| Summary: | Mammalian brains generate complex, dynamic structures of oscillatory activity, in which distributed regions transiently engage in coherent oscillation, often at specific stages in behavioural or cognitive tasks. Much is now known about the dynamics underlying local circuit synchronisation and the phenomenology of where and when such activity occurs. While oscillations have been implicated in many high level processes, for most such phenomena we cannot say with confidence precisely what they are doing at an algorithmic or implementational level. This thesis presents work towards understanding the dynamics and possible function of large scale oscillatory network activity. We first address the question of how coherent oscillatory activity emerges between local networks by measuring phase response curves of an oscillating network in vitro. The network phase response curves provide mechanistic insight into inter-region synchronisation of local network oscillators. Highly simplified firing models are shown to reproduce the experimental data with remarkable accuracy. We then focus on one hypothesised computational function of network oscillations; flexibly controlling the gain of signal flow between anatomically connected networks. We investigate coding strategies and algorithmic operations that support flexible control of signal flow by oscillations, and their implementation by network dynamics. We identify two readout algorithms which selectively recover population rate coded signal with specific oscillatory modulations while ignoring other distracting inputs. By designing a spiking network model that implements one of these mechanisms, we demonstrate oscillatory control of signal flow in convergent pathways. We then investigate constraints on the structures of oscillatory activity that can be used to accurately and selectively control signal flow. Our results suggest that for inputs to be accurately distinguished from one another their oscillatory modulations must be close to orthogonal. This has implications for interpreting in vivo oscillatory activity, and may be an organising principle for the spatio-temporal structure of brain oscillations.
|
|---|
