Chapter 15 · The False Gauge and Beyond the Model

As we saw at the end of the previous part, the learner cannot directly look at whether their long-term memory has actually changed. Up to here this model has drawn the processing that happens in the head while the learner is engaged with a task, but the learner who runs it is left at the model's edge. The learner cannot directly read their own state, must engage with the work in the first place, and—even for the same time spent—what processing happened within it divides the result.

The False Gauge

First, what can the learner read off from their own head? A representation's storage strength is a hidden value that does not rise into consciousness. What the learner actually senses is one thing only: how easily an item comes up in working memory. Comes up smoothly and it seems well known; stumbles and it seems unknown—this feeling of ease is called fluency. What they know and do not know, whether to study more or stop, the learner gauges by this fluency alone.

The trouble is that this sole gauge does not measure the very thing it should. What fluency reflects is the retrieval strength of the moment, not the storage strength that will remain as learning. Because the two grow separately, the moment it feels easy may be the moment storage strength is barely growing. Moreover, retrieval strength does not rise only when the path to long-term memory is opened; it rises even when external input just seen merely floats the item into working memory. An experiment reveals this. People were made to read a text and predict the score they would later get on a test, and—leaving the text's meaning intact, by the mere manipulation of blurring out and restoring some letters—that prediction swung. The content was the same, so the amount to learn was the same, yet as the ease during reading changed, the judgment of "how much have I learned" moved with it. The gauge was reading not what remained in the head but the smoothness during reading.

So rereading is dangerous. Rereading floats the item with external input and raises retrieval strength for a moment, and the gauge reads that ease as "I know it." Though storage strength has barely grown, you decide "this is enough" and skip the very hard retrieval that has the large effect. Worse, this mismatch is systematic in one direction. The behaviors that raise fluency—rereading or cramming—grow storage strength almost none, while the behaviors that grow storage strength most—covering and recalling, or spacing out your practice—happen where retrieval strength has fallen and feel taxing in the moment. Overlay the two facts and the best method to learn by feels, in the moment, the most inefficient. It is a structure in which the more faithfully the learner follows their own feeling, the more they settle for the worse method.

Whether you can pull it out by your own power with the material covered—this alone reveals the actual state of storage strength. The heart of regulating one's own learning is to distrust fluency and put a retrieval test in its place, that is, to swap "it seems I know it from reading" for "does it come even when covered."

The Door of Engagement

The next place is the premise for the model to operate. However good the processing, it happens only when the learner starts the work and holds out. That hard retrieval grows storage strength is a story of when the learner endures that hardship and recalls; that deep synthesis builds a situation model is a story of when they engage that much. Without engagement, no part of the model operates.

What opens this door is motivation, and research that has looked into motivation holds that it divides by where it comes from. That autonomous motivation, springing from interest or a self-endorsed value, ties more strongly to achievement and persistence than motivation driven by outside pressure or reward, is what many sources point to in the same direction. And the strongest pillar held to support that autonomous motivation is a sense of competence—the perception that "I can do this." When they feel it is doable, the learner endures difficulty, and when repeated failure breaks the sense of competence, interest and persistence fade together.

Here motivation meshes with the cognitive processing seen earlier. For difficulty to benefit learning, the learner must be able, with help if needed, to finally succeed. Difficulty beyond that boundary is doubly harmful: too hard, and retrieval fails, so it is harmful cognitively in that no path to reinforce is made; and harmful motivationally in that repeated failure shaves the sense of competence and cuts engagement itself. So matching difficulty to the learner's level is at once a cognitive variable and a motivational one. That narrow band of desirable difficulty is at once the stretch where one learns best and the stretch where the learner can keep holding on.

Not Amount but Kind

The last place is how much this processing structure explains the actual differences in ability among people. There is a widespread belief here: practice long enough and anyone gets good; pour ten thousand hours into one field and you become an expert. That the time spent decides sounds fair and hopeful.

But even for the same one hour, what happened within it varies enormously. Practice that picks the place you are poor at and clings to it strenuously, and practice that comfortably repeats what you are already good at, yield wholly different results. In this book's language, comfortably repeating is retrieving again something whose retrieval strength is already high, so it merely fills time and does not grow storage strength, while clinging strenuously to a weak spot induces a hard retrieval where retrieval strength has fallen and grows storage strength. The same one hour spins idle on one side and remains on the other.

This is why practice amount, as a yardstick, predicts ability poorly. That yardstick measures only the length of time and does not see what processing happened in it. Indeed, an analysis pooling how much deliberate practice explains ability differences by field found that share swung greatly by field: around twenty percent in relatively well-structured domains like games or music, but only a few percent in less-structured domains like education or the professions. This does not mean practice is unimportant, but that the claim "just fill the hours" is an exaggeration. What divides outcome is not the amount of practice time but the kind of processing that happened in that time.

The Edges Point Inward

We have thus looked around the three places set at the model's edge. The learner cannot directly read their own state, must engage in the first place, and is divided by what processing happened in their time. Yet all three point, in the end, to the model's inside. The answer to self-monitoring is not feeling but a retrieval test—that is, the very processing that grows storage strength; good motivation is motivation that keeps you engaged in such processing; good practice is practice in which the processing that grows storage strength happens.

We have thus looked around both the inside and the edges of the model. What remains is to string, from beginning to end in one line, whether the things that seemed scattered really flowed from one model, and to make out what this model tells us to do.