Categorical effects of working memory load on the selection of pop-out categorical oddballs

Recently, we have shown that the speed at which observers select and respond to a categorical oddball (house amongst faces or vice-versa) is strongly modulated by recent experience with members of that category (Lleras et al., VSS 2008). For instance, observers' RT to a face target is delayed by as much as 100ms if on the preceding trial, there was no target oddball (i.e., the search process “failed” to find a target) and all stimuli were faces, even though the oddball pops-out of the display. Here, we investigate the extent to which working memory load interacts with this categorical selection bias by asking participants to perform a working memory task concurrently with the categorical search task. In Experiment 1, participants alternatively viewed displays containing either one face at fixation (the target for the memory task, a modified one-back task) or three stimuli around fixation (the stimuli for the face/house oddball search task). In Experiment 2, the same design was used, but participants were asked to remember individual houses rather than faces. Importantly, the stimulus in the memory task was never used in the search task. Our results showed a category-dependent effect of working memory load on pop-out selection. When the target in the oddball search did not belong to the memory-task category, performance on the oddball search task was heavily modulated by previous experience with the category, replicating our previous findings. However, when the target in the oddball search task belonged to the memory-task category, the effect was eliminated. Interestingly, overall RTs were slowed down, suggesting there was a categorical effect of working memory load on oddball selection that amounted to a bias against selecting any stimulus that belonged to the memory-task category. Experiments 3 and 4 further explored the flexibility of this WM-bias on the selection of pop-out targets.