PIP: In a case-control study of low birth weight deliveries, comparison of hospital records with interviews conducted soon after fails to confirm the widely held view that concurrently recorded medical notes are superior to retrospective interviewing. This paper compares the information from interviews with women who had recently given birth with that derived from a detailed examination of their medical records, taking into account the internal inconsistencies that might emerge in the latter source. The study population included all births within Enfield and Haringey Health Districts during 2 time periods between 1980 and 1982. 122 singletons of low weight were selected, with an increased proportion in the lowest weight categories, plus a control group of 61 singletons in the birth weight range 3001-4000 g., and a group of 12 twin deliveries selected irrespective of birth weight. Although more episodes of bleeding occurred to mothers of low birth weight babies, the rate of disagreement did not differ according to whether mothers had babies of low birth weight (12%), of normal birth weight (15%), or twins (17%). Other effects for mothers of low birth weight babies are: 1) more episodes of vaginal discharge; 2) there were more disagreements in the low birth weight group, in which the mother's answer to the order of events disagreed with the hospital records, compared to the normal birth weight group; and 3) all 14 cases of diagnostic disagreement occurred in the low birth weight group. Overall, there was very good agreement between hospital records and the interview, although mothers were slightly more likely to under report abortions, especially induced ones. A possibility of bias exists in the use of hospital records for epidemiological research into outcomes associated with low birthweight. More generally, there is strong ground for disputing the assumption that concurrent recording in hospitals is always superior to the so-called "soft" information obtained retrospectively by interview.