Abstract
Several traditional accounts of visual search propose that search is completed in two separate sequential stages: a preattentive stage and a capacity-limited attentive stage. We challenge such accounts by showing that the first stage is an attentive stage of unlimited capacity. This stage gathers evidence in parallel at each location across the visual field regarding the likelihood that the item at that location is likely to be the target. Items unlikely to be the target are rejected from further consideration whereas locations of items likely to be the target are passed onto the second stage of attentive processing. The amount of evidence that needs to be gathered for each item is determined by the similarity of that item to the target template. These two properties (parallel unlimited capacity and similarity processing) produce search functions that increase logarithmically with the number of elements in the display such that the steepness of the log function tracks item-target similarity. Key differences between our proposal and past proposals are as follows. (I) Previous theories propose that the parallel stage is the same for present/absent tasks and target discrimination tasks. Here we show that present/absent search tasks require less evidence accumulation as indexed by the flattening of the log functions. (II) Previous theories propose that when items are rejected from consideration (filtered out), this tends to happen “en masse” (by spreading suppression or grouping, Duncan & Humphreys, 1989). Others have proposed that below a given level of activation, items won’t be inspected (Wolfe, 1994). Here we show evidence for multiple concurrent thresholds as each item’s similarity to the target template determines the evidence threshold at that location. Finally, (III) traditional theories propose an immutable boundary between parallel and limited capacity search stages, whereas we show that the two stages can tradeoff with one another.
Meeting abstract presented at VSS 2015