Definition
A Bongard problem is a visual classification puzzle: twelve boxed figures are presented, six on the left (Class I) and six on the right (Class II), and the puzzle is to discover the rule that distinguishes the two classes. The rule might involve shape, size, count, openness, convexity, or some subtler combination. Devised by Russian scientist Mikhail Bongard in the 1960s.
Why it matters
How it works
A solver scans the figures for candidate features (size, shape, color, count, openness, axis of symmetry). For each candidate, check whether the six left-side figures share it and the six right-side figures lack it (or vice versa). When you find one that perfectly separates the classes, you have a solution — but often the rule must be discovered, not selected from a fixed feature list. The cognitive move is invention of the right description, then verification.