A procedure for generating values of central processing uncertainty was developed from positive response data in a varied-set version of the Sternberg choice reaction task. This is a logical extension of a previously validated procedure for data from a fixed-set version of the same task. Both procedures provide information on the additive components of reaction time. It was concluded that S resolves more uncertainty in the varied-set than in the fixed-set situation. It was concluded also that S performs a rechecking operation prior to emitting a negative response, and this rechecking apparently involves less information than does the original testing for stimulus classification. This, in turn, suggests that rechecking is a self-terminating process with regard to display information. The results also imply that stimulus classification is partially serial and partially parallel, so a hybrid model may be appropriate for this task.