Catherine Halpern¹ (), Joy Geng¹; ¹UC Davis

The ability to localize and select targets during visual search depends on how well target-defining features are incorporated into the search template in memory. However, heuristics regarding probable target features (that do not define the target, but predict target identity) may also aid search guidance and decision making. We tested this hypothesis by having participants perform a visual search task for a target tilted ±15° from vertical among five distractors tilted ±15° from horizontal, with eye-tracking. On each trial, the six Gabors were split evenly between two colors (i.e., purple–yellow, green–purple, yellow–green). For each participant, one of the colors was randomly designated as the probable target color, which was predictive of target identity; this contingency was not disclosed. In the two color-pair displays where the probable color was present, the target always appeared in that color (100% color-orientation contingency), as did two distractor items; on neutral trials, (those containing the two non-probable colors) the target appeared in either color on 50% of trials. Preliminary analyses suggest that statistical learning of the predictive color strongly influenced attentional guidance. On displays including the probable color, initial fixations were disproportionately directed to items in that color first (μ= 76.5% σ= 8.1%), a bias that strengthened throughout the session (7.9% increase from the first half of the session to the second half). Additionally, fixations on non-probable colored distractor items (on probable color displays) had shorter durations during target evaluation, suggesting that participants more rapidly rejected these distractors. Together, the results suggest that statistically-learned, color-target contingencies can be used to guide attention by reducing the effective set size of the search display, and can bias target evaluation for decision making, even when the color that predicts the target does not define it.

Acknowledgements: UCD, JSMF