Spatial memory precision and learning in medial temporal and prefrontal networks
The medial temporal lobe and prefrontal cortex are known to play a critical role in tasks involving learned associations, such as between an object and a location. However, the exact neural mechanisms underlying the learning and retrieval of high-fidelity spatial associations are unknown. To address this gap, we tested presurgical epilepsy patients with bilateral depth electrodes implanted in the medial temporal lobe and prefrontal cortex on two variants of an object-location spatial learning task. In the first version of this task, subjects were shown a series of objects at random positions along the circumference of an invisible circle. At test, the same objects were shown at the top of the circle, and subjects used a dial to move the object to the location shown during encoding. Angular error between the correct location and the indicated location was recorded as a continuous measure of performance. By registering pre- and post-implantation MRI scans, we were able to localize electrodes to specific hippocampal subfields. We found a correlation between increased high-frequency gamma power (40-100 Hz), thought to reflect local excitatory activity, and the precision of spatial memory retrieval in hippocampal CA1 and dorsolateral prefrontal electrodes. Additionally, we found a directional relationship between these regions, suggesting that the dorsolateral prefrontal cortex is involved in post-retrieval processing. In order to examine the neural activity underlying spatial learning, we then tested presurgical epilepsy patients on a variant of this task in which subjects attempted to learn object-location associations over the course of three training blocks with feedback. At retrieval, we found increased medial temporal and dorsolateral prefrontal gamma power for low error trials, consistent with the results described above. At feedback, we found the opposite pattern of activity, with increased medial temporal and dorsolateral prefrontal gamma power for high error trials. Increased medial temporal gamma activity at feedback also predicted greater decreases in error from one training block the next, indicating that these error signals are involved in updating memory representations or modifying incorrect associations during learning. Finally, we examined the contributions of low frequency oscillatory power to performance on this spatial learning task as well as the relationship between the 1/f aperiodic slope and spatial memory retrieval and learning. We found that the aperiodic slope, thought to reflect the ratio of excitation to inhibition, decreased across training blocks, but did not predict error within training blocks, suggesting that decreased excitation and/or increased inhibition is associated with increased familiarity and/or decreased novelty. Low frequency oscillatory power did not predict error within blocks and did not change across blocks. Overall, these data suggest putative mechanisms for the learning and retrieval of high-fidelity spatial associations.