The second language acquisition of subject pronoun distribution in English, Italian and Polish: testing the Interface Hypothesis

This study examines the claims of the Interface Hypothesis (IH) (Sorace and Filiaci, 2006) which posit that while narrow syntax properties may be fully acquirable by end-state L2 speakers, syntax-discourse interface properties may not. However, research evidence to date is mixed and suffers from theoretical and methodological flaws. This study investigates the IH predictions by testing for the second language acquisition of subject pronoun distribution in structures posited to contain i) narrow syntax and ii) syntax-discourse interface properties in three languages which vary in their realisation of these properties: English, Italian and Polish. 118 speakers of L1 English, Italian and Polish at different levels of L2 proficiency, including near-native, were tested using five tasks including oral production, interpretation and acceptability judgement/error correction.The main findings were that L2 near-native speakers are able to fully acquire both interface properties as well as narrow syntax properties, contra the predictions of the IH, although at lower levels of L2 proficiency speakers may experience developmental delays. Evidence from the interpretation task further suggests a default preference to establish coreference with the matrix subject irrespective of pronoun form overriding Italian and Polish native speaker tendencies to use overt pronouns to corefer with non-subject antecedents. Moreover, in the structures containing interface properties,the Italian and Polish native speakers expressed an unexpected degree of flexibility in their pronoun preferences in contexts where an overt pronoun would be expected, suggesting that there may not be such a clear division of labour between null and overt pronoun preferences in null-subject languages, contra the assumptions of Sorace and colleagues. Since this study finds no support for the IH and since speakers expressed consistent patterns of behaviour across both interface and non-interface structures, there is arguably no motivation for separating linguistic properties into different categories based on L2 learnability predictions.