First, the plain or elongated outline of the dumb-bell appears with its handle on the final fixation-point (2 or 3). The image is plain and undistorted if the eye moves at just the rate of the pendulum, elongated if the eye moves more rapidly or more slowly. The point that concerns us is that the image appears with its handle. Two precautions must here be observed.
The eye does not perhaps move through its whole 42°, but stops instead just when the exposure is complete, that is, stops on either O or N and considerably short of P or P'. It then follows that the exposure is given at the very last part of the movement, so that the after-image of even the handle h has not had time to subside. The experiment is planned so that the after-image of h shall totally elapse during that part of the movement which occurs after the exposure, that is, while the eye is completing its sweep of 42°, from O to P, or else from N to P'. If the arc is curtailed at point O or N, the handle of the dumb-bell will of course appear. The fact can always be ascertained by asking the subject to notice very carefully where the image is localized. If the eye does in fact stop short at O or N, the image will be there localized, although the subject may have thoughtlessly said before that it was at P or P', the points he had nominally had in mind.
But the image 2 or 3 may indeed be localized quite over the final fixation-point. In this case the light is to be looked to. It is too bright, as it probably was in the case of Dodge's experiments. It must be further reduced; and with the eye at rest, the control (case I) must be repeated. In the experiments here described it was always found possible so to reduce the light that the distinct, entire image of the dumb-bell (2, Fig. 7) never appeared localized on the final fixation-point, although in the control, H, of Fig. 7:1, was always distinctly visible.
With these two precautions taken, the image on the final fixation-point is like either 3, 4, or 5. Shape 5 very rarely appears, while the trained subject sees 4 and 3 each about one half the times; and either may be seen for as many as fifteen times in succession.
Shape 4 is of course exactly the appearance which this experiment takes to be crucial evidence of a moment of central anæsthesia, before the image is perceived and during which the stimulation of the handle h completely elapses. Eight subjects saw this phenomenon distinctly and, after some training in timing their eye-movements, habitually. The first appearance of the handleless image was always a decided surprise to the subject (as also to the writer), and with some eagerness each hastened to verify the phenomenon by new trials.
The two ends (e, e) of the dumb-bell seem to be of the same intensity as in shape 2 when seen in reflex movement. But there is no vestige whatsoever of a handle. Two of the subjects stated that for them the place where the handle should have been, appeared of a velvety blackness more intense than the rest of the background. The writer was not able to make this observation. It coincides interestingly with that of von Kries,[21] who reports as to the phases of fading after-images, that between the disappearance of the primary image and the appearance of the 'ghost,' a moment of the most intense blackness intervenes. The experiments with the pendulum, however, brought out no ghost.
We must now enquire why in about half the cases shape 3 is still seen, whereas shape 5 occurs very rarely. Some of the subjects, among whom is the writer, never saw 5 at all. We should expect that with the intensity of H sufficiently reduced 4 and 5 would appear with equal frequency, whereas 3 would be seen no oftener than 2; shape 5 appearing when the eye did not, and 4 when it did, move at just the rate of the pendulum. It is certain that when 4 is seen, the eye has caught just the rate of the pendulum, and that for 3 or 5 it has moved at some other rate. We have seen above ([p. 27]) that to move with the pendulum the eye must already move decidedly more slowly than Dodge and Cline find the eye generally to move. Nothing so reliable in regard to the rate of voluntary eye-movements as these measurements of Dodge and Cline had been published at the time when the experiments on anæsthesia were carried on, and it is perhaps regrettable that in the 'empirical' approximation of the natural rate of the eye through 40° the pendulum was set to move so slowly.
In any case it is highly probable that whenever the eye did not move at just the rate of the pendulum, it moved more rapidly rather than more slowly. The image is thus horizontally elongated, by an amount which varies from the least possible up to 9 cm. (the width of the opening in T), or even more. And while the last of the movement (O to P, or N to P'), in which the stimulation of H' is supposed to subside, is indeed executed, it may yet be done so rapidly that after all H' cannot subside, not even although it is now less intense by being horizontally spread out (that is, less concentrated than the vanished h of shape 4). This explanation is rendered more probable by the very rare appearance of shape 5, which must certainly emerge if ever the eye were to move more slowly than the pendulum.
The critical fact is, however, that shape 4 does appear to a trained subject in about one half the trials—a very satisfactory ratio when one considers the difficulty of timing the beginning of the movement and its rate exactly to the pendulum.
Lastly, in some cases no image appears at all. This was at first a source of perplexity, until it was discovered that the image of the dumb-bell, made specially small so as to be contained within the area of distinct vision, could also be contained on the blind-spot. With the pendulum at rest the eye could be so fixed as to see not even the slight halo which diffuses in the eye and seems to lie about the dumb-bell. It may well occur, then, that in a movement the image happens to fall on the blind-spot and not on the fovea. That this accounts for the cases where no image appears, is proved by the fact that if both eyes are used, some image is always seen. A binocular image under normal convergence can of course not fall on both blind-spots. It may be further said that the shape 4 appears as well when both eyes are used as with only one. The experiment may indeed as well be carried on with both eyes.