Labeling bias is the distortion of diagnostic judgment as a function of prior diagnostic information. University students (N=102) were assigned to 16 treatment groups, with variables of: prior diagnosis attributed to a videotaped target person ("psychotic" versus none), setting in which the observed sample of his behavior was said to occur (clinical versus nonclinical), bogus "outcome measure" completed by subjects following the videotape (inferential versus descriptive), and subjects' training in psychology (advanced clinical and counseling graduate students versus undergraduates). Actual dependent variables were a mental health-illness rating scale and a diagnosis checklist, completed after the bogus outcome measure. Undergraduates exposed to the label "psychotic" gave more pathological judgments than those not given a prior diagnosis; in general they judged the target person to be more pathological than did graduate students. Subjects encouraged by the bogus outcome measure to make inferences about the target person gave more pathological judgments than those encouraged only to describe him; however, this effect was weak and appeared only in an interaction. Subjects viewing the target person in a nonclinical context ascribed more pathology to him than subjects viewing him in a clinical context; this effect of setting was significant only in an interaction. (Author/NRB)