Abstract
Attentional templates are optimally tuned according to the expected similarity between targets and non-targets. However, our current understanding of the underlying mechanisms that support such tuning is limited. We thus used EEG to track encoding, maintenance and target selection processes while individuals expected to encounter either low or high target-distractor similarity. Participants were asked to memorize low-contrast sinusoidal gratings that varied across twelve possible orientations. Following a delay, they then had to discriminate the target from a lure that differed from the target by ±45° (i.e., coarse discrimination) or ±22.5° (fine discrimination). Across experimental sessions, the proportion of coarse discrimination trials versus fine discrimination trials was varied; in a Mostly Coarse session, coarse discrimination trials outnumbered fine discrimination trials by 3:1, whereas in a Mostly Fine session, the reverse proportion was true. As optimal tuning accounts would predict, accuracy was greater for the Mostly Fine session relative to the Mostly Coarse session, with this difference being largest for fine discrimination trials (indicative of attentional sharpening). Underlying this effect, we observed an attentional difference at encoding and maintenance, marked by greater suppression of posterior-alpha when fine-grained target-distractor discriminations were expected. This attentional difference may support posterior maintenance, as stimulus-specific activity was recoverable from raw EEG at occipital sites over this period if fine-grained discriminations were expected, but not if coarse-grained discriminations were expected. Lastly, at target presentation, we observed a between-session difference in the amplitude of a sustained posterior contralateral negativity ERP component, likely reflecting greater efficiency in template-match evaluations when fine-grained discriminations were expected over coarse-grained discriminations. Overall, we demonstrate that the expectation of high target-distractor similarity produces a global difference in attention at encoding and maintenance, alters the manner by which target representations are encoded and held, and allows for more efficient target-match decisions at the time of response.