Remarks on Weak Cross-Over, Relational Nouns and Domain Restriction Daniel Blair University of Western Ontario 810 Whittier Place, NW Washington DC 20012 lcutler@earthlink.net Abstract. The recent interest in some of the phenomena traditionally associated with the context dependence of quantificational expressions (QPs) has centered around the idea that some constituents of a sentence might serve as the locus of domain restriction for QPs might be present but lack for overt manifestation. In this essay, one such argument – due to Stanley (2000) – is critically examined. Specifically, I will present a number of different kinds of constructions where the predictions of a theory based upon syntactically represented context variables are not confirmed. Introduction. There is perhaps no kind of expression within natural language which does not exhibit some sort of variability with respect to its interpretation. Cases where the nature of this variability is more or less rigidly predictable come from the domain of indexicals and demonstratives like ‘I’ and ‘that,’ but there are other plausible examples. In particular, quantificational expressions (‘QPs’) seem to exhibit at least some of the same general traits that other contextually variable expressions do. Stated informally, the exact denotation of a quantifier like ‘the tallest student’ seems to vary from one context of utterance to another. “The tallest student” can be understood in one context to denote the tallest student within, say, a small class of university students, understood on still other occasions to denote the tallest student within a given school 1 district. For many uses of quantificational expressions, the domain of the quantifier is understood as restricted just to certain sets of objects, even though no explicit restriction need appear alongside them. How the facts regarding the context sensitivity of quantificational expressions are to be captured within a semantic theory is a matter of some contention. There are at least two options. Perhaps the inexplicitness of the restrictions holds no semantic or syntactic secrets. Rather, domain restriction is a pragmatic process that essentially supplements, in one way or another, our understanding of what a particular utterance is intended to convey without necessarily telling us anything about what the sentence itself means. On the other hand, perhaps the silence is accidental and a more fine-grained analysis will disclose non-overt constituents to which we may localize the implicit restrictions. In this paper, I shall be concerned with one variety of this second option which has gained some notoriety recently, namely the view that the phenomenon of variability with respect to the interpretation of quantifiers can be explained by positing a particular syntactic formative which can be bound by other expressions or which may function like a demonstrative, picking up domain restrictions within a context of utterance. Since much of the plausibility and explanatory depth of theories along these lines depends upon the parallel behavior of the proposed null variables and explicit instances of pronouns, the discussion will be concerned with the degree to which the parallel holds across a series of different contexts. For concreteness, I shall be concerned with one set of arguments that purport to establish the existence of this kind of null formative, that of Stanley (2001, 2002) and Stanley and Szabo (2000). 2 The paper is structured as follows: The first section outlines the basic problem of quantifier domain restriction is outlined and a brief exposition of Nominal Restriction Theory (‘NRT’) is given. The second section presents in some detail the ‘Argument from Bindability’ that has been used to motivate the view that there exists a phonologically null syntactic constituent – what I shall call below a Null Domain Restriction (‘NDR’) within quantificational and nominal expressions which explains how it is that these expressions exhibit varying domains of quantification with respect to differing linguistic and non-linguistic contexts. The remaining fifth sections present various sorts of problem cases for this view. The upshot of this evidence is that NDRs, if they exist at all, would have to be quite unlike any other sort of expression, null or otherwise, which can be bound1. Let me stress that my argument is directed to the idea of there being a constituent of a sentence’s grammatical representation2 whose interpretation is keyed to the presence of items that serve the restrict domains of certain classes of nominal expressions. One could take the view that NDRs are constituents of semantic representations and leave the syntactic claim go. But that would rob the view of its interest as a defense of the claim that all truth conditional effects of context can be traced to elements in the Logical Form of a sentence. It would also take away from the motivation NRT draws from syntactic I shall not discuss at all any other sort of ‘implicit argument’ that has occasionally been posited with respect to certain constructions. Nominal Restriction Theory has also been critically discussed by Lepore and Cappelen (2002a, b), and Bach (2000), among others. However, the criticisms I make below differ from these: I am concerned wholly with the syntactic consequences of the NRT analysis. 1 2 More specifically: of the syntactic structure of a sentence over which semantic interpretation is defined, what is usually referred to within versions of generative grammar as ‘LF.’ See Hornstein, 1995 for an overview. This does not rule out that such variables may be present at other levels of syntactic representation, of course. 3 facts that are independent of the phenomena in question. If the guiding thought behind NRT is right, any formative like an NDR should be governed by syntactic constraints that are operative with respect to other syntactic categories and processes. My strategy below will be to press these analogies further, finding out where they break down. 1 The Context Sensitivity of Quantifiers and ‘NRT.’ Sentences like 1 - 4 present a challenge to many theories of quantification in natural language: 1. Every toy was broken 2. The tallest student got the highest grade 3. Fred’s parents decided to put everything under the Christmas tree at the last minute. 4. After everyone was gone, the thieves came out from their hiding places. The challenge consists in the fact that the most natural interpretation of 1 does not mean that every toy, near and far, was broken; rather, 1 is understood as asserting only that every toy within some contextually relevant domain of quantification is broken. Thus, the quantifier ‘every toy’ in 1 could be taken as equivalent to either 5 or 6 given the right circumstances, e.g. one in which John’s toys is the topic of conversation. (I use italics to designate domain restrictions.) 5. Every toy that belongs to John 6. Every toy I bought for Sam during the Holidays 4 Parallel considerations apply to 2 - 4. Although 2 could be used to make an assertion about the academic achievements of the tallest student in the universe, the most natural interpretation of 2 would take quantifier ‘the tallest student’ to have a considerably smaller domain, e.g., perhaps just the tallest student in a given class or university. For 3, the QP everything clearly does not denote absolutely everything in the universe. Rather, all that is meant, perhaps, is that every parent in the relevant domain of discourse put every present they had bought that year for Christmas under the Christmas tree in their home. Roughly, the challenge is to explain how the quantifiers in sentences like 1 and 2 are understood as restricted in the ways that the quantifiers in 5 and 6 are explicitly restricted, even when there is no overt restriction present3. There are a number of options. One could argue that restrictions like the italicized ones in 5 and 6 are present4, in some sense, in 1 and 2, and only lack for overt manifestation. That would imply that 1 and 7, for example, 7. Every toy that belongs to John is broken are truth conditionally equivalent within some range of contexts. Conversely, one might argue that quantifier domain restriction is a pragmatic phenomenon, a matter of how ‘Roughly’ because explicit restrictions themselves may contain context sensitive expressions which in turn would need to be restricted and so on. By making the restrictions covert, Stanley and Szabo avoid this particular regress; see S and S (2000). 3 4 Whether or not one can assimilate these cases to more well understood instances of syntactic ellipsis is doubtful; see Stanley’s discussion of Sellars in Stanley, 2002. 5 sentences like 1 and 2 are used and, as such, has no bearing upon the truth conditions of the sentences in which quantifiers appear. I won’t review all the options here. Assuming that 1 and 7 are truth conditionally equivalent within some context, one attractive answer is to hold that the QP ‘Every toy’ is actually more structured than it appears to be. The extra dimension of structure comes from an unpronounced or phonologically null element, which restricts the quantifier’s domain, acting in much the same way that the italicized relative clauses above do. In a series of articles, Jason Stanley5 (2002a, b) has developed a theory of quantifier domain restriction along these lines. In particular, he holds that an unpronounced domain restriction is actually part of the grammatical structure of the quantificational phrases whose domains they restrict. The restriction acts in some ways like a demonstrative pronoun, picking up a contextually salient domain from either the surrounding discourse or from an ongoing conversation. In particular, the structure Stanley posits for a quantifier like ‘every toy’ is the following: QP 3 Q NP Toy C(i) Every The item marked ‘C(i)’ is what I have called a ‘Null Domain Restriction’ (‘NDR’ hereafter) in light of the fact that it is supposed to both have the semantic properties of overt restrictions and lack phonological content. 5 and with Zoltan Szabo (2000) 6 Significantly, Stanley holds that all nominal expressions have an NDR attached to them as part of their syntactic representation. By contrast, other proposals for handling some aspects of context dependence that posit ‘implicit arguments’ – e.g. Partee (1989) have only concerned a subset of nominals and adjectives, e.g., ‘relational nouns’ like ‘enemy,’ ‘foriegner’ or ‘opposite.’ In a quite direct way, NRT implies that quantifiers themselves do not display contextual relativity. Rather, it is the nominal head whose interpretation may vary and it is this variability that is inherited by the entire QP. Given this facet of NRT, a nice parallel between syntactic and semantic structure obtains; for every truth conditionally relevant way of restricting a domain, a grammatical element can be found which is the locus of that way of restricting the domain. Domain restriction is then a species of demonstrative reference, at least in many cases. 2. The Argument from Bindability. It is crucial to NRT that an NDR can pick up its content – the domain restriction – from the linguistic context in which it appears in much the same that pronouns and anaphoric noun phrases can pick up their antecedents within the very sentences in which they occur. For example, in 8. In all of John’s classes, he fails three students ‘Three students’ may be understood as denoting three students in each of John’s classes. On NRT, this interpretation comes about through the presence of an NDR on ‘student’ being bound by the sentence initial prepositional phrase: 7 9. In all of John’s classes, he fails three studentsC(i) The apparent parallels between contextual variation of this sort and the binding of pronominal elements by quantifiers is one of the main arguments for an approach like NRT, on which such parallels fall out as part of the general case. Approaches which do not posit such variables within the syntactic structure upon the semantic properties of a sentence are determined have no clear way of accommodating such cases, a point stressed by Stanley6. The availability of binding and of linguistic antecedents implies that the same conditions that apply to the binding of linguistic items like pronouns ought to apply to the binding of NDRs by linguistic antecedents. The facts concerning 8 suggest that the generalization holds. For example, just as the pronoun he can be understood as referring to John, the NO student can also be bound by an element standing in roughly the same structural relationship to it. More precisely, the NO student contains a variable that is bound by the PP In John’s classes7. Strikingly, phenomena like the above can replicated across a diverse range of data, e.g., comparative adjectives, … etc8. 6 2001. 7 I leave to one side the issue of the exact composition of the domain restriction when it is determined by a linguistic antecedent. 8 Showing that context dependence exhibits some of the properties of binding and so should be taken as testifying to the existence of elements within the logical form of sentences supporting such relations also serves Stanley’s polemical purposes against advocates of ‘unarticulated constituents.’ I will not have anything to say about Stanley’s larger purposes in this paper. 8 In his 2001, the argument from binding is pushed even further9. Consider the following: 10. [Every boy]i loves hisi mother 11. ?His mother is loved by [every boy]i 12. ?Whoi does hisi mother like? In 10, but not 11, binding between every boy and the genitive pronoun his is natural. This would just appear to be a case of a quantifier’s having to stand in a structurally superior position – from which is may ‘c-command’ the position in binds - to whatever elements it may bind. Yet in 12, who seems to be in a structurally superior position to the pronoun it would bind and the sentence is anomalous. One way of explaining these facts is as follows: with respect to its relevant structural properties, 12 is identical to 11, not 10. The object quantificational phrase every boy in 8 moves covertly to a position structurally higher than that of the subject QP his mother. On the supposition that QPs must move to a clause peripheral position in order to determine their scope10, that means that this quantifier will have a scope that includes that of the subject QP his mother. In the process of moving however, the pronoun his is passed by or ‘crossed over.’ Moving past an embedded, co-indexed pronoun like this is called a ‘weak crossover’ (‘WCO’) violation. A similar movement 9 See pgs. 423 – 4 of 2002. 10 Bear in mind that that Stanley assumes, following many others, that scope and binding are directly related to syntactic structure and hence, that WCO effects are to thought of in syntactic terms. 9 occurs in 12. 11 and 12 thus have structural representations in which a moved element crosses over a co-indexed pronoun: 11a. [Every Boy]i [hisi mother loves ti] 12a. Whoi [ does hisi mother [like ti]] No such problem arises with respect to 10 since the QP subject does not move past the phrase that contains a co-indexed pronoun. Now consider 13 and 14: 13. [Every reporter]i was sponsored by a [local barC(i)] 14. ?[A local barC(i)] sponsored [every reporter]i Stanley claims that what we find in 14 is the same thing we find in 12. In 14, the only reading which appears to be available is the one in which the NP a local bar is not bound by every reporter. To obtain a reading for a local bar on which it – or, more precisely, the NDR on bar - would be bound by the QP every reporter, a quantifier would have to cross a position containing an element that it binds. But then if 14 really is a case of WCO, there must be something that is being crossed over, like the genitive pronoun in 10 12. If the NP local bar contains an NDR, then we have an explanation of the contrast. This gives us something like an ‘existence proof’ for NDRs.11 The existence of examples like the above have been discussed previously in the literature. But, to the best of my knowledge, Stanley is the first writer to link them explicitly to cases of quantifier domain restriction of the sort that features in the argument from bindability. And surely the existence of contextual variation which appears to mimic the binding of explicit pronouns suggests that there is a deeper affinity between the two cases. Positing a null variable, capable of occurring either bound or free, makes the analogous patterning of data straightforwardly explicable as instances of the same general phenomena. Moreover, pursuing arguments from WCO effects with relational words makes sound methodological sense: to the extent that the same kinds of anomalies appear in both pronominal as well as relational cases and is predictable from the syntactic relationships each item bears to its antecedent, the case for NRT is made stronger. 3. Some Problematic Cases The argument regarding WCO effects implies that any NP in a position structurally parallel to the subject in 11 will trigger grammatical anomaly if another item is both the source of its restriction and occupies a position structurally similar to the WH-phrase in 12 and has been extracted from a similar position. The case that is particularly interesting is the one in which the object QP takes wide scope relative 11 This argument for the existence of a null formative having certain binding properties was first made by Jim Higginbotham (1996). (Stanley and Szabo cite an early, manuscript version of this paper, although the differences are not important here.) The goal of Higginbotham’s argument – like that of Partee’s very similar 1989 - is more restricted than the use to which they put it. Given the broadness of Stanley’s hypothesis – all NPs carry an NDR – I am uncertain to what extent the examples above are problematic for Higginbotham’s view. 11 to the subject QP. It is the unavailability of that reading which seemingly causes the problem with example 14 above. Alongside Stanley’s example though are cases that tell against the existence of an NDR. Here are a few, using ‘relational NPs12’ and adjectives: 15. A grandparent accompanied every student to his graduation. 16. A favorite story helped put every toddler to bed. 17. A friend helped every student move out of the dorms. 18. A feared competitor made every boxer to train harder. 19. An imminent deadline forced most reporters to cut lunch short In each of the above sentences, the subject QP can be understood – and is most naturally understood – as restricted by the object QP, which also has scope over it. Thus, in 15 the domain of grandparents that is relevant is the domain of grandparents of the students who are graduating. Supposing that the restricting QP every student takes wide scope relative to the subject QP a grandparent, these sentences should be anomalous in the same way that Stanley’s example is from above. That is, what we should have is the following: …..[every student]i …[.... a grandparentC(i) …[....ti 12 See Mitchell, 1986 12 parallel to the configuration of 11a. The same applies to 16 - 19. But, contrary to the prediction of the theory of NDRs, these sentences seem fine on the intended interpretations. Notice that there is a contrast with parallel sentences whose subject NPs contain overt pronouns bound by the object QP (the a examples) or with sentences featuring a potentially binding WH-phrase which overtly move past a subject NP containing a pronoun, the classic configuration for WCO effects: 15a’ b. 16a’ b.’ 17a’ b.’ 18a’ b. ?Hisi grandmother accompanied every studenti to graduation *[Which student]i did [hisi grandmother] accompany ti to graduation? ?Hisi favorite story helped put every toddleri to bed. *[Which toddler]i did [hisi favorite story] help put ti to bed? ?His friend helped every student move out of the dorms. *[Which student]i did [hisi friend] help move ti out of the dorms ?His competitor made every boxer train harder *[Which boxer]i did [hisi competitor] make train harder 19a.’ ?His deadline forced every reporter to cut lunch short b. *[Which reporter]i did [hisi deadline] force to cut lunch short. 13 Although judgments are a bit delicate here – hence the term ‘weak cross over13’ – there does seem to be some sort of contrast between 15 – 19 and 15’ – 19.’ NRT predicts that all of these examples ought to pattern alike. They don’t. Another class of examples show much the same thing: 20a. [Impending court appearancesC(i)] explained the absence of [every star player]i from the line-up. b. ?[Hisi court appearance] explained the absence of [every star player]i from the line-up. 21a. Approaching police sirens tipped off none of the hackers that time was running out. b. ?[Hisi mother] tipped off [none of the hackers]i that time was running out. In 20, the NP impending court appearances can be understood as restricted by the QP every star player. 20b shows that binding a pronoun in a comparable position is problematic. In 21a, approaching police sirens can be understood as denoting police sirens that are approaching the hackers. Again, we find the same contrast with respect to the adjective and a pronoun within a subject NP in 21b that we found in 20b. At the very 13 Although all speakers consulted registered the same contrasts between the a and b examples. 14 least, these contrasts tell against the symmetry between instances of binding of pronouns and the relation between relational nouns and their antecedents14. 4. More Cases Given the above facts, what is striking about NDRs, if they are syntactic constituents of NPs, is their exceptional behaviour with respect to binding. In fact, we can make a stronger point. In the above cases, what we found was that movement past a supposed NDR did not entail a WCO violation, contrary to the expectation of the theory. Next, we shall look at cases in which binding would be impossible on normal assumptions: the potential binder is not in a position to bind the potential bindee quite apart from any issue about context sensitivity. Nevertheless, the bindee can still be understood as restricted by the frozen item. To begin with, consider the following examples, where only the low scope reading is possible for the object quantifier every girl: 24. Some man expected to find every girl at the party Some man Every girl *Every girl Some man There are other mechanisms that could be used here besides syntactic movement to generate the right readings, e.g., the choice functions discussed by Yoad Winter and others. See Winter (2001). But NRT seems to be committed to scope and binding being determined through syntactic mechanisms. Recall that Stanley wants to support the thesis that effects of contextual variation on the truth conditions of sentences can be traced to elements within the logical form of a sentence. In the present case, this would imply that binders for choice function variables be present within the syntactic structure of a sentence. Although this may be a plausible view on some conceptions of what the syntactic structure of a sentence is, it does not appear to be a conception of syntactic structure congenial to the aims of NRT. 14 15 In 24, the quantifier every girl in the subordinate clause cannot take scope over the quantifier in the super-ordinate, matrix clause. Explanations vary here, but, to a first approximation, it seems that when a QP is contained within a syntactic phrase that is ‘too complex’ in some sense, its ability to take scope outside that phrase is compromised. Similar phenomena appear with respect to the extraction of WH-phrases: 25a. b. Why did John say that Mary left early Why didn’t John say that Mary left early 25a can be understood as a request for the reason for John’s having said that Mary left early or it can taken as a request for the reason that Mary left early, according to John. Both readings are possible. In 25b, only the first reading possible. The difference seems to be due to the interfering factor of negation. Similar facts hold for scope15: 26a. Some thief tried to rob every store last night 15 Notice that binding is out independently in these cases, since movement past a coindexed pronoun would provoke a crossover violation. Cases in which negation appears to block extraction have been much discussed over the last decade alongside other cases of so called ‘weak islands.’ See Ross, 1984, Szabolsci and Zwarts 1997 and Honcoop 1998 for discussion. 16 b. Some thief didn’t try to rob every store last night * In 26a, both scope orders are possible, although, when there is an intervening negation, the inverted reading is no longer possible. This suggests the following strategy for testing NRT: if the binder of a NDR is ‘trapped’ in a position from which binding is not possible, then it ought not to be possible for the putative NDR to be understood as bound by the trapped item. If the QP in the matrix clause is understood as restricted by the QP in the lower clause, then whatever it is that explains the restriction cannot be an instance of binding since binding requires that the binder be in a hierarchically superior position to the bindee. With this in mind, consider the following sentences: 28. Some manager managed to deceive every shareholder in the company 29. Some manager didn’t manage to deceive every shareholder in the company * 17 In 28, either scopal order seems possible. But in 29, negation appears to block the ability of the object of ‘deceive’ to take scope over the subject QP. Negation appears to freeze the downstairs QP in place. Note that, whether or not negation is present, it does not appears to be possible for a pronoun in the subject QP to be bound by the object of ‘deceive:’ 30. ?His hand-picked manager (didn’t) manage to deceive every shareholder in the company Suppose then that 29 is interpreted in the only permissible scope order, i.e. with ‘some manager’ having scope over ‘every shareholder.’ Keeping 30 in mind, what we would expect to find is that a variable within the nominal ‘manager’ can’t be bound by ‘every shareholder’ without engendering anomolous reading similar to what we found with respect to 14 above involving relational nouns. Recall that it was cases like that that seemed to reveal the presence of a null phrase. All the same though, it appears that ‘manager’ can be interpreted as ‘manager of the shareholder’s company’ without anomaly. Note that this is so even negation doesn’t block extraction the object QP. As 30 shows, with or without negation, binding into the subject QP by the downstairs QP shouldn’t be possible. The lack of a parallel here between binding and relational nouns thus tells against assimilating the two cases. One could argue that the interpretation of the NDR in the subject nominal is set independently in this case and so there is no substantial problem for NRT. The idea would be that the reason we don’t find a WCO-like violation here is that the NDR on 18 ‘manager’ takes on its value independently of the value of ‘every shareholder’ as, indeed, it must if a WCO violation is to be avoided. But this sort of defense comes at a price. In particular, one wonders why it isn’t available for every case, in particular for the case involving relational nouns that motivated the argument from WCO effects. Why doesn’t similar discourse level binding of an NDR rescue those cases? More importantly though, this sort of reply robs NRT of one of its most attractive features: binding of NDRs ought to pattern with binding of pronouns; where we find a problem in one case, a parallel difficulty ought to show up in the analogous case. Thus, both 28 and 29 ought to aptern with 30. The reply we are considering now would undercut NRT precisely at its strongest point, for it seem to say that binding violations of NDRs are rescueable by pragmatic factors. What principled reason then can be given for saying that all cases of alleged bound domains are governed by ‘pragmatic factors?16’ 5. Nominal versus Quantificational Restriction. Let us consider another aspect of NRT that is every bit as radical as the claim that there is a special null formative that is the locus of the contextual sensitivity of QPs. This is the claim that quantifiers themselves exhibit no contextual variability. Rather, the contextual variation we find with respect to QPs derives from their nominal complements. The strong reading of this claim is that all nominals, and not just those appearing with explicit determiners, have an NDR and hence are liable to contextual variation. There seem to be contexts though in which this claim would need to be qualified, perhaps by making the addition of an NDR optional in some 16 These last paragraphs were prompted by the comments of an anonymous referee, for which I am grateful. 19 respect. Left without some such qualification, the thesis would appear to allow for readings of nominal expressions that aren’t possible. For example, consider what such a thesis would entail for ‘bare plural17’ expressions like the NP ‘elephants’ in 31. John likes elephants This sentence does not seem to have a reading on which John would like only some particular set of elephants. Perhaps kind denoting terms like these are exceptions in some sense, but one would like to know why exactly18. As mentioned above, the generality of NRT dictates that quantifiers themselves do not have varying domains. Rather, it is the nominal head that does. This makes the predication that, e.g., conjoined NPs ought to be capable of having different domains. Thus, for example in 32. After leaving, the boys and girls did their home-work. NRT should allow both nominals to take distinct antecedents, as seems plausible in many cases. Perhaps we are discussing the studying habits of the girls from a local private school and the boys from a nearby high school. But in other cases, it does seem as 17 Cf. Carlson, 1977 18 One might say, in the spirit of the claim that NDRs are akin to incorporated items and that the presence of the plural ‘-s’ affix interferes with the ability of the NDR to incorporate onto the nominal head of the expression much as other null morphemes can’t be incorporated onto a word in the presence of an overt morpheme. We shall look at this claim in more detail below, but here it should be noted that the presence ‘-s’ itself can’t be blocking the availability of contextually sensitive readings, as attested by the examples of nominals like ‘the boys’ which do allow such readings. 20 though the boys and the girls have to belong to the same domain; it is hard – although perhaps not impossible - to see how one might be talking about the boys who left and the girls who stayed behind for example. More cautiously, although one can clearly imagine cases in which the domains do vary, if one of the nominals is restricted by a linguistic antecedent, the other one must be as well. One wonders why that is the case on NRT, since that theory would appear to allow for restrictions to vary in their reference as much as a demonstrative would. Still more troublesome are cases like the following: 33. Every boy in the garage and every boy in the basement formed a band 34. *Every boy and every boy formed a band. One wonders why 34 cannot have a reading like that in 33. It is very hard to see how there could be a nontrivial reading of 34 in this case. It is true that some coordinated QPs allow for independent domains, e.g., the following sentence does not seem especially unusual: 35. Some man and some man met in the park But it is curious why such a reading is out on 34 if domain sensitivity is solely a matter of the nominal and is not a matter of the quantifier or its properties19. For much the same Consider also in this light cases of ‘Determiner Sharing:’ i. After the graduation ceremony, the girls will drink whiskey and boys drink gin 19 21 reason, it seems like 36a has, to the extent that it is interpretable, the same interpretation as does 36b even though, on NRT, there ought to be an interpretation of 36a which is equivalent to that of 36c: 36a. Few policemen and few policemen respect each other b. Few policemen respect each other c. Few policemen from the 53rd precinct and few policemen from the 109th respect each other. The coordinated subject DP in 36a seems, at best, redundant and it is hard to see how 36a and 36c could have the same truth conditions. 7. Conclusion. In this essay, I have explored the idea that variables of the sort that have been posited to account for some of the facts concerning relational nouns and adjectives can be extended to the interpretation of quantifiers are constituents of syntactic representations. Despite some suggestive evidence for this position, the facts described throughout this essay indicated that such constituents behave in ways that seem quite exceptional. This does not, by itself, rule out the presence of such variables within the In this case, the NP boys appears without an overt determiner, although it seems to ‘share’ one with the girls. What’s remarkable about these contexts is that, if one understands the girls in question to be restricted to, say the ones that just graduated, it is hard not to apply the same restriction to the boys in question. In these cases, the nominal heads seem to have the same restriction. What both nominals seem to have in common is that they are tied to a single determiner, in some sense. This suggests that that domains restriction is not as independent from quantifier as one would expect on the NRT view. For more determiner sharing, see McCawley 1993, Johnson 1996, Lin 2000 22 syntactic structure of a sentence, although it does prompt serious questions about their nature. Thanks to Howard Lasnik, Ernie Lepore, Paul Pietroski, Sam Wheeler, Jason Stanley and the comments of two anonymous referees for comments on earlier versions of this paper. References: Bach, Kent: 2000, ‘Quantification, Qualification, and Context; a reply to Stanley and Szabo. Mind and Language 15.2, pp 262 - 83 Carlson, Gregory: 1977, Reference to Kinds in English PhD Dissertation, University of Massachusetts at Amherst Higginbotham, Jim: 1996, ‘A Plea for Implicit Arguments,’ in H. Bennis, et al, (eds.) Atomism and Binding, Kluwer Academic, Dordrecht, The Netherlands, pp. 183 – 203 Honcoop, Martin: 1998, Dynamic Excursions on Weak Islands Doctoral Dissertation, Leiden University. Hornstein, Norbert: 1995, Logical Form Blackwell’s, Oxford 23 Johnson, Kyle: 1996, ‘In Search of the English Middle Field,’ unpublished ms., University of Massschusetts. Lepore, Ernie and Herman Cappelen: 2002a, ‘Insensitive Quantifiers,’ in J. K. Campbell, et al., (eds.), Meaning and Truth, pp. 197 – 213, Seven Bridges Press, New York Lepore, Ernie and Cappelen, Herman: 2002b, ‘Indexicality, binding, anaphora and a priori truth,’ Analysis 62.4, pp 217 - 81 Lin, Vivian: 1999, ‘Determiner Sharing,’ MIT Working Papers in Lingusitics 33, pp 241 - 77 McCawley, James: 1993, ‘Gapping with shared operators,’ BLS 19. Mitchell, Jonothan: 1986, The Formal Semantics of Point of View PhD Dissertation, University of Massachusetts. Amherst, MA: University of Massachusetts, GLSA Partee, B. H.: 1989, ‘Binding Implicit Variables in Quantified Contexts,’ CLS 25, pp 342 – 65 Ross, John R: 1984, ‘Inner Islands,’ Proceedings of BLS. Berkeley Linguistics Society, University of California, Berkeley 24 Stanley, Jason: 2002, ‘Nominal Restriction,’ in Peters and Preyer (eds.) Logical Form and Language, pp. 365 – 88, Oxford University Press, Oxford Stanley, Jason: 2000, ‘Context and Logical Form,’ Linguistics and Philosophy 23, 391 – 434 Stanley, Jason Stanley and Szabo, Zoltan: 2000, ‘On Quantifier Domain Restriction,’ Mind and Language 15.2, 219 – 261 Szabolsci, Anna & Zwarts, Frans, (1997), “Weak Islands and an Algebraic Semantics od Scope Taking,” in, A. Szabolsci, ed., Ways of Scope Taking, 217 – 262. Dordrecht: Kluwer Winter, Yoad: 2001, Flexibility Principles in Boolean Semantics, MIT Press, Cambridge 25