Effect of Time and Sleep on the Transitive Inference Task

Abstract

Researchers seeking to understand the adaptive value of sleep have done so, in part, by assessing the cognitive effects of sleep on the performance of specific tasks. In this study we follow up on previous work reporting that the passage of time affects performance on the Transitive Inference (TI) task in subjects trained to below-ceiling levels on the premise pairs, and that sleep affected performance in distinctive ways. We discuss the possibility that these changes in performance are due to a shift in strategy from one in which intermediate premise pairs are coordinated as needed to respond to probes, to a strategy which relies on a gradient of preference amongst the stimuli. We argue that, in contrast to how previous work on the effect of delay and sleep on the TI task has been done, changes in performance should be measured within-subject, by testing the same subjects at multiple time points, and present evidence that this approach is valid with this task (as opposed to showing any learning effects with repeated testing.) We find, on average, no change in performance on this task over the course of 2.5 to 3 hours, with or without sleep. In contrast to previous work, we find in many subjects high levels of performance after only 20 minutes, and a pattern of performance that we expected to find only after sleep. We evaluate the use of an innovative technique to assess the preference gradient which has not previously been used with human subjects, and present evidence that the presence of a preference gradient determines how sleep affects changes in performance on this task.