Why Your Rules Can’t Pin Down an Infinite Number Line
A puzzle in a secret notebook
Imagine you want to describe the real numbers — all the decimals, from zero up to infinity — with absolute precision. You write a secret notebook of rules. Rule 1: there is a number zero, written 0. Rule 2: whenever you have two numbers, you can add them or multiply them. Rule 3: every number has a negative. You add more and more rules, until you think you’ve pinned down exactly one number system. No other system could possibly obey all your rules.
But then a mathematician friend reads your notebook and says, “Your rules are very good. But I found another universe of numbers that follows every single rule — and it’s bigger than the real numbers.” You stare at your perfect list. How can this be?
This is the puzzle at the heart of model theory. It asks: when you describe a mathematical world with a formal language, how many different worlds actually fit? And if there are many, can we still find order among them?
The language a structure understands
To talk about structures, mathematicians start with a signature. A signature is just a list of basic symbols: maybe one constant symbol (like 0), two function symbols (like + and ×), and one relation symbol (like <). Think of it like choosing ingredients for a recipe. You haven’t cooked anything yet — you’ve only named what you’ll use.
A structure of that signature then gives each symbol a real meaning. It picks a set of objects (the domain). The constant symbol 0 names a particular object in the set. The symbol + becomes an actual operation that takes two objects and gives back a third. If you chose a signature with just +, 0, and a minus, you could build the integers or the real numbers. The recipe gets a kitchen.
Now you can build sentences. A theory is a collection of sentences in this language. A model of a theory is a structure that makes every sentence true. Two structures that make exactly the same sentences true are called elementarily equivalent. They are twins from the language’s point of view.
The logician Alfred Tarski (1901–1983) figured out how to define truth for these languages. He then gave a complete description of all the sentences true in the field of real numbers. But that description, amazingly, didn’t nail down the real numbers uniquely. It allowed other structures — called real-closed fields — that are elementarily equivalent but not the same size. So something was loose.
How a language can multiply models
Now meet Thoralf Skolem (1887–1963). He proved two results that shook the foundations. Downward Löwenheim-Skolem: if you have an infinite structure and a smaller infinite size, you can always find an elementary substructure of that smaller size — a tiny copy that satisfies exactly the same sentences. Upward Löwenheim-Skolem: you can also blow any infinite structure up to any larger size while keeping all sentences true. Using the compactness theorem — which says that if every finite chunk of a theory has a model, then the whole theory has one — you can build enormous elementary extensions.
So any theory that has an infinite model has models of every infinite size. It can’t be categorical (having exactly one model up to isomorphism). The real numbers with addition and multiplication, for instance, have models of all infinite cardinalities. Your notebook rules could never lock the real numbers home.
A second chance: one model per size
Oswald Veblen (1880–1960) introduced the word categorical to mean a theory with exactly one model. After Skolem’s work, it seemed hopeless for infinite structures. But mathematicians asked a sharper question: what if we fix a specific infinite size? Could a theory have exactly one model of that cardinality?
Yes. Think of a dense linear order with no first or last element. It says: between any two points there’s a third, and there’s no end in either direction. For countable sets, all such orders turn out to be the same — the rational numbers. The theory is countably categorical. But it has many uncountable models, so it’s not categorical at larger sizes.
The real breakthrough came in 1965 when Michael Morley (1930–2020) proved a stunning theorem: if a countable first-order theory is categorical in some uncountable cardinal, then it is categorical in all uncountable cardinals. Suddenly, there were only three kinds of categoricity — totally categorical (all infinite sizes), countably categorical but not uncountably, and uncountably categorical. Morley’s work opened a floodgate. The question “How many models?” became a classification project.
From few to beautiful: the classification quest
Saharon Shelah (born 1945) took Morley’s fire and turned it into a science. He built classification theory, a way of sorting theories by the number and variety of their models. Shelah found sharp dividing lines. On one side sit stable theories: these have a well-behaved independence relation, much like linear dependence in vector spaces. In a stable theory, you can assign a “dimension” to a model, and that dimension determines the model up to isomorphism. Models are tame and few.
On the wild side, theories without stability often have huge numbers of non-isomorphic models, even for the same cardinality. You can’t tell them apart with normal tools. The slogan was “few is beautiful”: if a theory forces its models to be similar, the structures themselves must be elegantly geometric.
This programme led to two other born-from-it quests. Geometric model theory studied uncountably categorical structures and found deep connections with algebraic geometry. Later, o-minimality explored ordered structures (like the real line with exponentiation) that are so well-behaved that every definable set of numbers is just a finite union of intervals and points. These tame ordered worlds helped solve centuries-old problems about rational points on curves — linking logic to number theory.
Why this game still matters to you
You may never write down axioms for the real numbers in your notebook. But every day you rely on systems that work only because of model-theoretic ideas. When a computer programmer specifies how a banking app should behave, they’re writing a kind of theory. If that theory accidentally admits many unintended models, the app might do strange things nobody expected. Model theory helps designers spot such gaps before code goes live.
Beyond software, the same patterns appear whenever we try to pin down an idea with words. Think of setting rules for a club or a game. You want the rules to create exactly the one experience you imagine. But if your rules don’t nail down the right size or fairness, you might get very different game worlds. Model theory is the ultimate warning: when you describe infinite things with a finite set of rules, you always leave room for surprises. And that’s not a failure — it’s an invitation to explore the whole landscape of possible worlds that fit your words.
Think about it
- If you wrote down the perfect rules for a friendship, could those rules ever guarantee that only one kind of friendship exists in the world? Why or why not?
- A video game’s rulebook fits every copy of that game exactly. But what if the game has an infinite number of levels — could two different gamers, seeing different parts, ever be sure they’re playing “the same game”?
- Imagine a robot that can build any club that follows your written rules. If the robot builds two clubs of different sizes, are you to blame for not writing better rules, or does the limit come from the nature of rules themselves?
