Labels implying deviance or handicap can have stigmatizing effects. However, developmental theory and research suggest that such label effects may depend on cognitive processes that only develop in later childhood-processes such as trait inference and the logical linkage of attributions, expectancies, and behavior prescriptions. To test this possibility, we had third, sixth, and ninth graders watch a videotape in which a child failed a puzzle task; all of the students had identical information about the child's achievement scores and puzzle ability, but half were told that the child was mentally retarded. Consistent with key person perception and social cognition findings, the traidike retarded label had little impact on younger children but strong effects on older ones. Like adults in earlier research, sixth and ninth graders saw low effort as a less important cause of failure for the retarded than for the unlabeled child, and they saw less need to urge the retarded child to persist. Correlations showed no evidence of logical linkages among attributions, expectancies, and persistence-urging among third graders, but strong linkages among sixth and ninth graders. The findings suggest that label effects are mediated by cognitive processes dependent on developmental level. There is ample evidence that labels can activate sets that shape person perception. Labels such as warm versus cold can shape people's first impressions of a lecturer (Kelley, 1950) and their more carefully considered judgments about people in simulated job interviews (Huguenard, Sager, & Ferguson, 1970). The simple information that a child is feeling "under par" can significantly color the descriptions people write of that child's behavior (Rapp, 1965). Labels implying deviance, deficit, or a handicap may be especially important because their effects are often stigmatizing. Langer and Abelson (1974) showed clinicians a videotaped interview with a man labeled as either a job applicant or a (mental) patient. Clinicians given the patient label evaluated the interviewee as significantly more disturbed than did clinicians given the job applicant label. Finer and Kahle (1984) found that people performed better on a highly salient memory test when they thought their partner was a former mental patient than when they had no such information; the people were presumably attempting to cope with, and compensate for, perceived disability of their partners.