Direct Stimulation of Human Hippocampus During Verbal Associative Encoding Enhances Subsequent Memory Recollection

Abstract

Previous studies have reported conflicting results regarding the effect of direct electrical stimulation of the human hippocampus on memory performance. A major function of the hippocampus is to form associations between individual elements of experience. However, the effect of direct hippocampal stimulation on associative memory remains largely inconclusive, with most evidence coming from studies employing non-invasive stimulation. Here, we therefore tested the hypothesis that direct electrical stimulation of the hippocampus specifically enhances hippocampal-dependent associative memory. To test this hypothesis, we recruited surgical patients with implanted subdural electrodes to perform a word pair memory task during which the hippocampus was stimulated. Our results indicate that stimulation of the hippocampus during encoding helped to build strong associative memories and enhanced recollection in subsequent trials. Moreover, stimulation significantly increased theta power in the lateral middle temporal cortex during successful memory encoding. Overall, our findings indicate that hippocampal stimulation positively impacts performance during a word pair memory task, suggesting that successful memory encoding involves the temporal cortex, which may act together with the hippocampus.

Keywords: brain stimulation; hippocampus; lateral temporal cortex; memory enhancement; recollection; theta power.

Figure 1

(A) Location of stimulation. The green crosshair denotes the location of the stimulation electrode in the right mid-hippocampus (sagittal and coronal, respectively) in Patient 1 (left two panels, anode and cathode, respectively) and in the left mid-hippocampus in Patient 2 (right two panels). (B) The hippocampal subregions in each patient and one hippocampal electrode in the region of interest.

Figure 2

Timeline of the memory paradigm. The whole task consisted of three study periods: encoding, rest (distractor), and retrieval. Stimulation was delivered on-and-off at 50 Hz for 5 s during the encoding phase only and was randomly assigned to one of the two blocks in a single session. Lightning bolts denote periods of stimulation. In the encoding phase, the first word pair on the screen denotes “glass” (left) and “mirror” (right).

Figure 3

Memory performance. (A) Lines connect each patient’s “on” and “off” data point (left); data from all six patients were averaged (right). Accuracy refers to the proportion of correctly recognized words during the stimulation “on” and stimulation “off” periods. (B) Lines connect each patient’s “on” and “off” data for the recollection index (left); data were averaged across all six patients (right). (C) Familiarity index (D). Difference scores of the stimulation effect of the two conditions. FAM indicates the familiarity index and REC indicates the recollection index. *p < 0.05. Error bars indicate the standard error of the mean (SEM). n.s., not significant.

Figure 4

(A) Individual differences in theta power in the middle temporal cortex over 4 s for correctly remembered words during the “on” and “off” periods. (B) Individual differences in theta power in the middle temporal cortex over 4 s for incorrectly remembered words during the “on” and “off” periods. ***p < 0.01, **p < 0.05, corrected. Error bars indicate the standard error of the mean (SEM).

Figure 5

(A) Location of the target in the lateral temporal cortex in the sagittal plane following co-registration of preoperative high-resolution MRI and postoperative CT images (not illustrated), for patients 1–5, respectively. Spheres indicate the location of the recording site. Topographical maps of differences in theta power between the correct and incorrect responses during both the “on” and “off” condition. The yellow sphere denotes locations in which significant increases in theta power were observed in the “on” condition that were higher than those observed in the “off” condition. The yellow sphere indicates the region for the time-frequency map in panel B. Some electrode grids/strips were excluded because they were not visible. (B) Time-frequency map comparing correctly and incorrectly remembered pairs during the “on” and “off” period for five patients. On the x-axis, 0 s indicates the onset of the memory task. During the “on” condition, theta (3–7 Hz) power was significantly increased during the 4 s of word presentation (p < 0.05, corrected) for successfully recognized trials. In contrast, no significant increases in theta power were observed during the “off” condition (p > 0.05, corrected).