Ecological data on temporal trends suggest that the rising prevalence of obesity is, at least in part, attributable to declining population energy expenditure. However, population-level data on trends in physical activity are scarce. In longitudinal cohort studies individuals who report higher levels of leisure-time physical activity tend to be less likely to gain weight, but studies vary in their conclusions because of issues of confounding, reverse causality and measurement error. The majority of studies suggest that low levels of activity are only weakly associated with future weight gain. Questions about dose-response can only be properly addressed by studies including objective measures of activity with known measurement error. The observational studies leave uncertainties about the direction of causality, as individuals who are overweight are less likely to stay active. Adjustment for confounding can diminish the impact of known confounders, but only randomisation can deal with issues of unmeasured confounding. Although there are a large number of clinical trials on the treatment of individuals with obesity or the prevention of weight regain among weight losers, the updated review of trials to prevent weight gain de novo only reveals six trials published since 2000 in adults and eleven in children. Not only are these trials relatively few in number but, for various methodological reasons, they are uncertain in their conclusions about whether increasing activity will be effective in preventing obesity. Whilst efforts should continue to enhance the evidence base it is wise, in the meantime, to stick to the consensus public health advice of advocating 45-60 min moderate intensity activity daily to prevent obesity.