Gen. Physiol. Biophys. (2004), 23, 21—38 21 Different Types of Noise in Leaky Integrate-and-Fire Model of Neuronal Dynamics with Discrete Periodical Input V. Di Maio 1 , P. Lánský 2 and R. Rodriguez 3 1 Istituto di Cibernetica, “E. Caianiello” del CNR, Pozzuoli, Napoli, Italy 2 Institute of Physiology, Academy of Sciences of the Czech Republic, Vídeňská 1083, 142 20 Prague 4, Czech Republic 3 Centre de Physique Théorique, CNRS and Faculté des Sciences de Luminy, Université de la Méditerranée, Luminy-Case 907, F–13288 Marseille Cedex 09, France Abstract. Different variants of stochastic leaky integrate-and-fire model for the membrane depolarisation of neurons are investigated. The model is driven by a con- stant input and equidistant pulses of fixed amplitude. These two types of signal are considered under the influence of three types of noise: white noise, jitter on inter- pulse distance, and noise in the amplitude of pulses. The results of computational experiments demonstrate the enhancement of the signal by noise in subthreshold regime and deterioration of the signal if it is sufficiently strong to carry the infor- mation in absence of noise. Our study holds mainly to central neurons that process discrete pulses although an application in sensory system is also available. Introduction Generation of spikes and the spatio-temporal patterns they form represent the ba- sic mechanism by which information is exchanged by neurons. Spike sequences, recorded from neurons located in very different structures and under different ex- perimental conditions, suggest the presence of stochastic forces influencing neuronal activity. This random component, generally considered as a noise, can be contained either in the input signal or generated in the target neurons themselves. In commu- nication theories, the term noise usually denotes something negative and blurring the signal. In living systems, the noise can be a message by itself (Cecchi et al. 2000; Lánský and Sacerdote 2001) or its highly desirable part, as shown in articles on stochastic resonance (Moss et al. 1993; Longtin et al. 1994; Bulsara et al. 1996 and many others). A typical model of single neuron mimics the voltage of its membrane by using an equivalent electrical circuit described by a system of differential equations. The Correspondence to: Vito Di Maio, Istituto di Cibernetica “E. Caianiello” del CNR, Via Campi Flegrei 34, 80078 Pozzuoli, Napoli, Italy. E-mail: vdm@biocib.cib.na.cnr.it