The purpose of this study was to determine (1) whether comprehension, production, and spontaneous use of language are greater following language training by sign-alone, speech-alone, simultaneous communication, or alternating between speech and sign; (2) whether high- and low-verbal imitators benefit from the same mode of treatment; and (3) whether retention of the language skills differs among the four training conditions. Subjects were 60 moderate to severe autistic children randomly assigned to one of four treatment conditions. Subjects were divided into high- and low-verbal imitators based on verbal imitation performances. The results indicated that the high-verbal imitators did equally well in all four treatment conditions, while the low-verbal imitators did poorest in the speech-alone condition. The high-verbal imitators performed better than did the low-verbal imitators in all of the treatment conditions. Words or signs learned were retained for three months after treatment regardless of the treatment condition or level of imitative ability. The results were discussed in terms of the efficacy of using sign language with autistic children.