to give his judgment on the same distance ten times before passing to another. A glance at the diagram will show how pronounced the illusion is at first, and how it then disappears, and the judgment settles down to a uniform degree of accuracy. It will be seen that the short filled space is at first overestimated, and then, with the succeeding judgments, this overestimation is gradually reduced. In the case of the longer filled distances (which could not be conveniently reproduced here) the spaces were at first underestimated, and then this underestimation slowly decreased. [Illustration: FIG. 8.] [Illustration: FIG. 9.] None of the qualitative studies that have hitherto been made on this illusion have brought to light this significant wearing away of the illusion. VII. I have already spoken of the defects of the apparatus with which the experiments of the previous chapter were made. I shall now give an account of some experiments that were made with an apparatus designed to overcome these difficulties. This is shown in Fig. 10. The block _C_ was clamped to a table, while the block _A_ could be moved back and forth by the lever _B_, in order to bring up different lengths of filled space for judgment. For each judgment the subject brought his finger back to the strip _D_, and by moving his finger up along the edge of this strip he always came into contact with the first point of the new distance. The lever was not used in the present experiment; but in later experiments, where the points were moved under the finger tip, which was held stationary, this lever was very useful in producing different rates of speed. In one series of experiments with this apparatus the filled spaces were presented first, and in another series the open spaces were presented first. In the previous experiments, so far as I have reported them, the filled spaces were always presented first. [Illustration: FIG. 10.] In order to enable the subject to make proper connections with the first point in the filled space, when the open space was presented first, a slight depression was put in the smooth surface. This depression amounted merely to the suggestion of a groove, but it sufficed to guide the finger.