This paper reports on a study of the acquisition of reflexive and reciprocal forms in L2 Italian from the point of view of their argument structure. The results are interpreted with respect to the debate about whether transfer is monolithic (Schwartz and Sprouse 1996, Whong-Barr 2005) or modular (Montrul 1997), considering at the same time other relevant factors — innate mechanisms and L2 input. On the basis of the acceptability judgements of Serbian and English learners, compared to native speakers of Italian, it is argued that a modular approach is more realistic, as different types of transfer were identified for morphology (direct transfer of L1 marking) and argument structure (indirect influence of L1 general properties). Possible reliance on innate knowledge was detected at the argument structure level, while the input (examined through a corpus study) was found to influence both domains in different ways. It is also suggested that the three factors should be looked at jointly rather than in isolation.