Strain it to retain it

We tend to think of learning as a process of putting information into our minds. What we rarely notice is that pulling it back out is the part that makes learning last.

Try it now. Without looking, what was the last photo you took on your phone? Stay with the question for a few seconds before you check, and notice what happens: the blank, the half-memory, the guess, then the small correction when you look. That reaching, followed by the feedback, is retrieval practice, and it does more for durable learning than any amount of going back over the material.

Why it works starts with how memory takes anything in at all. We are not recording devices. New information lodges only when it is connected to what we already know, and making those connections is work the learner has to do. Rereading and highlighting feel like learning because they are smooth, and smooth is exactly what leaves that work undone.

Getting information back out is not playback either. Recall is a reconstruction, rebuilt each time from fragments, expectations and prior knowledge, which is why confident memories can contain details that were never there (Bartlett, 1932). Because recall rebuilds rather than replays, performing it changes what is being rebuilt. A memory that has been retrieved becomes more accessible afterwards than it would have been if left alone: retrieval is a memory modifier (Bjork, 1975). Time spent pulling information out produces more durable learning than the same time spent putting it back in (Roediger & Butler, 2011).

The active ingredient is effort. When recall is made so easy that success is guaranteed, by recency or by a cue doing most of the work, the benefit largely disappears (Landauer & Bjork, 1978). Difficulty is not a side cost of the method, it is the method. Even failing counts: eight seconds spent generating an answer that turns out to be wrong, followed by five seconds studying the correct one, beats thirteen seconds studying the correct answer alone (Kornell, Hays, & Bjork, 2009).

Over weeks the gap widens. Someone three weeks into a new job with forty names to learn can read the staff directory over and over, and it will feel like progress, because every pass is smooth and every face familiar. Then they meet a colleague away from their desk and produce nothing. Cover the names instead, try to produce each one, get a handful wrong and only then check: slower, more uncomfortable, and the names are still there a month later.

Most people do not see it this way. Asked why they test themselves, around seven in ten say it is to find out how well they have learned. Fewer than one in five say it is because testing teaches them more than rereading does (Kornell & Bjork, 2007).