Natural sound intensities can vary extensively over the long term. For short periods, and within a given environment, however, they more often fluctuate over a relatively limited range. An important function of hearing is to detect, and discriminate between, important acoustic signals against the background of interfering noise. Mammals can perform this task over a vast of sound intensities (120 decibels) with remarkable accuracy. Recently, using single neuron recordings, we have shown that neurons in the inferior colliculus of the guinea pig code sound intensity over this wide range by adjusting their firing rates to take account of the mean, variance, and more complex statistics of sound level distributions (Dean et al. 2005). These adjustments improve the accuracy of the neural population code close to the region of most commonly-occurring sound intensities, extending the range of sound levels that can be accurately encoded, fine-tuning hearing to the local acoustic environment. This adaptation to the stimulus statistics occurs rapidly; switching between different distributions of sound intensity reveals a time constant of approximately 100 ms. Currently we are exploring neural mechanisms that might account for this adaptive behaviour.