Coastline Paradox Explained: Why Length Changes

Quick answer: The coastline paradox is the fact that a coastline’s measured length depends on the size of the ruler used. A smaller ruler follows more bays, inlets and bends, so the measurement comes out longer: measuring Great Britain’s coast with 100 km steps gives about 2,800 km, while 50 km steps give about 3,400 km. Fractal geometry, developed by Benoit Mandelbrot, describes how this detail grows with scale using a fractal dimension. Real coastlines have a length at any stated resolution, but no single scale-free length.

Facts checked November 22, 2026.

How long is the coast of Great Britain? It sounds like a question with one answer you could look up. But official sources disagree by thousands of kilometers, and none of them is wrong. The reason is one of the most surprising ideas in geometry, and it led to the mathematics of fractals, now used in everything from phone antennas to the study of how AI models train.

The Paradox in One Table

Walk a pair of compasses set to a fixed step length along a map of Great Britain and count the steps:

Ruler (step) lengthMeasured coastline
100 kmAbout 2,800 km
50 kmAbout 3,400 km
Smaller and smallerKeeps growing

These are commonly cited illustrative figures, not official totals. Halving the ruler added about 600 km. A shorter ruler fits into bays and around headlands that a longer one skips. At the scale of rocks and pebbles there are still more wiggles, so the “length” never settles on one number. Published figures differ for exactly this reason, and because of rules about islands, estuaries and tides: the CIA World Factbook gives the United Kingdom (Great Britain plus Northern Ireland and the islands) 12,429 km of coastline, several times the 100 km ruler figure for Great Britain alone.

Richardson’s Discovery

Around 1951, the English mathematician and physicist Lewis Fry Richardson was studying whether the length of a shared border affects the chance of war between neighbors. He found that countries disagreed about their own borders: in the figures he collected, Portugal reported its border with Spain as 987 km, while Spain reported 1,214 km. The smaller country, measuring at a finer scale, got a longer border. Richardson found that measured length \(L\) changed with ruler length \(\varepsilon\) according to a simple rule:

\[ L(\varepsilon) \approx F \cdot \varepsilon^{\,1 – D}, \]

where \(F\) is a constant (with units) and \(D\) is a number that depends on how wiggly the line is. On a log-log plot of \(L\) against \(\varepsilon\), this is a straight line with slope \(1 – D\). For a smooth curve, \(D = 1\) and the length stops changing. For a coastline, \(D > 1\), so as \(\varepsilon\) shrinks, \(L\) grows without limit.

Mandelbrot: Fractional Dimension

On May 5, 1967, Benoit Mandelbrot published a short paper in Science titled “How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension.” He interpreted Richardson’s \(D\) as a dimension: a measure of how completely a shape fills space as you zoom in. Coastlines are self-similar: a small piece looks statistically like the whole. He reported \(D \approx 1.25\) for the west coast of Great Britain and about 1.02 for the much smoother coast of South Africa. Mandelbrot later coined the word fractal for such shapes (see the Mandelbrot set explained).

How to Calculate Fractal Dimension From Two Measurements

If \(L(\varepsilon) \approx F \varepsilon^{1-D}\) holds over a range of scales, two measurements give

\[ D = 1 + \frac{\ln(L_2 / L_1)}{\ln(\varepsilon_1 / \varepsilon_2)}. \]

With \(L_1 = 2{,}800\) km at \(\varepsilon_1 = 100\) km and \(L_2 = 3{,}400\) km at \(\varepsilon_2 = 50\) km:

\[ D = 1 + \frac{\ln(3400/2800)}{\ln(100/50)} \approx 1 + \frac{0.194}{0.693} \approx 1.28. \]

That is a rough estimate from two illustrative measurements of all of Great Britain, not Mandelbrot’s historical estimate of 1.25 for the west coast, but it is in the same range. With \(D = 1.25\), each halving of the ruler multiplies the measured length by \(2^{0.25} \approx 1.19\), about 19% longer every time.

The Koch Snowflake: A Coastline You Can Build

In 1904, Helge von Koch described a curve that shows the paradox perfectly. Start with a line segment. Replace its middle third with two sides of an equilateral triangle, making 4 segments, each \(\tfrac{1}{3}\) as long. Repeat on every segment, forever.

StepSegmentsLength of eachTotal length
0111
141/34/3 ≈ 1.33
2161/916/9 ≈ 1.78
\(n\)\(4^n\)\(1/3^n\)\((4/3)^n\)

Since \(\frac{4}{3} > 1\), \((4/3)^n\) grows without bound: the finished curve has infinite length, yet it fits inside a small box. Its dimension comes from the self-similarity: shrinking by a factor of 3 produces 4 copies, so

\[ 3^{D} = 4 \quad \Rightarrow \quad D = \frac{\log 4}{\log 3} \approx 1.262, \]

remarkably close to Britain’s coast. Three Koch curves joined in a triangle form the Koch snowflake: infinite perimeter around a finite area (exactly \(\tfrac{8}{5}\) of the starting triangle’s area).

Fractal Dimensions Compared

ShapeDimension
Straight line, smooth circle1
South Africa’s coastAbout 1.02
West coast of Great BritainAbout 1.25
Koch curve\(\log 4 / \log 3 \approx 1.262\)
Sierpinski triangle\(\log 3 / \log 2 \approx 1.585\)
Filled square2

For irregular shapes like real coastlines, scientists use box counting: cover the shape with a grid of boxes of side \(\varepsilon\), count the boxes \(N(\varepsilon)\) it touches, and estimate \(D\) from the slope of \(\log N\) against \(\log(1/\varepsilon)\), the same log-log trick used for AI scaling laws.

Fractals in Nature and Technology

  • Nature: river networks, lightning, ferns, Romanesco broccoli, blood vessels and the branching airways of the lungs all show self-similar structure over a range of scales.
  • Antennas: fractal-shaped antennas pack a long, wiggly conductor into a small space and work across several frequency bands.
  • Computer graphics: mountains, clouds and coastlines in films and games are often generated with fractal algorithms.
  • AI: in a 2024 paper, AI researcher Jascha Sohl-Dickstein found experimentally that, for the networks he studied, the boundary between training settings that converge and those that blow up is fractal, much like the edge of the Mandelbrot set. Tiny changes to the learning rate can flip the outcome. That is a finding about training, not a claim that AI systems themselves are fractals.

Do Real Coastlines Have Infinite Length?

No. The Koch curve has mathematically infinite length, but a physical coast is not infinitely detailed: at some point you reach grains of sand, and tides, erosion and map resolution all blur the line. Natural shapes behave like fractals only over a limited range of scales. So the useful answer is “this long, as measured at this scale.” A good figure gives the ruler length or map resolution along with the number. Real coastlines are not infinitely detailed (at some point you reach grains of sand, and tides move the line), so the true lesson is not “infinite length” but that length depends on scale, and fractal dimension describes how.

Common Misconceptions

“Coastlines are literally infinitely long.” Ideal fractals are; physical coasts are not, but within the scales we measure, length keeps growing as the ruler shrinks.

“A dimension must be a whole number.” Fractal dimension measures how a shape’s detail scales, and it can be any number between 1 and 2 for a curve.

“Someone measured it wrong.” Different official figures usually reflect different scales and rules, such as whether islands and estuaries are included.

Watch: The Coastline Paradox

If you learn better by watching, these videos are worth your time.

What Is The Coastline Paradox? (Veritasium). A short classic on fractals and the paradox.
Measuring Coastline – Numberphile (Numberphile). A hands-on demonstration of measuring a coastline with different ruler sizes.
The Coastline Paradox Explained (RealLifeLore). A map-based tour of why coastline figures disagree.

Check Your Understanding

Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.

Answer each question and press Submit to see your result right away.

1. What happens to a coastline’s measured length when you use a shorter ruler?
Found a bug? Let us know!
2. Each Koch step multiplies the length by what factor?
Found a bug? Let us know!
3. What is the Koch curve’s fractal dimension?
Found a bug? Let us know!
4. Britain measures 2,800 km at a 100 km ruler and 3,400 km at 50 km. Estimate D.
Found a bug? Let us know!
5. Why does our estimate of 1.28 differ from Mandelbrot’s 1.25?
Found a bug? Let us know!

Frequently Asked Questions

What is the coastline paradox?

The finding that a coastline has no well-defined length: the shorter the measuring unit, the longer the measured coast, because finer scales reveal ever more bays and bumps.

How long is the coastline of Great Britain?

It depends on the scale: about 2,800 km with 100 km steps and about 3,400 km with 50 km steps. The CIA World Factbook lists 12,429 km for the whole United Kingdom, measured with much finer detail.

Is a coastline infinitely long?

Not physically. The ideal Koch curve has infinite length, but real coasts only behave like fractals over a limited range of scales, so a coastline has a definite length at any stated measurement scale.

What is the fractal dimension of Britain’s coastline?

Mandelbrot reported about 1.25 for the west coast of Great Britain, using Lewis Fry Richardson’s measurements.

Why does the Koch snowflake have infinite perimeter?

Each step multiplies the length by 4/3, and (4/3)^n grows without bound, while the area stays finite at 8/5 of the starting triangle.

Who discovered the coastline paradox?

Lewis Fry Richardson noticed the effect around 1951 while studying borders, and Benoit Mandelbrot explained it with fractional dimension in a 1967 Science paper.

Keep Exploring

Sources and Further Reading

Related to This Article

What people say about "Coastline Paradox Explained: Why Length Changes - Effortless Math"?

No one replied yet.

Leave a Reply