In order to analyze factors that influence an interviewer's judgement of the validity of responses given by patients on the duration of their neoplastic signs and symptoms, 183 consecutive symptomatic patients hospitalized for a digestive tract neoplasm were personally interviewed. The validity of the answers was judged by the interviewers to be high in 156 cases (85%), and low in 27 (15%). The subjective validity of the interview (SVI) was inversely related to the time elapsed from first medical symptom to interview (TFMSI), even after adjusting for the duration of the interview (p < 0.05). SVI was not influenced by whether patient and interviewer agreed on the first symptom. SVI was inversely related to educational level (p < 0.01) and to occupational class (p = 0.04). Patients whose Karnofsky's Index (KI) was > or = 80 were over twice as likely to yield valid responses (TFMSI-adjusted odds ratio [OR] = 2.82, p = 0.037). Multivariate analyses selected education, TFMSI and KI as independent predictors of the interviewer assessment. The SVI of patients admitted to the hospital through the Emergency Department was lower than that of subjects whose admission was planned (OR = 6.49, p = 0.005). In this study SVI related in a logical manner to the characteristics of the interview, of the subjects and of their clinical course. It hence appeared to reasonably estimate the validity of data collected. Identifying factors that affect the reliability of patients' responses would help increase the validity of studies on the duration of cancer symptoms.