Educative interventions to promote primary prevention of cancer (PPC) should be based on psychosocial models and be tailored to attitude. Attitude is a difficult variable to measure. The aim of this study was to draw up a questionnaire to evaluate the attitude to PPC of relatives of patients affected by cancer. An interview of 48 patients, who answered open questions, allowed 67 items to be drawn up. Three hundred and fifty-five people answered the self-administered questionnaire of 67 items (69 of them did it twice, with an interval of 2 weeks). The factorial analysis revealed five factors that explain the 36% total and the 115% common estimated variance. The Cronbach's &agr; oscillated between 0.80 and 0.93. The intraclass correlation coefficient was 0.7721 (95% CI (confidence interval) 0.6311–0.8592,P<0.0001). The analysis based on Samejima's model ordered the items according to their discriminating power and the difficulty/ease with which people agree with the most preventive option. The average scores obtained in both surveys of 69 people did not show significant statistical differences (P=0.195). The definitive questionnaire obtained has 63 items that measure attitude to PPC behaviour in a reliable and stable way.