the arguments which show that a unified analysis is called for, but also leads us to an understanding of why it is the existential use of the bare plural behaves in the pecuJ iar ways it does when opposed to the quantified NP's. The analysis given here predicts these properties exactIy in the way in which they in fact occur, and as such adds further support to the not ion that .'I correct analysis of the bare plural lays along the lines of a unified analysis. To this point, formal ism has been largely ignored in favor of eXhibiting the intuitive significance of the sort of analysis offered here. We will move fai rly slowly, though in the direction of increasingly rigorous formal ization. (n the next section, other possible non-quantificational approaches are discussed. We will then follow with some further applications and extensions of the lilles of analysis
Carlson - 97
96 - Garlson pursued
and then present a prel iminary syntall and semantics .. The
prelimi~ary syntall and semantics will serve as a base from whlch to
go on
to further duscussion to the chapters that follow. 4.
Other possible Non-quantificational Approaches
4.0 Introductory remarks. I t has been proposed tha~ kinds of things are to be considered individuals ~n the sense o~t~ Ine~ above._ But this is just one of a number of posslble non-quantlflc~tlonal. re presentations that can be accorded the bare plural. In thls ~ectlon, we wi 11 briefly ellamine other possibi I ities th~t have been ralsed, and argue that the path chosen is, so far, the optimal one to take.
terms of the formal system presupposed here. First of all, if dogs denotes the intension of a set, then its extension at any given points of re fe ren ce should then be the set of all existing dogs. However, in the classically extensional contexts, one finds most often that the interpretation is SaME dogs, and clearly not the set of all dogs. In other extensional contexts, though, the interpretation of 'the class of all existing dogs' (or something along those lines) appears called far. By conventional criteria, the underlined NP in (72a) are in extensional contexts, yet In (72a) the interpretation is cJearly existential, and in (72b) it appears to be universal. (72)
4.1 Kinds as sets. One possiblity is that kinds of things are to be regarded simply as (characteristic functions of) sets. We may wi sh to represent what the NP dogs denote~ as si mp I y :he set of all dogs. Or, in more formal notation: ,\xdog (x) . . As pOlnted out by Gartwright (1975) and Mates (1968), this is simplY.lncorrect, as the set of all dogs will vary in membership from one time to the next {not to mention from one possible world to the next}. In.set theory, two classes or sets are identical just in ca se they conta~n the same members, otherwise, they are different sets. As Mates po~nts out, though this is not in accord with our intuitions concernlng the way we spe~k. It would be sheer nonsense, under this hypothesis. to say suppose men were less numerous than they are, for :hen we would be saylng something like Suppose the class of men, wh~ch bas so many members, bad less members then the class of members would be a DIFFERENT class, for it would have different members. But sup~o~ing men to be less numerous appears ta make perfeet 5en~e. In ad~ltlon, we would wish to be able to distinguish different k~n~s things even though these kinds of things rnay not be exempllfled In the real world. We know that 95-lb. house cats are quite different fra~ square circles, for example, and treating the bare plural as slmply a set would not distinguish these two.
0:
A far more reasonable approximation, then, would be to treat the denotation of dogs as being something that picks out all possible dogs, and not simply the existent ones. In terms of the.sernantic system used here, this would be to say that dogs ~enotes an I~ten sion of a set -- a function that picks out all eXistent dogs In any possible world at any given time. This would distinguish 95-lb. house cats from square circles, and it would allow us to suppose that ~n. n were less numerous in a rather straightforward manner (the SUpPOSltlO being, then, that there is a possible world in which the objects. picked out by men has fewer members than in the real w~rld). Th~S approach has an undeniable relev,ance to the study of kinds of things. However, it has at least two generalsorts of wea~nesses: ~o~h which have to do with actual implementation of thls posslbl I Ity Ln
0:
a.
Dogs are barking at me.
b.
Dogs are mamma!s.
So some ambiguity appears in extensional contexts where we would expect none, and the reading that appears most often is an existential rather than the predicted 'universal' reading (that iso meaning something like Ithe set of all dogs l ) . So treating dogs as the intension of a set makes some initially incorrect predictions with respect to extensionality. Oistributionally. and by all other known criteria. it is apparent that dogs is an NP, which in a Hontague framework denotes a set of properties. However, dogs simply denotes a property jf It Is the intension of a set, and not a set of properties. So we need to do something to the expression dogs to make it act semantically like an NP. There are two means of doing this that I can think of. The first involves additional complexity in the syntax and semantics that can be avoided by the approach that I am advocating, and the second, to be discussed in the next section, destroys the initial proposal entirely. One might propose that dogs is the set of higher-order properties that are associated with the property dogs, or the set of sets of which the intension of the set of dogs Is a member (rather than, in the case of John, the set of sets of which John Is a member (ignoring intensions)). That is, the NP dogs would be of type «5, , while John or every man would be of the type «5, . In order to be able to say anything meaningful about dogs, then, and treat it as sketched, we immediately give up the not ion that (in a Montague framework) dogs ls of category NP, for there is intended to be a functional relation between syntactic categories and semantic types. This not ion is violated by the treatment sketched. Even supposing that we grant dogs an NP stat~s, but allow it to have the suggested semantic type, It would ental I that the whole verbal system be completely restructured syntactically, along witn many other parts of the grammar. Instead of
Carlson - 99
98 - Carlson
just denoting sets of entities, .111 verbs (and adjectlves, and so on) would also have to be able to denote sets of properties as weIl. In order to say dogs run, run would have to be of type «s, rather than simply of type . That is, we would be making the claim that run in, dogs run is entirely different from run in John runs; and that be big in Dinosaurs were big is a different predicate from be big in Lincoln was big; and so on and so forth for every predicate for every argument position. The net effect.o: this program is simply to take a given predicate ~f the famll lar ~ort! find the 'innermost' occurrence of e, erase It, and replace It wlth I Y come about.
We find a similar difference arising in the case of the bare plurals with the individual-by-individual coreference required. (28)
a.
Goldfish like everyone who likes them.
b.11Everyone who likes them Is liked by goldfish. (28b can only mean that everyone who I ikes goldfish is I iked by goldfi sh).
1.2 Inductive Qeneralization. repeated here.
A third reason for thinking the pronouns of (25) to be cases of true bound variables is that bound variable NP's cannot be epithetics, while ca ses of non-bound variable pronouns can just as easily be replaced by some epithetic NP. A few examples of cases where such replacement is acceptable are as follows.
(29)
a.
One crook was attacked by several po 1 i I;emen who had
{him } the bastard . here means that {he idiot} has the
sworn that they would get b.
(25)
The fact that Bob i s
11
1
(JO)
a. -ItEach nan hates cops who attack the bastard. (OK: hirn). b. "Nobody was attacked by crooks who had sworn that they would get the idiot. (OK: hirn).
Quite clearly, all the sentences of (21) resist this sort of replacement. (25') a. b.
DOgs obey the little devil's masters. Cats clean the wretched creatures.
Goldfish like everyone who likes them.
On one reading, it means that goldfish I ike everyone who 1 ikes goldfish, a reading that would easily be accounted for by quantifying in the subject NP, binding the kind-level variable that appears in the relative clause. Schematically, the resultant translation would be of the following fOrm. (g is the individual kind 'goldfish').
1
1
forgotten his date wi th Susan aga in. But when we must have abound variable interpretation, replacement of the pronoun with an epithetic NP yields unacceptable results.
Consider a sentence like (25),
(31)
AXk[X likes everyone who likes x](g)
(31) ls equivalent to 01'). (31')
9 likes everyone who likes 9
However, there is the other reading of (25) in whieh it is required tllat a given goldfish I ike those people who I lke that goldfish (and not necessarily any others). There seems to be an object-Ievel predicate in (25) that is of the form of (32). (32)
AXO[X likes everyone who Ilkes
xl
(This is the property of being something that likes everyone who likes it). lntuitively, it is as if this predlcate here ls being applied to the bare plural. It might be initially thought of as being applied as in (33).
164 - Carlson
(33)
)"p~p(g)(xo[x likes everyone who 11kes x])
But of course (33) cannot be meaningfu11y interpreted, as 9 is a kindlevel entity, but the property is one that appl ies only to OBJECTS. (33) Is equivalent to (33').
(33')
o )"x [x I i kes everyone who 1 i kes x] (g)
Thls asserts that the kind-level entity 9 Is in the set of objects that likes,everyonewho likes it, whieh Is not meaningful, much less true.
Carlson - 165
tell you how to determine if a given objeet-Ievel predicate 15 to be truly generallzed to a given kind. These decisions require varlous acts of cognition, judgment, deduetion, and a host of other human activities in order to evaluate properly. I trema i ns to be shown how to der i ve a sentenee such as (25) in a reasonable rnanner with the rules of syntax and semanties we have before uso We will for the sake of brevity introduce another example for our inspeetion at this time. (35)
However, there appears to be an intuitive sense in which we would I ike to be able to assign the property of (32) to the set of properties of a kind of thing (e.g. goldfish). let us imagine first how it is we basical Iy may come to know whether or not goldfish 1 ike everyone who I ikes them (on the reading we are now interested in). Praeticalities aside, we come to know this to be true or false by observing individual goldfish. We observe that Goober likes everyone who likes her, Milton likes everyone who likes him, Aliee likes everyone who likes her ..... and so on. until we have observed enough goldfish to decide we can make a val id generalization. The generalization might be cast in the form 'goldfish like everyone who likes them', meaning that we have observed a sufficient number of objeets of whom the predicate holds to attribute that property genera'ly to the kind. But, as we have seen. the predicate in quest ion is not one that can be meaningfully applied to a kind of thing. However, there is a sense in which we were not applying the predicate directly to the kind 'goldfish', for we made this attribution as a result of inductive generalization, based On the truth of the predicate in quest ion with respect to particular object-Ievel goldfish.
Cats I ike themselves.
The reading we are interested in 15 where any (7) individual cat likes i tse If. \.Je mus t add a coup le of ru I es to the grammar in order for us to be able to derive this sentence. First of all, we must have some mechanism for obtaining reflexive pronouns, the statement of whlch I forego at th ist i me. We a I so need a means of gett 1ng G i nto the i nterpretation. As G in Engl ish has no surface reflex, the rule that syntactically introduces it is the identity function that applies to VP's. These approximate the final vers ions to be found in the second fragment.
There is one further syntactic rule which will be adopted here. This is Partee's (1976a) rule of Oerived Verb-Phrase formation. The effect of this rule is to take a sentence with a free variable subject, and make that sentence into a derived VP by forming an abstract on the subject of that sentence. The version of Derived VP presented below is a modification of Partee's original suggestion. Derived VP: If (m), 1>(I) •...• 1>(r) is true for enough times, and n,m.I, ...• r are stages of b. then we consider GC1»(b) to hold. Let us ca!l this process 'generalization', thus the choice of G for the means of representing this predicate. This is a cognitive process, and wi I I not be entirely represented in the grammar. In particular, there is no mention made of a necessary and sufflcient number of times for some stage-level predicate 1> to hold of stages of x to say G('(m) and 1>(n) botn hold. Thus, if Bill smokes, then there have to be two or more occasions on which B~ll is actually smoking. Thls condition will be formalized as a meanlng postulate and discussed at a later time.
Carlson - 169
let us then see how we might represent the ambiguity of Bill smoked. The generic reading would be as follows.
Syntactically, in order to obtain this formula, we wi 11 have to introduce a new syntactic rule for G. It would be as foliows. 521.
If Cl ( P
and Cl is of the form [[ß1Vt-r)1, then F (Cl) I9 1V ( P IV ' wnere F (Cl) = [Cl]IV' I9
Here, IV and IV' are syntactically differentiated 'levels' in the VP, semantically nondistinct. IV' but not IV, now may combine with subjects to form sentences. Semantically, the operation that corresponds to this syntactic rule would be: T21.
If Cl translates as a'
,
The event reading of Bill smoked (continuing to ignore tense) cannot be represented as in Chapter 3 any longer, as we are taking smoke as a predicate that basically applies to stages. If we left this predicate unchanged and added a subject, we would obtain an undefined formula as the subject is in all cases the property set of some INDIVIDUAL, and not astage. The crucial element that is mlssing here is the means by which we say tnat there is some stage of the subject that smoked. To accompl ish this, we add a rule that works on verbal IV's that serves the same semantic function as be does with non-verbal IV 2 phrases. and Cl is of the form [[sI (6)], then F (o.) v 21 IV c P IV ' where F21 (Cl) = [a]IV'
523.
If Cl (P
Tl3.
F
21
(Cl) translates as AX
i
3 zS[R(z,x)
& a' (z)]
So then the event reading of Bill smoked comes out represented in this way. (40) 3zs[R(z,b)
&
smoke' (z)]
One will note that when Gis semantically added to the IV phrase, it creates an intensional context for that IV phrase. This is a desirable result. as we have seen. On the event reading, when there is an existential claim made about some stage of the subject, no intensional context is i ntroduced , and hence the extensional ity of certain verbs in the event sense, but there is apparent intensional lty on the generic reading. It is noted he re that G syntactically patterns as part of the aspectual system of Engl ish. It is not represented as a modal, as G
170 - Carlson
Carlson - 171
may syntactically appear even in the presence of another modal or ~n places where we can find no rnodals (infinitives and gerunds). It IS not represented as a verb operator for the reasons stated above. Having it as part of the aspectual system of Engl ish is at this tim: more of a hunch than a rnotivated, arguable analysis. Nonetheless, It should be pointed out that G seemS to be incompatible with th~ presence.of the perfective aspect (and clearly the present progressive), and In those few cases I know of where it seems that a bare IV phrase appears no ambiguity between generic and existential shows up. The underscored IV phrases do not appeal" to have a generic interpretation. (41)
a.
Bill likes to see Sally run around in circ1es.
b.
Sam helps Fred load trucks.
However, it is recognized that the exact nature of these constructions is not entirely understood, and that the judgments are of a very subtle nature. In spite of this, we wi I I hold to the analysis proposed.
Iv
2,3 Predicates that basical hold of objects. This proposed analysis whereby we introduce G into the aspectuaJ system makes the prediction that any verb whatsoever that basically ~pplies t~ stages may be generically attributed to both objects and kinds. Thls generalization appears to be correct. This still leaves us in doubt about the gap in the distribution of predicates mentioned earlier, whereby there appeal"ed to be no predicates at all that applied exclusively to objects, and not to kinds as weil. Let us intuitively characterize the general ization function on ce again, so that we might then have an intuitive grasp on why the predicted gap does not appeal". We have in our ontology a h'lerarchy of stages, kinds, and objects. The G function allows one to gen~ralize a ~re~i cate from one that appl ies to stages to one that applles to an IndIvidual that the stages are of. A possible situation might be as foliows. a is a kind, band c are objectsj m. . . . t are stageS. F i gure
The fact that holds of the stages allows the possibility of generalization to occur. attributing G(A..xs ... p(y
3z s [R(z,y)
Adamisbuiltbybeavers.
b. ;o!Beavers build adam.
(I'm not sure this is even grammatical)
The general analysis will have the NP beavers in subject position with a generalized predicate being associated with it, 'built by beavers'. As the passivized NP now appears in subject position, it may have 01 general ized predicate associated with it that would otherwise be unavailable in object position. Accompl iShing this goal requires the introduction of a Passive transformation. In particular, ! make use of the form of the passive rule proposed in Partee (1976) in its strict form (i .e. only a variable may be moved out of object position). The form of the passive ru1e is
the
f:;~~W;::.,
oed, ;, of the foem ["I
,P
& build+(x,z)])
(,l,P"P(d») = G(x s 3z s [R(z,d) & build+(x,z»{b) Since the general ized predicate is being attributed to the kind beavers here. psychologically this represents the noting of 'enough' realizations of beavers building dams to make the generaJization warranted. We are incl ined to express this criterion in terms of a quantification over the subject ("all beavers") in informal discourse, but nothing of the sort is impl ied here in the translation presented. There is no general ized predicate attributed to the direct object dams, but only 1 stages of dams are mentioned. Hence, there Is nothing said about 'all dams. One wil I note that we could alternatively say a beaver builds dams and be saying almost exactly the same thing. It is further noted that beavers bulla adam is not naturally interpretable as being a statement about 'alT'dams.
to [Iß)[[OIV[{ T
t
Let us represent a translation of (66a). We treat build as a predicate that basically applies to stages of the subject, and takes realizations of the object (an extensional TV). The G function mediates. (69)
a.
him;}I $thern-'
(n)]IVJ. then F (.;.) s Pt where F {3w
The mouse is very quick.
(wrong reading)
o
k
k
&~3w 3 x 3y [x
This prohibition against the articles is not total, however. A may appear naturally if it is in predicate nominal position. This is in accord with our treatment of predicate nominals, where a there is not to be identified semantically with the indefinite article. (30)
The shrew i s not a kind of mouse.
And if an adjective or a relative clause modifies the noun, the sentences become quite acceptabJe. (31)
a.
A kind of beaver that lives in Siberia is fast disappear i ng.
b.
The kind of beaver that tinct.
This, too, holds with simple CN's. (32)
I
was speaking of is now ex-
find (32) entirely acceptable.
a.
A beaver that I ives in 5iberia is fast disappearing.
b.
The beaver that I was just speaking of is nowextinct.
believe the case of the definite article has a rather straightforward sort of explanation. It does not seem at all plausible, as Jackendoff (to appear) points out, to think of this as a case that cculd be readily handled by positing a determiner source for the restrictive relative clause, and have the kind CN's subcategorized for a Det S sequence. Instead, this distribution fallS out of the semantics of the definite article itself, in concert with the disjunction condition. We wi I I treat the definite article as asserting the uniqueness of some entity. Let us first look at what would be the full translation cf an NP such as the kind of mouse.
k
[mous~'
"y
k
&
0 (w ) &
~R'(wo,xi)J
k k 5 (Y ) & 5 (x ) 0
0
&
R'{wo,x k ) & R'(wo,l)J) This denotes the property set of that unique entity such that (a) that entity is a member of 5 (b) there are possibly some mice that do not realize that unique ent?ty (e) 5 is a subset of all the possible 0 kinds of mice (d) all members of 5 are disjoint kinds (e) necessari-
!~d :~~:y~~~~~. th~~ ~~~~~~s t~~~ ~~~d h:~~: :!s~a~:a~~z~~o~s~i:ds~~
50 5 that nas one member. 5uppose we choose a set 5 with two members tRat fulfill the requirements. Then there 'o'Jould be no unique xi member of 5 , as the only requirement is that 5 contain disjoined subordinate~kinds. Given this interpretatio~, ~he only way to Find a unique Xl in 50 is to let 5 be a one-member set. But if 50 is a one member set, all realization~ of the member of 5 must nece.ssarily be identical with the set of mice, and if somethin~ is a mouse, it must real ize the member of 5. Thus, the unique member of 5 would have to be the same aS the k?nd mice itself. But this confl?cts with the requirement that the kinds in 5 be subordinate kinds. Thus. we arrive at a contradiction. as tge sole member of 5 must and must not be identical with the kind mice. 0 Let us contrast this result with what happens when we have a restrictive modifier present, as in (34). (34)
The kind of mouse that roars.
Note here that the relative clause must be interpreted as modifying the CN kind oE mouse, and not just the CN mouse. If we illustrate with an adjective instead. the position of the adjective el iminates the ambiguity, and we find contrasts such as (35).
218 - Carlson (35)
a.
Carlson - 219 3. Such
The migratory kind of penguin.
b. *The kind of migratory penguin. Below is illustrated the translation of (34) with respect to the syntax presented.
(361
Syntax:
the kind of mouse that roars
'----
/" the
3.0 Introductory remarks. The English ward such represents an extremely interesting and recalcitrant phenomenon from both a historical and a synchronie point of view. This is due in part to the fact that it appears to participate in more than one sort of construction, and in part is due to the fact that the usage of this word appears to be in the process of change, leaving many 'fossi 15' behind that appear in present-day Engl ish.
kind of mouse that roars
/'
kind of mouse kind
/'
\
mouse
'----kt I
Such in many cases is ambiguous between a meaning that speaks of quant i ty, and another tha t speaks of qua I i ty or of character i st i cs. A sentence of the following sort is clearly ambiguous.
roars
.-1-
Such a Inversion
(AP) Ctl
(57)
a.
almost as s~art/ ~almost such
b.
just a Mt too quiek/
~just
a bit such
However, this lack of modification may follow from treating such as having (semantically) a p']t1CN"
where 01>'
CN
1.
and
where F25 ,n,m(a,qo)
is gotten from . Figure 2
What this rule will do is take a CN' like such men, and a clause like Bill sah' such men and der 1ve a CN' of the form such men as Bill sah'.
NP
,
~
AP ,
such
CN
,
men
Semantically, F
25 ,n,m
It creates a predicate out of the as-clause which has as ar9ument the free variable in the matrix occurrence of such. Formally,
$'
a~S
T27.
/'--.VP
NP
~ ,
CN
such
men
AP
gives us a derived phrase of type .
,
be'he,e
let us il lustrate with a concrete example how the system works. The example il lustrated is: All such men as Bill sah'. The translation of Bill sah' here is abbreviated as SB'.
Carlson - 243
242 - Carlson Syntax:
a! I such
/' all
S
such
men as Bi 11 saw
S
"men as Bi 1I
It would not be an understatement to note the complexity of the resulting formula, so we will in words here try to describe what the translation of the NP all such men as Bill saw says. The NP is of course a set of properties, in this case, properties of objects. It is the property set associated with any object that is areal ization of a kind, represented by the free variable x~. The context of use
saw
SUCh~ Bi 11 s~w SUCh 3 men SU{h ~en Bill .~ see~such men S \
/
see
suCh
I
such
/
such
3 3
all object-Ievel real izations of men
\men
yet another restriction.
Semant i cs:
Axo[Vz'b[R'(Zo,x~)
such
-+
S
man'{zo)] &
-+
(wo.x~)])
Translation of the expression resulting from rule F
2S.S,3 :
(zo) & R' (yO.x;)J
(xo) & '\X~[SB(tyk Oiwo[R' (wO,yk) __ Vzo [R'
(zo.x~)
-+
!!@!!'(zo) & R'
x~
x~
must be
is subject to
There is some kind of thing that Bill saw, Cl
kind all ofwhose realizations are men ( /
s
slnce they have all the same real izations.
3
(zO.x~) -+~'
must be men, and thus
S
R'(Xo,x~)]
SB (1/ DVw O [R' (wo, / ) __ [Vz°D[R' (zo, x~)
'\xo[[,\yOrVzOOR'
First of all,
too, is a kind of man) an by removing the first oecurrence of he~
Brv'/IV'
([oftenlAdv' [always]Ad)
or him Y and ehanging all subsequent occurrences to he or o
BIV / IV I T
BA =
A
([in 2 ]p' [near 2 Jp )
hirn respectively.
if A is not one of the eategories I isted above.
56.
If ct
Syntax. By PA Is meant the set of phrases of CAT A. Thls set is defined as folIows. Again. outermost braekets are omitted in the specifications of form, and parentheses around variables in these specifications of form indieate optional presenee. At the end of the syntaetie rules is abrief deseription of the funetion of eaeh.
57.
~
PA for every eategory A;
PCN' and ct iotf $B. then F (a) 5
I':
PT' where F {a) 5
If CI. ( PCN/CN and B ( PCN ' then F6 (a,B) ( PCN' wnere F6 (a,S) '" [$aS]T if B iotf $6, and "" [aB]r if B is not of the form $6
5a.
BA
I':
[$a]r
=
SI.
4> ( P , tnen F
4>' is gotten fram 4> by removing tne first oeeurrence
[expeet 1 } 2 V
{[be21V' [will 1V' can 2 ]V' [ ean ]V} +pres 2 +pres +pres 3
BIV'/IV
jf
= [$Cl[tnat 4>'J t J CN' wnere
( PCN' wnere F ,n,y(Cl,4» 4
{[believe]v' [expeetllV)
B 1Y / t
If a ( P and a iotf $B, and eN
If a (P
the proper braekets are assigned.
CN
"
then F (a) (P ' where F (a) = [al lV if a rv 7 7
iotf SB, and = [an ctllv if a is not of the form $B Thus, if a ( BA' by this ruJe [aJ
52.
If a
E
P
A
[. PA. 59.
and a iotf [B]N' then F (a) [. P ' where f (a) CN l 1 CN
= [$a]U-l 53.
If a ( PCN' then f 2 (0.) [. PCN' where F2 (0.) = [ct]CN'
54.
If a [. P ' and a iotf $B. and if E Pt' then f (o.,.) T4 IV/t
E
P IV "
527.
If f
517.
0. E
P IV ' where f 14
E
I5
Plv,then F
E
I5
(o.,S)
E
Prv ' where
528.
(0.,S) = [aB]rv
If a E P and S rv [o.S] IV unless
t:
Pr' then F
(a,S) E P 1V ' where F 16 (o.,S) 16
5Z9·
iotf [o]V[Y] IV' then F'6(o.,S)
0.
If
5Z0.
If Cl
0. E
E
PIV/INF and S
5Zl.
If a
E
P
where F 5ZZ. 5Z3.
If a
t:
If a
E:
1V
I9
P P
and
(0.)
IV IV
=
If. (.)
525.
5z6.
If Cl
E
E
P INF , then F I4 (Cl,S) E
P IV
t:
Pr' then F I8 {CL.S)
iotf [S]V(ll), then F'9{a)
t:
E:
Pt'
Ifa
E
CN
PCN' and Cl iotf ($)such n [B1 CN ' and if 0 PCN' where F26 ,n(o,o) = [aras
E
PT and B
t:
then F
PIV"
Z7
(o,B)
E
E
Pr' then
611 cN ,
Pt' where F
27
(Cl,B)
[0. to BIt unless ß is not of the form ([6]Adv)[[y]V(n)],
PIV/IV'
P
and
t/t
0
jotf [BI E:
v,
and if ]t Z9 E:
PI V' , where F (a,S} = 30
E
PCN ' then FZ3 (0.,S)
533.
If a, ß
t:
Pr' then F (o.,S) 31
E
Pr' where F
534.
If q"
I.j.>
E
Pt' then F (.'W) 3Z
E
Pt' where F
535.
If a. ß
t:
P I V" then F33 {a,ß)
then F22 ,n,y
IV ,where F" ,n,y (4)) = [oliV
PCN'/CN and S
t:
[Cl and S] IV'
and Cl iotf [slV(o), then FZ1 (a) [0.]
If 0.
if Cl is not of the form [S]V. then
[aliV'
Pt and. iotf [hen]T to [[Cl]IV]IV" P
E
ct
E
= [BI
6 to y, then F (a,4» 28
[o.S]IV/IV
' then FZO(o.)
where F (Cl) = 21 524.
E
530.
= [to[allv,JHIF
PIV/IV I T and S
where F I8 {a,S)
E
in which case F (a,s) = [Clto[[be ]VBIIV' J l Z7 t
P INF , where F I7 ,n,y{4»
519.
If 0.
=
If. E Pt and
"p[ .3x ["iy["S (z)J (y) ...... x=yJ & "p(x)]
; «s, ..xk[Many yO[R' (y,x) & Vp(y»))
2
always transTates as >..P>..xk(Vyo[R(y,x) j.
kind translates as >..Q>..xk[So(x) &
p (y)1J V
4
k.
such
o
&~R'
o
;
n.
in
and near
0
o[ Place' (z)
°
0.
E: PeN' and translates as a', then FS(cd translates
If u (. PeN , and translates as Cl', then F (a) translates as (1' 7
T9.
If a
P ,V and translates as Cl', then F (a) translates as a'
t:
8
If a
translates as (I', then k{(I) translates as k k >..x [5 (x) & Dis'(5 )«(1')1. Conditions: For every x in
there must
TII.
If
a member of BeN such that (a)
0.
++ R'
(xo, xk)l
T12_
113.
114.
"
, 'CN
aod
translates as
then F (a) translates as a' I
T).
and trans lates as (1', then F2 «(1) translates as (1' and B' respectIf o , PT /eN' and B E: PeN , and translate as
T4.
rf
"
0.',
T16.
ively, then F (a,B) translates as (1' (~B') 3
P 1V and translate as a' and B' resp_, a'(~B')
B'
«(1,B) translates
P'V'/IV and B
E
I2
E
3S (1'(~B')
P,v and translate as Cl' and
B'
resp.,
(Cl,B) translates as a'(~S')
If a E P'V'/IV' and S E P,v, and translate as (1' and
If a
resp.,
B' resp.,
F (a,s) translates as (l'('S') 13 E
P'V/t and 4>
E
Pt and translate as a' and 4>' resp.,
translates as o.'(~4>')
If a (. PTV/IV and SE: P 1V and translate as
(1'
and S' resp.,
then FIS(a,s) translates as a'(-S')
T17.
" , 'CN
"'
If a
l1
then F «l,4» I4
as 9«(1). If
F
then
Any expression of BA not translated above is in the
T2.
E
If " E P'V/IIIV and B E P 1V and translate as a' and
then F
11S.
domain of 9 such that if a is in its domain, (1 translates
PIVIIIV and B
f;
(l
then FIO(u,B) translates as
O[!fxo[T'(x) .... (I'(x)]] wnere T' is the translation of T (b) \fx°O[T' (xl
P 1V / IV and B (. P and translate as Cl' and B' resp., 1V
t:
9
The rule is as follm'ls:
°be
u' (zo»))
then F «(1,B) translates as u' (~B')
Cl
° 50
-+--+
If u (. PCN/CN and S E: P and translate as a' and S' resp., eN
then If
' and.;p' resp.,
T8.
respectively
We assume a lexical rule that maps object-Ievel nouns that are members of B to kind-level nouns that are also CN members of BeN. let us posit a function k which is of type «e , t>,..x Vz
If
as >"pvp(lXk\fz°O[R' (zO,/)
(wo, xk)]J
at' (y,z)]J
m.
T6.
translates as >..Q[>..x o[ Vz O[R' z, x k )
R' (xo, x k )}&.-{}-9wO["Q(w) I.
(
IfuE:Pc"and~E:Ptaodt rans I ate as
then F4 ,n,y(u,q,) translates as Ax~[(1'(X) & 4>'1
~'(50)(Q)1
(where ~' is the disjunction condition)
°
T5.
If
0
E
PTV and S
E
Pr and translate as Cl' and S' resp.,
tnen F (Cl,S) translates as I6 118.
(I'
(-S,)
If 4> E Pt and translates as 4>', then F as h Y[4>'] o
I7
,n,y(4)) translates
262 - Carlson
T19·
If
0
E PIV/INF an d B E P INF and translate as et' and B' resp.,
then F
HO.
Carlson
I4
F
(et,ß) translates as Cl'
I8
T2 J. T22.
If et E P,V and translates as a' • then
T23.
If et E P
1V
and translates as a'
AX i 3z s [R(z,x)
T33.
.
translates as
T36.
If a E PCN'/CN and 8 E PCN an d trans I a te as et' and B' resp.,
137.
If et E P
CN
' and translates as a', then F
T27.
a'
If et
24
25 ,n,rn
(a,cp) translates as Axo[a' (x) &
T38.
If Cl E PCN' and B E PT and translate as a' and B' resp., 26
,n(et,s) translates as AX °1 Cl' () x
If a E PT and ß
E:
P
1V
&
"yk B , (zo[R' (z,y)])
' and translate as a' and
B' resp.,
If U E P t/t an d "'+' E P t an d tc.nslate as et' and cp' resp., then F (et,CP) translates as et' (~cp') 28
131.
If 1/1,
v
resp., then
1jI'
P ,V ' and transJate as
Cl'
and B' resp., then
(a,B) trans lates as )"xi [u' (x) vB' (x)J BE PT and translate as
0:,
0:'
and B' resp., then
34 (et,B) translates as AP[a' (P) v ß' (p)]
If a
PT and cP
E
E:
Pt and translate as a' and use, etc.). On the other hand, for most people, it is strange to talk of some oriole as being very comman, as it is generally unknown as to whether there are lexically-entered subordinate kinds of orioles. It seems reasonable that a simi lar state of affairs holds concerning the relationship between such nouns as virtue and courage. Notice that we Cc.:ted
Moravcsik. J. M. E. 1970. et al. 1973, 263-285.
P