This study describes the effects of ability-paired student interactions on achievement as fifth graders (10-year-olds) worked on laboratory activities relating to balance. Achievement gains were assessed (n = 83) by analysis of pretest–posttest differences on the lever concept test. Audio recordings and field notes (n = 30) were analyzed for the following laboratory behaviors: number of words spoken, tinkering, block moving, turns speaking, incidence of helping and distracting behavior. Results revealed that: (1) low-ability student achievement is greater when students are paired with high-ability partners; (2) low-ability students speak more words, exhibit less distracting behavior, and move blocks on the lever less when they are paired with a partner of high ability; (3) high-ability students speak more words, take more turns speaking, and exhibit more helping behaviors when they are paired with low-ability students rather than with other high ability students; and (4) there are no achievement differences for high-ability students regardless of the ability level of their partner. These findings suggest that heterogeneous grouping of students in science can be beneficial to low-ability students partnered with high-ability students, without being detrimental to the high-ability partners.