SummaryThe adaptation of the various mammals to all gradations between a totally diurnal to a totally nocturnal pattern of activity is noted and the nature of the determinants of the high proportion of species which are crepuscular or nocturnal is discussed. It is seen that detailed study of the activity patterns has been confined to a small proportion of species, principally some of the smaller rodents. Ruminants are unusual in remaining vigilant during normal sleep, loss of awareness of external events being confined to periods of paradoxical sleep which occupy a very small proportion of the 24 hours.The 24 hour rhythm of activity has an endogenous basis. Under normal conditions it is entrained by the changes in light intensity at dusk and dawn, but circadian activity rhythms can be entrained to periodicities longer or shorter than 24 hours. Relatively small variations in the intensity of the lighting between day and night are effective in maintaining entrainment, but under arctic conditions in summer the 24 hour rhythm becomes much less pronounced and activity may even develop a free–running pattern.A short–term rhythm with a regular periodicity of a small number of hours has been reported to occur by many authors in species of small eutherian mammals feeding on a bulky diet, that is insectivores and the more herbivorous of the small rodents. A summary account is given of information, much of it derived from unpublished research, indicating that in fact the patterns of diel activity of such rodents vary greatly with the environmental conditions. The duration of daylight is apparently only one of many factors affecting the distribution of their activity over the 24 hours. It is noted that versatility of this nature also occurs in some large mammals.It is concluded that an advance in knowledge of the mechanisms determining the activity patterns of mammals will require a greater emphasis on field observations of mammal behaviour and on laboratory experiments in which the variety of conditions found in nature is closely simula