Sydney George¹ (), Xiaoli Zhang¹, Melisa Menceloglu¹, Taosheng Liu¹; ¹Michigan State University

Many studies have shown that multiple features can be used to guide attention, but it is still debated precisely how multiple templates are maintained in visual working memory. In the current study, we aimed to differentiate two hypotheses: a multiple-template model where more than one template is maintained simultaneously; or a single-template model, where only one template can be active at a given time with the possibility of switching. We used a brief stimulus duration to prevent template switching in order to examine these competing hypotheses. Participants were shown a dot array (33 msec) with multiple colors and were asked to decide whether there was an over-represented color. In separate blocks, participants were either given no cue, one cue, or two cues at the beginning of the trial. The cue(s) were 100% predictive of the over-represented color when presented. Analysis of detection sensitivity suggests that participants can use either one or two features to guide attention. We then examined accuracy for target present and absent trials separately and found that compared to the no-cue baseline, the hit rate for the one-cue condition was the highest, followed by the two-cue condition, which is predicted by both hypotheses. Given our design, the multiple-template model predicts two weaker templates in the two-cue condition as compared to one strong template in the one-cue condition, which would result in a higher false alarm rate (FA) in the two-cue condition.​​ On the other hand, the single-template model would predict similar FA rates if switching is not possible. The results showed that FA rates were significantly higher in the two-cue condition than the one-cue condition. This indicates that two weak templates were maintained in the two-cue condition, which supports the multiple-template hypothesis. We propose that attentional guidance of multiple features may be enabled by simultaneously active templates.

Acknowledgements: This work was supported in part by grants from NIH (R01EY032071) and NSF (2019995).