Best practice assessment of childhood ADHD includes behavior ratings from multiple sources across multiple environments. However, adolescents in secondary schools interact with several teachers each day, and research has shown that teacher perceptions of the same child can be highly inconsistent. As a result, rating scale data can be equivocal, depending on which teachers are selected. The intent of the present study was two-fold: 1) to assess the consistency between teacher behavior ratings of adolescents with ADHD, and 2) to explore predictors of rater leniency or severity (i.e., sources of bias). Results suggest that interrater reliability within our sample was moderate, consistent with previous research. Further, teacher characteristics, including sex and age, were related to biases on ratings of student hyperactivity-impulsivity. Specifically, women and younger teachers provided significantly more severe ratings on average than did men and older teachers. Implications for the interpretation and statistical norming of ADHD rating scales are discussed.