Why nothing sticks
Two things conspire. The first is fluency: text you're reading feels understood while it's in front of you, and that feeling doesn't distinguish between comprehension now and retrievability later. The second is that reading has no failure signal. You can read four pages while thinking about something else and arrive at the bottom with no indication anything went wrong.
Compare that to writing a summary from memory, which fails loudly and immediately. The intervention that works is always the one that converts silent failure into loud failure, early enough to do something about it.
Four interventions, ranked by cost
| Intervention | Cost | What it buys |
|---|---|---|
| Question before reading | 60 seconds | Turns passive reading into a search. Attention has a target. |
| Close-and-recall per section | 2 minutes per section | The single highest-yield habit. Converts reading into retrieval. |
| Explain the section out loud | 3 minutes | Finds the sentence you can only say in the author's words. |
| Convert to questions afterwards | 10 minutes per chapter | The only step that makes the reading survive past this week. |
None of these is a reading technique. All four are output steps wedged between reading steps, and that's the whole trick.
The method, applied to one chapter
- 1
Spend one minute predicting before you read anything
Read the headings and the first line of each section, then write down three questions you expect the chapter to answer. Reading with questions is a search; reading without them is a scroll. This is the cheap version of the survey step in SQ3R.
- 2
Read one section — not the chapter — then stop
Sections are the natural unit. A chapter is too long to hold and a paragraph is too short to be worth recalling.
- 3
Look away and say what it said, out loud, in your words
Twenty to sixty seconds. Out loud matters: subvocal summarising is much easier to fake, and you'll let yourself get away with a vague gesture at the meaning. If you stall, you've found the sentence you didn't understand.
- 4
Write one line — the claim, not the topic
"Supply shocks raise prices and lower output, unlike demand shocks" is a claim. "Supply shocks" is a topic and is worth nothing later. One line per section, no more, or you're transcribing.
- 5
At the end of the chapter, answer your three questions from memory
Closed book. This is the chapter-level retrieval and it takes four minutes. What you can't answer is what to re-read — specifically that, not the whole chapter.
- 6
Convert the claim lines into questions the same day
Each claim becomes a question with a single answer. Now the reading is reviewable, which is the difference between remembering it this week and remembering it in the exam.
Note-taking that helps versus note-taking that doesn't
Copying while reading is the most common and least useful habit. It feels like engagement, produces a document, and requires no processing — the words go from page to page through your hand without stopping anywhere useful.
- Notes written with the book closed are worth roughly ten times notes written with it open. The closure is the entire mechanism.
- Notes in your own words force translation, which requires understanding. Verbatim notes require only transcription.
- Notes shaped as questions get reused. Notes shaped as summaries get read once and never opened again.
- Fewer notes is usually better. Ten good lines per chapter you'll actually revisit beats four pages you won't.
- Highlighting is not note-taking. It marks the text as important and does nothing to encode it — why highlighting doesn't work covers the evidence.
Different material, different reading
| Material | Read for | The output step that fits |
|---|---|---|
| Textbook chapter | Structure and mechanism | Close-and-recall per section; convert to questions. |
| Journal paper | Claim, method, limitation | Four-line summary: what they asked, did, found, and why it might be wrong. |
| Legal case | Facts, issue, holding, reasoning | Case brief from memory — see how to read a legal case. |
| Novel or primary source | Argument and evidence you'll cite | Quotation log with one line on what each quote proves. |
| Lecture slides | The examinable claims | Straight to questions; slides are already compressed. |
Reading speed is the wrong target
Speed-reading courses mostly train you to move your eyes faster, which improves how many pages per hour you cover and does nothing for retention — often less than nothing, since comprehension drops and the output steps get dropped first.
The honest optimisation is reading less, chosen better. A ranked reading list where you read the top three properly beats twelve read fast, and for most courses the marginal reading contributes far less than the marginal retrieval. Reading list strategy is about making that choice deliberately rather than by running out of time.