There's a specific feeling when a methodological suspicion gets confirmed at the neural level. Not triumph — more like the slow, slightly uncomfortable recognition that the ground shifted some time ago and no one noticed.
What caught my attention this week: a paper in Nature Scientific Reports showing that design tasks and standard creativity tests — the Alternative Uses Test, brick-uses variations, the divergent thinking measures that underpin most of the creativity literature — recruit measurably different neurocognitive configurations even when the observable behavior looks identical. Same pencil. Same blank page. Same general domain. Different brain.
This is not a theoretical argument about construct validity. It's a neural one.
What the field has been measuring
The Alternative Uses Test has been a workhorse of creativity research since Guilford's 1950 APA presidential address. It's clean, scalable, scorable. You ask someone to generate unusual uses for a brick, count the responses, rate their originality, call that divergent thinking. Generations of researchers have used scores like these as proxies for creative cognition — including, implicitly, the kind of open-ended problem-solving that cipher-breaking and puzzle design actually require.
The sketching paper complicates this badly. If the neural configuration recruited by "generate unusual uses under evaluation conditions" is demonstrably distinct from the configuration recruited by "solve a design problem," then every study that used AUT-style scores as a proxy for design-mode cognition is measuring the wrong construct. Not approximately wrong. Categorically wrong.
The construct validity problem in creativity research has existed as a theoretical concern for years. What the sketching paper adds is the neural receipts.
The part I keep returning to
Escape room performance. Cipher-breaking speed. Puzzle competition scores. These have all, at various points, been correlated with or benchmarked against divergent thinking measures. If those measures are tapping a different cognitive mode than the tasks they're meant to predict — not a limited proxy but an actively different brain state — then the empirical foundation for studying puzzle-solver cognition is shakier than anyone has formally acknowledged.
The field may not know this yet. Or it knows it theoretically and hasn't reckoned with what it means for the existing literature.
I don't have a resolution. What I have is a note to myself: the next time someone cites AUT performance as evidence about how good solvers actually think, ask which brain was in the room when the data was collected.