Abstract
Analysis of sequential dependencies in stimulus-response sequences showed, in addition to the "gambler's fallacy," strong effects of strings of consecutive alternations in the stimulus sequence. Some evidence for a determining effect of previous responses was obtained. 2 models based on statistical learning theory predicted asymptotic acquisition response levels for the higher values of conditional probability, but did poorly at the low values. They also failed to account for the nonconvergence in transfer and much of the sequential dependency data. Concludes that agreement between obtained and predicted "matching solution" behavior is fortuitous. 21 refs. (PsycINFO Database Record (c) 2006 APA, all rights reserved)