Within the insect order Lepidoptera (moths and butterflies), the so-called nonditrysian superfamilies are mostly species-poor but highly divergent, offering numerous synapomorphies and strong morphological evidence for deep divergences. Uncertainties remain, however, and tests of the widely accepted morphological framework using other evidence are desirable. The goal of this paper is to test previous hypotheses of nonditrysian phylogeny against a data set consisting of 61 nonditrysian species plus 20 representative Ditrysia and eight outgroups (Trichoptera), nearly all sequenced for 19 nuclear genes (up to 14 700 bp total). We compare our results in detail with those from previous studies of nonditrysians, and review the morphological evidence for and against each grouping The major conclusions are as follows. (i) There is very strong support for Lepidoptera minus Micropterigidae and Agathiphagidae, here termed Angiospermivora, but no definitive resolution of the position of Agathiphagidae, although support is strongest for alliance with Micropterigidae, consistent with another recent molecular study. (ii) There is very strong support for Glossata, which excludes Heterobathmiidae, but weak support for relationships among major homoneurous clades. Eriocraniidae diverge first, corroborating the morphological clade Coelolepida, but the morphological clades Myoglossata and Neolepidoptera are never monophyletic in the molecular trees; both are contradicted by strong support for Lophocoronoidea + Hepialoidea, the latter here including Mnesarchaeoidea syn.n. (iii) The surprising grouping of Acanthopteroctetidae + Neopseustidae, although weakly supported here, is consistent with another recent molecular study. (iv) Heteroneura is very strongly supported, as is a basal split of this clade into Nepticuloidea + Eulepidoptera. Relationships within Nepticuloidea accord closely with recent studies based on fewer genes but many more taxa. (v) Eulepidoptera are split into a very strongly supported clade consisting of Tischeriidae + Palaephatidae + Ditrysia, here termed Euheteroneura, and a moderately supported clade uniting Andesianidae with Adeloidea. (vi) Relationships within Adeloidea are strongly resolved and Tridentaformidae fam.n. is described for the heretofore problematic genus Tridentaforma Davis, which is strongly supported in an isolated position within the clade. (vii) Within Euheteroneura, the molecular evidence is conflicting with respect to the sister group to Ditrysia, but strongly supports paraphyly of Palaephatidae. We decline to change the classification, however, because of strong morphological evidence supporting palaephatid monophyly. (viii) We review the life histories and larval feeding habits of all nonditrysian families and assess the implications of our results for hypotheses about early lepidopteran phytophagy. The first host record for Neopseustidae, which needs confirmation, suggests that larvae of this family may be parasitoids.