Circadian clocks have evolved to synchronize physiology, metabolism and behaviour to the 24-h geophysical cycles of the Earth(1). Drosophila melanogaster's rhythmic locomotor behaviour provides the main phenotype for the identification of higher eukaryotic clock genes(2,3). Under laboratory light-dark cycles, flies show enhanced activity before lights on and off signals(4,5), and these anticipatory responses have defined the neuronal sites of the corresponding morning (M) and evening (E) oscillators(6,7). However, the natural environment provides much richer cycling environmental stimuli than the laboratory, so we sought to examine fly locomotor rhythms in the wild. Here we show that several key laboratory-based assumptions about circadian behaviour are not supported by natural observations. These include the anticipation of light transitions, the midday 'siesta', the fly's crepuscular activity, its nocturnal behaviour under moonlight, and the dominance of light stimuli over temperature. We also observe a third major locomotor component in addition to M and E, which we term 'A' (afternoon). Furthermore, we show that these natural rhythm phenotypes can be observed in the laboratory by using realistic temperature and light cycle simulations. Our results suggest that a comprehensive re-examination of circadian behaviour and its molecular readouts under simulated natural conditions will provide a more authentic interpretation of the adaptive significance of this important rhythmic phenotype. Such studies should also help to clarify the underlying molecular and neuroanatomical substrates of the clock under natural protocols.