The overwhelming majority of spiders are solitary and territorial. Of the handful of web-sharing social species, most belong to the cobweb genus Anelosimus Simon, 1891 (Theridiidae). Anelosimus species, especially those from the Americas, have therefore become model organisms in the study of spider sociality. However, lack of a phylogeny and outdated taxonomy have hindered progress in understanding the evolution of social behaviour. The identity of many species studied behaviourally is in doubt, and choice among the competing hypotheses on the course of evolution of sociality in Anelosimus requires a robust phylogeny. This paper offers a revision of the New World ‘eximius lineage’ containing the most intensely studied Anelosimus species, and a phylogenetic study including worldwide exemplars. Previous taxonomic work on the group was incomplete and oversimplified. Some species-level taxa, e.g. A. jucundus (O. P.-Cambridge, 1896) and A. studiosus (Hentz, 1850), as previously circumscribed represent a compendium of species and are here re-examined. Eight new species are here described: A. arizona, A. baeza, A. octavius and A. puravida, of the ‘jucundus group’, and A. gucamayos, A. oritoyacu, A. pantanal and A. tungurahua of the ‘studiosus group’. Furthermore, Enoplognatha dubia Chamberlin, 1916 and Theridion tosum Chamberlin, 1916 previously synonymized with A. jucundus, and Anelosimus fraternus Bryant, 1948, previously synonymized with A. studiosus, are here again considered valid. Enoplognatha dubia becomes a junior secondary homonym of Brattia dubia Tullgren, 1910 (= Anelosimus dubius) and the replacement name Anelosimus elegans Nomen Novum is here provided. The parsimony analysis of the morphological matrix (43 taxa, 147 characters) resulted in two equally most parsimonious trees, with four trichotomies in the strict consensus. Three of these lack character evidence to resolve them; one is a result of character conflict. One of the two trees is optimal under successive weighting. The New World Anelosimus are not monophyletic, but rather form three clades, the eximius lineage (20 species), the ‘rupununi group’ (two species) and the ‘ethicus group’ (six species). The phylogenetic results corroborate previous transfer of species to Kochiura and Selkirkiella. The following additional species are removed from Anelosimus: Styposis camoteensis (Levi, 1967) (comb. nov.), Styposis tepus (Levi, 1967) (comb. nov.), Chrosiothes episinoides (Levi, 1963) (comb. nov.) and Stemmops osorno (Levi, 1963) (comb. nov.). Four species are here treated as nomina dubia, Anelosimus nigrobaricus Barrion & Litsinger, 1995 (type in very bad condition, original description lacks sufficient detail for identification), and A. salaensis Barrion & Litsinger, 1995, Theridion fasciatum Holmberg, 1876 and T. sordidum Holmberg, 1876 (types lost, original descriptions lacks sufficient detail for identification). The results corroborate previous findings of convergent evolution of permanent sociality in the genus. However, instead of sociality evolving twice as previously suspected, the current phylogeny suggests no less than six, independent origins. Each time, the evolution of sociality seems to be responsible for a dramatic shift in population structure from outbred panmictic to strongly inbred subdivided populations. Perhaps as a consequence, once they are permanently social, species seem to fail to diversify; all social clades are small (one or two species) and usually smaller than their sister clade. No losses of social behaviour are inferred. The maternal care route hypothesis is again supported. To explain sociality in Anelosimus it seems sufficient to hypothesize a temporal extension of the juvenile web-sharing, co-operation and conspecific tolerance, displayed in basic maternal care, coupled with depression of dispersal. Given the most parsimonious phylogeny, the basal-most Anelosimus species occur in the Old World, and three Anelosimus clades occur in the New World. Sufficient data are not available to estimate the age of the Anelosimus lineage accurately, but the sparse fossil record hints at a relatively recent origin (20–40 mya). If true, vicariance could not account for this distribution; rather, the pattern may suggest three independent colonization events of Old World Anelosimus in the Americas. Support for most branches within Anelosimus is relatively low, especially the support for the relationships within species groups. Thus, although the forgoing conclusions are clearly implied by the phylogeny, weak support limits their force. © 2006 The Linnean Society of London, Zoological Journal of the Linnean Society, 2006, 146, 453–593.