Poster D64, Monday, March 26, 8:00-10:00 am, Exhibit Hall C
Scene-specific cortically distributed activation patterns predict mnemonic reactivation
Benjamin R Geib1, Erik A Wing1, Marty G Woldorff1, Roberto Cabeza1; 1Duke University
Early studies of memory reactivation during retrieval focused on region-specific univariate activation differences. With the advent of multivariate methods (e.g., representational similarity analyses), however, this type of approach has fallen out of favor. While multivariate methods are often more sensitive, a drawback of them is that they assume all stimuli are represented in a given region, a conclusion contrary to perceptual models that suggest some degree of cortical stimulus-specificity. The current fMRI study utilizes novel analysis techniques to address this discrepancy. During memory encoding, subjects were presented with 96 scenes with associated scene labels while making typicality judgements, and during retrieval the subjects were presented with the scene labels alone (e.g., beach). Subjects’ encoding data was pooled to create 96 scene-specific activation maps. Representational similarity analyses revealed that semantically similar scenes had similar activation maps, and that a left-out subject’s activation maps (for all scenes) could be predicted from the other subjects, confirming that distributed scene-specific patterns of activity exists across subjects. Additionally, and more importantly, scene-specific activity was reliably recapitulated in scene-specific regions during retrieval, with greater activity predicting better memory. These results suggest that while localized patterns of reactivation are predictive of memory, more distributed patterns of activity across the brain are also predictive. The relationship between these local and distributed activation patterns will be explored in future analyses.
Topic Area: LONG-TERM MEMORY: Episodic