Effects of training on response latency and accuracy for a difficult visual discrimination

The learning of a difficult discrimination task was investigated in a laboratory simulation of a quality control task. In an earlier experiment, observers were trained to discriminate between 2 brief, masked stimulus patterns. Then, a trained and an untrained group of observers were tested on the same stimulus-discrimination task. Response latency as well as P(Dis) were measured in a forced-choice reaction time paradigm. Untrained observers responded with a mean probability of correct discrimination, P(Dis), = .65 for stimuli with 30-msec. duration and mean P(Dis) greater than or equal to .90 those stimuli with 100-msec. duration. Trained observers produced discrimination scores of P(Dis) greater than or equal to .90 for all stimulus durations tested. Despite this large, reliable training effect of P(Dis), response latencies did not differ reliably between the two groups but instead were a negative function of stimulus duration for both trained and untrained observers.