Criterion changes in relative discrimination tasks

The thesis begins with accounts of various models of discrimination and of previous studies of sequential effects. It is mainly concerned with the study of discriminations in serial response tasks in which the difficulty of the discriminations varies from trial to trial. All the experiments reported were concerned with the discrimination between two lines of different length which were presented simultaneously in a display. There were always two levels of difficulty of discrimination, which were mixed in random sequence. The first two experiments demonstrated that the RT's to displays were affected by the average difficulty of the preceding displays ("local difficulty bias effect"). In Experiments 3, 4, 5 and 7 it was shown that the overall difficulty of a run of displays also affected the mean RT to all the displays in a run ("overall difficulty bias effect"). It was seen that the direction in which the RT's were affected by local difficulty bias was sometimes different from the direction of the effect of overall difficulty bias, suggesting that subjects may be aware of both local and overall average difficulty levels. Experiments 4, 5 and 6 showed that instructions to be fast modified the effects of bias, but that instructions to be cautious did not increase the effect of bias. Experiment 7 investigated the effect of increasing the actual difference between the easy and difficult discriminations - but was rather inconclusive. In Experiment 8 the RSI was varied and showed no major effect on local difficulty bias, showing that these effects were due to other factors than the time between trials. It also supported an hypothesis considered in earlier experiments that when a criterion is not being used (either during the interval between decisions, or while another criterion is being used) it will drift towards a value that will give rise to more cautious decisions. When displays were easy to discriminate subjects were able to track their speed to the errors they made, but they were not able to do so when the discriminations were more difficult. Whether the displays were easy or difficult, subjects were able to track their speed to the average difficulty of the displays. In general, as the average difficulty of the displays increased ( as the proportion of difficult displays in a mixed run of easy and difficult displays increased ), so the mean RT of the immediately following displays became slower (local effect). It was suggested that subjects' adjustments in performance which were induced by difficulty bias or instructions could be explained by changes in the values of criteria.