Homer Never Said Blue: How Language Carves Up Color

In 1858, the future British prime minister William Gladstone counted every color word in the Iliad and the Odyssey. Black showed up about 170 times and white more than 100. Red appeared 13 times, yellow 10, and green once. Blue did not appear at all.

Homer's sea is famously "wine-looking," not blue. Gladstone drew a dramatic conclusion from the gap. Then a century of linguists, psychologists and field researchers spent decades arguing over what the gap really meant.

The argument still matters if you work with color. Every time you name a swatch in a design system, you are drawing a line through a continuous spectrum. This article follows that line from Homer to a lab in Boston to a Namibian village, and ends with a traffic light in Tokyo.

The Count That Started It

Gladstone's tallies came from the third volume of his Studies on Homer and the Homeric Age. Significance magazine lists them: about 170 for black, over 100 for white, 23 for violet, 13 for red, 10 for yellow, 2 for ochre and 1 each for rose and green. There was no orange and no blue.

The words that did appear behaved oddly. Homer uses oinops, "wine-looking," for the sea and for oxen. He uses ioeis, "violet," for the sea, for wool and for iron. Chloros, which later meant green, describes honey, faces and wood. Gladstone read this as a system built mostly on dark and light, not on hue. The Spectator's account notes that "wine-dark" itself is a later translator's phrase, from Andrew Lang.

Gladstone concluded that "the organs of colour and its impressions were but partially developed among the Greeks of the heroic age." Put plainly, he thought the ancients saw the world with a defective eye.

German philologist Lazarus Geiger pushed the idea further. In 1867 he observed that the Vedic hymns, like Homer, never mention a blue sky, and neither does the Hebrew of the Bible. A theory took shape: color words arrive in a set order, and blue comes last.

Egypt is an awkward fact for that story. Its workshops were making a synthetic blue pigment thousands of years ago. They could clearly see and produce the color. Having the pigment is not the same as having a basic word that sorts it from everything else.

The eye theory did not survive. The words probably did not mean what Gladstone assumed, and the vocabulary question turned out to be far more interesting than a vision defect.

The Survey That Found a Pattern

In 1969, Brent Berlin and Paul Kay at Berkeley tested the order idea properly. They interviewed speakers of 20 languages. Each speaker looked at a board of 320 Munsell color chips plus nine grays, and marked which chips each of their basic color words covered, and which chip was the best example.

Their claim had two parts, as Kay and Richard Cook later summarized it. The referents of basic color terms in all languages appear to be drawn from a set of eleven universal perceptual categories. And those categories enter a language in a partially fixed order.

Twenty languages is a small sample, so in 1976 Kay's team began the World Color Survey. Fieldworkers working with SIL International collected data on 110 unwritten languages from 45 families. Each participant named 330 color chips one at a time. The average was 24 speakers per language.

The result confirmed the broad pattern and loosened the details. Languages differ in how many basic terms they have, but they do not carve up the spectrum at random. The exact 1969 sequence needed revising, and Kay and Cook describe a revised version from 2009.

Something about human vision pulls color words toward certain regions. The question left open is how far the words then pull back.

The Russian Blue Test

English has one basic word for blue. Russian has two: siniy for darker blues and goluboy for lighter ones. Russian makes the choice obligatory.

In 2007, Jonathan Winawer and colleagues turned that into an experiment, published in PNAS. They built 20 blue squares that spanned the siniy and goluboy range. On each trial, a person saw a target square and two choices, and had to say which choice matched the target as fast as possible.

Some trials paired two squares from the same Russian category. Others paired squares that sat on opposite sides of the siniy/goluboy line. For English speakers, all of them were simply blue.

Russian speakers were faster when the two colors fell into different categories. With no interference, they averaged 1,164 milliseconds on cross-category trials against 1,288 on within-category trials. That is a gap of about an eighth of a second. English speakers shown the same squares showed no such advantage.

The experiment had a further twist. When Russian speakers silently rehearsed a string of digits while doing the task, the advantage vanished. When they held a spatial pattern in mind instead, it stayed. The authors concluded that language was acting online, during the seeing, and that tying up the verbal system took it away.

Here is the usual wrinkle. This was not a case of two groups seeing different worlds. The lab measured a speed difference of a fraction of a second on hard comparisons. And a 2020 attempt to repeat it, by Jasna Martinovic, Galina Paramei and Joseph MacInnes in Cognition, found a speed advantage at the light blue to green boundary, but not at the dark blue to light blue one. They also found that the siniy/goluboy boundary moved depending on which shades the participants had seen most often.

So the Russian result is real in some form and fragile in another. That is a fair summary of the whole field.

Here is a ramp you can try on yourself. Where would you draw the line?

An English speaker might say "light blue" for the first two and "blue" for the rest, or might not feel a break at all. A Russian speaker has a word waiting at the point where the ramp feels as though it changes.

What a Himba Herder Sees in a Green Square

The Russian test compares two languages that are both spoken in labs and cities. The harder question is what happens in cultures with very different color vocabularies.

Debi Roberson, Ian Davies and Jules Davidoff first studied the Berinmo of Papua New Guinea, whose language has five basic color terms. Their 2000 paper in the Journal of Experimental Psychology: General found that recognition of colors tracked the Berinmo word boundaries, and that the best example of a category was not remembered better than other colors once response bias was controlled.

A skeptic could say that a tiny remote community proves little. So Roberson, Davidoff, Davies and Laura Shapiro went to southern Africa and tested the Himba, a semi-nomadic cattle-herding people. Their paper reports that Himba also has five basic color terms, although one of them, burou, is a recently borrowed word from Herero.

The Himba divide the green and blue territory in a way English does not. Roberson's team picked sets of Munsell chips with the same lightness and saturation. One set crossed the Himba line between dumbu and burou. Another crossed the English line between green and blue.

In a recognition task, Himba adults did better on pairs that crossed the dumbu/burou boundary. They showed no such advantage across the English blue/green boundary.

The two sets contained chips of the same physical difference. What changed was whether a word sat between them.

The Himba and the Berinmo each have five basic terms. Yet the paper concludes that two systems that look alike on the surface still lead their speakers to encode, remember and discriminate colors differently. Counting terms does not predict behavior. Where the lines fall does.

A Blue Light That Is Green

Japan gives you the same effect in daily life. The word ao traditionally covered what English speakers split into blue and green. Modern Japanese has a separate word for green, midori, but ao still stretches over some green things.

Japanese traffic signals have been called ao, blue, in official documents for nearly a century. According to 99% Invisible, a 1973 cabinet order resolved the clash with a compromise. Lights must use the bluest shade of green possible. They are still technically green, but blue enough to keep the legal name.

Here is the same ambiguity as a ramp.

The government bent the physical signal toward the name, not the name toward the signal. We cover the wider story in our piece on the history of signal colors.

So Did Homer See Blue?

Almost certainly. Nothing in the evidence suggests that Greek eyes differed from yours. The modern view, which the Spectator links to Guy Deutscher's Through the Language Glass, is that color vocabulary reflects language and culture, not biology.

What the experiments add is a narrower claim. Words do not change the light that reaches your retina, and they do not make you blind to a distinction. They tilt a hard decision. When two shades are close, a name can tip the choice, speed up the answer or sharpen a memory. Remove the name, by giving someone a verbal task, and the tilt goes away.

That is a modest effect, and it fits the data. It is also enough to matter to you.

What This Means for Your Palette

Three practical lessons follow.

• Names can split what looks like one color, and join what looks like two. If your product serves Russian, Japanese or Himba-speaking users, the line between your "blue" and your "green" may fall somewhere else for them. Check your semantic colors with the color semantics across cultures in mind, and run your status colors through the color blindness simulator too.
• Don't rely on a name to separate close shades. If two tokens differ by a hair, a team that calls them the same thing will treat them as the same. Name them for their job, not their hue, as we argue in the design token reference.
• Test with real speakers. A boundary that feels natural in your language may feel arbitrary in another. The named colors reference shows how many labels English already stacks onto the same region.

The best summary comes from the experiments themselves. Colors are a spectrum. Languages are cuts across it. And the cut you grew up with decides which two neighbors feel like strangers.

Sources

• Gladstone's Homeric color counts, Significance magazine
• Fathoming the wine-dark sea, The Spectator
• World Color Survey, Kay and Cook, Encyclopedia of Color Science and Technology
• Russian blues reveal effects of language on color discrimination, Winawer et al., PNAS 2007
• Russian blues reveal the limits of language influencing colour discrimination, Martinovic, Paramei and MacInnes, Cognition 2020
• Color categories: Confirmation of the relativity hypothesis, Roberson, Davidoff, Davies and Shapiro
• Red for stop, grue for go, 99% Invisible