Correcting spirometric indexes to BTPS assumes that the spirometer has a short cooling time constant. This assumption was challenged by testing a rolling seal spirometer (RSS), a water seal spirometer (WSS), and a heated pneumotachygraph (PT) with simulated spirograms from a computer-driven servo-controlled pump whose internal temperature could be regulated. When tested with ambient air (19 degrees C) the devices recorded forced expiratory volume in one second (FEV1) (4.0 L) to within 1%. With air at 37 degrees C the precision of the PT was little changed but the RSS and WSS (corrected to BTPS) overestimated FEV1 by 6 and 5%, respectively. The first moment of the spirogram (mean transit time) was 8 and 15% underestimated by the RSS and WSS, respectively, whereas the PT was precise to within 1%. These errors changed with ambient temperature and time constant of the spirogram. Clinical judgments concerning individual patients are unlikely to be seriously influenced by these errors. However, for the precise recording of spirograms required in epidemiologic and research work, and especially for studies on the shape of spirograms and flow-volume curves, these are important errors.