The role of L1 typology and L2 proficiency in null subject transfer

This dissertation is a large-scale investigation on the omission of subject pronouns by learners of null-subject languages. A null-subject language is one where pronouns are either optional or unnecessary to make reference to a primary subject. In these languages, (which includes, Spanish, Italian, Arabic, Mandarin Chinese, Russian, Swahili, and many others) dropping a pronoun does not result in ambiguity. However, English is a non-null-subject language and a subject pronoun is always required in referring expressions. The aim of the present study is to determine if this difference in subject parameters means that learners are more prone to omit pronouns in second language (L2) English production. The first research question of the present study asks, ‘do null subjects in the L1 transfer to L2 English?’ This question becomes more complex when we consider that there are more than one type of null-subject language. How null subjects are licensed is in ‘canonical’ null-subject languages like Spanish is quite different from ‘radical’ null subject languages like Japanese. Typological classification, then, becomes an additional consideration and the basis of research question two. The literature has suggested, however, that regardless of L1, the effects of language transfer are still persistent at advanced levels of L2 proficiency. This means, that learners are still omitting pronouns in L2 English—beyond the B2 level—just as they would in their null-subject first language (L1). Research question three queries whether proficiency, rather than L1, is the key factor in accurate L2 grammatical production.

The majority of the research on these questions has focused on smaller-scale studies of L1 Spanish or Italian learners of English. To address this, I use data from the EFCAMDAT learner corpus to sample a large and typologically diverse set of learners. This corpus contains over a million writing samples from 174,000 learners representing 198 nationalities. For this study, writing samples from learners with L1 Arabic, L1 Chinese (Mandarin), and L1 Russian were selected and analysed for their accuracy in pronoun production and omission. The results of chi-square and logistic regression analysis indicate that null-subject transfer is limited. There was a clear effect of L1, but not typology. L1 Arabic and L1 Chinese learners had nearly identical rates of accuracy (p <.001) but L1 Russians performed far worse in comparison. A quadratic (U-shaped) function in performance by proficiency was observed for all L1s, however. Learners did not omit pronouns at the elementary (A1) level. By A2, there was a downward trend, indicating an increase in pronoun omission, which continued to the lower-intermediate (B1) level. By the upper-intermediate (B2) level, accuracy in pronoun production matched that at the initial A1 level, before fully resolving at the advanced (C1/C2) level. These results are contrary to those in previous studies on null-subject transfer which found a more linear pattern in accuracy. The empirical data provides evidence that L1 influences the acquisition of certain features in L2. This has implications on the learnability of certain features in SLA. The data also suggests that language development is non-linear, and learners may struggle with transfer at the A2/B1 threshold. However, in the absence of experimental data to support these findings, the implications are tentative pending further investigation.