In the last lesson we watched the man with a hammer flail: one model, applied to everything, wrong most of the time. The obvious fix is “get more models.” But a pile of models is just a pile of hammers. The interesting question is what you do with them — how they hang together in your head so they’re actually there when you need them.
Charlie Munger had a word for the answer, and it’s the whole point of this course.
Before you read — take a guess
Munger's 'latticework' is best described as…
The latticework, defined
Picture a garden trellis: a lattice of crossed slats fixed to a wall. A climbing plant that meets bare brick just flops on the ground. Give it a lattice and it grabs on, climbs, and spreads — every tendril hooked to a slat, held up by the whole structure. Now picture the alternative: a heap of loose sticks in the corner. Same wood, no structure, nothing to climb.
Your experience is the plant. Models are the lattice.
Munger’s latticework of mental models is his central idea from the 1994 talk A Lesson on Elementary, Worldly Wisdom. In his words: “you’ve got to have models in your head, and you’ve got to array your experience — both vicarious and direct — on this latticework of models.” The home discipline here is really cognitive psychology dressed in a gardener’s metaphor: memory is associative, so a fact tied to many other things is easy to retrieve, and a fact tied to nothing is nearly impossible to retrieve. Models are the “many other things.” They’re the hooks.
Two words in that quote do a lot of work. Vicarious experience is what you learn from others — history, biographies, other people’s disasters. Direct experience is what happened to you. Munger’s claim is that you should hang both on the same lattice, because your own life is too short to supply enough data on its own.
A worked example: two minds, one fact
Suppose both of us learn the same fact: “A discount airline cut prices, filled its planes, and its profit went up.”
| Mind A: loose sticks | Mind B: latticework | |
|---|---|---|
| How it’s stored | One isolated fact about one airline | Hung on several models at once |
| Models it touches | none | supply & demand, fixed vs. marginal cost, incentives, feedback loops |
| A year later | ”Some airline did something with prices?" | "Right — near-zero marginal cost per seat, so filling empty seats is almost pure profit; that’s the same reason software and hotels discount the last unit.” |
| Transfers to a new problem? | No — it’s about airlines | Yes — it’s about any business with high fixed and low marginal cost |
Same input. Mind B didn’t memorize harder; it had somewhere to put the fact. The fact became a case of several general ideas, so it can be recalled through any of them and reapplied anywhere those ideas show up.
The lattice does double duty. It helps you remember (more retrieval routes to the same fact) and it helps you transfer (a fact filed as an instance of “high fixed cost, low marginal cost” reappears the moment you meet software, hotels, or bridges). A fact alone is inert. A fact on the lattice is a tool.
The pitfall
The failure here isn’t ignorance — it’s a full head that’s badly wired. Trivia champions and the crammer who forgets everything after the exam both have plenty of facts and no lattice. Loading more facts onto loose sticks just gives you a bigger heap. Munger’s fix is structural, not quantitative: build the slats first, then hang things on them.
When to use it
Use the lattice test whenever you learn something you actually want to keep: don’t ask “did I memorize it?” — ask “what models does this hang on?” If the honest answer is “none,” you’ve stored a loose stick that will roll away by next week. Give it at least two hooks before you move on.
Why cross-disciplinary breadth beats depth in one field
Now, which models make up the lattice? Munger is blunt: “you must know the big ideas in the big disciplines and use them routinely — all of them, not just a few. Most people are trained in one model — economics, for example — and try to solve all problems in one way.” That last sentence is the man with a hammer, named and shamed.
The home discipline for this claim is something like the economics of learning, and it rests on a Pareto point: within any field, a small handful of big ideas explains most of what matters. The tenth-most-important idea in physics rarely helps a non-physicist; the first three help everyone. So the high-return move isn’t to go a mile deep in one field — it’s to grab the top few models from many fields. You don’t need to be a physicist. You need the physicist’s handful of big models, used routinely.
Depth-in-one-field has sharply diminishing returns for a generalist decision-maker; breadth-across-fields does not, because each new field’s big ideas cover situations your current fields simply can’t see.
A worked example: a business problem cracked by importing a model
A subscription business is bleeding: it spends heavily to acquire customers, but they keep leaving, and the finance team keeps re-running the same spreadsheet, tuning the discount, getting nowhere. That’s the loose-stick move — more of the one model they own.
Then someone reaches across disciplines and asks a biologist’s question: what’s the population dynamics here? Model customers like a population with a birth rate (new signups) and a death rate (churn). Suddenly the math is obvious: if the monthly death rate exceeds the birth rate, no acquisition budget on Earth wins — you’re pouring water into a leaky bucket. The lever was never price; it was the death rate. Fix retention and the same acquisition spend now compounds.
| Lens applied | Question it asks | Verdict |
|---|---|---|
| Finance only (the hammer) | “What discount maximizes this quarter’s revenue?” | Local, keeps losing |
| Biology: population dynamics | ”Is birth rate above death rate?” | Reframes the whole problem — fix churn first |
| Physics: compounding / leaky integral | ”Does retained value accumulate or leak out?” | Confirms retention is the real lever |
Nothing new was learned about finance. The breakthrough came from outside finance — a model from biology that the finance-only mind couldn’t reach.
The pitfall
Breadth curdles into name-dropping if you collect models like trivia and never use them. “Breadth” here means the big models loaded, connected to real cases, and reached for by default — not a glossary you can recite. A shelf of unread field guides doesn’t make you a naturalist.
When to use it
Deliberately widen the lens when your usual model keeps failing or keeps producing the same non-answer. That stuckness is the signal that the situation lives partly in a discipline you’re not looking at. Before adding one more field to yourself, make sure you’ve genuinely grabbed the top few models of the fields you already claim.
The handful of high-value models that pay across domains
You don’t need hundreds. You need a compact set of models whose home is one discipline but whose reach is every discipline. Here’s a starter set — the kind Munger meant by “the big ideas in the big disciplines.” Note the third column: each model is really a type of question you can ask of any situation.
| Model | Home discipline | The kind of question it answers |
|---|---|---|
| Incentives | Economics / psychology | ”Who benefits from the current behavior, and what are they being rewarded to do?” |
| Base rates | Probability / statistics | ”Ignoring this vivid story, how often does this kind of thing usually turn out this way?” |
| Second-order effects | Decision-making | ”And then what? What does the reaction to the effect cause?” |
| Supply & demand | Economics | ”What happens to price and quantity when this becomes scarcer or more wanted?” |
| Feedback loops | Systems theory | ”Does this outcome feed back to amplify itself or damp itself down?” |
| Natural selection | Biology | ”What is being selected for here over time, and what dies out?” |
| Critical mass | Physics | ”Is this below the threshold and fizzling, or past it and self-sustaining?” |
| Margin of safety | Engineering | ”How much room for error before this fails, and have I left enough?” |
| Opportunity cost | Economics | ”What’s the best thing I’m giving up by choosing this?” |
| Inversion | Mathematics / problem-solving | ”Instead of how to win, how would I guarantee failure — then avoid that?” |
| Compounding | Mathematics | ”What does this look like after it feeds on itself for a long time?” |
Read the middle column and the point of the whole course jumps out: these homes are scattered across the map. Economics, biology, physics, psychology, math, engineering. A latticework isn’t a stack of finance ideas. It’s a genuinely cross-disciplinary grid — which is exactly why it can catch situations that any single field would drop.
Match each mental model to the home discipline it comes from. (Some models earn a living far outside home — the point is to learn where each was born.)
A worked example: five hooks for one headline
Say the headline is: “City caps rents; a year later there are fewer apartments for rent than before.” Run the handful over it and watch the same fact hang on hook after hook:
- Supply & demand: a price ceiling below market clears more demand than supply — shortage, by construction.
- Incentives: landlords now earn less per unit, so some convert apartments to condos or stop building. They’re rewarded to withdraw supply.
- Second-order effects: the first-order goal (cheaper rent) triggers a reaction (less supply) that undercuts the goal.
- Feedback loops: fewer units → longer waitlists → more pressure for more controls → still fewer units.
- Base rates: price ceilings have produced shortages across many cities and eras; the base rate of “this ends in shortage” is high, so this outcome is unsurprising.
One event, five models, five reasons it makes sense — and if any one model had failed you, the other four would still catch it. That redundancy is the whole payoff of a lattice.
The pitfall
Two traps live here. First, treating the home discipline as a fence: “supply & demand is an econ thing, and I’m not doing econ today.” The whole trick is that these models trespass — critical mass explains viral posts, natural selection explains which memes and companies survive, margin of safety belongs in your calendar as much as your bridge. Second, over-collecting: a table of forty models you never actually run beats a table of ten, on paper only. Ten you reach for by reflex win.
When to use it
Keep a short, memorized checklist — this table, roughly — and run it deliberately over any important situation, the way a pilot runs a pre-flight list. The goal is that reaching for several of these becomes automatic, so you notice when a situation is really about incentives or base rates instead of the one model you happened to open with.
How you actually build one
So the lattice isn’t handed to you; you grow it. Three moves, in order:
- Learn the big models. Grab the top few ideas from each big discipline — roughly the table above — and understand each well enough to state the question it asks. You’re building slats, not memorizing a syllabus.
- Attach real experience to each. For every model, hang at least one concrete case you actually know — a story, a project, a headline you thought hard about. The rent-control case is your hook for supply & demand; the leaky-bucket startup is your hook for population dynamics. A model with no case attached is a slat with nothing climbing it.
- Reach across disciplines on purpose when stuck. When your default lens stalls, deliberately ask: “What would a biologist see here? An engineer? A psychologist? A physicist?” Rotating the question is how you find the model that was hiding in a field you weren’t looking at — exactly the move that cracked the churn problem.
The mechanism underneath all three is the same, and it’s worth saying plainly: a fact hung on several models is recallable and transferable; a fact alone is inert. Every hook you add gives the fact another way to be remembered and another situation to be reused in. That’s why breadth beats depth for a decision-maker — more disciplines means more hooks per fact.
Drag each model to the discipline it was born in. (Out in the world they roam everywhere — but each has a birthplace.)
- Social proof
- Natural selection
- Incentives
- Critical mass
- Supply & demand
- Opportunity cost
- Margin of safety
- Population dynamics
Don’t try to build the whole lattice at once. Pick three models this month, and each time you meet a real situation, force it onto one of them and attach the case. Slats get built one plant at a time; in a year you’ll have a wall.
The pitfall
The classic building mistake is stopping at step 1 — learning the models and never attaching experience. That’s a beautiful empty trellis: impressive on the wall, nothing growing. You’ll be able to define incentives and still fail to notice them in the room you’re standing in, because you never hung a real case where you could feel it. Definitions live in step 1; instincts live in steps 2 and 3.
When to use it
This isn’t a one-time project — it’s a standing habit. Treat every important experience as raw material: as it happens, ask which slats it belongs on, and hang it. Do that for years and the lattice thickens until multi-model thinking stops feeling like effort and starts feeling like just… seeing the situation clearly.
Recap: the latticework
Why do facts 'hung on' models survive in memory when isolated facts fall out?
Check your answer to continue.
Where this goes next
You now have the picture: a lattice of big, cross-disciplinary models with your experience hung on it. But “run several models” is still a slogan until you know how several models actually combine on one situation — sometimes they agree and reinforce, sometimes they disagree and force a call, sometimes they multiply into something none of them predicted alone.
That’s lesson 3, Three Ways Models Combine — where we stop admiring the lattice and start operating it.