Stop Rereading and Start Retrieving

In 2006, Henry Roediger and Jeffrey Karpicke at Washington University in St. Louis gave college students a short science passage and extra time to study it. Some spent that time rereading the passage. Others read it once and spent the rest of the time writing down what they could remember with the passage out of sight. A week later, the students who had practiced recalling it remembered about 61% of the passage, and the students who had reread it remembered about 40% (Roediger & Karpicke, 2006). The rereaders had also felt more confident that they would remember it.

Rereading won for five minutes.

Both groups worked with the same material for the same total time. The only difference was how they used it. The students then took a final test five minutes, two days or one week later.

At five minutes, the rereaders were ahead, at roughly 80% against 70%. By the one-week test the result had reversed. Out of every hundred ideas in the passage, the students who practiced recall kept about 21 more than the students who reread.

Most students study the way the rereading group did. When Karpicke, Andrew Butler and Roediger surveyed college students about how they study, 84% said they reread their notes or textbook, and about one in ten said they tested themselves (Karpicke, Butler & Roediger, 2009). In physical therapy education I have watched very hard-working students do exactly this in the weeks before a licensing exam. When those students struggled, the cause was usually how they studied.

Rereading feels like learning because it is easy.

The second time through a page, the sentences are already familiar and your eyes move quickly. Researchers call that ease fluency, and people tend to read it as a sign that they know the material. Asher Koriat and Robert Bjork described the result as an illusion of competence: the conditions that make something easy to read during study are different from the conditions that make it easy to recall later, and the difference can’t be felt from the inside (Koriat & Bjork, 2005).

I think of it as the difference between recognizing a face and remembering the name that goes with it. Recognition is quick and effortless. Recall means searching a large, disorganized store and coming back with one specific thing. Rereading practices recognition, while an exam question or a clinical instructor’s question asks for recall.

Two kinds of memory strength.

Robert and Elizabeth Bjork at UCLA proposed that every memory has two separate strengths. One is how well the memory is learned. The other is how easily it comes to mind at this moment. Rereading raises the second for a short time and does little for the first, which is why the rereaders led at five minutes and fell behind at a week.

The Bjorks also proposed that successfully retrieving a memory strengthens it, and that a harder retrieval strengthens it more. Mary Pyc and Katherine Rawson tested that directly and found that more difficult correct retrievals led to better memory later (Pyc & Rawson, 2009). The difficulty has to end in success, though. Robert Bjork called these desirable difficulties: struggle helps when you eventually reach the answer or are shown it afterward. Struggle that ends without the answer teaches very little.

Students can’t tell which method is working.

In 2008, Karpicke and Roediger taught college students 40 Swahili words and their English meanings. Once a student had recalled a word correctly, one of four things happened to it: it stayed in both study and testing, it was dropped from study only, it was dropped from testing only, or it was dropped from both. A week later, the two groups that kept testing remembered about 80% of the words. The two groups that stopped testing remembered about 35%. Studying a word again after getting it right added almost nothing, while testing it again more than doubled what students kept. Students in all four groups predicted about the same score, and their predictions had no relationship to how they actually did (Karpicke & Roediger, 2008).

A fair objection is that rereading is a weak comparison. Karpicke and Janell Blunt compared recall practice with concept mapping, a well-regarded method in which students diagram how the ideas in a text connect. A week later, the recall group did better on factual questions and on questions that required inferences the text never spelled out. They did better even when the final test asked them to build a concept map (Karpicke & Blunt, 2011).

The effect holds up outside the laboratory.

Christopher Rowland pooled the research comparing testing with restudying and found a clear advantage for testing (Rowland, 2014). In practical terms, switching a student at the middle of a class from rereading to self-testing would move them to roughly the top third. A later review of the research found a somewhat larger benefit (Adesope, Trevisan & Sundararajan, 2017). Effects of that size are uncommon in education research.

Classroom studies conducted with middle school students in Columbia, Illinois, found the same pattern on the teachers’ own chapter and semester exams: material that had been quizzed during the unit was remembered better than material that had not (Roediger et al., 2011; McDaniel et al., 2011). Andrew Butler showed that the benefit extends to new questions that apply the same ideas, including questions from a different subject area (Butler, 2010). And in middle and high school classrooms that used frequent low-stakes quizzes, most students reported that the quizzes made them less nervous about unit tests (Agarwal et al., 2014).

The one thing to do.

At the end of the next chapter you read or lecture you attend, close everything. Take a blank sheet of paper and write down everything you can remember for five to ten minutes, in whatever order it comes back. Then open the book and check what you missed.

The checking is half of the method. Getting something wrong and then seeing the right answer is useful: Nate Kornell and colleagues found that a failed attempt to recall an answer, followed by the correct answer, improved later memory (Kornell, Hays & Bjork, 2009), and errors made with high confidence are the ones people correct most reliably (Butterfield & Metcalfe, 2001). The blank page will feel worse than rereading. That feeling is not a reliable guide to how much you are learning.

If what stops you is staying with something while you are still bad at it, the post on the week-three wall covers that.

What this does not show.

Testing yourself without checking the answer loses much of its value when you can’t recall the item. If you can’t remember something and never see the answer, you have mostly practiced not knowing it.

There is an open debate about complex material. Tamara van Gog and John Sweller argued that the benefit shrinks, and may disappear, when learners must hold several interacting ideas in mind at once, as in a multi-step clinical case (van Gog & Sweller, 2015). Karpicke and William Aue argued that the evidence does not support that conclusion (Karpicke & Aue, 2015). The effect is best established for facts and clearly defined concepts.

Timing also matters. Rereading usually wins in the very short term, so for a test 90 minutes away it is a reasonable choice. For anything you need weeks or years from now, it is a poor trade. Most of the laboratory studies above were conducted with college students and short texts, and John Dunlosky and colleagues’ review of ten common study techniques rated practice testing and spaced practice highest and rereading and highlighting among the lowest (Dunlosky et al., 2013).