In the psychophysical experiments the symmetrical bright lines of various length and curvature are presented against the dark background on the monitor. The subjects are asked to adjust the test-spot by moving it horizontally or vertically into the invisible arc linking the endpoints of two visible segments of the given stimulus. The data have shown that the precision of curvature estimation depended on type of the curves: it is higher for the centered circle, less for the exponential spiral and still less for the shifted circle. The precision is in direct proportion with the length of the visible segment and in reverse proportion with the invisible segment of the stimulus. The curvature estimation does not depend on the stimulus symmetry, orientation, and position in the visual field.