In the clinical setting, the observation of undisturbed spontaneous respiration and direct measurement of an "unbiased" respiratory rate is not a practical option. However, the respiratory modulation of the heart rate (respiratory sinus arrhythmia, RSA) offers a possibility of recording the respiratory rate indirectly. The present work introduces a novel procedure for extracting the respiratory rate from the ECG, and compares the results thus obtained with those of a previously applied method. The latter is based on the determination of dominant peaks in the amplitude spectrum of a five-minute RR interval time series. The new method presented here is based on the band pass filtering of a one-minute RR interval time series with subsequent analysis of the maxima, minima and zero crossings of the resulting RSA curve. The methods were compared by determining the respiratory rates of 16 healthy male subjects measured in 15 different experimental situations, including spontaneous and controlled breathing. The respiratory rate as determined by thermistor recording of nasal breath temperature was used as control. The mean deviation from directly measured respiratory rate was 2.9 min-1 for the new and 7.2 min-1 for the previous method. The new method proved superior under all conditions except controlled breathing at low frequencies. The present study thus demonstrates the high precision of respiratory rate determined indirectly by analysing the RSA curve. Since it requires no special instrumentation, this procedure is ideally suited for clinical application, and permits the continuous recording of respiratory rate over long periods of time.