The extent and genesis of uncertainty of gestation has been studied in a total obstetric population, using in all cases a best clinical estimate of gestation made according to a specific protocol. In 73.2% of patients, the estimate was assessed as certain, in 19.7% as approximate, and in 7.1% as uncertain. Uncertainty assessed in this way has several antecedents, and coding of only of these (presence or absence of a precise date for the last menstrual period) is insufficient in data to be used for research in which gestation length is to be used as a variable.