The primary purposes of this investigation were to analyze the effect of rater bias and to determine the degree of reliability of teacher evaluation ratings from a school district using a career ladder advancement system for teachers. This analysis attempted to establish the degree to which the evaluations reflected true differences in teaching performance rather than biases held by the rater or the inability of raters to produce consistent results over time;Approximately 34,000 completed copies of the formative appraisal instrument used by a large urban school district during the 1985-86 school year were obtained for analysis. Samples were drawn from evaluators at the elementary level to test various hypotheses. This instrument was found to be consistent internally, as a high rate of agreement was found between two subgroups of criteria;The effect of rater biases was tested using the characteristics of gender, race, years of experience, and level of training. Female evaluators were found to give significantly lower ratings to teachers than male evaluators. Also, lower evaluation scores occurred for teachers evaluated by members of racial minorities (black and Hispanic). No significant interaction effect was found for rater/ratee gender or race;Evaluation scores became consistently more severe as the amount of formal training of the evaluator increased. A curvilinear relationship was found between evaluation ratings and the years of experience in education of the evaluator, with mid-career evaluators assigning stricter ratings than less or more experienced raters;A high degree of inter-rater agreement was found between the evaluation scores assigned by a teacher's building principal and a second appraiser, who was a principal from a different elementary school in the same school district. Also, a significantly high correlation was found between appraisers' first and second semester evaluations of a common group of teachers. These findings suggest that the evaluation system used by this district produced highly reliable results.