Representation and processing of grammatical genderanalysing the gender congruency effect

Resumo

Grammatical gender is one of the most controversial issues in the study of language, having frustrated the work of many philosophers, linguists, and more recently, psycholinguists. Indeed, in recent decades special attention has been devoted to the link between grammatical gender and cognition. Yet, although the picture seems rather clear in terms of gender acquisition and retrieval during language comprehension, in the area of language production we are close to what we might call scientific havoc. In short, many of the on-going debates here involve disputes about the mandatory character of an agreement context for the retrieval of a noun’s gender value, the automatic or competitive-based nature of the mechanisms behind gender selection, and different views on the link between morphophonology and gender processing. These matters have traditionally been assessed through the picture-word interference (PWI) paradigm, in which native speakers of different languages have to name aloud pictures using target nouns while ignoring superimposed written distractor words, and, more recently, through naming and forward translation tasks with learners of second languages (L2s). In all these tasks, conditions of gender congruency and incongruency are created through the matching or mismatching of the gender values of either targets and distractors or the equivalent translations of the L2 and the first language. Variations on the response times of the participants are expected depending on the congruency/incongruency between gender values. More specifically, authors were looking for a gender congruency effect (GCE), by which pairs of nouns of the same gender speed up the response. However, as regards native language production, the presence and type of gender effects seems to depend on the language family and the type of response given by participants. In the case of Germanic and Slavic languages, the GCE is only observed with noun phrases that include elements of agreement, such as the singular definite articles in German (“der” [masculine] vs. “die” [feminine]). Critically, Caramazza et al. (2001) argue that because the so-called GCE depends on the presence of determiners, the effect reflects competition for selection between determiner forms of different gender (a determiner congruency effect), rather than between gender values themselves. Hence, they conclude, gender values do not compete for selection and are retrieved automatically. For Romance languages, results are mainly null, except for a gender incongruency effect that is sometimes observed in Spanish and Italian when bare nouns are used to name the pictures (gender incongruency effect, GIE, facilitation for nouns of different gender). In bilingualism, the scenario is different, since a cross-linguistic GCE is consistently obtained regardless of the language family and the type of naming response involved (noun phrases vs. bare nouns). Hence, this reflects genuine effects of gender competition. The main aim of this thesis is to shed light on this intricate group of findings and to discover the nature of the mechanisms underlying gender selection. To do so, we defined our object of study from both linguistic and psycholinguistic perspectives. We then proposed a clear definition of gender transparency based on gender phonology and morphology, as well as a standardized method that seeks to operationalize the degree of gender transparency of languages through an adaptation of Audring’s (2019) concept of relative complexity of gender systems. We then conducted two meta-analyses, of the PWI paradigm and of naming/translation tasks, respectively. We did so by comparing the effect sizes related to the gender congruent and incongruent conditions through the computation of the Hedges’ g measure. Results showed that the determiner congruency effect is of medium size and is highly significant, whereas the cross-linguistic GCE is small but also significant. Interestingly, the GIE was the smallest effect. We then conducted three studies with speakers of a hitherto unexplored language in this context, European Portuguese, an interesting choice in that it is a member of the Romance family and exhibits the same transparent gender cues as Spanish and Italian, but has one-to-one gender-determiner mapping, as in Germanic languages. Our first study included two experiments featuring PWI tasks to test the presence of gender effects with bare nouns, as well as the role of the animacy of the targets. A significant GCE, although one which was not supported by Bayesian analysis, was restricted to the condition featuring only inanimate nouns. Our second study included three PWI tasks that attempted to replicate the GCE obtained in Study 1, and also to test the determiner congruency effect with noun phrases, taking into consideration the role of ortho-phonological gender transparency. We failed to replicate the GCE but obtained a determiner congruency effect that interacted with the gender transparency of both targets and distractors. Finally, our third study consisted of a forward and backward translation task of bare nouns with early bilinguals of Portuguese and German. A cross-linguistic GCE was obtained that depended on the proficiency of the two languages. Results were interpreted through the classical models of language production, the WEAVER++ (Levelt et al., 1999) and the Independent Network models (Caramazza, 1997), as well as through the Double-Selection Model (Cubelli et al., 2005), the Late Selection Hypothesis (Miozzo & Caramazza, 1999), and our own adaptation to language production of the Dual-Route Model (Gollan & Frost, 2001). As a whole, the body of results can be seen to be better in line with the notion that gender values do compete. The case is clear when it comes to bilinguals, for whom the tasks used are quite simple, and for whom there is an increased flow of activation reaching gender nodes from multiple lexical entries. However, the answer is not so clear in the case of monolinguals: we detected notable limitations on the PWI paradigm, the complexity of which may render it incapable of capturing certain small effects of competition, such as those concerning gender values. Indeed, we note that (1) there is evidence that using more fine-grained techniques such as fMRI can result in competition between gender values being detected, even in German (see Heim et al., 2009); (2) there is an urgent need for the replication of findings in the field with new studies using larger participant samples, better stimuli control, and more precise measuring techniques; (3) the absence of effects of the PWI paradigm cannot be interpreted as evidence for the absence of effects of competition between gender values (Vasishth & Gerlman, 2021); thus (4) it might ultimately be the case that the field even requires a paradigm shift in terms of the experimental approach taken.