Ensemble perception during multiple-object tracking

Abstract

Multiple-object tracking studies consistently reveal attentive tracking limits of approximately three to five items. How do factors such as visual grouping and ensemble perception impact these capacity limits? Which heuristics lead to the perception of multiple objects as a group? This work investigates the role of grouping on multiple-object tracking ability, and more specifically, in identifying the heuristics that lead to the formation and perception of ensembles within dynamic contexts. First, we show that group tracking limits are approximately four groups of objects and are independent of the number of items that compose the groups. Further, we show that group tracking performance declines as inter-object spacing increases. We also demonstrate the role of group rigidity in tracking performance in that disruptions to common fate negatively impact ensemble tracking ability. The findings from this work contribute to our overall understanding of the perception of dynamic groups of objects. They characterize the properties that determine the formation and perception of dynamic object ensembles. In addition, they inform development and design decisions considering cognitive limitations involving tracking groups of objects.

Keywords: Attention; Grouping and segmentation; Visual perception.

Fig. 1

Depiction of the general multiple-object tracking experimental design. Target groups were highlighted for 2 s (here, the highlight is indicated by white-colored dots, in the actual experiment, the dots flashed), followed by a 7-s tracking phase. Participants then selected the target groups (Experiment 1)/ dots (Experiments 2 and 3) using the mouse cursor. Note variations in the group configurations for Experiments 2 and 3 as described in the text

Fig. 2

Stimuli for Experiment 1. Left panel presents a trial of group size 2, two groups to track, and enclosed groups. Right panel presents a trial of group size 4, four groups to track, and non-enclosed groups

Fig. 3

Overall proportion correct across participants for Experiment 1, non-enclosed groups. Error bars denote standard errors of the means

Fig. 4

Overall proportion correct across participants for Experiment 1, enclosed groups. Error bars denote standard errors of the means

Fig. 5

Stimuli for Experiment 2. Schematic representation of the near, intermediate, far, and separate group spacings. White-colored dots did not appear in the actual experiment but are used for representation of groups in the figure

Fig. 6

Overall proportion correct (dots selected) across participants for Experiment 2. Error bars denote standard errors of the means

Fig. 7

Overall proportion correct groups (all four dots selected) across participants for Experiment 2. Error bars denote standard errors of the means

Fig. 8

Overall proportion incorrect (one to three dots of distractor groups selected) across participants for Experiment 2. Error bars denote standard errors of the means

Fig. 9

Stimuli for Experiments 3a and 3b. Experiment 3b consisted of a 40% increase in movement eccentricity. White-colored dots did not appear in the actual experiment but are used for representation of groups in the figure

Fig. 10

a Overall proportion correct (dots selected) across participants for Experiment 3a. Error bars denote standard errors of the means. b Overall proportion of correct groups (all four dots selected) across participants for Experiment 3a. Error bars denote standard errors of the means

Fig. 11

a Overall proportion correct (dots selected) across participants for Experiment 3b. Error bars denote standard errors of the means. b Overall proportion correct groups (all four dots selected) across participants for Experiment 3b. Error bars denote standard errors of the means. Note: Low-jitter eccentricity in Experiment 3b corresponds to Intermediate-jitter eccentricity in Experiment 3a

Fig. 12

Overall proportion incorrect (one to three dots of distractor groups selected) across participants for Experiment 3b. Error bars denote standard errors of the means