The authors of the theory of reasoned action (TRA) and its extension, the theory of planned behaviour (TPB) recommend that researchers who use these theories to investigate the determinants of a given behaviour should first conduct an elicitation study to identify the modal salient beliefs in the target population. In spite of the importance accorded to salient beliefs by the TRA/TPB, the elicitation stage has received little research attention. This paper reports a detailed analysis of beliefs about "being more physically active in the next 12 months." A general population sample of 213 adults completed a questionnaire while attending a research centre for a series of tests. The findings showed that the beliefs that were elicited by questions designed to prompt affective outcomes (like or enjoy, dislike or hate) differed systematically from those that were elicited by the traditional questions designed to prompt instrumental outcomes (advantages and disadvantages). Whether this resulted in different final sets of modal salient beliefs was found to depend on the particular decision rule that was employed. An alternative decision rule is proposed, based on maximizing the degree of overlap between the modal set and the full set of salient beliefs generated by the sample. The index of overlap can be used to gauge the adequacy of using a modal set of a given size to represent the salient beliefs of the whole sample. In the current dataset, the optimal modal set for "advantages and disadvantages" was associated with only 26 percent overlap with the salient beliefs of the whole sample, which was judged to be insufficient. In such cases, a better strategy may be to ask participants to generate and rate their own beliefs.