Reasoning isn't one act of thinking harder—it evolves in stages.

When quick intuition feels uncertain, people stop to think more. Scientists wanted to know what actually happens during this extra thinking time and why it leads to better answers. The researchers used large language models to score recordings of people speaking their thoughts out loud. This tracked confidence and four mental jobs: coming up with candidate answers, justifying them, controlling intuitive responses, and regulating effort. Thinking followed a set rhythm. People generated answers early on. Later, they spent more effort justifying ideas, controlling impulses, and regulating their focus. Correct answers involved more justification, control, and regulation than biased answers. Also, feeling less confident prompted people to do more thinking, while regulating effort led to higher confidence later.

Why it matters

This matters because it shows that successful reasoning depends on the timing of different mental processes, not just effort. It reveals a predictable sequence from generating ideas to justifying and controlling them, which predicts whether the final answer is correct. Understanding this could improve how we teach critical thinking or design AI that reasons more like humans.

Questions answered