Grammatical gender in language production: psycholinguistic evidence from Greek

Abstract

This thesis is concerned with the representation and processing of grammatical gender in Greek. It addresses two issues. The first concerns the conditions under which gender priming can be obtained; the second concerns the relationship between gender and other nominal categories, particularly case and number. These two issues bear upon the more general question of how lexical-syntactic properties are stored, retrieved and used during grammatical encoding, and how various consequences of the grammatical make-up of words are evident in the fluency of speech. Furthermore, insofar as grammatical gender constitutes a point of divergence a cross different languages the thesis uses Greek data to examine the scope of a particular production theory originally developed for typologically distinct languages.

The theoretical framework for this thesis is the production model of Levelt, Roelofs and Meyer (1999). A particularly attractive feature of this model is that it makes highly articulated proposals about the content and the mechanisms of access to lemma level representations that is, of abstract representations of words and morphemes. The critical claims for this thesis concern the distinction between inherent grammatical properties and diacritic parameters, and the conditions under which these properties are selected or merely activated.

The empirical part of the thesis comprises two sets of experiments employing primed picture naming. The first set (Experiments 14), which focuses on gender alone, investigates the linguistic contexts in which gender priming can be obtained. These include bare noun or definite determiner + noun primes and colour adjective or indefinite determiner + noun targets. The picture of gender that arises from these experiments is largely compatible with the Levelt et al. treatment of gender insofar as it shows that gender refers primarily to an abstract lexical-syntactic property. Also in line with what has been previously observed for other languages, gender selection, which occurs when agreement has to be computed between a gender controller and an agreement target, proves to be a pre-requisited or gender priming.

The second set (Experiments 5-8) focuses on the relationship between gender and case on the one hand, and gender and number on the other. Two possible accounts of this relationship (independent features and feature clusters) provide plausible yet extreme hypotheses about the way gender, case and number may be interrelated. Experiments 5-8 show that when other nominal properties are varied, gender priming can be obtained only with a particular type of prime noun phrase (i. e., definite determiner + noun). Furthermore, the effect of gender relatednessis more readily apparentw hen a single selection process, namely gender selection, has to be carried out on-line during the retrieval of the target lemma. In all, the present results from Greek converge with related evidence in the language production literature insofar as they show that prior access to gender information affects subsequent word retrieval. More important, the conditions under which this effect is observed confirm two basic assumptions of the Levelt et al. model concerning first, the abstract nature of lexical-syntactic properties and second, the processing distinction between activation and selection, and thus strengthen the viability of this account. Finally, the apparent sensitivity of the production system to the coordination of different selection processes e.g., genders election, cases election and number selection, suggests that although gender is independently represented its effect on word production is determined, at least in part, by the relative effect of the other selection processes.