Highlighting is the most popular study method and the least effective — a combination that makes sense once you see what it trains. The yellow marker moves across the page, the important-looking sentences get flagged, and you feel like you're processing the material. What you're actually doing is marking recognition: "this looks important" is a judgment about the text, not a production of it. Close the page and the highlight's promise evaporates.

The fix is converting the highlight into a generated answer. For each yellow streak, write a question, hide the page, and produce the answer in your own words. The highlight marked what looked important; the generation proves you can reconstruct it. One is a souvenir of the reading session; the other is the skill the exam tests.

Color is not memory. The page that looks like a rainbow isn't a study record — it's a collection of spots where you meant to come back and never did.

Why highlighting fails the exam

The Dunlosky review rated highlighting as low utility — not useless, but near the bottom of what works, because it trains the wrong skill. Recognition ("I remember this was important") is what highlighting produces; retrieval ("I can produce this point without the page") is what the exam requires. The gap between them is why students who highlighted the whole chapter blank on the test.

Generation is retrieval — producing the point without the text in front of you. It's the testing effect applied to your own markup: the highlight tells you where the content was, the closed page is where you find out if it stuck. Fluency from color is the illusion of competence — the page looks processed, the material feels familiar, and neither one means you can produce a single highlighted claim.

The highlight-to-question workflow

Use your existing highlights. A chapter you already marked up — the yellow streaks are the prompt list. Each one is a spot where you flagged something as important; now it's a spot where you prove you can produce it.

Turn each streak into a question. "Legitimacy is not popularity" becomes "what's the distinction between legitimacy and popularity, and can you give an example?" The highlight was the marker; the question is the test.

Answer before opening. Hide the page, produce the answer in your own words — the claim, the contrast, the example. The highlight told you where to look; the closed page tells you whether it's there. If you can only say it while looking at it, the highlight was doing the remembering.

Name the fake-progress version. For each item, say what another color would have missed: "I highlighted the legitimacy-popularity line in yellow and pink, but can't produce the contrast." The contrast makes the gap visible — more colors is more recognition, not more memory.

Retest with a new highlight. Same chapter, different marked passage — the 1/3/7-day spacing holds, and the new prompt prevents memorizing the generated answer instead of learning to generate.

A 40-minute session on the clock

The weak answer is the highlight's promise broken — "legitimacy means people like the government" is the exact misconception the highlighted sentence was there to correct. The strong answer produces the distinction (popularity is approval; legitimacy is the recognized right to rule) and invents a new example. That's what the highlight was pointing at — and what couldn't be retrieved until you tried to generate it.

Why the page has to close

The highlight's entire function is external memory — it marks the spot so you don't have to hold it. Which means the highlight is only useful if you come back and retrieve; the marking itself stores nothing in you. A page full of highlights is a page full of debts: spots where you promised to return and learn, and never did.

Generation is the repayment. Each highlight becomes a question, each closed-book answer is the debt cleared — and the misses are where the real work is, because those are the spots where the "important" flag was recognition all along. The page closes because the exam's page will be closed; the highlight was never the point, the retrieval was.