1. To investigate how intrathoracic pressure affects left ventricular function during normal breathing and positive-pressure ventilation, beat-by-beat measurements of left ventricular stroke volume (LVSV; pulsed Doppler ultrasound), heart rate and blood pressure were made in five normal subjects breathing actively and then undergoing passive positive-pressure ventilation. 2. To assess the within-breath effects of positive-pressure ventilation on left ventricular function in the clinical setting, further measurements were made in six patients sedated, paralysed and ventilated because of brain injuries, who had no disease of the heart or lungs. 3. In the normal subjects, there were minimal phasic variation in LVSV during positive-pressure ventilation with the subject passive. Heart rate and blood pressure also stayed relatively constant throughout the ventilator cycle. During active breathing at the same depth and rate, there were large phasic respiratory variations in LVSV, with the lowest values occurring during inspiration. 4. In the paralysed and unconscious patients, an increase in LVSV was associated with the increase in airway pressure which occurred during lung inflation; however, the phase lag between the rise in airway pressure and the rise in LVSV varied widely between patients. These changes occurred whether or not sinus arrhythmia was present. 5. Addition of 4.6 +/- 1 cmH2O of positive end-expiratory pressure (PEEP) did not increase the within-breath LVSV variations, but resulted in a mean depression of LVSV of 5.4% (paired t test, P = 0.035). 6. The smaller variations in LVSV during positive-pressure ventilation compared to normal breathing in the conscious subjects may reflect (a) the smaller magnitude of positive, compared to negative, pleural pressure excursions which accompany a given tidal volume, and (b) an asymmetry between the effects of positive and negative pleural pressure on the heart. 7. The prominent effects of positive-pressure ventilation on LVSV in unconscious patients, compared to the minimal effects seen in ventilated normal subjects, may result from reduced lung compliance and a degree of pulmonary vascular congestion in the patients which was undetectable clinically or radiologically.