Clever Hans, the horse that could apparently count. · published 11 September 2026
They were right — there was no trick, and the horse still could not count; Hans was reading the involuntary tension in the questioner’s body and stopping the instant it released, which is why the real failure was in the question the panel asked, not in their honesty, and why double-blind procedure exists at all.
These are the working notes, unedited. Everything the video
states as fact comes from here, and nothing else may be stated as fact.
The sections on limits and on what the finding does not show are
part of the brief, not an afterthought — see
the method.
The claim: a horse in Berlin, owned by a retired schoolteacher named Wilhelm von Osten, who appeared to do arithmetic. Asked a question, Hans answered by striking the ground with his hoof — the right number of taps, then a stop. He was credited with addition, subtraction, telling the time, identifying colours and spelling out words. The exhibitions were free; von Osten never charged, which is part of why so many people believed it.
The first investigation got it wrong. In September 1904 the psychologist Carl Stumpf assembled a panel of thirteen — a veterinarian, a cavalry officer, schoolteachers, a circus manager, the director of the Berlin zoological gardens — to look for trickery. They found none, and said so. That verdict was correct and useless: there was no trick, and the horse still could not count.
What actually settled it. In 1907 Oskar Pfungst, working in Stumpf’s laboratory, varied the things the earlier panel had left constant. He changed who asked the question, whether that person knew the answer themselves, how far away they stood, and whether Hans could see them at all.
The numbers. When Hans could see the questioner he answered correctly on 89 per cent of trials — 50 of 56. When he could not, he was right on 6 per cent — 2 of 35. The pattern held regardless of who was asking, and it collapsed whenever the questioner did not know the answer.
The mechanism. As the taps climbed toward the correct number, the person asking grew very slightly tense — a change of posture, of expression, of breathing — and released that tension the instant the right number was reached. Hans had learned to start tapping and to stop on that release. It is an extraordinary piece of animal perception, just not the one being claimed.
Von Osten was not a fraud, and this matters. He had no idea he was signalling anything, and he never accepted Pfungst’s explanation; he went on exhibiting Hans until his death in 1909. The cue was invisible to the man producing it.
What it established. The case is the origin of routine double-blind procedure, and of the observer-expectancy effect — the experimenter’s knowledge of the expected answer leaking into the result through channels nobody has noticed. An artefact of this kind is still called a Clever Hans effect, and it is why work on animal cognition and on human subjects is designed so the person collecting the data does not know what they should be seeing.
Limits
Pfungst never managed to reproduce Hans’s errors reliably by producing the cues deliberately himself, which is a genuine loose end in an otherwise convincing account. The exact figures above come from Pfungst’s own trials. And this is one horse and one investigation: it demonstrates that a striking result can have an unnoticed channel behind it, not that animals cannot learn anything difficult.
What this does not show
It is not that Hans was stupid, or that the people who believed it were fools. Reading a human being’s involuntary posture accurately enough to stop on the right beat, across different questioners, is a real and rare ability. The thirteen experts who cleared him were asking the right question — is there a trick? — and the answer was no. The failure was in the question, not in their honesty, and that is the part worth taking seriously: an experiment can be run in good faith by careful people and still measure the wrong thing.