Maintenance and transformation of representational formats during working memory prioritization

Visual working memory depends on both material-specific brain areas in ventral visual stream (VVS) that support the maintenance of stimulus representations and on regions in prefrontal cortex (PFC) that control these representations. These two areas putatively support bottom-up information processing vs. top-down control functions that rely on oscillations in the gamma and beta frequency bands, respectively. Recent studies identified stimulus-specific working memory contents via representational similarity analysis (RSA) and analyzed their representational format using deep neural networks (DNNs) as models of the multi-layered hierarchy of information processing. How executive control prioritizes specific working memory contents and whether this affects their representational formats remains an open question, however. Here, we addressed this issue using a multi-item working memory task involving a retro-cue that prompted participants to either maintain one particular item or the entire item sequence. We exploited the excellent spatiotemporal resolution of intracranial EEG (iEEG) recordings in epilepsy patients and analyzed activity at electrodes in VVS (n=28 patients) and PFC (n=16 patients). During encoding, we identified category-specific information in both VVS and PFC. During maintenance, this information only re-occurred in VVS but not in the PFC - suggesting transformation of PFC representations from encoding to maintenance, putatively driven by the prioritization process. We thus used RSA in combination with two different DNN architectures to investigate the representational format of prioritized working memory contents. When we applied a feedforward DNN, we found that representational formats during maintenance could be well explained in VVS, but not in the PFC. We therefore employed a recurrent DNN architecture (BL-NET) to evaluate the representational formats in PFC during maintenance. We observed that representations in the deepest layer of this network matched the representational format of prioritized stimuli in the PFC. This effect was transient, locked to the presentation of the retro-cue, and specific to the beta (16-29Hz) frequency band. Together, these results suggest that PFC represents prioritized working memory information in an abstract representational format, which may dynamically control working memory representations through top-down mechanisms. ### Competing Interest Statement The authors have declared no competing interest.