OBJECTIVE: To examine the influence of an individually derived versus modal subjective norm component of the Theory of Planned Behaviour (TPB) in predicting attendance at breast screening. DESIGN: A prospective, longitudinal design was employed using a postal questionnaire at Time 1 and objective attendance data from screening records at Time 2. METHOD: Questionnaires were sent to 1000 women from East Kent due to be invited for X-ray mammography under the UK's National Health Service Breast Screening Programme. The TPB constructs were used to predict attendance, with a focus on subjective norm. Two versions of the questionnaire were distributed: one assessed subjective norm using individually generated normative beliefs (individual condition) and the other provided a list of modal referents (modal condition). RESULTS: The mean number of normative referents in the individual condition was just two, almost all of whom were family members, in contrast to the modal condition where 'GP' led the rank table. The results suggest that a more senstive measure of subjective norm is attained using individually derived beliefs. Structural equation modelling revealed that, contrary to theoretical prediction, the effect on attendance of the individually derived subjective norm was not entirely mediated by intention: an independent contribution to attendance was shown. In the modal condition the construct was entirely mediated by intention. CONCLUSIONS: For health behaviours that have implications for others as well as self, measuring individually salient normative beliefs may both improve theoretical prediction and yield information of potential value for designing intervention programmes.