This article presents an experimental study investigating the compounding parameter in the L2 Spanish interlanguage of English and French NSs in light of the Subset Principle and its predictions for the process of L2 development. The compounding parameter (Snyder, 1995, 2001) argues that languages permit complex predicate constructions like verb particles, resultatives, and double objects if and only if they can productively form N-N compounds. English exhibits the plus value of the parameter, allowing N-N compounds and the related constructions, whereas in Spanish and French these compounds and constructions are ungrammatical. Because English also allows periphrastic constructions of the same meaning, which are the only option in French and Spanish, English represents the superset parameter value to the Spanish and French subset value. At issue is whether L2 learners are able to acquire the subset value of the compounding parameter based on the naturalistic input they receive. In this case, the learning task involves realizing that some L1 constructions are unavailable in the L2. Results indicate that the learners initially transfer the L1 (superset) value and do not start with the subset value of the parameter. Findings also inform the debate on whether negative evidence can engage UG-related acquisition. Ten of the 26 advanced subjects were able to successfully reset the whole parameter based on negative data for only two of the four constructions in the cluster. This fact suggests that it is not impossible for negative evidence to be utilized in grammar reorganization.