Four retarded adults were exposed to a multiple variable interval-extinction schedule of reinforcement. The 2 min components were presented in a random order. Two subjects acquired the discrimination with an errorless procedure. The two other subjects learned the discrimination with errors. The schedule generated sequential contrast effects under both training procedures: response rates during periods of reinforcement were higher when a reinforcement period followed an extinction period than when it followed another reinforcement period. These results were confirmed in a second observation where eight retarded children were exposed to a multiple fixed ratio-extinction schedule of reinforcement. This study is the first demonstration of sequential contrast effects with human subjects during the acquisition of a discrimination without errors. The results are discussed in terms of Terrace's theory of errorless learning.