Abstract
Visual attention is thought to select task-relevant objects using attentional templates tuned toward relevant features. When we look for ripe raspberries, templates are tuned to red. But what information are these guidance templates tuned toward: perceptual interpretations or physical values? Are they tuned to physical red (wavelength roughly 620-750 nm) or to what appears red in the current context due to color constancy mechanisms? We addressed this question using a variation of the feature distribution learning paradigm (FDL; Chetverikov et al., 2016). Our observers (N=15) searched for an odd-one-out colored diamond among 47 distractors and reported which of the four corners of the target diamond had a notch (4AFC). Importantly, observers performed mini-blocks of visual searches within displays as seen through different color filters. Each mini-block consisted of three to four learning trials and one test trial. Throughout the learning trials, the target, distractors, and color filter were constant, promoting the formation of a specific attentional template. Importantly, on test trials, we manipulated context via color filters and/or the physical values of the stimuli to introduce a physical change or a perceptual change in target and distractor colors. The results showed that after the physical change (when perceptual values stayed the same), there were little or no differences in RT and error rates between the prime and test trials. Conversely, RT and error rates increased dramatically after the perceptual change. This means that perceptual, not physical changes affect the attentional selection and that attentional templates are formed according to perceptual interpretations, not physical values. This supports the idea that attentional guidance templates operate at higher levels of the visual hierarchy where perceptual interpretation has been resolved.