Search for shades of blue and you get a grid of forty or seventy swatches with names attached: cerulean, cobalt, periwinkle, glaucous, YInMn. The reasonable suspicion is that most of them are the same color wearing different labels, invented by someone who needed a longer post. That suspicion is testable, so I tested it.
I collected 67 named blues, every blue and cyan in the CSS named-color set plus the shades that recur across the mainstream lists with a settled hex code, and ran all 2,211 pairs through CIEDE2000, the perceptual color-difference formula this site uses to score guesses. The result went against my expectation. Only five pairs sit below the threshold where two colors stop being distinguishable. The lists are broadly honest.
The interesting failure is somewhere else. Side by side, under good light, with both swatches touching, almost every name earns its place. That is not the situation anyone uses color names in. You use them from memory, across a room, an hour after seeing the reference. Rerunning the same measurement at memory tolerance instead of side-by-side tolerance turns 67 names into 18. Below are the names, the hex codes, the numbers, and what the gap between those two figures is actually telling you.
The named shades of blue with hex codes
Grouped by lightness rather than alphabetically, because lightness is what separates them in practice. L* is CIELAB lightness on a 0 to 100 scale. If the six-digit notation needs unpacking, how to read hex color codes covers what the digits do.
Dark blues, L* below 35
- Oxford blue, #002147. L* 12.9. The darkest named blue in the set.
- Navy, #000080. L* 13.0.
- Dark blue, #00008b. L* 14.8. Measured 1.56 from navy, which is below the threshold at which anyone can see a difference. Two CSS names, one color.
- Midnight blue, #191970. L* 15.9.
- Space cadet, #1d2951. L* 17.7.
- Federal blue, #0a2472. L* 18.1.
- Prussian blue, #003153. L* 19.3. Named for the pigment, the first modern synthetic, made by accident in Berlin around 1706.
- Yale blue, #00356b. L* 22.3.
- Medium blue, #0000cd. L* 25.0. A misleading name: it sits with the dark blues, not the middle ones.
- International Klein Blue, #002fa7. L* 26.2.
- Indigo dye, #00416a. L* 26.2.
- Egyptian blue, #1034a6. L* 27.6. Measured 1.42 from Klein blue, so the two are the same swatch with four and a half thousand years between them.
- Dark slate blue, #483d8b. L* 30.8.
- Blue, #0000ff. L* 32.3. The darkest of the six pure hues by a long way, and the reason this whole family behaves the way it does.
- Cobalt blue, #0047ab. L* 32.8.
- Ultramarine, #3f00ff. L* 34.6.
- YInMn blue, #2e5090. L* 34.7. Found at Oregon State in 2009 when a graduate student heated manganese oxide and got a pigment nobody expected.
Mid blues, L* 35 to 65
- Sapphire, #0f52ba. L* 37.3.
- Denim, #1560bd. L* 41.5.
- Majorelle blue, #6050dc. L* 43.3.
- Savoy blue, #4b61d1. L* 45.1.
- Slate blue, #6a5acd. L* 45.3.
- Han blue, #446ccf. L* 47.6. A synthetic barium copper silicate used in China from around 500 BC.
- Royal blue, #4169e1. L* 47.8.
- True blue, #0073cf. L* 48.0.
- Cerulean, #007ba7. L* 48.2.
- Teal, #008080. L* 48.3.
- Medium slate blue, #7b68ee. L* 52.2.
- Dark cyan, #008b8b. L* 52.2.
- Steel blue, #4682b4. L* 52.5.
- Glaucous, #6082b6. L* 53.8. The color of a blue-gray plant bloom, the frost on a grape.
- Azure, #007fff. L* 54.4. Not to be confused with the CSS color also called azure, #f0ffff, which is a near-white. One name, two official hex codes, 44 units of lightness apart.
- Air force blue, #5d8aa8. L* 55.4.
- Bleu de France, #318ce7. L* 57.2. The exact median of the whole set.
- Tufts blue, #3e8ede. L* 57.7. Measured 1.01 from bleu de France, the closest non-identical pair in the list.
- Dodger blue, #1e90ff. L* 59.4.
- Cadet blue, #5f9ea0. L* 61.2.
- Carolina blue, #4b9cd3. L* 61.6.
- Blue-gray, #6699cc. L* 61.6.
- Cornflower blue, #6495ed. L* 61.9.
- Vista blue, #7c9ed9. L* 64.8.
Light blues, L* above 65
- Iceberg, #71a6d2. L* 66.1.
- Tiffany blue, #0abab5. L* 68.4.
- Deep sky blue, #00bfff. L* 72.5. The name says deep, the measurement says light. Names are unreliable in this family.
- Dark turquoise, #00ced1. L* 75.3.
- Maya blue, #73c2fb. L* 75.6.
- Medium turquoise, #48d1cc. L* 76.9.
- Pale cerulean, #9bc4e2. L* 77.2.
- Light steel blue, #b0c4de. L* 78.5.
- Sky blue, #87ceeb. L* 79.2.
- Baby blue, #89cff0. L* 79.7. Measured 1.43 from sky blue. Interchangeable in every practical sense.
- Light sky blue, #87cefa. L* 79.7.
- Cornflower (Crayola), #9aceeb. L* 80.2.
- Turquoise, #40e0d0. L* 81.3.
- Periwinkle, #ccccff. L* 83.6.
- Light blue, #add8e6. L* 83.8.
- Columbia blue, #b9d9eb. L* 85.0.
- Powder blue, #b0e0e6. L* 86.1.
- Pale turquoise, #afeeee. L* 90.1.
- Cyan and aqua, both #00ffff. L* 91.1. Not similar. Identical. CSS ships two names for one value, which is the only genuinely redundant entry in the entire set.
- Electric blue, #7df9ff. L* 91.4.
- Aquamarine, #7fffd4. L* 92.0.
- Celeste, #b2ffff. L* 95.2.
- Alice blue, #f0f8ff. L* 97.2.
- Light cyan, #e0ffff. L* 97.9.
- Azure (CSS), #f0ffff. L* 98.9. A blue so pale that against paper it reads as white with a fault.
Which named blues are actually the same color
A difference of ΔE00 2.3 is the standard just-noticeable-difference threshold, the point below which two adjacent patches stop being separable by a normal observer. Mahy and colleagues established it in 1994 and it has held up as a working figure ever since. Across all 2,211 pairs in the set, exactly five land under it.
- Aqua and cyan, ΔE00 0.00. The same hex code twice.
- Bleu de France #318ce7 and Tufts blue #3e8ede, ΔE00 1.01.
- Egyptian blue #1034a6 and International Klein Blue #002fa7, ΔE00 1.42. Yves Klein registered his formula in 1960 as a color nobody had achieved. The Egyptians were firing theirs in kilns around 2500 BC and the two are visually the same.
- Sky blue #87ceeb and baby blue #89cff0, ΔE00 1.43.
- Dark blue #00008b and navy #000080, ΔE00 1.56. Both are CSS keywords, both are in every browser, and no display can show you the difference.
Another 33 pairs fall between 2.3 and 5, the band where a difference exists but only survives direct comparison. Light sky blue and sky blue are 4.55 apart. Light blue and powder blue are 4.27. Turquoise and medium turquoise are 5.00. Put any of those side by side and you will see it. Put a wall between them and you will not.
So the honest verdict on the shades-of-blue lists is better than expected: 62 of 67 names describe a color you can distinguish from every other name on the list, provided you are looking at both at once.
The number that matters: 18
Side-by-side discrimination is a laboratory condition. What you do with a color name is hold one color in your head and compare it to a color in front of you, which is a different task with a much coarser tolerance. How much coarser depends on the delay, the size of the patch and the light, so rather than pretend there is one figure, I clustered the 67 names at three tolerances and let the shape of the collapse speak. Each name merges into the first cluster it falls within.
- ΔE00 2.3, side by side under good light: 62 distinct blues.
- ΔE00 5, a glance apart, or across a paint chip border: 42 distinct blues.
- ΔE00 10, from memory, which is how names are actually used: 18 distinct blues.
Eighteen. That is what the vocabulary is worth in practice, and the buckets are unflattering. Cornflower blue, dodger blue, azure, air force blue, Carolina blue, iceberg, blue-gray, glaucous, bleu de France, Tufts blue and vista blue collapse into a single bucket: eleven names, one rememberable color. Cobalt, sapphire, denim, Yale blue, indigo dye and YInMn collapse into another. Navy, dark blue, medium blue, midnight blue and federal blue collapse into a third.
This is not an argument that the names are worthless. Cobalt and sapphire are genuinely different if you are holding the two paints. It is an argument that the names carry far less information than the length of the list implies, and that anyone using them to communicate a color across a phone call is working with an 18-item vocabulary while believing they have 67.
Why blue in particular collects so many names
Three things are true of blue that are not true of the other families, and they compound.
It is the darkest hue in the gamut
Pure blue #0000ff has a CIELAB lightness of 32.3. Pure yellow is 97.1, cyan 91.1, green 87.7, magenta 60.3, red 53.2. Blue is not slightly darker, it is 21 units below the next darkest and 65 below yellow. Every light blue is therefore a heavy compromise, a large amount of white mixed in, and the mixing ratios where useful colors live are spread thinly across a wide range. Twenty-seven of the 67 named blues sit below L* 50 and the mean is 57.3, so the family fills lightness far more evenly than the red or yellow families can.
It occupies more perceptual room than its share of the cube
I sampled the sRGB cube at every third value, 636,056 colors, discarded the near-grays, and greedily packed each hue family at the ΔE00 2.3 threshold to count how many distinguishable colors each one holds. Blue, taken as HSV hue 195 to 255 degrees, is 16.7 percent of the saturated cube and yields 1,953 distinguishable colors. Green covers 90 degrees and 25.0 percent of the cube and yields 1,951, essentially the same number from half again as much space. Red yields 1,478 and yellow 1,186. Blue packs more perceptual variety into less of the color cube than any other family, which gives naming more to work with.
Those counts are floors, not ceilings. Refining the sample to every second value raises the blue count to 2,181 and it is still climbing, so the true figure is higher. The comparison between families holds because all four were counted the same way. If the general question interests you, how many colors can you see works through the whole-gamut version of it.
Language spent longer on it
Berlin and Kay’s survey of basic color terms found blue to be one of the last to appear as languages develop their vocabularies, arriving after black, white, red, green and yellow. Late arrival meant it entered vocabularies at a point when trade and dye technology were already producing distinctions worth naming. Russian went further and never settled on one term at all, splitting light blue and dark blue into goluboy and siniy as separate basic colors. Winawer and colleagues showed in 2007 that Russian speakers are measurably faster at discriminating blues that cross that boundary, which is the rare case of a color vocabulary changing what the eye does rather than just what the mouth says.
Where the hue slider lies to you
This one is practical and it costs people points in the color matching game constantly. HSV hue degrees are not perceptually even, and the blues are where the unevenness is worst.
Walking the fully saturated hue circle and reading the CIELAB hue angle at each step:
- HSV 180, #00ffff, CIELAB hue angle 196.4
- HSV 195, #00bfff, 247.5
- HSV 210, #0080ff, 284.8
- HSV 225, #0040ff, 301.2
- HSV 240, #0000ff, 306.3
- HSV 255, #4000ff, 307.6
- HSV 270, #8000ff, 311.7
From 180 to 225, forty-five degrees of slider, the perceived hue swings 104.8 degrees. From 225 to 270, the identical forty-five degrees of slider, it swings 10.5. The back half of the blues is a plateau. Drag the hue control from cyan towards blue and the color changes fast; keep dragging past #0040ff and almost nothing happens to hue, while lightness and chroma keep moving underneath you.
The practical consequence is that when a blue target is wrong, hue is rarely the reason. Somewhere past 225 degrees the hue control has run out of authority and the error is in saturation or brightness. Players who keep nudging hue on a deep blue are adjusting the one axis that has stopped responding. Training your eye for color covers how to build the reflex of checking lightness first.
Navy against black, and other things people get wrong
The standing complaint about navy is that it looks black. Measured, navy and black are 29.70 apart on CIEDE2000, which is not marginal, it is thirteen times the just-noticeable threshold. Midnight blue against black is 27.33. Oxford blue, the darkest name in the set, is still 18.97 from black.
So the confusion is not a color-difference problem, it is a light problem. In dim conditions cone response falls away and the eye shifts towards rod vision, which carries no color information at all, and the blue simply stops being available. Anyone who has picked a navy sock and a black sock out of a drawer at 6am has run the experiment. Under a lamp the two are obviously different, which is also why buying navy under shop lighting goes wrong so often. The underlying mechanism is the same one behind color constancy, the eye discounting the illuminant and getting it wrong at the edges.
The other systematic mistake is treating light blues as usable text colors. Contrast ratios against white, where WCAG asks for at least 4.5:1 for body text:
- Oxford blue #002147, 16.05:1. Comfortable.
- Navy #000080, 16.01:1. Comfortable.
- Blue #0000ff, 8.59:1. Fine, and this surprises people.
- Dodger blue #1e90ff, 3.24:1. Fails body text.
- Sky blue #87ceeb, 1.74:1. Fails everything.
- Baby blue #89cff0, 1.71:1.
- Powder blue #b0e0e6, 1.43:1.
- Alice blue #f0f8ff, 1.07:1. Effectively invisible.
Because blue is the darkest hue, pure blue clears the accessibility bar on white while pure red at 4.0:1 does not. That is a direct consequence of the L* 32.3 figure and one of the few places where blue’s darkness is an advantage.
Tints, shades, and the word itself
Strictly, a shade is a color with black added and a tint is a color with white added, so most of the list above is tints of blue rather than shades of it. Sky blue is a tint. Navy is a shade. The usage collapsed long ago and arguing about it is a waste of a good afternoon, but the distinction is worth keeping when you are building a palette, because tinting and shading move a color along different paths and produce different problems. Monochromatic color schemes goes through the arithmetic of building a set from one hue, and color theory covers where the blue family sits relative to the rest of the wheel.
One last thing worth knowing about this family. YouGov polled ten countries across four continents in 2015 and blue came first in every single one, between 23 percent in Indonesia and 33 percent in Great Britain, ahead of the runner-up by eight to eighteen points. No other color led anywhere. A family that is the darkest in the gamut, holds the most perceptual variety per unit of color space, arrived late enough in language to get a rich vocabulary, and is universally the favorite, is going to end up with seventy names whether or not the eye can support them.
Using the list without fooling yourself
- Never specify a blue by name across a gap. The vocabulary is worth 18 buckets from memory. Send the hex code.
- Check lightness before hue. Two blues that feel wrong together are almost always wrong on L*, not on hue, because hue barely moves across the deep half of the family.
- Judge dark blues under the light they will live in. Navy against black is a 29.70 difference under a lamp and no difference at all in a dim hallway.
- Assume the name is wrong about lightness. Deep sky blue is one of the lighter entries at L* 72.5, medium blue is the second darkest CSS blue at L* 25.0, and dark turquoise at L* 75.3 is lighter than every color in the mid-blue section.
- Treat cyan and aqua as one color. Because they are.
Testing your own blue discrimination
Reading the numbers is not the same as owning the judgement. The place to find out where your personal threshold sits is a task that scores you rather than one that lets you self-assess. Every guess in the color matching game is scored with the same CIEDE2000 implementation behind every figure in this article, so a blue you thought you had matched comes back as a specific distance rather than a feeling. How long can you remember a color covers what happens to that accuracy as the delay grows, which is the mechanism behind the 67-to-18 collapse described above. If you would rather work on the vocabulary side, Name That Color runs the task in the opposite direction, showing a swatch and asking for the name.
How these numbers were produced
The 67 names are every blue and cyan keyword in CSS Color Module Level 4 plus the named shades that appear consistently across mainstream shade lists with an agreed hex value. All 2,211 unordered pairs were converted from sRGB to CIELAB against the D65 reference white and compared with CIEDE2000 following Sharma, Wu and Dalal’s reference implementation, which is the same code path this site uses to score gameplay. Clustering is greedy in list order, merging each name into the first existing cluster within the threshold; greedy clustering is order-dependent, so the bucket counts are approximate to within a name or two, though the shape of the collapse is not sensitive to it.
Distinguishable-color counts come from sampling the sRGB cube at every third value, discarding anything below 15 percent HSV saturation to keep near-grays out of the hue families, then greedily packing at ΔE00 2.3. These are lower bounds set by the sampling grid, as the step-2 rerun confirms. Hue-family boundaries are HSV hue bands, blue 195 to 255, green 75 to 165, red 345 to 15, yellow 45 to 75, which are conventional rather than perceptual and would shift the absolute counts if drawn differently. Contrast ratios follow the WCAG 2.2 relative luminance definition.