retrieval practice as the learning event, not the measurement
Testing yourself is not a way to check what you already know. It is a way to change what you know. Each time you pull a fact out of memory without prompts, the neural pathway to that fact is rebuilt and slightly strengthened, which is why retrieval practice consistently outperforms rereading in retention studies.
When a memory is retrieved, it briefly enters a labile state before being stored again, a process called reconsolidation. The re-stored version tends to be more durable than the original, particularly when retrieval required effort. Rereading skips this step. The information is re-encountered, recognised as familiar, and re-encoded in roughly the same form. Familiarity increases without the underlying trace getting stronger, which is why a third rereading feels even more comfortable than the first and yet predicts almost no improvement on a later test.
Recognition can be done by matching the surface form of a sentence; retrieval requires rebuilding the meaning from scratch. A useful test is to cover your notes and try to say the central claim of a paragraph in your own words. If you can, you have retrieved it. If you can only nod along when the page is open, you have recognised it. The comfortable feeling that comes from rereading is not evidence of learning; it is evidence of familiarity.
Testing only beats studying when the test is followed by feedback. A wrong attempt alone is mildly useful because it flags a gap, but a wrong attempt that is then corrected produces the largest gains. The failed retrieval marks the missing information as important, and the correction lands in a brain that has just been primed to notice it. Practising questions without ever checking the answers mostly exercises confidence, not accuracy. Free recall followed by checking is the core loop; both halves matter.
Suppose you have a chapter on cellular respiration in biology. After reading a section on the Krebs cycle, close the book and write down every intermediate molecule you can name, in order. Check against the text. Anything missing goes back in. Move to the next section, then at the end of the chapter, quiz yourself on the earlier sections alongside the new one. Over an evening, total reading time may drop to half, while time spent actively recalling roughly doubles. A week later, the recall-heavy session is the one that still holds up.
Retrieval is a weak tool when you are meeting material for the very first time and have no schema to anchor it to. A first read, even a sloppy one, is necessary before testing helps. It also does little for skills still being assembled, such as the first attempts at a new class of physics problem, where worked examples do more until the basic pattern is in place. Once the pattern exists, retrieval takes over. Testing is not a replacement for initial exposure; it is what you do after the exposure to make it stick.