Abstract
Previous studies have demonstrated that the contents of visual working memory (WM) can be decoded from the spatial patterns of brain activity in visual cortex (e.g., Serences et al., 2009; Harrison & Tong, 2009; Rahmati et al., 2017). Results such as these support the sensory-recruitment model of WM, which posits that the neural populations in visual cortex involved in stimulus perception are also involved in the maintenance of WM representations. Feedback signals from frontal and parietal cortex, accordingly, are thought to help keep these sensory representations in a state that can be easily used to guide behavior contingent on WM (Curtis & D'Esposito, 2003; Postle, 2006). Here, we test the hypothesis that the maintenance of a non-spatial feature (i.e., orientation) is supported by retinotopic codes. We used fMRI and population receptive field mapping to identify retinotopic visual maps in occipital cortex and along the dorsal parietal cortex (Mackey, Winawer, & Curtis, 2017). We then examined WM-related activity in these visual maps as participants performed a delayed orientation discrimination task. On each trial, participants maintained the orientation of a single sample Gabor over a 10.5 second delay period. After the delay, they compared the memorized orientation to that of a probe Gabor presented in the quadrant diagonal to the sample. We used two inverted encoding models (IEM), one to reconstruct the orientation of the sample (Ester et al, 2013) and one to reconstruct the spatial locus of WM from the population activity in each visual map (Sprague et al, 2014). Although location information was irrelevant to task performance, occipital and parietal areas tracked the sample and upcoming probe locations. Importantly, orientation could be decoded best from the original sample location, suggesting that spatial signals may help keep non-spatial sensory representations in a state that can be accessed for memory-guided decisions.
Meeting abstract presented at VSS 2018