This study of a role change intervention, a type of cross-age tutoring, was conducted as a "true" experiment. Forty tutors from four low-achieving ninth-grade general math classes and 68 tutees from three fourth-grade classes were randomly selected. The classes were in inner city schools in which about ninety percent of the students were black and the remainder Spanish-surnamed. For three weeks, each ninth-grade tutor worked on a one-to-one basis daily with a fourth-grade tutee, teaching fractions. In addition to pretests and posttests, retention tests were given almost three months after the end of the project. In addition to the tutoring versus no tutoring manipulation, there were planned variations within the tutoring condition. Time-allocated was one manipulated variable. Another manipulated variable was the method of assigning tutees to tutors. Tutees randomly assigned to be either matched on sex and ability with their tutors, or mismatched on both variables. Posttest and retention test data indicated significantly higher achievement for students in the tutoring treatment. Tutoring had a positive impact on the attitudes of ninth graders toward tutoring. Among tutors, it was those students who were rated by their teachers as being the less well-behaved in the regular classroom who were the more keen to continue tutoring. Residual gain analysis applied to tutee posttest scores indicated a slight tendency for lower achieving tutors to be the more effective teachers. (Author/AM)