A sample of 367 actual jurors provided demographic and attitudinal information and verdict preferences on four simulated trials. Three of the trials were criminal trials for first degree murder, rape, and robbery. The fourth trial was a civil suit against a common carrier for damages arising from an accident. Each trial was presented on audiotape and was in the format of an actual trial, including opening and closing statements by attorneys, witness testimony, and judge's instructions. The average trial length was 25 minutes. Juror responses were analyzed (using regression analysis) to determine whether jurors' verdict preferences could be predicted from their attitudes and personal characteristics. The regression models could account for only 5 to 16 percent of the variance in the four cases. The failure to account for substantial portions of the variance in jurors' verdict preferences suggests that social scientific methods of juror selection based on public opinion surveys (in which verdict preferences are not available) are not an effective method of selecting juries and pose no threat to the impartiality of the jury system. Tabular data, reference notes, and about 30 references are provided. (Author abstract modified)