I’m just going to say it!
Dividing colors into “warm” and “cool” categories is a common way to organize them and think about color theory. (It’s a scheme I use to organize my dot card swatches, for example.) But it never really made a lot of sense to me, and now I know why. It’s inherently confusing!
What’s the big idea?
Think about a color wheel with six colors: the primaries (magenta, yellow, and cyan), and the secondaries (the colors made by mixing them: magenta + yellow = orange, yellow + cyan = green, cyan + magenta = purple).
You can divide these six slices evenly into two categories: three warm colors (red, orange, yellow) and three cool colors (green, blue, purple).

You’ve probably seen this before; it feels pretty intuitive, right?
Take a second to notice where the three primaries fall here. Since there are only three of them, they are really divided unevenly between warm and cool; two of them (magenta and yellow) are warm, and only cyan is cool. This will cause problems later.

Of course, the color wheel is really a spectrum. We just arbitrarily divide it into six wedges by convention. You could divide it into any number of wedges. Within each wedge, there will also be a spectrum of colors. For example, the “purple” wedge covers from a blue-purple to a magenta-purple. The “magenta” wedge really covers from a purple-magenta to a red-magenta. And so on.

When you talk about a “cool” or a “warm” version of color, you are talking about where it falls along its color wheel wedge: more toward the cool side or the warm side? A “cool” purple is more blue, and a “warm” purple is more magenta.
The Problem
The problem is that this system only works for the four color wheel slices along the warm-cool divide. For orange and cyan, it doesn’t work at all.
What’s a warmer cyan: one with more magenta in it (a purple-blue), or one with more green in it (a turquoise)? You could make the argument for either, since both magenta and green are warmer than cyan. By convention, purple-blues are typically called “warm” and green-blues are typically called “cool,” but this doesn’t feel intuitive or logical to me; I just memorized it.
Orange is another condundrum. If this system works, you should be able to talk about any color having a “warm” or a “cool” version. But no matter which direction you go for orange – more red or more yellow – you are still warm. There is no such thing as cool orange.
These are the kinds of problems that happen when you try to impose an arbitrary binary on a trinary system.
The arbitrary warm and cool system results in some major issues.
Major Issue #1: Temperature and Color Mixing
One bit of received wisdom that I tried to make work for a long time is this: you get more vibrant color mixes when you mix within a temperature. For example, mixing a cool with a cool, or a warm with a warm, is supposedly more vibrant than mixing a cool with a warm. This sounds good on its face, but I found it didn’t really stand up to testing.
I tried all sorts of ways of painting oranges, for example, and found that while warm red and warm yellow worked best for creating a vibrant orange (what you’d expect), all the other combinations looked pretty similar (and not too bad really).
Meanwhile, while I ought to have been able to paint a warm purple by combining warm red and warm blue, I found that any combination with warm red was terrible and just turned brown.
Through experimentation, I landed on the following “rules”, none of which seemed to have any unifying factor:
- vibrant orange = warm red + warm yellow
- vibrant green = cool yellow + cool blue
- vibrant purple = cool red + warm blue
A different rule clicked, however, once I read this article on color temperature from the small batch paint company Greenleaf & Blueberry. Instead of thinking about colors in terms of “warm” or “cool,” think about which color wheel neighbor they tend toward.
- Red: Instead of “cool to warm,” think “purple-toned to orange-toned”
- Yellow: Instead of “cool to warm,” think “green-toned to orange-toned”
- Blue: Instead of “cool to warm,” think “green-toned to purple-toned”
This removes the “warm blue” problem! It also works for the secondaries (or any other color): greens are yellow- or blue-toned; purples are red- or blue-toned; and oranges are red- or yellow-toned.
And now the color mixing rules make total sense and are easy to remember:
- vibrant orange = orange-toned red + orange-toned yellow
- vibrant green = green-toned yellow + green-toned blue
- vibrant purple = purple-toned red + purple-toned blue
Isn’t that so much simpler?
Major Issue #2: The Split Primary Palette
A common way of designing a minimalist, six-color watercolor palette is called “split primary”: a warm and cool version of each primary color. For example:
Color Family | Cool | Warm |
---|---|---|
Red | Cool Red (Magenta) Quinacridone Rose, Purple Magenta, Carmine | Warm Red (Orange Red) Vermilion, Cadmium Red, Pyrrol Scarlet |
Yellow | Cool Yellow Lemon Yellow, Hansa Yellow Light | Warm Yellow (Yellow Orange) Hansa Yellow Deep, New Gamboge |
Blue | Cool Blue (Cyan) Phthalo Blue (Green or Red Shade), Phthalo Turquoise, Cerulean | Warm Blue (Violet Blue) Ultramarine Blue, Indanthrone Blue, Cobalt Blue |
The Daniel Smith Essentials set, which I started off with, contains one paint from each category, for example. This is the system I originally advised in How to Build a Watercolor Palette from the Ground Up (I’ve since changed it).
This system has some conceptual problems. Aside from the issues with “cool” and “warm” blue identified above:
- It inherently relies on a “red”, “yellow”, “blue” primary system which is not accurate. Magenta isn’t “cool red,” it is a primary. Similarly, cyan isn’t “cool blue,” it is a primary. A split primary system based on the actual primaries wouldn’t work as well because it would contain near-primaries instead of the primaries themselves (e.g. plum purple and fire engine red instead of magenta). The reason this works is accidental: because it misidentifies the primaries and insists on containing near-primaries, it accidentally contains the primaries!
- The one primary it correctly identifies is yellow, and that’s the reason yellow doesn’t work as well: this system doesn’t contain a middle yellow, which would be a better mixing color than either “cool yellow” or “warm yellow.”
The system also has some practical problems:
- As noted above, both of the yellows are non-ideal. One of the first modifications I made to this palette is switching to a middle yellow (Pure Yellow) which I found a lot more useful and versatile than either of the split primary yellows. I still have yet to use New Gamboge for much.
- The second modification I was eager to make? Adding green! Green is so abundant in nature, and it’s weirdly hard to mix with this set. Sure, it’s blue and yellow, but not all the blues and yellows in the split primary set even mix green at all (New Gamboge + Ultramarine = ugly gray). Second, you don’t always want the bright neon green mixed by Hansa Yellow Light and Phthalo Blue GS, so you have to temper it with a red – meaning you’re mixing all three colors in some way just to get a major, important color. Life is a lot easier when you just add a green or two.
Switching to the Secondary Palette
I had an “ah-ha” moment when reading Bruce MacEvoy’s takedown of the split primary palette and proposed solution: the secondary palette. The secondary palette is an alternative way of designing a six-color minimalist or starter palette that is a lot more like the six-step color wheel I created above.
As usual, you begin with the real primaries (magenta, yellow, and cyan), but your next three colors – instead of “warm versions” of those – are the secondary colors in between them. The idea is to reduce the amount of steps across the color wheel you need to cover to mix the color you want.
MacEvoy is more precise in locating those mid-steps than I was in my color wheel above; they are not simply orange, green, and purple, but more accurately: red-orange, green, and blue-violet.
Here’s a look at the slots and some options for them in this system:
Primary | Secondary |
---|---|
Magenta Quinacridone Rose, Purple Magenta, Carmine | Red-Orange Vermilion, Cadmium Red, Pyrrol Scarlet |
Yellow Pure Yellow, Hansa Yellow Medium, Nickel Azo Yellow | Green Phthalo Green (Blue or Yellow Shade), Viridian, Sap Green, Hooker’s Green |
Cyan Phthalo Blue (Green or Red Shade), Phthalo Turquoise, Cerulean | Blue-Violet Ultramarine Blue, Indanthrone Blue, Cobalt Blue |
One thing to immediately notice about this system is that it’s functionally super similar to the split primary system! It just switches the cool yellow to a middle yellow and swaps out relatively useless orange-yellow for a super-useful green. It solves all the functional problems of the split primary palette while also working better theoretically because it is no longer beholden to the fake idea of warm and cool.
What Is the Cool and Warm Binary Good For?
The “cool and warm” framework isn’t totally useless. There are useful some things you can do with it.
Evoking an Emotion
Warm colors tend to be more energetic and to evoke more stimulating, “hot” emotions. My first associations with warm colors are positive: happiness, optimism, joy, passion. But there are also negative warm emotions, like anger, hostility, and fear. Cool colors, by contrast, tend to be calmer. I associate cool colors with sadness and depression, but also with peace and relaxation.
Creating a Sense of Depth
In a painting, warm colors will tend to jump out or come forward, while cool colors will tend to recede to the back. This is because particles in the air create a cool grayish haze between you and faraway objects. This is also why things look lighter further back.
Identifying the Light Source
The side of an object that’s closer to the light source (e.g. the sun) will tend to be warmer-colored in addition to being lighter in value, while shaded sides will tend to be bluer.
Conclusion
Cool and warm colors can be used effectively to create various effects in paintings, and it’s a useful dichotomy to know about. But when it comes to grouping paints or theorizing about mixes, it does more harm than good. From now on, on the blog, when I talk about different shades of a color, I will try to be specific about what I mean in terms of the color bias.
Reminds me of John Muir Laws asserting that the primaries are c, y, m not r, g, b. You seem to be starting from there and going even further.
I recommend his “value does the work, colour gets the credit” video. Which might shed some light on his palette choices for your other blog on that topic.
Your blog is so much fun, is there a way to subscribe? I think you meant to put the phthalo greens in the green box, and not the phthalo blues. I love the secondary palette, Bruce is very convincing and it does just make more sense to use colors with a wider spread in hue.
Correct!! I updated the list of colors.