Interference Induces Improvement across Wake but not over Sleep in the Weather Prediction Task

Abstract

The ability to make predictions based on information extracted from patterns and frequencies of past events is a critical aspect of learning. This study investigates the effect of a nap on performance in a task that has been used as a measure of learning and memory of this type--the Weather Prediction Task (WPT). Here, two similar versions of the task are used to investigate the effects of interference after sleep and across wake. In the WPT, subjects learn to associate presented card combinations with fictitious weather conditions of sun or rain, where each card has a specific probability of occurring with each weather condition. The essential probabilistic nature of the task is not revealed to the subjects. Yet, even when explicit knowledge of the problem is limited, subjects do improve on the task with training. Earlier work with the WPT has shown that an implicit memory route is activated as well during learning, so that effective performance on the task (and in probabilistic category learning in general) involves both implicit and explicit memory. One earlier study demonstrated improvement in performance based on original score on the WPT with sleep, suggesting that at least some memory consolidation occurs over a night of sleep. Here, improvement on the WPT was measured after a 90-minute nap with and without interference. Interference training is thought to disrupt incompletely stabilized memories of a task through the application of additional training sufficiently similar as to use the same rules of memory consolidation as the original training, but which is also sufficiently different so as to be an effective competitor with the original learning scheme. This was expected to impair retest scores on the original scheme for all subjects. But since sleep has been show to support off-line improvement, an only somewhat reduced effect of interference was expected in the subjects who slept compared to those who did not. In this study, interference training with an alternate card set is administered at 2.5 hours after the first training session. For the sleep group subjects, this is immediately after a 90-minute nap. (Control subjects are awake for this period, and both groups are divided into interference and non-interference groups.) Based on data in earlier studies showing complete memory stabilization after six hours, interference is applied during that window in order to maximize the chance of seeing an interference effect. As expected, subjects showed a trend toward improvement in performance after sleep without interference. With interference, however, a significant improvement was induced across wake as compared with sleep. This is contrary to expectations and provokes questions about the effects of interference. Since wake group subjects in particular showed improved performance on the Weather Prediction Task following interference training, it is possible that the effects of interference are not modulated by memory stabilization state alone. Earlier work shows that sleep alters the quality of memory in the WPT. I have shown here that in a similar probabilistic task, sleep does not offer protection from the destructive effects of interference. Further, an unexpected benefit is seen with interference across wake in the WPT in observation mode that earlier work suggests may derive from a perturbation of implicit and explicit memory components of the task, though this remains a question, as does the mechanism for the differential effects of interference seen between sleep and wake.