Demonstratives, definiteness effects and the type-token distinction 1

. Definiteness effects have been attested in the literature for some DPs complementing HAVE (= English have and their counterparts in other languages). In this paper, which focuses on English and Spanish, demonstrative DPs are shown to be affected by the type - token distinction in HAVE contexts: the internal argument of HAVE receives a type reading in these cases and rejects a to-ken interpretation. The “type restriction on demonstrative DPs” (TRD) is shown to follow from the need for narrow focus NPs complementing HAVE to receive a hearer-new reading, a well-known property of presentational structures. Type readings of demonstrative DPs are shown to meet this condition. Apparent exceptions to the TRD effect include (i) DPs in so-called “remainder contexts”; (ii) structures in which the relevant DP is the subject of a small clause (sometimes with a non-overt predicate); and (iii) structures containing a number of anti-assertive operators. The contexts in (i) are shown to be hearer-new, in spite of being definite. Those in (ii) and (iii) provide DPs which escape the narrow focus interpretation that gives ride to this variety of the definiteness effect.


Introduction
Suppose that you are in front of the window of a car dealer and you point to a certain car. you could then use sentence (1a), in which the NP this car would receive an interpretation of "type" or "kind" (type reading).
b. This care is mine [*type reading, token reading].
The sentence would therefore mean "I have a car like this one" or "I have a car of this type". In that scenario, you could only use sentence (1b) if you had purchased that particular car (token reading). Conversely, if you are at your garage and you point your finger at a certain vehicle, you could use (1b). you could not use (1a) -except in very restricted circumstances, which I will address shortly-, since the relevant interpretation of the NP this car is "token" in this scenario, not "type". Now suppose that you are at the place of some friend or yours to whom you lent a book some time ago. If you point to a certain book on her bookshelf and utter sentence (2a), (2) a. I have this book. [type reading, *token reading].
b. This book is mine. [*type reading, token reading].
you will be saying that you own another copy of that book, and not (dropped hints and ironic intentions aside) that this is the book you lent to your friend. Conversely, if you use sentence (2b) in such a situation, you will be suggesting to your friend (perhaps somewhat rudely) that she should return the borrowed book to you. As we see, the meanings expressed in these situations are quite different. 2 The two ways of expressing possession illustrated in (1) and (2) have long been associated in the literature, including proposals which derive the (a) sentences above from their (b) counterparts, or relate them through various mechanisms which insure their supposed equivalence. As is well known, the meaning of HAVE (= English have plus its equivalents in other languages) may be expressed in many languages with BE (= English be and its equivalents in other languages) followed by dative or genitive NPs. Alternatives include possessives and PPs encoding the meanings expressed by these morphological case marks. PPs headed by wITH denoting contact or transitory (i.e., stage level) possession provide another option, as in I am with the flu ~ I have the flu.
There is little doubt that the opposition between these two constructions is related to informational structure: the first pattern is often associated with possession or ownership: we speak of a certain individual and we predicate that some entity belongs to him/her (or, rather, that he or she is temporally associated with it). In the second pattern, which expresses membership, belonging (sp. pertenencia; Port. pertença) or mere adscription, a certain entity is spoken of, and it is pointed out to which individual it corresponds. In a large number of cases, the result may be relatively equivalent in strict truth-value terms, as in (3): (3) a. I own the apartment in the third floor.
b. The apartment in the third floor is mine.
But, as shown in (1) and (2), such equivalence does not necessarily hold when demonstratives are involved. The irregularity of type readings in the membership pattern may be due to a pragmatic factor, since one may attribute the possession of entities or that of groups of entities to individuals, but hardly that of classes of entities. By contrast, it is not so obvious how exactly we can derive the restriction that forces type readings in (1a) or (2a). This restriction will be called type restriction on demonstratives here (henceforth, TRD): (4) In unmarked contexts, demonstrative NPs acting as internal arguments of HAVE receive a type interpretation, as opposed to a token reading.
The TRD effect, as formulated in (4), is restricted to demonstratives. Interestingly, this interpretation is preserved in interrogative sentences: (5a) is a question formed by targeting the NP complement in (1a). Just as its Spanish counterpart (namely, ¿Qué coche tienes?), (5a) asks about a type of car (then, some branch or some car model). This utterance does not ask the hearer to identify his or her vehicle from within some set of cars, as opposed to (5b). Moreover, if a wh-phrase unmistakable selects for tokens in some scenario parallel to (1), HAVE will be excluded, as shown in (6a). Again, type readings naturally license parallel uses of HAVE, as in (6b): (6) a. which one among these three cars {is yours / *do you have}?
b. Now that we talk about allergies, which one do you have?
All this confirms that ownership and belonging, as introduced above, prove to be rather different relations from a grammatical perspective. In this short paper I will attempt to schematically characterize the TRD, set out its conditioning factors, and suggest its plausible place in the grammatical system. Although this article was originally intended to analyze the TRD in Spanish, I will use English in many examples, given that the results obtained are practically identical for both languages. Even so, Spanish demonstratives -unlike their English counterparts-can be prenominal (este libro 'this book') or postnominal (el libro este, lit. "the book this"; approx. "the book here"). I will not address the possible extension of the TRD to postnominal NPs in the pages to follow.

Definiteness effects with HAVE
I will assume some familiarity with the so-called Definiteness Effect (henceforth, DE). The DE with HAVE has been addressed in fewer studies than its counterpart with THERE-BE. Analyses of the former include Gutiérrez-Rexach (2000, 2007, Tham (2006), Leonetti (2006), Bassaganyas-Bars (2017) and Bassaganyas- Bars & McNally (2020). For a state of the art on the DE in THERE-BE contexts -deeply scrutinized for about thirty years-I simply refer to the overviews included in Reuland & ter Meulen (1987), Lumsdsen (1988), Lyons (1999), Landman (2004), Fisher et al. (2016), McNally (2016 and the references therein. It is worth remembering that numerous interlinguistic differences on the DE have been noted, and also that a large number of exceptions to it have been pointed out for individual languages. As a matter of fact, the literature mentioned above amply demonstrates that the presence of definite determiners and strong quantifiers in the complement of THERE-BE sentences does not guarantee that the DE will be attested at all, since it can be obviated by several resources that give rise to quantificational readings obtained compositionally. Exceptions to the DE in the grammar of Spanish are summarized in Leonetti (1999: 814 and ff.) and RAE-ASALE (2009: 15.6i and ff.). The TRD is not attested in existential sentences. otherwise, it would be possible to say Here is this Volkswagen intending something like 'Here is a car of the type of a Volkswagen', contrary to fact. But there is little doubt that the type-token distinction is relevant for the analysis of THERE-BE contexts. Definite NPs denoting types may escape the DE in Spanish presentational sentences, whereas their token counterparts may not. Sometimes the nouns tipo 'type' or clase 'class' are sufficient to obviate the DE, as in (7) 3 Even so, from the pattern of (8) -more frequently attested in literary language-one cannot deduce that NPs headed by definite determiners or strong quantifiers denoting individual types always escape the DE in Spanish. Most of my informants who accepted (8)  This connection between the TRD and the DE in THERE-BE sentences should not obscure the fact that the DE with HAVE is much more restrictive than its counterpart with THERE-BE. The former occurs characteristically with HAVE in NPs expressing kinship relations, as in (9a), and other forms of inalienable possession, particularly those of a meronymic or part-whole nature, as in (9b) Although both constructions express the existence of some entity relative to a spatio-temporal context, the contrasts of (9)-(10) are extremely marked. The explanation usually given to them lies in the fact that the cases of meronymy and inalienable possession often coincide with those that give rise to the interpretation of HAVE that Keenan (1987) called "existential"; that is, the one in which sentences with HAVE admit paraphrases with THERE-BE sentences: (11) a. This book has {a / *the} preface. [DE with have] b. There is {a / *the} preface in this book. [DE with there-be] Even so, these paraphrases are known to be appropriate for nouns expressing meronymic (or integral: Hornstein et al. 1996) relationships, but not quite so for those denoting kinship relations. Both sorts of possession are expressed through relational nouns; that is, nouns (typically biargumental) denoting realities that cannot be conceived of as independent of the entities in which they are included or to which they are attributed. This makes it difficult for them to become autonomous discourse referents (intuitively, it is difficult to identify the reference of "a brother" in reality, as opposed to that of "a tree" or "a car"). A more complete classification of Spanish relational nouns is presented in RAE-ASALE (2009: 12.10c).
Bassaganyas-Bars and McNally (2020) note that NPs formed from nouns such as type or kind, escape the DE with HAVE, as in (12a). Nouns that receive a "kind(-like) interpretation" (in their terms) do so as well, as in (12b): (12) a. Jan has this kind of sister (cf. *Jan has this sister).
b. Jan has the same intelligence as Kim (cf. *Jan has the intelligence).
The meaning of (12b) may be paraphrased with expressions that would include NPs such as "the same type, degree or amount of intelligence". on the effect of Sp. mismo 'same' on the cancellation of the DE in Spanish, see RAE-ASALE (2009: 15.6n). Again, these sentences admit paraphrases with indefinite determiners, such as Jan has this kind of sister ~ Jan has a sister of this kind. They coincide with (1a) in that similar paraphrases are admitted in the latter case (John has a car of this type), but the fundamental difference lies in the fact that car is not a relational noun, but a sortal noun. Bassaganyas-Bars and McNally (2020) detect more exceptions to the DE with HAVE with relational nouns modified by certain complements, and show that the interpretation obtained in these cases is also that of type. This is shown in their example (13): (13) Jan has the body of an athlete, which means "Jan has the type of body that an athlete has". 4 Again, there is some relationship with the TRD effect, but also a crucial difference, since the TRD is not restricted to relational nouns nor existential HAVE.
NPs attesting the DE in sentences with existential HAVE may escape it through a well-noun resource: definite NPs impose a familiarity condition that requires us to identify a single individual in the domain of some referential expression. This information is often obtained through endophoric means, including relative clauses and adjectives or participles expressing reference to prior knowledge of individuals (usual, typical, expected, required, known, etc.), as in (14). The asterisk outside the parenthesis indicates that the information it contains cannot be omitted.
(14) a. This book has {the preface that it was supposed to have / the usual preface}.
b. you have the *(typical) autumn cold.
c. I have the luck *(of having you with me).
d. He has the wife *(that he deserves).
e. John has the *(required) age.
f. I do not have the *(necessary) heigh.
g. you had the good sense *(to speak to him).
Notice that many abstract nouns (age, luck, sense, etc.) typically behave like relational nouns as regards this pattern. This link is discussed in relation to the DE for Spanish in RAE-ASALE (2009: § 12.20c and § 15.6q and ff.). 5 we can, then, deduce from (14) that, in unmarked contexts, most definite NPs with relational nouns show a certain resistance to identify their referent in HAVE contexts through the mere presence (in the awareness of speakers) of the entities being mentioned. Endophoric identification provides some clues for the hearer to acknowledge their background status. The minimal pair in (15) gives an idea of this difference between "sortal nouns" (such as book) and "relational nouns" (such as luck): (15) a. I have the book (of John's).
b. I have the luck *(of being here). 5 In any case, it is conceivable that the distinction required turns out not to be a strictly lexical one, especially in cases in which the information provided by the complements marked in (14) can be deduced from the immediate context. we may compare (11a) and (14a) with Finally, your dissertation has the (necessary) preface; so, you may now deposit a copy in the department, in which the adjective necessary could be omitted.
It is known that anaphoric devices tend to be most effective with sortal nouns when some small clause contains a secondary predicate, as in I have the book with me or You have the car in the garage. The effect of this secondary predicate may be lexical, in the sense that HAVE no longer denotes ownership, but rather some locative association temporally restricted or circumscribed to a particular environment. Thus, from Tengo el coche esperándome 'I have the car waiting for me' (Leonetti's 2006 example), one cannot infer Tengo el coche 'I have the car'. In a similar vein, Gutiérrez-Rexach (2000) notes that locative adjuncts override Hornstein et al.'s (1996) integral reading of predication, so that My truck has a Ford engine in the back seats does not imply My truck has a Ford engine. 6 All this suggest that, when HAVE expresses accidental, circumstantial or stage-level possession, it does not properly select for an NP, but for some small clause, even is its predicate is not always overt. This might provide a syntactic explanation for the fact these sentences apparently constitute exceptions to the TRD: If I point out to a set of documents and I say The police has these proofs, I will certainly not mean that the police has proofs of the same type as the ones I am pointing at, but rather that they have these specific (i.e. token) items. Notice that this interpretation will not be different if the demonstrative is identified anaphorically, rather than indexically or ostensively.
I shall return to the important role of secondary predicates in HAVE sentences, but before going into that, let me narrow down the effects of the TRD.

Refining the restriction
In this section I will take a schematic look at the fundamental factors that override, or at least strongly condition, the TRD.

Pragmatic accessibility and circumstantial possession
The type reading has to be pragmatically accessible, in the sense that it presupposes the existence of objective classes or types in the entity we are 6 Interestingly, Juan es largo de piernas 'Juan is long-legged' does not imply Juan es largo 'Juan es long', as pointed out by Español-Echevarría (1977) in his study of the inalienable prepositional construction in Spanish and its relationship with its HAVE counterpart (in this case, Juan tiene las piernas largas 'Juan has long legs'). Recall that secondary predicates are never omissible in these contexts (therefore, *Juan tiene las piernas 'Juan has the legs'). talking about. This allows us to compare the naturalness of (1a) with the strangeness of ?I have this umbrella or ?John has that bottle. These differences are not strictly grammatical. In fact, they are not even lexical, since it would be impossible to construct the paradigm of nouns that group with book and car, and those that do so with umbrella or bottle.
Although one might argue that the inaccessibility of a "type reading" in sentences such as I have that phone number is due to pragmatic reasons (that is, the difficulty of identifying "types of phone numbers"), I suggest that these cases pattern with the accidental or circumstantial possession cases that I have just mentioned. Their capability to cancel the TRD may be either lexical (if we assume that they contain a particular reading of HAVE not relevant for the TRD), or rather -and perhaps preferably-syntactic, if we suppose that HAVE does not select for NPs in these cases, but for small clauses allowing the contextual retrieval of their missing predicate.

Contrastive environments and reminder contexts
Contrastive patterns may introduce demonstrative NPs that provide possible substitutes of a certain entity that has been mentioned in the immediately preceding discourse, then cancelling the TRD effect. There is little doubt that this stool receives a token reading in (16) The result is similar to the one provided by the so-called reminder contexts (RCs). These contexts have been studied in Rando andNapoli (1978), woisetschlaeger (1983), Hannay (1985), Lumsden (1988), ward & Birner (1995, Abbott (1993), and Leonetti (2016), among others, in relation to THERE-BE sentences. Notice that the DP this car does not receive a type interpretation in (17), but a token reading, as does this stool in (16): (17) -you own nothing.
-That's not true. I have this car.
RCs are not mentioned in the literature as contexts licensing the token interpretation of expressions such as this car in (17), but rather as contexts which escape the DEs in English THERE-BE sentences. For example, the utterance Yes, there is that problem would be natural if someone wants to acknowledge that some particular problem is a possibly forgotten item in a list of relevant factors. 7 In a broad sense, RCs are contrastive contexts, such as the one in (16), but the patterns in (16) and (17) cannot be reduced to a single class, since RCs are typically associated to some (possibly) forgotten content, whereas in (16) some unknown -not forgotten-information is introduced.
NPs such as those characterizing RCs in the examples above are definite, but they have not been previously mentioned. As ward and Birner (1995: 730) note about RCs, "there are sufficient grounds for the speaker to believe that the entity has been (at least momentarily) forgotten [...]. For this reason, the speaker is licensed to treat the referent as though it were hearer-new". I conclude, along these lines, that RCs rescue definite NPs from the DE in existential contexts because, although their referent is definite and refers to a particular entity, it is also new in the relevant domain.
In my view, the paraphrases with indefinites that characterize type readings in demonstrative contexts (i.e., I have this car ~ I have a car of this type) can be extended to RCs. I will suggest that one possible way to carry out such an extension is to attribute what we may call a sample reading to this car in (17). In this interpretation, a paraphrase of I have this car in this context will be close to "I have something relevant (that you forgot); namely, this car".
Interestingly, sample readings are obtained in other contexts. In Bosque (2010) it is argued that so-called inclusive disjunction is not properly inclusive, but rather an instance of a relatively similar "sample interpretation", one of the possible instantiations of free-choice indefinites in intensional environments. That is, if the speaker pronounces coffee or tea in (18) in a single intonational group with a final rising pitch contour, 7 It is worth remembering that there are strong interlinguistic differences on the relevance of RCs contexts as regards DEs. The example (ia) -from Lumsden (1988: 110)-would be natural in English if one wants to suggest that certain university should be considered in relation to some particular purpose.
(i) a. There's the University of Stoke. b. *Hay la Universidad de Stoke. Its translation to Spanish, which appears in (ib), is strongly ungrammatical, regardless of the context in which it might be used, since estar 'be' is the unmarked verb required in presentational sentences with definite NPs.
(18) would you like coffee or tea?, he or she would not be suggesting the hearer to have both coffee and tea (a rather absurd suggestion, although somehow implied by the so-called "inclusive reading of disjunction"), but rather asking a question such as Would you like something, such as coffee or tea? This analysis -but not the strictly "inclusive" interpretation of disjoined NPs-naturally accounts for the fact that the hearer might answer the question by saying Thank you, I'll have a soft drink.
Interpreting demonstrative NPs in RCs as instances of the "sample reading" suggested above has another interesting consequence: it easily accommodates the fact that I have this car in (17) admits paraphrases with at least, among other things, and other similar expressions suggesting a set of which only one representative member is mentioned by way of mere illustration. It is, thus, implied that there are other members that need not be mentioned. In fact, NPs in RCs are sometimes claimed to receive a "list reading", as in Lyons (1999: 239). For Rando and Napoli (1978), they do not constitute the focus of the assertion in THERE-BE contexts. This focus is provided -they argue-by the set containing those individuals, then by some quantificational entity.
Finally, I suggest that the "sample reading" analysis can be extended to contrastive contexts such as (16), which -as I have just pointed out-do not conform to the TRD either. The contextual interpretation of I have this stool in (16), would then be close to "I have something similar to what you want; namely, this stool".

Anaphoric or textual contexts
Most demonstrative NPs built from relational nouns do not have referents able to be identified through ostensive or indexical means. This leaves anaphora as the only resource for them to find their reference: Even so, anaphoric contexts may be equally compatible with both type and token readings of these NPs. Notice that that problem means "a problem of that type" in (20a), but also "that particular problem", then giving rise to a token interpretation. The former reading is harder to obtain with sortal nouns, as (20b) suggests.
(20) a. John couldn't leave his home at night, but I didn't have that problem.
b. Mary then showed a blue notebook. I now have that notebook.
As I have suggested in section 2, some secondary predicate is understood in the latter sentence, such as "with me" or "here". In many cases, it is simply not possible to distinguish type and token readings in anaphoric uses of definite NPs, especially if abstract nouns, such as problem in (20a), are involved. Similarly, the NP this discussion in Every time we have this discussion admits paraphrases such as "a discussion of this type" (type reading), but also "this specific discussion" (token reading), assuming that repeated discussions are instances of the same notion.
The use of demonstratives sometimes called evocative does not clearly distinguish these two senses either, as in The food had that special flavor, in a context in which no previous mention has been made of any flavor. Even so, the evocative use of demonstratives could perhaps be reduced to the endophoric pattern, assuming that the former freely admits adjectives such as typical or characteristic.

Non-focal interpretations
The last factor, among those that cancel the TRD, is informational: the TRD disappears when the narrow focus of the predication does not fall on the NP complementing HAVE, but on some other constituent external to it. The relevance of this factor is illustrated in (21) The TRD also disappears when the NP complement is under the scope of an anti-assertive operator, in the sense developed in Bosque & Sáez (2017). In this paper, negative, modal-intensional or presuppositional operators are shown to provide the proper contexts for the licensing of Sp. tan(to) 'so much, so many' in non-overt comparisons, as in Sp. Siento que Juan trabaje tanto 'I am sorry that Juan works so hard' vs. *Creo que Juan trabaja tanto 'I think Juan works so hard'. Their capacity to cancel the TRD is shown in (22) 4. Assembling the pieces of the puzzle A classic problem of the DE is the fact that it is often formulated as a filter or a negative generalization. The DE tells us which specific situations do not occur, a strange way of formulating scientific generalizations in any discipline. There are several ways to turn the DE upside down so that it becomes formulated in positive terms. Certainly, the DE cannot be formulated as a constraint against the occurrence of definite NPs, proper names and strong quantifiers in existential contexts, as the literature has repeatedly made clear. one way to reformulate it in positive terms, as applied to HAVE, could be the one suggested in (23): (23) Narrow focus NPs complementing HAVE must be hearer-new.
we may, in fact, interpret the TRD as a natural outcome of (23). Notice that the TRD nicely falls within this generalization if we assume that type readings are hearer-new, as I have suggested. Moreover, in the previous section I have argued that the "sample interpretation" of demonstrative NPs in reminder contexts is also hearer-new, and also that it may be related to the fact that these readings allow for paraphrases with indefinites, which gives them a natural place in (23) as well.
As Tham (2006) argues, sentences with HAVE in which an ownership relation is denoted are associated with presentational focus. As a natural consequence, these NPs introduce new information into the discourse. we may suppose that the canceling effect described in § 3.4 is an extension of the one which cancels the DE with HAVE in small clauses. In simple sentences such as I had that book at home (which does not imply I had that book, as I have explained), the small clause prevents that book from occupying the narrow focus of the main predicate. Since adjunct modifiers attract the focus of the predication in these sentences, the focus no longer falls on the demonstrative NP. An immediate consequence, according to (23), is that these definite NPs do not have to conform to any quantificational reading.
In order to clarify this idea, we may consider the following contrast from Spanish: In principle, the omission of a temporal adjunct should not make a simple sentence ungrammatical. But there is nothing anomalous about (24): the TRD rules out the token reading for este resfriado in (24a). we may suppose that the hearer tries to get the type reading for this NP, but he or she fails as well (cf. *Tengo un resfriado como este 'I have a cold like this one'). Notice that the reference of the demonstrative NP in (24a) cannot be retrieved anaphorically (as opposed to the one in Tengo este problema 'I have this problem'; recall section 3.2.). The "hidden secondary predicate" (such as with me) that might be available por HAVE in some small clauses is not plausible either. Since none of these options is suitable, the sentence cannot be interpreted, unless some overt reminder context is provided. 8 In contrast, in (24b) the demonstrative NP is outside the narrow focus of the predication; the type reading ceases to be mandatory and the token (i.e., non-quantificational) interpretation becomes admissible, according to (23). The contexts that Bosque & Sáez (2017) call anti-assertive, as in (22), play a similar role, as their name indicates. Those operators place the relevant NPs under modal or presuppositional environments able to assimilate their content to the information already known by the hearer. This makes them free of the requisite imposed by (23). I realize that (23) does not explicitly mention demonstratives. As we know, demonstrative NPs typically retrieve their referent in ostensive contexts (deixis ad oculos); they reject associative anaphora and they do not give rise to scope-related ambiguities, as opposed to their counterparts with definite articles. Certainly, demonstratives do not reject anaphoric contexts, as we saw in (20), but sortal nouns suggest type readings in those environments, or rather add (or suggest the need for) secondary predicates (here, with me, at home, ready, etc.), as I have recalled. Relational NPs do not present this problem because of their facility to anaphorically or endophorically identify the entities they designate. In addition, in many cases they do not discriminate type readings from token readings. we can therefore assume that (25) is on the right track: (25) Demonstrative NPs with sortal nouns and token readings complementing HAVE are referential expressions requiring a contextual identification.
If this is the case, it is entirely natural that these NPs do not contribute the "new information" demanded in (23). I am afraid I will not be able to fully develop (23) here in order to cover the interpretation of both demonstrative and non-demonstrative definite NPs within a positive formulation of the DE in HAVE contexts. I will just recall one relevant factor that should be taken into account in such a task: distinguishing between two senses of the same verb HAVE (approximately, existential and non-existential) constitutes a lexical solution to tell apart "membership-localization" and "possession-ownership", two concepts systematically related in many languages. There seems to be a general agreement on the idea that noun classes (basically "relational vs. sortal" nouns) determine the appropriate interpretation of these two large groups. But at-