Evaluating Holland's and Gati's vocational-interest models: A structural meta-analysis.

Abstract

Summary and Implications Gati's (1991) conclusion that Holland's model was flawedand that his own three-g'roup-partition model of RIASEC typeswas superior is questionable on both empirical and logicalgrounds. To the contrary, we found considerable support for thesuperiority of Holland's model. The results of the three separateanalyses, each examining Gati's and Holland's models fromvarying perspectives, provide evidence of the superiority ofHolland's models (both the order model and the circumplexmodel) over Gati's three-group partition. Support for the superi-ority of Holland's models was provided using the methods ad-vocated by Gati (1991, p. 319) of testing predictions on the rawdata itself (i.e., correlation matrices) and of using a variety ofdifferent types of analyses obviate the idiosyncrasies of eachanalysis. Gati's Criticism of Holland's Model Gati (1991) posed a number of criticisms of Holland's model,for example, issues of dimensionality, interpretation of dimen-sions, inconsistency of Holland's ordering of types, and vari-ance in the distances between adjacent types. For each criti-cism, Gati selected several studies support his argument,ignoring, in most cases, studies that were not supportive ofthese criticisms. Although we differ with the interpretation andadequacy of these selecte d studies, it seems that most of thesecriticisms would have been better addressed empirically, usinga quantitative approach synthesize the literatur e rather than anarrative one. Our analyses countered many of the problemswith Holland's model claimed by Gati (1991). However, twoissues in Gati's criticism of Holland's model that merit com-ment are his equating of Roe's and Holland's models and thepresence of the hole in the of the circumplex model.Although we focused only on Gati's (1991) criticism of Hol-land's model of RIASEC interests, Gati also examined thestructure of interests using Roe's (1956) fields. However, insteadof examining the merits of his model in relation Holland'smodel on RIASEC types and his model in relation Roe's onRoe's fields separately, Gati (1991) linked Holland's and Roe'smodels by labeling these two structures the hexagonal-circularmodel, thereby creating th perception that criticisms of Hol-land's structure were applicable Roe's structure and viceversa. For example, Gati (1991) stated, to conclude, the combi-nation of the conceptua l an d empirica problems face byth e circular-hexagonal structur just discussed raise s doubtregarding the model's adequacy [emphasis added ] and indicatesthe nee d for an (p . 312) Grante that both Hollanand Roe proposed spatial models and that several attempts(Lunneborg & Lunneborg, 1975; Meir & Ben-Yehuda, 1976)have been made document that the two models parallel eachother, Holland and Roe had different vocational-interest catego-ries and measures assess these categories, and most impor-tant, researchers have yet replicate Roe's hypothesized circu-lar order of interest categories (Rounds & Zevon, 1983). In com-parison, it is rare find a study that has not replicatedHolland's RIASEC order (Gati, 1991).Furthermore, when Gati (1991, p. 310) focused on Roe'smodel, he failed make it clear whether he was evaluatingRoe's circular structure or the different representation of Roe'seight fields proposed by Meir (1973) and Lunneborg and Lunne-borg (1975). And meanwhile, yet another inventory-based or-dering of Roe's categories has been proposed (Knapp & Knapp,1984; Knapp, Knapp, & Buttafuoco, 1978). Usually, when atheoretical model fails fit the empirical data, the next step isto evaluate the measures developed assess the constructs(Cronbach & Meehl, 1955) rather than revise or reject thetheory. Because only a few studies (Gati, 1991) have been con-ducted on Roe's model—using primarily the Ramak (Meir,1975), one of three vocational-interest measures proposed tomeasure Roe's interest categories—it seems premature reviseor propose yet another alternative model, such as Gati's hierar-chical model. Instead, research should focus on how the eightinterest constructs are being assessed.Gati (1991) criticized Holland's model, and circular modelsin general, with respect the presence of a hole in the mid-dle (p. 311). Because there are two dimensions underlying thecircumplex model, he claimed there should be types that wouldlie in the middle of the spatial depiction (i.e., moderate on bothdimensions instead of having all scores distributed in a circularorder, equally distant from the origin). Gati's criticism was inerror in that it confused the factor space with the subject space.