The diagnostic accuracy of conventional exercise electrocardiography was compared with exercise algorithms specifically aimed at application in females, with R wave and heart rate adjusted ST-segment changes, and with discriminant analysis. A symptom-limited bicycle exercise test was performed by 189 females without previous infarction with a normal electrocardiogram (ECG) at rest. Frank lead ECG was computer processed. ST-segment amplitudes in lead X were superior to measurements in lead Y, but their sensitivity amounted only to 50% at a specificity of 90%. Correction for R wave amplitude marginally increased sensitivity. It was confirmed that leads with inferiorly directed vectors such as standard leads II, III, aVF or Frank lead Y should not be used for diagnostic purposes in females. Exercise-induced increases in R-wave amplitude of at least 0.1 mV occurred in only 8% of females with coronary disease. Previously described discriminant function using attained heart rate, ST-segment depression and a sex-dependent ST segment correction factor yielded a sensitivity of 60% at a specificity of 90%. Sensitivities of heart rate adjusted ST-segment changes and of a previously proposed discriminant function were greater than 70% at a specificity of 90%. It is concluded that the diagnostic yield of exercise testing in females can be improved by use of more sophisticated electrocardiographic and exercise variables.