The Giant anteater (Myrmecophaga tridactyla) is one of the most specialized mammalian predators, subsisting exclusively on a diet of ants and termites. This study examines the feeding and food preference of captive and wild Giant anteaters in central Brazil. Eight species of mound‐building termites were offered to anteaters in the Brasilia Zoo. The anteaters showed marked preferences for some species over others. The overall pattern of preference did not correlate with prey size, prey nutritional quality or the soldier‐based defence type of the different prey species.Preferences of wild anteaters in Emas National Park, Goias State, Brazil for an equivalent set of termite species were then determined. The preferences of wild animals did not correlate with the preferences of captive ones. This discrepancy results principally from the fact that the conditions under which anteaters encountered prey in the captive trials were very different from the conditions under which prey were encountered in the wild. A Giant anteater incorporates three major factors in its decision to feed on a particular termite species: nutritional value, availability (or mound defence) and response to attack.The very short feeding periods that typify anteater feeding are caused by the rapidly decreasing prey value of the social insect colony following predator attack. The different types of defence exhibited by the different prey species all serve to cause this rapid decrease and thus limit, but do not prevent, predation.SummaryThe purpose of this study was to investigate the relationship between the Giant anteater as a predator and termites as prey. Giant anteaters have long been known to eat only ants and termites, but little work has been done on which species are consumed and why those are chosen over others. In two‐way choice trials the present author offered captive Giant anteaters eight species of central Brazilian mound‐building termites. Myrmecophaga preferences were determined by the amount of time spent feeding on each termite species. The preferences of captive animals were then compared with the preferences of wild anteaters in Emas National Park, Brazil. Myrmecophaga showed marked preferences for some species of termites, both in captivity and in the wild. However, the ranking of prey species differed between the two situations. Most dramatically, one genus, Velocitermes, which was least‐preferred in captivity, was most‐preferred in the wild. By analysing the differences between wild and captive animals, it has been shown that three major factors seem to influence Giant anteaters food preferences: (1) the type of soldier‐based defence; (2) the type of mound defence or availability and (3) the nutritional value. Each of these factors can serve as a defence against anteater predation, and the different termite species incorporate these defences in different ways with different emphases.Giant anteaters feed on all prey species in a similar way: by feeding for short periods of time at many different locations. This pattern of feeding is forced on the predator by the nature of its prey defences, and is not a form of prudent predation as has been previously suggested.I would like to thank the following organizations for funding my research: Organization of American States, National Geographic Society, Harvard University, Sigma Xi and Friends of the National Zoo. The Brazilian Institute of Forestry Development granted me permission to work in Emas Park and the Director, Heber Silva de Oliveira and Assistant Director, Antonio Malheiros da Cruz, provided me with help while in the Park. Ms Sonia Pereira provided invaluable service in helping me follow anteaters. The members of the Order of Saint Benedict in Mineiros, and especially Father Eric Deitchman, helped me greatly. The Brasilia Zoo, and particularly Raimundo David Monteiro Lima, kindly gave me access to their animals and greatly aided my work. The Laboratories of Zoology and Ecology at the University of Brasilia provided working space and support; I would particularly like to thank Drs Cleber J. R. Alho, Tony Raw, Helen Coles de Negret and Tom E. F. Lacher. Dr Alan Mill kindly shared his knowledge of Brazilian termites and identified the termites from Emas Park, while Dr E. O. Wilson helped me with the ants.Dr Lee‐Ann Hayek and Jim Craig helped me with statistical analyses; Ms V. Garber and Ms B. L. Stanton typed the manuscript. This manuscript was greatly improved by J. Shaw, K. Sebens, A. W. Crompton, B. Holldobler, N. T. Wheelwright, S. D. Thompson, J. F. Eisenberg and, particularly, P. L. Shaw and J. G. Robinson. I would also like to thank my parents, R. H. Redford and L. C. Redford, and Peter Eisner for helping me to recover and continue.