Choice reaction performance following an error

Some data are presented showing the changes in choice-reaction time (CRT) and the probability of error (PE) on the five trials following an error in a two-choice experiment. The magnitudes of the increase in RT and the decrease in PE are different according as the error-trial stimulus is re-presented or the alternative stimulus is preservec. In addition, RT is quicker to recover its equilibrium value than is PE.This pattern of changes implicates two distinct adjustments to the underlying decision process following an error: (i) a selective outward adjustment of the boundary at which the error was uttered, and (ii) a delay, with respect to presentation of the reaction stimulus, of the epoch at which the subject begins sampling information from the stimulus display. When these ideas are concatenated with the sequential probability-ratio-test model for CRT, a theoretical relation can be derived between the difference in mean RT for an error and for the same response given correctly, on the one hand, and the decrease in PE on the trial following an error, given that the alternate stimulus is presented, on the other. This relation is satisfied by the data.