Journal of Counseling Psychology Copyright 1996 by the American Psychological Association, Inc. 1996, Vol. 43, No. 3, 310-329 0022-0167/96/$3.00 Cross-Cultural Structural Equivalence of RIASEC Models and Measures James Rounds and Terence J. Tracey University of Illinois at Urbana-Champaign A structural meta-analysis was conducted to evaluate the fit of J. L. Holland's (1985a) circular order model, I. Gati's (1982) three-group partition, and an alternative three-class partition on vocational interest correlation matrices drawn from the cross-cultural literature. The randomization test of hypothesized order relations (L. Hubert & P. Arabie, 1987) was used to evaluate the model fit for 20 U.S. ethnic matrices, 76 international matrices (representing 18 countries), and a U.S. benchmark sample of 73 matrices. The cross-culture structural equivalence of Holland's circular order model was not supported. Both Gati's partition and the alternative partition fit the U.S. benchmark and international samples equally well. None of the 3 models were found to be an adequate representation of the structure of vocational interests for U.S. ethnic samples. There is an expanding interest in the problems of applying models and measures generated on one population to an- other, especially when this generalization involves applica- tions across cultures (Ben-Porath, 1990; Eysenck, 1986; Hesketh & Rounds, 1995; Hui & Triandis, 1985; Irvine & Carroll, 1980; Lonner, 1990; Lonner & Ibrahim, 1989; Malpass & Poortinga, 1986; Van de Vijver & Poortinga, 1991; Watkins, 1989). Researchers have frequently studied cultural differences, for example, in work values (Hofstede, 1980a), job satisfaction (Griffeth & Horn, 1987), work centrality (MOW International Research Team, 1987), and managerial attitudes (Haire, Ghiselli, & Porter, 1966) by comparing group means on measures across cultures. Such practices can be highly questionable given the possible different underlying meaning ascribed by each culture to the methods and items, leading to possible differences in the structural properties of the measures. Without equivalent structures, mean score comparisons across cultures are pointless: The differences or similarities are essentially un- interpretable. The purpose of our study was to examine the cross-culture equivalence of Holland's (1985a) model of vocational interests. Our study involved both cross-ethnic (U.S. cultural groups) and international comparisons. Holland's Model of Vocational Interests Not since E. K. Strong's (1943) pioneering interest mea- surement research has one approach, that of J. L. Holland James Rounds and Terence J. Tracey, Department of Educa- tional Psychology, University of Illinois at Urbana-Champaign. This article has benefited from the comments and suggestions of Lawrence Hubert. We thank John L. Holland, Bernard Tttreau, Elchanan I. Meir, Beryl Hesketh, Stanley H. Cramer, Rent V. Dawis, William E. Sedlacek, Nadya A. Fouad, L. Suzanne Dancer, Thomas F. Harrington, Gregory J. Boyle, and Sar B. Khan for sending RIASEC correlation matrices and manuals. Correspondence concerning this article should be addressed to James Rounds, 210 Education, University of Illinois at Urbana- Champaign, 1310 South Sixth Street, Champaign, Illinois 61820. Electronic mail may be sent via Internet to j-rounds@uiuc.edu. (1959, 1966, 1973, 1985a), had such a vast influence on how psychologists conceptualize and assess vocational in- terests. Simply put, Holland asserted that there are six personality types: realistic (R), investigative (I), artistic (A), social (S), enterprising (E), and conventional (C), referred to collectively as RIASEC; and that structural relations among the types are best represented by a circular order of R-I-A- S-E-C (often called the hexagonal model). Holland's struc- tural hypothesis specifies that types that are more proximate are more similar than types that are more distant. As doc- umented by Borgen (1986), the impact of Holland's pro- posal has had far-reaching effects on vocational interest assessment. With the merger of Holland's and Strong's systems (Campbell & Holland, 1972), for example, the Strong Interest Inventory (Hansen & Campbell, 1985) has developed scales to assess the RIASEC types and has used Holland's model to organize assessment results and inter- pretation. New interest inventories such as the Career Decision-Making Interest Survey (Harrington & O'Shea, 1993) and Career Assessment Inventory (Johansson, 1986) are also based on Holland's RIASEC model. Major voca- tional assessment programs including the American College Testing Program (1988) and the United States Employment Service (U.S. Department of Labor, 1979) have either ex- plicit scales to assess RIASEC types or methods to convert interest scale scores to Holland's system. Procedures have even been suggested to link Kuder's General Interest Sur- vey scales (Kuder, 1988), Jackson's factor-derived General Occupational Themes (Jackson, 1986), and scales built on Roe's theoretical model (Lunneborg, 1981; Meir & Ben- Yehuda, 1976) to Holland's RIASEC types. With such widespread acceptance of Holland's RIASEC model in the United States, it is not surprising that research- ers and practitioners have adopted his model and measures internationally. Holland and Gottfredson (1992) reported that the Strong Interest Inventory (SII) has been translated into 17 languages, and the Vocational Preference Inventory (VPI; Holland, 1985b) and Self-Directed Search (SDS; Hol- land, 1985c) have been translated into 20 languages. Test publishers in the United States have recently begun to 310