Within-class experimental designs (with experimental and control groups in the same classroom) are subject to diffusion effects whereby both experimental and control students benefit from the intervention, thereby contaminating the control group and biasing evaluations of intervention effects. In support of diffusion effects, this study shows that a classroom intervention resulted in systematically higher academic self-concepts for internal (within class) controls compared to external (between class) control groups. The construct validity of the interpretation of this difference as a diffusion effect was supported by observer and teacher comments and ratings of teacher success in focusing the intervention on experimental students and different patterns of results for teachers who were more or less successful in maintaining this focus. Potential dangers in sole reliance on internal within-class control groups may outweigh advantages of this expedient experimental design. (Contains 19 references.) (Author/SM)