Abstract
Recent work in attention suggests that selection doesn’t simply highlight a single feature value but takes into consideration the visual context. The current study provides a comparison between two theories which account for relationships between targets and non-targets items. The relational theory proposes that instead of tuning to a specific feature value attention is directed towards the relationship between the target and its environment; leading attention to be directed to the relationally strongest items, e.g. the ‘reddest’ or the ‘largest’. In contrast, optimal tuning accounts suggest that this non-veridical tuning arises through shifting of the attentional template. This theory poses that when targets and non-targets are similar the visual system shifts criteria to an exaggerated target feature allowing for more efficient selection. Probe trials show that shifted-feature items are more likely to be recognised as target items. It is, however, unclear if these results represent early selection or perceptual decision making (and associated memory representations). In the current study, participants performed a visual search in a high similarity (between target and non-targets) or low similarity condition, with the target always the reddest orange amongst orange non-target. On some trials singleton distractors were presented ranging from red to yellow. Saccades directed to these distractors allowed for inferences on which feature values captured attention. Masked probe trials were also displayed requiring identification of the search target; measuring perceptual and memory representations. The visual search results showed that saccades were driven solely by relational context, i.e. all redder distractors reliably attracted gaze. In probe trials the shifting effects were replicated, in the high-similarity context shifted distractors were selected more frequently than the target. This suggests that optimal tuning is not an attentional effect effecting early selection, but is instead related to memory and perceptual representations of the target item.