A content-based representational scaffold for naturalistic event memories

Abstract

Real-world events are complex, featuring elements that may be unique to, or shared across, multiple situations. In the present study, we used fMRI to identify how different event components are represented in real-time and during memory retrieval. Twenty participants viewed and recalled eight videos depicting real-world events, combining people, contexts, and context types. Multi-voxel pattern similarity analyses revealed specific person representations, persistent across contexts, in regions of an Anterior-Temporal Network. Conversely, we found specific context representations, persistent across people, in regions of a Posterior-Medial Network. We also found schema-like generalization across contexts in medial prefrontal cortex, and episodic specificity in the hippocampus. Event patterns were reinstated during recall, and hippocampal reinstatement predicted the number of details retrieved. Finally, we observed distinct representational timescales across the hippocampus and cortical regions. These findings reveal mechanisms for scaffolding different aspects of lifelike event representations in cortico-hippocampal networks as experiences are observed and recalled.

Schema representations in distinct brain networks support narrative memory during encoding and retrieval

“…Given the strong involvement of mPFC during schematic encoding, it has been hypothesized that mPFC may play a role at retrieval by providing schematic cues for memory search (van Kesteren et al, 2012). While some studies have found that schema-related activity in mPFC during retrieval benefits memory (Brod et al, 2015;van Kesteren et al, 2010b;Müller et al, 2020;Webb and Dennis, 2019;Raykov et al, 2021), others have not (van Buuren et al, 2014;Webb et al, 2016;Guo and Yang, 2020;van der Linden et al, 2017;Reagh et al, 2021). In our study, we did not observe strong mPFC schema reinstatement, nor were able to relate it to a behavioral memory benefit.…”

“…Because mPFC has been frequently implicated in previous schema research (e.g., van Kesteren et al, 2010a(e.g., van Kesteren et al, , 2013(e.g., van Kesteren et al, , 2014(e.g., van Kesteren et al, , 2020Baldassano et al, 2018;Raykov et al, 2020Raykov et al, , 2021Reagh et al, 2021) -in particular with regard to integrating new knowledge into existing schemas (Preston and Eichenbaum, 2013;Gilboa and Marlatte, 2017;Tse et al, 2007;Wang and Morris, 2010;van Kesteren et al, 2012) -we predicted that robust mPFC schema representations at encoding would lead to improved subsequent memory for the narrative. Based on prior work implicating the hippocampus in schema representation (van Kesteren et al, 2013(van Kesteren et al, , 2014(van Kesteren et al, , 2020Brod et al, 2015;Liu et al, 2017;Raykov et al, 2020;Webb et al, 2016;van der Linden et al, 2017;Bonasia et al, 2018), we also hypothesized that hippocampal schema representations at encoding would support subsequent memory; more specifically, based on work showing that hippocampus has a coarse-to-fine gradient of representations along its long axis (Collin et al, 2015;Guo and Yang, 2020;Audrain and McAndrews, 2020;Poppenk et al, 2013;Brunec et al, 2018;Sekeres et al, 2018), we predicted that anterior hippocampus (which has coarser and thus more general representations than posterior hippocampus) would contain schematic representations that contribute to subsequent memory, whereas posterior hippocampus would contribute to subsequent memory by representing story-specific details.…”

“…The schema literature in fMRI has been mostly split between studies that investigate the role of schemas at encoding and those that investigate their role in retrieval (but see Bonasia et al, 2018;Sommer, 2016;van der Linden et al, 2017;Raykov et al, 2021;Reagh et al, 2021). Of the studies focusing on the encoding phase, schemas have been operationalized by contrasting conditions in which participants have relevant prior knowledge vs. when they do not have this knowledge (Maguire et al, 1999;van Kesteren et al, 2010avan Kesteren et al, , 2014Raykov et al, 2018Raykov et al, , 2020Keidel et al, 2018;Liu et al, 2017Liu et al, , 2018Sommer, 2016) or by using stimuli that are consistent vs. inconsistent (Bonasia et al, 2018;Raykov et al, 2021;Reagh et al, 2021). In contrast to this past work, our design employed naturalistic, temporally-extended schema-consistent stimuli that were then paired with realistic unpaced verbal recall.…”

“…The schema literature in fMRI has been mostly split between studies that investigate the role of schemas at encoding and those that investigate their role in retrieval ( (Bonasia et al, 2018;Raykov et al, 2021;Reagh et al, 2021). In contrast to this past work, our design employed naturalistic, temporally-extended schema-consistent stimuli that were then paired with realistic unpaced verbal recall.…”

The hippocampus constructs narrative memories across distant events

“…It is also possible to generate testable hypotheses about many other brain areas that might be involved in this task. For completeness, we have included post hoc analyses (see STAR Methods) that test for narrative coherence effects, within two key cortical networks Article that interact with the hippocampus [52][53][54][55] : the posterior medial (PM) network, which supports information about event contexts (e.g., types of situations); and the anterior temporal (AT) network, which supports information about particular entities within events (e.g., people or objects). Because stories were presented aurally, we also investigated auditory cortex.…”