To investigate the high frequency oscillations observed in the inspiratory activity of respiratory motor nerves of decerebrate cats, we applied a signal processing technique, power spectral analysis, to the electrical activity of the phrenic and recurrent laryngeal nerves. We found two peaks in the phrenic nerve power spectral densities, one at 88.1 +/- 6.4 Hz (mean +/- S.D.) and the other at 37.1 +/- 9.7 Hz, and two peaks for the recurrent laryngeal nerve, at 87.4 +/- 10.1 Hz and at 55.4 +/- 5.1 Hz. We identified 3 factors affecting the peaks. Anesthetics reduced or eliminated the 88 Hz peak and produced new low frequency peaks in the phrenic and recurrent laryngeal nerves. Increasing end-tidal CO2 decreased the bandwidth of the 88 Hz peak and increased its amplitude relative to that of the low frequency peak. Decreasing body temperature from 38 to 30 degrees C reduced the frequency of the 88 Hz peak by 5.0 Hz/degrees C. The power spectral density of the phrenic nerve activity differed from that of the recurrent laryngeal nerve activity because the single fibers in each nerve had different power spectral densities. About 70% of the fibers recorded in a nerve had power spectral densities similar to that of the whole nerve. A minority of the phrenic nerve fibers had the same low spectral peak as the recurrent laryngeal nerve, and conversely, a minority of the recurrent laryngeal fibers had the same low spectral peak as the phrenic nerve. Bilateral removal of the dorsal respiratory group eliminated the high frequency peak in the power spectral density of the phrenic nerve and the peripheral reflexes, but rhythmic bursts of inspiratory activity remained. From these findings we hypothesized that there are two central respiratory pattern generators in the brain stem with parallel pathways to the respiratory motoneurons.