Failures to Find Probability Matching,

Estimates of the asymptotic response probability are presented for individual Ss from thirteen different groups run in two-choice noncontingent probability learning experiments. These estimates are shown to exhibit more variability than is predicted by either of two experimenter-controlled-events linear models. One of these models has previously been shown to provide a generally satisfactory fit to probability learning after extensive training Freidman et. al., 1964. The implications of this variability for decision theoretic models and for other learning models are discussed. Author