(FORTUNE Magazine) – Our title alludes of course to ''banding.'' Wait. Strike the ''of course.'' The honest-to-God truth is that your servant had not heard of banding -- at least not in the affirmative-action context -- until he wandered the other day into sultry Washington and noted it was mainly what U.S. government civil rights specialists talk about when power-lunching at Duke Ziebert's. Banding is the linear successor to ''race-norming,'' now definitively discredited. As expounded in this space several years ago, race-norming represented an effort to hire and promote more black and Hispanic workers in the face of the fact that, on average, they scored lower than whites on valid job-related tests -- on, for example, the Labor Department's General Aptitude Test Battery (GATB). Top-down hiring of applicants with the highest test scores repeatedly resulted in underrepresentation of minorities. Depraved solution adopted by numerous state employment agencies using the GATB: Rank workers only against other members of their own ethnic groups, and then report these race-normed percentile scores to employers (who would often not know that the minority bus driver being referred for the job had in fact scored far lower than other applicants). Republicans made an issue of race-norming during the Senate debate on the Civil Rights Act of 1991, and Democrats ultimately stopped trying to defend it. The practice was outlawed by the act. Banding represents a new effort to wriggle out of top-down hiring decisions, and it comes at you with some fancy statistical footwork. A fellow looking for a decent college-level course in probability theory could do worse than start with the briefs in a banding case now before the courts, which involves hiring and promotions in the San Francisco Police Department. Point mainly at issue: In deciding which cops get promoted to sergeant and assistant inspector, can the department legally eschew top-down procedures and opt instead for hiring within a certain band? Suppose, for example, that the highest score attained by anyone taking the test was 90. Would it be okay for the department to rule that all testees who scored between, say, 80 and 90 were more or less equally qualified -- and then provide a measure of preference for minorities within that band? Judge Robert Peckham of the federal district court said such arrangements would be fine with him. His opinion upset the San Francisco Police Officers Association and is now being reviewed in the Ninth Circuit Court of Appeals. Banding fans argue that, depending on the width of the band and the test's ''validity coefficient'' (a measure of the accuracy with which it predicts job performance), the procedure is appropriate. They note correctly that within some bands, you could not really be sure that Worker A, who scored a bit higher than Worker B, would actually do better on the job. If you're not sure about it -- if your confidence level is below the familiar 95% standard -- why should test scores decide B's fate? Why not just give a break to the minority applicant? The argument against banding begins by pointing out that a lower level of $ confidence is not the same as no confidence. In the absence of other information, the higher-scoring testee is always the better job prospect. The nondiscriminatory decision would still be for A. And when you move beyond A and B, and begin talking of hundreds of job decisions within a band, the law of large numbers takes effect and confidence levels soar toward certainty. Almost 200 jobs are involved in the San Francisco case. A psychologist testifying for the police association put the probability at less than one in 10,000 that a group selected for promotion via banding would be as good as one selected via top-down scoring. The odds in the courts are more iffy.