PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES Evaluating Replicability of Factors in the Revised NEO Personality Inventory: Confirmatory Factor Analysis Versus Procrustes Rotation Robert R. McCrae, Alan B. Zonderman, and Paul T. Costa, Jr. Gerontology Research Center, National Institute on Aging, National Institutes of Health, Baltimore, Maryland Michael H. Bond Chinese University of Hong Kong Sampo V. Paunonen University of Western Ontario Despite the empirical robustness of the S-factor model of personality, recent confirmatory factor analyses (CFAs) of NEO Personality Inventory (NEO-PI) data suggest they do not fit the hypothe- sized model. In a replication study of 229 adults, a series of CFAs showed that Revised NEO-PI scales are not simple-structured but do approximate the normative 5-factor structure. CFA goodness-of- fit indices, however, were not high. Comparability analyses showed that no more than 5 factors were replicable, which calls into question some assumptions underlying the use of CFA. An alternative method that uses targeted rotation was presented and illustrated with data from Chinese and Japa- nese versions of the Revised NEO-PI that clearly replicated the 5-factor structure. If an aeronautical engineer announced that the latest super- computer simulation proved that monoplanes cannot fly, we would not rush to ground the airfleets of the world. We would instead conclude that the computer simulation was fatally flawed. It is the essence of empiricism that conceptual models, no matter how mathematically elegant, are abandoned when they fail to lead to accurate predictions of known facts. In this article we argue that maximum likelihood confirmatory factor Robert R. McCrae, Alan B. Zonderman, and Paul T. Costa, Jr., Lab- oratory of Personality and Cognition, Gerontology Research Center, National Institute on Aging, National Institutes of Health, Baltimore, Maryland; Michael H. Bond, Department of Psychology, Chinese Uni- versity of Hong Kong; Sampo V. Paunonen, Department of Psychology, University of Western Ontario, London, Ontario, Canada. This research was supported in part by Social Sciences and Humani- ties Research Council of Canada Research Grant 410-91-1282. Por- tions of this article were presented at the 102nd Annual Convention of the American Psychological Association, August 1994, Los Angeles. We thank Yoshiko Shimonaka for allowing us to report analyses of the Jap- anese NEO-PI-R; Ross R. Vickers, Jr., Robert Cudeck, and A. Timothy Church for consultation on LISREL analyses; Ralph L. Piedmont for supplying the SPSS version of the program in the Appendix; and Carmi Schooler, Lewis R. Goldberg, John M. Digman, Gerard Saucier, Clar- ence McCormick, Peter Borkenau, Stephen G. West, Kevin B. Lanning, and Marco Perugini for helpful comments. Correspondence concerning this article should be addressed to Robert R. McCrae, Personality, Stress, and Coping Section, Gerontology Research Center, 4940 Eastern Avenue, Baltimore, Maryland 21224. Electronic mail may be sent via the Internet to jeffm@mvx.grc.nia.nih.gov. analysis (CFA), as it has typically been applied in investigating personality structure, is systematicallyflawed:Its statistical in- dices reject models that are empirically replicable (see Study 1) and accept models that are not (see Study 2). We also propose an alternative approach to the statistical evaluation of factor replicability that yields more reasonable results. The Five-Factor Model and Its Statistical Evaluation Over the past decade, studies of natural language adjectives (Goldberg, 1990; Ostendorf, 1990), California Q-Set (Block, 1961) items (Lanning, 1994; McCrae, Costa, & Busch, 1986), and personality questionnaires (McCrae, 1989; Paunonen, Jack- son, Trzebinski, & Forsteriing, 1992) have converged on a repre- sentation of personality trait structure in terms offiveorthogonal factors—the Big Five, orfive-factormodel (FFM). The Revised NEO Personality Inventory (NEO-PI-R; Costa & McCrae, 1992c) was designed to operationalize the FFM by measuring six specific traits, or facets, that define each of thefivefactors: Neurot- icism (N), Extraversion (E), Openness to Experience (O), Agree- ableness (A), and Conscientiousness (C). Exploratory factor analyses that have used varimax rotation of principal components in younger and older, White and non- White, and male and female subsamples have shown very similar five-factor structures that supported the theoretical model (Costa, McCrae, & Dye, 1991). Furthermore, that structure has been es- sentially replicated in self-reports from independent samples of adults (Costa & McCrae, 1992a; Piedmont & Weinstein, 1993) and college students (Costa & McCrae, 1994), in observer ratings from peers (Costa & McCrae, 1992d) andfromspouses (McCrae, Journal of Personality and Social Psychology, 1996, Vol. 70, No. 3, 552-566 In the public domain 552 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.