The hippocampus remains activated over the long term for the retrieval of truly episodic memories

Abstract

The role of the hippocampus in declarative memory consolidation is a matter of intense debate. We investigated the neural substrates of memory retrieval for recent and remote information using functional magnetic resonance imaging (fMRI). 18 young, healthy participants learned a series of pictures. Then, during two fMRI recognition sessions, 3 days and 3 months later, they had to determine whether they recognized or not each picture using the "Remember/Know" procedure. Presentation of the same learned images at both delays allowed us to track the evolution of memories and distinguish consistently episodic memories from those that were initially episodic and then became familiar or semantic over time and were retrieved without any contextual detail. Hippocampal activation decreased over time for initially episodic, later semantic memories, but remained stable for consistently episodic ones, at least in its posterior part. For both types of memories, neocortical activations were observed at both delays, notably in the ventromedial prefrontal and anterior cingulate cortices. These activations may reflect a gradual reorganization of memory traces within neural networks. Our data indicate maintenance and strengthening of hippocampal and cortico-cortical connections in the consolidation and retrieval of episodic memories over time, in line with the Multiple Trace theory (Nadel and Moscovitch, 1997). At variance, memories becoming semantic over time consolidate through strengthening of cortico-cortical connections and progressive disengagement of the hippocampus.

Figure 1. Experimental design.

On day 1, participants were instructed to rate the valence of a series of emotional (positive and negative) and neutral colored pictures and to memorize them. Memory performance was tested on day 4 () and day 90 (±7 days, ) using a recognition task associated with the Remember/Know/Guess paradigm. Functional MRI data were acquired during these recognition sessions. After each session, all pictures that received an R response were presented again outside the scanner during a debriefing interview, to obtain from subjects a justification of their R responses with contextual details. Note that the same learned (“old” pictures) were used for both retrieval sessions allowing to track the fate of memories and their possible qualitative change over time. Thus, we distinguished consistently familiar or semantic memories (KK), consistently episodic memories (RR) and those that were initially episodic and became later semantic (RK). Distracter stimuli were different between the two recognition sessions. The pictures used in the figure are not the original IAPS images, but similar items used for illustrative purposes only.

Figure 2. Brains areas subserving retrieval of consistently episodic (RR responses) or initially episodic and later semantic (RK responses) memories compared to correct rejections (CR), 3 days (a) and 3 months (b) after encoding.

Activations are displayed at p<0.001 (uncorrected) on sagittal or coronal sections of the MNI template. Note that the right hippocampal activation did not survive correction for multiple comparisons.

Figure 3. Common and specific activations for RR and RK memories, 3 months after encoding.

Common activations (top) are displayed at p<0.05 (FWE corrected) and specific activations (bottom) at p<0.001 (uncorrected), on axial, sagittal and coronal slices of the MNI template. The red circle illustrates the specific activation of the hippocampus for RR memories.

Figure 4. Results of ROI analyses in the hippocampus.

A. Hippocampal activity (mean ± SEM) according to the nature of memories (RR vs RK responses) and delay (3 days vs 3 months). *: p<0.05, ***: p<0.001. B. Hippocampal activity (mean ± SEM) for RR responses according to the delay (3 days vs 3 months) distinguishing the anterior and posterior parts. ***: p<0.001.