Research on stimulus-stimulus pairing to induce novel vocalizations in nonverbal children has typically employed response-independent pairing (RIP) procedures to condition speech sounds as reinforcers. The purpose of the present study was to evaluate the effects of a response-contingent pairing (RCP) procedure on the vocalizations of three nonverbal boys diagnosed with autism spectrum disorder. During RCP, adult-delivered sounds that were either paired with a preferred item (target sounds) or not (nontarget sounds) were presented contingent on a button-press response. In Experiment 1, RCP was compared with an RIP procedure, in which the timing of sound presentations was yoked to the preceding RCP session. RCP produced a greater effect on all participants' target vocalizations than RIP. Experiment 2 demonstrated the effects of differential reinforcement of the vocalizations induced in Experiment 1. The results suggest that RCP may develop vocalizations more reliably than RIP procedures.