The Guidance of Visual Attention Through Learned Feature Probabilities

Visual attention can be influenced through statistical learning of information in the environment, and over time, extracted visual patterns relevant to the current task can be used to guide attention. Specifically, within a visual search paradigm, statistical learning of feature information (e.g. color) can be implicitly extracted to make attentional guidance more efficient. In addition, we know that the system can use explicitly provided information, such as the contents of visual working memory (VWM) to bias attention. We hypothesized that feature-based statistical learning might interact with the contents of VWM. In the present study, participants searched through displays containing target features (i.e. shape) that were more likely to contain the target of the search than another feature. In order to examine the interaction between VWM and previously learned attentional biases, in Experiment 2, we trained participants on the same implicit learning task as Experiment 1. We introduced a color to be remembered that either matched the color of the target, of the distractor, or was a color that was not present in the search display. We found that when a color was stored in VWM and present in the search display, all attentional biases based on the previously learned statistics disappeared. Therefore, we hypothesize that VWM dominates attentional guidance. In other words, feature-based statistical learning disappears when there is a concurrent strong VWM bias.