We investigated the influence of the degree of similarity of two simultaneously presented geometrical figures on their recognition in tachistoscopical experiments on human beings. T he test stimuli w ere 16 nonverbal figures consisting of six vertical and horizontal line-segments. There were five groups of figure pairs (16 pairs in each group) that differed according the ,number of coincident line segments between two figures: two, three, four, five, or six (in the case of identical figures) . A pair of test figures was presented for 50 ms and after an individually determined interstimulus interval it w as masked by two rectangular frames of 12 line segments. On the first experimental situation, a special marker indicating which test figure must be drawn appeared after masking the patterns. ln the second experimental situation, the subjects were required to draw both test figures. Identical figures were recognized with higher probability than figures with a different degree of similarity. T he recognition probability of two simulultaneously presented figures also depended upon the practice level of the subjects. For experienced subjects we obtained a U-shaped function of recognition probability on the similarity level. Most of the inexperienced subjects did not show any recognition dependence on the similarity level. Different instructions in the second situation - to draw both figures - altered the experimental results. We concluded that the results in the situation. in which the subject was required to draw both figures reflect not the recognition accuracy of the two figures but the accuracy of restoration of the second figure from working memory.