Week 4 Seminar: Association

The first mechanism that can be systematically fooled, and why nothing inside it can tell the difference between a cause and a reliable coincidence.

Part II — Interior models18 March 2027Week 4 lectureSlidesSchedule

Once a system can link two events so that one predicts the other, it can learn. It can also be wrong in a new way, and the two facts are the same fact.

Associative learning has access to co-occurrence. It does not have access to causal structure, and there is no signal available inside the mechanism that distinguishes a real dependency from a coincidence that happens to be reliable. Skinner’s pigeons in 1948 acquired elaborate stereotyped routines in front of a food hopper on a fixed timer, and from the pigeon’s position the routine was working.

That is the shape to take forward. The failure is not noise in the system. It is what the system does when it is running correctly on input that happens to be misleading.

Clever Hans#

We spend a long hour on a horse.

Pfungst’s 1911 investigation of Clever Hans is on the reading list not as a curiosity but because it established something the rest of the course depends on: the experimenter is part of the apparatus. The horse was genuinely reading something. It was reading the questioner, who was genuinely unaware of signalling anything, and no amount of sincerity on anyone’s part prevented the result.

Every method you meet from here on — blinded scoring, preregistration, the adversarial collaborations in week 12 — is a descendant of that finding. It’s also why Markel’s blinding in the pea plant study, from last week, is not a technicality.

In the room#

You will be given three published animal cognition results and asked to design the Clever Hans objection to each. Two of them have one. Finding it is easier than you expect, which is the uncomfortable part.

Reading#

The paper this seminar is built around:

• Skinner (1948). ‘Superstition’ in the pigeon. Journal of Experimental Psychology.

The rest of the week’s reading:

• Kandel (2001). The molecular biology of memory storage. Science.
• Rescorla (1988). Pavlovian conditioning: it’s not what you think it is. American Psychologist — read this before the seminar. It is the clearest statement of what the mechanism actually has access to, by one of the two people who formalised it.
• Rescorla and Wagner (1972). A theory of Pavlovian conditioning. In Classical Conditioning II, Appleton-Century-Crofts — no DOI; the model matters more than the chapter.
• Pfungst (1911). Clever Hans (The Horse of Mr. von Osten). Holt — out of copyright and worth reading in full. It is a methods paper disguised as a story about a horse.

Grades are not attached to papers here, because a grade belongs to a claim rather than to the paper it appears in — see evidence grades. Several of these papers contain claims at two different grades.