Modulation of hippocampal memory

In studying learning and memory mechanisms, we make an important distinction between conditions required for learning (i.e., modulation of learning), and the actual stored content of learning. The conditions of pattern separation can be studied from the perspective of survival systems where emotional arousal is involved, and also from the perspective of reinforcement signals where feedback and prediction error are involved. We use both perspectives to examine the conditions that optimize hippocampal pattern separation and how these conditions change as a function of age.

Emotion arousal and hippocampal pattern separation

We examined the role of emotional arousal in modulating hippocampal function (cf. McGaugh, 2004) using an emotional pattern separation task where valence and interference (i.e., similarity) were parametrically varied. We found that emotion enhanced target recognition but impaired lure discrimination, suggesting that emotion’s enhancing effect was specific to gist information and not details. We also demonstrated that emotional pattern separation was associated with fMRI signals arising in the amygdala and dentate (DG)/CA3. We also found that depressed adults were better at discriminating negative lures and worse at discriminating neutral lures, an effect that was associated with a network imbalance where the amygdala’s response was heightened and the DG/CA3 response was suppressed. These results have been presented at the Society for Neuroscience annual meeting and have been submitted for publication. For more information on this topic, contact Stephanie Leal.

Reinforcement and hippocampal pattern separation

The difference between expected and actual outcomes creates a prediction error that motivates learning about the present situation. One mechanism by which this is thought to occur is through the interactions between the ventral tegmental area (VTA) and the hippocampus. Given our enhanced understanding of hippocampal computations, we asked the question: how does prediction error facilitate this process? We modified our standard object pattern separation task by giving subjects a small monetary reward (non-contingent on their responses) on a random subset of trials during the study phase. We found that during a subsequent test, items previously rewarded were more likely to be correctly discriminated from similar lures compared to unrewarded items. Interestingly, this potentiation was not present when pattern separation consistently failed (items too similar) or was not necessary (items too different). Identical results were obtained when subjects suffered a monetary loss instead of a gain, suggesting that this memory facilitation is associated with an unsigned prediction error (Ji and Yassa, CNS Abstr. 2012). We are now extending this work to: (1) parametrically vary the monetary value of reward and punishment to see if this relationship is sensitive to value, (2) parametrically vary the expectancy of the outcome to examine the role of surprise, and (3) to assess VTA-hippocampal functional connectivity using high-resolution fMRI during performance of this task. For more information on this project, contact Jared Roberts and Stephanie Leal.