While it seems like an effortless task to make sense of sounds, it is in fact a formidable challenge to turn the continuous stream of pressure variations that reach our ears into meaning. This process relies on the ability to detect spectral sequences, but our understanding of how the brain accomplishes this has remained elusive. This thesis seeks to change that by uncovering the neural mechanisms which allow the auditory system to detect sequences in the stream of pressure variations we generally call sounds. The work presented in this thesis utilises two approaches to shed light on how spectral sequences can be detected by the brain’s neural network. Publications 1 and 2 draw on computational modelling of hierarchical neural networks, while Publications 3 and 4 present new computational methods for analysing data from single-cell recordings. The first approach illustrates that synaptic depression, a form of short-term plasticity of connections between neurons, facilitates sequence selectivity in simulated hierarchical neural circuits in a manner consistent with experimental findings. Sequence selectivity emerges for sequences of increasing duration as the hierarchy is traversed, and single-cell and population responses exhibit stimulus-specific adaptation and mismatch responses, respectively. The second approach, in turn, resulted in new models (context models) for describing neural responses to arbitrary stimuli. These models can detect synaptic depression and sequence selectivity in single-cell recordings, two effects that the commonly used spectro-temporal receptive field is unable to capture. The context models were also found to be superior when compared to another commonly used model of neural behaviour, the multi-filter LN model. This observation held for both simulated data and real data from complex neurons in the visual cortex. In conclusion, this thesis 1) highlights the idea that synaptic depression in a hierarchical network might be one of the underlying mechanisms which lets our brain detect sequences, and 2) yields new tools that could be used for investigating sequence selectivity in the brain. These results may therefore be useful for advancing our understanding of how populations of neurons can make sense of sounds.
|Publication status||Published - 2018|
|MoE publication type||G5 Doctoral dissertation (article)|
- temporal binding
- context models
- synaptic depression
- mismatch response
- stimulus-specific adaptation