1. The responses of units in the ventral cochlear nucleus in anesthetized guinea pigs have been measured to best-frequency tones, noise bands geometrically centered around the unit best frequency, and noise bands asymmetrically positioned around the best frequency. 2. Each unit isolated was characterized using peristimulus time histograms (PSTHs) to best-frequency tones at 20 and 50 dB suprathreshold, frequency-intensity response areas and rate-versus-level functions in response to best-frequency tones and wideband noise. The data reported here are derived from full analyses of 5 chopper units and 17 onset units. The onsets were divided into onset-I (OnI), onset-L (OnL), and onset-C (OnC) by the criteria described by Winter and Palmer: the PSTHs of OnI units show only an onset response, OnL units respond with a single spike at onset followed by a low level of sustained activity, and OnC units have PSTHs with one to four onset peaks and low levels of sustained discharge. 3. In response to geometrically centered noise bands of constant spectral density, the discharge of chopper units and one OnI unit increased over a relatively narrow range of bandwidths, corresponding to the equivalent rectangular bandwidth calculated from their response area, and then became constant. In contrast, OnL and OnC units showed increases in discharge rate with noise bandwidth over very wide ranges of bandwidth. The growth of the discharge rate with noise bandwidth was approximately linear on double logarithmic axes and therefore could be described by a power function with an exponent of 0.37. This relation held even for noise levels near threshold. 4. When noise bands with constant spectral density (at the input to the earphone) were presented with one edge fixed at the unit's best frequency, the discharge rate of most chopper units and the one OnI unit increased over a narrow range of bandwidths and then became constant. This pattern was observed irrespective of whether the second edge of the noise was progressively increased above, or decreased below, the best frequency. For two of the chopper units, in which lateral inhibitory sidebands could be demonstrated, increasing the noise bandwidth led first to increases and then to decreases in the discharge rate as the noise energy impinged upon the sideband. The chopper units act like energy detectors with a filter corresponding to their single tone response area, but, for some units, with the addition of inhibitory sidebands. 5. For the OnL and OnC units, increasing the noise bandwidth above or below best frequency caused progressive increases in the discharge rate over wide ranges of bandwidth. These increases occurred even for low noise spectral densities. The growth in discharge rate for these onset units was well fitted at all spectral density levels by power functions: one above best frequency and one below. At levels of the noise 40 dB above the unit threshold, the point at which the discharge rate reached 90% of its maximum was, on average, about 2 octaves below best frequency and 1 octave above. For some onset units, changes in the discharge rate were seen as the noise bandwidth was varied over about 14 kHz, which is about one-third of the total frequency hearing range of the guinea pig. 6. The data for onset units is consistent with the hypothesis that onset units in the ventral cochlear nucleus achieve their precision in the temporal domain by integration of the inputs from auditory nerve fibers with a wide range of best frequencies. The range of frequency over which onset units integrate frequency matches that of the inhibitory input to dorsal cochlear nucleus neurons, suggesting a possible role as an inhibitory interneuron.