Abstract
In visual search tasks performance can be manipulated by changing the probability of occurrence of the upcoming stimulus. This “frequency effect” is generally attributed to a form of conscious expectancy. Contrary to this belief, we show that frequency effects can be completely accounted for by an unconscious memory-based mechanism.
Subjects discriminated the cut-off side of an odd-color diamond in a three-element display. We changed the ratio of red- to green-target stimuli from 20% to 100%. Across all conditions the run-lengths of red- and green-target stimuli were kept constant at 3 by interleaving irrelevant blue-yellow stimulus combinations with relevant red-green trials (the clustering manipulation). We observed a large frequency effect with a single irrelevant stimulus: e.g., responses to red targets became faster when red appeared 100% than when it appeared 20% of the time relative to green. The effect became smaller with 3 irrelevant stimuli and was completely abolished by inserting 7 irrelevant stimuli.
It was previously shown that in this task each trial leaves an implicit memory trace for the target defining feature that lasts ∼8–10 trials, and that individual traces sum. In the absence of clustering, frequency effects are due to the fact that more frequent stimuli have longer run lengths, thus more memory is allowed to accumulate because of repetition of the same target feature, leading to response facilitation. The clustering manipulation enabled us to control the amount of memory that accumulates, consequently modulating the frequency effect. A single irrelevant stimulus allowed most memory accumulation and thus a large frequency effect. With seven irrelevant stimuli there was no accumulation and responses to the target appearing 100% and 20% of the time did not differ.
These results demonstrate that implicit working memory, with its inherent unconscious coding of frequency distributions, can completely explain probability effects in visual search.
NIH EY13155 grant to V. Maljkovic