Open any used textbook and you will find them: stripes of yellow and pink across half of every page. Highlighting may be the most widely used study technique in the world. It is cheap, it feels productive, and it leaves visible proof that you did something. According to the largest review of study techniques ever conducted, it is also one of the least effective ways to learn from what you read.
Ten techniques, one clear verdict
In 2013, John Dunlosky and colleagues published a monograph in Psychological Science in the Public Interest reviewing the evidence behind ten popular learning techniques, from summarization to practice testing (Dunlosky et al., 2013 (opens in new tab)). Highlighting and underlining landed in the bottom category, rated low utility alongside rereading. The authors' summary is blunt: in most situations examined, with most participants, highlighting does little to boost performance.
The pattern held across strikingly different groups. In a follow-up article for educators (opens in new tab), Dunlosky notes that highlighting has failed to help learners of every kind tested, including children, undergraduates, and Air Force trainees. If you have read our piece on why rereading doesn't work, the shape of the problem will be familiar: the most popular techniques are popular because they are easy and feel good, not because they build durable memory.
Sometimes worse than nothing
The more troubling finding concerns comprehension. Dunlosky's review flags one study in which students who highlighted while reading went on to perform worse on questions that required drawing inferences across the text. The likely mechanism is simple. Marking a sentence is a judgment about that sentence in isolation, so the act pulls your attention toward individual facts and away from how ideas connect. Inference questions ask precisely about the connections.
Dunlosky is careful not to demand anyone's highlighter back. He calls the tool a security blanket for reading and studying, and his advice is that highlighting is only the beginning of the journey: after you read and mark, the real studying still has to happen with more effective strategies.
The nuance the headlines skip
In fairness, the verdict has softened at the edges. A 2021 meta-analysis by Donoghue and Hattie (opens in new tab) put numbers on the same ten techniques and found underlining averaged an effect size around d = 0.44, a real if modest benefit, roughly half of what practice testing and distributed practice deliver. Their conclusion is gentler than the 2013 rating: even the weakest techniques are effective enough to keep in the toolbox. The same analysis adds a telling detail about quality: text marked by experts helped readers more than marking it themselves, and inheriting a previous owner's highlights helped least of all. Where the marks land matters as much as making them.
Strong claims in this literature deserve scrutiny in both directions. You may have heard of the 2014 Mueller and Oppenheimer study finding that handwritten notes beat laptop notes because typists transcribe verbatim instead of processing ideas. A direct replication by Morehead, Dunlosky, and Rawson in 2019 failed to reproduce that advantage (opens in new tab). The durable lesson from the note-taking debate is not about pens or keyboards. It is that shallow, verbatim capture does little for understanding in any medium. That is exactly what most highlighting is: capture without processing.
What separates marking from learning
Cognitive psychologists have a name for the missing ingredient: generative processing. In an influential review, Fiorella and Mayer identified eight strategies that promote generative learning (opens in new tab): summarizing, mapping, drawing, imagining, self-testing, self-explaining, teaching, and enacting. Every one of them forces you to reorganize the material and integrate it with what you already know. Highlighting appears nowhere on the list because it asks for none of that. It is pure selection: you flag information for a future self to process, and the processing rarely happens.
The techniques at the top of every ranking are generative through and through. Retrieval practice works because pulling an idea out of memory reshapes and strengthens it. Distributed practice works because returning to material after partial forgetting forces effortful reconstruction. A highlighter can feed those techniques. It cannot replace them.
How to mark a book so it actually helps
None of this means throwing the highlighter away. It means demoting it from a study technique to a filing system, then adding the processing the research says is missing.
- Read first, mark second. Finish the section, then decide what deserves ink. Marking on the first pass means marking before you can tell what matters, and discriminating selection is the one condition under which Dunlosky's review found marking can help.
- Ration it hard. Try one marked sentence per section. A page that is half yellow records no decisions, and the decision is the useful part.
- Turn every mark into a question. For each highlight, write a prompt on a note or a flashcard: what claim did this support, what evidence backed it? This converts passive marks into self-testing, the strategy the rankings put on top.
- Summarize the section in the margin, in your own words. Translation into your own phrasing is generative; copying and coloring are not.
- Come back on a schedule. Answer your questions from memory a few days later, then again after a longer gap. Marks you never revisit are decoration.
The verdict on highlighting is really a verdict on passive reading. The marks are not the learning. What you do with them is.