Cultivation of fungus for food originated about 45-65 million years ago in the ancestor of fungus-growing ants (Formicidae, tribe Attini), representing an evolutionary transition from the life of a hunter-gatherer of arthropod prey, nectar, and other plant juices, to the life of a farmer subsisting on cultivated fungi. Seven hypotheses have been suggested for the origin of attine fungiculture, each differing with respect to the substrate used by the ancestral attine ants for fungal cultivation. Phylogenetic information on the cultivated fungi, in conjunction with information on the nesting biology of extant attine ants and their presumed closest relatives, reveal that the attine ancestors probably did not encounter their cultivars-to-be in seed stores (von Ihering 1894), in rotting wood (Forel 1902), as mycorrhizae (Garling 1979), on arthropod corpses (von Ihering 1894) or ant faeces in nest middens (Wheeler 1907). Rather, the attine ant-fungus mutualism probably arose from adventitious interactions with fungi that grew on walls of nests built in leaf litter (Emery 1899), or from a system of fungal myrmecochory in which specialized fungi relied on ants for dispersal (Bailey 1920) and in which the ants fortuitously vectored these fungi from parent to offspring nests prior to a true fungicultural stage. Reliance on fungi as a dominant food source has evolved only twice in ants: first in the attine ants, and second in some ant species in the solenopsidine genus Megalomyrmex that either coexist as trophic parasites in gardens of attine hosts or aggressively usurp gardens from them. All other known ant-fungus associations are either adventitious or have nonnutritional functions (e.g., strengthening of carton-walls in ant nests). There exist no unambiguous reports of facultative mycophagy in ants, but such trophic ant-fungus interactions would most likely occur underground or in leaf litter and thus escape easy observation. Indirect evidence of fungivory can be deduced from contents of the ant alimentary canal and particularly from the contents of the infrabuccal pocket, a pharyngeal device that filters out solids before liquids pass into the intestine. Infrabuccal pocket contents reveal that ants routinely ingest fungal spores and hyphal material. Infrabuccal contents are eventually expelled as a pellet on nest middens or away from the nest by foragers, suggesting that the pellet provides fungi with a means for the dispersal of spores and hyphae. Associations between such "buccophilous" fungi and ants may have originated multiple times and may have become elaborated and externalized in the case of the attine ant-fungus mutualism. Thus, contrary to the traditional model in which attine fungi are viewed as passive symbionts that happened to come under ant control, this alternative model of a myrmecochorous origin of the attine mutualism attributes an important role to evolutionary modifications of the fungi that preceded the ant transition from hunter-gatherer to fungus farmer.