Color Memory Game
By John K.··14 min read

Different shades of green: 80 color names, hex codes, and the measurement showing why the middle of green is the flattest ground on the color wheel

Every different shade of green with its name and hex code, from the dark greens to the yellow-greens, measured pairwise with CIEDE2000. The 80 names span 127 degrees of hue, four times the width of the red vocabulary, and the reason is that 34 of those degrees are perceptually indistinguishable from each other.

Sage, emerald, viridian, phthalo, celadon, feldgrau. The shades-of-green lists run longer than the lists for any other family, and the usual assumption is that this is padding. I ran the same measurement I ran on the reds and the blues to find out, and green turned out to be the family where the numbers tell a completely different story from the other two.

I collected 80 named greens, every green, teal and cyan-adjacent keyword in the CSS named-color set plus the shades that recur across mainstream lists with a settled hex value, and ran all 3,160 pairs through CIEDE2000. Only four pairs sit below the threshold where two colors stop being distinguishable. That is the cleanest result of the three families: 99.9 percent of the pairs are genuinely different colors, against 99.3 percent for red and 99.8 percent for blue.

The finding worth the article is what happens when you plot where those names sit. The red names all fit inside a 58 degree wedge of the color wheel and separate almost entirely by lightness and chroma. The green names span 127 degrees, from olive at hue 60 to midnight green at 187, more than a third of the wheel and more than four times the red arc. On the surface that reads as a richer, more finely divided vocabulary.

It is the opposite. Green names spread that wide because they have to. The middle of the green wedge is the flattest terrain on the entire color wheel, and a vocabulary that stayed inside it would have almost nothing to name.

The 34 degree dead zone in the middle of green

Take a fully saturated color and rotate its hue by one degree. How much does it visibly change? Sweeping that measurement around the whole wheel gives the answer, and the curve is nowhere near flat.

  • At hue 119, one degree of rotation costs 0.030 ΔE00. That is the flattest single degree on the wheel, roughly one seventy-fifth of a just-noticeable difference.
  • At hue 185, out where green surrenders to cyan, the same one degree costs 1.945 ΔE00, sixty-five times more.
  • Averaged across hues 90 to 150, the heart of green, a degree of rotation is worth 0.243 ΔE00. Across hues 0 to 60, the red-to-yellow run, it is worth 1.041, more than four times as much.

Stated the way it actually matters: starting from pure green at hue 120, full saturation, full value, you can rotate the hue anywhere from 100 to 134 before the result becomes distinguishable from where you started. Thirty-four degrees of the color wheel, all of it one color to the eye. Run the same test on pure red and you get 14 degrees. On pure blue, 21 degrees.

That is the constraint the green vocabulary is working around. If you want a name that means something different from green, you cannot get it by nudging the hue, because thirty-four degrees of nudging produces nothing. You have to leave the neighbourhood entirely, out toward olive and chartreuse on one side or toward teal and viridian on the other, and that is exactly where the names went.

The names confirm it

Splitting the 80 names by whether they fall inside the 100 to 134 dead zone and measuring what separates them from each other:

  • Inside the dead zone, 20 names. Hue accounts for 7 percent of the average separation between them. Chroma does 57 percent and lightness 36 percent. These names are a brightness and saturation ladder, the exact structure the red family has.
  • Outside it, 60 names. Hue accounts for 29 percent, four times the share, and the ladder structure loosens accordingly.

So the green vocabulary is really two vocabularies wearing one label. In the middle it behaves like red naming, stacking dark green, green, forest green, lime green, pale green and honeydew up a lightness ladder at identical hue. Out at the edges it behaves like nothing else, using genuine hue distance to separate olive from chartreuse from emerald from teal.

The four indistinguishable pairs in the whole set make the same point from the other direction. Every one of them is crowded into the bright corner where the flattening is worst: harlequin against neon green at 0.51, lawn green against chartreuse at 0.65, lime against neon green at 0.89, lime against harlequin at 1.22. A hundred and twenty-seven degrees of naming effort produced no duplicates anywhere except in the one place where hue stops working.

The different shades of green with names and hex codes

Grouped by hue rather than by lightness, which is the reverse of how I grouped the reds, because hue is what does the separating here. L* is CIELAB lightness on a 0 to 100 scale and C* is chroma, the distance from gray, where higher means more saturated. If the six-digit notation needs unpacking, how to read hex color codes covers what the digits do.

Yellow-greens, hue 60 to 100

The half of the family that leans toward yellow, and the half most often mislabelled. Several of these are closer to yellow than to green.

  • Olive, #808000. Hue 60, L* 51.9, C* 58.1. A CSS keyword and the extreme edge of the family. At hue 60 it is exactly halfway between red and green in RGB terms, which is why it reads as neither.
  • Army green, #4b5320. Hue 69, L* 33.5.
  • Apple green, #8db600. Hue 74, L* 68.8, C* 77.7.
  • Moss green, #8a9a5b. Hue 75, L* 61.0.
  • Artichoke, #8f9779. Hue 76, C* 17.1. One of the grayest names on the list, and a good example of a name doing work no hue-based name could do.
  • Olive drab, #6b8e23. Hue 80, L* 54.7, C* 57.1. The CSS keyword, inherited from X11 rather than chosen.
  • Dark moss green, #4a5d23. Hue 80, L* 36.8. Measured 4.33 from army green, which is a real difference but a small one.
  • Yellow green, #9acd32. Hue 80, L* 76.5, C* 76.7.
  • Avocado, #568203. Hue 81, L* 49.4.
  • Dark olive green, #556b2f. Hue 82, L* 42.2.
  • Green yellow, #adff2f. Hue 84, L* 92.0, C* 97.2. The lightest high-chroma entry in the set, and one end of the widest pair in the whole measurement.
  • Rifle green, #444c38. Hue 84, C* 13.1. Almost a gray with a memory of green in it.
  • Olivine, #9ab973. Hue 87, L* 71.4.
  • Sage, #9caf88. Hue 89, L* 69.1, C* 22.5. The interiors-industry favourite, and structurally a desaturated yellow-green rather than a green.
  • Chartreuse, #7fff00. Hue 90, L* 89.9, C* 109.5. Named for the Carthusian liqueur, which has been made at Voiron to a recipe the order has never published.
  • Lawn green, #7cfc00. Hue 90, L* 88.9. Measured 0.65 from chartreuse. These are the same color.
  • Sap green, #507d2a. Hue 93, L* 47.7. Originally a fugitive pigment made from buckthorn berries that faded within years, which is why surviving medieval greens are rarely the green the painter mixed.
  • Asparagus, #87a96b. Hue 93, L* 65.4.
  • Basil, #579229. Hue 94, L* 54.8.
  • Laurel green, #a9ba9d. Hue 95, C* 17.0.
  • Pistachio, #93c572. Hue 96, L* 74.4.
  • Nyanza, #e9ffdb. Hue 97, L* 97.6. Barely a color.

True greens, hue 100 to 134

The dead zone. Twenty names occupying 34 degrees that the eye reads as a single hue, which is why almost all of the separation here is lightness and chroma rather than hue.

  • Clover, #3f7d20. Hue 100, L* 46.7. Measured 2.76 from sap green.
  • Tea green, #d0f0c0. Hue 100, L* 91.5.
  • Kelly green, #4cbb17. Hue 101, L* 67.4, C* 86.2.
  • Harlequin, #3fff00. Hue 105, L* 88.2, C* 117.0.
  • Juniper, #4a6741. Hue 106, L* 40.5.
  • Fern green, #4f7942. Hue 106, L* 46.6.
  • Neon green, #39ff14. Hue 111, C* 116.0. Measured 0.51 from harlequin and 0.89 from pure lime. Three names, one color.
  • Dark green, #006400. Hue 120, L* 36.2, C* 60.3.
  • Green, #008000. Hue 120, L* 46.2. The CSS keyword, and one of the sixteen original 1996 web colors that has never changed value.
  • Forest green, #228b22. Hue 120, L* 50.6.
  • Lime green, #32cd32. Hue 120, L* 72.6, C* 91.0.
  • Dark sea green, #8fbc8f. Hue 120, L* 72.1, C* 29.9. Identical lightness to lime green and a third of the chroma, which is the whole distinction.
  • Lime, #00ff00. Hue 120, L* 87.7, C* 119.8. The most saturated color in the set and the most saturated green sRGB can express.
  • Screamin’ green, #66ff66. Hue 120, L* 89.5.
  • Pale green, #98fb98. Hue 120, L* 90.8.
  • Light green, #90ee90. Hue 120, L* 86.5. Measured 2.75 from pale green, just over the threshold.
  • Russian green, #679267. Hue 120, L* 56.5.
  • Honeydew, #f0fff0. Hue 120, L* 98.6, C* 9.3. Eleven of these twenty names share the identical hue value of 120 and are told apart purely by how light and how vivid they are.
  • Celadon, #ace1af. Hue 123, L* 84.9. Named for the Chinese stoneware glaze, itself named after a character in a seventeenth-century French pastoral novel who wore green ribbons.
  • Hunter green, #355e3b. Hue 129, L* 36.1.

Blue-greens, hue 134 to 187

Thirty-eight names, the largest group, sitting where a degree of hue rotation finally buys something. This is where the vocabulary gets its precision back.

  • Xanadu, #738678. Hue 136, C* 11.1.
  • Cal Poly green, #1e4d2b. Hue 137, L* 28.8.
  • Emerald, #50c878. Hue 140, L* 72.5, C* 59.5. The gemstone gets its color from chromium or vanadium in beryl, and the hex value in circulation is a rendering of the stone rather than a measurement of one.
  • Malachite, #0bda51. Hue 140, L* 76.6, C* 89.4. Ground malachite was the standard green pigment from Egypt through the Renaissance, and it is a copper carbonate, chemically a cousin of verdigris further down this list.
  • Feldgrau, #4d5d53. Hue 143, C* 9.2. The second least saturated name in the set. Adopted by the German army in 1907 as a field color precisely because it disappears.
  • Parakeet green, #03c04a. Hue 143, L* 68.2.
  • Sea green, #2e8b57. Hue 146, L* 51.5.
  • Medium sea green, #3cb371. Hue 147, L* 65.3.
  • Spring green, #00ff7f. Hue 150, L* 88.5, C* 90.1.
  • Mint cream, #f5fffa. Hue 150, L* 99.2, C* 4.3. The least saturated and lightest entry in the whole set.
  • Seafoam green, #93e9be. Hue 150, L* 86.1.
  • Dark spring green, #177245. Hue 150, L* 42.2.
  • Phthalo green, #123524. Hue 151, L* 19.2. The darkest green in the set. The pigment is copper phthalocyanine, synthesised accidentally at Scottish Dyes in 1928 and now so cheap and so overpowering that painters keep it away from the rest of the palette.
  • British racing green, #004225. Hue 154, L* 23.8. Never an officially specified color, which is why the hex value varies more across sources than almost anything else on this list.
  • Dartmouth green, #00693e. Hue 155, L* 38.6.
  • Shamrock green, #009e60. Hue 156, L* 57.4.
  • Medium spring green, #00fa9a. Hue 157, L* 87.3.
  • Mint, #3eb489. Hue 158, L* 66.1.
  • Jade, #00a86b. Hue 158, L* 60.8. Measured 3.21 from shamrock green.
  • Medium aquamarine, #66cdaa. Hue 160, L* 75.7.
  • Aquamarine, #7fffd4. Hue 160, L* 92.0.
  • Viridian, #40826d. Hue 161, L* 49.7, C* 26.9. The brightest green on the list that still passes accessible contrast on white, which makes it more useful than its obscurity suggests.
  • Brunswick green, #1b4d3e. Hue 162, L* 29.1.
  • Hooker’s green, #49796b. Hue 163, C* 20.0. Named for the botanical illustrator William Hooker, who mixed it to paint leaves.
  • Jungle green, #29ab87. Hue 163, L* 62.8.
  • Illuminating emerald, #319177. Hue 164, L* 54.3.
  • Castleton green, #00563f. Hue 164, L* 31.8.
  • Bottle green, #006a4e. Hue 164, L* 39.3.
  • MSU green, #18453b. Hue 167, L* 26.1.
  • Wintergreen dream, #56887d. Hue 167, L* 53.0.
  • Persian green, #00a693. Hue 173, L* 61.1.
  • Pine green, #01796f. Hue 175, L* 45.4.
  • Myrtle green, #317873. Hue 176, L* 46.1. Measured 3.82 from pine green.
  • Light sea green, #20b2aa. Hue 177, L* 65.8.
  • Verdigris, #43b3ae. Hue 177, L* 66.9. Measured 2.31 from light sea green, which is right on the threshold. The name is the corrosion product of copper, and it is the reason old church roofs are this color.
  • Teal, #008080. Hue 180, L* 48.3. Named for the duck, which has a stripe of it on its head.
  • Dark cyan, #008b8b. Hue 180, L* 52.2. Measured 4.03 from teal, so the two CSS keywords are distinguishable but close.
  • Midnight green, #004953. Hue 187, L* 27.8. The far edge of the family, and by most reckonings a blue.

The dark shades of green, L* below 40

Cutting the same 80 names by lightness instead of hue pulls out the group people usually mean by dark green, and it is worth listing separately because it is the one part of the family where the hue grouping above actively misleads. These fifteen come from all three hue bands and read as one color family anyway, which is what happens once lightness drops far enough to swamp everything else.

  • Phthalo green, #123524. L* 19.2. The darkest green measured, and the only one in the set below L* 20.
  • British racing green, #004225. L* 23.8.
  • MSU green, #18453b. L* 26.1.
  • Midnight green, #004953. L* 27.8.
  • Cal Poly green, #1e4d2b. L* 28.8.
  • Brunswick green, #1b4d3e. L* 29.1. Measured 3.26 from MSU green, so the two are close to interchangeable.
  • Rifle green, #444c38. L* 31.1, C* 13.1. The dullest of the dark greens by a wide margin.
  • Castleton green, #00563f. L* 31.8.
  • Army green, #4b5320. L* 33.5.
  • Hunter green, #355e3b. L* 36.1.
  • Dark green, #006400. L* 36.2, C* 60.3. Twice the chroma of hunter green at the same lightness, which is the whole difference between them.
  • Dark moss green, #4a5d23. L* 36.8.
  • Feldgrau, #4d5d53. L* 37.9, C* 9.2.
  • Dartmouth green, #00693e. L* 38.6. Measured 3.20 from dark spring green.
  • Bottle green, #006a4e. L* 39.3.

Every one of these clears the 4.5:1 contrast threshold on white, which is the practical argument for the dark end of the family. If a green has to carry text, it is going to come from this list.

What the numbers say when you stop looking side by side

Side-by-side comparison is not the situation anyone uses a color name in. You use it from memory, later, without the reference in front of you. Rerunning the clustering at wider tolerances collapses the list the same way it collapsed for red and blue:

  • At ΔE00 2.3, the side-by-side threshold, 77 clusters from 80 names. Nearly everything survives.
  • At ΔE00 5, 56 clusters.
  • At ΔE00 10, 22 clusters.
  • At ΔE00 15, roughly memory tolerance for an untrained observer, 14 clusters. Three of them swallow 49 of the 80 names.
  • At ΔE00 20, 8 clusters.

The three big clusters at memory tolerance are worth naming, because they are what green actually reduces to in the head. One is the bright yellow-green group, 19 names from lime and chartreuse through kelly green and pistachio. One is the dark group, 16 names from dark green and hunter green through avocado and Cal Poly green. One is the mid blue-green group, 14 names from sea green and jade through viridian and Hooker’s green. Everything else is a handful of grays, a handful of near-whites, and the olive corner.

Which means that despite occupying four times the hue range of the red vocabulary, green collapses to almost the same number of memorable buckets. All that extra territory buys very little once the reference swatch is gone. That is the honest summary of the green naming problem, and how long you can remember a color covers why the tolerance widens as fast as it does.

Why green gets so much of the RGB cube and so little payoff

The flatness is not an accident of the naming. It comes from the geometry of sRGB itself, and the numbers are lopsided enough to be worth stating plainly. Sampling the sRGB cube and discarding near-grays:

  • Green occupies 24.8 percent of all saturated sRGB colors, against 16.4 percent for blue, 16.9 for magenta, 8.5 for red and 8.3 for yellow. It is by a wide margin the largest region.
  • Packing each region at the just-noticeable-difference threshold, green yields 1,951 distinguishable colors. Blue yields 1,953 from a region two thirds the size.
  • Per degree of hue band, green returns 21.7 distinguishable colors, blue 32.5, magenta 30.7, yellow 39.5 and red 49.3. Green is last by a distance.
  • Only 1.26 percent of sampled green points survive the packing, against 2.84 percent of red points. Green space is more than twice as redundant.

There is a practical version of this that anyone who has fought a color picker will recognise. Nudge one RGB channel by 8 out of 255 and measure how much the color visibly changes. In red territory the average change is 2.18 ΔE00, near the threshold of visibility. In green territory it is 1.53. The same slider movement does 30 percent less visible work in green, which is why matching a green by eye feels like the numbers are lying to you.

The underlying reason is that human vision spends most of its luminance budget on green. The sRGB luminance coefficients are 0.2126 for red, 0.7152 for green and 0.0722 for blue, which follow the photopic sensitivity curve peaking around 555 nanometres. Green carries nearly three quarters of perceived brightness, so a change in the green channel mostly reads as a change in brightness rather than a change in color. The hue barely moves because the eye is interpreting the signal as light level.

The same fact shows up in hardware. The Bayer filter on essentially every digital camera sensor allocates half its photosites to green and a quarter each to red and blue, on Bryce Bayer’s 1976 reasoning that green is where luminance detail lives. It is also part of why chroma-key screens are green: the green channel is the least noisy and most densely sampled channel a sensor has, so a green key separates more cleanly than a blue one.

Which greens are safe for text

Green is the family most likely to fail an accessibility check while looking fine, because the brightness that makes a green look rich is exactly what destroys its contrast on white. Of the 80 names, only 28 clear the 4.5:1 ratio that WCAG requires for body text on a white background.

  • Viridian #40826d, 4.53:1 on white. The lightest green in the set that passes, and the one to reach for when you want a green that still looks like a color rather than a near-black.
  • Green #008000, 5.14:1 on white, 4.09:1 on black. The CSS keyword passes on white and narrowly fails on black.
  • Teal #008080, 4.77:1 on white, 4.40:1 on black. Close to the only green that nearly works on both.
  • Forest green #228b22, 4.39:1 on white. Fails, by less than anyone would guess.
  • Sea green #2e8b57, 4.25:1 on white. Also fails.
  • Olive #808000, 4.20:1 on white. Also fails.
  • Dark green #006400, 7.44:1 on white, 2.82:1 on black.
  • Lime #00ff00, 1.37:1 on white, 15.30:1 on black. On a white background this is invisible.
  • Emerald #50c878, 2.13:1 on white, 9.87:1 on black.

The cluster of near-misses is the useful part. Forest green, sea green and olive all land between 4.2 and 4.4, which means the standard “tasteful” greens are the ones most likely to fail an audit by a margin too small to see. Viridian is the nearest passing replacement for all three, and the cost is not the same in each case. Sea green to viridian is a ΔE00 of 9.10, which reads as the same green gone slightly deeper. Forest green to viridian is 17.55 and olive to viridian is 25.79, both of which read as a genuine change of color rather than a tweak. If you have a green in running text, check which of the three you are on before assuming the swap will go unnoticed.

Which of the 80 names are worth keeping

My own opinion, after staring at these numbers longer than is reasonable: green is the family where the names earn their keep, but not evenly, and the useful ones are almost all at the edges.

  • Keep the edge anchors. Olive, chartreuse, green, emerald, teal. Five names spanning the full 127 degrees, each far enough from its neighbours to survive memory. Someone who knows only these can locate any green in the family by interpolation.
  • Keep the chroma outliers. Feldgrau at C* 9.2, xanadu at 11.1, rifle green at 13.1 and artichoke at 17.1 describe greens with the color taken out, which no hue-based name can do. Sage belongs here too, and is the reason it took over interior paint charts.
  • Keep the ones with a fact attached. Phthalo, malachite, verdigris and viridian carry pigment histories worth knowing, and a name that teaches you something survives being redundant.
  • Collapse harlequin, neon green and lime. All three sit within 1.22 of each other. One name is enough, and it should be lime, because that is the one the CSS spec will keep.
  • Collapse lawn green into chartreuse. ΔE00 0.65. It is chartreuse.
  • Stop treating hue words as precise inside the dead zone. Kelly green, fern green, forest green and hunter green all promise a different green and deliver a different lightness. Nothing wrong with them, but specify the number if it matters.
  • Accept that teal, midnight green and dark cyan are blues. They sit at the far end of the arc and behave like the blue family. The shades-of-green lists claim them because green has the wider wedge, not because they belong.

Testing your own green discrimination

Reading the numbers is not the same as owning the judgement, and green is the family where the gap is most punishing, because the flat middle trains bad habits. People learn that adjusting green hue does nothing, stop adjusting it, and then miss badly at the edges where it suddenly matters again. Every guess in the color matching game is scored with the same CIEDE2000 implementation behind every figure above, so a green you thought you had matched comes back as a number rather than a feeling. Watch what happens either side of hue 120 specifically: the scores go forgiving in the middle and unforgiving fast once you drift toward teal. Name That Color runs the task backwards, showing a swatch and asking which name it deserves, which is where the fourteen-cluster result stops being an abstraction, and Hue Sort makes the flat zone obvious the moment you try to order a row of greens that all look the same.

For the theory underneath, what is CIEDE2000 explains the formula, color theory covers where the families sit relative to each other, what color do two colors make handles mixing, and how many colors can you see goes further into the packing counts this article uses.

How these numbers were produced

The 80 names are every green, teal and cyan-adjacent keyword in CSS Color Module Level 4 plus the named shades that appear consistently across mainstream shade lists with an agreed hex value. Names that duplicate another entry’s hex exactly were excluded rather than double-counted. All 3,160 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.

Hue values quoted per name are HSV hue in degrees, which is what the naming intuition tracks. The 127 degree span is the arc containing all 80 names after removing the largest empty gap on the circle. In CIELAB hue angle, which is perceptual rather than device-referred, the same set spans 113.6 degrees, so the conclusion does not depend on which hue definition is used.

The one-degree hue-step figures are measured at full saturation and full value, stepping HSV hue by one degree and taking CIEDE2000 between neighbours. Flatness varies with saturation and lightness, so the 34 degree dead-zone width is specific to the fully saturated case and narrows for duller greens. The component attribution splits use plain CIELAB deltas rather than the weighted CIEDE2000 terms, so they attribute raw difference rather than perceptual weighting. Using the weighted terms shifts the magnitudes but not the ordering.

Clustering is greedy in list order, merging each name into the first existing cluster within the threshold, 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, then greedily packing at ΔE00 2.3. These are lower bounds set by the sampling grid. Hue-family boundaries are HSV bands, red 345 to 15, yellow 45 to 75, green 75 to 165, blue 195 to 255, magenta 285 to 345. These bands are conventional rather than perceptual and are unequal in width, which is why the per-degree normalisation is quoted alongside the raw counts. Contrast ratios follow the WCAG 2.2 relative luminance definition.