Abstract
Ample evidence has shown that people are sensitive to the mean size and the variability of a set of items in a display (Ariely, 2001). Through simulations, we show that neither parallel access to all items (e.g., Chong & Treisman, 2003) nor random subsampling of just a few items (Myczek & Simons, 2008) is sufficient to allow accurate estimations of size variability. In four experiments, we examined how variability discrimination is achieved. Participants compared two arrays of circles with different variability in size. In the first 2 experiments, we manipulated the congruency of the range (smallest and largest items) and the variance of the two arrays. We showed that participants were more accurate when range and variance conveyed congruent information. This indicates a reliance on range as an approximation for variability [Experiment 1: F(1,33)=28.5,p< 0.001; Experiment 2: F(1,42)=6.7,p=0.01]. In Experiment 3, we replicated Experiment 2 using outlined, instead of filled, circle displays. By removing most of the texture information from the filled circle displays, we showed that the range heuristic is a general one, being utilized in different scenarios [F(1,31)=11.4,p< 0.002]. In Experiment 4, we directly tested the use of texture information on variability discrimination. On each trial, one of the circle arrays was filled, the other was outlined. We showed that participants were more accurate when the more variable array was filled [F(1,29)=14.8,p< 0.001]. They also relied more heavily on the range heuristic when texture information was not available [F(1,29)=3.4, p=0.075]. These experiments indicate that range and spatial texture information are both utilized as proxies for variability discrimination, and people are flexible in adopting these strategies whenever they are available. Importantly, these "smart" subsampling strategies are at odds with the claim of parallel processing and random subsampling strategies previously proposed in the literature.
Meeting abstract presented at VSS 2018