What this measures
Choose a combination of ingredients and brew. The potion turns red if your combination satisfies a hidden rule, blue if it does not. The rule is never stated. Work it out, then produce it on demand.
This is rule induction from feedback — closely related to the Wason 2-4-6 task and to concept-learning experiments generally. It measures something the other games here do not: how well you design experiments.
The rules you might be facing
There are four ingredients, so sixteen possible combinations. The hidden rule is one of five kinds:
- Inclusion — a particular ingredient must be present
- Exclusion — a particular ingredient must be absent
- AND — two specific ingredients together
- OR — either of two ingredients
- Count — a certain number of ingredients, regardless of which
Only rules that produce at least three red and three blue combinations out of sixteen are used, so you always have enough signal to work with. Empty combinations do not consume an experiment, so you can test freely.
How to test properly — and how most people fail
The classic error, and the one Wason built his task to expose, is confirmation bias in experiment design: once you have a hypothesis, you test combinations you expect to succeed. Those tell you almost nothing.
- Change one thing at a time. If A+B+C is red and you want to know whether C matters, test A+B. Changing two ingredients at once produces a result you cannot attribute.
- Try to make it fail. A combination you expect to be blue is worth more than one you expect to be red, because a surprise there kills your hypothesis immediately.
- Count as well as identify. If two-ingredient combinations succeed and three-ingredient ones fail regardless of which, the rule is about number, not identity. People rarely check this.
- Hold more than one hypothesis. Early on, several rules usually fit. Pick the test that splits between them rather than the one that confirms your favourite.
Done well, three or four experiments narrow sixteen possibilities down almost every time. Done by intuition, five is often not enough.
Why this is the hardest game here for many people
Because it punishes the thing that usually works. Most tasks reward acting quickly on a plausible guess. This one rewards deliberately looking for evidence against your own guess, which is unnatural enough that Wason's original experiments found most educated adults failing it.
It is also the closest thing here to actual scientific reasoning, which is why variants appear in assessments aimed at analytical roles.
Frequently asked
Q. Can I brute-force all sixteen?
A. Not within the experiment budget. The task is designed so that information has to be
bought efficiently.
Q. I found the rule but scored low.
A. Then you spent too many experiments getting there. Score reflects efficiency, not just
the final answer.
Q. Are the rules the same each round?
A. No — a new rule is drawn each round, so there is nothing to memorise between attempts.
For a deduction puzzle with a fixed information budget, try ball weight ordering.