Neutral Geometry: Before You Choose Parallels

A switch you can leave off

In guide 2 you watched Hilbert rebuild geometry on five families of axioms: incidence (which points lie on which lines), order (which point is between which others, tightened by Pasch's axiom), congruence (when two segments or angles match), continuity (the Archimedean and completeness axioms that fill the line's pinholes), and finally parallels. The first four families build the stage and the actors. The fifth alone decides one thing: through a point off a line, how many parallels run.

Neutral geometry is the deliberate decision to keep the first four families and switch off only the fifth. You assert incidence, order, congruence, and continuity in full — and then, about parallels, you say nothing at all. Not "exactly one" (Euclid's choice), not "many" (the hyperbolic choice), not "none". You refuse to take sides. Neutral geometry is sometimes called absolute geometry, the older name from Bolyai, and the two words mean the same restraint.

What survives without the fifth

Here is the part that startles people: an enormous amount of school geometry never needed the parallel postulate at all. Vertical angles are equal, the side-angle-side and angle-side-angle congruences hold, the isosceles triangle theorem is true, every segment has a unique perpendicular bisector, and the triangle inequality |AB| + |BC| > |AC| stands firm. All of these are neutral theorems — proved using only incidence, order, congruence, and continuity. You have been doing neutral geometry for years without knowing its name.

The single most important neutral theorem is the exterior angle theorem, and it is worth stating exactly because its precise wording is the whole point. The exterior angle theorem of neutral geometry says: an exterior angle of a triangle is strictly greater than either of the two remote interior angles. Notice what it does not say — it does not say the exterior angle equals their sum. That stronger, tidier equality is the version you learned in school, and it quietly smuggles in the parallel postulate. The neutral version is the honest one, and it is a strict inequality.

Why does that gap matter? Because the angle-sum of a triangle, m(angle A) + m(angle B) + m(angle C) = 180 degrees, is not a neutral theorem. In neutral geometry you can prove only the weaker statement that the three angles sum to at most 180 degrees (the Saccheri-Legendre theorem). Whether the sum hits exactly 180, or always falls short, is precisely the question the parallel postulate answers — and a neutral geometer is forbidden from answering it.

Two parallels do exist — that much is neutral

It is easy to misremember neutral geometry as a place with no parallels, but that is wrong, and the correction is illuminating. In neutral geometry you can prove that at least one parallel exists. Take a line and a point P not on it. Drop a perpendicular from P to the line, then erect a second perpendicular to that segment at P. The exterior angle theorem guarantees this new line never meets the original — so a parallel through P exists.

          P
          |  \
          |    \  m   (second perpendicular at P; parallel to L)
          |      \
   -------+----------------  L   (original line)
          F

  PF _|_ L   (drop a perpendicular, foot F)
  m  _|_ PF  at P
  =>  m never meets L            (exterior angle theorem)
  =>  at least one parallel through P EXISTS  -- this is NEUTRAL

  HOW MANY such m are there?  <-- the parallel postulate decides
     Euclidean:  exactly ONE        Hyperbolic:  INFINITELY MANY

So the disputed point is never whether a parallel exists — it always does. The disputed point is how many. This is exactly why Playfair's axiom — "through a point off a line there is exactly one parallel" — is the clean modern face of Euclid's fifth postulate: the word doing all the work is one. Drop "exactly one" to "at least one" and the statement becomes a neutral theorem you can prove. Insist on "at most one" and you have re-added Euclid's choice.

Saccheri's quadrilateral: probing the gap

How do you study a question your axioms forbid you to answer? You build a neutral object whose behavior would settle it, then see what neutral geometry alone can and cannot pin down. In 1733 the Jesuit Giovanni Saccheri did exactly this with a now-famous shape. A Saccheri quadrilateral is built on a base segment AB, with two equal sides AD and BC both perpendicular to the base (so m(angle DAB) = m(angle CBA) = 90 degrees), |AD| = |BC|, and a top DC joining their ends.

Using only neutral tools — congruence and the isosceles triangle theorem — Saccheri proved the two summit angles at D and C are always equal to each other. That much neutral geometry guarantees. But it cannot tell you their size. Three cases stare back: the summit angles are right (each 90 degrees), or obtuse, or acute. Each case is internally consistent so far as neutral geometry can see, and each is a different geometry waiting to be chosen.

1. Right-angle case: the summit angles equal 90 degrees. This is equivalent to the parallel postulate — it gives you Euclidean geometry, where the quadrilateral is a true rectangle and triangle angles sum to exactly 180 degrees.
2. Obtuse-angle case: the summit angles exceed 90 degrees. This forces triangle angle-sums above 180 degrees — the world of elliptic and spherical geometry, where (with the right adjustments) lines always meet.
3. Acute-angle case: the summit angles fall below 90 degrees. This gives triangle angle-sums below 180 degrees — hyperbolic geometry, where many parallels pass through a point.

Saccheri's own goal was the opposite of what he achieved. He wanted to prove Euclid's fifth postulate by assuming the acute case and deriving a contradiction — turning the postulate into a theorem. He killed the obtuse case (it conflicts with neutral results about lines extending without bound), but the acute case refused to break. He squeezed strange, unfamiliar conclusions out of it and finally declared them "repugnant to the nature of the straight line" — but a repugnant feeling is not a contradiction. Without realizing it, Saccheri had been proving theorems of hyperbolic geometry, a full century before anyone dared to call it real.

Why the neutral trunk is exactly the right strategy

Recall from guide 3 how we test an axiom for independence: an axiom is independent of the others if there is a model where all the rest hold but it fails. Neutral geometry is the natural launch pad for that test. Because the neutral trunk fixes everything except parallels, you can graft on either parallel rule and get a consistent system — Euclidean from "exactly one", hyperbolic from "many". The mere existence of both proves the parallel postulate is independent: it cannot be a hidden consequence of the other four families, or you could not have built a model that denies it.

This reframes a two-thousand-year embarrassment. From antiquity onward, geometers suspected Euclid's fifth was really a theorem in disguise and tried to derive it from the first four. Every such attempt failed — and now you can see why it had to fail. Any proof of the fifth from the first four would be a neutral proof, valid in every neutral model, including the hyperbolic one where the fifth is false. A valid proof of a false statement is impossible. The two millennia of failure were not bad luck or dull wits; they were the universe quietly insisting the postulate is a free choice.