When one analyzes transvalvular and venous flow velocity patterns, it is important to relate them to respiration. For this reason a nasal thermistor technique is often used, although it is known that this signal is delayed in relation to intrathoracic pressure changes. The magnitude and variation in delay have not been investigated previously and were, therefore, studied in a model experiment in 10 normal subjects, in 10 patients with obstructive, and in 10 patients with restrictive pulmonary disease. Esophageal pressure variations measured with an air-filled balloon served as a gold standard for intrathoracic pressure changes. During basal conditions there was, for both patient groups and normal subjects, a considerable delay of the thermistor signal. The average delay for all subjects was 370 msec with a wide variation (from 120 to 720 msec). At higher breathing frequencies the delay shortened to 310 msec (P < 0.01) but there was still a wide variation (ranging from 200 to 470 msec). Theoretic calculations show that the delay caused by the respiratory system accounts for only a minor portion of the total delay. Model experiments confirmed that the response characteristics of the thermistor probes limit the accuracy in timing of respiration. The total delay with the investigated thermistor technique is too long and variable to fulfil clinical demands.