The Semantics of Determiners

a hypothesis that M. has left not Cl. has passed through head nobody.Dat
‘?A hypothesis that Maria had left didn’t cross anybody’s mind.’

(The definite and indefinite articles are bold faced for perspicuity.)

Another serious problem for the familiarity theory of the definite article, noted in Abbott (1999), is that, at least in English, stressing the definite article stresses the uniqueness rather than the familiarity property of the discourse referent. The argument is hard to reproduce for Romanian because the definite article cannot bear contrastive stress. In English, however, and in other languages where the definite article can be stressed, an example such as Abbott’s (2), repeated here as (31), with bold-facing indicating stress,

(31) That wasn’t a reason I left Pittsburgh, it was the reason.

is interpreted as conveying the information that the referent of that is the unique reason the speaker left Philadelphia.

Dynamic uniqueness approaches to definiteness have been proposed in the literature in Kadmon (1987), Roberts (2003), and Farkas (2002a). The dynamic uniqueness notion in Farkas (2002a), called determined reference, is given in (32).

(32) A discourse referent x in a DRS K_e to be merged with an input DRS K has

determined reference iff for every f that embeds K relative to M it holds that
every f’, f” that extend f and which satisfy the conditions imposed on x by the
DP introducing it in K_e, f’(x) = f”(x)

The requirement of determined reference amounts to ruling out any choice in assigning a value to a discourse referent once the conditions the nominal introducing it are satisfied relative to the input context.

An advantage of thinking of uniqueness in these terms is that it allows us to treat definite descriptions (DPs with a definite article), Proper Names and definite pronouns as having determined reference. In the case of definite descriptions, determined reference is achieved because the description identifies a singleton set relative to the context; in the case of definite pronouns, the nature of their anaphoric condition ensures determined reference, while in the case of Proper Names it is the special connection between the name and its referent that ensures determined reference. To make sense of the hierarchy in (24), ways of achieving determined reference can be distinguished based on the connection between the nature of the conditions imposed by different types of DPs and determined reference. We will come back to the special nature of partitives below.

I follow here Farkas (2002a) in assuming that the definite article in both English and Romanian is a mark of determined reference. Determined reference is uniqueness-based, and therefore examples like (29) are not surprising. Accommodation can be seen as triggered by the determined reference requirement associated with definite descriptions: if after I have mentioned a house I mention the roof, a natural way for you to understand this DP as having determined reference is to assume that it is the roof of the house that has been just mentioned. Accommodation works just as in the case of familiarity theories but it is triggered by the determined reference requirement. You accommodated the roof of the house in the input context and process the definite DP the roof relative to this enhanced input. In the new context the determined reference requirement of the definite article is satisfied if the definite DP is interpreted as anaphoric to the accommodated discourse referent. Given the central role familiarity plays in cases when determined reference is established by an anaphoric link, as well as in most of the ordinary cases of definite description usage, when the Restrictor denotes a singleton only relative to context, it is not surprising to see familiarity and definiteness connected in some principled way. The connection can be seen as a preference for contextual uniqueness, in the spirit of Hawkins (1991). In Farkas and de Swart (2007) it is expressed, in O(ptimality) T(heoretic) terms, as a constraint against using novel definite DPs.

So far then we have proposed to see the ordinary definite article as a marker of dynamic uniqueness and have also suggested that familiarity and definiteness are connected. In concrete terms then, -ul/the have the feature [DEF], which signals determined reference. More generally, the definite/indefinite distinction, I suggest, is sensitive to the determined reference parameter. The question that is still open is the issue of the symmetric vs. asymmetric nature of the distinction. In concrete terms, is the ordinary indefinite article un marked for non-determined reference or is it unmarked for it? The unmarked nature of ordinary indefinites is argued for in Hawkins (1991) and taken up in recent work (see Abbott 1999, Farkas (2007)). Treating the distinction as asymmetric should be the null hypothesis because of simplicity reasons. The non-determined reference character of indefinites can be derived from the existence of definite forms by a Horn-scale implicature. If we take the ordinary definite to bear the [DEF] feature while the indefinite to be unmarked for it, the two articles form the Horn scale < the, a >, where the former is more informative than the latter with respect to the referential properties of the discourse referent just introduced. The choice of the less informative, indefinite form, implicates that the use of the more informative definite was not warranted. The blocking effect here involves choosing the referentially more informative form over the referentially less informative one.

Heim (1991) adopts an asymmetric view but proposes a new principle, Maximize Presupposition, to account for the blocking effect just mentioned. The two views are not incompatible since Maximize Presupposition would need to be motivated and such motivation could well involve the maximization of referential and contextual information.

We conclude here that the distinction between the ordinary definite and indefinite article is asymmetric, with the indefinite unmarked for determined reference.¹⁵ Under the view of definiteness as a marker of determined reference, the crucial issue concerns the relationship between the discourse referent associated with the DP, the input context and the model. In the case of singular definite descriptions, this boils down to contextual singleton denotation of the Restrictor.

We turn now to some stubborn problems for any kind of uniqueness analysis of definite descriptions, which are the same in both English and Romanian. They fall into several categories. One category is exemplified in (33):

(33) Maria a luat autobusul

M. has taken bus.Def
‘Mary took the bus.’