Clinicians continue to question the usefulness of microscopic examination of Gram-stained sputum specimens ("Gram staining") and sputum culture for diagnosis of pneumonia. We analyzed the sensitivity of these techniques in 105 patients with pneumococcal pneumonia proven by blood culture. Gram staining revealed gram-positive cocci in pairs and chains, and culture yielded pneumococci in only 31% and 44% of all cases, respectively. However, sputum specimens were never submitted for examination in 31 cases; in 16 others, the specimen was inadequate and a culture was not done. Excluding these cases, the sensitivities of Gram staining and culture were 57% and 79%, respectively. If patients receiving antibiotics for >24 h had been excluded, Gram staining would have suggested pneumococci in 63%, and culture results would have been positive in 86%. Sensitivity increased in inverse proportion to the duration of antibiotic therapy (P<.05). Microscopic examination of sputum samples before antibiotics were administered and performance of culture within 24 h of receipt of such treatment yielded the correct diagnosis in >80% of cases of pneumococcal pneumonia.