Examining the nature of variability in gender and number agreement in native and non-native Spanish

We explored the nature of morphological variability in English-speaking learners of Spanish, examining whether the variability is systematic and whether performance is impacted by task demands. McCarthy (2008) proposed that `default' errors, referring to the incorrect extension of masculine/singular morphemes to contexts requiring feminine/plural agreement, are more frequent than the opposite pattern, called `feature clash' errors. According to representational accounts of morphological variability, the source of the errors in L2 learners is the inability to acquire features absent from the L1. In contrast, computational accounts argue that this variability in learners is due to processing issues. In order to test these hypotheses, our L2 leaners took either a Speeded Grammaticality Judgement Task (GJT) or an Untimed GJT. Three groups of Spanish natives took a Speeded GJT at three different rates of presentation (Speeds 1, 2, and 3) in order to examine whether natives under pressure perform similarly to learners (Hopp, 2010). The results show quantitative effects of speed for both natives and learners with decreased rates of accuracy in both groups. Natives were significantly less sensitive to errors as speed increased. For the learners, all proficiency levels showed increased sensitivity in the untimed task, with individuals in the advanced group showing target-like performance. With respect to the qualitative nature of the errors, differences between `default' and `feature clash' errors emerged for the natives in Speed 2 and the intermediate L2 group in the Speed GJT. The results for both groups showed more sensitivity to feature-clash than default errors for number. For gender, L2 learners showed the opposite pattern, while no differences were observed in native speakers. Overall, our results are more consistent with computational accounts of morphological variability