How to take notes from a textbook without copying it out

Read one section, close the book, and write the main points from memory in your own words. Then reopen the page, fix what you missed, and turn each heading into a question you can answer later. Highlighting and rereading feel productive but rate low in learning research; testing yourself rates high (Dunlosky et al., 2013).

The short version of the studies below:

Should you take notes while you read, or after each section?

After each section, from memory. If you write while your eyes are on the page, you end up copying sentences, and copied notes do worse than notes in your own words.

The clearest test is a 2009 study by Mark McDaniel, Daniel Howard and Gilles Einstein. They compared rereading, taking notes, and a method they called read-recite-review, or 3R: read the passage, set it aside, recite everything you can remember, then read it once more to check. In two experiments with college students, 3R beat both rereading and note-taking on free recall, right away and a week later. With longer passages on engineering topics, it matched note-taking on multiple-choice and problem-solving questions, and it took less study time.

So can you read a textbook and take notes at the same time? You can, but it pulls you toward copying. A 1979 experiment by Bretzing and Kulhavy had students take summary, paraphrase or word-for-word notes on the same passage, and the groups that had to work out the meaning did better on the tests. Those were high school students, so treat it as a nudge, not proof.

The one thing worth writing during the first read is a page number next to anything you didn’t follow. It works like the timestamps in the chunk-and-write method for video lectures: it tells you where to look when your recall comes up short.

What’s the most effective way to take notes from a textbook?

Turn each section into questions, answer them from memory, and check your answers against the page. In order:

1. Before you start, read the review questions at the end of the chapter, or your professor’s study guide. Don’t answer them yet.
2. Read one section, from one heading to the next, pen down.
3. Close the book. Write what you remember in your own words: the main idea, the terms, any numbers, and how they connect.
4. Open the book and check. Add what you missed and fix what you got wrong in a second color, so your gaps stay visible.
5. Turn the heading into a question and write it above your notes.
6. A day or two later, answer those questions with the book shut. Then check again.

Step one works, with a limit. A 2025 meta-analysis by King-Shepard and colleagues found that questions answered before studying improved learning of the material they asked about and did essentially nothing for the rest of the reading. So the end-of-chapter questions earn their two minutes for the points they cover, and no more.

Steps three and six are where the learning happens. In a 2006 experiment by Henry Roediger and Jeffrey Karpicke, college students either studied a prose passage once and then tried to recall it on three tests, or studied it four times. A week later, the recall group remembered 61% of the passage and the rereading group 40%, even though the rereaders had read it 14.2 times against 3.4. It holds in real classrooms too: across 222 studies and 48,478 students, a 2021 review by Yang and colleagues found that quizzing raised achievement by a medium amount.

Step six is also spacing. In a 2008 study of more than 1,350 people, Cepeda and colleagues found that when the test was a week after the review, the best gap between first study and review was about 20 to 40% of that week. In practice: come back to a section one to three days after your first pass, not the night before the exam. (The translation into days is mine.)

What does this look like on a real chapter?

Take section 8.1, “How Memory Functions,” from the memory chapter of Psychology 2e, a free intro psych textbook from OpenStax at Rice University.

Before reading, open the chapter’s Review Questions. One of the 12 asks for the three functions of memory, so watch for a list of three. Copy the section’s learning objectives down as questions too.

Now read the first subsection, “Encoding,” and close the book. Your recall might come out like this:

Encoding = getting information in. Automatic processing (time, space, word meanings, no effort) vs effortful processing (new material). Three types: semantic, visual, acoustic. Meaning sticks best.

Reopen the page and check what you left out. Two easy ones to miss are the experiment behind “meaning sticks best” (Craik and Tulving’s study with words) and the self-reference effect: information you connect to yourself is easier to remember. Add both in your second color.

“Storage” is where the section’s key diagram lives: Figure 8.4, the Atkinson-Shiffrin model. Don’t copy it. Look at it for a minute, close the book, and draw it from memory, then compare. Check the details that are easy to drop, like the rehearsal loop on short-term memory and the exits where information that isn’t passed along is lost.

Short-term memory comes with numbers, and the book gives two for its capacity: Miller’s 7 plus or minus 2 from 1956 and Cowan’s more recent 4 plus or minus 1. Write both with the names attached, plus the 15 to 30 seconds the section gives for how long short-term memory lasts. Almost right is wrong on a multiple-choice question.

The last subsection, “Retrieval,” separates recall, which has no cues, from recognition. That’s the difference between this method and rereading: rereading asks you to recognize the page, and closing the book makes you recall it.

For the whole section you end up with about a page: five or six questions, short answers in your own words, one diagram drawn from memory and a few lines in the second color. That page is what you study from next week, not the chapter.

Does highlighting count as taking notes?

Not really. It helps a little with remembering and not with understanding. A 2022 meta-analysis of 36 articles by Ponce, Mayer and Méndez found that highlighting you do yourself improved memory a little but not comprehension. The memory benefit showed up for college students, not for younger students.

Highlighting done for the reader, where someone who knows the material has already marked what matters, improved both memory and comprehension. The closest thing in a textbook is what the authors already marked, the bold key terms and boxed definitions. Read those carefully before you add your own colors.

If you do highlight, do it after the recall step and only on what you missed, so the color marks your gaps instead of half the page. The Dunlosky review rated highlighting and rereading low utility because they don’t consistently improve performance, and recommended practice testing instead.

Is rereading the chapter a waste of time?

Rereading straight away mostly is. Rereading a week later is not.

In four experiments with textbook chapters and a Scientific American article, Aimee Callender and Mark McDaniel summed up the result: “With only several exceptions, rereading did not significantly increase performance on the assessments.” Timing changes the picture. When Katherine Rawson and Walter Kintsch had 423 college students read a text once, twice in a row, or twice a week apart, rereading in a row helped only on an immediate test. On a test two days later, only the spaced rereading beat a single read.

So why does rereading feel so productive? Because it wins the only test you give yourself while doing it, the one five minutes later. In Roediger and Karpicke’s experiment, the students who only reread recalled more than the testing group after five minutes (83% against 71%), and they were the most confident they’d remember the passage in a week. A week later they recalled the least.

Most students run on that feeling. When Karpicke, Butler and Roediger surveyed 177 college students, 84% listed rereading as a study strategy and 55% said it was their number one. Only 11% (19 of 177) said they practiced recalling the material. The authors’ name for this is “illusions of competence.” One caveat: the survey was done at Washington University in St. Louis, where students averaged above 1400 on the SAT.

Does the SQ3R method actually work?

Parts of it do. As a complete system, its evidence is much thinner than its reputation.

SQ3R stands for Survey, Question, Read, Recite, Review, and educators have recommended it since the 1940s. A 1980 review found that support for it rested more on opinion than on evidence, and a 1982 review found it no better than students’ own methods. A 1989 study with an intro psych textbook found the opposite.

The most recent head-to-head test I found, by Jairam and colleagues in 2014, pitted SQ3R against a newer system called SOAR. SOAR students learned 20% more relationships, 14% more facts and 13% more concepts. Two things to weigh: the study had 25 undergraduates, and SOAR was developed by one of its authors, Kenneth Kiewra.

ERIC, the federal index of education research, returns 142 records for SQ3R. Ninety-eight of them are from before 1990, and only 20 are tagged as research reports. I looked for a recent, well-powered trial of SQ3R as a whole package and didn’t find one.

What survives is the part that overlaps with stronger evidence. Recite and Review are retrieval practice under older names, and Survey and Question work like the pre-questions above. Keep the questions and the reciting, and skip the ritual of doing all five steps in order.

How long should textbook notes be?

Short enough to quiz yourself from in one sitting, which for most sections means under a page. I couldn’t find a study that sets a right length, so that number is mine, but one finding points the same way.

In a 2005 study by Igo, Bruning and McCrudden, college students who could paste only a small amount of text at a time into a note chart recalled more facts, recognized more concepts and inferred more relationships than students who could paste as much as they liked.

Layout matters too. In two experiments with 153 college students, Robinson and Kiewra found that students who studied graphic organizers learned more of the relationships in a text than students who studied outlines, and applied the knowledge better. When a chapter compares things (types of memory, say, or cell organelles), make a table with the things as rows and their features as columns. The students in that study were handed their organizers, so building your own goes a step past what was tested, but an empty cell shows you exactly what you haven’t learned.

A quick test: if your notes for a section are nearly as long as the section, you’re copying.

What do you do with worked examples and diagrams?

In a math or science textbook, cover the worked example’s solution and try the problem first. Then go through the book’s solution and explain each step to yourself: why this step, and why here.

The explaining part has good support. A 2018 meta-analysis by Bisra and colleagues pooled 69 effect sizes on prompting students to explain material to themselves, including while studying worked problems, and found a clear benefit on average. Generic “why is this true?” questions are shakier. In a 2016 study of 198 undergraduates reading a lesson on posterior probability, answering those questions led to worse learning than simply reading the lesson twice, possibly because the students’ answers were poor.

Diagrams are the part students skip. In a think-aloud study, Cromley and colleagues followed 91 beginning biology majors through an eight-page passage from their own textbook with seven complex diagrams, and the students “often skipped over or superficially skimmed diagrams.” Diagrams matter most when the topic is new to you: in three experiments with 300 college students, illustrations helped when they explained how a system works and readers knew little about it.

Treat a key figure the way you treated Figure 8.4 above: look, close the book, redraw it, compare. A 2018 review by Fiorella and Zhang found that drawing consistently beat reading alone for comprehension and transfer. If you type your notes, keep paper next to the keyboard. In a lecture study, Luo and colleagues found that laptop note takers recorded fewer images than longhand note takers.

Can ChatGPT take notes from a textbook for you?

It can, if you paste in the chapter. Whether that helps depends on whether you still do the remembering.

The strongest objection goes first: writing your own notes is where the learning happens, so a tool that writes them takes away the useful part. There’s real evidence for it. In a preregistered 2025 study, Etkin and colleagues had 195 US readers aged 18 to 22, recruited online, read passages from the ACT, some with an AI summary in place of the passage. Strong readers’ quiz scores fell from 82.5% with the passage to 66.5% with the summary. And in six experiments by Pan and colleagues, undergraduates who made their own flashcards from a text did better on a test two days later than students given premade cards, in five of the six.

The Etkin result needs two caveats. The quiz came right after reading, so it says nothing about next week. And readers were sorted into strong and weak by their score on the passage they read without AI, which exaggerates the gap, because an unusually good score tends to come down next time anyway. (That caveat is mine, not the authors’.) When they re-sorted readers by SAT or ACT score, the summary still hurt strong readers, by less than half as much. Weaker readers did better with the summary in the first split (44.9% to 55.4%) but not in the second. What helped them in both was a chatbot that asked them questions.

So the objection is right about replacement. Where it goes wrong is in assuming the notes were the learning. In McDaniel’s study, reciting from memory beat note-taking. In a 2011 Science paper, Karpicke and Blunt found that 101 of 120 students (84%) learned more from recalling a science text than from building a concept map of it, and 90 of them (75%) had expected concept mapping to work as well or better. In their words, whether a task is considered “active” is “not diagnostic of how much learning the task will produce.”

The practical rule: an AI-written note is useful as the answer key you check your own recall against, and as a source of questions you answer before you look. It does harm when it replaces the reading. A 2025 experiment with 405 English students aged 14 and 15 found the same split three days later: note-taking, with or without an AI chatbot, beat the chatbot alone. If you know you’d read the summary instead of the chapter, don’t get one.

What if your textbook is a PDF on your laptop?

Print the chapter if you can. If you can’t, keep your notes somewhere other than the PDF, and don’t copy and paste from it.

Paper has a small, real edge. A 2019 meta-analysis by Virginia Clinton of 33 studies with 2,799 readers found that reading from screens lowered reading performance a little compared with paper, most of all for informational text like textbooks. Screen reading wasn’t reliably faster, and readers judged their own understanding more accurately on paper.

Whether you then handwrite or type matters less than you’ve heard. The famous finding that laptop note takers did worse on conceptual questions (Mueller and Oppenheimer, 2014) didn’t hold up in direct replications by Morehead, Dunlosky and Rawson and by Urry and colleagues, and a 2024 meta-analysis of 24 studies found only a small edge for taking and reviewing handwritten notes. All of those were lecture studies (more in taking notes and listening at the same time). With a book you set the pace, so the risk isn’t falling behind. It’s copying.

If you have the chapter as a PDF you’re allowed to use, NoteAlong, which is our product, can turn it into a study note. The book’s license and your course’s rules still apply, and some textbook terms forbid AI tools, so check those first. Each page’s image stays in the note, figures are cut out of the page and placed where the note discusses them, and a citation like (p. 12) opens that page, so you can check any line against the book. The same note gives you flashcards and a quiz, and the cards export to Anki. On the free plan a note holds up to 30 pages, with 50 pages a month, which covers a chapter, not a whole book. Do your recall pass first, then check yourself against the note.

Where NoteAlong is the wrong tool

If your book lives only inside a publisher platform such as VitalSource Bookshelf, Pearson+, McGraw Hill Connect or Cengage MindTap, and you don’t have it as a file, there’s nothing to upload, and NoteAlong doesn’t try to get around a publisher’s protection. Read it there and use the method above with paper.

It’s also the wrong tool if your course bans AI tools or uploading course materials, so check the syllabus. A seminar where you annotate the reading together on Perusall, and your comments are the assignment, needs your own words. Some books’ terms rule it out too: OpenStax says its books may not be “ingested into large language models or generative AI offerings” without its prior written permission.

For a 10-page reading, skip the software. A pen and the read-recite-review loop cost nothing, and they’re what McDaniel’s study tested. A whole textbook on the free plan doesn’t fit either: a free note holds up to 30 pages, so it’s a chapter at a time. The limits are on the PDF notes page.

If you change one habit this semester, make it this one: close the book before you write anything down.