Although interview information is usually the sole source of data in case-control studies, the accuracy of such data is infrequently assessed. We compared interview data on selected medical conditions and surgical procedures with medical records of subjects with chronic lymphocytic leukemia. We examined agreement by type of respondent (self or surrogate), age, sex, race, and type of hospital. The strength of agreement between the two data sources (as measured by kappa statistics) was substantial kappa greater than 0.6) for splenectomy, appendectomy, asthma, and systemic lupus erythematosus; moderate kappa greater than 0.4) for tonsillectomy/adenoidectomy, tuberculosis, diverticulitis, hepatitis, rheumatic fever, and drug allergy; and poor kappa less than 0.3) for chronic bronchitis, chronic sinusitis, psoriasis, rheumatoid arthritis, and most other types of allergy. In general, self respondents had more accurate recall than surrogate respondents. Among self respondents the strength of agreement tended to be greater for males than females, for whites than blacks, and for subjects from referral hospitals than for community hospitals. No consistent patterns were apparent by age. Despite a number of limitations, the findings of the study provide an addition to the scant epidemiologic literature on this topic, and suggest that for certain conditions medical record data collection may be needed to supplement interview information.