The present study examines the relationship between alternation behavior and performance, and in addition, the effects of reinforcement configuration and relevant dimension upon the use of alternation strategies in probability learning. Also investigated is the hypothesis that children's errors at terminal levels of performance in a two-choice probability learning situation are attributable to errors resulting from attention to irrelevant dimensions. Subjects were 31 third grade and 47 fourth grade children who were given a two-choice probability learning task in which one relevant dimension (brightness or position) and one irrelevant dimension (position or brightness) were present. One of two reinforcement probabilities (2:1 or 5:1) was utilized. Children's preferences for aspects of the irrelevant dimension were found to account for more errors than would be expected by chance. The use of alternation strategies was differentially affected by both the reinforcement probability and the relevant dimension employed in the task. The results offered some support for the hypothesis that errors at terminal levels of performance can be accounted for by attention to irrelevant dimensions. An important finding was that the relationship between alternation strategies and performance was jointly determined by the relevant dimension and the reinforcement configuration employed. (Author/MS)