Theoretical background of syntax of pre- and postnominal adjectives

To explain the syntax of pre- and postnominal adjectives I adopt the basic assumptions of the Minimalist framework as outlined in Chomsky (2001a, b). In this approach the fundamental syntactic operations are “Merge” and “Move.” (External) Merge, the operation of basic-structure building, is an operation imposed by the recursive nature of language: it takes two syntactic objects, A and B, and creates a new object consisting of the two {A, B}. Move (also called “internal Merge”) takes an element B already constructed by external Merge and places it under the c-command of a probe A. As to the motivation for Move, it is considered to be necessary in derivational approaches to express the fact that certain elements appear in non-theta positions for reasons of scope, or to manifest informational or discourse-oriented properties.

Internal and external Merge are both facets of so called “Set-Merge.” In addition to Set-Merge, another way of yielding syntactic objects out of already constructed units can be conceived. In fact, “Pair-Merge” or adjunction is such an operation. Pair-Merge is asymmetric: it takes two elements (one of which is already built) and adjoins one of them to the basic projection, taking its label. Pair-Merge “has no selector and is optional” (Chomsky 1998: 51).¹

In a strict minimalist system, every device employed has to be sustained on “conditions of computation efficiency and the interface conditions that the [linguistic] organ must satisfy for it to function at all” (Chomsky 2001b: 3). If this quite strong position holds, any syntactic derivation D should provide a pair of forms legible by the Phonetic and Semantic levels or interfaces which are respectively accessed by the sensory motor and the conceptual–intentional (C-I) systems. These systems impose conditions on the operations which are active in narrow syntax. As for Merge and its relation to the semantic module SEM, if it “comes free” (Chomsky 2001a: 3), and if it has to provide units easily mapped onto the interface level SEM, it is conceivable that there could be a correlation between the semantic properties required by the C-I interface and the structures provided by the operations in narrow syntax.

The fundamental interface semantic properties discovered and elaborated along the history of formal grammar are (i) properties related to theta theoretic relations, namely, to predicate–argument and predicative relations; (ii) properties deriving from the “composition of predicates;” and (iii) discourse-related properties. Theta-theoretic relations express s-selection and obey c-command; composition of predicates expresses the necessity of a predicate to act as an A    operator binding a variable when no subject is available; discourse-related properties are A or A      ́́            relations resulting from the way information is organized.

It appears that the category labeled Adjective manifests these three semantic properties and therefore gives rise to the three operations. The proposal I will elaborate on in the following section shows three possibilities.² The first one is based on the idea that certain adjectives (externally merged in DP) will be interpreted as expressing, roughly speaking, theta-theoretic requirements: they will be predicates selecting Ns as their “subjects.” This is the case of predicative adjectives – usually postnominal in Romance languages . As a second possibility, other adjectives, those which interact with a functional category above NP (nP) as their adjuncts, will have the semantic properties associated with the “composition of predicates”; I understand as such the operation of one-place predicates that modify elements in N. This is the case of non-predicative adjectives – usually prenominal ones . In the third possibility, adjectives with a predicative interpretation may be moved from NP to the edge of nP to receive a focus interpretation.

The semantic relations between adjectives and nouns result then from the configuration obtained when certain lexical categories are merged with the appropriate heads. In other words, the interaction between lexical semantic interpretable (valued) properties of adjectives and the configurations in which they appear provides the interpretation of DPs containing adjectives. I claim, moreover, that adjectives come from the lexicon encoding uninterpretable formal features (gender and number) and (valued) semantic features. Adjectives will end up with a specific semantic interpretation, predicative or non-predicative: [+p] or [−p], according to their position after external Merge. Of course, [+p] or [−p] are only convenient ways to represent the many nuances of the two distinct but not always univocal semantic interpretations that adjectives may receive in pre- and postnominal position.

Regarding the structure of DP, I assume a quite strict – although not total – parallelism with VP and I propose that, besides NP and DP, there must be an nP category whose head is a light n (Carstens 2000; Adger 2003). This head contains uninterpretable phi-features and may project a possessive “agent” in its Spec. This head becomes a probe to delete the (un)interpretable gender and number features of a goal N. I assume that, aside from nP and from conceptually and empirically necessary functional projections like, perhaps, DemP, there are no other functional categories in DP.

1 According to Chomsky (2001b: 16) the interface condition which imposes Pair-Merge appears to be the necessity to produce “composition of predicates.”

2 Demonte (2005) contains a more comprehensive elaboration of this hypothesis.