In an attempt to investigate the relative importance of endogenous and exogenous factors in determining food intake, 14 healthy subjects were studied while living in an Isolation Unit (where external time cues were absent) for eighteen 28 h "days" (equal to 21 solar days). The subjects were free to spend their waking time as they chose, and they had a free choice of what they ate and when they ate it. The only restrictions were that no naps were allowed in the "daytime," that some time was required to perform a variety of tests at regular intervals throughout the 18.67 h waking periods, and that any food preparation had to be performed by the subjects themselves. Core (rectal) temperature and activity were monitored throughout, and the subjects answered a questionnaire on their eating habits at 3 h intervals during the waking periods. The questionnaire investigated reasons for eating or not eating a meal during the previous 3 h and, if a meal had been eaten, its type, the factors influencing that choice, and the subjects' subjective responses (hunger before, enjoyment during, and satiety after) to it. The results were analyzed (two-way ANOVA) in terms of both the imposed day length (the exogenous component) and the free-running period of the temperature rhythm (the endogenous component). Results indicated that by far the main reason for eating/not eating was hunger/lack of hunger rather than factors such as food availability and time-pressure. There were statistically significant effects of time within the imposed waking periods upon the type of meal eaten--"breakfast" tending to be a snack, "lunch" a small hot meal, and the "evening meal" a large hot meal. Hot meals (whether small or large) were associated with more hunger before the meal, more enjoyment of the meal, and a greater degree of satiety afterward than were cold meals. These effects suggest that the individuals adjusted their eating habits to fit in with the imposed wake times. By contrast, the effect of circadian phase upon food intake, the type of meal eaten, and subjective responses to the meal was much weaker, and either statistically nonsignificant (P > 0.10) or only marginally so (0.10 > P > 0.05). For example, a large hot meal was chosen as readily for an "evening meal," and subjective responses to it were the same, at whatever circadian phase it was eaten. We conclude that food intake during forced desynchronization is dominated by the waking schedule rather than by circadian influences; some of the implications of these findings when eating habits and the metabolism of food are concerned, particularly in night workers, are considered briefly.