Abstract
Previous work has shown that statistical regularities can bias working memory encoding towards locations where targets are more likely to appear, even in groups of subjects that cannot report the likely target positions. Recent techniques using EEG and inverted encoding models have shown that the topography of alpha-band (8-12 Hz) activity can provide a sensitive index of where covert attention is oriented, as well as the positions of items stored in visual working memory. Thus, we examined whether this neural index of spatial selection would track the spatial bias towards likely target positions, or whether this bias taps into a mode of selection that is not indexed by alpha. Participants were briefly presented with a display of eight colored squares equidistant from fixation. After a brief blank delay, the squares reappeared. On half of the trials, all squares remained the same color. On the remaining trials, there was a change in the color of one of the squares. Unbeknownst to subjects, change trials had a seventy-five percent chance of occurring in one predetermined (dominant) quadrant, which was counterbalanced across participants. We replicated the finding that change detection performance was significantly higher for items in the dominant quadrant relative to items in the non-dominant quadrants, which suggests that memory encoding was biased towards the dominant quadrant. However, spatial tuning functions estimated using alpha activity did not reveal a bias towards the likely target locations during either pre-trial or delay period epochs, suggesting that this spatial bias is implemented by a distinct kind of visual selection, or that it reflects a bias in the efficacy of decision processes.