How to Actually Remember the Things You Read

Most people remember almost nothing from the books they read. This is true even of books they loved, books they recommended at dinner, books they finished a month ago and could not now describe in three sentences if asked. Stop a thoughtful adult on the street and ask them to summarize a nonfiction book they read last year, and the typical answer is a blurry impression of theme, a fragment of one anecdote, and a feeling about the author. The pages went in. Very little stayed.

This is not a moral failure. It is a predictable result of doing the thing that feels like learning rather than the thing that produces learning. The two are not the same, and the gap between them is one of the best-replicated findings in cognitive psychology. Highlighting feels productive and is close to useless. Rereading feels diligent and barely moves the needle. The techniques that actually build durable memory are slightly uncomfortable in the moment, which is the main reason almost no one uses them. This piece is about what those techniques are, why they work, and what to do instead of highlighting.

The single most important finding in the memory literature

The finding has a boring name and very dramatic consequences. It is called the testing effect, or in its more general form, retrieval practice. The idea is simple. The act of pulling information out of memory does more to strengthen that memory than the act of putting information in. Reading a passage three times is less effective than reading it once and then trying to recall it twice. The retrieval is the lesson. The reading is the setup for the lesson.

The canonical paper is Henry Roediger and Jeffrey Karpicke in 2006 in Psychological Science, titled Test-Enhanced Learning. Students studied short prose passages under two conditions. One group reread the passage four times. The other group read it once and then took three short recall tests. Five minutes after the session, the rereaders did slightly better. A week later, the testers remembered roughly fifty percent more. The pattern repeats in study after study. The group that feels like it is learning more often learns less, and the group that feels like it is struggling often learns more.

Karpicke and Janell Blunt extended this in 2011 in Science with a paper that quietly stunned the education research community. They compared retrieval practice against elaborative concept mapping, which had been considered a gold-standard study method for decades. Retrieval practice won, not by a small margin, and not on a trivial measure. Students who closed the book and tried to write down what they remembered outperformed students who built elaborate concept maps of the same material on meaningful inference questions one week later. The students themselves predicted the opposite outcome. They thought the mapping had taught them more. The test did not agree.

Why retrieval works, in one sentence

Memory is not a recording. It is a route. Every time you successfully travel a route in your head, the route gets a little easier to travel next time. Rereading does not make you walk the route. It hands you a map and asks you to look at it. Retrieval makes you walk it, which is the only thing that paves it.

Conditions of learning that make performance improve rapidly often fail to support long-term retention, whereas conditions that introduce difficulty often produce learning that lasts.

The generation effect

A close cousin to retrieval is what Norman Slamecka and Peter Graf called, in 1978, the generation effect. In their original experiments, participants either read word pairs (read: hot, cold) or generated the second word from a cue (generate: hot, ?). The generators remembered the pairs better, sometimes substantially so, even though they had spent the same amount of time on the task. The act of producing the answer, rather than receiving it, was the difference.

The practical consequence for reading is that any time you can rephrase a passage in your own words rather than simply absorb it, you are running a small generation task, and the material is more likely to stick. A reader who closes the book at the end of a chapter and tries to explain the argument aloud to an imagined skeptic is generating. A reader who underlines a sentence and moves on is not. The first feels like effort. The second feels like progress. The first is the one that works.

Elaborative interrogation: the why question

Michael Pressley and colleagues showed in 1989 that a deceptively small intervention produced large gains in retention. They called it elaborative interrogation. The intervention was to pause and ask, of any fact you encounter, the question: why would this be true? Not in a debunking spirit. In a connecting spirit. The act of generating a plausible explanation links the new fact to a network of things you already know, and the link is what makes the fact recoverable later.

John Dunlosky and colleagues, in a major 2013 review in Psychological Science in the Public Interest, surveyed the entire literature on study techniques and ranked them by evidence. Elaborative interrogation came out near the top, alongside retrieval practice and distributed practice. It is also one of the most portable techniques in the set. It costs nothing. It requires no flashcards, no app, no schedule. It is a small habit of pausing and asking why, applied to whatever you are reading, and it disproportionately rewards the reader who develops it.

Interleaving: mix the topics, even when it feels worse

The intuition most people bring to study is to block. Practice one type of problem until you have it, then move on to the next type. This is what nearly every textbook does, and it is what nearly every student does. It is also, on a careful look at the evidence, the wrong move.

Doug Rohrer and Kelli Taylor showed in 2007 that math students who interleaved different problem types within a session learned less during the session and remembered far more on a delayed test. Nate Kornell and Robert Bjork showed in 2008 that the same pattern held for painters: students learning to identify the styles of unfamiliar artists did better when they studied paintings in a mixed order than when they studied them in blocks by artist, even though almost every student in the study predicted, after the fact, that blocking had worked better. Their belief was wrong. The mixed condition won.

The reason appears to be that interleaving forces you to discriminate between similar cases. Blocked practice lets you settle into one mode and apply it on autopilot. Interleaved practice keeps asking which mode this is, which is the same question the real world will ask later, when the practice is over and the problem arrives without a label.

Levels of processing: depth beats repetition

Fergus Craik and Robert Lockhart proposed in 1972 that memory is a side effect of processing depth. Shallow processing, like noticing the typeface of a word, leaves little trace. Deep processing, like considering the meaning of a word and connecting it to something else, leaves a strong one. The level at which you process the material is a better predictor of how well you remember it than how many times you saw it.

The framework is a useful diagnostic for any study habit. Rereading is mostly shallow. Highlighting is shallow plus a small motor action. Summarizing in your own words is deeper. Explaining a concept to someone who does not understand it is deeper still. Trying to apply it to a problem you care about is the deepest end of the pool, and it is also where most of the memory is made.

The Bjork principle: if it feels easy, you are probably not learning

Robert and Elizabeth Bjork, working at UCLA across several decades, gave the whole field its organizing slogan. They called it desirable difficulty. The idea is that techniques that feel harder in the moment, that produce more errors, slower performance, and more frustration, often produce better learning that lasts. Techniques that feel smooth, fluent, and confident in the moment often produce learning that evaporates within a week.

This is the deepest reason highlighting persists. It feels good. The page looks more studied after a session with a yellow marker. The reader feels diligent. The brain has done almost no work. The principle does not say that all difficulty is desirable. Some difficulty is just confusion. The kinds that help are the ones that force retrieval, generation, interleaving, and depth. Those are uncomfortable in a specific, useful way, and the discomfort is a signal that memory is being made.

What does not work, briefly and honestly

The Dunlosky 2013 review is the cleanest summary available. The techniques that the evidence does not support, at least not at the level most students believe, include highlighting and underlining, rereading, and summarization as commonly practiced. Highlighting on its own does almost nothing. Rereading produces a short-term fluency bump that fades within days. Summarization can help if it is done as a generation task from memory, with the book closed, but the kind most students do, copying or paraphrasing with the book open, mostly does not.

The trade-press summary of all of this is the 2014 book Make It Stick by Peter Brown, Henry Roediger, and Mark McDaniel. It is short, plainly written, and accurate to the literature, which is rare. If you want one volume to sit with, that is the one.

Further reading

The papers and books worth knowing: Roediger and Karpicke, Test-Enhanced Learning, Psychological Science, 2006; Karpicke and Blunt, Retrieval Practice Produces More Learning Than Elaborative Studying with Concept Mapping, Science, 2011; Slamecka and Graf, The Generation Effect, Journal of Experimental Psychology: Human Learning and Memory, 1978; Pressley, Wood, and Woloshyn, on elaborative interrogation, 1989; Dunlosky, Rawson, Marsh, Nathan, and Willingham, Improving Students' Learning With Effective Learning Techniques, Psychological Science in the Public Interest, 2013; Rohrer and Taylor on interleaving in mathematics, 2007; Kornell and Bjork on interleaving in inductive category learning, 2008; Craik and Lockhart, Levels of Processing: A Framework for Memory Research, 1972; and the Bjorks on desirable difficulty across many papers from the 1990s and 2000s. For the trade-press summary: Brown, Roediger, and McDaniel, Make It Stick, Harvard University Press, 2014.