Because citizens who sit on juries are not a random sample of the adult population, the most common form of statistical underrepresentation involves comparing the proportion of the group in the total population to the population called to serve. The paper describes three statistical methods used to measure the disparity in representation: relative difference in proportions (RD), relative chance (RC), and the odds ratio (OR). Unlike RD, the RC and OR statistics work well when the underrepresented group is a small part of the eligible population as when it constitutes a larger segment. Since both are more understandable than the RD approach and because they do not rest on a retrospective point of view, they appear preferable to the statistics currently used in court opinions. Other issues that arise in measuring underrepresentation are deciding the appropriate total population for comparison purposes and the extent of disparity before discrimination can be found. The article cites several cases illustrating these problems and then discusses statistical techniques for detecting subtle forms of discrimination and aggregating data. It uses the cases of United States ex rel. Barksdale v. Blackburn to demonstrate how an appellate court that attempts to be its own statistician can be led astray. Tables, formulas, and 49 footnotes are included.