Clever Hans: The Cue Nobody Knew They Were Sending

Between 1904 and 1907 the most famous animal in Europe was a horse in a Berlin courtyard who could do arithmetic.

Wilhelm von Osten, a retired mathematics teacher, had spent years teaching his stallion Hans, and the results were extraordinary. Asked what four plus three was, Hans tapped his hoof seven times. He could handle fractions, tell the time, identify musical intervals, spell using a letter-to-tap conversion, and answer questions from strangers with von Osten absent. A commission of thirteen investigators, including a circus trainer and the director of the Berlin zoo, examined him in 1904 and found no evidence of fraud. There was none. Von Osten was entirely sincere.

Oskar Pfungst, a young psychologist, was given the problem in 1907 and solved it with two experimental manipulations.

First: does the questioner know the answer? When the person asking the question did not know the answer themselves — because it had been whispered to the horse's other side, or split between two people — Hans's accuracy collapsed.

Second: can the horse see the questioner? With blinkers restricting his view, or with the questioner behind a screen, accuracy collapsed again.

Pfungst then worked out what Hans was reading. As a questioner watched the tapping and approached the correct count, they made a minute involuntary movement — a slight relaxation, a small upward tilt of the head, a change in breathing or posture. The movement was so small that Pfungst could barely produce it deliberately once he knew what it was, and questioners were entirely unaware of making it. Hans, who had been rewarded with carrots for stopping at the right moment for years, had become an extremely good reader of it.

The horse was not doing arithmetic. He was doing something more impressive.

The general phenomenon is ideomotor cueing: expectation produces involuntary micro-movement, and a sufficiently sensitive receiver reads it. It is not confined to horses.

Facilitated communication, in which a facilitator supports the hand of a non-speaking person at a keyboard, was shown in controlled trials to be producing the facilitator's words rather than the client's — the messages failed precisely when the facilitator did not know the answer, exactly Pfungst's first manipulation. Dowsing rods move because the dowser's hands move. Ouija planchettes move for the same reason, and the participants' sincerity is not in question. And the cold reader in chapter 27 is often genuinely reading the client's involuntary reactions rather than fabricating from nothing, which is why a good one seems uncanny in person and unimpressive on a transcript.

The scientific consequence was permanent: this case is why blinding became standard procedure. If the experimenter's expectation can travel through a channel neither party can perceive, then the experimenter must not know which condition a subject is in. Chapter 23's replication failure — where the priming effect appeared only when the experimenter expected it — is the Clever Hans effect surviving into modern social psychology.

The personal application: if you are questioning someone and you already have an answer in mind, you are transmitting it. The interview protocols in Part 4 that require the interviewer to withhold what they know are not only about catching liars. They are about not accidentally supplying the answer.