OBJECTIVE: To review evidence as to the precision and accuracy of clinical examination for aortic regurgitation (AR). METHODS: We conducted a structured MEDLINE search of English-language articles (January 1966-July 1997), manually reviewed all reference lists of potentially relevant articles, and contacted authors of relevant studies for additional information. Each study (n = 16) was independently reviewed by both authors and graded for methodological quality. RESULTS: Most studies assessed cardiologists as examiners. Cardiologists' precision for detecting diastolic murmurs was moderate using audiotapes (kappa = 0.51) and was good in the clinical setting (simple agreement, 94%). The most useful finding for ruling in AR is the presence of an early diastolic murmur (positive likelihood ratio [LR], 8.8-32.0 [95% confidence interval [CI], 2.8-32 to 16-63] for detecting mild or greater AR and 4.0-8.3 [95% CI, 2.5-6.9 to 6.2-11] for detecting moderate or greater AR) (2 grade A studies). The most useful finding for ruling out AR is the absence of early diastolic murmur (negative LR, 0.2-0.3 [95% CI, 0.1-0.3 to 0.2-0.4) for mild or greater AR and 0.1 [95% CI, 0.0-0.3] for moderate or greater AR) (2 grade A studies). Except for a test evaluating the response to transient arterial occlusion (positive LR, 8.4 [95% CI, 1.3-81.0]; negative LR, 0.3 [95% CI, 0.1-0.8]), most signs display poor sensitivity and specificity for AR. CONCLUSION: Clinical examination by cardiologists is accurate for detecting AR, but not enough is known about the examinations of less-expert clinicians.