The stethoscope is used as first line diagnostic tool in assessment of patients with pulmonary symptoms. However, there is much debate about the diagnostic accuracy of this instrument. This meta-analysis aims to evaluate the diagnostic accuracy of lung auscultation for the most common respiratory pathologies. Studies concerning adult patients with respiratory symptoms are included. Main outcomes are pooled estimates of sensitivity and specificity with 95% confidence intervals, likelihood ratios (LRs), area under the curve (AUC) of lung auscultation for different pulmonary pathologies and breath sounds. A meta-regression analysis is performed to reduce observed heterogeneity. For 34 studies the overall pooled sensitivity for lung auscultation is 37% and specificity 89%. LRs and AUC of auscultation for congestive heart failure, pneumonia and obstructive lung diseases are low, LR- and specificity are acceptable. Abnormal breath sounds are highly specific for (hemato)pneumothorax in patients with trauma. Results are limited by significant heterogeneity. Lung auscultation has a low sensitivity in different clinical settings and patient populations, thereby hampering its clinical utility. When better diagnostic modalities are available, they should replace lung auscultation. Only in resource limited settings, with a high prevalence of disease and in experienced hands, lung auscultation has still a role.