The Semantics of Determiners

3.2 Marked indefinites
It is now time to turn to the motley group of marked indefinite determiners. So far we are assuming that ordinary indefinites do not have the feature [DEF], while ordinary definites do. The ‘marked indefinites’, just like their unmarked sister, are not specifically marked for being indefinite. They are marked, we will see, for other special features that distinguish them from one another. We are justified in calling all these determiners indefinite not only because they involve, morphologically, the indefinite determiner or a morpheme close to it, but also because they do not involve the determined reference requirement associated with definiteness under our view.
3.2.1 The partitive determiner
Partitives in Romanian are made up of the partitive indefinite (bold-faced below) followed by a prepositional phrase headed by the preposition din(tre) ‘of’:

(53) Un-ul dintre copii a plecat.

Part of child.Pl has left
‘One of the children left.’
The partitive DP here is understood as referring to an element of the set of children referred to by the object of the preposition. We call the object of the preposition in a partitive DP the domain nominal since it provides the set from which the value of the discourse referent introduced by the partitive determiner is chosen. The domain nominal in partitive constructions in Romanian obeys the partitive constraint: it must be definite. (In (53) we don’t see the definite article because we have an unmodified nominal as object of a preposition; were we to use a more complex nominal, it would have to be definite: unul dintre copiii de pe stradă ‘one of the children on the street’.) In Romanian the partitive determiner is bi-morphemic, made up of the definite article (in (53), un) followed by the definite enclitic article (in (53) -ul).

The partitive determiner introduces a discourse referent whose value has to be chosen from among the members of a sum-level (plural) discourse referent introduced by the definite domain DP. The constraint it imposes then is of the form x  y, where x is the discourse referent introduced by the partitive determiner and y is the sum-level discourse referent introduced by the domain DP. Partitive DPs do not have determined reference since any member of the domain DP can legitimately count as a potential value for x, and therefore partitives count as indefinite in our terms. On the other hand, their referent counts as familiar in the sense that, as part of the definite domain DP, it must have already been introduced in the discourse, implicitly or explicitly. In Farkas (2002a) it is argued that the reference of partitives is more determined than that of unmarked indefinites because the domain of partitives is explicitly restricted to a previously introduced set. As a DP type then, partitives are between definites and unmarked indefinites with respect to determinacy of reference. If DOM is sensitive to determinacy of reference, it should not be surprising if partitives are, cross-linguistically, stronger DOM triggers than ordinary indefinites but weaker DOM triggers than definite DPs, which is what appears to be the case.

Note also that partitives turn out to be akin to possessives in the sense that their referent is explicitly connected to another, already introduced referent. The difference, however, is that in the case of partitives, determined reference is ruled out: the domain DP must refer to a sum entity. It is, therefore, not surprising that we do not find definite non-unique partitives on a par with definite non-unique possessives. In fact, if one cannot use a definite non-unique possessive because the context has already introduced the set of entities possessed, the partitive construction tends to be used.

(54) Ion are trei fete. Una din fetele lui Ion e la Timişoara.

I. has three girl.Pl one of girl.Pl.Def.Fem of I. is at T.
‘Ion has three girls. One of Ion’s girls is in Timisoara.’

In this example, the definite possessive fata lui Ion ‘girl.Def of Ion’ would be inappropriate in the second sentence.

We conclude that partitive determiners have the feature [PART] whose content, in DRT terms, is the requirement of the presence of a constraint of the form x  y on the discourse referent x introduced by the determiner.

Interestingly, the presence of a partitive determiner in the language does not block a partitive interpretation of the unmarked indefinite, on a par with what happens in the case of the unmarked definite/indefinite competition:

(55) Ion are trei fete. O fată e la Timişoara, alta la Cluj.

I. has three girl.Pl a.Fem girl is at T. other.Def.Fem at C.
‘Ion has three girls. A girl is in Timisoara, another in Cluj.’

The indefinite o fată ‘a girl’ here can, and in this context, would be, interpreted as referring to one of Ion’s three girls. We will see below that none of the marked indefinites block the use of the unmarked form. Exactly how the different determiners block one another is an interesting open issue.