While agreeing with Lardiere that the "parameter-resetting" approach to understanding second language acquisition (SLA) needs rethinking, it is suggested that a more construction-based perspective runs the risk of losing deductive and explanatory power. An alternative is to investigate the constraints on feature assembly/re-assembly in second language (L2) grammars. A model of grammatical organization is adopted from Ramchand and Svenonius (2008) in which properties of the conceptual-intentional (C-I) module of mind are universal, and variation between languages is determined by the extent to which such properties are grammaticalized or determined by context. Predictions are then made about the degree of difficulty involved in determining the appropriate mapping from the C-I module to grammar or context when a learner's first language (L1) is similar to or different from the L2. (Contains 5 footnotes, 2 tables, and 1 figure.)