MMPI protocols of 424 active members of the San Antonio Police Department (Texas) were used in the study. Two contrast groups of 23 each were selected for study from the total sample. Both judges had extensive experience with the MMPI, and one had 2 years experience as a screener for police candidates. The judges were asked to discriminate between unacceptable and acceptable police officers and worked at four tasks, collaborating on one. The judges used markedly different selection strategies and disagreed on nominal placement in about one-third of the cases. They reached the near-maximum non-chance agreement possible, given that they used the nominal classes at different rates. Neither judge was more accurate in his selections (when current officer performance records were compared with judges' determination of acceptability) despite different selection strategies. They could not improve their overall individual performances when collaborating. Thus, results suggest the vulnerability of this screening process to judge bias and ultimately weaken the validity of the MMPI as a sole screening method. Criterion variables require improved definition before the search for valid predictors can be objectified. Twelve references are included.