Context effects (assimilation and contrast) are examined in relation to accuracy in judgments of stimuli. Context effects are distinguished from context errors. This is shown to depend on one's definition of true scores, rater tendencies (leniency-severity) relative to true scores, and the direction and magnitude of observed context effects. The framework is illustrated empirically in a study of contrast effects involving performance judgments. Implications for reliability, validity, and agreement of behavior judgments in practice are explored.