Abstract
Parallel processing in efficient search tasks involves a rejection of non-target items via a process of
evidence accumulation. This evidence accumulation process results in a logarithmic increase in response
times as a function of set size. The slope of this logarithmic function indexes the rate of accumulation;
the greater the target-item similarity, the slower the rate, the steeper the slope. Although almost all of
visual search in the real world involves items against backgrounds, evidence accumulation thus far has
only been examined without backgrounds. Here, we examined the effect of background information in
efficient search tasks. In Experiment 1, search stimuli were displayed against a background that was
either a scene, phase-scrambled, or solid-colored. When target-distractor similarity was low, there was
no effect of background type on both the slope and intercept of the logarithmic function. When target-
distractor similarity was high, the slope for the scene background was steeper than that for the
scrambled background, which was in turn steeper than that for the single-colored background. Thus, the
greater the complexity of the background, the slower the rate of evidence accumulation of individual
items. In Experiment 2, we examined the effect of meaningful but unstructured backgrounds by
replacing the solid-colored background with an upside-down scene. Regardless of target-distractor
similarity, the upside-down background produced the shallowest slope (fastest accumulation rate) and
the highest intercept. Consistent with previous findings, our results suggest that processing of scene gist
is automatic. When the scene is meaningful but unstructured, a constant processing cost (increased
intercept) is incurred. This could arise either from the discounting of the inconsistent background, or a
longer time to obtain scene gist. However, object segmentation is easier since objects do not blend in
with the upside-down scene structure, resulting in a faster accumulation rate.