Models of the jury selection process in a Federal District Court are developed from court records, legal standards, and other information and then compared to determine sources of bias. The jury selection plan developed by the court has been operationally reformulated using 1970 census data to identify bias resulting from even a fair application of the plan. This normative model is also used to test for biases in the procedures and the results of the 1970 plan implementation in Eastern Massachusetts based on voter registration lists. Findings support the hypothesis that the juror selection process effectively discriminates against the poor, the young, racial minorities, women, and persons with low and high educational attainment. Sources of bias are found in the use of outdated voter registration lists, unreturned jury qualification questionnaires, and the excuse process. Recommendations are given for ways to reduce the extent of this bias.