The paper reports the results of an experimental study of second language (L2) acquisition of Russian verbal morphology by the American formal classroom learners of Russian. It contributes to a better understanding of the psycholinguistic mechanisms underlying the processing of inflectional morphology in two areas. First, it provides empirical data on how the Russian verbal system is represented in the mind of native speakers and L2 learners, and addresses the issue of whether the one-stem verb system (Jakobson, 1948) is psycholinguistically real. And second, it explores the role of linguistic probabilities in the processing of complex inflectional morphology.
The reported data make it possible to test certain claims made in the framework of psycholinguistics and formal linguistics. Formal linguistics and psycholinguistics appear at the opposite poles of the spectrum of views on the role of learning, input frequencies, and probabilities in the processing of inflectional morphology. Thus, linguistic theory claims that (a) linguistic processing is based on symbolic rule computation, (b) symbolic rules are applied regardless of frequencies and probabilities, and (c) minor rules are used before more general rules, and block their application. In contrast, psycholinguistics (a) does not make a commitment to the exclusive use of symbolic rules in linguistic processing and allows for the use of associative patterning, analogies, etc., (b) claims that frequencies and statistical probabilities play an important role in linguistic processing, (c) assumes that more general and frequent rules are applied first. The paper sheds new light on the on-going debate between the proponents of dual- and single-system approaches to inflectional morphology. The dual-system approach, which stems out of formal linguistics, postulates that symbolic rule computation involved in the processing of regular morphology is independent of input frequencies, while the single-system approach, which is linked to connectionism, claims that regular morphological processing is based on input frequencies and phonological associations, and does not engage the use of abstract rules.
The data were collected in a real and nonce verb generation task. The material involved the verbs belonging to the –aj-, -a-, -i-, and –ova- classes according to the one-stem verb system and ranging in type frequency, productivity, and morphological complexity (degree of regularity). The subjects were beginning and intermediate American college-level students of Russian and adult native Russian controls. The results demonstrate the influence of input frequencies both on adult L1 and L2 processing of regular verbal morphology. At the same time, L2 processing relied less on the default conjugational pattern than native processing, which can be accounted for by the differences in native and L2 input frequencies. The response patterns to the different verb classes indicate that the choice of the conjugational pattern is rule-based rather than analogy-based. Taken together, the facts that regular verb processing is rule-based, and at the same time is influenced by statistical probabilities, lend support for the proposed Rules and Probabilities Model of morphological processing, which combines symbolic rules and linguistic probabilities, and thus challenges both the dual-system and single-system theories.