A sari and a sports bike are both exercises in the same curve

A draped pallu and a superbike's flank are both drawn by the same law: a bent elastic strip settles into the curve with the least bending energy. Here is the maths that links six yards of silk to a Ducati fairing, and where the link snaps.

Side view of a red Ducati 916 SPS sports motorcycle showing one long flowing line from the nose to the tail
Photo: ScuderiaAssindia, CC BY-SA 4.0, via Wikimedia Commons

A sari and a sports bike share a real mathematical curve: the line a thin elastic strip takes when it bends with the least possible energy. Silk falls into that curve because physics forces it to, and motorcycle designers draw it on purpose with splines and Bézier curves descended from a wooden strip held down by lead weights. Both have continuous curvature, which is why the eye reads them as one flowing gesture, and both break the rule in the same place: a deliberate crease.

Key takeaways
  • Bending moment = B × curvature: in any draped cloth, curvature cannot jump, so a hanging pallu is automatically curvature-continuous (G2).
  • 1946: I. J. Schoenberg named the mathematical "spline" after the draughtsman's bendy strip that shipwrights used.
  • G2 or G3 continuity is what car and bike stylists call a Class A surface.
  • 67.2% vs 50.6%: people liked curved versions of objects more than sharp-angled ones in Bar and Neta's 2006 experiment.
  • c ≈ l/2: Peirce's cantilever test turns a fabric's overhang into its bending length, the scale of every fold it makes.

This is a tier-4 "collision": two things that have no business together, and a check of whether the link is real. The claim to test is specific. A Nivi-draped sari and the side profile of a sports bike are both built from one continuous line that looks like motion while standing still. If that is true, there should be one equation underneath both. There is, and it is about 300 years old.

What curve do a sari and a sports bike share?

They share the elastica, the shape of a thin flexible strip bent under load, and its engineering descendant, the cubic spline. The elastica minimises total bending energy, and the spline is its small-deflection approximation, which is the tool every car and bike body is drawn with.

The story

Before computers, curves for ship hulls were drawn with a flat spline: a long, thin strip of wood pinned at a few points by lead weights that draughtsmen called "ducks". The strip settled into the smoothest shape it could, and the draughtsman traced along it. In 1946 the mathematician I. J. Schoenberg borrowed the tool's name for the piecewise polynomials that copy it. A decade later two French car companies needed the same thing in numbers. Paul de Casteljau worked out his algorithm at Citroën in 1959, and Pierre Bézier, independently, used the same polynomials in the 1960s to define the bodywork of Renault cars. Every Bézier pen tool in design software descends from that bent strip of wood.

Wikipedia: Flat spline and Wikipedia: Bézier curve

A sari is that same strip, made of silk. The difference is who holds the ducks. On a bike, the designer places the control points. On a body, the shoulder, the waist tuck and gravity do it.

A naval architect's lines drawing of a hull, with many long smooth curves drawn through stations
A hull lines drawing. Curves like these were drawn with a flat spline held by weights, the physical ancestor of the Bézier curve.Photo: Matahmoush, CC BY-SA 4.0, via Wikimedia Commons

Why can a draped cloth never make a sharp corner?

Because a sharp corner would need infinite bending moment. In a thin sheet the internal bending moment equals bending rigidity times curvature, and with no point-loads along the free hang, the moment changes smoothly, so curvature changes smoothly too.

The worked derivation

Treat a hanging strip of the pallu as a thin rod of unit width. These are my assumptions:

  • the fabric bends but does not stretch (silk is far stiffer in tension than in bending);
  • it carries only its own weight \( w \) per unit length, with no knots or pins along the free part;
  • bending rigidity per unit width is \( B \), constant along the strip.
  1. Constitutive law (Euler–Bernoulli). The internal moment is proportional to curvature \( \kappa = d\theta/ds \):
    \[ M(s) = B\,\kappa(s) = B\,\frac{d\theta}{ds} \]
  2. Moment balance on a small element. With tension and shear forces \( \mathbf{F} \) and no applied couples,
    \[ \frac{dM}{ds} + \left(\mathbf{t}\times\mathbf{F}\right)\cdot\mathbf{k} = 0 \]
    Forces are finite everywhere along a free hang, so \( dM/ds \) is finite and \( M(s) \) is continuous.
  3. Conclude. Since \( \kappa = M/B \) and \( B \) is constant, \( \kappa(s) \) is continuous. The curve is G2: position, tangent and curvature all join smoothly.
  4. The size of the folds. Comparing bending stiffness with weight gives the only length scale in the problem:
    \[ c = \left(\frac{B}{w}\right)^{1/3} \]
    This is Peirce's bending length, which textile labs still measure with the cantilever test in ASTM D1388.

That last line matters most. Folds much tighter than about \( c \) cost too much bending energy; folds much larger than \( c \) are pulled flat by gravity. So the drape of a sari has a built-in scale set by the fabric itself.

A sari is a strip of silk solving a minimisation problem in real time, with gravity as the computer.

Measuring the bending length

Peirce's cantilever test pushes a strip about 25 mm wide (the Ijaz study cut 6 × 1 inch specimens) off a ledge until its tip droops to a line about 41.5° below horizontal, the angle at which the formula conveniently gives half the overhang. The overhang \( l \) at that angle gives

\[ c = l\left[\frac{\cos(\theta/2)}{8\tan\theta}\right]^{1/3} = l\left[\frac{0.935}{8\times0.884}\right]^{1/3} \approx 0.51\,l \]

so the bending length is about half the overhang. Rigidity per unit width then follows from the fabric's mass \( W \) in g/m², using the formula Ijaz and colleagues used:

\[ G\,[\mu\mathrm{N\,m}] = W\,[\mathrm{g/m^2}] \times c^3\,[\mathrm{mm^3}] \times 9.807\times10^{-6} \]

For a 145 g/m² silk–polyester fabric (the mass of the silk blend in their study) and an assumed bending length of 20 mm (my estimate for a soft sari silk), \( G = 145 \times 8000 \times 9.807\times10^{-6} \approx 11.4\ \mu\mathrm{N\,m} \).

Evaluation of Bending Length, Rigidity and Modulus of Woven and Knitted FabricsMehreen Ijaz, Namood-e-Sahar, Zohra Tariq · Proceedings of the Pakistan Academy of Sciences A 60(3), 57–61 · 2023

Six fabrics from Nishat Mills, three woven and three knitted (125 to 229 g/m²), were tested on a Shirley stiffness tester under ASTM D1388, three times in warp and weft on both faces. The authors computed bending rigidity and modulus from the measured bending length and found woven fabrics draped better than knits, with the silk-blend plain weave (60% silk, 145 g/m²) giving the most favourable result. The limit: six samples from one mill, and the per-fabric values appear only in charts, so treat it as a method reference rather than a silk database.

Read the paper (free)

Use this to check what you just read. Drop a strip of your own fabric off a table edge, measure the overhang when its tip reaches about 41.5° (a protractor app works), and plug it in.

Bending length and stiffness of a fabric (Peirce cantilever test)

ledge41.5° lineoverhang lfabric stripc = 0.51 × lG = W × c³ × 9.807×10⁻⁶ (μN·m)Stiffer cloth: longer l, bigger folds
Peirce's cantilever test. The overhang when the tip droops to 41.5° gives the bending length, the natural size of every fold the fabric makes.

How do designers draw the same curve on a motorcycle?

They use splines and Bézier curves on purpose and check them for curvature continuity. Body surfaces that pass this check are called Class A surfaces, and they are usually held to G2 or G3.

A cubic Bézier segment runs between two end points steered by two control points:

\[ \mathbf{B}(t) = (1-t)^3\mathbf{P}_0 + 3(1-t)^2t\,\mathbf{P}_1 + 3(1-t)t^2\,\mathbf{P}_2 + t^3\mathbf{P}_3,\quad 0\le t\le1 \]

Joining two segments with the shared point and its neighbours in a straight line gives tangent continuity (G1). Curvature continuity (G2) needs one more condition on the second derivative. That is exactly what the bent strip of wood did for free, because a thin elastic strip pinned at knots is a cubic spline to first order: it minimises

\[ E = \frac{B}{2}\int \kappa^2\,ds \]

and the minimiser of that integral between pins is a piecewise cubic with continuous second derivative. The draped sari minimises the same integral plus a gravity term. Same objective, different constraints.

ContinuityWhat joins smoothlyWhat the eye seesSariSports bike
G0Position onlyA visible kinkPleat edges, the waist tuckPanel gaps, seat-to-tank break
G1TangentNo kink, but reflections jumpPressed pleat foldsSharp character creases (deliberate)
G2CurvatureOne continuous lineThe free hang of the palluClass A tank and fairing flanks
G3Rate of curvature changeHighlights flow without wobbleNot controlledPremium Class A surfaces
A Raja Ravi Varma oleograph of a woman in a draped sari, the fabric falling in long smooth folds
Raja Ravi Varma's Vasantika. The painter followed the cloth's free hang as one long line, with tight folds only where the body pins the fabric.Photo: Raja Ravi Varma, Public domain, via Wikimedia Commons

Why does a continuous curve read as motion?

Two separate lines of research point the same way: people prefer curved contours at a glance, and still images that imply motion activate the brain's motion areas. Neither proves "curves equal speed", but together they explain why a G2 line feels alive.

Humans prefer curved visual objectsMoshe Bar, Maital Neta · Psychological Science 17(8), 645–648 · 2006

Fourteen people made like or dislike judgements on 140 pairs of real objects and 140 pairs of meaningless patterns, each shown for 84 ms, with the pairs matched except for sharp versus curved contours. Curved objects were liked 67.2% of the time against 50.6% for sharp ones (d = 0.94); for patterns, 37.9% against 24.8%. Reaction times were the same, so the authors argue the bias is not just easier processing. The limit: 14 participants and greyscale images, with no test of real products.

Read the paper (free)
Share of images liked at an 84 ms glance: curved vs sharp contours
See the data as a table
Curved (%)Sharp-angled (%)
Real objects67.250.6
Meaningless patterns37.924.8

Source: Bar and Neta, Psychological Science (2006) · 14 participants, 140 pairs of each kind, like or dislike after an 84 ms exposure.

Impact of contour on aesthetic judgments and approach-avoidance decisions in architectureOshin Vartanian et al. · PNAS 110 (Suppl. 2), 10446–10453 · 2013

Eighteen adults in a 3-tesla fMRI scanner saw 200 photos of interiors, half curvilinear and half rectilinear, balanced for ceiling height and openness, for 3 seconds each. Curved rooms were judged beautiful more often (Wilcoxon Z = −2.13, p < 0.05), and pleasantness explained 58% of the variance in beauty ratings; curvature lit up the anterior cingulate cortex. It did not change whether people chose to enter. The limit: an exploratory study that did not vary the degree of curvature.

Read the paper (free)

The motion half comes from Zoe Kourtzi and Nancy Kanwisher, who showed in 2000 that photos of athletes caught mid-action drove the brain's motion area MT/MST harder than photos of athletes at rest. A sweeping line from headlight to tail is the designer's version of that frozen action. William Hogarth called the S-shaped version the line of beauty in his 1753 book The Analysis of Beauty, arguing that serpentine lines look lively and straight ones look still.

G1 joint (tangent only)G2 joint (curvature too)curvature jumps: reflection breakscurvature ramps: one flowing lineκ
Same end points, two joints. The top curves look alike; the curvature plots below show why only the right one reads as a single gesture.

The link is robust for the free hang and fragile everywhere the cloth is pinned. Here is the sensitivity check, one assumption at a time.

  • Stiffer fabric. Bending length scales with \( B^{1/3} \). An 8 times stiffer Kanjivaram silk only doubles \( c \), so folds grow from roughly 2 cm to 4 cm but stay G2.
  • Heavier fabric. Doubling the mass per area shrinks \( c \) by \( 2^{1/3} \approx 1.26 \): folds get about 20% tighter, still smooth.
  • Pinned points. At the waist tuck and pleats, a concentrated force or a pressed crease is applied. There \( dM/ds \) is no longer finite, curvature can jump, and the curve drops to G1 or G0. These are the sari's "panel gaps".
  • Bike side. A designer can choose any continuity. Many modern sports bikes deliberately use sharp G1 creases, and Bar and Neta's data say sharp contours read as more threatening, which a superbike may actually want.
A sari draped on a shop mannequin, showing pleats at the waist and a smooth pallu over the shoulder
Pleats at the waist are pinned folds (G0 to G1); the pallu across the shoulder hangs free and stays curvature-continuous.Photo: Suyash.dwivedi, CC BY-SA 4.0, via Wikimedia Commons

Where does the analogy break?

It breaks in three places, and they are worth knowing because they are also design lessons.

  1. Two dimensions versus three. A sari is a developable sheet: it can bend in one direction at a time without stretching. A fuel tank has double curvature, which fabric cannot form without wrinkles. This is why a sari falls in long folds and a tank bulges.
  2. Physics versus intent. Silk is G2 because it has no choice; a fairing is G2 because a surfacing engineer checked it. The same curve, but only one of them can be wrong.
  3. Motion versus stillness. A sari actually moves; a bike only implies it. The brain research covers implied motion in photos, not drape in a breeze.

The graphic on top has to follow the surface underneath. On my KTM 390 Adventure's graphics I kept asking that each panel's original shape stay intact and the design follow its contours. The panel already had its curve, and a decal that fights it breaks the line.

Verdict: true, with one change

The link is real for the line, not the surface. A freely hanging sari and a Class A bike flank are both curvature-continuous (G2) curves that minimise \( \int \kappa^2\,ds \), the same objective a draughtsman's spline solved, and the eye responds to that continuity. The link fails at pinned points (pleats, tucks, panel creases) and in double curvature, which cloth cannot make. Change the claim to "the free line of a sari and the designed line of a bike are the same curve", and it holds.

If you want to try it on your own work, the Bézier tool in any vector app is this exact curve; vector vs raster for logos explains why that matters for print. The proportion myths that often get attached to "beautiful curves" are taken apart in the golden ratio in design, and the way shapes age is in why logos look dated. For bikes as objects rather than drawings, see rake and trail explained and the Yantra world.

What if next?

Questions people ask

What is the S-curve in design called?

William Hogarth called it the line of beauty in The Analysis of Beauty (1753). Engineers would describe a good one as a curve with continuous curvature, meaning its bend changes gradually with no sudden jumps.

What does G2 continuity mean?

Two curves meet with the same position (G0), the same direction (G1) and the same curvature (G2). G2 surfaces reflect light as one unbroken highlight, which is why car and bike bodies are checked for it.

Why does silk drape better than cotton?

Drape depends mainly on bending length and shear stiffness. A fabric with a short bending length for its weight forms smaller, softer folds; the 2023 Ijaz study found the silk blend among its six fabrics gave the best result.

Do people really prefer curved designs?

In controlled tests, yes, on average. Bar and Neta (2006) found curved versions of objects liked more often at an 84 ms glance, and Vartanian et al. (2013) found curved rooms judged beautiful more often. The effect is moderate and context-dependent.

Sources

  1. Wikipedia: Flat spline (source for ducks, shipbuilding use and Schoenberg naming splines in 1946)
  2. Wikipedia: Bézier curve (source for de Casteljau at Citroën in 1959, Bézier at Renault, and the cubic formula)
  3. Wikipedia: Class A surface (source for G2/G3 requirements in automotive surfacing)
  4. Ijaz, Namood-e-Sahar and Tariq: Evaluation of Bending Length, Rigidity and Modulus of Woven and Knitted Fabrics, Proc. Pakistan Acad. Sci. A (2023) (source for ASTM D1388 method, the rigidity formula and the 145 g/m² silk blend)
  5. Carson and Worthington: Stiffness of Paper, J. Res. NBS (1952) (source for Peirce's bending length, flexural rigidity and cantilever method)
  6. Bar and Neta: Humans prefer curved visual objects, Psychological Science (2006) (source for 67.2% vs 50.6% liking)
  7. Vartanian et al.: Impact of contour on aesthetic judgments, PNAS (2013) (source for the fMRI study of curved rooms)
  8. Kourtzi and Kanwisher: Activation in human MT/MST by static images with implied motion, J. Cognitive Neuroscience (2000) (source for implied motion activating MT/MST)
  9. Wikipedia: Line of beauty (source for Hogarth's 1753 serpentine line)
  10. Moto.it: EICMA 2025 visitor record (2025)
curvature continuityelasticadraughtsman's splineBezier curveClass A surfacefabric drapePeirce bending lengthline of beauty
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Artblog.

Keep reading in Artblog

From the other blogs