The effects of unstructured peer-tutoring procedures on the math performance of fourth- and fifth-grade students were investigated. Students' performances in two daily math sessions, during which they worked problems of the same type and difficulty, were compared. When students tutored each other over the same math problems as they subsequently worked, higher accuracies and rates of performance were associated with the tutored math sessions. The use of consequences for accurate performance seemed to enhance the effects of tutoring on accuracy. The results from an independent-study control condition, which was the same peer-tutoring except that students did not interact with each other, suggested that interactions between students during the tutoring procedure were, in part, responsible for improved accuracy and rate of performance. When students tutored each other over different but related problems to those that they were subsequently asked to solve, accuracies and rates during tutored math sessions were also higher, suggesting the development of generalized skills in solving particular types of math problems.