We used digital filtering techniques and segmental analysis to dissect a series of respiratory variables into three components: (a) outlying values, including deep breaths or sighs; (b) random variation; (c) non-stationary baseline variation. 2. Records of about 30 min breathing were obtained from normal adults at rest and at 50 W exercise. 3. Deep breaths were defined as having a tidal volume greater than 2.5 sd above the mean. 4. We related these deep breaths to preceding trends in tidal volume and end-tidal partial pressure of CO2. 5. At rest, there was no relation between deep breaths and tidal volume, but the deep breaths were significantly clustered around the troughs in end-tidal partial pressure of CO2. 6. At 50 W exercise, there was no relation between deep breaths and end-tidal partial pressure of CO2, but the deep breaths were significantly clustered around tidal volume troughs. 7. Results obtained by pneumography were concordant with those obtained by using a mouthpiece to measure ventilation.
