Attorney-conducted voir dire is not an effective procedure for selection of impartial juries. The evidence demonstrates a consistent lack of impressive attorney performance. Attorneys disagree substantially about what information to rely on and which jurors to select. They consistently produce low levels of accuracy in judging juror verdict preference prejudices. The heightened power of prediction of statistical models also demonstrates comparatively low levels of success in forecasting juror verdict preferences. Those studies that yielded some accurate predictors, usually through statistical analyses, typically found these predictors to be "obvious" background characteristics of the jurors. The best current prediction methods provide only slightly more accuracy than the attorneys' judgments. 88 footnotes