The personality patterns of cancer patients as found in retrospective case-control studies are often suspected to be consequences of the disease. In this study an attempt was made to remove the bias arising from the disease itself by taking into account two indicator variables for the subject's anticipation of the subsequently established diagnosis. Seventy-five women with breast cancer were compared to 75 benign controls, matched in pairs for age and "reason for consultation" (the first indicator). Relative risks of 14 psychosocial scales were estimated in turn by logistic regression analysis for matched sets. The analysis showed 13 scales being positively or negatively related to cancer risk in accordance with the hypotheses. After adjusting for "fear of breast cancer" (the second indicator), five scales showed a significant association. In a previous report on this study, neither indicator variable was accounted for and the associations were generally found to be weaker, suggesting that they were masked by the malignant and benign subjects' differences in their degree of anticipation of a cancer diagnosis. As there were no a priori hypotheses regarding these indicator variables, the statistical significance of the results should be treated with caution.