The need for continuous instruction of five severely or profoundly impaired (SPI) students (12-18 years old) was examined in light of recent court decisions regarding extended year schooling. SPI Ss and a control group of normal IQ Ss were taught new discriminations. After a 3-month break, the discriminations were taught with procedures, methods, and materials identical to the initial learning phase. Analysis of number of trials necessary to reach criterion for each of the discriminative stimuli revealed that, on the average, SPI Ss took longer to relearn the discriminations than they took to learn them intially. However, there was a wide range of individual variations. Control Ss relearned the discriminations in fewer trials than were required for initial learning. Findings suggested that generalizations about regression among SPI groups are not warranted. For two SPI Ss discrimination relearning was not impeded by the time lapse any more than would be expected for normal IQ Ss. (CL)