Molecular sequence analyses have yielded many important insights into diatom evolution, but there have been few attempts to relate these to the extensive fossil record of diatoms, probably because of unfamiliarity with the data available, which are scattered widely through the geological literature. We review the main features of molecular phylogenies and concentrate on the correspondence between these and the fossil record; we also review the evolution of major morphological, cytological and life cycle characteristics, and possible diatom origins. The first physical remains of diatoms are from the Jurassic, and well-preserved, diverse floras are available from the Lower Cretaceous. Though these are unequivocally identifiable as centric diatoms, none except a possible Stephanopyxis can be unequivocally linked to lineages of extant diatoms, although it is almost certain that members of the Coscinodiscophyceae (radial centrics) and Mediophyceae (polar centrics) were present; some display curious morphological features that hint at an unorthodox cell division mechanism and life cycle. It seems most likely that the earliest diatoms were marine, but recently discovered fossil deposits hint that episodes of terrestrial colonization may have occurred in the Mesozoic, though the main invasion of freshwaters appears to have been delayed until the Cenozoic. By the Upper Cretaceous, many lineages are present that can be convincingly related to extant diatom taxa. Pennate diatoms appear in the late Cretaceous and raphid diatoms in the Palaeocene, though molecular phylogenies imply that raphid diatoms did in fact evolve considerably earlier. Recent evidence shows that diatoms are substantially underclassified at the species level, with many semicryptic or cryptic species to be recognized; however, there is little prospect of being able to discriminate between such taxa in fossil material.