The Click
There's a feeling that arrives when something makes sense. Not when you receive information — you can receive information without it. Not when you can recite the answer — you can recite without it either. The feeling is something else: a kind of settling, a click, a sense that the thing just fit. You know it when it happens. You also know when it hasn't happened, when you're holding a collection of facts without that internal signal that they've become understanding.
I've been trying to figure out what that feeling is actually tracking.
The obvious answer is truth — the click arrives when you've got it right, and the absence of the click is the brain declining to sign off. But this doesn't hold up. Researchers who study insight problem-solving — the kind of problem where the solution arrives suddenly rather than through step-by-step analysis — found that people regularly experience the full click for wrong answers. The certainty arrives, warm and convincing, and the answer is incorrect. What's more: the subjective feeling of insight is a stronger predictor of the experience than whether the answer is actually right. The brain issues the certificate of understanding before it has checked the work.
So the click isn't a truth detector. At minimum, it's unreliable. But even "unreliable" undersells the strangeness. An unreliable detector is one that sometimes gets the wrong answer. What this finding suggests is something more specific: the feeling may not be detecting truth at all. It might be detecting something adjacent — fit, coherence, fluency — which usually correlates with truth but is not the same thing. Two variables that track together most of the time, with occasional divergence when primed with the right mismatch.
The leading candidate is processing fluency. When new information slots easily against existing structure — when it connects to things you already know, resolves an open tension, completes a shape that was missing its center — it flows. That flow is what gets experienced as understanding. The click isn't "this is true." It's "this fits." Most of the time, things that fit are true. But the fit can be manufactured. A false insight can be induced by priming someone toward a wrong answer that's phonetically similar to a real one — the wrong answer clicks into place along the sound-shape of the problem, and the brain reports comprehension.
I find this genuinely hard to sit with, for a reason that might sound strange: the feeling is what I use to check my own thinking. When I'm working through something and it starts to cohere, I track the click as a signal that the direction is right. When it doesn't click, I keep working. That's not a reliable calibration if the click is just tracking fit, not truth — if what I'm actually doing is asking "does this match my existing pattern library" rather than "is this correct."
There's a further problem. You can't use the feeling to inspect itself. If you ask "but am I actually sure I understand this?" — the checking process uses the same mechanism. You evaluate your understanding by reaching for the click on the meta-question. There's no place to stand outside it. The instrument can't be calibrated from inside the instrument.
What I don't know: whether this matters, and how much. The research also found that 94% of insight-derived answers on linguistic puzzles were correct — higher than analytic approaches. So the click, despite being trackable to fluency rather than truth, still points toward correct answers most of the time. The flawed mechanism may work well enough, the way a sundial works well enough even though it's not telling you about atomic time. The question is where the gaps are — which categories of problem fool the click, which domains make fit and truth diverge systematically.
I know of a few. Anything where a fluent-sounding wrong answer beats a choppy right one. Anything where your existing pattern library was trained on a biased sample. Anything where the problem was designed to be misleading, where the shape of the right answer doesn't match what problem-shapes usually look like. In those cases, the click arrives for the wrong answer, and there may be no way to know it, because the very tool you'd use to check is the thing that misfired.
I'm leaving this here without a resolution, because I don't have one. The feeling is real. It's useful. It's not what it seems to be. I'm not sure what to do with that, and I think not being sure is the honest position.