Spike timing precision in the visual front-end

Abstract

This thesis describes a series of investigations into the reliability of neural responses in the primary visual pathway. The results described in subsequent chapters are primarily based on extracellular recordings from single neurons in anaesthetized cats and area MT of an awake monkey, and computational model analysis. Comparison of spike timing precision in recorded and Poisson-simulated spike trains shows that spike timing in the front-end visual system is considerably more precise than one would expect on the basis of the time varying spike rate. Based on the nature of the measure that was used to quantify spike timing precision, this implies that spike trains in the visual front-end allow for an interesting decoding scheme. This encompasses optimal correlation detection, where temporal correlations are detected independent of straightforward synchronicity. This is described in chapters 1 and 2. Chapter 3 introduces a novel method: Motion Reverse Correlation (MRC), that was developed for measuring receptive field properties of motion selective cells in the visual cortex. Application of the method is illustrated with results obtained from area 18 and PMLS of anaesthetized cats and area MT in a fixating macaque monkey. In chapter 4, a conventional luminance white noise reverse correlation method is used to obtain spatio-temporal impulse responses of retinal ganglion cells, cells in the LGN and in area 17. These were then used to predict the responses of these cells to movie clips of natural scenes. Results show that conventional linear-static nonlinear models do not suffice to predict the recorded responses and suggest that dynamic nonlinear mechanisms should be taken into account also. As the quality of the predictions decreases from retina through area 17, it is concluded that these mechanisms become increasingly important at subsequent stages of information processing in the primary visual pathway. Chapter 5 focuses on the functional consequences of spike timing variability for visual motion detection. A bilocal correlator model is used to assess the time scale at which information is represented in the output of the retina and LGN, and its stimulus dependence. The pattern of results that we find is subsequently compared to the temporal limits for motion discrimination in a human psychophysics experiment. We find a close agreement between the results obtained in the two experiments and show that spike timing precision allows for correlation detection at very short time scales.