← Blog

Right Brain · Reading

Does Language Change the Colours You See?

This is the question where popular science is most likely to lie to you, in both directions. One version says speakers of a language without a word for blue cannot see blue. The other says language is just labels and perception is untouched underneath. Neither survives the data. What survives is smaller, weirder, and much more interesting: a modest speed advantage, that disappears when you occupy the language system, and that shows up in only half of your visual field. Take in the whole. Before the parts.

By ComunAI · 7 September 2026 · ~6 min read

The starting position: naming is not arbitrary

In 1969 Brent Berlin and Paul Kay surveyed basic colour terms across many languages and reported something nobody expected from a purely cultural account: the vocabularies were not scattered at random across colour space. Languages with few terms tended to carve it up in similar places, and the terms appeared to be added in a broadly consistent order.¹ Moderate

Their methods have been criticised for decades and the strict universal sequence has not held up. But the core observation did. Terry Regier, Paul Kay and Naveen Khetarpal later showed computationally that the attested colour vocabularies sit close to optimal partitions of perceptual colour space — the divisions a system would choose to maximise how well a name predicts a shade.² Moderate

Edward Gibson and colleagues added the missing half: across many languages, colours are named more precisely when they are more useful to talk about, with warm colours — the colours of objects — carrying more efficient names than cool background colours.³ Moderate

The result that put a number on it

Russian has no single basic term covering the English range of blue. It has two: goluboy for lighter blues and siniy for darker ones, and they are as basic to a Russian speaker as blue and green are to an English one.

Jonathan Winawer, Nathan Witthoft, Lera Boroditsky and colleagues used that as a natural experiment. Participants saw three blue squares and had to say which of the lower two matched the top one. Russian speakers were faster when the two options fell on opposite sides of the goluboy/siniy boundary than when both fell inside one category. English speakers, whose language draws no line there, showed no such difference.⁴ Moderate

Two details do most of the work in interpreting this, and both are usually dropped.

• The effect was a matter of milliseconds, not of visibility. Everyone got the answer right. Russian speakers got there slightly sooner across a category line. Nobody was unable to see anything.
• It disappeared under a verbal load. When participants had to hold a string of digits in mind while judging the colours, the advantage went away — while a comparable spatial load left it intact. The language system was participating in the perceptual judgement in real time.

That last point is what makes the finding worth reporting at all. It is not evidence that Russian speakers have different eyes. It is evidence that a colour judgement recruits linguistic categories on the fly, and that you can knock them out by giving language something else to do.

Then it got strange: the left half of the world

Aubrey Gilbert, Terry Regier, Paul Kay and Richard Ivry ran a search task in which a target square differed from surrounding squares either across a colour-name boundary or within a name. The category advantage appeared when the target was in the right visual field — the half processed first by the left hemisphere, where language is typically lateralised — and was absent or much weaker in the left.⁵ Moderate A verbal interference task removed it, as in the Russian study.

Guillaume Thierry and colleagues then found an electrophysiological version. Greek has separate basic terms for light and dark blue; Greek and English speakers watching colour changes they were not asked to attend to showed differing early brain responses, before any deliberate judgement was possible.⁶ Moderate

Regier and Kay's summary of the field is the line worth keeping: Whorf was half right.⁷ Language appears to influence colour processing, and it does so in a way that is lateralised, task-dependent, and small — which is exactly what neither the strong nor the sceptical version of the story predicted.

Where it broke, and it did break

Two corrections, because a post that stopped here would be doing the thing this blog exists to avoid.

Debi Roberson, Ian Davies and Jules Davidoff worked with speakers of Berinmo, a language of Papua New Guinea whose colour vocabulary divides the space differently from English, and found memory and categorisation performance tracking the local categories rather than any universal set.⁸ Moderate That result cuts against the strong universalist reading of Berlin and Kay, and it came from fieldwork rather than a lab — which is both its strength and, in sample-size terms, its limitation.

And the lateralised effect itself has had an uneven record. Christoph Witzel's review of the area argues that several widely repeated claims about colour categories — including the robustness of categorical perception effects — are weaker or more method-dependent than their reputation suggests.⁹ Contested

Our reading: the direction of the Whorfian effect has replicated often enough to take seriously, and its size, reliability and boundary conditions are genuinely unsettled. Anyone telling you this is a closed question — in either direction — is ahead of the evidence.

Where the evidence stands

• Colour vocabularies are non-arbitrary and partly predictable from perception and use. Moderate^1,2,3
• Having two names for a range speeds discrimination across the boundary. Moderate — milliseconds, not visibility.⁴
• The advantage depends on the language system being free. Moderate — verbal interference removes it.^4,5
• The effect is stronger in the right visual field. Moderate — striking, and not uniformly replicated.⁵
• Language-specific differences appear in early, unattended brain responses. Early — one paradigm, small samples.⁶
• Categorisation follows the local vocabulary, not a universal set. Moderate — field evidence against strong universalism.⁸
• The size and robustness of categorical colour effects. Contested — a live methodological dispute.⁹
• Speakers of a language without a term for a colour cannot see it. No evidence — nothing here supports it; discrimination is intact in every study above.
• Learning more colour words improves your everyday vision. No evidence — never tested, and we are not going to imply it.

What this means for Right Brain

Right Brain is named for a metaphor, not a hemisphere — the open, whole-first way of looking, as against the narrow, part-by-part one. This literature is an awkward and useful guest, because it is the one place where the part-by-part machinery reaches down and touches the whole.

A name is a part. It is a discrete label imposed on a continuous surface. What the Whorfian results suggest is that once you have that label, it does not sit politely on top of the experience — it gets into the judgement, early, and it speeds you up in the direction the label points. That is a good deal when the category is the thing you need. It is a cost when what is in front of you does not fit the categories you happen to own.

This is the same shape as why optical illusions fool everyone and why what you hear changes what you see: perception arrives already committed, and the commitments were made using whatever the system had to hand — priors, other senses, and here, apparently, your vocabulary. Pair it with why people saw The Dress differently, which is the same subject one level down, at the point where the illumination assumption is made rather than the name.

The perception games in the app are that idea shrunk to a few quiet minutes: no score, no verdict about your abilities, just your own inference machinery caught briefly in the act. The standing limit applies as always — there is no good evidence that playing perception games improves your everyday seeing, and none at all that anything here changes your colour vision. Broad transfer is the weak link in this whole field, and we covered it properly in Do Brain-Training Games Actually Work? Contested