The genus Epithelantha (Cactaceae, Cactoideae, Cacteae) is native to Mexico and the southern USA and includes several uncertain species and ambiguous phylogenetic relationships. We applied multivariate analysis to a set of nine quantitative characters and six qualitative characters and we used Bayesian inference and maximum likelihood to reconstruct the phylogeny of Epithelantha using molecular data from four chloroplast regions (petL-psbE, psbA-trnH, trnL-F, and trnQ-rps16). Eleven taxa were collected in 39 localities where digital images of five individuals were taken. A correlation analysis allowed us to eliminate correlated characters. A discriminant canonical analysis (DCA) for quantitative characters and principal components analyses with mixed data (PCA mix) for quantitative and qualitative characters identified the length of the flower, number of spines, areole length/width ratio, the expansion of the hilar region of the seed, multicellular sculpture, and the relief of the periclinal wall as the most variable characters. The partial least squares-discriminant analysis (PLS-DA) allowed us to recognize nine of eleven taxa. As a complement, the molecular data matrix included 68 substitutions and 12 indels between the four markers. The 39 terminals analyzed were recovered in nine strongly supported clades, confirming the results of multivariate analyses. Epithelantha was recovered as a monophyletic group strongly supported by four substitutions and an indel, with Turbinicarpus being the sister group. Based on our results, we include a taxonomic synthesis of Epithelantha where we recognize ten species, propose two new combinations, designate seven lectotypes, and provide a key for species.