Bengt Grua and Dr. C. Brock Kirwan, Departments of Psychology and Neuroscience
The goal of this study was to determine how sleep affects memory recall in an academic setting with relation to memory discrimination and memory generalization. Memory generalization depends on the computational process of pattern completion, which is the brain’s ability to retrieve a memory representation based on a partial or modified stimulus (such as viewing a false statement as true if it has many true elements embedded in it). Memory discrimination depends on the computational process of pattern separation, which is the brain’s ability to store a memory as distinct and separate from other, similar memories. Current models of sleep and its effects on memory posit that during sleep, memory representations are shifted from the brain’s hippocampus into the cortex. Furthermore, memories stored in the cortex appear to be more generalized and stable, but less specific than those stored in the hippocampus. Based on this previous research, it was our hypothesis that when recalling information learned hours prior, sleep will make a person tend towards generalization rather than discrimination.
Using a database of several hundred pairs of similar images, we first used a group of 35 students to rate all of the images on a Likert scale of similarity where a 1 represented images almost identical, and a 5 represented very distinct images (See Figure 1).
The experimental procedure was composed of two parts. In the first part, participants would study the first image in the pair for one hour. After a twelve-hour delay, the participants returned and were asked to view a series of lure images (some of which were identical, some which were similar pairs) and asked whether the image was the same or merely similar to the image they viewed earlier. We used BYU’s Sona System to recruit 50 participants and divided them evenly into two groups. Each group followed the same experimental procedure, the difference being whether they slept between. In other words, 25 participants studied the images at 9am and returned to be tested on the images at 9pm with no sleep in between. The other half began at 9pm and then slept a full night before being tested at 9am. Finally, we compared the two groups’ ability to correctly distinguish between old or similar images across all 5 levels of similarity.
We looked at the proportion of correctly identified different (or similar) responses across the 5 levels of similarity. We expected that as the objects became more distinct (moving up on the Likert scale) there would also be a higher proportion of correct “similar” responses. The curvilinearity of the trend line however indicates the neural processes at work. If there were no additional mental processing, we would expect a straight line of increases discrimination as the images become less similar. If the curve were concave, that would suggest additional discrimination computation, while a convex curve would suggest memory generalization.
The critical comparison was the proportion of “different” or “similar” responses between those in the wake and sleep conditions (see Figure 2). Those in the sleep condition had an increased ability to correctly distinguish between old and similar images as compared to those in the wake condition. Furthermore, contrary to our predictions, the wake condition shows a convex trend indicative of memory generalization, while the sleep conditions shows a much more linear relationship.
Discussion & Conclusion
Contrary to the predictions of current research and computational models, our findings suggest that there is a greater bias toward memory generalization following a waking delay than one involving sleep. A possible explanation is that sleep has been shown to prevent memory loss due to interference from the formation of new memories. Both conditions lost some of the image detail (indicated by neither group achieving a perfect 100% “Different” response) but more detail was clearly lost by those who stayed awake. This result suggests that sleep has a protective quality for the formation of short-term memory, in contrast to the hypothesis that sleep would decrease memory specificity and increase generalization. Further study is clearly needed to reconcile these opposing predictions. Perhaps a study increasing the delay between the study and test phases would show whether the protective effect of sleep on memory is merely a short or long-term phenomenon.