Abstract
Working memory (WM) extends the temporal period within which neural representations can be integrated and transformed, enabling a vast array of cognitive abilities. Conversely, WM has severe capacity limitations, and varies widely between individuals and across the lifespan. Psychophysical studies and computational models indicate that random noise corrupts the quality of WM representations (Wilken & Ma, 2004; Bays, 2015). Here, we combine computational modeling, fMRI, and TMS to test hypotheses about the neural basis of WM limits. First, we simulated the fidelity of WM in various sizes of neural networks and found that the size of the network population affected WM precision. Second, we used population receptive-field mapping (Mackey, Winawer, & Curtis, 2017) to estimate the size of the precentral sulcus (sPCS) visual map across participants. Consistent with the neural network results, we found a correlation between the size of sPCS and WM precision. Finally, we applied TMS to the sPCS during the delay period of a WM task to simulate the addition of noise in the population. We found that sPCS map size mediated the detrimental effects of TMS on WM accuracy. Specifically, TMS applied during the retention interval caused a greater reduction in WM accuracy in subjects with smaller sPCS maps. Interestingly, subjects with large maps were resilient and were hardly affected by TMS. Together, these results indicate that 1) the sPCS is necessary for accurate WM, 2) it's size may place a hard constraint on WM resources, and 3) individual differences in it's size may predict one's resilience or the degree to which WM representations are corrupted by noise.
Meeting abstract presented at VSS 2018