Abstract
Recall of familiar people or places elicits activation in distinct sets of cortical regions, including ventral temporal cortex, medial parietal cortex, and posterior parietal cortex. To investigate the temporal dynamics of visual recall from memory, we collected magnetoencephalography (MEG) data while participants (N=30) visualized highly familiar people or places requiring retrieval of long term internal representations. Conditions were personalized to each participant and they each provided the names of six personally familiar people (e.g. Aunt Sanika) and places (e.g. gym). In a retrocue paradigm, two names were presented sequentially on the screen (800 ms each with a 200 ms gap) followed by a blank screen (400 ms) and then the presentation of a retrocue (500 ms). This retrocue was either the number 1 or 2, indicating that the participant should visually recall either the first or the second item shown. Participants then visualized the cued item as vividly as possible for 4000 ms. Each condition was presented 32 times and cued 16 times over the course of 192 trials. Trials were broken into 8 runs in which each name was seen 4 times and cued for recall twice. Data was sampled across 272 channels at 1200 Hz with whole-brain coverage, then downsampled to 200 Hz. Principal component analysis was implemented to retain the components explaining 99% of the variance. Using both multivariate pair-wise classification and representational similarity analysis on the responses measured across sensors, we found significant decoding of people versus places starting around 900 ms after retrocue onset. This decoding persisted throughout the visual recall period. Further, we observed a tendency toward decoding individual people and places. These results demonstrate the ability to decode visual representations recalled from long term memory in the absence of any preceding visual stimulation.