8% 77% 81% 77% 82%

It will be seen that the number of correct interpretations is quite high
and in none of the cases does it deviate far from the mean average of
82%.

I based my judgment as to the direction of the subject's eyes, upon an
imaginary line perpendicular to the center of the cornea. (This
perpendicular does not always coincide with the subject's line of
vision, which was the thing I was after, but this cannot be directly
obtained. This, of course, was what made the judgment a rather difficult
matter.) My judgment as to the direction of the head I based largely on
the direction of the nose, (to express it more accurately: upon the
direction of the median plane.) I purposely noted only the position of
the experimenter and not the movement which led up to it. When I tried
to do the latter, the results were not always satisfactory, because the
head and eyes of the person would frequently, in the process of
adjustment, move beyond the goal and thus lead me into error. An attempt
was made to make each judgment as independent as possible of the
preceding one. But usually, after a few tests, an unintentional
association became established between certain attitudes and the
different places in the series of papers. Often all that was necessary
was to observe the experimenter in order to know which of the places he
had in mind, it was not necessary to look at the papers at all. Every
change in the position of the person would, of course, make the
association thus established, useless.

Later, the subjects and I changed roles, I took the part of the
experimenter and they the part of the "horse". The number of tests in
each case was 200 as before. Here, too, errors were, with but one
exception, never more than of one place to either side. Whether the
error was one place to the right or one place to the left appeared to
depend upon the position of the person making the judgment, i. e., it
depended on whether he stood at my right or at my left. The following
results were obtained:

Subject ("horse"): v. A. B. C. Mrs. v. H. K. Miss v. L.
Correct inferences: 76% 79% 75% 81% 77% 74%

A certain agreement can be seen in these results. The average of correct
inferences is somewhat lower than that which was obtained by me (page
135), 77% as over against 82%. This is probably due to the fact that the
subjects had had so little practice compared with me.

With one of