| Summary: | <p>The auditory system must represent sounds with a wide range of statistical properties. One important property is the spectrotemporal contrast in the acoustic environment. The level of some sounds varies only a little over frequency and time, while the level of other sounds can vary a lot. This raises a contrast problem for neural coding: auditory neurons, with a limited dynamic range of firing rates, must be able to efficiently encode the sound level fluctuations in both low and high contrast sounds. In this thesis, I show how the contrast problem is solved in the primary areas of the ferret auditory cortex (A1/AAF).</p> <p>One hypothesis is that different neurons specialise for representing sounds of different contrasts. I find little evidence for such specialisation in the auditory cortex. Rather, the system adapts its coding to operate under different contrast conditions. I demonstrate that neurons in A1/AAF rescale their gain to partially compensate for the spectrotemporal contrast of recent stimulation. When contrast is low, neurons increase their gain, becoming more sensitive to small changes in the stimulus, without changing their tuning. I quantify these gain changes and find that they resemble divisive normalisation, a phenomenon observed in many other neural systems. In a given stimulus context, auditory cortical neurons determine their gain predominantly on the basis of spectrotemporally local statistics. By comparing neural responses in A1/AAF with those in the inferior colliculus (IC), I show that this adaptive strategy of the auditory cortex is not simply inherited from the IC. When stimulus contrast changes, IC neurons undergo a variety of different changes in coding, which are on average weaker than in cortex. </p> <p>Together, these results suggest that the auditory cortex attempts to divide out contrast in its representation of an ongoing stimulus. This appears to be a novel property of the higher auditory pathway. </p>
|
|---|