In the old days, in murder mysteries, the detective's task was to find out the truth. Proving the guilt was a bother generally dispensed with, for the murderer, once confronted by Jane Marple or by Philo Vance, could be counted on to confess and conscientiously take poison. The evidence leading to the solution did not need to do double duty and stand up in adversarial proceedings. Howard Wainer's (1989) stimulating, thoughtful comments on statistical adjustment set me thinking how much harder the job of the detective has become. In this age of Adam Dalgleish and Virgil Tibbs, finding the truth is the least of tasks. The work is in building the case, building it out of admissible evidence to survive in a court of law or judgment. As detectives, we as statisticians are obliged more and more to go beyond trying to find the truth, and engage in building a case. It seems to me that this is an important distinction for thinking about Howard Wainer's example of SAT scores. If I wished solely to satisfy myself as to whether higher per-pupil expenditures really do show up in higher SAT scores, I would go in for exploratory data analysis of a sort I shall sketch in later paragraphs. But if I needed to build a case worthy of convincing a skeptical, reflective voting public (or myself as reader, not writer) such an approach would be the wrong one. When better modeling means more complicated modeling, as it often does and as it does in Howard Wainer's examples, better modeling may be of limited use for creating consensus among an audience with differing viewpoints and conflicting interests. My own panacea for the problems of statistical adjustment is a trite one. Still, it is one that seems to me worth emphasizing at every opportunity. The only real solutions to the difficulties of adjustment come from direct empirical research on determinants of the particular nonresponse and nonparticipation for which adjustment is desired. With the SAT example, I think there is great scope for exploratory data analysis in a search for the truth of the matter. I should be inclined myself to look at the upper tails of the distributions of SAT scores state by state. I would be hoping to find similarities in shape from state to state in the upper tails, and if I did, I would try to account for differences in the lower