Pronominal Anaphora and Sentential Antecedents

Clausal arguments (henceforth argument CPs) have been assumed to lack any specification for the features of Person, Number and Gender (henceforth PNG features). Subject and object CPs have also been assumed to lack case (Nominative or Accusative). If so, the conclusion follows that arguments of predication are of two different types with respect to the application of agreement mechanisms, computational or interpretive: those that can participate in them (nominals) and those that can not (argument CPs).

The previous assumptions and their subsequent conclusion are very problematic. On the one hand, anaphoric relations between some types of pronouns and their CP antecedents can not be properly accounted for. Anaphora involves agreement in PNG features (or a subset of these) between a linguistic antecedent and the anaphoric element linked to it. If CPs lack any specification for PNG features, CPs should not be able to function as the linguistic antecedents of any pronoun, contrary to fact. On the other hand, the claim that argument CPs lack also specification for case content renders the computational system very complex under Minimalist guidelines (Chomsky 1995, and subsequent literature). Note that syntactic operations can not be assumed to apply blindly between arguments and functional categories, as they should. Therefore, either the mechanism should be provided with an unwelcome look ahead requirement or some ad hoc stipulation must be adopted to account for the sensitivity of the system to the categorial label of the elements bearing a thematic role.

In this study, I offer empirical evidence suggesting that the assumption that argument CPs lack both specification for PNG features and a value for case content is incorrect. Constructions involving anaphora with possessive pronouns, floating quantification, sloppy readings in VP-ellipsis contexts and the distribution of interrogative elements are examined. The interest of these structures lies in the fact that they involve relations that are sensitive to the value of (at least one) the PNG features. It is shown that an anaphoric relation must involve matching in PNG value(s) between an anaphor and its linguistic antecedent. The contrasts between grammatical and ungrammatical constructions show that argument CPs must be endowed with PNG content. I also suggest the hypothesis that subject and object CPs must have specification for case. This hypothesis can account for the distribution of abstract agreement mechanisms in all types of embedded subjects.

In conclusion, the present proposal simplifies the computational system and can account for the facts involving anaphora between CPs and pronouns. It is also shown that a distinction should be made between the formal feature content of a syntactic object and the morphological expression of that content.

References

Chomsky, Noam (1995) The Minimalist Program, Cambridge, Mass., MIT Press.