In this work, a method to reveal dissension in a measurement sequence based on the analysis of the distribution of an auxiliary random value is offered. This value — the number of sign changes of the difference between the current measurement and the respective estimate in a moving window — is formed in view of intermediate results obtained during execution of the stochastic approximation algorithm. The relation between the measure of the difference between two histograms of the specified random value and reference one (symmetric with respect to zero) and the dissension parameters is shown. It is proved that as such a measure it is possible to choose the initial statistical moment of the third order or the value of the *2 goodness-of-fit criterion.