Cladistic analysis of 43 foliar characters was used to explore the phylogeny of genera in Phyllanthoideae and to test the systematic value of foliar morphology. Character polarities were assessed by outgroup analysis treating other subfamilies of Euphorbiaceae and primitive genera in Flacourtiaceae as outgroups. Results indicated that several major tribes recognized in the most recent classification of the subfamily are essentially monophyletic, including Aporuseae, Bridelieae, Drypeteae, and most of Wielandieae, whereas others, notably Antidesmeae, are probably paraphyletic. Conflicts between leaf-based cladograms and the previous classification appeared in the relationships of Astrocasia, Didymocistus, Neowawraea, and genera in Amanoeae, but relationships were clarified for genera like Bischofia, Uapaca, and Hymenocardia, whose positions have previously been obscure. These results verify the systematic value of foliar characters. However, the rates of homoplasy found in this study are higher than found in most cladistic analyses, possibly indicating that many foliar characters are of most value at lower taxonomic levels. Relatively conservative characters include outer vein loop size, tertiary vein pattern, areolation development, unicellular trichome type, and tanniniferous epidermal cells. In the last fifteen years, the success paleobotanists have had using leaf architectural and epidermal characters to classify fossil leaves (Hickey 1971, 1977; Wolfe 1972, 1973; Dilcher 1974; Doyle and Hickey 1976) has led some systematists to explore the usefulness of these characters with extant plants (e.g., see Baranova 1972; Hickey and Wolfe 1975; Raju and Rao 1977; Rury and Dickison 1977; Den Hartog and Baas 1978; Hickey 1980; Bedel 1981). In general, these workers found close concordance between relationships suggested by foliar characters and traditional classifications, a result that seriously challenges the widely held belief that leaves are of little systematic value, especially at higher taxonomic levels (for the traditional view, see Stebbins 1950:515-516; Good 1964:312; Cronquist 1968:39). Before foliar characters could be used successfully, character sets that both adequately describe the range of leaf morphology and can be used with modern analytic techniques had to be developed. Particularly important contributions were made by Stace (1965), who catalogued epidermal characters (see also Dilcher 1974), and Hickey (1973), who developed a system of terminology for describing leaf architecture. Later, Dolph (1976) and Hill (1980) attempted to apply numerical phenetic techniques with these and other leaf character sets. Using collections of leaves randomly drawn from trees of various families, they found that within species leaves varied little enough that they clustered separately from those of other species, but that the patterns of similarity represented by the leaf-based phenograms did not correspond to generic or familial relationships among the plants. As part of an ongoing study of the foliar morphology of Euphorbiaceae, I analyzed leaf architectural and epidermal characters of Phyllanthoideae, putatively the most primitive subfamily, using a numerical phenetic method called Similarity Graph Clustering (Wirth et al. 1966; Legendre and Rogers 1972). My results (Levin 1986b) contrasted markedly with those of Dolph (1976) and Hill (1980). Using sections or undivided genera as my taxonomic units, I found that the major groups suggested by similarity of leaf characters corresponded remarkably well with higher taxa in the most recent classification of the subfamily (Webster 1975), which was based on characters more widely recognized as being of systematic value, particularly floral morphology, pollen morphology, and wood anatomy. There were, however, specific genera that did not cluster with other members of the tribes to which Webster assigned them. In many such cases, other data suggested that the classification was questionable and instead corroborated the leaf evidence. In the remaining cases, the source of the