In Study 1, 10-, 13-, and 16-year-olds were assigned to conditions in which they were instructed to think logically and provided alternative antecedents to the consequents of conditional statements. Providing alternatives improved reasoning on two uncertain logical forms, but decreased logical responding on two certain forms; logic instructions improved reasoning among adolescents. Correlations among inferences and verbal ability were found primarily when task conditions created conflict between automatic and controlled inferences. In Study 2, when the cognitive demands of the logic instructions were reduced, 10-year-olds made more logically correct inferences, but only when a conditional's consequents were strongly associated with alternative antecedents. Discussion focuses on the ability to inhibit invited inferences and the role of automatically activated memories.