Cerebral reflections of memory modification following episodic prediction errors

1073 

Abstract Submission 

Sophie Siestrup^1,2, Ricarda Schubotz^1,2, Benjamin Jainta¹, Sen Cheng³

¹University of Münster, Münster, Germany, ²Otto Creutzfeldt Center for Cognitive and Behavioral Neuroscience, Münster, Germany, ³Ruhr University Bochum, Bochum, Germany

Sophie Siestrup  
University of Münster|Otto Creutzfeldt Center for Cognitive and Behavioral Neuroscience
Münster, Germany|Münster, Germany

Ricarda Schubotz  
University of Münster|Otto Creutzfeldt Center for Cognitive and Behavioral Neuroscience
Münster, Germany|Münster, Germany

Benjamin Jainta  
University of Münster
Münster, Germany

Sen Cheng  
Ruhr University Bochum
Bochum, Germany

Episodic memories do not always accurately reflect our experiences, but can be modified in favor of new information. Such changes are potentially fueled by mnemonic prediction errors which arise when there is a mismatch between what was expected based on memories and what is experienced in a new situation [1,2]. While it is believed that such memory changes allow us to maintain valid predictions in a dynamic environment [2,3], little is known about how the brain adapts existing memories in response to such prediction errors. Therefore, the aim of the present study was to identify brain regions which are involved in the establishment of updated episodic representation.

Thirty-six healthy female participants (age: 22 ± 2.78 years), all right-handed, took part in the study. First, participants encoded 24 different episodes by playing short toy stories in the laboratory. During two further sessions, one day and one week after the encoding, participants went through further video-based retrieval sessions to aid memory consolidation. Another week later, participants returned for a functional Magnetic Resonance Imaging (fMRI) session, during which episodic retrieval was cued by videos showing the original episodes or slightly modified versions thereof to induce prediction errors. Modified videos deviated from originally encoded ones by a single aspect which was either the order of action steps or an individual object. Imaging was conducted on a 3T Siemens Prisma MR tomograph (TR/TE= 2000/30 ms, FA= 90°, FOV= 192 x 192 mm², 33 slices, slice thickness= 3 mm). Lastly, participants completed a post-fMRI memory test during which each video was presented in the original and one modified version in pseudo-randomized order. Participants were asked to indicate on a four-point Likert scale (with 1= yes to 4= no) whether or not each video showed an episode that had been originally encoded. FMRI data were preprocessed (slice time correction, realignment, co-registration of functional to structural scans, normalization, and smoothing) and analyzed with SPM12 using a general linear model. Behavioral data were analyzed using R Studio.

In the post-fMRI memory test, participants were significantly more prone to falsely accept modified episodes as truly encoded when they had experienced prediction errors during the fMRI session (p = .001). Additionally, they also had an increased tendency to reject originally encoded episodes as such (p < .001). Together, these behavioral findings demonstrate that as expected, memories were modified in response to prediction errors. To investigate the neural effects of memory modification, we analyzed the parametric increase in brain activation to modified episodes that later elicited false alarms in the post-fMRI memory test. On the whole-brain level, increasing activation was detected in several areas, including ventrolateral prefrontal cortex, precuneus, anterior and posterior cingulate cortex, and middle temporal cortex (p < .01, FDR corrected). Further, we performed region of interest analyses in areas which are involved in memory formation, hippocampus and parahippocampal gyrus. In hippocampus, a significant increase of activation was detected for later false alarms (p = .01). In parahippocampal gyrus, we detected a significant decrease in activation for later correct rejections in the post-fMRI memory test (p = .04) and additionally found that contrast estimates were significantly higher for later false alarms compared to later correct rejections (p =.034).

The findings of the current study support the idea that episodic memories can be modified in response to prediction errors. For the first time, we demonstrated that false alarms in an episodic memory test were preceded by parametric increases in brain activation during the processing of prediction errors. Consequently, we suggest that this activation might be reflective of the establishment of updated episodic representations.

Long-Term Memory (Episodic and Semantic) ¹

Learning and Memory Other

Activation (eg. BOLD task-fMRI) ²

FUNCTIONAL MRI

Memory

MRI

Other - Episodic memory;Prediction error;Memory modification