This study investigated the peer acceptance of integrated students with disabilities as a function of severity of disability and amount of overall student variance in the class. The study involved 14 classrooms and 285 students. Of the 44 students with disabilities, 30 had been classified as mildly disabled and 14 as severely disabled. A sociometric nomination procedure was used to measure peer acceptance. Classroom composition variance was calculated by taking the classroom mean of (1) the classroom percentage of ethnic minority students, (2) the proportion of students identified for special education, and (3) the proportion of students with a low reading performance. Results indicated that significant severity of disability by classroom composition created variance interaction effects regarding social desirability and peer acceptance in a play context, and peer nominations of frequent playmates. Although generally there was low acceptance of students with severe disabilities, the most severely disabled students were generally the most accepted. The paper suggests that the students who were more definitively severely disabled were more likely to attain protected deviant status and differentiated expectations. (Contains 12 references.) (DB)