The learnability of gender agreement in Spanish bilinguals

Achieving high levels of proficiency in a second language (L2) represents a great challenge. One of the domains where second language learners show persistent difficulties is morphosyntax, and specifically grammatical gender. This chapter will examine whether distributional regularities can improve the learnability of complex grammatical constructions, such as gender agreement dependencies. Previous behavioral studies seem to suggest that formfunction regularities (e.g. the presence of consistent correspondences between the gender of a word and its ending) facilitate L2 grammar acquisition and might have an impact on L2 gender agreement learnability. We will focus on this topic by examining the impact of gender-to-ending regularities on the time course of agreement processing in monolinguals and bilinguals. Four EEG (electroencephalography) experiments will be taken into account with four different experimental samples: Spanish monolinguals, Basque-Spanish bilinguals, Spanish deaf readers, English-Spanish bilinguals. Experimental sentences were presented in Spanish, a Romance language with a strong gender-to-ending consistency. In all experiments gender-to-ending consistency and grammatical gender agreement were manipulated. Nouns that were transparent (the ending is a reliable gender cue) and opaque (the ending is uninformative of gender) were presented embedded in Spanish sentences. These target nouns created grammatical and ungrammatical gender agreement dependencies. Event-related potentials (ERPs) were derived for each condition to test whether formal gender cues could affect agreement computation. The ERP results showed that while Spanish formal gender cues are not always crucial in native agreement processing, the role of gender distributional regularities becomes more important when Spanish is the non-native language. In non-native Spanish speakers, having a weak lexical representation of grammatical gender might lead to treating transparent and opaque nouns differently and over-relying on form-function regularities to compute agreement dependencies. These results suggest that the weight of formfunction regularities during agreement processing is not fixed and can change depending on language experience.