Abstract
Working memory (WM) increases the duration with which stimulus representations are available for further processing. The canonical theory of WM posits that persistent activity in neurons in lateral prefrontal cortex (PFC) stores WM representations. While electrophysiological data from nonhuman primate studies supports this hypothesis, data from fMRI studies have largely failed to find persistent activity in human lateral PFC. One possible explanation for this puzzling discrepancy stems from recent analyses of the anatomical distribution of neurons tuned to stimulus features across macaque PFC (Miller et al., 2022; Xiang et al., 2023). Namely, the resolution of previous fMRI measurements may be too coarse relative to the fine-grained spatial scales of tuned neurons. To address this potential limitation, we used high-resolution (900 micron isotropic voxels) fMRI at 7T to measure delay period activity during a memory-guided saccade task within a partial slab covering lateral PFC (N=5; TR 2100 ms, 60 slices, partial Fourier 6/8, in-plane acceleration 2, multiband acceleration 2). Custom pre-processing methods were implemented to achieve high spatial accuracy, including EPI undistortion, head motion correction, and run-wise nonlinear anatomical coregistration. We also performed population receptive field mapping to identify topographically organized visual field maps and spatially selective voxels. In our preliminary analyses, we observed persistent activity in several areas of the PFC. Moreover, in some voxels this activity contained information about the location of the memoranda. These results suggest that high-field fMRI with sub-millimeter spatial resolution can bridge the gap between nonhuman primate neurophysiology and human neuroimaging research.