The Army Selection and Classification Project has produced a comprehensive examination of job performance in 19 entry‐level Army jobs (Military Occupational Specialties) sampled from the existing population of entry‐level positions. Multiple methods of job analysis and criterion measurement were utilized in a subsample of nine jobs to generate over 200 performance indicators, which were then used to assess performance in a combined sample of 9,430 job incumbents. An iterative procedure involving a series of content analyses and principal components analyses was used to develop a basic array of up to 32 criterion scores for each job. This basic set of scores formed the starting point of an attempt to model the latent structure of performance in this population of jobs. After alternative models were proposed for the latent structure, the models were submitted to a goodness‐of‐fit test via LIS‐REL VI. After accounting for two components of method variance, a five‐factor solution was judged as the best fit. The implications of the results and the modeling procedure for future personnel research are discussed.