This study was carried out to test the main and interaction effects of training condition and pretreatment-elicited verbal imitation ability when predicting spoken language use during language training of 60 minimally verbal autistic children. Subjects were randomly assigned to Speech Alone, Sign Alone, Simultaneous Presentation of Sign and Speech, and Alternating Presentation of Sign and Speech training conditions. Speech Alone, Simultaneous Presentation, and Alternating Presentation condition facilitated more child-initiated speech during treatment than did the Sign Alone condition. Regardless of training condition, pretreatment verbal imitation ability positively predicted the size of child-initiated spoken vocabulary observed during training. Exploratory analyses indicated that, in addition to verbal imitation, pretreatment age and IQ may also predict spoken language developed during training.