Pupil Dilation Reflects Task Relevance Prior to Search

Abstract

When observers search for a specific target, it is assumed that they activate a representation of the task relevant object in visual working memory (VWM). This representation - often referred to as the template - guides attention towards matching visual input. In two experiments we tested whether the pupil response can be used to differentiate stimuli that match the task-relevant template from irrelevant input. Observers memorized a target color to be searched for in a multi-color visual search display, presented after a delay period. In Experiment 1, one color appeared at the start of the trial, which was then automatically the search template. In Experiments 2, two colors were presented, and a retro-cue indicated which of these was relevant for the upcoming search task. Crucially, before the search display appeared, we briefly presented one colored probe stimulus. The probe could match either the relevant-template color, the non-cued color (irrelevant), or be a new color not presented in the trial. We measured the pupil response to the probe as a signature of task relevance. Experiment 1 showed significantly smaller pupil size in response to probes matching the search template than for irrelevant colors. Experiment 2 replicated the template matching effect and allowed us to rule out that it was solely due to repetition priming. Taken together, we show that the pupil responds selectively to participants' target template prior to search.

Keywords: attentional priority; pupil response; template; visual search; working memory.

Figure 1

Schematic depiction of a trial in Experiments 1 A) and Experiment 2 B) Participants memorized one centrally presented color (Experiment 1, A) or two colors (Experiment 2, B) presented to the left and right from fixation. In Experiment 1, the memory item was automatically the search template, target in the search task. In Experiment 2 an arrow retro-cue indicated which color was the target. During the delay period, a probe display was presented, which contained one colored disk. Participants were asked to ignore the probe display and wait for the search display. The task was to search for the memorized color in the search display, and report the orientation (left or right) of the arrow on it. In the example the arrow on the template color (red) points to the left. While in Experiment 1 the search display remained in the screen until response, in Experiment 2, participants had up to 1200 ms to respond.

Figure 2

A) The time course of the pupil response for Experiment 1: The probe was presented at time zero. The thick lines indicate the mean pupil diameter, and the shaded areas indicate the standard error (SEs) across subjects. Colored (horizontal) bars indicate clusters of significant modulations; black line indicates significant differences between colored traces (p < 0.05 cluster corrected, N = 19). B) Average pupil effect per participant: A negative value indicates that the pupil was smaller when the probe matched the relevant color. Per subject we calculated the mean pupil size in the interval 740 to 2410 ms in the Relevant condition and subtracted the mean size of the Not Presented condition.

Figure 3

A) The time course of the pupil response for Experiment 2: The probe was presented at time zero. The thick lines indicate the mean pupil diameter, and the shaded areas indicate the standard error (SEs) across subjects. Colored (horizontal) bars indicate clusters of significant modulations; black line indicate significant differences between the Relevant and Irrelevant probe conditions (p < 0.05 cluster corrected, N = 19). B) Average pupil effect per participant: A negative value indicates that the pupil was smaller when the probe matched the relevant color. Per subject we calculated the mean pupil size in the interval 820 to 2570 ms in the Relevant condition and subtracted the mean size of the Irrelevant condition.

Figure 4

A) The time course of RT by pupil response correlation for Experiment 1 and B) Experiment 2: The probe was presented at time zero. The thick lines indicate the mean Spearman rank correlation, and the shaded areas indicate the standard error (SEs) across subjects. Colored (horizontal) bars indicate clusters of significant correlation (i.e., rs different from zero). Black bar in B corresponds to the difference between the Relevant and Not Presented condition (p < 0.05 cluster corrected).