Dissociating parallel versus serial mechanisms responsible for the inefficiency of conjunction search - a goliath task!

Visual search is less efficient when a target is accompanied by multiple distractors with target-matching features, but the cause for these costs remains contentious. According to serial models, target-matching distractor features interfere with attentional guidance by frequently attracting attention before it is allocated to the target. According to parallel models, these features generate attentional biases concurrently with the target, thereby delaying the moment when biased competition is resolved in favour of the target object. Because these alternatives are notoriously hard to dissociate based on behaviour alone, we employed N2pc components as electrophysiological markers of attentional biases in visual cortex. In two search experiments, lateral targets that were defined by a colour/shape conjunction could be accompanied by two partially target-matching distractors on the opposite side. Any attentional capture by these opposite-side competitors (OS-Cs) should be reflected by an early N2pc with opposite polarity relative to the subsequent target N2pc. The presence of OS-Cs slowed search and delayed N2pcs to target objects. Target N2pcs were further delayed on OS-C trials with slow versus fast target reports. However, this median split did not reveal any sign of early attentional capture by OS-Cs even on slow trials. This was confirmed in Experiment 2, where the number of OS-C displays was increased to enable RT-based quartile and octile splits. Target N2pc onsets and target report delays remained closely associated, but there was no sign of any reversed early N2pc, not even for the slowest 12.5% of all trials. Our results strongly support a parallel biased competition account of attentional selectivity in visual search, but provide no evidence for serial attentional reallocation processes as the main mechanism responsible for inefficient conjunction search.