variable

**natural deduction** metalanguage, practical foundations

**type theory** (dependent, intensional, observational type theory, homotopy type theory)

**computational trinitarianism** = **propositions as types** +**programs as proofs** +**relation type theory/category theory**

In logic, a variable is a symbol that stands for an arbitrary instantiation of some given type (although in single-typed theories the type is trivial). Thus, every variable of type $T$ is a term of type $T$ (although typically there are other terms).

Logicians have developed several ways to precisely specify what variables are and how to keep track of them, and the only thing more annoyingly pedantic than learning one of these is translating between different ones.

If we keep track of context, every introduction of a variable changes the context. Thus, whatever terms of type $T$ there may (or may not) be in a given context $\Gamma$, in the context $(\Gamma, x\colon T)$ —which is $\Gamma$ extended by a **free variable** $x$ of type $T$— there is (at least) one more term of type $T$, and that term is $x$ itself.

Conversely, we may be able to take a term $a$ in the context $(\Gamma, x\colon T)$ and apply some operation $f$ to it to create a term $f(a)$ (possibly of a different type) in the context $\Gamma$; then any appearances of the variable $x$ in the term $a$ have become **bound variables** in the term $f(a)$. That is, the appearances of $x$ are bound by the operator $f$, and $x$ has no meaning outside of its scope.

So for example, if (say in something simple like Peano arithmetic, supplemented with the abbreviations used in practice) I write $2x$, this is a term (for a natural number) in a context with a free variable $x$, whereas if I write $3 \sum_{x=0}^4 2x + 5$, this is a term (also for a natural number) in a context with no free variables; here, $x$ is bound. The variable $x$ has a meaning after the operator $\sum_{x=0}^4$ (whose notation also specifies the variable that will have a meaning inside it) and before the plus sign (which, following the standard order of operations, marks the end of the scope of $\sum_{x=0}^4$) but has no meaning outside of that (where the $3$ and $5$ are, or even where the $0$ and $4$ are). And indeed, the ultimate meaning of $2x$ depends on what $x$ is, but the ultimate meaning of $3 \sum_{x=0}^4 2x + 5$ is simply $65$.

In the internal language of a category $\mathcal{C}$ a morphism

$f\colon B \to A$

is a *term* $f(x)$ of *type* $A$ where $x$ is a **free variable** of *type* $B$, which in symbols is given by

$x\colon B \vdash f(x)\colon A
\,.$

We may think of the *free variables* here as being placeholders for all the generalized elements $U \stackrel{x}{\to} B$ of $B$.

Then the assertion

$x\colon B \vdash f(x)\colon A$

indicates that with $B \stackrel{f}{\to} A$ given we may send $U \stackrel{x}{\to} B$ to the composition

$(U \stackrel{x}{\to} B \stackrel{f}{\to}A ) = (U \stackrel{f(x)}{\to} A)
\,.$

A free variable becomes a **bound variable** after application of a quantifier: for instance the image of $f\colon P \to X$ under base change

$\mathcal{C}_{/X} \stackrel{\to}{\stackrel{\leftarrow}{\to}}
\mathcal{C}$

represents, for the left adjoint the type

$\exists x\colon X , P(x)$

(existential quantifier), and the the right adjoint the type

$\forall x\colon X , P(x)$

(universal quantifier) in which now $x$ is a *bound variable*.

A textbook account with an eye towards applications in computer science is in section 1.2 of

- Robert Harper,
*Practical Foundations for Programming Languages*(web)

An exposition on the relation between free variables and universal quantification is in

- Andrej Bauer,
*Free variables are not “implicitly universally quantified”!*(web)

Revised on December 11, 2017 05:39:38
by Urs Schreiber
(178.6.238.60)