The devices for prediction of police job performance were compared as to their validity in relation to criteria of job performance derived from paired-comparison ratings by superiors, peers, and subordinates. Data analysis initially found contradictory findings on superior, peer, and subordinate ratings. Results from the current research showed serious deficiencies in interrater reliability across all three types of ratings. In addition, significant levels of convergent validity between the various types of ratings were found, which contradicts previous research evidence. A methodology suggested by Guion (1965) and Thorndike and Hagen (1961) was used to produce revised ratings for the three types, with few differences being shown in the relative validities of the four predictors. The MSCS was found to be valuable for use in higher rank promotion decisions in conjunction with other predictors. Biodata may, in a much more indepth form, provide certain important information for use in selection and promotion decisions. The PVQ appears to be valuable for describing the value system of police officers and identification of concepts that influence their behavior, and it appears to be predictive of measures of higher level managerial job performance. The 'limited version' assessment center currently used was deficient in interrater reliability and exhibited no relationship to measures of job performance of assessees. Recommendations for improving the study methodology are offered. Detailed explanations of the devices tested are appended, and tabular data are provided, along with about 200 bibliographic entries. (Author abstract modified)