OBJECTIVE Pulmonary embolism (PE) gives rise to alveolar dead space, which can be measured with a single breath test for CO2 (SBT-CO2). The characteristics of the SBT-CO2 are different in PE and other common conditions giving rise to alveolar dead space, notably airways disease. An analysis of alveolar dead space focusing on the late part of the breath (fDlate) has been suggested as a method for diagnosis of PE. Our aim was to evaluate this technique by comparison with lung scintigraphy. METHODS We randomly selected patients with clinical suspicion of PE. SBT-CO2 and lung scintigraphy were performed on the same day. The scintigraphies were reviewed and classified as high, intermediate and low probability of PE. RESULTS Out of 223 patients able to be evaluated, there were 20 of the high, 29 of the intermediate and 174 of the low probability category. There were large differences between the means of fDlate in the high and the intermediate and in the high and the low categories. We obtained a sensitivity of 85% and a specificity of 93% for diagnosis of PE, based on high and low probability categories. If a patient with previous PE, but no scintigraphic evidence of current PE, is excluded the sensitivity increases to 90%. CONCLUSIONS This study provides further support for the measurement of fDlate by the SBT-CO2 as a diagnostic test in patients with suspicion of PE. The test should be especially useful in small hospitals without access to pulmonary scintigraphy or pulmonary angiography.