The recruits were tested within the first week of their training classes, their performance was judged after 2 years and after 4.5 years on the job. During the first judgment period, the sergeants of the squads containing the subjects rank-ordered all the men in the squad (both subjects and nonsubjects) and rated them on a 9-point scale. During the second judgment period, the sergeants, senior constables, and staff sergeants performed the ranking and rating. In both instances, the rank-order measure was used to divide the subjects into three performance groups (high, average, and low performers). Differences among the three criterion groups were discovered by a step-wise discriminant function analysis which selected the most predictive personality variables. After 2 years, 12 subjects had left the force; after 4.5 years, 4 more had left -- all were categorized either as low or average performers. Study findings support the hypothesis that the predictability of police work to personality variables changes over time. For this study, the striving qualities of 'capacity for status' were good predictors of performance 2 years later. After 4.5 years, the best predictors were those that indicated a stable, responsible, socially skilled individual. These results should be replicated to establish the stability of the predictive equations. Four tables and 19 references are included.