It might be contended, however, that Series I is an example of optical illusion, that the card was overestimated for that reason, and the interspaces necessarily made wider. To avoid this difficulty another series was made. Two sets of cards (125 mm. wide) were prepared; one with five black strips at equal distances apart as before (excepting that the strips were made heavier), the other with six strips. The card with an odd number of strips had thereby a strip in the middle upon which the attention could centre,--possessed a kind of balance. The card with an even number of strips had, moreover, no such central line but only a space, thus preventing repose of attention, and making the unit more pronouncedly rhythmic. (It will be noticed in the foregoing table that one subject, V, made narrower interspaces in I than in II. He said he felt the units as centring around the group of three lines in the centre, not as proceeding equally to the edge. The unit became thus for him symmetrical instead of rhythmic, which could easily account for the difference in estimation.) The results in the present case are an average of three trials: I, 5 strips II, 6 strips J. 129 137 S. 125 129 U. 132 133 H. 125 131 R. 127 138 W. 126 133 V. 123 129 Average difference in estimate of both series = 6.1. Mean variation = 2.1. Since both these figures represent an effect usually explained by optical illusion, that factor may be counted out, and the difference in the estimate be accounted for by the difference in the rhythm of the units. The difference in estimation between the two rhythmic units, differing only in odd and even number of strips, is greater than between the rhythmic and more strictly symmetrical, and yet the two were more comparable in construction. It would seem, then, that the greater overestimation of II is due to the rhythmic movement which is not limited or driven back to a central line as in I, but, by continuing over the limits, produces a greater feeling of breadth. The same question was experimented on in another way. Smaller strips of cardboard all 50 mm. wide, but with different designs, were hung behind the narrow window previously used. Four of each set were hung at a time behind the window, and subjects arranged them so that the interspaces appeared to equal the strips. These designs were to illustrate different points in question. The difference in estimation for an empty card, and a filled one; the difference according to the strongly centred, or rhythmic, or slant lines of the filling. These experiments were not so complete as the former ones, since the subjects were scattered; hence they represent only one trial or an average of two. But the results conform with what we have been led to expect. [Illustration: Fig. 26]