Processing of Grammatical Gender, Number and Case: Experimental Studies on Russian

Agreement processing is essential for parsing and speech comprehension, especially in languages with rich inflection. One of the questions about agreement processing concerns the differences in processing of morphological features. Previous studies on various languages mostly focused on processing of number and gender in sentences with agreement mismatches. Some studies observed that processing of gender agreement mismatches is more costly than processing of number mismatches. Other experimental accounts reported an opposite effect or found no differences between number and gender. Case was almost never added to the picture. We conducted three experiments using both behavioral methods (self-paced reading) (Experiments 1-2, N = 88 and 68 participants respectively) and electrophysiological methods (ERPs) (Experiment 3, N = 14 participants). The aim of our study was to compare gender, number and case in real-time processing of Russian sentences with postpositive participial modifiers (Experiments 1, 3) and to compare gender and number in predicative agreement (Experiment 2). Experiment 1 was a continuation of the first year’s work (we present an experiment with a larger participant sample). Experiment 2 was designed during the second year of the MA program. Experiment 3 was a pilot ERP study conducted in collaboration with the colleagues from IEPhB RAS. We observed that case mismatches were the least salient in the behavioral experiment and required additional processing before being recognized as an agreement error in the ERP experiment. Number mismatches were the most salient in the ERP experiment and were more salient than gender in the behavioral experiment on predicative agreement. Gender was between number and case with respect to saliency in the ERP experiment and its effect occurred later than the effect of number in the experiment with predicative agreement. Interestingly, the effect of gender lasted longer for attributive agreement. Our stimuli also allowed to compare grammemes. In predicative agreement, gender mismatches in sentences with masculine heads induced more processing cost than in sentences with feminine heads. Surprisingly, similar asymmetry was not observed for number grammemes: number mismatches were always salient. We attempted to contextualize our results with respect to previous theoretical and experimental works and outlined some future research directions.