Many cells in the auditory brainstem 'phase lock' to tone stimuli. From the changing phase relationship between the stimulus and the neural response in phase-locking cells, the delay between them can be estimated. This delay, however, is consistently greater than the latency measured in response to click stimuli, an important discrepancy. In this paper the different measures of delay, namely phase delay, group delay and signal-front delay are re-examined. An improved method for computing the average group delay is presented, which accounts for the cyclical nature of the phase data. Data were collected from units in successive processing sites of auditory pathway: the auditory nerve, the cochlear nucleus, the trapezoid body and the medial nucleus of the trapezoid body. Low-characteristic frequency (CF) units gave multimodal post-stimulus-time histograms in response to clicks, and showed stepwise decreases in latency with increasing intensity, with the appearance of earlier peaks in the response, rather than shifts in the timing of the peaks. The separation of peaks corresponded to the inverse of the unit's CF. High-CF units also showed a decline in click latency with intensity, but to a lesser degree than low CF units. We present an analysis which explains the difference between click latency and delay, and which in contrast to previous accounts is experimentally testable. We demonstrate that this new framework accounts for the discrepancy between the two measures of delay, and in addition accounts for the observed stepwise shifts in click latency for low-CF units.