Abstract. A probability model for a combined gamma generator with nonuniform noise in a resynchronization mode is studied. We consider the problem of testing hypotheses about the distribution of a random binary vector X (0) (the state of a combined gamma generator) by using a sampled binary sequence whose signs depend on X (0) in a specified way and on certain other random parameters. We obtain a nonasymptotic upper bound for the mean error probability of a Bayesian criterion for testing the hypotheses mentioned above. The setting of the problem and main resultsLet n, L, and t be natural numbers, ∆ be a nonempty subset of the setand let g (1) , . . . , g (t) be an arbitrary sequence of n-dimensional Boolean vectors. Further,n , l ∈ {1, . . . , t}, be jointly independent random vectors such that (a) the vector X (0) has the uniform probability distribution in the set V nL , (b) the vector C (l) has the uniform probability distribution in the set a 1 ), . . . , y n (a n )) = s} for all s ∈ V n , 2000 Mathematics Subject Classification. Primary 94A60; Secondary 94B70.