Patient-rated questionnaires are increasingly used to assess health-related quality of life. We studied one aspect of the validity of such measures that has rarely been investigated do patients interpret questionnaires in the same way as do the researchers reporting the results? If not, there may be a problem. We employed the EORTC QLQ-C30 quality-of-life questionnaire to study 95 cancer patients and measured the agreement between (1) the patient's self-assessment and (2) an observer's rating of the patient's open-ended responses to the same questionnaire administered as an interview. The observer made qualitative recordings describing potential misinterpretations. The agreement between patients' and observers' ratings was high (median kappa = 0.85, range 0.49-1.00). The qualitative data revealed a few minor validity problems. One of these, selective reporting, may lead to systematic errors; some patients reported only what they considered "relevant" symptoms. The combination of quantitative and qualitative methods proved useful for questionnaire validation.