Relative Consistency and Independence

This page draws the logical moral of the whole synthetic half of the spine. The models of hyperbolic geometry are not merely convenient pictures: they are a proof, in the strict metamathematical sense, that the parallel postulate can be neither proved nor disproved from Euclid's other axioms. This is the definitive end of the two-thousand-year program described in discovery-history.md, and it is a clean early instance of the model-theoretic method for proving independence — the same method that governs the independence results in logic and set theory.

References: Greenberg, Euclidean and Non-Euclidean Geometries, ch. 7; Hilbert, Grundlagen der Geometrie; Nagel & Newman, Gödel's Proof (for the general consistency-via-models idea); Bonola, Non-Euclidean Geometry.

1. Relative Consistency

A theory is consistent if no contradiction can be derived from its axioms. Absolute consistency proofs are rarely available; what models give is relative consistency — a reduction of one theory's consistency to another's.

Relative Consistency Theorem. If Euclidean geometry (equivalently, the theory of the real numbers) is consistent, then hyperbolic geometry is consistent.

Proof. Suppose hyperbolic geometry were inconsistent — that some statement $\phi$ and its negation $\neg \phi$ were both provable from the hyperbolic axioms. Take any model, say the Poincaré disk. In the model every hyperbolic axiom is a true theorem of Euclidean geometry (each "point," "line," and "congruence" is a defined Euclidean object, and the axioms translate into provable Euclidean facts). A proof is just a finite chain of logical inferences, so translating the hyperbolic derivation of $\phi \wedge \neg \phi$ through the dictionary produces a Euclidean derivation of the translated contradiction ${\phi }^{\ast }\wedge \neg {\phi }^{\ast }$. Thus Euclidean geometry would be inconsistent too. Contrapositively: if Euclidean geometry is consistent, so is hyperbolic geometry. $■$

The identical argument, run on the sphere sitting inside ${\mathbb{R}}^3$, gives the relative consistency of elliptic geometry. And since Hilbert modeled Euclidean geometry itself inside the real numbers (the Cartesian plane ${\mathbb{R}}^2$), all three classical geometries stand or fall with the consistency of $\mathbb{R}$:

$$\mathrm{Con}(\mathbb{R})⟹\mathrm{Con}(\text{Euclidean})⟹\mathrm{Con}(\text{hyperbolic}),\mathrm{Con}(\text{elliptic}).$$

2. From Consistency to Independence

Consistency of the denial of an axiom is exactly what it takes to prove that axiom independent.

Independence of the Parallel Postulate. The parallel postulate is independent of the neutral axioms (Hilbert's Groups I–III + V): it can be neither proved nor disproved from them.

Proof. Two models do the work:

• Not provable. The Poincaré disk (or any hyperbolic model) satisfies all the neutral axioms and the negation of the parallel postulate. If the postulate were provable from the neutral axioms, it would hold in every model of them — but here it fails. So it is not provable.
• Not disprovable. The ordinary Euclidean plane ${\mathbb{R}}^2$ satisfies all the neutral axioms and the parallel postulate. If the postulate were disprovable (i.e. its negation provable) from the neutral axioms, its negation would hold in every model — but here the postulate itself holds. So it is not disprovable.

Having exhibited a model of neutral-geometry-plus-postulate and another of neutral-geometry-plus-negation, the postulate is logically independent of the neutral core. $■$

This is the precise, final answer to the ancient question "can the fifth postulate be proved from the other four?" — No, and the proof is a pair of models. The centuries of failed proof attempts were doomed not by insufficient cleverness but by a theorem.

3. The General Method: Independence via Models

The structure of the argument is completely general and worth isolating, because it recurs throughout foundations:

Model method for independence. To show a statement $S$ is independent of an axiom system $\mathcal{A}$, exhibit

• a model of $\mathcal{A}\cup \{S\}$ (so $\mathcal{A}/⊢\neg S$), and
• a model of $\mathcal{A}\cup \{\neg S\}$ (so $\mathcal{A}/⊢S$).

Hyperbolic geometry was the first great success of this method — historically prior to, and a template for, its most famous later applications:

• The independence of the Axiom of Choice and the Continuum Hypothesis from ZFC (Gödel's constructible universe $L$ and Cohen's forcing supply the two models) — see the set-theory page and the remark on what justifies Con(ZFC).
• The independence of the parallel postulate is the geometric ancestor of all such results: replace "ZFC" by "neutral geometry" and "CH" by "the parallel postulate."

The connection to model theory is exact: a model is an interpretation making the axioms true, and "$\mathcal{A}/⊢S$" is established by producing a model of $\mathcal{A}\cup \{\neg S\}$ — the soundness direction of the completeness theorem.

4. What Independence Does Not Say

Two cautions keep the result in proportion.

• Independence is not "meaninglessness." That the postulate is independent does not make it arbitrary or contentless. Each choice yields a rich, specific geometry; the choice is settled empirically for physical space (by measuring curvature), not by logic. This is the hinge into the philosophy of geometry: the conventionalist reads the choice as a matter of stipulation, the empiricist as a matter of fact about the world — but neither reads it as provable a priori. General relativity later made it a measurable, dynamical field.
• Relative, not absolute. The proof reduces hyperbolic consistency to Euclidean (hence real-number) consistency; it does not prove the real numbers consistent from nothing. By Gödel's second incompleteness theorem no sufficiently strong theory can prove its own consistency, so an absolute proof is not to be had. The geometry is exactly as safe as arithmetic — which is safe enough for every working purpose, and no safer.

5. Geometry and Model Theory

The independence proof is one point of contact between geometry and model theory — the study of the relationship between formal theories and the structures that satisfy them. The connection runs deeper than this single result, and it is worth collecting the threads, because geometry is both the founding example of the model method and one of its richest modern subjects.

• Models as independence proofs (this page). The argument of §2–3 is exactly the model-theoretic method: to show a statement independent of a theory, exhibit two models. It is the soundness direction of the completeness theorem, and hyperbolic geometry is its historical prototype — prior to and a template for the ZFC independence results.
• Elementary geometry is a decidable first-order theory (Tarski). First-order Euclidean geometry, in Tarski's axiomatization, is complete and decidable via quantifier elimination for real closed fields — the landmark recorded under quantifier elimination and definability. "Elementary plane geometry" and "the first-order theory of $(\mathbb{R},+,\cdot ,<)$" are, model-theoretically, essentially the same object — the coordinatization worked out in full on the next page. Lacking a first-order definition of "$n$ times," geometry cannot encode $\mathbb{N}$ and so escapes the Gödel incompleteness that governs arithmetic.
• Categoricity needs second-order logic. Hilbert's axioms determine the Euclidean plane ${\mathbb{R}}^2$ uniquely only through their second-order continuity axiom (Group V). By Löwenheim–Skolem and the failure of first-order categoricity, the first-order theory has non-standard models — including non-Archimedean ones with infinitesimal segments, the same phenomenon that yields the hyperreals of non-standard analysis.
• The Cayley–Klein construction is a model-theoretic interpretation. Building hyperbolic geometry inside projective / real-field geometry by fixing an absolute conic is a definable interpretation of one structure in another — the general mechanism by which relative-consistency transfers, and what turns the Beltrami–Klein model into a proof rather than a picture.
• O-minimality: tame geometry as model theory. The real field is o-minimal (see model-theory.md), meaning definable subsets of the line are just finite unions of points and intervals. O-minimality is essentially a model-theoretic axiomatization of geometric finiteness (Grothendieck's "tame topology"), with definable sets admitting cell decompositions and finitely many components.
• Projective geometry recovered intrinsically. In classification theory the algebraic-closure operator on a strongly minimal set forms a combinatorial geometry (pregeometry); the Zilber trichotomy forces it to be trivial, affine/projective over a division ring, or the geometry of an algebraically closed field — a modern echo of the classical coordinatization theorems in which Desargues and Pappus recover the coordinate ring.

In one line: model theory studies the theory ↔ structure relationship, and geometry supplies both its founding example (the parallel-postulate independence proof) and one of its deepest ongoing subjects (real closed fields, o-minimality, tame geometry).

6. Summary

The models convert the synthetic strangeness of the hyperbolic plane into a rigorous consistency proof; consistency of the denial converts into independence of the parallel postulate; and the independence argument is the prototype of the model method used across logic. With this the synthetic and logical story is complete. The remaining pages change register from axioms to metrics, recasting all three geometries as the constant-curvature cases of Riemann's differential geometry — where the parallel postulate reappears, quantitatively, as the single equation $K=0$.