Revision of Continuous Subjective Probability Distributions

Abstract
Subjects observed sequences of data drawn from binomial populations. After each observation in a data sequence Ss divided the continuum of proportions, from 0 to 1, into three intervals such that each interval was equally likely to contain the population proportion. The boundaries of the subjective intervals were generally quite similar to the corresponding boundaries of the Bayesian posterior distributions, especially after the first few observations in each data sequence. However, a theoretical conservatives S, accumulating information at the rate of only one-half datum per observation, also generated boundaries quite near the Bayesian boundaries. This degree of conservatism in the revision of a continuous subjective probability distribution does not preclude a relatively high degree of accuracy in the placement of credible interval boundaries.