This investigation examined the effects of deviancy labels on teachers' expectations of child behavior and their ability to evaluate child behavior objectively. One hundred elementary school teachers were randomly assigned to one of four label groups. Each group dealt with one label (emotionally disturbed, learning disabled, mentally retarded, normal), and each group participated in two separate treatment phases. During Phase I teachers identified behaviors they expected to be displayed by hypothetical children characteristic of the label condition. They were asked to complete a referral form for either a hypothetical normal, mentally retarded, learning disabled, or emotionally disturbed child. During Phase II, each group saw the same videotape of a normal fourth grade boy and completed a second referral form based on the behaviors displayed during this presentation. Experimental procedures were identical for the four groups, except each group was told the child was a member of a different category. Results indicated that teachers hold negative expectancies toward children categorized with a deviancy label and maintain expectancies even when confronted with normal behavior, behavior inconsistent with the stated label. Maintenance of this bias is sufficient to cause teachers to misinterpret actual child behavior, resulting in a halo effect. Results further indicated that the label of educable mentally retarded generated a greater degree of negative bias than did the labels learning disabled or emotionally disturbed, although all three deviancy labels produced negative expectancies and halo effects significantly different from those found under control conditions.