Contenu du sommaire : La syntaxe de la coordination, sous la direction de Danièle Godard et Anne Abeillé

It is widely accepted within Principles and Parameters theory that coordinate structures are Conjunction Phrases, in which the first term is the specifier, the conjunction the head, and the second term the complement. They differ from typical specifier-head-complement structures in various ways, so at most they could be exceptional instances of such structures. A consideration of the distribution of coordinate structures, the number of conjuncts (or terms), the nature of conjuncts (or terms), and the basic structure of coordinate structures suggests that there are no significant similarities between coordinate structures and specifier-head-complement structures and a consideration of certain unbalanced coordinate structures points to the same conclusion. It seems, then, that the Conjunction Phrase analysis is untenable.

It is well known that “coordinating conjunctions” form a subconstituent with the phrase that follows them ([conj X]). We show that such a “conjunct phrase” can have different grammatical functions: it can be coordinated in a coordinated phrase, it can be adjoined to several categories, it can also appear as an independant sentence. We focus on French incidental and emphatic conjunct phrases:Jean est parti, et rapidement. (‘Jean has left, and rapidly')Jean est en avance, ou je me trompe. (‘Jean is early, or I'm wrong')Contrary to their properties as an element of a coordination, these conjunct phrases are mobile, do not trigger plural agreement and do not obey the constraint on parallel extraction. We analyze them as syntactic adjuncts, and extend the same analysis to phrases introduced by car in French, and to asymmetric verbal “coordinations” in Korean. We formalize our analysis with the phrase structure grammar HPSG, and propose to analyze conjunctions as “weak heads”, sharing most of their syntactic features with the following phrase.

Two kinds of coordinate structures can be distinguished in French according to the distribution of conjunctions. In simplex coordinations, the conjunction introduces the last conjunct and optionally appears before non-initial conjuncts (Paul (et) Jean et Marie) while in correlative coordinations, which are the focus of this paper, the conjunction appears before each conjunct, including the first one (et Paul *(et) Jean et Marie). We present arguments against asymmetric analyses according to which the initial conjunction should be analyzed as an adverb (more precisely focus-sensitive adverbs), homonymous with the conjunction and adjoined to the coordinate phrase. Instead, we adopt a construction-based variant of the symmetric analysis according to which each conjunct is introduced by one and the same conjunction. We couch our analysis in the phrase structure grammar HPSG. We assume [conj X] phrases to be head-complement-constructions while coordinate structures are a type of non-headed construction. We define two subtypes of coordinations to account for the distribution and specific properties of simplex and correlative constructions.

This article examines French NPs of the form N Conj N (Il a rangé livres et cahiers ‘He put away books and notebooks'). Since French does not allow NPs without a determiner as semantic arguments of a predicate in a general way, it is clear that their acceptability derives from the presence of the coordinating conjunction, an intriguing phenomenon. Moreover, such NPs raise serious problems for the widespread assumption according to which French is a language with no “Bare Nouns”. On the basis of a cross-linguistic analysis comparing both their distributional and their semantic properties with their (non)-coordinated equivalents in Romance languages like Italian and Germanic languages like English, it is shown that the presence of the conjunction leads to striking and unexpected similarities of the French N Conj N construction with Bare Nouns of the Germanic type.

The terms of a coordination can differ to a certain extent by certain properties, e.g. part of speech, case or person. We propose that the correct description of such phenomena involves radical underspecification. We allow, for example, that an NP can be resolved as [CASE case] or [CASE direct-case], even if, in the language in question, case and direct case can be further specified as nominative or accusative. Moreover, constraints can require that CASE values can be, for example, “at least as specific” as dative, when there exists the possibility of conjunctive types (e.g. accusative&dative). We show how a number of puzzling data in French, English, German or Polish can be accommodated by general rules of coordination and the interaction of constraints affecting the upper and lower bounds of CASE values.