Whitin-object conjunction visual search: evidence for grouping and binding before and after practice

It has been suggested that within-object conjunction search can be parallel and preattentive provided that feature binding (by specific junction detectors) and nontarget grouping (based on regularity) occur. We tested the binding hypothesis by investigating the time course of learning in visual search for conjunction of features. Two learning sessions (10 and 5 blocks each) took place three months apart. Results show that: (i) in within-object conjunction (inverted V among upright Vs) but not in across-object conjunction search (red among green and red), learning endured from one session to the other, suggesting that it is located at an early level of processing in the central visual system; (ii) this enduring effect in within-object conjunction disappeared when both target and distractors were rotated to modify their absolute but not their relative orientations, as predicted by the binding hypothesis. We tested the grouping hypothesis in two ways. In perceptual learning experiments we showed that the enduring effect disappears when collinearity is perturbed. In standard psychophysical experiments we showed that temporal thresholds (stimulus duration for 75% correct), measured independently for 'present' and 'absent' target conditions as a function of background numerosity (4 to 64 elements), increase in within-object conjunctions and decrease in simple search when collinearity is perturbed.