36. Level response changes in IC: adaptation or instantaneous non-linearity?

Phillipp Hehrmann1 hehrmann@gatsby.ucl.ac.uk Isabel Dean2 i.dean@ucl.ac.uk Misha B. Ahrens1 ahrens@gatsby.ucl.ac.uk Nicol S. Harper2 nicol.harper@ucl.ac.uk David McAlpine2 d.mcalpine@ucl.ac.uk Maneesh Sahani1 maneesh@gatsby.ucl.ac.uk

1Gatsby Computational Neuroscience Unit, University College London, London, UK
2EAR Institute, University College London, London, UK

Recent work has demonstrated that the responses of non-linear systems can appear sensitive to changes in the stimulus statistics, even when the parameters of the underlying system remain constant. A Reichardt motion detector, for example, exhibits a form of dynamic gain control closely resembling the apparently adaptive behaviour of the H1 neuron in the fly visual system. In the auditory domain, neurons in the inferior colliculus (IC) of anaesthetised guinea pigs have been shown to adapt their response characteristics to the sound-level distribution of an amplitude-modulated noise stimulus. Specifically, the rate-level functions of these neurons change in a way that improves the population coding accuracy around the most commonly occurring stimulus sound levels. We asked whether, and to what extent, this stimulus adaptation in IC could be explained as an inherent effect of an essentially unadaptive but nonlinear system being probed with varying stimuli (as has been suggested for H1), or whether it is a consequence of more slowly-acting adaptive processes in the system.

Following the former hypothesis, we first extended a parametric model originally proposed to characterize the responses of A1 neurons to short amplitude transients, so as to model the responses measured in the IC experiments. The longest membrane time constants in the model are on the order of 10 ms, much shorter than the duration of individual sound-level stimuli in the experiments (50 ms). Nevertheless, we found that for appropriate parameter settings, the rate-level functions obtained from the model were indeed sensitive to stimulus statistics, and in a way that closely resembled several key characteristics of the changes observed in IC. We then returned to the original data to quantify the relative contributions of the instantaneous effect of changes in the probe stimuli on the one hand and long term adaptive effects on the other. This analysis revealed a continuum of response characteristics: While instantaneous effects failed to account substantially for the observed changes in a majority of cells, we also found several cells for which the model provides a reasonable explanation as well as cases where cell responses seemed to be determined by both instantaneous and long-term adaptive effects.

Thus, although an established non-linear level-response model has the potential to reproduce the experimental findings, this mechanism alone does not provide a sufficient explanation. Firstly, we want to extend our simple model to incorporate slower adaptive processes. Secondly, we are going apply our current analysis to similar data from IC neurons in the domain of interaural time differences.