It would mean, then, that the apperception of the repeated group in I (if it accords with the subject's own introspection) consists in repeated fluctuations of attention over the five strings, with no repose on any one more than another. The movement is back and forth from edge to edge, and hence needs more of an end to finish it than in a series of symmetrical units where the movement is not back and forth, but balanced and resting on the central point. In other words, in Group I there is a rhythm of movement within the group itself, as well as of the whole, while in II it is balanced and coördinated from the centre of each group, out and back, so that a longer, or at least more important end of some description is necessary to break the rhythm, and stop the series in I than in II. It is noticeable also that H. and S. thought in both cases they were making the end spaces equal to the interspaces; but after Series I, _a_ was made 102 and 109; and after Series II 95 and 104 respectively. This naturally raised another question: Does a series of groups, with repetitions within each, tend to make one overestimate distances between or at the end, or at least does one overestimate these distances in comparison with a series of symmetrical units? The subjects were so unanimous in preferring a larger "embankment" after Series I than II, that it was useless to test them further on that point, and the experiment was changed to the other question according to the suggestion above. A series of eight cards was prepared (125 mm. wide) on four of which five heavy black strips were drawn equally distant from each other, and on the others a much wider strip in the centre with another on each side near the edges. Of the two series just made, one was composed of what we have called "rhythmic units" and the other of "symmetrical units." [Illustration: Fig. 25] The subjects were asked to arrange the two separate series so that the interspaces should be exactly equal to the units. It will be observed that the rhythmic unit had a black strip on each edge, thereby apparently decreasing its size, while the edge of the symmetrical unit was white. In this respect the comparison was hardly fair, but the result was the following. The figures represent an average of two trials, and stand for the size estimate of the interspaces for each subjects respectively, in the two series. I II J. 132.5 130 S. 125 122 U. 133 129 H. 128 124 R. 126 122 W. 136 133 V. 129 131 Average difference of estimate of both series = 2.64. Mean variation = 1.37.