Closing · Everything That Flowed from One Model

In the Opening we began with one riddle and one discontent. The riddle that a smoothly read text does not, in fact, remain; and the discontent that learning methods always come as a scattered list of tips. Instead of memorizing the tips, we promised to build one model that breeds them. Now, tracing that promise back, we string in one line whether all those many things really flowed from one.

The Model Strung in One Line

We placed in the head two stores opposite in character: a narrow, volatile working memory and a broad, long-lasting long-term memory. Between them we placed three processes: what working memory builds is written into the store, consolidated, and read again. What piles up in long-term memory is a network of representations, and on each representation two forces hang—retrieval strength, the degree of reaching now, and storage strength, which sets how long it lasts and how fast it revives.

From here on it was corollaries. From the one thing that the two forces interlock crosswise—that storage strength grows greatly only in a hard retrieval where retrieval strength has fallen—flowed the testing effect, spacing, and desirable difficulty. From the one thing that working memory is narrow came depth failure and expertise; from the three sharing the same resource came cognitive load and the levers that turn it. From retrieval groping long-term memory by surface cues alone came the failure of transfer, schemas, interference, and misconceptions. And from the learner being unable to read storage strength directly and feeling only working memory's ease came the false gauge that judges one's own learning in reverse.

How to Distrust the Surface

There is one thing this model pointed to again and again: the surface is not an index of depth. The smoothness during reading cannot measure what the text left in the head; the amount of time spent in practice cannot see what processing happened in that time; the fluency of something coming up easily reflects not storage strength but the retrieval strength of the moment. The things we readily reach for as measures of learning—smoothness, time, ease—happen all to be surface, while what actually remains as learning grows where that surface cannot show.

So to read depth, you must distrust the surface and take up a different ruler. Read smoothly does not mean it remained; held a long time does not mean it was mastered. What actually remained is revealed only by a retrieval test that covers the material and pulls it out by oneself. To trust the test instead of the feeling is the heart of governing one's own learning.

Open a Space, but Only as Much as Can Be Closed

This model tells a direction for writing and teaching, too. Understanding and learning are not handed over intact, packed in text or speech, but assembled inside the one who receives. The situation model stands on the reader's inference that fills gaps, and storage strength grows when the learner endures a hard retrieval. Both happen only when the receiving side does it themselves. So teaching well is less a matter of spoon-feeding everything than of leaving a space for the receiver to assemble themselves.

But a heavy condition attaches to this prescription: open a space, but only as much as the learner can close. A gap grows a situation model only for a reader who has the background knowledge to fill it; to a reader with no material to fill it, it is just an empty hole. A hard retrieval grows storage strength only when the learner can finally succeed; cross that boundary and it not only fails to make a path to reinforce but shaves the sense of competence and cuts engagement. So a good gap, a good difficulty, is one fitted exactly to the learner's current state. Spoon-feed too much and there is nothing to assemble and nothing remains; leave it too empty to manage and they cannot assemble and it collapses. To open exactly that much in the narrow band between—just what the learner can fill and recall and close themselves—is the place this model points to.

Once You Hold the Model

Now, whenever you meet a new learning method, instead of receiving it as an item to memorize, you can ask it this. Which of the two forces does this push, and in which direction? Does it raise only retrieval strength for a moment, or does it grow storage strength? Does it gather resource into synthesis, or scatter it in vain? Does it bind to a surface cue, or detach into structure? If you can answer this, when that method works and when it does not follows from the model.

How do people understand and learn? And do you, right now, hold a situation model of how people understand and learn?