Lingvisticæ Investigationes - Volume 13, Issue 1, 1989

Within the last twenty years and in the framework of transformational grammar at least seven kinds of se have been proposed: spurious se, reflexive se, reciprocal se, pronominal se, impersonal se, passive se and se moyen. Each of these se's shows its own syntactic and semantic characteristics. In this article, in the framework of the Theory of Government and Binding, an eighth type of se, the impersonal passive se, is proposed. Similar to the passive se, it is passive in interpretation; however, unlike passive se, and similar to the impersonal se, it has an impersonal subject. The different syntactic properties this new se exhibits are shown to follow from the principles of universal grammar.

RELATIVES AND NOMINALIZATION: WHERE GUILLAUME TURNS TRANSFORMATIONALISTIt is a little-known fact that by 1940 G. Guillaume had already developed a genetic syntax in which most embedded clauses were generated transformationally. But while the mechanism he postulated — internal or external nominalization — claims to reflect mental reality, an examination of the data shows, at least as far as relative clauses are concerned, it does not meet even the minimum requirement of observational adequacy. Although the falsification of an inadequate hypothesis is a step forward for any theory, the fact that in this case it has taken nearly fifty years to make it denotes a disturbing lack of rigour in the methodology of Psychomechanics.

The constitution of electronic dictionaries, i.e., directly usable by a computer, leads to several difficulties both linguistic and computational. The content of such dictionaries (entries and rules) must be defined in a precise and exhaustive way, using linguistic criteria, and must be based on an extensive description of the French language (or at least on a description of a specific domain of French, e.g. sports). Information must be systematic and consistent. When this analysis is done, there remains a problem with the representation of data. Different solutions are considered, such as lists or matrices. The method suggested here for representing compound words uses the simple mathematical model of a finite automaton.

The article presents a critical review of the various solutions that have been proposed, within the framework of generative phonology, regarding the problem of the morphophonological relations between the sounds [e, e, e] in French. In order to remedy the flaws in the above mentioned approaches that have been examined, a new model is developped, which may be considered as an extension of lexical phonology, invented by Kiparsky (1982) and Mohanan (1984). The model makes use of paracyclic rules beside the cyclic rules. It will be demonstrated that the recourse to paracyclic rules offers some advantages with respect to the postcyclic rules advocated by Booij & Rubach (1984, 1987).