Biased perception of distributions: anchoring, interpolation and smoothing  as potential causes

Perceiving the degree of variation in the social and non-social environment is a cognitive task that is important  for many judgments and decisions. In the present research, we investigated cognitive underpinnings of how  people estimate the average value of segments of a statistical distribution (e.g., what is the average income of the  richest 25% of a population?). In three experiments (total N = 222), participants learned about the values of  experimentally created distributions of income values and city sizes and later estimated the mean value of the  four quarters of values. We expected participants to draw on heuristic shortcuts to generate such judgments.  More specifically, we hypothesized that participants use the endpoints of the distributions as anchors and  determine the mean values by linear interpolation. In addition, we tested the contribution of three further  processes (Range-Frequency adjustments, Normal Smoothing, Linear Smoothing). Quantitative model tests  suggest that anchoring and Linear Smoothing both affected mean interquartile judgments. This conclusion is  corroborated by tests of qualitative predictions of the models under consideration.