Evidence for negative feature guidance in visual search is explained by spatial recoding

Abstract

Theories of attention and visual search explain how attention is guided toward objects with known target features. But can attention be directed away from objects with a feature known to be associated only with distractors? Most studies have found that the demand to maintain the to-be-avoided feature in visual working memory biases attention toward matching objects rather than away from them. In contrast, Arita, Carlisle, and Woodman (2012) claimed that attention can be configured to selectively avoid objects that match a cued distractor color, and they reported evidence that this type of negative cue generates search benefits. However, the colors of the search array items in Arita et al. (2012) were segregated by hemifield (e.g., blue items on the left, red on the right), which allowed for a strategy of translating the feature-cue information into a simple spatial template (e.g., avoid right, or attend left). In the present study, we replicated the negative cue benefit using the Arita et al. (2012), method (albeit within a subset of participants who reliably used the color cues to guide attention). Then, we eliminated the benefit by using search arrays that could not be grouped by hemifield. Our results suggest that feature-guided avoidance is implemented only indirectly, in this case by translating feature-cue information into a spatial template.

Figure 1

Example of trial events and search array for a negative-cue trial used in Experiment 1 (replication of Experiment 1A from Arita et. al (2012)).

Figure 2

Manual response time results from Experiment 1 plotted as a function of cue condition (negative, neutral, or positive). Mean correct RT was faster in the positive-cue condition (M=1105.69) than in the neutral-cue condition (M=1420.01), but there was no RT advantage for the negative-cue condition (M=1355.05) compared with the neutral-cue condition. Error bars represent within-subject 95% confidence intervals (Morey, 2008).

Figure 3

Positive cue benefit (neutral RT – positive RT) was strongly correlated with negative cue benefit (neutral RT – negative RT; r=.67, p=.001). Half of the neutral trial RTs were used to calculate the positive cue benefit and the other half were used to calculate the negative cue benefit.

Figure 4

Participants were split into two groups based on magnitude of positive cue benefit: high group (greatest positive cue benefit, N=13), low group (least positive cue benefit, N=13). Participants in the high group demonstrated a reliable benefit from the negative cue (M=1229.36), relative to the neutral condition (M=1525.19), which replicated the pattern of results found by Arita et al. (2012). Participants in the low group demonstrated no benefit from the negative cue. Error bars represent within-subject 95% confidence intervals (Morey, 2008).

Figure 5

Example search arrays illustrating the segregated (top panel) and mixed (bottom panel) conditions used in Experiment 2. All other trial events (fixation, cue, ISI) were the same as used in Experiment 1.

Figure 6

Manual response time results from Experiment 2 plotted as a function of cue type (color, location) and collapsed across array type (segregated, mixed) since this did not result in any significant effects. When given a location cue, participants were faster to respond to the target item in both the negative-cue (M=1107.97) and positive-cue (M=1044.60) conditions compared with neutral (M=1255.94). When given a color cue (as in Experiment 1), participants demonstrated a positive-cue benefit (M=1087.84) compared with the neutral-cue condition (M=1315.81), but not a negative-cue benefit (M=1354.44). Error bars represent within-subject 95% confidence intervals (Morey, 2008).

Figure 7

Similar to the results from Experiment 1, the magnitude of the positive cue benefit (neutral RT – positive RT) was strongly correlated with the magnitude of the negative cue benefit (neutral RT – negative RT; r=.62, p=.006). Half of the neutral trial RTs were used to calculate the positive cue benefit and the other half were used to calculate the negative cue benefit.

Figure 8

As in Experiment 1, participants were split into two groups based on magnitude of positive cue benefit: high group (greatest positive cue benefit, N=9), low group (least positive cue benefit, N=9). Participants in the high group demonstrated a reliable negative cue benefit (M=1212.32), relative to the neutral condition (M=1459.93), but only when the different colored items were segregated by hemifield (panel A). When the different colored items were mixed within each hemifield (panel B), participants in the high group no longer demonstrated a benefit from the negative cue (M=1378.57), relative to the neutral condition (M=1397.77). Participants in the low group did not demonstrate a negative cue benefit for either array type. Error bars represent within-subject 95% confidence intervals (Morey, 2008).