Generic limits and relationships in tribe Cereeae are reviewed and reassessed using cladistic methodology. Based on field and herbarium studies, and on reliable published observations, 20 characters are discussed in detail, polarized into plesiomorphic (primitive) and apomorphic (derived) states and then employed in the generation of a series of phylogenetic hypotheses using the PAUP computer program. Melocactus, Coleocephalocereus (incl. subg. Buiningia but excl. subg. Lagenopsis), Micranthocereus and Austrocephalocereus (excl. Espostoopsis dybowskii) represent a monophyletic group, here provisionally designated subtribe Melocactinae F. Buxbaum, and defined by the following synapomorphies: (1) loss of well-developed stem wood, (2) sunken-lateral or terminal cephalium, (3) nonblackening floral remains, and (4) derived fruit apex morphology (superficial floral remains). Uebelmannia, whose placement in tribe Cereeae is at best provisional, may also belong to this group, unless its similarities are due to convergence. Coleocephalocereus and Austrocephalocereus lack obvious autapomorphies distinguishing them from Melocactus and Micranthocereus, respectively. The superficial pseudocephalium present in most species of Micranthocereus may have been derived from the sunken-lateral type characteristic of this group. The remaining genera of the tribe, i.e. the Cereinae Britton & Rose, are more loosely related with relatively few synapomorphies. Pilosocereus sens. str. (incl. Pseudopilocereus) is clearly defined by the apomorphies of loss of well-developed stem wood (in parallel with the Melocactinae) and more or less depressed, dehiscent fruits (fruit dehiscence evolved in parallel in Cereus sens. lat.). ‘Lagenopsis’ (Cereus luetzelburgii) may have common ancestry with Stephanocereus, which is allied to Arrojadoa by the synapomorphic ring cephalium. Assessment of the precise relationship and appropriate classification (rank) of these three taxa, and a potentially related but inadequately known undescribed species, must await more comprehensive data, since at present it is not possible to be certain whether they have had a close common ancestry or are in part the product of convergence. Espostoopsis dybowskii is inadequately known, but does not appear to belong in Austrocephalocereus or Melocactinae; it may even be out of place in Cereeae (cf. Trichocereeae, especially Espostoa). Brasilicereus has no synapomorphies with, and various autapomorphies distinguishing it from, Cereus, and should be recognized as a genus. Cipocereus is defined by indehiscent fruits with colourless, watery pulp and should not be sunk into Pilosocereus. It is amplified to include Floribunda pusilliflora, Cereus crassisepalus and Pilosocereus bradei, besides C. minensis (C. pleurocarpus) and an undescribed species. Cereus sens. lat. (incl. subg. Ebneria, Mirabella minensis, Praecereus and Subpilocereus) is a primitive complex that cannot be resolved in terms of synapomorphies at present. Either of Cereus, Cipocereus or Brasilicereus—the genera with fewest synapomorphies—may be closest to the ancestor of the tribe. A key to the principal taxa of tribe Cereeae is provided.