In an effort to assess the reliability of patient history in excluding pregnancy, we studied the correlation between specific historical factors and the presence of a positive qualitative serum beta-human chorionic gonadotropin assay. Two hundred eight patients were studied, and information was collected prospectively on a variety of historical criteria. Three historical variables were statistically less likely to be associated with pregnancy: last menstrual period that was on time, the patient thinking she was not pregnant, and the patient stating there was no chance she could be pregnant (P less than .001). There was, however, still at least a 10% chance of the patient being pregnant. Combinations of historical criteria were likewise unsuccessful at totally excluding pregnancy. These data support the contention that patient history is an unreliable method of excluding pregnancy in emergency department patients and supports the liberal use of pregnancy tests.