CORTICAL SYNAPTIC DEPRESSION AND AUDITORY PERCEPTION

6. Introduction There are many aspects of auditory perception, such as the growth of loudness with duration and the effects of masking, which indicate that the auditory system performs some sort of temporal integration in processing incoming acoustic signals. However, the auditory system is also capable of fine temporal resolution, as evidenced by gap detection, double click discrimination, and also in the short latency and lack of jitter of onset responses in cortex [28]. This has been termed the resolution-integration paradox, i.e. how is it possible for a system to integrate information over long periods while retaining fine temporal resolution. Most accounts which satisfy the integration criterion use long time constants and therefore fail to behave swiftly enough to explain fine temporal resolution, and vice versa [28]. The time constants typically associated with sub-cortical processing differ substantially from those in the cortex. In comparison with the speed and precision associated with processing in the auditory periphery, the temporal response properties of neurons primary auditory cortex (AI) can appear to be surprisingly sluggish. For example, in the thalamocortical transformation of incoming signals a great deal of the temporal fine structure is lost [5], best modulation frequencies measured in AI are generally below 15 Hz [24], and the effects of a masker on a probe tone can be detected up to 400 ms after masker offset [3]. The focus in this paper is therefore on the temporal response properties observed in AI. What gives rise to these phenomena and can they be explained by some common mechanism? As yet there have been no models proposed which can satisfactorily explain the observed behaviour of neurons in AI. Explanations in terms of intracortical inhibitory circuits have been proposed but inhibition does not provide an adequate account, at least in the case of forward masking which is unaffected by the application of a GABA antagonist [3]. On the other hand, simple threshold neural models cannot replicate such behaviour without some form of inhibition or by means of very long time constants operating on the input signals, which as discussed above, would then prevent the model from satisfying the requirements for good temporal resolution. Recently it has become apparent that cortical synaptic dynamics may be an important factor affecting the behaviour of biological neurons [17,18,1,25,23]. When synapses are repeatedly activated they do not simply respond in the same way to each incoming impulse and synapses may …