This study investigated the relationship between immediate learner repair in response to interactional feedback and learning targeted forms. In particular, the study examined and compared the relationship between two major types of repair: learner-generated self-repair following elicitations and teacher-generated repair following recasts. It also examined the relationship of repair that involved repetition of the recast versus those that involved incorporation of the recast into new and expanded production. Participants were 42 adult ESL learners interacting with two native speaker teachers. The amount of each type of repair during interaction was examined and was then related to the degree of correction of the same error in individualized posttests immediately after interaction, and then after two weeks. The results showed that overall learners were able to recognize and correct after interaction over half of the errors that they had repaired during interaction. The results also showed that teacher-generated repair and self-repair led to similar degrees of immediate post-interaction correction. However, whereas the effects of self-repair were maintained, the effects of teacher-generated repair were reduced over time, particularly when the repair involved repetition rather than incorporation. The findings suggest that different types of repair may have different impacts on language learning.