Schaette, Roland, Tim Gollisch, and Andreas V. M. Herz. Spike-train variability of auditory neurons in vivo: dynamic responses follow predictions from constant stimuli. J Neurophysiol 93: 3270 -3281, 2005. First published February 2, 2005 doi:10.1152/jn.00758.2004. Reliable accounts of the variability observed in neural spike trains are a prerequisite for the proper interpretation of neural dynamics and coding principles. Models that accurately describe neural variability over a wide range of stimulation and response patterns are therefore highly desirable, especially if they can explain this variability in terms of basic neural observables and parameters such as firing rate and refractory period. In this work, we analyze the response variability recorded in vivo from locust auditory receptor neurons under acoustic stimulation. In agreement with results from other systems, our data suggest that neural refractoriness has a strong influence on spike-train variability. We therefore explore a stochastic model of spike generation that includes refractoriness through a recovery function. Because our experimental data are consistent with a renewal process, the recovery function can be derived from a single interspike-interval histogram obtained under constant stimulation. The resulting description yields quantitatively accurate predictions of the response variability over the whole range of firing rates for constant-intensity as well as amplitude-modulated sound stimuli. Model parameters obtained from constant stimulation can be used to predict the variability in response to dynamic stimuli. These results demonstrate that key ingredients of the stochastic response dynamics of a sensory neuron are faithfully captured by a simple stochastic model framework.
I N T R O D U C T I O NSensory neurons provide an animal with the primary representation of its environment and internal state. Any computation and behavioral decision has to be based on this representation. Yet even under controlled laboratory conditions, repeated stimulus presentations often lead to a considerable trial-to-trial variability of neural responses. This is particularly true for cortical neurons, which typically discharge with high variability under in vivo conditions (Buracas et al. 1998;Holt et al. 1996;Shadlen and Newsome 1998;Stevens and Zador 1998). As expected for Poisson-like random processes, coefficients of variation near unity and spike-count variances equal to or even above the mean count have been found (Softky and Koch 1993;Teich et al. 1996).On the other hand, neural response characteristics depend sensitively on the temporal features of the stimulus pattern (Mainen and Sejnowski 1995), which is of special importance for sensory neurons receiving highly structured dynamic stimuli (de Ruyter van Kara et al. 2000;Machens et al. 2001;Warzecha et al. 2000). Indeed, for appropriate inputs, individual spikes can be highly reliable and precisely timed (Berry et al. 1997;Reinagel and Reid 2002), resulting in spike-count variances far below the mea...