PURPOSE: To evaluate the learning effects of a deductive language-teaching procedure when teaching a novel gender agreement verb inflection to children with language impairment. METHOD: Thirty-two 6-8-year-old children with language impairment were randomly assigned to either a deductive (N = 16) or an inductive (N = 16) treatment group. In the deductive treatment, the examiner presented a rule guiding the novel inflection to be learned as well as models of the inflection. In the inductive treatment, only models of the verb inflection were presented. Learning was assessed in 3 different production contexts during each of 4 treatment sessions. RESULTS: Significantly more participants in the deductive group than the inductive group acquired the novel morpheme based on a teaching probe (10 vs. 3), generalization probe (10 vs. 3), and maintenance probe (7 vs. 2). Task performance was not significantly influenced by language ability or nonverbal intelligence. CONCLUSIONS: The deductive teaching procedure was found to be efficacious when teaching a novel grammatical inflection. However, this effect was limited because treatment gains varied across participants, testing contexts, and sessions. Future studies should continue to examine the efficacy of deductive procedures when integrated into traditional implicit approaches for children with language impairment.