Protracted presynaptic activity can induce long-term potentiation (LTP) or long-term depression (LTD) of the synaptic strength. However, virtually all the experiments testing how LTP and LTD depend on the conditioning input are carried out with trains of stimuli at constant frequencies, whereas neurons in vivo most likely experience a stochastic variation of interstimulus intervals. We used a computational model of synaptic transmission to test if and to what extent the stochastic fluctuations of an input signal could alter the probability to change the state of a synapse. We found that, even if the mean stimulation frequency was maintained constant, the probability to induce LTD and LTP could be a function of the temporal variation of the input activity. This mechanism, which depends only on the statistical properties of the input and not on the onset of additional biochemical mechanisms, is not usually considered in the experiments, but it could have an important role to determine the amount of LTP/LTD induction in vivo. In response to a change in the distribution of the interstimulus intervals, as measured by the coefficient of variation, a synapse could be easily adapted to inputs that might require immediate attention, with a shift of the input thresholds required to elicit LTD or LTP, which are restored to their initial conditions as soon as the input pattern returns to the original temporal distribution.
