Factor analyses to create the Multidimensional Pain Inventory (MPI) warrant further examination due to small sample size, the use of separate factor analyses for each MPI section, and lack of evidence for the replicability of the factor model. The present study randomly assigned 972 respondents to one of three samples. Study 1 used data from Sample 1 (n = 452) to examine the replicability of the MPI factor structure reported by Kerns, Turk, and Rudy (1985), and to measure the internal consistencies of MPI scales. In Study 2, confirmatory factor analyses using Sample 2 data (n = 267) verified a revised MPI factor structure that was generated in Study 1; cross-validation was achieved using data from Sample 3 (n = 253). The revised MPI model differed from the original model in its number of scales and pattern of scale loadings. Recommendations for the refinement of the MPI are proposed. © 2001 John Wiley & Sons, Inc. J Clin Psychol 57: 765–783, 2001.