Purchase this article with an account.
M.J. Lankheet, B.G. Borghuis, W.A. van de Grind; Spike Timing Precision: Stimulus Dependence and Predictability . Invest. Ophthalmol. Vis. Sci. 2005;46(13):5684.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: Retinal spike trains can encode motion information on a time scale of about 10 ms, which is on the order of the interspike interval. Limits in spike timing precision therefore limit information capacity of a spike train. We asked whether spike timing precision depends on stimulus contrast and temporal dynamics, and assessed whether a stochastic model can account for the observed precision. Methods: Spike trains were recorded from the optic tract and LGN of anaesthetized cats. Using a novel method, we quantified spike timing precision of the cells' responses to drifting sine wave gratings of varying contrast and temporal frequency. We fitted a Poisson model to the response and compared firing precision of recorded and simulated spike trains. Results: Spike timing precision ranged from about 70 ms at low contrast and low temporal frequency to about 2.0 ms at high contrast and high temporal frequency. At low firing rates, precision of recorded and simulated spike trains was the same. At high firing rates however, precision of recorded spike trains exceeded precision of simulated spike trains. This discrepancy remained, even when absolute and relative refractory periods were accounted for. We then fitted a Noisy Leaky Integrate and Fire model to the recorded responses so as to reproduce the recorded firing rate. By manipulating the amplitude of the additive noise source, timing precision of the recorded responses could be accurately reproduced. Furthermore, the model accounted for dynamic changes in precision over the course of the response and predicted the observed dependence on grating contrast and temporal frequency. Conclusions: Timing precision of retinal spike trains is strongly stimulus dependent. The observed timing precision is accurately described by a Noisy Leaky Integrate and Fire model, but not by a Poisson model – even when refractoriness of the spike generator is accounted for.
This PDF is available to Subscribers Only