All this goes to show that Hans did not know the names of the colors (to say nothing of the symbols on the placards). It was plain that here also, as in all the other cases, he was controlled by signs made by the questioner, the nature of which I soon discovered. Standing erect, Mr. von Osten always turned head and trunk in the direction of the cloth or placard desired. Hans, keeping his eye on his master, would proceed in that direction. Even after he had already started out, thanks to his large visual field one could control his direction by turning slightly more to the right or to the left. If, however, he had already arrived at the row of placards or cloths, this method ceased to be effective, for then he could no longer see the experimenter. It made no difference whether the cloths lay on the ground, or were suspended, like the placards.
The following fact justifies the conclusion that the bodily attitude of the questioner was the effective signal. The more numerous the cloths, or the nearer they were placed together, the more difficult one would expect it to be for the horse to select the one indicated by the experimenter. Such was indeed the case, for the number of errors increased with the number of cloths presented.
But no matter how many cloths there might be, or how closely they might be placed, it was always possible to indicate either end of the row, for in that case one had merely to turn to the extreme left or the extreme right, and might even turn beyond the row. Hans seldom failed in these cases, whereas he made many errors when cloths or placards within the series were wanted.
To turn from the nature and number of Hans's errors, to their distribution,--observation proved the hypothesis that the nearer two cloths lay together, the greater was the chance of their being mistaken one for the other. If we designate as "error 1" all those cases in which Hans went to cloth II instead of to cloth I, cloth III instead of cloth II, to V instead of IV, etc., and as "error 2" when he mistook III for I, IV for II, in fine, whenever he went two places too far to the right or left, and as "error 3" whenever he went three places too far to either side of the cloth desired, we find the following grouping of errors: