The article reports the results of experiments on human ability to estimate time intervals separately or in groups of two intervals as elements of a three-interval array presented by two light signals separated by a dark interval. The first light interval is shown to be systematically overestimated compared with the second one. Moreover, the subjects usually ascribe a constant value to the dark interval. The double interval (light + dark or dark + light) proves to be underestimated compared with the case when estimation of the same intervals was required without estimation of a third interval. The additional task changes the estimations of the subjects. It may be claimed that the complexity of the problem is connected with definite changes in the average estimations of the subjects.