We investigate the phylogeny, biogeography, time of origin and diversification, ancestral area reconstruction and large-scale distributional patterns of an ancient group of arachnids, the harvestman suborder Cyphophthalmi. Analysis of molecular and morphological data allow us to propose a new classification system for the group; Pettalidae constitutes the infraorder Scopulophthalmi new clade, sister group to all other families, which are divided into the infraorders Sternophthalmi new clade and Boreophthalmi new clade. Sternophthalmi includes the families Troglosironidae, Ogoveidae, and Neogoveidae; Boreophthalmi includes Stylocellidae and Sironidae, the latter family of questionable monophyly. The internal resolution of each family is discussed and traced back to its geological time origin, as well as to its original landmass, using methods for estimating divergence times and ancestral area reconstruction. The origin of Cyphophthalmi can be traced back to the Carboniferous, whereas the diversification time of most families ranges between the Carboniferous and the Jurassic, with the exception of Troglosironidae, whose current diversity originates in the Cretaceous/Tertiary. Ancestral area reconstruction is ambiguous in most cases. Sternophthalmi is traced back to an ancestral land mass that contained New Caledonia and West Africa in the Permian, whereas the ancestral landmass for Neogoveidae included the south-eastern USA and West Africa, dating back to the Triassic. For Pettalidae, most results include South Africa, or a combination of South Africa with the Australian plate of New Zealand or Sri Lanka, as the most likely ancestral landmass, back in the Jurassic. Stylocellidae is reconstructed to the Thai-Malay Penisula during the Jurassic. Combination of the molecular and morphological data results in a hypothesis for all the cyphophthalmid genera, although the limited data available for some taxa represented only in the morphological partition negatively affects the phylogenetic reconstruction by decreasing nodal support in most clades. However, it resolves the position of many monotypic genera not available for molecular analysis, such as Iberosiro, Odontosiro, Speleosiro, Managotria or Marwe, although it does not place Shearogovea or Ankaratra within any existing family. The biogeographical data show a strong correlation between relatedness and formerly adjacent landmasses, and oceanic dispersal does not need to be postulated to explain disjunct distributions, especially when considering the time of divergence. The data also allow testing of the hypotheses of the supposed total submersion of New Zealand and New Caledonia, clearly falsifying submersion of the former, although the data cannot reject the latter. © 2011 The Linnean Society of London, Biological Journal of the Linnean Society, 2012, 105, 92–130.