Curiosity pays before the answer arrives

What the scans showed

Kang and colleagues put people in a scanner and showed them trivia questions they could not answer.

Regions associated with anticipated reward, including the caudate, showed increased blood-oxygen-level-dependent signal for the questions people rated as most curious. That rise appeared while they were reading the question, before any answer arrived.

Whatever else curiosity is, it registers in the brain during the not-knowing, not only at the moment of finding out.

Where curiosity peaks

Kang and colleagues also found the relationship was not a straight line. Curiosity was highest when people were moderately confident about their guess, and lower at both ends — when they had no idea, and when they were sure.

That matches what George Loewenstein proposed decades earlier. On his account, curiosity is a response to a gap between what you know and what you want to know, so it needs some knowledge to exist at all. A gap you cannot perceive is not a gap.

If that is right, it says something about how a question has to be built. It cannot land on someone who has no footing in the subject, because they have nothing for the gap to open between. It also cannot land on someone who already has the answer. It has to arrive at the person holding one half of something.

The part that surprised me

Gruber and colleagues ran a study that went further.

People saw a trivia question, rated their curiosity, then waited. During that wait, an unrelated photograph of a face appeared. Then came the answer.

Afterwards, people remembered the faces better when they had been more curious about the question the face happened to follow. The curiosity was about the trivia. The memory advantage showed up on the face.

The authors link this to activity in the dopaminergic midbrain and the hippocampus during those anticipation windows.

What I take from it

Here the evidence stops and my reading of it starts.

A feed and a rabbit hole can feel identical while you are inside them. Both are wanting. But a feed is built out of items that resolve — each one finishes, and nothing points at the next.

A rabbit hole is built the other way. Each answer opens another gap, so the anticipation the studies describe keeps getting renewed rather than closed.

I do not know of work that has tested that comparison directly, and I would not claim the difference has been measured. But if anticipation is when the encoding advantage appears, then how often something reopens the gap is at least worth asking about — and it is the difference I notice an hour later, in what I still have.