all hand movements, including the "wonderfully effective thrust of the
hand into the pocket filled with carrots", brought no response. I might
also change my position and walk forward and then backward some distance
behind the horse, but the back-step would only occur in response to the
characteristic stimulus. After what has been said it is easy to
understand how vain were Mr. Schillings' attempts to disturb the horse
and how naturally he might conclude that Hans was not influenced by
visual signs. Mr. Schillings simply did not know which signs were
effective.

While the horse could thus be interrupted in the process of tapping by
movements which were executed at the level of the questioner's head, yet
movements below this level had the opposite effect. If Hans showed that
he was about to cease tapping before it was desired, it was possible to
cause him to continue by simply bending forward a trifle more. The
greater angle at which the questioner's trunk was now inclined caused
the horse to increase the rate of tapping. The rule may be stated thus:
The greater the angle at which the body inclined forward, the greater
the horse's rate of tapping, and _vice versa_. It was noticeable that
whenever Mr. von Osten asked for a relatively large number--in which
case he always bent farther forward than in the case of smaller
numbers--Hans would immediately begin to tap very swiftly. Not being
entirely satisfied with these observations, the following more exact
measurements were taken. I asked the horse to tap 20. From 1 to 10 I
held my body at a certain constant angle, at 10 I suddenly bent farther
forward and retained this posture until 20 had been reached. If there
existed a relationship between the angle of inclination and the rate of
tapping, then the time for the last ten taps ought to be less than for
the first ten. Of 34 such tests 31 were sucessful. The following are two
specimen series.

The first series consisted of ten tests of 15 taps each. In all cases my
head was bent at an angle of 30 deg. to the axis of the trunk, but I
constantly changed the angle of inclination of the trunk. It was not
possible to measure this angle accurately on account of the rapidity
with which the whole test had to be made. I was able, however, to
differentiate between them with enough accuracy to designate the
smallest angle (about 20 deg.) as belonging to Grade I, and the greatest
angle (about 100 deg.) as belonging to Grade VII. By fi