Abstract
The paper provides a quantitative analysis of L1 interference in Italian as an L2. The study is based on a corpus of intermediate and advanced interlanguages of Italian, collected from a sample of immigrant Spanish-speaking women. As in the corpus the variation between L1 borrowings and target-like forms seems a permanent feature, the paper tries to account for optionality in L2 as a stable and structural phenomenon in non native grammars. The variable analyzed is the couple of function words de/di and the hypothesis tested through multivariate analysis is what factors affect the presence of the Spanish variant de. Using the computer program VARBRUL, it has been developed a statistical model formed by four variables: the degree of acquisition, the social network structure, the preceding phonological context, and the syntactic function. The results of the analysis show that 'degree of acquisition' and 'social network' are interconnected and, as a cross-product variable, they are the most important factors affecting the presence of the L1 variant de.