Health-related information from multiple sources was collected on 334 women living in two predominantly white, affluent retirement communities near Los Angeles as part of a case-control study of a serious chronic disease (cancer of the breast) conducted in 1977-1978. There was no evidence on interview of cases preferentially recalling more drug use or past diseases than controls. Agreement between interview and medical record for all disease conditions studied as well as for height, weight, and most menstrual and reproductive variables was of the order of 90% or better. Age at last menstrual period as reported on inerview did not correspond particularly well with that recorded on the medical record. Agreement between data sources for ever/never drug use varied considerably with the type of drug studied, from a low of 69% for use of barbiturates and related drugs to a high of 87% for use of antihypertensive medications. Estrogen usage information was collected in detail. Better correspondence was observed between medical record and interview than between either medical and pharmacy records or interview and pharmacy records.