Data on 21 performance tests, entry level and promotional, motor and verbal, were evaluated for 263 applicants for city government jobs over a 17‐month period. The heterogeneity in applicants was matched by a similar heterogeneity in raters. Thorough training of raters, clear performance standards, and content valid tests resulted in average inter‐rater reliabilities of .93 (promotional), .87 (entry level), .91 (motor) and .89 (verbal) for the different types and levels of tests. There was no evidence of adverse impact, and turnover decreased from 40 percent to less than 3 percent in the 9‐26 months following the introduction of performance testing. In comparison to paper and pencil tests, performance tests were more cost effective, more face valid, and more acceptable to applicants in this situation. Considering the growing problems in public sector management, performance testing may be viewed as a rose among thorns. Copyright © 1979, Wiley Blackwell. All rights reserved