Croatian Mixed-Gender Conjunct Agreement: An ERP Study

In a recent elicited-production study with native speakers of Slovenian, Marusic, Nevins, and Saksida (2007) and Marusic, Nevins, and Badecker (2015) show that there are three distinct variously attested gender-agreement grammars. In this study, the high temporal-resolution of the ERP (event-related potential) technique was used to detect neurological components and measure the processing cost of the three gender-computing mechanisms. The study is comprised of two acceptability-judgment experiments, using a factorial design with nonmasculine mixed-gender conjuncts. Experiment 1 contrasts two strategies, Distant-(DCA) and Closest-Conjunct Agreement (CCA), to question whether the linear distance between a participle and the two conjuncts is language-or memory-related. The Experiment 1 results show behaviorally an overall significant effect of gender; and neurologically a memory-related component, the P300. Experiment 2 sets out to detect alternations to the processing cost when default (Def) agreement is added to the experimental paradigm. The Experiment 2 results indicate no gender effects; instead, two language-related components, N250 and N450, were observed, statistically picking out DCA once again. We argue that in an ecologically valid environment where all three grammatical options are made available, processing of DCA is no longer supported by a general cognitive mechanism, such as memory, but is rather computed by language-related processes.