Which van die dingen?! * Bert Le Bruyn

In this paper I present a semantic and syntactic analysis of the Dutch van die N (of those N) construction and answer the following questions: (1) How is it possible that this PP-like expression can appear in full argument position; (2) Where does its idiomatic interpretation come from; (3) Why is it that only demonstratives can appear in this construction. The argumentation depends on Chierchia’s type-shifting analysis, blocking and economy.

0.Introduction The topic of this paper is the semantics and syntax of the Dutch van die N (‘of those N’) construction. 1 This construction, which has received some attention in the literature (see especially de Hoop, vanden Wyngaerd & Zwart 1990, de Hoop 1998, Oosterhof 2005a, b), is exemplified in (1): (1)Jan kocht van die kleffe koekjes. John bought of those doughy cookies John bought doughy (you know what kind I mean) cookies. Two properties stand out. The first is that van die N can be used in full argument position despite its PP appearance. The second is that its interpretation you know what kind I mean has an idiomatic flavour that presumably cannot be derived compositionally. In the course of this paper other properties will be highlighted when relevant for the argumentation.

1.The proposal in a nutshell In this section I will briefly introduce the two claims I make to account for the properties presented in the introduction. Please note that these claims will be worked out in the subsequent sections. My first claim is that van die N is part of an underlying full partitive construction. This claim will be worked out in section 2. For the moment I take for granted that full partitives can be represented as shown in (2) involving an upstairs noun deleted at PF. The way I assume the van die N construction fits into this structure is shown in (3).

*

Thanks are due to Henriëtte de Swart, the participants of SiNIV (especially Helen de Hoop and Louise McNally), Joost Zwarts, Albert Oosterhof , Huib Kranendonk, Ingmar van de Beld, Ewa Rudnicka and two anonymous reviewers. Although I have tried to implement as much as possible their comments and suggestions to improve the contents and readability of this paper I know that I have not fully succeeded in this task. 1 The demonstrative determiner die is a dummy for all demonstratives.

1

(2)

three of those books2

(3)

van die boeken of those (you know what kind I mean) books

(X =X is deleted at PF; e = empty) In (3) I make two assumptions. The first is that there is no upstairs determiner.3 This should not come as a surprise. Indeed, as can be seen in the full partitive structure in (2), there is no reason to assume that the determiner is PF-deleted together with the noun. The second assumption I make is that the upstairs noun is plural. The reason for this assumption is that bare plurals but not bare singulars are acceptable in Dutch in argument position. Note moreover that this provides a straightforward explanation for the aspectual properties of van die N, because bare plurals but not bare singulars or DPs can combine with foradverbials. (4)Hij heeft uren van die koekjes gegeten. He has hours of those cookies eaten. He ate (you know what kind I mean) cookies for hours. My first claim accounts for the first property alluded to in the introduction, viz. that van die N can appear in argument position despite its PP appearance. It moreover presupposes a syntax and semantics – viz. the ones of the full partitive – that will be at the basis of my account of the second property alluded to: the idiomatic you know what kind I mean. This brings me to my second claim. My second claim is that the downstairs DP in the van die N construction is a taxonomic DP. 4 Taxonomic DPs are DPs that refer to a subkind or subkinds of the kind corresponding to their noun (Krifka 1995) . The taxonomic aspect of the DP accounts for the kind part in the translation in (1). The fact that the taxonomic reading is obtained by a demonstrative will be shown to be at the basis of the you know what…I mean part. It will be argued that this follows 2

For ease of presentation and to save space I will use English examples wherever the argumentation doesn’t crucially depend on the Dutch version. 3 Nothing crucial depends on the choice whether bare plurals have an empty D or are simple NPs. 4 For clarity note that I do not make the same assumptions for full partitives in general.

2

from more general properties of the demonstrative. This second claim will be treated in section 3. In the sections to come I will defend the two claims introduced in this section. I will furthermore show that my analysis accounts for the fact that demonstrative taxonomic DPs are the only ones found in this kind of construction in Dutch (i.e. when either changing the reading of the demonstrative DP or replacing the demonstrative D by another D (on any reading) the same construction can no longer appear in full argument position). The most important contribution of this analysis compared to previous ones such as de Hoop, Vanden Wyngaerd & Zwart (1990), Oosterhof (2005a,b) is that it shows that the van die N construction is a completely regular phenomenon and that demonstrative DPs are the only ones that are semantically predicted to be able to appear in this construction.

2.The underlying full partitive construction 2.0 Introduction In the introductory section I presented two properties of the van die N construction that were in some way surprising. In this section I will account for the first, viz. the fact that despite its PP appearance the van die N construction is acceptable in full argument position (henceforth the PP-problem). In 2.1 I will argue that the PP-problem can best be accounted for under the assumption that the van die N construction is in fact a full partitive. In 2.2 and 2.3 I will respectively propose a syntax and a compositional semantics for full partitives. A summary and some concluding remarks will be presented in 2.4.

2.1 Why resort to the full partitive ? There are two ways to solve the PP-problem. We can either accept that the van die N construction is a PP and that it can exceptionally occur in argument position or we can assume that there is an underlying DP structure. Both options will account for the facts but the first will have little explanatory force if we cannot link the acceptability of the van die N construction to other PPs occurring in argument position. I will show that this condition is not met and conclude that the second option is preferable. The underlying structure that forces itself is the full partitive. One could in principle argue that the fact that van die N is a PP that can occur in argument position is not surprising. Indeed, as shown by the following examples other PPs share the same property (examples taken from Huddleston & Pullum 2002): (5)She wrote over fifty novels. (6)Under the mat is the place where we used to leave the key for the boys. (7)They won’t consider after Christmas, of course, to be soon enough. 3

It is however not difficult to demonstrate that the van die N construction cannot be linked to any of these PPs. Indeed, PPs that are known to occur in argument position come in three categories (cf. Huddleston & Pullum 2002): PPs containing a quantified DP (over fifty novels), locative PPs (under the mat) and temporal PPs (after Christmas). Van die N belongs to neither of these and would therefore constitute an exception. Another fact increases the exceptional PP status of van die N. As argued by de Hoop, Vanden Wyngaerd & Zwart (1990) van die N can be preceded by a variety of determiners without losing its typical interpretation. (8)Jan at vier van die koekjes. John ate four of those cookies John ate four of those (you know what kind I mean) cookies. Given that PPs in general and the PPs in (5-7) in particular do not combine with determiners the exceptionality of van die N as a PP increases once more. I conclude that the first option – viz. to analyze the van die N construction as a simple PP – is not defensible. A more elegant and more explanatory solution would be to link the structure in (8) to the Dutch full partitive which has the same surface structure. An example is given in (9). (9)Jan at vier van de koekjes. John ate four of the cookies. Two more analogies between the two constructions make this analysis worth exploring. The first is that the determiners banned from the upstairs determiner position of full partitives are also banned from the upstairs determiner position of the van die N construction: Van die N (10)*een/*desing,pl/*elke van die kleffe koekjes a/thesing,pl/every of those doughy cookies Full partitive construction (11)*een/*desing,pl/*elke van de koekjes die op de tafel lagen a/thesing,pl/every of the cookies that were lying on the table The second analogy concerns the possible range of determiners that can appear in the downstairs DP position of full partitives and that can be commuted for the demonstrative DP in the van die N construction. We know that indefinite determiners are banned from the downstairs DP position of full partitives (the partitive constraint of Jackendoff 1977). The same restriction applies to the DP in the van die N construction.5 5

This analogy might look suspicious in the sense that the demonstrative DP in the van die N construction can only be replaced by another demonstrative DP. The main problem is then that

4

2.2The syntax of full partitives In this subsection I will present the syntax I assume for full partitives in general and the van die N construction in particular. 6 It is not my aim to give a complete syntax but I think it is important to be clear about what I minimally assume. The surface realization of the full partitive three of the books leads me to assume that we minimally have the following structure:

This syntax accounts for the surface structure of full partitives and can serve as input for a compositional semantics (see Barwise and Cooper 1981, Farkas 2002). Nevertheless it is not entirely satisfying. The most important problem is that it predicts that all determiners are able to appear in the upstairs D position. This leaves the constraint on upstairs determiners in (11) repeated below as (12) for the semantics to account for: (12)*een/*desing,pl/*elke van de koekjes die op de tafel lagen a/thesing,pl/every of the cookies that were lying on the table There are three reasons why this is not desirable. The first is that the proscribed determiners don’t have a clear-cut semantic profile. Indeed, they are singular, plural, indefinite, definite and quantificational. The second reason is that these determiners become acceptable when followed by a noun despite the fact that the semantics does not change: 7,8,9 it cannot be replaced by other definite DPs such as those introduced by the definite article. I will however show in section 3 that this type of DPs can be ruled out on independent grounds. 6 The argumentation in this subsection can be found in a slightly different form in Jackendoff (1977), Hoeksema (1996) and Sauerland & Yatsushiro (2004). Similar proposals are found in Dean (1966), Jackendoff (1977), Abney (1987), Cardinaletti & Giusti (1992), Hoeksema (1996) and Chierchia (1997) (survey partly based on Cardinaletti & Giusti 2006). 7 One might object that the Dutch variants of the books of the books and the book of the books aren’t acceptable. This however has nothing to do with the syntax but with economy considerations in the former and presupposition (uniqueness) failure in the latter case. It suffices to modify the upstairs noun to make both acceptable. 8 λx(ONE(x)&BOOK(x)&≤[x, ιλx(PLURAL(x) &BOOK(x))]) is truth-conditionally equivalent to λx(ONE(x)&≤[x, ιλx(PLURAL(x) &BOOK(x))]) 9 The validity of this argument has recently been questioned by Martí i Girbau (2003). The problem she points out is that overt nouns can be modified while modification is impossible with covert nouns. The conclusion according to her is that the overt and the covert version have a different structure. I agree with her observations but note that the impossibility of modification is also present in other PF deletion contexts: *Sue read five books but John didn’t read one/two/… nice (see Perlmutter 1970). This is in line with what I predict.

5

(13) een boek van de boeken a book of the books The third reason is that the same determiner restriction applies in PF deletion contexts:10 (14)Sue read all these books, but John didn’t read all/most/three/those books. (Sauerland & Yatsushiro 2004, henceforth S&Y) (15)*Sue read all these books, but John didn’t read the/every/a book(s). (S&Y) In light of what precedes I propose to adopt the structure in (2) containing an upstairs noun that is deleted at PF. As a consequence (16’) will be assumed to derive from (16): (16)[DP three books [PPof [DP the books]]] (16’)[DP three books [PP of [DP the books]]] Note that to guarantee recoverability I will further assume, at least for the van die N construction, that the upstairs noun can only be deleted at PF if both nouns have the same extension. I leave aside how this referential identity should be rendered syntactically.

2.3 The semantics of full partitives In this subsection I will briefly present a compositional semantics for the full partitive construction. I postulate that the partitive constraint (Jackendoff 1977) applies and hence that only definite DPs are allowed in the downstairs DP position. I furthermore assume that definite DPs on their non-generic readings involve application of the ι-operator and that numerals as well as number inflection are analyzed as predicative conditions. I will present the semantics I assume by applying it to the following example: (17)three books of the books λx[BOOK(x)] λx[PLURAL(x)] λx[PLURAL(x)&BOOK(x)] λP[ιP] ιλx[PLURAL(x)&BOOK(x)] λx[≤(x,ιλx[PLURAL(x)&BOOK(x)]] λx[BOOK(x)]

book -s books the the books of the books book

10 More recent syntactic work on ellipsis proposes a feature approach to account for these facts (see especially Lobeck 2006). The crucial point is that we are not dealing with semantic constraints.

6

λx[PLURAL(x)] λx[PLURAL(x)&BOOK(x)] λx[PLURAL(x)&BOOK(x)& ≤(x,ιλx[PLURAL(x)&BOOK(x)] )] λx[THREE(x)] λx[THREE(x)& PLURAL(x)&BOOK(x)& ≤[x,ιλx[PLURAL(x)&BOOK(x)]]]

-s books books of the books three three books of the books

Given the structure I assume, there is nothing special about the semantics proposed.

2.4Summary and concluding remarks In this section I defended the claim that the van die N construction is in fact a full partitive. The argumentation was based on properties of the full partitive that were shared by the van die N construction. I further presented an explicit (limited) syntax and a compositional semantics for the full partitive that will serve as background for section 3. The following properties of the van die N construction can be accounted for based on the contents of 2.1 through 2.3: -

its syntax : [NP N [PP van[DP die[NP N]]]] its semantics : λx[[[N]](x)&≤[x,[[DP]]]] the restriction on its upstairs determiners: the van die N construction cannot be preceded by determiners such as de, een, elke one part of the restriction on its downstairs determiners: the van die N construction cannot have a downstairs indefinite DP

To conclude I will briefly go into the question why it is that not all languages that have a full partitive have developed a van die N construction. 11 The answer rests on the assumption that analyzing van die N as a full partitive instead of as a PP is an instance of (historical) reanalysis. 12 One of the necessary conditions for reanalysis is that the construction to be reanalyzed is potentially ambiguous. Applied to PPs such as van die N this condition can be interpreted as requiring PPs to occur in syntactic positions that not only allow PPs but also NPs. Dutch has such a syntactic position at its disposal: the object position of NP/PP selecting verbs. 13 The best known members of this class are verbs of bodily ingestion (eten ‘to eat’, drinken ‘to drink’):

11

With thanks to Helen de Hoop for asking me this at SiNIV with respect to English. The concept of reanalysis is taken from the grammaticalization literature. See especially Hopper & Traugott (1993). A famous example is the going to construction in English that – according to the authors – at first had the structure [[…be going][to N]] and was reanalyzed as […be going to V]. 13 For an analysis of this kind of verbs see Kupferman (1979). I assume the arguments he gives for the existence of this category of verbs in French carry over to Dutch. 12

7

(18) Jan at van die peren. Jan at van die peren John [VP[Vate] [PPof those pears]] / John [VP[Vate] [NPof those pears]] . John ate of the pears I’m pointing at / John ate of those (you know kind I mean) pears. Jan at peren. John [VP[Vate] [NPpears]]. John ate pears. The prediction that this condition makes is that languages which don’t have a similar environment will not be able to develop a van die N construction. The absence of the van die N construction in English can be attributed to this condition; verbs selecting both NPs and PPs are indeed a very marginal phenomenon in English and the class of verbs hasn’t been constant throughout history (cf. Visser 1963: 355-387, Hoeksema 1996: 16). It is important to note that the ambiguity condition is a necessary but not a sufficient condition for reanalysis. The other factors that play a role in the evolution of the van die N construction are yet to be discovered.

3.The downstairs DP 3.0Intro In the previous section I presented the semantics I assume for the van die N construction: (19) λx[[[N]](x)&≤[x,[[DP]]]] This cannot be the whole story though. Indeed, the specific (i.e. non-generic) reading of demonstrative DPs cannot lead to the intuitive interpretation rendered in the translation by you know what kind I mean. If the demonstrative DP referred to a contextually salient set of entities we would on the contrary expect an interpretation along the lines of you know which ones I mean. To remedy I take over Oosterhof’s (2005a,b) claim that the demonstrative DP in the van die N construction is kind-referring. I add to this that simple kindreference (i.e. reference to the kind corresponding to the noun) still does not lead to the right interpretation. The remaining problem is situated in the what kind part: this does not select the kind corresponding to the noun but a subkind of the kind corresponding to the noun. Under the assumption that reference to a subkind of the kind corresponding to the noun is the defining property of taxonomic DPs (Krifka 1995) I conclude that the downstairs DP in the van die N construction is taxonomic. The informal discussion based on the intuitive interpretation can be made concrete. Indeed, it can be shown that the van die N construction is truthconditionally equivalent to the downstairs DP (i) if the demonstrative DP refers to a contextually salient set of entities and (ii) if the demonstrative DP is kindreferring but not taxonomic. Under the standard economy principle that, all other things being equal, less complex expressions are preferred over more 8

complex expressions (i) and (ii) lead to the conclusion that the demonstrative DP in the van die N construction has to refer taxonomically. This economy principle will henceforth be referred to as the avoid complexity principle. In the remainder of this section I will defend the claims in (i) and (ii).

3.1 The demonstrative DP cannot be specific In this subsection I will defend the claim that if the downstairs DP of the van die N construction referred to a contextually salient set of entities, the resulting van die N would be truth-conditionally equivalent to the downstairs DP. Van die N with a specific demonstrative DP is then expected to be blocked by the demonstrative DP itself. I assume specific demonstratives – being definites – to involve application of the ι-operator. 14 The semantics of specific die N is then the following: (20)ιλx[PLURAL(x)&D(x)&P(x)] (D(x) is a tentative representation of the extra semantic restriction introduced by the demonstrative) The resulting semantics for van die N with a specific demonstrative is then as follows: (21)λx[PLURAL(x)&P(x)&≤[x,[ιλx[PLURAL(x)&D(x)&P(x)]]]] For (21) to occur in argument position it has to type-shift to an e or 〈〈e,t〉,t〉 type expression. Given that the van die N construction has no overt upstairs D I assume that the type-shift has to be covert. Two type-shifts come to mind: application of ι and ∃.15,16 The choice between the two is subject to the following considerations: (22) For any type shifting operation τ and any X: *τ(X) if there is a determiner D such that for any set X in its domain, D(X)= τ(X) (Chierchia 1998) (23) All other things being equal ι is preferred over ∃. This is due to the fact that ι is more meaning preserving: it does not add existential import. (This is the same argument as Chierchia 1998 uses to claim that the ∩-operator is preferred over ∃). The combination of these two considerations leads to the conclusion that the ιoperator will be preferred if there is no determiner available that has the same 14

The ι-operator is defined as follows for plurals : ι (PLURAL(x) &P(x)): the largest plurality of entities belonging to the set of Ps. (taken over from Chierchia 1998) 15 For an overview of generally accepted available type-shifts see Partee (1987). 16 For a definition of the ι-operator footnote 14. ∃ can be defined as follows for plurals: ∃P: λQ∃x[PLURAL(x)&P(x)&Q(x)]

9

function. Under the standard assumption that there is such a determiner – viz. the – one would initially expect ∃ to be preferred. In section 2.2 and 2.3 I however showed that the is independently blocked from the upstairs D position of the van die N construction. As a consequence ι becomes available again and will be preferred. The resulting semantics is the following: (24)ιλx[PLURAL(x)&P(x)&≤[x,[ιλx[PLURAL(x)&D(x)&P(x)]]]] This is truth-conditionally equivalent to the semantics of the downstairs DP: (25)ιλx[PLURAL(x)&D(x)&P(x)] (=20) Taking into account the avoid complexity principle I then conclude that van die N with a specific demonstrative DP is blocked by the demonstrative DP itself.17 (26)van die idiote cadeautjes (SPEC) blocked by die idiote cadeautjes (SPEC) of those stupid gifts those stupid gifts

3.2 The demonstrative DP cannot be a non-taxonomic kind-referring DP In this subsection I will defend the claim that if the demonstrative DP of the van die N construction were kind-referring but not taxonomic the resulting van die N would be truth-conditionally equivalent to the downstairs demonstrative DP. Van die N with a generic demonstrative DP is then expected to be blocked by the generic demonstrative DP itself.18 I assume kind-referring non-taxonomic demonstrative DPs to have the following semantics: (27) ∩P19 One might object that this corresponds to the kind-referring reading of Dutch bare plurals and Dutch kind-referring DPs introduced by the definite article. The observation is correct but it need not be an objection. In as far as kindreference is encoded on the DP itself and (27) represents the kind corresponding The attentive reader will have noticed that I have not treated the ∩-operator as a possible typeshifter. See however footnote 20. 18 Note that I am not convinced that demonstratives can refer to kinds without referring taxonomically. Given that I do not wish to make any claims about this I will however show that even if they existed they would still be blocked in the van die N construction. 19 The down-operator is defined as follows: ∩P : (For any situation/world s) λs ι Ps if λs ι Ps is in K, undefined otherwise (Ps is the extension of P in s), its inverse, the up-operator, is defined as follows: ∪d: (Let d be a kind. For any situation/world s) λx[x≤ds] if ds is defined, where ds is the plural individual that comprises all of the atomic members of the kind. The down-operator is taken over from Chierchia and Dayal’s work (see e.g. Chierchia 1998, Dayal 2004). 17

10

to P this is a defensible way of representing the semantics of any kind-referring non-taxonomic DP. The resulting semantics for van die N with a non-taxonomic kind-referring demonstrative is the following: (28)λx[PLURAL(x)&P(x)&≤[x,[∩P]]] For (28) to occur in argument position it has to type-shift to an e or 〈〈e,t〉,t〉 type expression. Given that the van die N construction has no overt upstairs D I assume that the type-shift has to be covert. Two type-shifts come to mind: application of ∩ and ∃. The choice between the two is subject to the following considerations: (29) For any type shifting operation τ and any X: *τ(X) if there is a determiner D such that for any set X in its domain, D(X)= τ(X) (Chierchia 1998) (30) All other things being equal ∩ is preferred over ∃. This is due to the fact that ∩ is more meaning preserving: it does not add existential import. (Chierchia 1998). The combination of these two considerations leads to the conclusion that the ∩operator will be preferred if there is no determiner available that has the same function. Under the standard assumption (cf. Chierchia 1998) that there is no such determiner one expects ∩ to be preferred. I assume accordingly that this is indeed the shift that takes place. The resulting semantics is the following: (31) ∩λx[PLURAL(x)&P(x)&≤[x,[∩P]]]] This is truth-conditionally equivalent to the semantics of the downstairs DP: (32) ∩P (=27) Under the avoid complexity condition this leads to the conclusion that van die N with a generic demonstrative DP will be blocked by the downstairs generic demonstrative DP itself.20 (33)van die idiote cadeautjes (GEN) blocked by die idiote cadeautjes (GEN) of those stupid gifts those stupid gifts 20

As in the previous subsection (see footnote 17) I have not considered all available type-shifts. In the previous subsection I did not consider the ∩-operator while in this subsection I have not considered the ι-operator. This omission can be overcome without further stipulations. It suffices to note that if the ∩-operator were applied to (21) it would give a semantics equivalent to (27) and if the ι-operator were applied to (28) it would give a semantics equivalent to (20). Both (27) and (20) correspond to less complex expressions which can therefore be taken to block the more complex van die N constructions.

11

3.3 The demonstrative DP can be a taxonomic generic DP In the previous subsections I defended the claims that (i) specific demonstrative DPs and (ii) kind-referring non-taxonomic demonstrative DPs are blocked from appearing in the van die N construction by themselves. In this section I will show that kind-referring taxonomic demonstrative DPs are not necessarily blocked. To do so I will first look more closely at taxonomic readings of demonstrative DPs. I assume that demonstrative DPs have two types of taxonomic readings. The first is shared by all determiners and stems from the shift from the standard domain consisting of individuals to the domain consisting of subkinds (cf. Dayal 2004). This type of domain can be represented as follows (demonstrated for the subkinds of lions):

I take the usual specific semantics of demonstratives to apply here (cf. Dayal 2004 for the same claim about definite articles): (34)ιλx[PLURAL(x)&D(x)&kind-of-P(x)] (D(x) is a tentative representation of the extra semantic restriction introduced by the demonstrative) This type of taxonomic reading – which will henceforth be referred to as the ιtaxonomic reading – is illustrated in (35) : (35)These animals are endangered. These kinds of animals, namely the cheetah and the lynx, are endangered. The second type of taxonomic reading, which – as far as I know – has not been treated in the literature before, is fundamentally different from the first. The initial motivation for positing a second type of taxonomic reading comes from the fact that (35) cannot only mean that more than one specific kind of animal is endangered but also that one specific kind of animal is endangered. Indeed, (36) is a correct gloss of (35): (36)This kind of animal, namely the cheetah, is endangered. The reading in (36) is entirely unexpected under the analysis in (34). It is moreover completely unclear how a plural can refer to just one subkind if we 12

take taxonomic reference always to involve a domain consisting of subkinds. To remedy I suggest we turn back to the standard domain consisting of individuals and look for other processes allowing plurals to refer to just one kind-like individual. This is not too difficult a task: standard kind-reference of bare plurals follows exactly this pattern. An analysis for demonstrative DPs along the lines of those of bare plurals can be given as follows: (37) ιλx[PLURAL(x)&D(x)&P(x)] Id(ιλx[PLURAL(x)&D(x)&P(x)])21 λx[PLURAL(x)&D(x)&P(x)] ∩ λx[PLURAL(x)&D(x)&P(x)] The only problem we encounter is that the ∩-operator cannot be applied directly to demonstrative DPs given that these are of type e. In (37) this is solved by applying the Id-operator first so that the demonstrative DP is turned into an 〈e,t〉 type expression.22 The crucial point to note is that we get taxonomic reference via application of the ∩-operator without losing the extra semantic restriction provided by the demonstrative. This type of taxonomic reading will henceforth be referred to as the ∩-taxonomic reading. Now that I have given an analysis of both types of taxonomic readings of demonstrative DPs let us look at how these fare in the van die N construction. From the discussion in subsection 3.1 it should be clear that the ι-taxonomic reading is blocked by itself in the van die N construction. In principle we expect the same for the ∩-taxonomic type. Indeed, given the semantics in (38) and the considerations in (29) and (30) repeated below as (39) and (40), we expect (38) to shift to (41) in order to be able to appear in argument position. The considerations are parallel to the ones made in 3.2. (38)λx[PLURAL(x)&P(x)&≤[x,[∩λx( PLURAL(x)&D(x)&P(x))]]] (39) For any type shifting operation τ and any X: *τ(X) if there is a determiner D such that for any set X in its domain, D(X)= τ(X) (Chierchia 1998) (40) All other things being equal ∩ is preferred over ∃. This is due to the fact that ∩ is more meaning preserving: it does not add existential import. (Chierchia 1998). (41) ∩λx[PLURAL(x)&P(x)&≤[x,[∩λx( PLURAL(x)&D(x)&P(x))]]] I assume that it is indeed ∩ that is applied. Consequently the resulting semantics would be truth-conditionally equivalent to the downstairs taxonomic DP and 21

The Id-operator is defined as follows for plurals : Id(x)=λx[x≤y] (taken over from Chierchia 1998). 22 In how far it is defensible to apply Id. in order to be able to apply the down-operator is yet unclear.

13

therefore blocked. There is however an exception: ∩ will only apply in generic contexts. In non-generic contexts ∃ will prevail. The reason is that in the opposite case there would be a sortal mismatch. Paraphrasing Krifka (2004): (42)Non-generic predicates expect ordinary objects, not kinds. This explains why van die N’s with a downstairs ∩-taxonomic DP are not entirely blocked by the downstairs DP itself. If this analysis is on the right track we expect the van die N construction to be blocked with kind-selecting predicates and to be acceptable with others. Examples like (1) (repeated below as (43) are sufficient to show that van die N’s are compatible with non-kindselecting predicates. Example (44), taken from Oosterhof (2005b), shows that van die N’s cannot combine with kind-selecting predicates: (43)Jan kocht van die kleffe koekjes (=1) John bought of those doughy cookies John bought doughy (you know what kind I mean) cookies. (44) De Hollanders hebben (??#van) die vogels uitgeroeid. The Dutch have of those birds exterminated The Dutch have exterminated those (you know what kind I mean) birds. The semantics we obtain is the following: (45)λQ∃x[Q(x)&PLURAL(x)&P(x)&≤[x,[ ∩λx( PLURAL(x)&D(x)&P(x))]]] One final remark whose relevance will become clear in 3.4.1 is that (45) allows us to verify that there can be no blocking by the upstairs silent bare plural. Indeed, the semantics of the upstairs bare plural is more general than that of the demonstrative DP. The crucial difference is the extra restriction ‘D(x)’ that stems from the reference to a subkind of the kind corresponding to the noun of the demonstrative DP.

3.4 Summary and conclusions In this section I identified two constraints on the downstairs demonstrative DP of the van die N construction: (46)The demonstrative DP in the van die N construction has to involve application of the ∩-operator. (cf. 3.1, 3.2, 3.3) (47)The demonstrative DP in the van die N construction has to refer taxonomically. (cf. 3.3)

14

These two constraints show that the demonstrative DP in the van die N construction has to be a ∩-taxonomic demonstrative DP.23 In the remainder of this concluding subsection I will do two more things. I will first show that the argumentation in 3.1 through 3.3 not only explains why the demonstrative DP has to refer taxonomically but why it can also be taken to explain the ban on non-demonstrative determiners. Afterwards I will suggest an answer to the question where the idiomatic meaning you know what … I mean comes from.

3.4.1The ban on non-demonstrative determiners In this paragraph I will show that non-demonstrative definite determiners will lead to similar blocking effects as those involved in the unacceptable readings of demonstrative DPs. I will show this for DPs introduced by the definite article24. Given that indefinite DPs are taken to be blocked by the partitive constraint (see section 2.2) I will be able to conclude that my analysis accounts for the fact that ∩-taxonomic demonstrative DPs are the only DPs that can occur in this kind of construction (i.e. van + DET +N). I will assume the following semantics for definite DPs: (48) specific: ιλx[PLURAL(x)&P(x)] (49) generic: ∩P (50) ι-taxonomic: ιλx[PLURAL(x)&kind-of-N(x)] (cf. Dayal 2004) (51) ∩-taxonomic: ∩λx[PLURAL(x)&P(x)]25 Note that the ∩-taxonomic reading is equivalent to the generic reading. I will therefore not treat it any further. If we introduce these expressions in the semantics of the van die N construction we obtain the following results: (48’)λx[PLURAL(x)&P(x)&≤[x,[ιλx[PLURAL(x)&P(x)]]]] (49’)λx[PLURAL(x)&P(x)&≤[x,[∩P]]] (50’)λx[PLURAL(x)&P(x)&≤[x,[ιλx[PLURAL(x)&kind-of-P(x)]]]]

When preceded by a determiner the van die N construction will not only allow for ∩taxonomic DPs to occur in its downstairs DP position. Indeed, as long as an existential interpretation is forced (by an indefinite determiner) and the upstairs bare plural has a larger extension than the downstairs DP I predict that the construction should be acceptable (at least as far as the semantics is concerned). The same applies to other definite DPs. 24 I assume with Chierchia (1997) that the part-relation induces a presupposition of plurality. This is taken to block singulars. 25 More elaborately: ∩(Id(ιλx[PLURAL(x)&P(x)])).

23

15

In the light of the argumentation in 3.1 and 3.2 we expect the following typeshifts to apply: (48’’) ιλx[PLURAL(x)&P(x)&≤[x,[ιλx[PLURAL(x)&P(x)]]]] (49’’) ∩λx[PLURAL(x)&P(x)&≤[x,[∩P]]] (50’’) ιλx[PLURAL(x)&P(x)&≤[x,[ιλx[PLURAL(x)&kind-of-P(x)]]]]

(ι>∃) (∩>∃) (ι>∃)

It can easily be verified that in the three cases the resulting semantics is equivalent to that of the downstairs DP. This however does not suffice to argue that all of these constructions are blocked by the avoid complexity principle. Indeed, parallel to the possibility of applying ∃ to the van die N construction containing a ∩-taxonomic demonstrative DP (cf. 3.3) we have to check what the resulting semantics of (49’) will be when ∃ is applied to it. The result is the following: (49’’) λQ∃x[PLURAL(x)&P(x)&≤[x,[∩P]]] It should not come as a surprise that this is truth-conditionally equivalent to the semantics of the upstairs (silent) bare plural when shifted appropriately. I can thus conclude that all possible readings of the downstairs definite DP lead to readings of the van die N construction that are truth-conditionally equivalent to (less complex) subparts of this construction. In light of the avoid complexity principle this leads to the conclusion that none of the readings of DPs introduced by the definite article lead to non-blocked readings when they are commuted to the demonstrative DP in the van die N construction. This explains the ban on this kind of DP.

3.4.2You know what…I mean In this paragraph I will sketch an answer to the question where the idiomatic you know what…I mean in the translations comes from. I claim that it is a pragmatic effect stemming from more general properties of demonstratives. In the language-philosophical literature on demonstrative DPs26 a recurring question is why demonstratives can be accompanied by a noun. The idea behind this question is that the indexicality inherent to demonstratives should suffice by itself to identify the referent. This has led people like Schiffer (1981) and Perry (1997) to assume that the only role played by the nominal is the pragmatic one of directing the hearer’s attention “to what the speaker is demonstrating with his use of ‘that’”(Lepore & Ludwig 2000). To see the implications of this point of view consider the following sentence: (52)That fly is annoying me. 26

I closely follow Lepore & Ludwig (2000:2) in their literature survey.

16

Under the point of view of Perry and Schiffer this sentence will have a truth value even if the object pointed at by the speaker is not a fly but a bee or a spider or even a horse. For communication to remain successful in this case the speaker will have to be convinced that the hearer knows what he is referring to. The resulting pragmatic implicature is that the speaker assumes that the hearer knows what he is referring to. I claim that it is this pragmatic aspect of demonstratives that has led to the idiomatic meaning you know what … I mean. The van die N construction is then the perfect expression to refer to something that is not easily describable: taxonomic reference allows to select a sufficiently vague noun that can be specified further by adjectives and even if the resulting vague description does not fit the referent this should not a priori be a problem. This explains the huge amount of Google hits for highly vague expressions that in fact refer to very specific things. Two examples will suffice to make this point: van die gekke dingen (2900 hits27) of those crazy (you know what kind I mean) things • van die kleine dingen (11700 hits) of those little (you know what kind I mean) things

•

4.Summary and conclusion In this paper I presented an analysis that accounts for two properties of the van die N construction: its ability to appear in argument position despite its PP appearance and the idiomatic you know what kind I mean interpretation. My analysis involved three claims: • • •

underlyingly the van die N construction is a full partitive the downstairs DP of the van die N construction has to be a taxonomic demonstrative DP involving application of the ∩-operator the you know what…I mean interpretation depends on a pragmatic implicature linked to the use of the demonstrative

The defence of the first claim was based on several analogies between the van die N construction and the full partitive (surface structure, constraint on upstairs Ds, constraint on downstairs Ds). The defence of the second claim was based on the semantics and syntax I proposed for the full partitive, Chierchia’s type shifting analysis and a standard economy principle. The third claim was based on considerations made in the language-philosophical literature on the demonstrative.

27

The number of hits was counted on September 8th 2006 and no distinction was made between simple PPs and the van die N construction. A random check of the examples revealed that approximately 90% was an instance of the van die N construction.

17

Four aspects of the analysis deserve special attention. The first is that it only involves independently motivated principles. The second is that the semantics is not only fully compositional but that it depends moreover on the presence of the upstairs bare plural that at first only seemed to have a syntactic function. The third aspect is that it is the first analysis that accounts for the generic/taxonomic constraint on the van die N construction. The fourth and final aspect is that it is the first analysis that at least sketches a way to account for the you know what…I mean part of the van die N construction.

References Steven Abney. 1987. The English Noun Phrase in its Sentential Aspect. MIT, PhD dissertation. Jon Barwise and Robin Cooper. 1981. Generalized quantifiers and natural language. Linguistics and Philosophy 4, 159-219. Anna Cardinaletti and Giuliana Giusti. 1992. Partitive Ne and the QP Hypothesis. In: Elisabetta Fava (ed.). Proceedings of the XVII Meeting of Generative Grammar. Turin: Rosenberg and Sellier. 121-141. Anna Cardinaletti and Giuliana Giusti. 2006. The Syntax of Quantified Phrases and Quantitative Clitics. The Blackwell Companion to Syntax. Oxford: Blackwell. Volume V. 23-93. Gennaro Chierchia. 1997. Partitives, reference to kinds and semantic variation. Proceedings of Salt VII. 73-98. Gennaro Chierchia. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339-405. Veneeta Dayal. 2004. Number marking and (in)definiteness in kind terms. Linguistics and Philosophy 27: 393-450. Janet Dean. 1966. Determiners and Relative Clauses. Unpublished manuscript, University of Geneva. Donka Farkas. 2002. Specificity Distinctions. Journal of Semantics 19: 213243. Donka Farkas and Henriëtte de Swart. to appear. Article choice in plural generics. To appear in Lingua. Helen de Hoop. 1998. Partitivity. GLOT International 3: 3-10. Helen de Hoop, Guido Vanden Wyngaerd and Jan-Wouter Zwart. 1990. Syntaxis en semantiek van de van die-constructie. Gramma 14, 81-106. Jacob Hoeksema. 1996. Introduction. In: Jacob Hoeksema. Partitives. Berlin: Mouton de Gruyter. 1-24. Paul J. Hopper and Elizabeth C. Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. 18

Rodney Huddleston and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.

− Syntax: A Study of Phrase Structure. Cambridge, Ray Jackendoff. 1977. X Massachusetts: MIT Press. [Chapter 5: NP Specifiers. 103-136] Ernest Lepore and Kirk Ludwig. 2000. The Semantics and Pragmatics of Complex Demonstratives. Mind 109: 199-240. Anne Lobeck. 2006. Ellipsis in DP. The Blackwell Companion to Syntax. Oxford: Blackwell. Volume II. 145-173. Nuria Marti i Girbau. 2003. Partitives: one or two nouns? Proceedings of the XXIX Meeting of Generative Grammar. Available at http://seneca.uab.es/ggt/Reports/GGT-04-6.pdf Albert Oosterhof. 2005a. Verbleekte partitieven: descriptieve, syntactische en semantische aspecten. Available at http://www.neerlandistiek.nl. Albert Oosterhof. 2005b. Dutch Bare Plurals, Faded Partitives and SubjectObject Asymmetry. Belgian Journal of Linguistics 19: 59-91. Barbara Partee. 1987. Noun Phrase Interpretation and Type-Shifting Principles. In: Jeroen Groenendijk et al. (ed.). Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers. Dordrecht: Foris. David Perlmutter. 1970. On the Article in English. In: Manfred Bierwisch and Karl Heidolph (ed.). Progress in Linguistics. The Hague: Mouton. John Perry. 1997. Indexicals and demonstratives. In: Bob Hale and Crispin Wright (ed.). A companion to the philosophy of language. Oxford: Blackwell. 586-612. Uli Sauerland and Kazuko Yatsushiro. 2004. A silent noun in partitives. Proceedings of NELS, 2004. Available at http://www.zas.gwzberlin.de/mitarb/homepage/sauerland/proceedings/nels04.pdf Stephen Schiffer. 1981. Indexicals and the theory of reference. Synthese 57: 43100. Fredericus Th. Visser. 1963. An historical syntax of the English language. Leiden: E.J. Brill. Robert Zamparelli. 2002. Dei ex machina. Unpublished manuscript, University of Bergamo. Available at http://wwwesterni.unibg.it/dsfc/pers/zamparelli/ling/bare-part.pdf

19