1. In order to investigate the possible neural mechanisms underlying delay-dependent facilitation in the bat's auditory cortex (18), the responses to single FM pulses of varying amplitude were examined. Analysis of amplitude-spike count functions revealed three distinct types: monotonic, simple nonmonotonic, and complex nonmonotonic. The complex nonmonotonic function had two separate amplitude peaks, with a clear notch or worst amplitude between them. Other units had spike count functions that were mainly monotonic or nonmonotonic, but showed some evidence for a second response region. 2. Examination of response latency revealed another novel response property, which has been termed the paradoxical latency shift. Units with this response property responded at a shorter latency to sounds of low amplitude than to sounds of high amplitude. The paradoxical latency shift also appears to be related to the twin-peaked complex nonmonotonic response function. Units with the most prominent twin-peaked response functions also had the clearest latency shifts. In these units, the high-amplitude peak corresponded to the long-latency response and the low-amplitude peak to the short-latency-response. 3. These curious spike count and latency observations can be explained if they are considered in relation to the temporal and amplitude pattern of the acoustic input during echolocation. In echolocation, a loud orientation pulse is followed by a weaker echo. In delay-dependent facilitation, this pulse-echo sequence is followed by a neural response if the pulse-echo delay is appropriate. The simplest model for delay-dependent facilitation assumes that a synchronization of excitatory inputs from the pulse and echo is needed for facilitation. Since the weaker echo occurs after the pulse, it is closer in time to the postulated synchronization point. Therefore, in order for this model to work, the echo input must reach the summation place with less of a time lag than the pulse input. This is exactly what is seen with the paradoxical latency shift; the loud "pulse" response is delayed relative to the weak "echo" response.