Women themselves are often the most convenient source of information regarding their pregnancy and birth outcomes such as prematurity. However, the ability of mothers to recall birth information and the accuracy of information they do recall has rarely been assessed. This study used a mail survey of women who delivered either term or preterm infants in Rochester, Minnesota, between 1980 and 1985. Maternal reports of circumstances and outcome of delivery were compared to data abstracted from the medical records. Maternal ability to recall and report events varied from 99.5% for smoking during pregnancy to 85% for infant's gestational age at birth. Agreement between medical record and maternal report was very high for perinatal events such as delivery by cesarean section, preexisting diabetes, and smoking. Percent negative agreement was quite high when comparing information on rare events such as placental abruption. Recall of gestational age was better for women delivering preterm infants but uncommon in all women. Maternal reports of perinatal events in which they directly participated such as cesarean section can be accurately and reliably reported 10 to 15 years after the birth. Gestational age is reported less accurately and with different rates of recall for mothers with term and preterm deliveries.