Similarity has bothered me for a long time. Why do all circles have the same formula for area â€” how do we know nothing sneaky happens when we make them larger? In physics, donâ€™t weird things happen when you scale things (particles, insects, small children) to gargantuan sizes? Youâ€™re saying that every circle has the same formula, yet a 300-foot honeybee cannot fly?

Hereâ€™s the deal:

Similar shapes are zoomed versions of each other. Because we canâ€™t tell them apart (read on for why), they must have the same internal formulas for area, perimeter, and so on.

However, items with the same formula arenâ€™t exchangeable! Sure, all humans (from children to NBA players) have the â€śformulaâ€ť that armspan = height. Fine â€” does that mean a 7 foot Sasquatch and my 18-inch nephew are equally good basketball players?

**My â€śaha!â€ť moment was separating the common formula (armspan = height) from the applicability of individual instances (infant vs. Sasquatch).**

## Why Scaled Objects Have The Same Formula

Thought experiments helped me realize that absolute size doesnâ€™t matter when figuring out whether a formula can hold for all instances of a shape. An insight is that perceptions of â€śsizeâ€ť are often determined by us, the observer, and not the shape itself.

**Field of Vision**

Imagine a triangle on a piece of paper. It takes up some amount of room in your field of vision â€” say, 30%.

Now, move closer to the paper, so the triangle takes up most of your view â€” say, 90%. What changed? The triangle is the same, but the sides appear much bigger. Yet we know that the core properties (area, perimeter, etc.) havenâ€™t changed â€” otherwise, weâ€™d need to know someoneâ€™s distance from a triangle when calculating area!

**Create a Tube**

This time, imagine a paper circle. Now, thicken the paper until you make a cylinder (of equal diameter) extending into the distant horizon. The part of the cylinder a few feet off â€ślooksâ€ť to be 1/2 the size of the disc before you, yet you know they are the same size. The ratios inside (circumference to diameter, area to radius, etc.) must be the same also.

**Photoshop Zoom**

Imagine a triangle on a computer screen. Make all sorts of formulas for area, perimeter, and so on. Now, zoom the triangle by 300% and measure again. What changed? Sure, everything was bigger on the second measurement, but does the triangle â€śknowâ€ť itâ€™s being zoomed and change itself to make the formulas different?

**Measurement Unit**

Suppose youâ€™re measuring ratios on a shape with your trusty ruler. You have your table full of figures: area to perimeter, diagonal to side, and so on. But whoops! It looks like you had written â€ścmâ€ť when you were actually measuring the sides in inches.

Do you need to redo your table because you were using the wrong unit? Does the shape know what units youâ€™ve been using?

## Enter Physics

My conundrum started when remembering a factoid from biology class: Godzilla couldnâ€™t exist because he would overheat. Big lizards canâ€™t do the same things little lizards can.

Why?

For simplicity, letâ€™s assume Godzilla is a lizard cube. For any cube of side â€śsâ€ť, the volume is s^{3}, and the surface area is 6 * s^{2}.

Now, letâ€™s assume that heat generated is proportional to volume (essentially your mass), and cooling is proportional to surface area (amount of skin you have exposed to that cool, breezy air). How much cooling do you have for each unit of ~~girth~~ mass?

For every unit of volume, we have 6 / s â€śsurface area unitsâ€ť available to cool it. If s = 1cm (for example), then we have 6 / 1 = 6 square centimeters to cool ourselves for each cm^{3} of volume. Great.

But what if s = 10cm? Uh oh. Now we have 6 / 10 = .6 square centimeters of cooling. And if s = 100cm we only have .06 sq cm of cooling. At some point, our cooling cannot balance our heat requirements and Godzilla falls over. (Suppose he needs at least .1 square cm of cooling for each cubic cm to stay alive).

Remember our insight:

- s
^{3}, 6 * s^{2}, and 6 / s are common patterns in all cubes, no matter the size - 6 / 1 = 6, 6 / 10 = .6, 6 / 100 = .06 are
**particular instances**of the â€śsurface area to volumeâ€ť formula. Some meet our heat requirements, others donâ€™t.

If youâ€™re interested, there are other structural problems with a lizard of that magnitude.

## Examples Abound

The idea of finding patterns in similar shapes (and separating them from specific examples) is ubiquitous in math and the sciences. Hereâ€™s a few examples of â€śsimilarityâ€ť which often arenâ€™t labeled as such.

**Discovering Pi**

Pi is the most famous example of similarity â€” all circles share the same ratio (circumference / diameter = pi). Again, no matter how much we zoom to make one circle appear like another, every circle has this fundamental trait.

**The Physics of Spheres**

A sphere is the most space-efficient shape â€” it gives the most volume for the least surface area. No matter if youâ€™re an elephant or mouse, youâ€™ll conserve the most heat by curling into a ball.

Planets and raindrops are spherical because of these unique properties â€” even though the scale difference for each example is enormous.

**Trigonometry**

Sine, cosine, and the rest of the trig family work off angles. And angles are perfect for similarity since size doesnâ€™t matter (how long do the sides of a 45 degree angle need to be? It doesnâ€™t matter!).

Because triangles with the same angles are similar, we can use the ratios inside one (i.e, triangles that fit inside the unit circle) and scale up the result for any example we need.

**Algorithm Running Time**

The running time of algorithms [O(n), O(n*log(n)), O(n^{2}), etc.] are based on finding a â€śsimilarity classâ€ť describing the runtime. An algorithm with O(n^{2}) should run 4x as slowly when the number of inputs are doubled.

However, for **specific instances**, the desired algorithm may be different: running 10 inputs with a O(n^{3}) algorithm can be faster than running 10 million inputs with an O(n).

**Object Oriented Programming**

In programming, members of the same class (â€śsimilarityâ€ť class!) may share formulas like Area = π r^{2}. However, each **instance** of that class may have a different value of â€śrâ€ť. The class provides the general patterns while the instances provide the details.

## Closing Thoughts

A few observations:

- Separate the common formula from particular instances of a shape. All circles are similar, but a bigger pizza is better than small one, right?
- Analogies help us remember. I have silly reminders about infant NBA players and â€śGodzilla cubesâ€ť that makes the â€śpattern vs exampleâ€ť concept more clear.
- The idea of similarity is broader than just geometry â€” itâ€™s about identifying classes of items that share the same internal properties.

The actual definition of similarity is more nuanced; you can reverse it and say shapes are similar if formulas based on their distance are always the same (they are uniformly scaled or dilated). But, those are fun diversions for another day â€” happy math!

## Leave a Reply

25 Comments on "Understanding Why Similarity Works"

This post get’s my up-vote because it uses the term, “lizard cube”.

@Bill: Hah — and just a few steps away from time cube :).

Enter: Equivalence class

Things get really interesting when you start to ask about why shapes are seen as similar in general eg when a child sees ‘M’ as a seagull in flight etc

http://www.foveola.com/demo/index.php

@Anon: đź™‚

@Patrick: Neat link! Yeah, the psychological aspect of similarity is an interesting one — our brains are pattern recognition machines.

I ask the question, “If I move closer to this page, will the words change?”

And the answer gives me more insight on similarity than any geometry class I’ve taken.

Thank you Kalid đź™‚

Similarity is in fact just a particular case of projective geometry (hence the link with perspective), and precisely the part which leaves angles unaltered.

In fact, nearly any kind of geometry can be seen as a particular case of projective geometry, that’s the “erlangen program”…

Quite a beautiful thing, and sadly one that is largely unspoken for in school!

@Prudhvi: Thanks, that’s what helped it click for me too — it doesn’t matter how far away we are, or whether it “looks” big or small :).

@Johann: Thanks for the background! I hadn’t thought of that, but you’re right — and the neat aspect is really that it’s all from the observer’s viewpoint (does this shape look similar to another one from the eyes of this other person?). Quite true, there’s so much beyond what we learn in school :).

Hey!

Lizards are poikilotherm (“cold-blooded”), they don’t produce any body heat!

@Anonymous: *slaps my forehead*. Um… imagine that GodZilla is really a giant human in a lizard suit.

interesting insights into things …like reading ur posts.

thanks and keep posting.

regards,

virender.

@Virender: Thanks, will do đź™‚

[…] scaled triangle (2x) and plop on another scaled triangle (times 3i). Even though it’s larger, similar triangles have the same angles — they’re just bigger (but don’t ask about its size, […]

No matter if youâ€™re an elephant or mouse, youâ€™ll conserve the most heat by curling into a ball.

Now i know how to stay warm! YES!

@Arnoldo: Haha, you got it! :).

6/s how insightful.

@Mark: Thanks! đź™‚

It’s sentences like “For simplicity, let’s assume Godzilla is a lizard cube” that really make me appreciate math as artful approximation.

@Mothra: Hah — that it is!

Next year when we start Common Core Geometry, this is exactly what we’ll be starting with. Thanks for the fresh perspective!

@AMSteele: Great, glad it helped!

[…] from similarity, ratios like “height to width” must be the same for these triangles. (Intuition: step […]

very interesting. but, isn’t lizard cold blooded and does not need to cool itself?

True :). But even so, the surface area would determine how quickly the outside air flowing by would cool the lizard cube.