A mental model for learning hard things, built around the idea that "hard" is usually a property of the gap between where you are and what you're trying to learn, not a property of the topic itself. The exception is bucket #2 below, where the topic itself fights you.
The Decomposition
Four kinds of "hard." Each has a different fix. Misdiagnosing which kind you're facing is where most time gets wasted.
These aren't fully independent: schemas functionally expand working memory (chess masters don't have bigger working memory, they have bigger chunks), and restructuring often surfaces as overload because you're holding the old frame and the new frame simultaneously. Sometimes a working-memory problem is really a schema problem in disguise. The buckets are a diagnostic, not a taxonomy.
Two of the four are additive: prerequisites and schemas. Investment compounds: every layer you build stays, every shape you learn joins the library. Two are transformative: restructuring and working memory. They have to be applied when the situation calls for them, and they don't stockpile. The frame-shift you did last year doesn't help with today's new frame; yesterday's paper doesn't reduce the load on today's problem. Misdiagnosis usually goes one direction: reaching for an additive fix when the situation needs a transformative one, because additive work feels more like progress. "Read another textbook" (meaning another book that explains it the same way) is the most common wrong answer to a restructuring problem; "solve more problems" is the most common wrong answer to a working-memory problem. Both are additive moves applied to transformative situations.
1. Prerequisites: the stack underneath (additive)
Before you can engage with an idea, you need the concepts it's built on. When something reads as gibberish, the first question is whether you're missing a layer.
The hard part isn't climbing back up; it's figuring out where the floor actually is. Named prerequisites are easy to look up. The traps are the tacit ones: notation conventions, the implicit problem the field is trying to solve, the thing everyone in the discipline treats as obvious and never says out loud. If a topic feels opaque and you can't name what's missing, the gap is probably tacit.
Fix: Trace backward until you hit ground you understand, then climb. When you can't find the missing piece by yourself, read introductions written for outsiders, or ask someone fluent what the field assumes you already know. Most "hard" is this, but it's rarely the kind that makes you want to quit. That kind lives in #2 and #4.
2. Restructuring: the frame you think in (transformative)
Some ideas don't slot into your existing model; they require you to rebuild the model. Recursion, limits, thinking in distributions, thinking in types. Prerequisites don't dissolve this kind of hard; the old frame has to break before the new one takes hold.
This is the bucket people most often misdiagnose. The signature is: you can follow each step but couldn't have generated it; you understand the words but the idea keeps slipping; you find yourself silently translating the new concept back into your old frame and wondering why it doesn't quite work.
Fix: "Sit in the confusion" is true but not actionable. The concrete moves:
- Find multiple explanations of the same idea. Different framings break the old frame faster than one good framing. If a single textbook isn't clicking, that's a signal to read three more, not to read it again.
- Work small examples that force the new frame. You can't trace recursion iteratively if the example genuinely requires recursion. Pick problems the old frame can't solve.
- Look for the historical motivation. Why did anyone invent this? What was broken about the previous way of thinking? The motivation usually contains the dissatisfaction you need to feel before the new frame becomes attractive.
- Catch yourself staying in translation. Analogies to the old frame are how the new one first gets purchase: recursion-as-Russian-dolls is fine as a foothold. The failure mode is when you're still translating after the idea should have settled. Eventually the new frame should feel different on its own terms, not like a re-skin of the old one.
These moves work because they hit the four conditions a frame-shift actually requires (from the conceptual change literature): dissatisfaction with the old frame, intelligibility of the new one, plausibility that it could be true, and fruitfulness, seeing it open up problems the old frame couldn't. Multiple explanations buy intelligibility; historical motivation surfaces dissatisfaction; forcing examples test plausibility; the whole exercise eventually demonstrates fruitfulness. If any of the four is missing, the restructuring stalls.
3. Working memory: how much you hold while thinking (transformative)
Even with prerequisites and the right frame, an idea can have too many moving parts to juggle in your head at once.
Fix: Externalize. Write it down, draw the diagram, work the small example. Paper isn't a crutch; it's what fluent learners actually do. The shame around needing to write things down is the thing to attack: treat paper as a working-memory extension and use it without apology.
4. Schemas: the library of problem-shapes (additive)
Once you understand a topic, using it on new problems requires recognizing which problems are which shape.
Schemas are built by solving the simple version of a pattern first, then complex versions, ideally with some failure mixed in, so retrieval gets tagged with "this is the kind of thing I missed."
Variation matters more than volume. Twenty problems with the same surface features just train pattern-matching on the surface. Five problems where the surface differs but the underlying structure is the same is what makes the schema portable. If the problems all look alike, you're memorizing, not building schemas.
Fix: Practice cold, and across varied surfaces. See a problem, ask "what shape is this?" before attacking.
The Diagnostic Plan
When stuck, ask in order: what kind of stuck am I? Each signature points to a different fix.
| Signature | Kind of hard | What to do |
|---|---|---|
| Individual words or symbols don't make sense, or words make sense individually but sentences don't compose into meaning | Prerequisite problem | Trace backward |
| Even after you understand it, the way of thinking still feels foreign; you keep wanting to translate it back into a familiar frame and the translation doesn't quite fit | Restructuring problem | Bucket #2 moves |
| You can hold each piece individually but pieces fall out when you try to keep them all in your head; writing things down or working a small example shrinks the confusion | Working memory problem | Externalize aggressively |
| You couldn't have generated the solution cold, but once you see it, it feels obvious in retrospect ("I'd recognize this next time"); theory is clear but you can't see which problems call for it | Schema problem | Solve varied problems cold |
If two diagnoses fit, try the cheaper fix first; externalizing costs nothing and often reveals which bucket you're really in. A "schema problem" sometimes turns out to be working memory once the pieces are on paper; a "restructuring problem" sometimes turns out to be a missing prerequisite once you try to write it down.
The metacognitive loop
Check in often enough that you catch yourself before the grind, not after. The right cadence varies: dense math wants every few minutes, absorbing a long paper can run much longer. Periodically ask: Is this working? What kind of stuck am I? If you can't articulate progress, switch modes. The cost of switching is much lower than the cost of an hour of grinding that goes nowhere.
Productive vs. wasted confusion
- Productive confusion is the kind you can articulate ("I see why step 3 follows from step 2, but I don't see why we'd think to do step 2 in the first place"). It's the engine of restructuring; let it sit.
- Wasted confusion is diffuse; you couldn't even say what's confusing you. It's a signal you're missing prerequisites or have hit a working memory wall. Don't romanticize it. If you can't name what's confusing you, that itself is the diagnosis: back up.
Frustration tracks wasted confusion, not productive confusion. The goal isn't to never struggle; it's to struggle on purpose, briefly, with the right tool.
Glossary
Working memory. The small set of items you can hold actively in mind at once while thinking, roughly 4±1 chunks for most people. Distinct from long-term memory, which has effectively unlimited capacity but is slower to access. The bottleneck in learning isn't usually how much you know; it's how much you can juggle while the idea is still wet.
Chunk. A unit of working memory. With expertise, what was once many separate items collapses into one: a chess master sees "Sicilian defense" where a novice sees twenty individual pieces, but both are using one slot. This is why schemas functionally expand working memory: bigger chunks, same number of slots.
Schema. A structured mental representation of a problem-shape or concept that lets you classify and respond to new instances without re-deriving from scratch. "This is a supply-and-demand problem," or, for programmers, "this is a sliding-window problem." Both are schemas in action. Once you have the schema, you stop solving the problem from first principles and start applying a known shape. Schemas are built through varied practice, not through reading about patterns.
Restructuring (conceptual change). Rebuilding a mental model rather than adding to it. The literature distinguishes three flavors: belief revision (updating a fact), mental model transformation (revising how parts fit together), and categorical shift (changing what kind of thing the concept is, e.g., heat as a process rather than a substance). Recursion, limits, and thinking in distributions usually require categorical shift, which is why prerequisites alone don't dissolve them.
Cognitive load. The total mental effort consumed at a given moment. Split into intrinsic (load from the material itself, set by how many elements interact and how much prior knowledge you have) and extraneous (load from how the material is presented: bad notation, bad explanations, irrelevant scaffolding). The same idea can feel hard or easy depending on extraneous load alone, which is why finding a different explanation often unlocks something that "isn't clicking."
Posner conditions. The four conditions a frame-shift requires, from Posner, Strike, Hewson & Gertzog (1982): dissatisfaction with the existing frame, intelligibility of the new one, plausibility that it could be true, and fruitfulness, seeing it open up problems the old frame couldn't. If restructuring stalls, one of these is usually missing. They're necessary but not sufficient: you can hit all four and still get stuck if the shift you actually need is a categorical one (changing what kind of thing the concept is) rather than a model transformation. In those cases the Posner conditions are the floor, not the ceiling.