Results of several analyses of error data from four different discrete-choice reaction experimental conditions (an equiprobable two-choice condition, a two-choice condition in which one of the stimulus events appeared with a probability of .70, an equiprobable four-choice condition, and a four-choice condition in which one signal appeared on 20% of the trials) indicate salient characteristics of error data which an adequate model should predict, as well as some clues as to the information processing microstructure which determine error responses. In general, the second and third-order sequential effects observed for the error probability data closely parallel the often reported sequential effects associated with choice reaction time (CRT) data. However, findings concerning CRTs associated with error and posterror responses are reported which raise a number of questions regarding adequacy of current theoretical formulations which assume that errors are generated by premature exit from an invariant sequence of additive and stochastically independent processing stages. In addition, several methodological implications of results are delineated and discussed.