This study aimed to investigate mothers' reporting of the nature, location, frequency and content of health care visits for their infants, as compared with data abstracted from the infants' medical records. It was part of a community-based parenting intervention designed to improve preventive health care utilisation among minority mothers in Washington, DC. Mothers >or=18 years old with newborn infants and with poor or no prenatal care were enrolled in the study. A total of 160 mother-infant dyads completed the 12-month study. Mothers were interviewed when the infants were 4, 8 and 12 months old, and were asked to recall infant visits to all health care providers. Medical records from identified providers were used for verification. The number and type of immunisations given, types of providers visited, and reason for the visits were compared. Only about a quarter of mothers agreed with their infants' medical records on the number of specific immunisations received. The mothers reported fewer polio (1.8 vs. 2.1, P = 0.006), diphtheria and tetanus toxoids and pertussis (DTP) (1.8 vs. 2.2, P = 0.002), and Haemophilus influenzae type b (HiB) (1.3 vs. 2.1, P < 0.0001) immunisations than were recorded. Similarly, about a quarter of the mothers were unaware of any polio, DTP or hepatitis B immunisations given, as documented in the medical records, and 38% did not know that their infant was immunised for HiB. Nearly half of the mothers recalled more infant doctors' visits than were recorded in the medical records (4.1 vs. 3.6 visits, P = 0.017). The mothers generally disagreed with the providers about the reason for a particular visit and reported fewer sick-baby visits (1.5 vs. 3.3, P < 0.0001) than the providers recorded. Mothers' reports and medical records matched in only 19% of the cases. In 47%, mothers under-reported and in 34% over-reported the total number of visits. The strongest agreement between mothers' reports and medical records was in the case of emergency room visits (63%). In conclusion, in this population, mothers' reporting did not match that of providers with respect to specific information: the number of immunisations, the location where services were provided, and the classification of sick- vs. well-baby visits. Future studies that evaluate health care utilisation data should take these discrepancies into consideration in their selection of information source, and in their interpretation of the data.