Athena Howell and Brock Kirwan, Department of Neuroscience/Psychology

Introduction

My project focused on the neural activity associated with memory consolidation, particularly the role of pattern separation and pattern completion over an extended period of time. Pattern separation occurs when the brain successfully encodes similar experiences into distinct memories (Deng et al. 2003); for instance, pattern separation allows you to correctly remember what you had for dinner last night as opposed to the night before last. Pattern completion, on the other hand, occurs when you reconstruct a memory from a partial memory (e.g. when asked what you ate last Wednesday night, you respond with the meal you consumed last Tuesday night) (Leutgeb & Leutgeb, 2007). The more time passes, the harder it is to access distinct, separate memories without interference. Computational models of the brain propose that the hippocampus (a brain structure critical for memory formation) establishes distinct memory representations through pattern separation (Norman & O’Reilly, 2003). Over time, however, these memory representations are shifted to other cortical regions and become more general and less distinct (Alvarez & Squire, 1994).

Previous human fMRI studies have studied pattern separation processes over very short time-scales, usually minutes to hours. These studies are useful because they allow us to recognize memory encoding as it is happening. However, though these studies show us what is happening, they fail to show us when it happens. The brain may wait a period of days or even weeks before having to recall an encoded memory. The behavioral observation of increased memory generalization (or, conversely, a loss of memory specificity) over time may be due to the shift from hippocampal-dependent memory representations to cortical-dependent representations that occurs during consolidation.

Methods

Participants for the study consisted of 15 young adults (7 female), with an age range from 18-30 years (mean 23.3). Before being scanned the first day, each participant was shown a list of pictures and asked to identify if the objects belonged indoors or outdoors; this setup facilitated implicit encoding of the stimuli, without particular attention to “studying” them. Afterwards, the participant were placed in the MRI scanner and shown another list of stimuli, consisting of repeat images, completely new images, and images similar in character to the encoding list. They were then asked to identify again whether the images belonged inside or outside. Once finished, participants returned at the same time for the next 4 days to be scanned, using the same procedure as above, but did not see the original list again. Upon completion of the study, participants were compensated for their time.

Results

Initially, we did a whole brain analysis, but none of our results were statistically significant; nothing survived the correction for whole brain analysis. We subsequently conducted an exploratory analysis of the vmPFC (known to be involved in consolidation). We conducted a linear regression with the average beta value as the dependent variable and day and stimulus type as independent variables. The test for the hippocampus did not return significant F(3, 296) = .06, p = .98, and the overall model for the vmPFC was not significant, F(3, 296) = 2.051, p = 0.1068, but there was a significant interaction between day and stimulus type (β=.10, p <.05).

Discussion

Activity for the hippocampus did not return significant, which is consistent with our hypothesis. The hippocampus is associated with pattern separation, and over time activity should switch from pattern separation to pattern completion in the cortex, leading to little activity in the hippocampus. Activity in the vmPFC increased over the five days, but not to a significant level. This did not support our hypothesis, as we predicted that the vmPFC, an area associated with pattern completion, would have a significant increase in activity. There was a significant interaction between the day and the stimulus type shown, which warrants further investigation.

Conclusion

For the most part, our results were inconclusive. Within the vmPFC, the change in activity looked promising but was ultimately not large enough to be significant. We believe with a larger subject pool and a longer testing period (e.g. scanning participants once a week for several weeks) we would see a significant increase in the vmPFC, reflecting memory consolidation. Further research is needed.