1Wolfson Institute of Biological Research, UCL, London, UK
2Gatsby Computational Neurosci Unit, London, UK
In primary sensory areas, repeated sensory stimuli are known to trigger variable spiking responses. It is unclear if this variability is a result of intrinsic noise or is actually encoding information (e.g. due to additional input from brain areas other than the afferent pathway). We have addressed this question using a bottom-up approach: studying the effect of small perturbations on the cortical network during a repeated stimulus. Theoretical analysis and simulations using highly recurrent network connectivity based on the mammalian cortex show that the effect of small perturbations can grow very rapidly and interfere with the effect of the driving incoming input from the whisker. We demonstrate that one can quantify this effect by measuring the mean increase in the firing rate of an average neuron in response to a single synaptic input. If the mean increase is sufficiently large, then perturbations grow, which can preclude precisely repeatable spike trains. To address this question experimentally, we have used in vivo patch-clamp recordings from cortical pyramidal neurons in the barrel cortex of anesthetized rats. We find that a perturbation consisting of a single extra spike leads to 25 extra and missing spikes in the network. Using theoretical calculations and modelling, we show that the mean increase in firing rate provides a quantitative lower bound on the precision at which spike timing can carry information. To validate these predictions, we are making simultaneous patch-clamp recordings from layer 2/3 pyramidal neurons and from multiple units using an extracellular multisite recording silicon probe. We stimulate the primary whisker of the barrel with a dynamic stimulus and assessed the reproducibility of spike trains. We then introduce a perturbation by injecting depolarizing or hyperpolarizing current into the patched neuron during the whisker stimulation and observe the evolution of activity in the network recorded with the extracellular probe. These experiments allow us to link the spiking of single neurons with precisely timed activity in tens of recurrently connected cells in order to validate and improve our models of the effects of noise on cortical dynamics.