Possible causes for the discrepancy between the ABT report on the national Project Follow Through and the Saint Louis, Missouri program were discussed: (1) ABT data were positively biased because of non-random, differential attrition from the program and control groups; (2) positive ABT data were due to use of a particular cohort group; or (3) local data were negatively biased because of the evaluation design. To examine sources of conflict, a databank was instituted, using old classroom rosters and available standardized test results through grade 5. The first hypothesis was upheld; each ABT group was above the district average and the non-ABT groups were below. The hypothesis that studying a particular cohort group caused the discrepancy was not confirmed. It was determined that previous local evaluation data provided clues to possible forms of bias. The poorest results for Follow Through Classes were centered on first grade students and led to speculation that students needed time to become acclimated to responsive education models. Two suggestions based on this study were made: affective based education may produce unusual patterns of academic growth, and the non-terminal nature of specific programs must be recognized. (MH)