APPLICATION OF A MODEL FOR NUMERICAL RESPONSE TO A PROBABILITY LEARNING SITUATION

127 female undergraduates predicted which of 21 numbers would occur next in a sequence of trials. 2 distributions of reinforcing numbers and 2 sets of instructions were used to test a linear model for numerical response. Sequential dependencies for both distributions had the predicted straight-line form under distance instructions in which S was ostensibly judging the distance between 2 points. Ss showed some nonlinearity under guess instructions. Some suggestion of the predicted numerical matching was also obtained under distance, but not guess, instructions. The slope of the 1st-order dependency was about .3 in all conditions, from which the model predicts a slope of about .21 for the 2nd-order dependency. The observed slopes were much less, and the pattern of results was similar to that of 2-choice probability learning. It was noted that a more general averaging model, with short- and long-term components, might be able to account for the data. (24 ref.) (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1969 American Psychological Association.