Intuitive averaging of categorized numerical stimuli

Abstract

Three experiments are reported concerning the ability of people to estimate the average of each of two distributions of numerical stimuli. On each study trial, a three-digit number was shown with one of two category names (nonsense syllables). About 50 stimuli for each category were shown rapidly (e.g., 0.5 sec) in a random sequence. The numerical stimuli were normally distributed in a large pilot study. The mean estimates of the average for each category given by experimental groups in the pilot study were quite close to the stimulus mean and were not affected by whether the two distributions had similar or different means and variances or by the stimulus presentation rate (0.5, 1.5 sec). The distributions of numerical stimuli were skewed in Experiment 1. Estimates of the average for each category seemed to reflect the mean of the stimuli. The distributions were bimodal in Experiment 2. The mean estimates of the average of a subset of the stimuli were observed in experimental conditions when the stimulus distributions were truncated (by instructions to the subject) after the stimuli had been shown. In Experiment 3, subjects produced frequency distributions reflecting the characteristics of the stimulus distribution. © 1983 Academic Press, Inc.