Effects of serial and concurrent task presentation on skill acquisition, generalization, and maintenance were compared. Two severely retarded females participated. During serial training, items of one response class, tracing, were trained to mastery before those of a second task, vocal imitation. In the concurrent method, training on two different tasks, tracing and vocal imitation, alternated within sessions for fixed periods of time. There were no major differences between the serial and concurrent methods of instruction in the number of steps attained per behavior or in the number of trials required to reach criterion levels of performance. It was found however, that concurrent training resulted in more generalization. Retention results were not consistently related to training method.