Several high inference ratings were made on a sample of 31 second- and third-grade teachers selected for study because of their consistency in producing student gains on the Metropolitan Achievement Test over three successive years. Data consist of scales, checklists, and percent estimates filled out by coders after or during classroom observations of the selected teachers. Both intra- and inter-rater reliabilities are reported. High and low correlations between these measures and a measure of teacher effectiveness in producing pupil achievement gains are presented and discussed. (Author)