A **Hilbert system** is a kind of deductive system characterized by the presence of many axioms and few rules. Usually the only judgment form is “$A \; true$” for some proposition $A$, which is usually written simply as “$A$”. Thus, a Hilbert-style deduction is a tree whose edges are formulas, whose leaves are axioms, and whose nodes are one of a very small number of rules. The most common rule is modus ponens:

$\array{\arrayopts{\rowlines{solid}} A \qquad A\to B \\ B }$

and in many Hilbert systems this is the only rule, although sometimes one finds others in use. Then each logical connective is described by imposing axioms.

In many cases, the necessary axioms are obtained fairly straightforwardly from the corresponding rules by replacing derivability and entailment by the implication connective. For instance, the rules for disjunction are

$\frac{\Gamma\vdash A}{\Gamma\vdash A\vee B}\qquad
\frac{\Gamma\vdash B}{\Gamma\vdash A\vee B} \qquad
\frac{\Gamma\vdash A\vee B \quad \Gamma,A\vdash C \quad \Gamma,B\vdash C}{\Gamma\vdash C}$

and the corresponding Hilbert-style axioms are

$A\to (A\vee B) \qquad B\to (A\vee B) \qquad (A\vee B) \to (A\to C) \to (B\to C) \to C.$

Of course, this gives a special role to implication. Its axioms are

$\array{ P \to P \\
P \to (Q\to P) \\
(P \to (Q\to R)) \to ((P \to Q) \to (P \to R)).
}$

Note that these are precisely the types of the basic combinators? in combinatory logic. In fact, under the propositions as types correspondence, Hilbert systems correspond to combinatory logic in the same way that natural deduction/sequent calculus corresponds to lambda-calculus.

In a Hilbert system one often writes rules and deductions using a turnstile $\vdash$ instead of a horizontal bar. Thus, modus ponens would be written

$A, A\to B \vdash B$

This can be somewhat confusing when comparing a Hilbert system to other systems such as sequent calculus or type theory in which $\vdash$ appears as part of *each judgment* to specify the context of that judgment. Often the two meanings of $\vdash$ can be conflated, but not always. For instance, in substructural logic such as linear logic or relevance logic, we may have a rule (not an axiom) such as

$\array{\arrayopts{\rowlines{solid}} A \qquad B \\ A\& B }$

(where $\&$ represents the additive conjunction), which in a Hilbert system would be written as $A,B\vdash A\& B$. But the *sequent* $A,B\vdash A\& B$ is not valid in the usual sequent-calculus presentations of such logics, because the comma in the sequent context (on the left) is usually a judgment-level version of the multiplicative conjunction.

Last revised on July 11, 2016 at 20:19:33. See the history of this page for a list of all contributions to it.