This dissertation provides a comprehensive analysis of the L2 acquisition of Spanish psych-verbs (e.g. gustar `to like') across four different proficiency levels. In particular, psych-verbs constitute a testing ground for the predictions of the Interface Hypothesis (Sorace and Filiaci, 2006; Tsimpli, Sorace, Heycok & Filiaci, 2004; Sorace, Serratrice, Filiaci & Baldo, 2009; inter alia), one of the most influential theories in current generative second language acquisition. Its main claim is that properties that hinge on external interfaces (i.e. those that require the interaction between a linguistic module and a cognitive module) are more problematic for learners than those that do not hinge on that interface (i.e. internal interfaces/narrow syntax). In order to assess the empirical adequacy of the IH, this project encompasses five experiments that test different syntactic properties of psych predicates as well as phenomena that belong to both internal and external interfaces. The results of this study indicate that clitic and verb agreement is the most problematic aspect of psych-verb acquisition in accordance with the previous literarture in the field (e.g. Montrul, 1998, 2001). As for the issue of interfaces, this project is only partially consistent with the proposals of the IH. Whereas external interfaces present a certain level of difficulty for some groups of L2 learners, the low-proficiency participants are sensitive to pragmatic factors in spite of their lack of mastery of the morphosyntax of these constructions. Thus, external interfaces are problematic for L2ers but not more so than internal interfaces. Additionally it is not a necessary condition that syntax will precede the understanding of pragmatic phenomena. Instead, pragmatics can come for free in L2 acquisition while the learner still struggles with the target syntactic templates. Because of these inconsistencies with the IH, I turned to a more articulated model, the Integrative Model of Bilingual Acquisition (Pires & Rothman, 2011), that accounts for the differences between native and non-native speakers by resorting to the interplay of a series of factors (i.e. formal complexity, L1-L2 parameter mapping, processing resources and primary linguistic data). I argue that this more sophisticated model not only is able to more successfully account for the patterns found in this dissertation but it is also a more integrated explanation for the intricacies of the acquisition process.