Objective.  The objective was to test, in a trial cohort of sedentary adults at risk of Type 2 diabetes, whether theory of planned behaviour (TPB) cognitions about becoming more physically active predicted objective and self‐reported activity levels and change.Design.  Participants of a randomized controlled trial underwent measurement at baseline, 6 and 12 months.Methods.  Participants (N= 365, 30–50 years) were recruited via their parent or family history registers at 20 general practices in the UK. Energy expenditure was measured objectively at baseline and 1 year. Participants completed questionnaires assessing physical activity and beliefs about becoming more physically active over the next year at baseline, 6 and 12 months.Results.  Between baseline and 12 months, objective energy expenditure in the cohort increased by an average of 20 minutes of brisk walking per day. Based on the 252 participants who provided complete data, affective attitude and perceived behavioural control consistently predicted intention, but intention and perceived behavioural control failed to predict physical activity levels or change (p‐values>.05).Conclusions.  Failure of the theory to predict behaviour and behaviour change may be due to inapplicability of the theory to this at‐risk population or to trial participation and intensive measurement facilitating behaviour change without affecting measured cognitions, or lack of correspondence between cognitive and behavioural measures. A wide range of potential personal and environmental mediators should be considered when designing physical activity interventions among at‐risk groups. High‐quality experimental tests of the theory are needed