Three series of experiments were carried out on 21 subjects. In the first series the subjects were presented with two temporal intervals--500 and 800 ms, limited by electrocutaneous stimuli of 1 ms duration. According to the instruction, the subjects reproduced them, pressing the left or right button, correspondingly. In the other two series, the subjects made the prognosis of presentation of the one or the other interval by corresponding reaction, after which they were presented with the real interval (at P = 0.5/05 in the 2nd series and P = 0.2/08 in the 3d series), which also had to be reproduced. In reproduction of temporal intervals without preliminarily prognosis, the number of wrong reactions reached 12%. In conditions of preliminary prognosis at equally probable signals distribution, the number of wrong reactions was equal to the number of correct ones, and at different probabilities, for rather improbable signals it exceeded twice,--and for more probable--it was twice less than the number of correct responses. Besides, the accuracy of interval reproduction depended on the prognosis of signals distribution but it could correspond to the really perceived interval and not to the interval chosen at prognosis. The reaction time, as the experiments showed, at prognosis was significantly greater than the reaction time of interval reproduction.