Does selective attention utilize physical feature values or perceptual interpretations?

Visual attention is thought to select task-relevant objects using attentional templates tuned to task-relevant features. When we look for strawberries, templates are tuned to red colors. But to what exactly are these templates tuned? Physical values (wavelength ~620-750 nm) or perceptual interpretations (what appears red due to color constancy mechanisms)? We addressed this using the priming of pop-out paradigm (Maljkovic & Nakayama, 1994). Observers (N=8) reported the tilt of an odd-one-out colored square (2AFC) presented along with two distractors. Participants performed mini-blocks (prime+test trials) within displays as seen through different color filters in a VR environment. We tested the nature of observers' representations on test trials by preserving perceptual color characteristics while changing the physical ones (via altering the filter color but keeping the object colors, “Perceptual” condition) or preserving physical values while changing the perceptual ones (via changing the filter and object colors to match the physical characteristics of the prime colors, “Physical” condition). Importantly, we rejected the assumption of perfect color constancy and introduced an adjustment task before the search task where participants reported the perceptual matches for all colors which we then used for the “True perceptual” condition. We calculated priming effects: RT differences between “Swap” test trials (target and distractor colors were swapped compared to the previous prime trial) and “Repeat” trials (target-distractors were repeated). All conditions produced significant priming but the magnitude was largest for the True perceptual condition (~70 vs. 40-50 ms). This shows that attention utilizes a color constancy mechanism - attentional templates are based on perceptual, not raw physical colors. This also highlights the importance of measuring true perceptual matches since color constancy is often far from perfect.