A series of 80 patients underwent continuous electrocardiography by Holter monitoring (ECG-H) for 24 hours to detect myocardial ischaemia. Fifty five patients were not on anti anginal therapy. The results of ECG-H were compared with those of exercise electrocardiography (ECG-E) (33 cases) and coronary angiography (50 cases). The ECG-H was positive in 31 of 43 patients (72%) with clinical (5 patients) or angiographic (38 patients) signs of ischaemic heart disease. The ECG-H was negative in 11 out of 12 patients (92%) with normal coronary; angiography. The sensitivity and specificity of ECG-H (57% and 92%) were inferior to those of ECG-E (75% and 100%) in the 33 untreated patients undergoing all three investigations. Twenty five recordings were compared with the ECG-E to assess anti anginal therapy. In asymptomatic patients ECG-H showed pathological ST depression in 10 cases, the ECG-E being positive in 1 7 cases. Anginal chest pain was induced on ECG-E in 5 out of 7 cases with a positive ECG-E and negative ECG-H. The lower sensitivity of the ECG-H compared to the ECG-E is related to several factors: 1) the sensitivity of the ECG-E increases with the number of exploratory electrodes; 2) reduced levels of physical activity decrease the sensitivity; in false negative cases the heart rate on ECG-H was only 74 +/- 7% of that corresponding to the threshold of positivity of the ECG-E, compared to 97 +/- 16% of the threshold heart rate in true positives (p less than 0,001); 3) the sensitivity of the ECG-H and ECG-E depends on the severity and distribution of the coronary lesions; false negative results were commoner in single vessel disease (57%) than in double or triple vessel disease (24%) (p less than 0,01). Anginal pain during the test increased the sensitivity to 92%. The specificity of the ECG-H is partially dependent on the recognition of positional variations of the ST segment. These were observed in 10% of cases but were generally easy to distinguish by their beat-to-beat appearances. The satisfactory specificity of the ECG-H in this study is also related to the high incidence of coronary artery disease in the population under study (80%). The predictive value of a positive test (Bayes theorem) was 97%, but that of a negative test was only 41%.