A new method for evaluation of phasic spike responses is proposed. It is based on defining the time interval deltat in which spikes occur after the stimuli so often that the probability of finding a similar or more "suspicious" interval when the response is known to be absent (i.e., probability of error) is minimal. The offered response measure combines on the probability basis two different parameters: the time interval deltat and the number of trials containing at least one spike in deltat. The deltat interval may have any duration and location acceptable in the physiological sense, and is not supposed to be chosen a priori. This approach has been proved to be adequate and unambiguous. A technique is also proposed for estimating the error probability without using the characteristics of distribution of interspike intervals, which for this reason can be applied to any neurone. By means of this technique a table for non-parametric estimation of significance of spike responses has been compiled.