at _A_. When the
subject's eye is watched, it is found that in this case it moved
either too soon or too late, so that when the exposure was made, the
eye was resting quietly on one of the fixation-points and so naturally
received the same image as in case 1, except that now it lies in
indirect vision, the eye being directed not toward _A_ (as in case 1)
but towards either _P_ or _P'_.

Second, the image correctly localized may be like 2 (Fig. 7), and then
it is seen to move past the opening _ON_. The handle _h_ looks as
bright as _e_, _e_. This appearance once obtained generally recurs
with each successive swing of the pendulum, and scrutiny of the
subject's eye shows it to be moving, not by separate voluntary
innervations from _P_ to _P'_ and then from _P'_ to _P_, but
continuously back and forth with the swing of the pendulum, much as
the eye of a child passively follows a moving candle. This movement is
purely reflex,[20] governed probably by cerebellar centers. It seems
to consist in a rapid succession of small reflex innervations, and is
very different from the type of movement in which one definite
innervation carries the eye through its 42 deg., and which yielded the
phenomena with the perimeter. A subject under the spell of this reflex
must be exercised in innervating his eye to move from _P_ to _P'_ and
back in single, rapid leaps. For this, the pendulum is to be
motionless and the eye is not to be stimulated during its movement.

[20] Exner, Sigmund, _Zeitschrift f. Psychologie u. Physiologie
der Sinnesorgane_, 1896, XII., S. 318. 'Entwurf zu einer
physiologischen Erklaerung der psychischen Erscheinungen,'
Leipzig u. Wien, 1894, S. 128. Mach, Ernst, 'Beitraege zur
Analyse der Empfindungen,' Jena, 1900, S. 98.

These two cases in which the image is localized midway between _P_ and
_P'_ interest us no further. Localized on the final fixation-point,
the image is always felt to flash out suddenly _in situ_, just as in
the case of the 'correctly localized' after-image streaks in the
experiments with the perimeter. The image appears in one of four
shapes, Fig. 7: 2 or 3, 4 or 5.

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