Multiplicative effect of task-irrelevant feature variability on variability judgments

Pooling within the neural population processing a specific feature, or population coding, has been proposed as a mechanism for ensemble judgments of simple features (e.g., mean orientation judgments, orientation variability judgments). According to the population coding model, task-irrelevant features have no influence on ensemble judgments, and if they do, the influence would be additive since they can only influence judgments after the target ensemble property has been estimated. In present study, we investigated whether task-irrelevant feature variability has additive or multiplicative influence on variability judgments. To this end, participants were asked to compare the orientation variability between the two arrays of tilted oval shapes. Tilted oval shapes were randomly placed within a 4-by-5 grid on one of the two arrays (random array). In the other array, they were placed so that their orientations were sorted from counterclockwise to clockwise starting from the top left corner of the grid (sorted array). We expected participants to judge the sorted array as less variable if the orientation variability was similar between the two arrays. Furthermore, we manipulated the color variability of the two arrays. In half of the trials, oval shapes had a uniform color within each array, while in the other half, they had variable colors. If the influence of task-irrelevant color variability is additive, differences in the perceived variability between the random and sorted arrays would decrease. On the other hand, if the influence is multiplicative, differences in the perceived variability would increase. Participants consistently judged the sorted array as less variable when the two arrays had the same orientation variability, and the introduction of task-irrelevant color variability led to increased differences between the perceived variability of random and sorted arrays. These results suggest that task-irrelevant feature variability influences ensemble judgments prior to the estimation of target ensemble property.