Perceptual prominence and morphological processing in initial Second Language Acquisition

Within the field of Second Language Acquisition, the present contribution intends to evaluate the factors affecting the perceptual prominence of case endings in initial learner varieties. Research has so far demonstrated that initial learners are able to identify regularities in the speech stream even after minimal exposure (Gulberg et al. 2012) and that in attempting to process the input they recur to all the available linguistic knowledge deriving from their L1 and any other known languages (Rast 2010). This study aims at advancing research on the initial stage of input processing, i.e. perception, by evaluating the accuracy with which case endings are produced as a function of frequency in the input, target item transparency and word order. The data are part of the Italian sub-corpus of the VILLA project (Varieties of Initial Learners in Language Acquisition), a multinational initiative aimed at exploring the very first stages of second language acquisition (Dimroth et alii to appear). 31 carefully selected participants with Italian L1 and no previous experience of Slavic languages were exposed to a 14-hour course of Polish: in order to enhance cross-linguistic comparability of the data, the course was taught by the same trained native speaker in all the countries involved and in the most uniform way as possible. The input was recorded and transcribed, so as to account for the crucial parameter of “frequency”. Throughout the course, participants also took a series of tests aimed at evaluating their developing passive and active competence in the second language and at measuring psycholinguistic variables. This contribution is based on an Elicited Repetition which the learners took twice, after 9 and 13.5 hours respectively, in order to correlate learner’s performance with item frequency and trace the development of interlanguage phonology. The target items were sentences featuring SVO or OVS word order, e.g. artystka pozdrawia tłumaczkę ‘the artist (f.) greets the translator (f.)”. Learners were asked to listen to each sentence, draw a simple geometrical figure (as a distracting pause of ca. 5 seconds), and then imitate the sentence heard. The same target lexical items were repeated four times in the test sentences. In the focus of this study are the feminine nominal endings /a/ and /e/, corresponding to the values NOM and ACC respectively. Polish exhibits a rich nominal morphology which expresses number, gender and, crucially, case: it is hypothesised that the absence of the latter in the source language, Italian, will make case oppositions problematic for learners to perceive and acquire (Rast & Wątorek 2011). The results show that in spite of its negligible frequency in the input, the ending /e/ (ACC) was identified and reproduced by all learners, which confirms their sensitivity to the phonic substance of the stimulus. However, the difference in the accuracy rate of the two endings proved to be statistically significant, most errors involving /e/ (ACC) rather than /a/ (NOM). Secondly, word order appears to play a significant role in that errors are more common in OVS sentences rather than SVO ones, hinting at both universal constraints (Peters 1985) and the relative frequency of such syntactic structures in the input (Rast 2008). Item transparency also proved to be strongly significant. No effect of period (t1, t2) has been observed. Finally, possible correlations with working memory resources (Juffs & Harrington 2011) are explored. The data suggest that learners are more likely to perceive differences in case endings under specific conditions, i.e. learners do find specific contexts more prominent than other: maximum prominence (minimum error rate) may be expected for the ending /a/ (NOM) within transparent words constituting the first noun of SVO sentences, whereas minimum prominence (maximum error rate) may be expected for the ending /e/ (ACC) within non- transparent words constituting the first noun of OVS sentences. It is hypothesised that the significant decrease in accuracy rate observed in the least prominent contexts is the result of a sort of “deafness” to case endings, which in turn is determined by specific combinations of frequency, item transparency and word order values.