Primary, Secondary, and Tertiary Shapes
By Neil Blevins
Created On: Sept 28th 2006
Updated On: Oct 31st 2017

Go here to read this tutorial in Indonesian.

Go here and here for a fantastic talk on the subject by Gleb Alexandrov.

This theory in composition comes in many names.

• Primary, Secondary, and Tertiary Shapes
• Big, Medium, Small
Whatever the name, the idea is the same, if your image has a nice distribution of big (primary), medium (secondary) and small (tertiary) shapes, the resulting image will tend to be more pleasing to the eye.

First, a definition. Primary shapes are your big shapes. If you squint at an image, the details tend to disappear and you're left with only your big shapes.

Secondary shapes are the smaller shapes that either sit ontop of, or help make up the primary shapes.

Tertiary shapes are again smaller than the secondary shapes.

Here's three 2D diagrams showing primary, secondary and tertiary shapes in a fake composition. Even though this is 2d, the exact same rules apply to 3d.

First off, what should the size of these details be? Well, that depends on personal taste. I like having the primary details being huge, almost the size of the image itself. And I like Tertiary details being really small, sometimes the size of a pixel or two. From there you can decide the appropriate size for the Secondary Details. Of course, not all details should be the same size, it's more like a range of sizes, so not all secondary details should be the same size, but should be in a similar range.

So these could be primary, secondary and tertiary shapes....

But so can these...

And these...

Next comes the distribution of these details. Take a look at this image, it has all 3 levels of detail, but the tertiary shapes are in one uniform block...

Many images suffer from this problem, huge blocks of small repeating patterns. For example, you get this effect if you place a procedural noise with no variation on a surface.

This does not provide the eye any spot to rest. So in general you want to vary the placement of your tertiary details so that they are more random, and have several areas of detail, and several areas of no detail.

Notice how this is easier to digest. The eye can look at the detailed lower right corner, then move to the center of the image and rest there for a moment (in the big area of little detail) before looking at more tertiary details elsewhere.

This is another reason why natural patterns generally look better than out of the box procedural patterns. Natural patterns tend to have areas of lots of detail, and areas of very little detail, and they're placed in a more random fashion. Of course, through mixing and what not, you can get a more "organic" quality from procedurals, but procedurals right out of the box generally don't give you that kind of variety. See my Complex Shaders Using Procedurals, The Green Tongue lesson for a discussion on how to get more variety from your procedurals.

Here's a photo of a rock in greyscale. Notice the variety in the patterns.

Here's another example of building up shapes. First, lets start with a cylinder, maybe this will be a pipe for a robot.

Now lets place some secondary shapes, rings around the pipe.

Now that's more interesting, but the regular pattern is kind of boring. Lets vary the position of the secondary rings...

Better, but all the secondary shapes are the same size, lets vary the sizes a bit...

Looking better. We now have variety in the sizes of the secondary shapes, and we have areas with lots of secondary shapes, and areas with no secondary shapes. Lets add a few tertiary shapes...

Now we have a well balanced composition.

Here's a few paint overs of some of my images showing the primary, secondary and tertiary shapes...

In this case, the giant tower is the primary shape (a cylinder). That shape is made up of many secondary shapes, pipes and large pieces of concrete, and the entrance. The tertiary shapes are the tiny ships, and tiny lights along the surface of the tower, many of which are just a few pixels large. This helps give the piece a good sense of scale.

The primary shape is the shape made by the creature against the background. Then there's some larger tentacles, which are secondary shapes. Then there are more tiny tentacles, and grass, etc., which are tertiary shapes. If the image had no tertiary shapes, it would feel like something's missing.

Robots are a great opportunity for practicing "Big, Medium, Small". In this image, the larger shape of the creature is the primary shape. Then that has secondary shapes on it like the front panel or eye. Then those have tiny tertiary shapes on them as well, such as nuts, panel lines, small raised panels, etc. Also note that there aren't tertiary shapes everywhere on the secondary shapes, there are areas of lots of detail, and areas of no detail. Also notice just how small some of the tertiary shapes are compared to the secondary shapes.