UCL logo

UCL Discovery

UCL home » Library Services » Electronic resources » UCL Discovery

The influence of statistical context on the neural representation of sound

Williamson, R.S.; (2012) The influence of statistical context on the neural representation of sound. Doctoral thesis, UCL (University College London). Green open access

[img]PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
5Mb

Abstract

Models of stimulus-response functions have been used for decades in an attempt to understand the complex relationship between a sensory stimulus and the neural response that it elicits. A popular model for characterising auditory function is the spectrotemporal receptive field (STRF), originally due to Aertsen and Johannesma (1980); Aertsen et al. (1980, 1981). However, the STRF model predicts auditory cortical responses to complex sounds very poorly, presumably because the model is linear in the stimulus spectrogram and thus incapable of capturing spectrotemporal nonlinearities in auditory responses. Ahrens et al. (2008a) introduced a multilinear framework, which captures neuron-specific nonlinear effects of stimulus context on spiking responses to complex sounds. In such a framework, contextual effects are interpreted as nonlinear stimulus interactions that modulate the input to a subsequent STRF-like linear filter. We derive various extensions to this framework, and demonstrate that the nonlinear effects of stimulus context are largely inseparable, and fundamentally different for near-simultaneous and delayed non-simultaneous sound energy. In two populations of neurons, recorded from the mouse auditory cortex and thalamus, we show that simultaneous sound energy provides a nonlinear positive (amplifying) gain to the subsequent linear filter, while non-simultaneous sound energy provides a negative (dampening) gain. We demonstrate that this structure is largely responsible for providing a significant increase in the predictive capabilities of the model. Using this framework, we show that nonlinear context dependence differs between cortical fields, consistent with previous studies (Linden et al., 2003). Furthermore, we illustrate how such a model can be used to probe the nonlinear mechanisms that underly the ability of the auditory system to operate in diverse acoustic environments. These results provide a novel extension to the study of receptive fields in multiple brain areas, and extend existing understanding of the way in which stimulus context drives complex auditory responses.

Type:Thesis (Doctoral)
Title:The influence of statistical context on the neural representation of sound
Open access status:An open access version is available from UCL Discovery
Language:English
UCL classification:UCL > School of BEAMS > Faculty of Maths and Physical Sciences > CoMPLEX - Maths and Physics in the Life Sciences and Experimental Biology

View download statistics for this item

Archive Staff Only: edit this record