The nature of syntactic gender processing in Spanish: An ERP study

Abstract

Syntactic gender as a lexical feature has been studied via picture-word interference paradigms in many languages. While effects have been found for noun phrase production in many languages, no effects have been found in Spanish, despite the fact that articles, nouns, and adjectives have a syntactic gender. Cubelli et al. (2005) found inhibitory effects in bare noun production in Italian which led to the hypothesis that such effects could be found for Spanish. Experiments 1 and 2 represented attempts replicate Cubelli et al.'s findings (Experiment 1 used an auditory distractor word and Experiment 2 a visual distractor), however no gender congruency effects were found. Experiment 3 attempted to generate congruency effects by requiring subjects to utilize gender-marked demonstratives and adjectives but still no effects were found. The lack of effects gave rise to the proposal that gender is not accessed during noun phrase production in Spanish and that the extreme regularity of the gender-marking system makes an article-plus-noun phrase more akin to a single lexical unit that can be accessed without an explicit synthetic process. Experiment 4 contrasted simple noun phrases that might be directly retrieved to constructions with long-distance dependencies, for which access to abstract gender features is relevant to parsing hierarchical sentence structure and aimed to distinguish these distinct cognitive processes via event-related potentials. The hypothesis was that a local gender violation in a sentence like "la piano" (the-fem piano-masc) would elicit a LAN as compared to the correct alternative, while a long-distance violation like "el piano que compré ayer es antigua" (the-masc piano-masc that I bought yesterday is antique-fem) would elicit a P600. All violations elicited a LAN and all violations involving adjacent segments elicited a P600; critically, the long-distance violation did not elicit a P600. It was concluded that the P600 reflects a repair process which occurs when such repair is not costly to the parser. Experiment 5 was a behavioral study using the stimuli from Experiment 4 with an error detection task which confirmed that subjects were sensitive to all error types.