ANALELE UNIVERSITATII BUCURESTI

Anul 2003

Anul 2004

Anul 2005

ANALELE UNIVERSITATII BUCURESTI
  

LIMBI ŞI LITERATURI STRĂINE

2 0 0 5

Models and metamorphoses of determination

ALEXANDRA CORNILESCU

1. Aim of the paper

The paper presents two semantic models of determination that have recently been proposed in the tradition of formal semantics, in an attempt to maximize isomorphism between logical semantics and natural language syntax. These are the model of generalized quantification and that of relational determiners. The paper also illustrates some consequences of these proposals for the interpretation of DPs.

2. The generalized quantifier model, the functional analysis

Interest in general quantification theory as a tool for studying natural languages arose because of the insufficiency of the formal means and substantive concepts offered by the lower predicate calculus (=LPC). Thus, in LPC, quantifiers are interpreted syncategorematically, while in NLs quantifiers are members of the well-defined class of determiners and pronouns. Secondly, following Fregean lines, LPC drastically separates referential NPs (viewed as constants) from quantifiers, as is apparent in comparing the representations in (1) and (2) below. The problem is made worse if quantifier phrases, rather than simple quantifiers are involved, since in this case what is a simple sentence in English appears as a complex formula in LPC (cf. (3)). The class NP/ DP is quite coherent as far the syntax of NLs is concerned and it should be coherent in natural language semantics as well. Thirdly, NLs mostly make use of quantifier phrases, since in NLs quantifiers have restricted domains, while in LPC, quantifiers are unrestricted.

(1)     a.   Everything flows.                "x flows (x)

         b.   Nothing flows.                     Ø$x flows (x)

         c.   Something flows.                 $x flows (x)

(2)     a.   John works.                        works (j)

         b.   Bill met Sue.                       meet (b, s)

(3)     a.   Every river flows.                "x [ river (x) ®flows (x) ]

         b.   No man lives forever.          Ø$x [ man (x) Ù live forever (x)]

         c.   Some students work.           $x [ student (x) Ù works (x)]

2.1. Generalized quantification: Restoring the unity of the category NP

Generalized quantification is the theory of expressions which denote families of sets. The insight of generalized quantification theory is that all NPs (terms and quantifier phrases) can be given a unitary denotation, but this denotation is of a higher type: namely all NPs denote generalized quantifiers. A generalized quantifier is a particular semantic object, namely, a set of properties ascribed to a noun phrase (Barwise & Cooper, 1981).

(4)     NP Quantifier Universal (Barwise and Cooper) "Every natural language has syntactic constituents (called noun phrases) whose semantic function is to express generalized quantifiers over the domain of discourse".

The semantic type of NPs, under their generalized quantifier reading is: <<e,t>,t>, that is, they denote functions from properties into truth values. The intuition expressed by generalized quantifiers is that individuals can be viewed as denoting the sum of their properties, so that individuals are property sets. A referential expression like John, denotes the set of all the properties of the individual John, while a quantifier term like every boy likewise denotes the set of properties that all boys share. Here are translation rules employing generalized quantifiers denotation:

(5)     a.   John = lR R (j)

         b.   every man = lR"x [man (x)® P(x) ] = lR [MANÍ P ] or {P: MANÍ P }

         c.   some students  = lR$x [student (x) Ù P(x) ]= lR [STUDENT Ç P ]¹Æ

         d.   no students = lR~$x [student (x) Ù P(x) ]= lR [STUDENT Ç P ] = Æ

         e.   most students = lR| students Ç P | > | students – P|

Sentences like (6a, b) now receive a unitary analysis: their denotation is simply constructed by saying that the property expressed by the predicate (VP_{<e, t>}) is included in the property set of the subject. In other words, quantifier phrases (i.e., NPs) are viewed as properties of VPs, so that the predicate becomes the argument of the subject, "through a 180° revolution of the natural subject – predicate structure" (cf. van der Does & de Hoop, 2001: 3). This is apparent in the type theoretic analysis in (7), where the subject is defined as a function from predicate (= VP_{<e, t>}) denotations) into truth values:

(6)     a.   Every student walks.

         b.   John walks.

Through the simplification rules of the calculus the translations of (6) in (8) are reduced to formulas familiar from LPC, such as (8b) and (8d). Notice that the proper name is reduced to an <e> denotation eventually. This indirectly shows that the proper name may denote both a generalized quantifier (type <<e, t> t> ) and an entity (type <e>).

(8)     a.   Every student walks   à     lP"x [student(x) ® P(x)] (walk) à

         b.                                             "x [student(x) ® walks (x)]

         c.   John studies.              à     lP P(j) (walk) à

         d.                                             walk (j)

As shown in these analyses, the generalized quantifier approach to determination gives a functional view of quantifier phrases in agreement with their syntax: there is a binary subject predicate structure at the syntactic, as well as at the semantic level.

2.2. Importance of generalized quantifiers for NL semantics

The model of generalized quantification has several favorable consequences: The essential gain is that generalized quantification assigns a uniform denotation type to all NPs: all NPs are sets of properties; it thus eliminates the split between referential phrases, which refer to specific objects, and quantificational phrases, which do not refer. The NP, which is a well established natural language category, is paired with a single suitable denotation. The compositional semantics of all subject predicate sentences is homogeneous and it is isomorphic with their syntax. As remarked by commentators, the NP – universal in (4) may be given a stronger or a weaker interpretation:

a.   Weak version of (4)

All languages have NPs and all NPs can be interpreted as generalized quantifiers.

b.   Strong version of (4)

All languages have essentially quantificational NPs, i.e., NPs which can be analyzed as generalized quantifiers, but not reasonably as referential <e>, or predicative <e, t>, expressions.

The choice between these interpretations of (4) is a matter of empirical research. Partee (1987) argues that NPs are open to several readings: All of them may denote generalized quantifiers (type <<e, t>t>>), but they may equally well denote entities (type <e>) and predicates (type <e, t>). According to Partee (1987), it is the weak reading of (4) which is correct, in the sense that all NPs may express generalized quantifiers, but they need not do so in all contexts. In principle, interpretations of lower type (<e>, or <e, t>) are preferred to the generalized quantifier interpretation <<e, t>t>, whenever a lower type reading is contextually justified; see (9a) or (9b).

(9)     a.   A tall man entered the room. He looked tired. < e>

         b.   Bill is a tall man. <e, t>

         c.   One teacher and every student in the class attended the event <<e, t> t>

The hypothesis that the nominal phrase systematically possesses several types of denotations does not mean that all NPs in all contexts are open to all denotations. Constraints on the interpretation of NPs arise: a) from the position of the NP in the sentence (cf. Diesing's (1991) Mapping Hypothesis); b) from the internal constitution of the NP. Thus proper names favor an <e> type reading, while NPs like nobody, most students have only the generalized quantifier reading. However, strong support for the generalized quantifier approach derives from the fact that all NPs, even proper names, must be interpreted as generalized quantifiers at least sometimes. Such is the case of any sentence involving the coordination of a proper name and a traditional quantifier:

(10)   a.   John and everybody else walked out of the class.

         b.   Either Peter or nobody will solve this problem.

         c.   Peter and no one else understands this theory.

A second, semantically interesting aspect of generalized quantifier theory is the possibility to regard the DP as a higher order predicate which the VP as an argument, in a manner which is very different from the ordinary view that treats the DP as an argument of the VP.

A third advantage of generalized quantification regards the syntax semantics interface. The generalized quantifier analysis of nominal denotations may be felicitously coupled with the DP syntactic analysis of nominal phrases advanced by Abney (1987), which proposes that Determiners, rather than Nouns head the nominal phrase. Nominal phrases are DPs, as in (11).

It is possible to assign the different denotation types of the DP to the NP and respectively to the complete DP. One may separate the lexical NP component in the DP, which denotes a predicate and which is an expression of type <e, t>, from the DP constituent as a whole, which denotes an entity <e> or a generalized quantifier <e, t>t>>. A highly widespread opinion is that is the Determiner layer of the DP which is responsible for reference (cf. Zamparelli, 1995).

3. A few illustrations

3.1. The interpretation of the prepositional Accusative in Romanian

It is known that in Romanian, the Accusative is either a bare NP/DP, or it is accompanied by the preposition pe 'on'. In terms of general quantification theory, it may be shown (cf. Cornilescu, 2001) that the difference between bare and prepositional Accusatives lies in the range of the denotations they allow. Of the range of possible DP denotations mentioned above, Romanian prepositional Acc. allow only the generalized quantifier denotation, <<e, t>t>, and the entity denotation <e>; therefore prepositional Acc. allow only object-level denotations, but they lack the predicative, property denotation <e, t>. The occurrence of pe 'on' with generalized quantifier <<e, t>t> denotations is immediately apparent in the fact that pe is required in the accusative of all [+person] bare quantifiers like cineva 'somebody', nimeni 'nobody', cine 'who', etc.

The characterization of the semantic effects of pe ‘on’ we have proposed allows us to predict when pe is obligatory and when it is excluded. The following empirical generalizations can be stated:

(12)   Generalization 1

Pe 'on' is obligatory with those DPs which allow only e-type readings.

Consequently, pe 'on' is obligatory with proper names which have only e-type denotations.

(13)   Generalization 2

Pe 'on' is impossible in those contexts where a property reading or a kind reading is the only possible one. Consequently, the Prep Acc is not possible with verbs that require a property reading of their object.

While Generalization 1 has long been known, though not stated in these terms, Generalization 2 is first proposed in Cornilescu (2001). Let us now briefly review some of the evidence for the two generalizations above

3.1.1. Proper names. The characteristic property of pe DPs is that they have e-type (object-level) denotations, but not <e, t> (property) denotations, so that pe 'on' is obligatory with those DPs which have only e-type reference. This is the case of (person-denoting) proper names and of personal pronouns. Indeed, the only category of lexical nouns with which pe is obligatory in contemporary Romanian is that of proper names.

Evidence that proper names have only the object-level readings is the fact that they are excluded from contexts which require the property reading, precisely because they lack the property reading. One construction where the property reading is the only available one is the se (reflexive) passive. Significantly, proper names, as well as personal pronouns, which lack the property reading, are excluded from this construction, though they are allowed as subjects in fi 'be'-passives. Thus, the bare plural in (14a), which typically has a property reading (cf. Dobrovie-Sorin & Laka, 1996), is a felicitous subject in the SE passive, while the proper name in (15a) is not.

(14)   a.   S-au adus cămăşi.

               SE-have brought shirts.

         b.   Au fost aduse cămăşi.

               Have been brought shirts

               'Shirts have been brought.'

(15)   a.   *S-a adus Ion la judecată

               SE-has brought Ion to trial.

         b.   Ion a fost adus la judecată.

               Ion has been brought to trial.

               'Ion has been brought to trial.'

3.1.2. PE DPs lack the property reading. Romanian disposes of one type of nominals which can only express properties, these are the bare singular NPs (cf. Kalluli, 1999). There are predicates whose object position allows only a property denotation, so that the verb and the direct object make up a derived complex predicate. Thus, in the intended reading only bare singulars or singular indefinites (which both allow the property reading) may be used as objects of such verbs. An example of this kind is the verb avea 'have'. Notice that, while the non-prepositional indefinite Accusative is perfect, the pe indefinite is excluded (see (16c)), even though the lexical head is person-denoting.

(16)   a.   Ion are nevastă tînără.

               Ion has wife young.

         b.   Ion are o nevastă tînără.

               Ion has a wife young.

         c.   *Ion (o) are pe o nevastă tînără.

               Ion has PE a wife young.

               'Ion has a young wife.'

Other verbs that belong to this paradigm are the following: a pretinde 'claim, demand', cere 'request', dori 'wish', vrea 'want'. Notice that they are all intensional verbs. With some of these intensional verbs pe DPs are possible, but the meaning is significantly different (see (18)).

(17)   a.   Ion pretinde/ cere/ vrea/ doreşte nevastă tînără.

               Ion claims / requests/ wishes wife young.

         b.   Ion pretinde/ cere/ vrea dorešte o nevastă tînără.

               Ion claims / requests/ wishes a wife young.

         c.   *Ion (o) pretinde/ cere/ vrea doreşte pe o nevastă tînără..

               Ion (her)-claims / requests/ wishes PE a wife young.

               'Ion claims / requests / wishes a young wife.

(18)         Ion a cerut-o pe o fată frumoasă.

               Ion has requested-her PE a girl beautiful.

               'Ion has proposed to a beautiful girl.'

To conclude, generalized quantification theory offers a plausible characterization of the use of the prepositional Accusative in Romanian.

3.2. The weak / strong contrast

3.2.1. The distinction. The weak/ strong distinction was introduced by Milsark (1977) to account for some observations made with respect to the distribution of NPs in existential contexts.

1. Weak NPs are defined by Milsark as those NPs which can be felicitously used in existential there-sentences, whereas strong NPs are strange or ungrammatical in this context.

(19)   a.   There is a cat / no cat /*the cat /*every cat /*neither cat in the garden.

         b.   There are cats /two cats/ seven cats / many cats / no cats /at most two cats / *the cats /*all cats /* most cats in the garden.

         c.   ‘There aren't two cats in the garden.’

         A.  ‘There isn't any, actually, there are two dogs in the garden.’

         B.   ‘There are three cats, not two.

(20)   Strong Ds:   the, this, that every, each

         Weak Ds:   a, one, many, much, little, several, cardinals; some, any, no

Sentences in (19) assert something about the size of the set of cats present in the garden, distinguishing between situations where this number is two, seven, at most two, etc. The semantics of sentences (19) is intersective; the truth of any of these sentences depends on the cardinality of just one. More exactly, the required set consists of the intersection of the restrictor set (the common noun) and the scope of the determiner (the cardinality predicate). Neither of these sets is presupposed; either of them can be targeted by negation (cf. the examples in (19c)).

In sum, occurrence in there– sentences is one means of identifying weak determiners, in contrast with strong ones.

2. A second identifying criterion is semantic. Unlike weak determiners, strong determiners are presuppositional, in the sense that the set which is the restriction of the quantifier is presupposed to be non-empty. A typical example is (21a). Demontrative determiners are always strong. Determiners which are weak, and may appear in there-sentences also have strong readings. The reading of the strong DP is proportional, equivalent with a partitive construction. The difference between the weak reading of and the strong reading of some comes out very clearly in the two sentences below:

(21)   a.   Those of the students who have studied will get good grants

         b.   There are some ghosts in the castle.

         c.   Some (of the) ghosts are in the pantry.

3.            Various other criteria pick out the same class of NPs.

Information packaging: strong quantifiers make good topics, while weak DPs may often be part of the focus in the sentence. The relation between strong quantification and topichood is probably clear: both are presupposed, discourse given entities. Thus, the examples in (22) illustrate topics that have strong readings, partitive, generic, specific. Notice in contrast the examples in (23). They all start with a frame space or time adverbial, which actually represents the topic of the sentence. The subject is part of the focus phrase and the subject DPs receive weak interpretations. Thus the indefinite phrase in (23a) is [- specific], the bare plural in (23 b) is existential, etc.

(22)   a.   A cousin of mine is in the navy. (+specific)

         b.   Students are intelligent (generic)

         c.   Two of the students like semantics (partitive)

         d.   No cats like milk. (generic)

         e.   Few (of the) students like semantics [partitive/ proportional]

(23)   a.   Later, a student called and asked for your telephone number.

         b.   Right now, dogs are playing in the front yard. (existential)

         c.   Under the tree, two children were waiting for the bus.

         d.   At this hour of the day, few children are still playing in the street.

As far as the contrast between de re/ de dicto interpretation is concerned, weak DPs are open to both de dicto / and de re readings, as can be seen in (24 a). Strong, quantificational NPs always get a de re interpretation, as can be seen in (24 b):

(24)   a.   Eve seeks a new book on semantics / two students interested in semantics / no unicorn [de re or de dicto].

b.       Eve seeks every new book on semantics / most students interested in semantics / neither unicorn [only de re].

The contrast between weak and strong determiners holds cross-linguistically and is highly systematic. Moreover, in all languages, weak determiners may also have strong readings, while the strong quantifiers do not have weak readings in any language.

3.2.2. Generalized quantification theory is relevant for the strong / weak DP contrast since it again proves to be a convenient tool for formalizing the distinction between strong and weak interpretations of determiners phrases. Within generalized quantification theory, the difference between strong and weak DPs may be expressed by assigning them different types of denotations. Weak interpretations denote cardinality predicates or properties; strong interpretations denote generalized quantifiers or entities.

(25)   Weak DPs              <e, t>

         Strong DPs             <e> or <<e, t> t>

Saying that weak DPs have predicate / property readings accounts for their behavior: a) they are not presupposed; they function as predicates; they are not de re; they may represent new information. It is less transparent, however, why strong DPs have the properties described above. It will be seen that the relational view of determination offers a superior account of the properties pf strong/ weak DPs.

4. The relational view of determiners

4.1. Determiners as relations between sets

The functional view of the quantifiers, which treats the DP as a unit: every student, is opposed to the relational one, which views Ds as independent binary operators expressing relations between sets.

The relational view of determiners starts with a feeling of dissatisfaction about the translation rules offered to DPs in generalized quantifier theory, as illustrated in () above. Notice that determiners are still not assigned independent denotations, but are treated as parts of determiner phrases. This may not be a problem, if a language possessed a small or a finite list of determiners. As a matter of fact, however, determiners represent an open-ended set if complex determiners (in the sense of Keenen & Stavi, 1987) are also taken into account. Complex determiners (illustrated in (26) are not only complex syntactically, but also compositional semantically. This means that they should be analyzed semantically, rather than be given en bloc translations:

(26)   cardinal determiners: between five and ten, two dozen, a large number of, an even number of..

         exceptional determiners: all but ten, all but almost ten, every … but John, no…but Mary

         value judgment determiners: too many, a few too many,  surprisingly few, etc.

         proportionality determiners: exactly half the.., two out of three,  (not) one …. in ten, less than half the..

         conjoined determiners: at least two but not more than ten…, most but not all…, at least a third and at most two thirds

To solve this problem and assign the determiner an independent denotation, it is profitable to once more consider the manner of syntactic composition:

While the generalized quantifier in line (27b) combines with one property (set <e,t>) to produce a proposition, the determiner in line (27c) first combines with a property (set <e,t>) to form a generalized quantifier << e, t>t>. This means that the determiner successively combines with two properties to form a generalized quantifier and then a sentence, so that its denotation is that of a function of type <<e, t> <<e, t> t>>. We conclude that determiners are relations between two sets of individuals, <e, t> or between two properties. This is the relational analysis of determiners. Here are examples highlighing the relation between the interpretation of the DP and that of the D:

(28)   a.   [lP"x [student (x)--> P(x)]]                  every student <<e, t>t>

               lQ [lP"x [Q(x)® P(x)]]                     every <<e, t> <<e, t>t>>

         b.   [lP $x [student(x) Ù P(x)]]                    some student

               lQ [lP $x [S(x) Ù P(x)]]                      some (Q) (P)

         c.   [lP Ø$x [student(x) Ù  P(x)]] (S)          no student

               lQ[lP Ø$x [S(x) Ù P(x)]]                     no (Q) (P)

Every, for example, is the subset relation, that is, every (Q)(P) holds true just in case Q is a subset of P. The sentences in (29) are analyzed as shown, using the abbreviations proposed in (27 e).

(29)   a.   lQ [lP"x [S(x) ® P(x)]] (S)(W)                   Every student walks.

         b.   lQ [lP $x [S(x) Ù P(x)]] (S) (W)                   Some student walks.

         c.   lQ[lP Ø$x [S(x) Ù P(x)]] (S) (W)                 No student walks.

In (30), (31), we present an example of derivation, employing the abbreviations suggested in (27e); simplification rules of the l-calculus again lead to the familiar LPC translation in (31b), for the sentence in (31a):

(30)   a.   Every student             à     lQ [lP"x [Q(x) ® P(x)]] (student)         à

         b.                                    lP"x [student (x) ® P(x)]

(31)   a.   Every student walks.           lP"x [student(x) ® P(x)]] (walks)  à

         b.                                    "x [student(x) ® (walks) (x)]

Conclusions: 1. The relational analysis of determiners views them as propositional operators, relating two sets of individuals, therefore relations between predicates: Det (A) (B), one set being that of the (semantic) subject and the second being that of the (semantic) predicate. 2. Determiners are expressions of type <<e, t><<e,t>t>>.

4.2. The tripartite structure: operator, restrictor, nucleus

The analysis of Ds as relational expressions with two arguments Det (A) (B) has led to a ternary semantic analysis of quantified propositions, accepting the premise that quantification in natural language is restricted quantification. This ternary model of analysis (proposed in Heim 1982) is generally accepted because of its descriptive adequacy and flexibility. The tripartite structure: operator + restriction + nuclear scope is typical of all quantificational structures as may be noticed in the examples below:

(32)   Operator          Restrictor                                       Nuclear scope

         Always             if a man owns a donkey                   he beats it

         Sometimes        if a car falls from the fifth floor        it survives

         Every               man                                                arrived

         Some                students                                          speak French

This description of determiners is sheds light on the common semantic properties of (the two types of) quantificational operators (above), determiners (every, some) and adverbs (always, sometimes). Like quantificational adverbs, Ds are treated as propositional operators, with an important part to play in determining the logical form of the sentence. Modal verbs may also be quantificational operators in this ternary structure (cf.33f). Notice that a quantificational D or Av may at times bind all the variables in its scope. Thus often and many in (33a) and (33b) bind the individual variable x as well as the event variable of the nuclear VP. This has lead to the concept of unselective binding: an operator may bind more than one variable and it may bind variables of different logical types: individual variables, event variables, etc.

(33)   a.   Operator                   Restrictor                           Nucleus (matrix)

         b.   Often             if a cat_x falls from the V^th floor            it survives_e

         c.   Many             cats_x that fall from the V^th floor           survive_e

         d.   Seldom          if a person falls from the V^th floor        she survives

         e.   Few               persons that fall from the V^th floor       survive.

         f.    Must              if this is a bird                                     it have wings

4.3. The Mapping Problem

In the mapping hypothesis in (32), the determiner semantically functions as a propositional operator with scope over the subject set and the predicate set. An interesting syntactic problem has arisen: How does the determiner come to be a propositional operator, or more exactly, how does it come to take scope over the VP. General syntactic projection principles produce structure (34a), while semantically a configuration more like (34b), where the D is the head taking two arguments is needed. If DP is an argument of V and VP is simultaneously an argument of D, how can we find a structure satisfying this joint requirement?

Recently there have been several proposals of relating the syntax of determiners with their interpretation as propositional operators. Of these, the simplest and most explicit, to my knowledge is that in Larson (2004), which we sketch below:

4.3.1. While Ds are usually considered the nominal analogues of Inflection or Complementizers, Larson considers that Ds most closely resemble verbs. In his view, like verbs, Ds have arguments; they are inherently relational concepts and have a- structure like verbs. The content of the Ds is best characterized by the relational view of determiners in (36), with illustrations in (37).

(36)   Relational view of determiners.

         Determiners express relations between predicates.

(37)   a.   All whales are mammals.

         b.   ALL ([ x: x is a mammal], ( x: x is a whale)).

According to (36) Ds express the relations specified in the translation between the restrictor set and the nuclear set. Larson proposes that the restrictor set and the nuclear scope are the two arguments in the a-structure of a D: There are then the following thematic hierarchies for DPs and VPs, where roles are arranged in terms of their prominence. It is further assumed that the roles determined by a predicate are projected according to the thematic hierarchy, starting from the least prominent role:

(38)   a.   D      Nuclear Scope > Restrictor

         b.   V      Agent > Theme > Goal > Oblique

In (38), it appears that a determiner looks like a transitive verb, whose internal argument is the restrictor and whose specifier is the nuclear scope argument. The syntactic problem that arises is how to construct an LF where the nuclear scope is an argument (= the specifer) of the determiner, since clearly that is not the case in overt syntax.

Larson’s proposal is that at merge, Ds take a nominal complement and a free position in Spec DP, reserved for the Nuclear Scope (matrix VP). The nuclear scope argument of the determiner is an independent, inaudible pro-predicate element pro, licensed by D, and projected in Spec DP (19). Its semantic value is determined at LF from the derived predicate that is sister to DP in (34):

(39)   [_DP pro          [D’                NP    ]]

         q- nuclear scope      q-restrictor

What happens is that at LF, the VP moves to the position Spec DP and gets interpreted there, in the manner sketched in (40) and (41) below:

In (41), the determiner is in a configuration where it expresses a relation between the restriction, projected as its complement and the nuclear scope, which has ended up in the specifier position of the determiner. Further research is needed in this area to secure a better syntax-semantics mapping.

5. Re-interpreting the contrast between week and strong determiners

The relational view of Ds may very naturally express the contrast between weak and strong DPs, taking into account that in the ternary representation of a proposition, the restriction is a discourse-given non-empty set:

Remember that strong Ds are presuppositional and have partitive, proportional readings. This means that with strong Ds the NP complement of the D is part of the restriction and it is discourse given. In contrast, with weak DPs, the D itself, as well as its NP complement, is part of the nuclear scope. The variable the D predicates on is bound by a default existential quantifier. The contrast between the two types of readings represented in (42):

(42)      _Dstrong                     NP                         [_VP nuclear scope ]

                                    $                           [ _VP D weak_x+ NP +…..]

6. Conclusions

1. General quantification theory and the theory of relational determiners are successive attempts at better integrating syntactic and semantic representations. The two theories are in line with traditional linguistics in acknowledging the semantic coherence of the DP and D syntactic categories.

2. From a semantic point of view Ds are propositional operators; their interpretation closely resembles the interpretation of logical quantifiers.

MODELE ŞI METAMORFOZE ALE DETERMINĂRII

Rezumat

Lucrarea prezintă două modele semantice ale determinării, recent apărute în tradiţia semanticii formale, într-o încercare de a maximiza izomorfismul între semantica logică şi sintaxa limbii naturale. Acestea sunt modelul cuantificării generalizate şi cel al determinanţilor relaţionali. Lucrarea ilustrează de asemenea câteva consecinţe ale acestor propuneri de interpretare a grupurilor nominale cu determinant.

REFERENCES

BARWISE, J., and COOPER, R. (1981) “Generalized Quantifiers and Natural Language”, Linguistics and Philosophy, 4.2, 159-221.

CORNILESCU, A., 2001, “Notes on the Interpretation of the Prepositional Accusative in Romanian”, Bucharest Working Papers in Linguistics, Bucharest.

DIESING, M., 1990, Indefinites, MIT Press, Cambridge, Mass.

DOBROVIE-SORIN, C., and LAKA, B. (1996) “Generic Bare NPs”, Ms. Univeristy of Paris 7.

VAN DER DOES, J., and DE HOOP, H. (2001) “Generalized Quantifiers”, Glot International, 5.1., 3-12, Utrecht.

HEIM, I. (1982) The Semantics of Definite and Indefinite Descriptions, doctoral diss., MIT.

KALLULI, D. (1999) The Comparative Syntax of Albanian. On the Contribution of Syntactic Types to Propositional Interpretation, doctoral diss., University of Durham.

KEENAN, E., and STAVI, J. (1986) “A Semantic Characterization of Natural Language Determiners”, Linguistics and Philosophy 9, 253-326.

LARSON, R. (2004) Adjectival Modifiers, (to appear).

MILSARK, G. (1977) “Towards the explanation of certain peculiarities of existential sentences in English”, Linguistic Analysis 3, 1-29.

PARTEE, B. (1987) “Noun Phrase Interpretation and Type-Shifting Principles”, in J. Groenendijk, D. de Jongh & M. Stokhof (eds.), Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, 114-144.

ZAMPARRELLI, R. (1995) Layers in the Determiner Phrase, doctoral diss., University of Rochester.