A canine model in which unilateral Streptococcus pneumoniae pneumonia was induced was used to evaluate the sensitivity, specificity, and risk of various technics proposed for the diagnosis of pulmonary infection. Control and infected animals were studied during spontaneous and mechanical ventilation. We found that three procedures provided a reasonable degree of combined specificity and sensitivity: transthoracic needle aspiration (TTN), catheter-brush biopsy (CBB), and transbronchial biopsy (TBB). Transtracheal aspiration, although sensitive, was consistently low in specificity, whereas TTN provided the highest sensitivity/specificity. Both CBB and TBB provided almost the same results in intubated animals, but specificity was lower in nonintubated animals. Gram-stained smears obtained by TTN, CBB, and TBB were highly specific, but often negative. The only risk encountered was pneumothorax. The TTN was associated with a 20 to 30% risk of pneumothorax; CBB and TBB did not appear to induce pneumothorax. These studies provide the first animal data regarding the comparative yield/risk of these diagnostic approaches. As such, they may prove useful in the design of future human investigations.