Jun-Ming Yu¹ (), Zoe(Jing) Xu², Simona Buetti¹; ¹University of Illinois, Urbana Champaign, ²University of Washington

When targets differ from distractors along multiple feature dimensions, the attention system can adaptively assign greater weight to the dimension providing stronger guidance signals. Earlier, we quantified this dynamic weighting process with distractor colors sampled from both directions on the color circle relative to the target. Distractor colors were twice as likely to be picked from one direction than the other, introducing potential weighting biases. Here, we selected distractor colors from only one direction. Search performance was modeled when shape had a stronger guidance signal than color (shape easier condition) and vice-versa (color easier condition), with distractors differing from the target in both color and shape. The strength of the shape and color guidance signals was quantified in single-feature-dimension search tasks. In the shape-easier condition (N = 20; color search slopes 190 – 318 ms/log-unit, shape search slopes 101 – 140 ms/log-unit) we found a shape weight of 1.56 and a color weight of 0.44, indicating that participants prioritized shape over color. Unexpectedly, in the color-easier condition (N = 20; color search slopes 108 – 176 ms/log-unit, shape search slopes 263 – 365 ms/log-unit), we found a color weight of 0.53 and a shape weight of 1.47, indicating that participants did not prioritize color over shape. We hypothesized that the lesser reliance on color arised from participants subjective similarity judgments, perceiving target and distractors colors as overly similar and therefore choosing not to rely on color information. As a follow-up, we increased the color guidance strength (color much easier condition; N = 21; color slopes 39 – 78 ms/log-unit, shape slopes 101 – 140 ms/log-unit). The color weight increased to 1.00, indicating that participants shifted attention towards color when the subjective dissimilarity was sufficiently large. The results suggest that participants are not necessarily sensitive to search efficiency, rather favoring subjective similarity judgements.