Fifteen children (mean age 10 years) who were defined as learning disabled (LD) by school authorities and were enrolled in special LD classrooms were matched with 15 children from regular school classrooms on the basis of sex, chronological age, and mental age. Each child performed individually in a two-hour reaction time (RT) session, during which four different RT tasks were administered. All children received the following tasks, ordered as listed: Simple RT (100 trials), Selective RT (100 trials), Choice RT (100 trials), and Search RT (100 trials with one digit in the memory set, 100 trials with memory-set size of 3, and 100 trials of memory-set size of 5). For Simple RT, subjects pulled a reaction trigger with their dominant hand following the symbol []; for Selective RT, the symbol [ or ] was presented and subjects pulled the dominant-hand trigger to [; and for Choice RT, subjects pulled the dominant-hand trigger to [_] and the nondominant-hand trigger to I-I. Prior to each block of 100 trials during the Search RT task certain digits were defined as critical memory-set items (i.e., 1 for Set-Size 1; 0, 2 and 5 for Set-Size 3; and 1, 3, 6, 8 and 9 for Set-Size 5); and the child was requested to pull the dominant-hand trigger if the stimulus presentation was included in the memory set and to pull the nondominant-hand trigger for stimuli not included in the memory set. For each task the stimulus presentation was preceded by a "ready" buzzer and then a random interval ranging from .5 to 1.5 sec. The RTs were equivalent between groups for the Simple RT task, but were significantly slower for the LD children during the Selective, Choice, and Search RT tasks. The between-group, RT difference was maximum for the Selective RT task, and therefore an application of Donders' Subtractive Model reflected that stimulus encoding was the only cognitive dysfunction among the LD children. Furthermore, the functions relating Search RT to Memory-Set Size demonstrated equivalent memory search rates for the normal and LD children. Such findings suggest that RT may be used to diagnose particular cognitive deficiencies or to demonstrate particular facilitative effects of cognitive-remediation programs. The finding that consecutive blocks of 25 trials did not interact with the latencies in any RT task, suggests that individual sessions to obtain diagnostic RT can be as short as 15 minutes (i.e., 25 trials per RT task). The present RT results are discussed with reference to other RT studies that compared reaction latencies between normal and abnormal subjects. / Ed. D.