Abstract
Spatial working memory (WM) is thought to operate by enhancing the response of neural populations that selectively respond to the remembered stimulus location (Awh & Johnides, 2001). However, maps of visual space are widely distributed across regions of occipital, parietal and frontal cortex, and the relative contribution of each region to the maintenance of precise spatial information during WM has not been established. Here, we reconstructed 2D spatial representations of remembered spatial locations using a forward encoding and BOLD fMRI (similar to Sprague & Serences, VSS, 2012). The model was trained using independent scans during which participants viewed a spatially arrayed set of flickering checkerboards. Each WM trial began with the presentation of two small (0.23°) peripheral stimuli. After 0.5 s, a post-cue instructed subjects to remember the location of a single stimulus, the locations of both stimuli, or to simply wait for the next trial to begin (i.e., passive fixation). After an 8 s delay period, a single probe item was presented and subjects made a 2AFC response indicating whether the probe occupied precisely the same location as the matching remembered stimulus. Probe/target separations were manipulated using the method of constant stimuli, which allowed for estimation of spatial WM precision during scanning. We then reconstructed the spatial representations of the remembered locations using patterns of activation measured during the memory delay period. Robust reconstructions of remembered – but not forgotten – positions were observed in a multitude of frontal (e.g., sPCS, the human homolog of macaque FEF), parietal, and occipital (V1-V3) regions. These results suggest that spatial WM is supported by the operation of widely distributed cortical maps that extend well beyond retinotopically mapped regions of early visual cortex.
Meeting abstract presented at VSS 2013