Multimodal imaging reveals the spatiotemporal dynamics of recollection

Abstract

Functional MRI research suggests that different frontal and parietal cortical regions support strategic processes that are engaged at different stages of recollection, from pre-retrieval processing of a cue to post-retrieval maintenance and evaluation of recollected information. Whereas some of these regions respond in a domain-general way, other regions are sensitive to the type of information being recollected. However, the low temporal resolution of fMRI cannot distinguish component processes at the time-scale at which recollection occurs. We therefore combined fMRI with the excellent temporal resolution of source localised EEG/MEG to investigate the spatiotemporal neural dynamics of recollection. fMRI and EEG/MEG data were collected from the same participants in two sessions while they retrieved different types of episodic information. This multimodal imaging approach revealed striking consistency between the regions identified with fMRI and EEG/MEG, providing novel evidence of how these brain areas interact over time to support source recollection. For domain-general recollection, results from both modalities converged in showing the strongest activations in medial parietal cortex, which according to EEG/MEG was reliable at a late retrieval stage. Domain-specific source recollection increased fMRI and EEG/MEG activation in the left lateral prefrontal cortex, which EEG/MEG indicated also to be recruited during a post-recollection stage. The findings suggest that although medial parietal and left lateral prefrontal regions mediate functionally different retrieval processes, they are both engaged at a late stage of episodic retrieval.

Fig. 1

Stimuli examples and procedure overview. During study phases, pictures of famous faces were presented on the left or right of the screen, either slightly nearer or further out from the centre. Participants were first pseudo-randomly cued to make either a pleasant/unpleasant or a British/non-British judgement about each face. Second, they made a near/far perceptual judgement on the picture location. During the test phases, participants were pseudo-randomly cued to make context memory or semantic control decisions about stimuli. In context memory conditions, decisions required recollection of either whether the pleasant/unpleasant or British/non-British task had been undertaken (Task recollection) on whether that stimuli had been shown on the left or right side of the screen (Location recollection). In the Semantic control condition, participants made actor/non-actor occupation decisions about experimentally novel famous faces.

Fig. 2

Grand average ERPs (A), and magnetometer (B) and gradiometer (C) ERFs. ERPs/ERFs are plotted for the three conditions at representative sensors where experimental effects were maximal, and topographic maps below depict ERP/ERF amplitude differences between conditions, averaged over successive 200 ms time-windows between 0 and 2 s post-stimulus (Semantic Control subtracted from Location Recollection [L − S], Semantic Control subtracted from Task Recollection [T − S] and Location Recollection subtracted from Task Recollection [T − L]). The gradiometer maps were constructed by computing the root mean square (RMS) of gradiometer pairs for each condition and subtracting these condition RMS maps. All sensor-types showed domain-general late slow-drifts over parietal sensors and more frontally distributed domain-specific late slow-drifts.

Fig. 3

Topographic plots of SPM F-statistics for scalp-level ERPs (A & B) and Magnetometer ERFs (C & D). Maps have been thresholded at P < 0.001, corrected for cluster extent at P < 0.05, and averaged over successive 200 ms time-windows between 0 and 2 s post-stimulus. For domain-general recollection (A & C, testing both types of recollection versus semantic control using inclusive masking), both modalities showed a highly significant parietal effect that was maximal between ~ 1000 and 1400 ms post-stimulus for ERPs (A) and between ~ 600 and 1600 ms post-stimulus for ERFs (C). For domain-specific recollection (B & D, testing the simple difference between Task and Location recollection) both modalities showed a highly significant left temporal/frontal effect that was maximal between ~ 800 and 1200 ms post-stimulus for ERPs (B) and between ~ 800 and 1800 ms for ERFs (D).

Fig. 4

Fused EEG/MEG (MEEG) source localisation results indicating the underlying neural generators of domain-general MEEG recollection effects. Effects are thresholded at P < 0.001 (uncorrected), with a minimum cluster size of 10 voxels. A, an inclusive masking analysis showing regions where both types of episodic recollection produced significantly enhanced MEEG power compared to the Semantic control condition, analysed in successive 200 ms time-windows and overlaid on bilateral posterior (left column) and left medial (right column) inflated canonical cortical surfaces. Domain-general effects were significant at this threshold in a medial and superior PP region between ~ 600 and 1600 ms post-stimulus. B, a direct comparison between the domain-general MEEG localisation in the 800–1000 time-window and the corresponding domain-general fMRI activations overlaid on bilateral posterior (left), left medial (middle) and right medial (right) inflated canonical cortical surfaces, showing substantial spatial overlap in the medial PP cortex. C, estimated MEEG source activity from a 10-voxel sphere centred in the Precuneus region (− 6 − 73 46) that overlapped between MEEG and fMRI modalities (arb. unit).

Fig. 5

Fused EEG/MEG (MEEG) source localisation results indicating the underlying cortical generators of domain-specific MEEG recollection effects. Effects are thresholded at P < 0.001 (uncorrected), with a minimum cluster size of 10 voxels. A, a simple contrast showing regions where Task recollection produced significantly enhanced MEEG power compared to Location recollection, analysed in successive 200 ms time-windows and overlaid on a left lateral inflated canonical cortical surface. Domain-general effects were maximally significant in left LPFC between ~ 1000 and 1600 ms post-stimulus. B, a direct comparison between the domain-specific MEEG localisation in the 1200–1400 ms time-window and the corresponding domain-specific fMRI activations, showing substantial spatial overlap in the left LPFC cortex. C, estimated MEEG source activity from a 10-voxel sphere centred in the LPFC region (- 42 11 49) that overlapped between MEEG and fMRI modalities (arb. unit).

Fig. 6

An inclusively masked analysis showing regions that were significantly active in both fMRI and MEEG modalities during domain-general (left) and domain-specific (right) recollection with a joint probability of P < 1.5 × 10^–5 or less. This analysis confirmed that the Precuneus and left dorsal PFC showed highly significant convergence between fMRI and MEEG for both functional contrasts.