Alcohol intoxication seems to be an important contributing factor to cyclist injury and death in traffic. The first aim of this study was to examine how much of the variance in intention to cycle when mildly alcohol intoxicated during the coming year could be explained by only three direct measures of attitude, subjective norm, and perceived behavioural control, and by indices constructed from the belief-based measures, respectively. The second aim was to identify the belief composites that contributed to the prediction direct measures of attitude, subjective norm, perceived behavioural control, and intention, regarding cycling when mildly alcohol intoxicated during the coming year. A sample of cyclists (n = 196) completed a postal questionnaire including both direct and belief-based measures of the latent variables in the theory of planned behaviour. The results indicated that the three direct measures of attitude, subjective norm, and perceived behavioural control contributed more to the prediction of drivers’ intention to cycle when alcohol intoxicated than did the 33 indices constructed from belief-based measures. This suggests that three items is sufficient to predict peoples intention. Standard multiple regressions then showed which of the belief composites contributed to the prediction of attitude, subjective norm, and perceived behavioural control, as well as intention. These identified beliefs could be used to, for example, develop effective road safety campaigns, but it is important to remember that these beliefs might not be directly transferable to other groups of people and that all campaigns must be based on the beliefs of their specific target audience.