Pulse oximetry, combined with spirometry, was evaluated as a method of selecting chronic obstructive pulmonary disease (COPD) out-patients requiring definitive arterial blood gas analysis for long-term oxygen therapy (LTOT) assessment. A relatively high screening arterial oxygen saturation by pulse oximetry (SaO2) level was set, in order to maximize sensitivity. All 113 COPD out-patients attending the hospital clinic over a 6 month period were screened. Sixty had a forced expiratory volume in one second < 1.5 l and 26 had an SaO2 < or = 92%. These 26 underwent arterial blood gas analysis. Nine had an arterial oxygen tension < 7.3 kPa all with an arterial carbon dioxide tension (PaCO2) > 6 kPa. A further eight had a PaO2 < 8 kPa. This produced a sensitivity of 100% and specificity of 69% for oximetry in the detection of PaO2 < 7.3 kPa determined by direct arterial puncture and 100% and 86% respectively for detecting a PaO2 < 8 kPa. Although the poor specificity of oximetry in the crucial PaO2 range makes it unsuitable, when used alone, for prescription of LTOT, it may prove valuable in selecting patients who require further definitive arterial blood gas analysis.