Windle and Scheidt (2004) answer the question posed in their title by suggesting that two subtypes are not ‘sufficient’ for understanding the heterogeneity of alcoholism. Although they identify four subtypes, the authors replicate the consistent finding of a core group of highly dependent individuals, who score toward the extreme end on dimensions of risk and problem severity, tend to evidence antisocial personality and, to a lesser extent than in some prior studies, tend also to be male. The authors partition the remaining pool of less severely affected individuals somewhat differently than other investigators; nevertheless, there are notable consistencies with earlier studies (e.g. negative affect as a distinguishing characteristic). Although this study replicates and extends prior research, it also underscores the fact that much more work is needed to demonstrate the utility of alcoholic typologies. As in many other studies, Windle and Scheidt used cluster analysis to determine the number and characteristics of subtypes. As the authors acknowledge, the selection of a clustering solution, in the absence of definitive statistical criteria, is somewhat impressionistic. However, even with firm statistical guidance, clustering-derived typologies will vary in subtype number and features, depending on sample size and attributes, contextual (cultural and historical) factors, method of analysis, and the predilections of investigators. Thus, in addition to the question of how many subtypes, there are important conceptual, theoretical, and analytic issues in need of thoughtful consideration. First, the present research, and other replications of the Babor et al. (1992) study, have used multiple risk/severity measures, including concomitant psychopathology, as input variables for generating subtypes; personality/temperament variables were then examined in relation to the empirically derived subtypes. However, the theoretical rationale for differentiating variable sets in this way remains to be fully articulated. Careful attention should be given to the fundamental questions of what domains should serve as the basis of classification, and of how independent evidence can be used for validation. Within this context, the interplay of both personal and environmental factors in influencing the development of alcohol dependence should be borne in mind. Following the Windle and Scheidt observations, candidate variables for defining and/or validating subtypes might be expanded to include factors studied by molecular geneticists, as well as measures of affect regulation and of early family and peer experiences. Second, most typology studies have used treatment samples recruited in the USA, which has obvious limitations in terms of identifying subtypes. Greater inclusion of non-treatment participants, and samples from other cultures, would undoubtedly improve understanding of the role that personal attributes and contextual factors play in producing differences among alcoholics in etiology, course, and outcome. Third, investigators need to confront the limitations of conventional clustering techniques and expand their use of alternative approaches for explicating data structure and examining process (Del Boca 1994). Because subtypes should enhance our understanding of the etiology and course of alcoholism, researchers should make greater use of longitudinal data and methods of analysis, such as latent growth mixture modeling, which can identify multiple processes of change. In contrast to conventional clustering procedures, such methods have the further advantage of probabilistic assignment of individuals to classes. Finally, like the current research, many typology studies make use of data initially collected for other purposes. This practice limits the investigator's ability to assess theoretically important measures, to recruit appropriate samples, and to use alternative analytic approaches. Advances in understanding the heterogeneity of alcoholism will require investigators to design studies that more directly serve this purpose. Alcohol & Substance Use Research Institute Department of Psychology University of South Florida 4202 East Fowler Avenue Tampa FL 33620-8200 USA E-mail: fdelboca@cas.usf.edu