I II III IV V J. 49 57 52.5 52 53 S. 51 56 50 49 51 U. 53 54 49 50 48 H. 53 52 49 51 R. 52 52 51.5 48 51 W. 50 55 50 50 52 V. 51 53 49 48 51 Average = 51 54 50.6 49.4 51 Mean Variation = 1.5 1.4 1.2 1.06 .9 These results point to the fact that there is a tendency to overestimate the strips unless there is a strong central accent, which draws the attention back to the middle of the strip, in which case it is slightly underestimated. This would seem to be contradicted in I, where the centre is strongly marked by slant lines coming toward it. But the subjects, instead of taking the lines as pointing _towards_ the centre, in almost every case felt them as leading _away_ from it, and the oblique lines gave an appearance of greater breadth, which result was carried out by the greater overestimation of II. In this case, in addition to slant lines, there was no central accent, and the overestimation was proportionately large. III and IV were intended to illustrate the difference in estimate of rhythmic and symmetrical units, but although a slight difference is apparent, the subjects did not feel III as strongly rhythmic, because the black lines on the ends of the strip were ignored against the black background, and only the two central lines were taken. This made it more a balanced than a rhythmic unit, so it is not a fair type of the point in question. We may say, in conclusion, that oblique lines (which involve a more complex muscular adjustment to perceive them) give an impression of greater distance traversed, hence are overestimated; of two rhythmic groups, the one containing an even number is more overestimated than the odd, since the movement across is unchecked, and not balanced around a central line; a series of strongly centred groups is more correctly estimated as to its interspacing, and even slightly underestimated, because of the check imposed by the centre of fixation in each group. Although these results are very uniform, a more complete series of experiments should be done on this subject, to make the conclusions thoroughly valid. [Illustration: Fig. 27]