Attention in visual search: Distinguishing four causes of a set-size effect

Palmer, J. (1995). Attention in visual search: Distinguishing four causes of a set-size effect. Current Directions in Psychological Science, 4(4), 118-123.

Abstract:
Analyzes set-size effects on search accuracy, using signal detection theory. Examples of experimental paradigms are presented which determine when the 4 kinds of processing (preselection, selection, postselection, and decision) in a generic attention model contributes to a set-size effect.

Note:
A general information processing model is proposed by Broadbent (1958) to explain the set-size effect in the context of visual search.

p.s. The set-size effect means the reaction time increases a s a function of the increase of distractors in the search display.

Four kinds of processing stages were defined by the concept:
Pre-selection: early visual processing before selection
Selection: voluntary attention control
Post-selection: it depends on the selection theory which is limited by capacity
Early selection: most of perception is post-selection
Late selection: all of perception occurs before selection
Decision: integrate information from all of the relevant stimuli to determine a single response

Using four stages to account for the set-size effect:
Pre-selection: early visual processing is constrained by the eccentricity and the target-distractor similarity. To control the confound of eccentricity, all stimuli was at a equal distance from the fixation. The contrast increment threshold was measured to excluding the influence of target-distractor similarity. With a cuing paradigm, the results still showed n effect for the relevant display set-size (cued location). It is suggested that pre-seleciton did not matter the set-size effect.
Selection: All target and distractors were selected. The selection process did not matter the set-size effect.
Decision: Internal representation is independent of size. Assume (1) representation is noisy (2) with most optimum of common rule. A signal detection approach was used to model the data (distribution of distractors v.s. distribution of target + distractors). With the increase of distractors, the distribution of distractors shifts to right with higher probability to be mistaken as target. Simple decision model fits well.
Post-selection: Posner (1980) proposed a limited-capacity model. The model did not fit the data well, but fit others in the other task context in which participants did not know what they were searching for and had to remember all the stimuli.

The decision integration model is sufficient to account the set-size effect in the search task such as detecting an increment in contrast.