Peer collaboration can be a useful tool in a school classroom to help students perform at their best. With whom should students be paired, though? Previous research yields inconsistent findings regarding whether the benefits of peer collaboration depend on the gender or friendship of collaborators. We paired students with a same-gender friend or a nonfriend in their classroom to examine whether friendship and specific dimensions of relationship quality were important for understanding adolescent (N = 132 high-school students) boys' and girls' performance on a scientific reasoning task. Dimensions of relationship quality were related to task performance with greater perceived conflict predicting poorer performance. Girls outperformed boys, but the difference was marginal and nonsignificant after accounting for dimensions of relationship quality. Friends' and nonfriends' performance was similar. Results are informative for educators who use collaboration as an instructional technique and for other professionals who work to support the development of effective reasoning and problem-solving skills among adolescents. (Contains 3 tables.)