The present thesis presents a series of experiments investigating potential (syntactic) attrition effects in Italian beyond traditional interface structures. This is motivated by a critical review of the previous literature on L1 attrition primarily focusing on the Interface Hypothesis (Sorace & Filiaci 2006). From this, we argue for two relevant patterns: (i) the often discussed relaxation of discourse constraints for specific forms (e.g., overt pronouns in null-subject languages) and (ii) a tendency, which we believe has so far gone largely unnoticed, for speakers to rely more heavily on other (interpretive) biases already present in L1. To explore this second pattern and disentangle the role of interpretive/processing biases from the syntax-discourse interface, we compare pronominal subjects which are often taken to sit at the syntax-discourse interface with (pseudo-)relative clause attachment ambiguities. We argue that this is an interesting comparison as such attachment ambiguities involve two ambiguities which are both conditioned by principles of computational efficiency at the level of the parser and are not related to discourse. In a language like Italian, when encountering an embedded CP of the relevant type, we will assume that the parser must decide whether the string contains a relative clause (RC) or a pseudorelative (PR) which are structurally distinct but string identical (PR-FIRST HYPOTHESIS, Grillo & Costa 2014). Moreover, in the case of a RC reading with multiple possible attachment sites, the parser must additionally decide where to attach the embedded CP (LATE CLOSURE, Frazier 1978). Should structures at more external interfaces (e.g., the syntax-discourse interface) be more vulnerable under attrition, we would expect the two phenomena to be affected deferentially under attrition. However, should attrition lead to an increased reliance on certain biases regardless of interface status, we might expect the two phenomena to pattern together.

To compare the effects of attrition on the offline interpretive biases for the two phenomena, we carried out a sentence interpretation task with globally ambiguous items of both types in which we manipulated the pronominal forms and pseudorelative availability. To compare online processing biases, we conducted a self-paced reading task in which we additionally forced pronominal resolution and ‘RC' attachment. Given some unexpected results from the control group for the self-paced reading task, we additionally present a sister eye-tracking-while-reading task as a validation (albeit with a different methodology) of the control baseline. For the experiments in the present thesis, all participants reported growing up monolingually in Italy. However, experimental groups differed from control groups in that the former had immigrated to a majority English-speaking country (e.g., the UK) during adulthood and had lived there for a minimum of 5 years. Recent re-exposure to the L1 was controlled for.

Results from the offline interpretation and online processing of both phenomena indicated that, where attested, attrition always surfaced an increased reliance on L1 biases. Strikingly this holds even for the prime syntax-discourse interface phenomenon, overt subject pronouns in a null-subject language like Italian. As such, we argue that future accounts for the attrition of syntactic phenomena should not focus on the specific cognitive demands of particular interface phenomena (e.g., pronominal forms and the integration of contextual information) but instead should seek a more parsimonious account couched in terms of increased reliance on L1 biases in ambiguity resolution more generally.