A distinction among O-features: The role of gender and number in the retrieval of pronoun antecedents

We investigated the use of two O-features, gender and number, in pronoun antecedent identification in six experiments in Italian. In a previous work (Di Domenico & De Vincenzi 1996) we found that 1000 msec after the presentation of a pronoun only number information is used to restrict the set of possible antecedents for the pronoun, whereas gender information is not. Here we add two experiments with onset of the target word at 500 msec, which show that neither number nor gender are used. The conclusion is thus that while number information is used within 500 and 1000 msec to restrict the reactivation of pronoun antecedents, gender is not. The results are compared to previous experiments conducted in English by Nicol (1988). The author found an immediate use of number information but a less clear result for gender, possibly due to the fact that in English gender is not marked on the same way as number is. Our study in Italian, with nouns where gender and number are equally overtly and regularly marked, suggests that the different use of gender and number is not specific to English nor has to do with the surface markedness of features, but, on the contrary, is to be attributed to an intrinsic difference between the two features. This difference is characterized in terms of a different syntactic representation of the two features, as assumed in recent proposals in linguistic theory, and of the parser's modular use of linguistic information.*