How to lie using visual proofs

TL;DR

Unwrapping curved surfaces into flat shapes can change edge geometry; Gaussian curvature means you can’t flatten without losing information.

Briefing Cornell Notes

Briefing

A run of “visual proofs” goes spectacularly wrong in three different ways—showing that convincing pictures can hide fatal assumptions about geometry, limits, and even where a constructed point actually lies. The most striking example starts with a sphere sliced into vertical wedges, then “unwrapped” into a shape treated like a rectangle. That setup yields an area of π²R², even though the true surface area of a sphere is 4πR². The mismatch isn’t a minor arithmetic slip; it comes from flattening curved geometry into flat space without preserving the wedge shapes’ true edge behavior.

The sphere argument begins by taking the sphere’s surface and cutting it into many orange-like slices. Unraveling the northern hemisphere wedges and the southern hemisphere wedges and then interlacing them creates a target shape whose base is taken as the unraveled equator length (2πR) and whose height is treated as a quarter of a circumference (πR/2). As the slices get thinner, the picture suggests the shape should converge to a perfect rectangle, giving area (2πR)(πR/2)=π²R². But the curvature of the sphere matters: when a wedge is unwrapped while trying to preserve its geometry, its edges don’t behave linearly. The wedge’s width changes with latitude according to a sine relationship (proportional to 2πR·sin(φ)), so the flattened edges should bulge outward rather than remain straight. Because the edges are non-linear, the northern and southern pieces overlap in a way that does not vanish under finer slicing—so the “limit” doesn’t rescue the calculation.

A second fake proof targets a different kind of subtlety: limits. A classic construction starts with a unit circle and inscribes a square whose perimeter is 8. Then a sequence of jagged curves is built so each iteration still has perimeter 8 while the curves get closer to the circle. The visual claim is that the circle’s circumference must also be 8, implying π=4. The key mathematical correction is that even if the curves converge pointwise to the circle, there’s no guarantee that the lengths of the approximating curves converge to the length of the limiting curve. Here, the approximations remain inefficient—jagged in a way that preserves perimeter at 8—so the “limit of lengths” and “length of the limit” come apart.

The third example is a Euclid-style congruence chase that tries to prove every triangle is isosceles (and even equilateral). It constructs an angle bisector and perpendicular bisector, then uses side-angle-side and related congruence arguments to conclude AB must equal AC. Every congruence step can be made valid, but the proof collapses at the end: it assumes a constructed point E lies between A and C, so that AC=AE+EC. In many triangles, the intersection point actually falls outside the segment, making that final addition invalid. The result is a reminder that hidden assumptions—about where points land, about what “straight” really means, and about what limits preserve—are where visual certainty becomes mathematical error.

Taken together, the three failures map to three lessons: curved surfaces can’t be flattened without losing geometric information (Gaussian curvature), convergence of shapes doesn’t automatically preserve length, and diagram-based constructions can smuggle in positional assumptions. Visual intuition is powerful, but it never replaces rigor and edge-case checking.

Cornell Notes

Three “visual proofs” fail for three distinct reasons: curvature, limits, and hidden positional assumptions. A sphere-slicing argument unwraps curved wedges as if they form a rectangle, but the wedge edges change nonlinearly with latitude, causing persistent overlap and the wrong area (π²R² instead of 4πR²). A second construction approximates a circle with jagged curves that all have perimeter 8, tempting a conclusion that π=4; the flaw is that the limit of curve lengths need not equal the length of the limiting curve. The final Euclid-style congruence proof claims all triangles are isosceles, but it breaks when a constructed point lies outside a segment, invalidating a final length addition (AC=AE+EC). These cases show why rigor must check geometry, limit behavior, and hidden assumptions.

Why does the sphere-slicing “rectangle” argument produce π²R² instead of 4πR²?

Unwrapping the sphere’s wedges treats their edges as if they behave linearly, so the northern and southern pieces interlace without accounting for curvature. In reality, the wedge width across a line of latitude depends on latitude angle φ through a sine relationship: the relevant radius is R·sin(φ), so the latitude circumference is proportional to 2πR·sin(φ). Because that dependence is not linear, the flattened wedge edges should bulge outward (not remain straight). When the northern and southern unwrapped pieces are interlaced, their non-linear edges overlap by an amount that does not disappear as slices get thinner, so the limiting-rectangle intuition fails.

What is the precise mathematical trap in the “π=4” perimeter construction?

Even if a sequence of curves converges pointwise to the circle, it does not follow that the perimeters (lengths) of the curves converge to the circle’s circumference. Here, every approximating curve is constructed to have perimeter 8, while the curves get closer to the circle visually. But jaggedness remains at every stage: the approximations are inefficient paths that keep total length fixed. The example emphasizes that “length of the limit” and “limit of lengths” are different operations; calculus-style limiting arguments require extra justification about how errors in length behave.

How can a Euclid-style congruence proof be logically valid in its triangle comparisons yet still end up wrong?

The congruence steps can be correct, but the final algebraic step can rely on an unspoken geometric placement. The proof constructs an intersection point E using perpendiculars and angle bisectors. It then concludes AC=AE+EC, which is only true if E lies between A and C. For many triangles, the constructed point E falls outside the segment AC. In that case, AE+EC does not equal AC, so the final conclusion (AB=AC, and thus “all triangles are isosceles/equilateral”) collapses even though the congruence reasoning earlier was sound.

What role does Gaussian curvature play in the failure of the sphere visual proof?

Gaussian curvature means the intrinsic geometry of a curved surface cannot be flattened into the plane without distortion. When a curved surface is “unwrapped” as if it were flat, geometric information is lost unless the flattening preserves the correct metric. The sphere example fails because the wedge pieces are only made thin in one direction; the curvature persists in the other direction, so the local flatness needed for a safe limiting argument never fully applies.

Why does the pizza-wedge argument for the area of a circle work even though it also “unravels” wedges?

The circle-area “pizza slice” method is valid because the rearrangement effectively preserves the relevant geometry in the limit: the pieces can be straightened into a shape whose area matches πR². In contrast, the sphere method fails because the unwrapped wedge edges are not linear and do not interlace without overlap. The sphere’s non-linear edge behavior creates a persistent discrepancy, while the circle’s wedge rearrangement aligns in a way that yields the correct area.

Review Questions

In the sphere argument, what specific functional relationship (involving sin(φ)) causes the wedge edges to behave nonlinearly after unwrapping?
Give an example of why pointwise convergence of curves does not guarantee convergence of their lengths.
In the triangle proof, what hidden condition about the location of point E makes the final step AC=AE+EC fail?

Key Points

1
Unwrapping curved surfaces into flat shapes can change edge geometry; Gaussian curvature means you can’t flatten without losing information.
2
A sphere-slicing “rectangle” estimate fails because wedge widths vary nonlinearly with latitude (proportional to sin(φ)), creating overlap that persists under finer slicing.
3
Convergence of shapes does not automatically preserve length; “limit of lengths” can differ from “length of the limit.”
4
A perimeter-preserving sequence of jagged curves can converge to a smooth curve while still having the wrong limiting length.
5
Euclid-style congruence arguments can be correct while the overall proof fails due to a final step that assumes a constructed point lies between two others.
6
Visual proofs often smuggle in hidden assumptions about straightness, containment, or segment order; checking edge cases is essential.

Highlights

The sphere wedge method gives π²R² by treating unwrapped wedges as if they form a rectangle, but non-linear wedge edge behavior causes persistent overlap.

The “π=4” construction shows that even when curves converge to a circle, their lengths can stay fixed at 8—so length does not follow from pointwise convergence.

The triangle proof’s congruence steps can be valid, yet the conclusion fails because a constructed point E may lie outside segment AC, breaking AC=AE+EC.

The core theme across all three examples: pictures can look decisive while hiding assumptions about curvature, limits, and point placement.

Topics

Fake Proofs
Sphere Surface Area
Limits of Length
Euclid Congruence
Gaussian Curvature

Mentioned

Henry Reich