This paper compares explicit instruction in second-language Spanish with a control treatment on a written picture description task and a timed auditory grammaticality judgment task. Participants came from two intact, third-year US high school classes, with one experiencing a week of communicative lessons on the Spanish clitic "se" ("n" = 15) and the other exposed to "se" only incidentally ("n" = 20). Explicit instruction consisted of grammar rules with sentence-level examples, followed by communicative tasks. Three test versions were administered within a split-bloc design as a pretest, immediate posttest, and delayed posttest 6 weeks after instruction. The instructed group increased targetlike uses of "se" on both tasks and sustained gains through the delayed posttest, although first-language transfer errors persisted. Meanwhile, overgeneralization errors centered on semantic and syntactic contexts similar to the instructional object, aligning with the unergative-unaccusative distinction among intransitive verbs. It is argued that the data provide evidence for the permeability of second-language implicit knowledge to explicit instruction and against total encapsulation as a model of the mind.