Memory signals are temporally dissociated in and across human hippocampus and perirhinal cortex

Abstract

In the endeavor to understand how our brains enable our multifaceted memories, much controversy surrounds the contributions of the hippocampus and perirhinal cortex (PrC). We recorded functional magnetic resonance imaging (fMRI) in healthy controls and intracranial electroencephalography (EEG) in patients during a recognition memory task. Although conventional fMRI analysis showed indistinguishable roles of the hippocampus and PrC in familiarity-based item recognition and recollection-based source retrieval, event-related fMRI and EEG time courses revealed a clear temporal dissociation of memory signals in and across these regions. An early source retrieval effect was followed by a late, post-decision item novelty effect in hippocampus, whereas an early item novelty effect was followed by a sustained source retrieval effect in PrC. Although factors such as memory strength were not experimentally controlled, the temporal pattern across regions suggests that a rapid item recognition signal in PrC triggers a source retrieval process in the hippocampus, which in turn recruits PrC representations and/or mechanisms, evidenced here by increased hippocampal-PrC coupling during source recognition.

Figure 1

Methods and design. a. fMRI regions of interest (highlighted in yellow) for hippocampus and perirhinal cortex of an example participant. b. iEEG electrode locations. Left: Medial temporal lobe implantation scheme used for all participants. Middle: Hippocampus and perirhinal cortex contacts of an example participant, shown on the post-implantation MRI scan. Right: Same contacts shown on the co-registered pre-implantation MRI scan. Anatomical images in a and b are normalized for comparability. c. Schematized experimental design (iEEG version). During the Study phase, participants saw concrete nouns together with an associated “source” (a color in half the runs and a scene in the other half of the runs) and indicated whether or not the combination was plausible. During the Test phase, from which the present data were analyzed, studied (old) words were shown along with unstudied (new) words, and participants indicated, with one button press, whether they remembered the word from the study phase and whether they remembered its associated source. Conditions of interest were correct rejection of new items (CR), correct identification of old items, without remembering the associated source (item recognition, IR) and correct identification of old items plus remembering the associated source (source recognition, SR). Note that procedural details differed slightly between fMRI and iEEG versions of the experiment (see Methods) to allow for the different signal characteristics and to match behavioral performance across controls and patients.

Figure 2

fMRI results for hippocampus (a) and perirhinal cortex (b). Left: Results from modeling conditions in a conventional analysis using an assumed HRF. Bar graphs represent mean (+ s.e.m.) parameter estimates. Note that, though differing with respect to baseline, item effects (CR vs. IR) and source effects (SR vs. IR) are indistinguishable within or across regions. Middle: Average (+/− s.e.m.) fMRI BOLD time courses versus baseline for the three conditions of interest. Right: Statistical development of the item effect (differential evoked response for CR vs. IR) and source effect (SR vs. IR), showing t-values for each effect across time. Points above the dashed line correspond to P < .05, two-tailed. Note the temporal dissociation of item- and source effects within and across regions.

Figure 3

iEEG results for hippocampus (a) and perirhinal cortex (b). Left: Stimulus-locked ERPs. Right: Response-locked ERPs. Shaded areas show the two time windows used for statistical analysis.

Figure 4

Temporal sequence of iEEG item- and source effects in perirhinal cortex (bottom) and hippocampus (top). ERPs are identical to Fig. 3 (left). Vertical lines demarcate the onset of the first statistically reliable item- and source effect in each region. Darker portion of vertical lines highlights the relevant condition differences.

Figure 5

Functional coupling between hippocampus and perirhinal cortex increases during source recognition (SR vs. IR). a. MTL regions showing a psychophysiological interaction (PPI) using participants’ individually drawn hippocampi as seed regions (schematized in the sagittal view, right panel). Results are shown at P < .001 (uncorrected) for display purposes. Left peak: x = −27, y = 2, z = −35; right peak: x = 21, y = 2, z = −38. b. Time/frequency clusters of significant iEEG coherence differences between SR vs. IR (P < .05, corrected for multiple comparisons). Color reflects absolute t values for SR vs. IR (only significant t values shown). c. Average (+/− s.e.m.) time course of coherence in the significant clusters. Left: Coherence in the low gamma band (30-35 Hz) is enhanced for SR relative to IR between ~700 to 800 ms. Right: Coherence in the alpha band (8-12 Hz) is reduced for SR relative to IR between ~500 to 900 ms.