Problems involved in reducing the achievement gap among ethnic groups in the Seattle (Washington) public schools were studied. An 11-member team selected 25 elementary schools and 216 classrooms for the study, which included classroom observations, teacher questionnaires, and measurement of ethnic achievement gap reduction through California Achievement Test scores. Achievement test results and analysis suggest that the ethnic achievement gap is more directly modifiable by school and classroom practices than is overall achievement level. In the intermediate grades, findings suggest that the optimal setting for reading gap reduction is a classroom with considerable one-on-one interactive instruction. A similar situation fosters mathematics gap reduction, although language gap reduction suggests that situations in which the teacher interacts with the whole class are beneficial. At the primary level, the best setting for gap reduction appears to be a classroom in which students can wander off task, but in which they must remain in their seats. Off-task behavior may be a form of active and productive cognitive processing in "play." Seven tables present study findings. (Contains 23 references.) (SLD)