The statistical tests used in these analysesvaried from very simple two-group t tests orbivariate correlations to complex multiple-regression and structural equation models.Because some of the variables areintercorrelated and issues of mediation,moderation, suppressor effects, andinteraction terms are likely, the mostsophisticated statistical model appropriateto the data should be used. Humanbehaviour is likely very complex; therefore,the presence of interaction terms seems tobe likely, and these terms should be sought.Although each model has something tooffer obesity prevention, the early promisecan be achieved only with substantialadditional research in which these modelsare applied to diet and physical activity inregard to obesity. The most promisingavenues for such research seem to beusing the latest variants of the Theory ofPlanned Behaviour and Social Ecology.Synergy may be achieved by taking themost promising concepts from each modeland integrating them for use with specificpopulations. Biology-based steps in aneating or physical activity event areidentified, and research issues aresuggested to integrate behavioural andbiological approaches to understandingeating and physical activity behaviours.Social marketing procedures have much tooffer in terms of organizing and strategizin