Incidental probability learning: Effects of task- relevant vs. irrelevant stimulus dimensions

Abstract

Relative effects of probabilities associated with task-relevant and irrelevant stimuli were studied in a probability learning paradigm. Four groups of 24 subjects each received different training experiences prior to sequentially predicting 200 occurrences of a green or yellow color. For 200 training trials subjects predicted the shape of arrows outlined on a green or yellow background. During training the two arrow shapes and background colors varied on independent probability distributions (i.e., each dimension either 75:25 or 50:50); during testing the colors varied on a 75:25 schedule. Regardless of the distribution of background colors during training, subjects who received a 75:25 distribution of arrow shapes during training predicted significantly more occurrences of the more probable color during testing than did subjects who received a 50:50 distribution of arrow shapes. Thus, the frequency bias associated with task-relevant stimuli in a probability learning paradigm influenced subsequent predictions of another stimulus dimension, but the probability distribution of the irrelevant stimulus dimension in the first task did not affect subsequent prediction strategies for the same dimension.