This article critically reviews the literature from statistical attempts to derive empirical classifications of alcohol and drug abuse. Consideration is given to the search for discrete subtypes using cluster analysis, as well as to attempts to identify major dimensions along which individuals are ordered using factor analysis. Several important conclusions are drawn. First, this research is at an early stage of development and refinements are needed in methodology before it is likely to have broad impact. Next, only limited content areas have been examined to date. For instance, the clustering research on alcoholism has curiously avoided a direct focus on akohoi use patterns, but instead has predominantly studied personality variables. Moreover, the empirical research has not been well integrated with either theory or clinical practice. Although statistical methods have not been instrumental for the identification of new syndromes, these techniques do provide a powerful technology for refining provisional syndromes as well as for making diagnoses more objective and reliable.