Short answer
Not in the sense of what your eyes can resolve: when the smallest detectable colour differences were measured, English and Korean speakers did not differ, and neither group was more sensitive at its own category boundaries. What language measurably affects is the speed and pattern of judgements made about colours that are easy to tell apart. In the best-known experiment Russian speakers were about a tenth of a second faster to match blues that fell on opposite sides of their goluboy–siniy boundary, an advantage that vanished when they had to do a verbal task at the same time, and a later re-run of the experiment did not find it.
The experiments do not ask people what they see. They time them, or ask which of three colours is the odd one out, or test what they remember. In 1984 Kay and Kempton devised a colour experiment that detected an effect of language on judgement, proposed a mechanism for it, and ran a second experiment designed to block that mechanism, in which the effect disappeared. In 2007 Winawer and colleagues showed Russian and English speakers three blue squares and asked which of the lower two matched the top one. For close shades, Russian speakers answered in 1,164 milliseconds when the pair crossed their own goluboy–siniy boundary and 1,288 when it did not, a 124-millisecond advantage. English speakers, tested on the same blues, showed no such difference under any condition.
About 21 speakers per group remained after exclusions. Each person's boundary was found first by having them sort twenty blues; the Russian and English boundaries fell at almost the same step.
If words reshaped perception itself, people should detect smaller differences where their language draws a line. Roberson, Hanley and Pak measured this directly. They found the just-noticeable difference for English speakers was no smaller at the blue–green boundary than within either category, even though the same speakers showed the usual boundary advantage when the colours were further apart. They found no difference in thresholds between English and Korean speakers, whose languages divide the green region differently. Their conclusion was that the effect is categorical and not perceptual: it arises when a colour is classified, not when it is registered. That is the single most useful result for anyone wondering whether an unfamiliar vocabulary means an unfamiliar visual world.
The Russian advantage was fragile in an informative way. When participants had to carry out a verbal task at the same time, the advantage disappeared and even reversed slightly; with a spatial task in its place it stayed, at 108 milliseconds. That points to language being used during the task, not to a permanent change in how blue looks. It was also larger for hard discriminations than easy ones. The Korean study found speakers faster across the boundary between yeondu and chorok, which English does not mark, and the effect depended on response speed: slower responders showed it in both halves of the visual field, faster ones only in the right. The authors took that as evidence that the left hemisphere's language system mediates the effect wherever it appears.
Two of the headline findings have not held up cleanly. In 2020 Martinovic, Paramei and MacInnes ran the Russian tasks again and found no speed advantage at the siniy–goluboy boundary, though they found one between goluboy and green; the blue boundary also shifted according to which shades had been shown more often. And the claim that the effect is confined to the right visual field, reported in 2006 and central to the argument that 'Whorf was half right', was tested in ten versions of the original experiments with 230 observers and controlled colour rendering. Search was faster across categories, but equally in both visual fields. Neither result abolishes the category effect. Both show that its size and conditions are less settled than the first reports suggested.
One study measured brain potentials instead of responses. Greek and English speakers did a colour oddball task; a component of the signal that reflects automatic change detection was larger for blue changes than green ones in Greek speakers, whose language separates ghalazio from ble, and similar for both in English speakers. That is evidence of a difference early in processing, from one study. A study of Greek–English bilinguals shows how movable such differences are: how sharply they separated the two blues depended on how readily the Greek words came to them and on how long they had lived in an English-speaking country. Whatever a vocabulary does to colour judgement, it is acquired with use and fades with disuse.
Three things have to be kept apart. The measurements are reaction times, similarity choices, memory scores and brain potentials, each from tens of participants. The interpretations belong to the researchers: Roberson's group reads the pattern as showing a tight relationship between language and cognition, with linguistic categories shaping judgement; Regier and Kay argue that language affects colour judgement in some conditions while naming itself follows shared tendencies. Colourwise's reading is narrower still. The effects are real, small, task-dependent and at category boundaries. They do not show that speakers of one language see a colour that speakers of another cannot, and no study cited here claims that.
| Study | Speakers | What was measured | Result | Limit |
|---|---|---|---|---|
| Kay and Kempton (1984) | Not stated in the abstract read | Colour judgements in a test designed to be sensitive to naming | An effect of language was detected; it disappeared when the proposed mechanism was blocked | Details beyond the abstract were not checked |
| Gilbert and colleagues (2006) | Not stated in the abstract read | Visual search time by visual field | Faster across categories in the right visual field only; removed by a verbal task | Not reproduced in the right field only by the 2011 study |
| Winawer and colleagues (2007) | Russian and English | Time to match blues across or within the goluboy–siniy boundary | 124 ms faster across the boundary for Russian speakers on close shades; none for English | About 21 per group; not found in the 2020 re-run |
| Roberson, Pak and Hanley (2008) | Korean and English | Visual search across a Korean-only boundary | Korean speakers faster across it; English speakers not | Pattern across visual fields depended on response speed |
| Roberson, Hanley and Pak (2009) | Korean and English | Smallest detectable colour difference | No lowering at boundaries; no difference between groups | Two language groups |
| Thierry and colleagues (2009) | Greek and English | Brain potential indexing automatic change detection | Larger for blue than green changes in Greek speakers; similar in English speakers | One study; a neural measure, not a report of appearance |
| Witzel and Gegenfurtner (2011) | 230 observers | Ten versions of the lateralised search experiments | Category effect present in both visual fields equally | Tests lateralisation, not cross-language differences |
| Martinovic, Paramei and MacInnes (2020) | Russian and English | The 2007 matching and sorting tasks | No speed advantage at siniy–goluboy; boundary moved with context | Corrigendum published in 2026, content not checked |
Why: A reaction-time advantage of about a tenth of a second on one task was retold as a difference in what is seen.
Fix: Report the measure: faster matching across a named boundary, in one study, removed by a verbal task and absent in a later re-run.
Why: Studies of naming and of similarity judgements were retold as tests of vision. The published work on Himba and Berinmo speakers measures where category effects fall, and the threshold study found no difference between language groups.
Fix: Go to the paper and check what task was run. If the measure is naming, sorting or reaction time, it does not support a claim about what can be seen.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
FactModerate evidence
In a speeded matching task, Russian speakers discriminated close blues in 1,164 ms when the pair crossed their goluboy–siniy boundary and 1,288 ms when it did not, a 124 ms advantage; English speakers showed no category advantage.
Measured: About 21 Russian and 21 English speakers after exclusions. Method: Three-square matching task on twenty blues; each participant's boundary set by their own sorting.
Caveat: One study; a 2020 re-run of the tasks did not find the advantage at this boundary.
Source: Russian blues reveal effects of language on color discrimination (PNAS 104: 7780–7785, 2007)
FactModerate evidence
The Russian speakers' advantage was eliminated by a concurrent verbal task (a 65 ms reversal) and not by a concurrent spatial task (a 108 ms advantage remained).
Source: Russian blues reveal effects of language on color discrimination (PNAS 104: 7780–7785, 2007)
FactStrong evidence
Discrimination thresholds were no lower at the blue–green category boundary than within categories for English speakers, and did not differ between English and Korean speakers, although both groups showed category effects with larger colour differences.
Source: Thresholds for color discrimination in English and Korean speakers (Cognition 112: 482–487, 2009)
FactModerate evidence
A 2020 re-run of the Russian tasks found a boundary speed advantage between goluboy and green but none between siniy and goluboy, and found that the siniy–goluboy boundary shifted with the frequency of lighter or darker shades.
Caveat: A corrigendum to this paper was published in 2026; Colourwise confirmed it exists but could not read what it changes.
FactStrong evidence
Ten versions of the original lateralisation experiments, with 230 observers, found faster visual search across colour categories in both visual fields equally, not in the right field only as reported in 2006.
Source: Is there a lateralized category effect for color? (Journal of Vision 11(12): 16, 2011); Whorf hypothesis is supported in the right visual field but not the left (PNAS 103: 489–494, 2006)
FactModerate evidence
Regier and Kay argued in 2009 that the Whorf hypothesis is 'half right' for colour: that language influences colour perception mainly in the right visual field, and that naming is shaped by both universal and language-specific forces.
Caveat: This is the authors' interpretation in a review. The lateralisation half rests on findings that a larger study did not reproduce.
Source: Language, thought, and color: Whorf was half right (Trends in Cognitive Sciences 13: 439–446, 2009)
Colourwise interpretationModerate evidence
The published evidence shows that colour words affect the speed and pattern of judgements at category boundaries; it does not show that speakers of different languages differ in which colours they are able to see.
Based on: Reaction-time and similarity effects in the cited studies appear at boundaries, depend on verbal resources and task, and coexist with identical discrimination thresholds across two language groups.
Caveat: Thresholds were compared for English and Korean speakers only; the conclusion is extended to other language pairs by argument, not by measurement.
Reviewed 1 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.