This paper reports 20 unsolved problems in educational measurement and points toward what seem to be promising avenues of solution. The first group of concerns involves validity, posing the problems of obtaining suitable validity criteria, determining and measuring the predictor variables that best characterize the traits and proficiencies of interest, and determining the validity of a test being used for selection. Other unresolved issues are those of data insufficiencies, statistical adjustment, test fairness, domain coverage, reliable scoring, and standard setting. A number of technical issues remain to be resolved, including considerations of difficulty, problems of estimation, response time, and the complexity of psychometric models. The final question is that of attracting, training, and retaining technically-talented individuals in the field of educational measurement. (Contains 44 references.) (SLD)