Three blue squares appear on a screen, one above the other two. The task is to choose the lower square that matches the upper one. You can look at all three at once. In an influential experiment on linguistic relativity, this small decision became a way to investigate a much larger question: whether the language we speak changes how we think.
The attraction of the idea is easy to understand. Learning another language gives you new ways to describe familiar things. It can be tempting to assume that a new description must come with a new experience. But a useful word, a faster answer and a change in perception are different findings. Following the experiment into its later reevaluation makes those differences harder to overlook.
Linguistic relativity vs determinism: what is the claim?
Linguistic relativity concerns the possibility that differences between languages influence cognition. A stronger claim, linguistic determinism, holds that language sets limits on what someone can think. The distinction matters because evidence of an influence does not establish a limit. An entry in the Stanford Encyclopedia of Philosophy treats these as separate positions.
The familiar label Sapir-Whorf hypothesis often brings these questions together. For a reader assessing an experiment, however, the broad label does less work than a precise question. Did language affect a person’s choice of name? What they attended to? Their speed on a task? Or the early processing of the visual signal itself?
A participant has to process the display, compare the squares and choose a response. Timing the final button press does not, by itself, isolate which part of that sequence language influenced.
The blue squares that became a linguistic relativity example
Russian has separate basic color terms for lighter blue, goluboy, and darker blue, siniy. In contrast, English speakers can describe the same distinction, but usually do so within the basic category blue.
In the 2007 experiment, Russian speakers responded faster when a difficult color comparison crossed their naming boundary than when it stayed within one category. English speakers did not show the same boundary advantage. Asking participants to hold a string of digits in mind removed the Russian advantage; a spatial memory task did not.
This was a specific pattern, not generally superior color vision. In fact, English participants were faster overall. The authors also allowed more than one explanation: language might feed back into visual processing, or help at a later decision stage.
That leaves an interesting result without turning it into a claim that English speakers cannot see something Russian speakers can. The experiment measured how quickly people matched colors under particular conditions. It did not test the outer limits of either group’s imaginable world.
What changed when researchers revisited the experiment?
A 2020 reevaluation did not find the same response-speed advantage across the Russian light-blue/dark-blue boundary. Naming still showed a sharper transition between the Russian categories, but where people placed that boundary shifted with the colors’ frequency and presentation context.
For linguistic relativity, the distinction between naming and discrimination was the point. A crisp division in the words people choose need not produce a matching division in the speed of their visual decisions.
There was a complication: this was not an exact replication. For example, the researchers could reproduce only 15 of the original 20 colors, and other features of the displays and procedure differed. Their proposed explanations for the earlier result therefore remain explanations to evaluate, rather than a demonstrated account of what caused it.
A 2026 correction matters to that comparison. In the 2020 paper, the brightest stimulus from the original experiment was reported as 55.94 cd/m². The correct value was 84 cd/m²; 55.94 belonged to the later experiment itself. The notice corrects that reported brightness value. However, it does not announce a new analysis or a reversal of the study’s findings.
The disagreement cannot be resolved simply by choosing the newer publication. To understand what changed, we have to look at what each experiment asked people to do and what it put in front of them.
Does switching languages change the result?
Comparing language groups leaves language intertwined with other experiences. Bilingual research offers another way to test linguistic relativity: examine how the same people perform when a different language is active.
In a 2024 study, 31 Lithuanian-Norwegian bilinguals completed color tasks alongside a digit-memory task. A blue-category advantage appeared when Lithuanian was active, but not when Norwegian was active. Lithuanian, like Russian, has distinct basic terms for light and dark blue.
The result is consistent with language context influencing task performance. It also comes with an important design limit: participants did the Lithuanian condition before the Norwegian condition. Because the order was fixed, the language change cannot be cleanly separated from order-related effects.
It would be an overstatement to describe the participants as entering a different visual reality each time they changed language. The experiment asks a more specific and useful question about the resources available during a particular task.
What linguistic relativity evidence can establish about perception
If the question concerns early vision, researchers need evidence closer to that stage of processing. A 2026 PNAS paper used three EEG experiments to investigate color responses. It revisited an earlier line of work comparing Greek and English, using Russian speakers and additional color-contrast controls.
The investigators did not find a reliable extra early response attributable to the relevant color category under their tested conditions. Physical contrast and adaptation helped explain the brain-response differences. As a result, this challenges a particular interpretation of early visual signals, not every possible effect of language.
However, these recordings answer a different question from the 2007 button-press experiment. A language-related difference in response time and an absent category effect in an early electrical signal are not interchangeable outcomes. The useful question is where a proposed influence operates, and how convincingly an experiment isolates it.
For readers encountering linguistic relativity examples, that is a practical way to assess the claim. Begin with the measured outcome. Then check whether the explanation has quietly expanded from faster responding to changed perception, or from changed perception to a limit on thought.
A language can matter without containing the whole mind
Evidence beyond color makes the strongest language-as-a-boundary claim difficult to sustain. In a 2024 Nature Perspective, Evelina Fedorenko, Steven Piantadosi and Edward Gibson review cases in which people with severe language impairment retain substantial nonverbal reasoning abilities, including mathematical reasoning. They argue that language primarily serves communication rather than being a prerequisite for complex thought.
That does not mean language impairment leaves every ability unaffected. Nor does their argument make language unimportant. The distinction is between needing language for every act of thinking and using it to communicate knowledge that other people can learn from.
Taken together, these linguistic relativity experiments show why the task belongs in the explanation. A familiar distinction may help someone choose a response under one set of conditions without producing the same advantage when the display, interference or measurement changes.
A new word can make a distinction easier to discuss without proving that the underlying experience was previously unavailable. Sapien’s exploration of Sonder offers a related question about naming an experience and what we can know about it.
Calling a shade light blue does more than report a private impression: it gives another person a distinction they can question, refine or use. That shared activity matters even when the stronger claim about altered perception remains unsettled. The reach of a language includes what people can communicate with it, not just what a laboratory task can detect.
References
- Sapir E. The Status of Linguistics as a Science. Language. 1929;5(4):207-214. Opening quotation, p. 209.
- Margolis E, Laurence S. Concepts, section 4.3: Linguistic determinism and linguistic relativity. Stanford Encyclopedia of Philosophy. Revised 2019.
- Winawer J, Witthoft N, Frank MC, Wu L, Wade AR, Boroditsky L. Russian blues reveal effects of language on color discrimination. PNAS. 2007;104(19):7780-7785.
- Martinovic J, Paramei GV, MacInnes WJ. Russian blues reveal the limits of language influencing colour discrimination. Cognition. 2020;201:104281. Corrigendum: 2026;275:106566.
- Sinkevičiūtė A, Mayor J, Vulchanova MD, Kartushina N. Active Language Modulates Color Perception in Bilinguals. Language Learning. 2024;74(S1):40-71.
- Martinovic J and colleagues. Cognition does not automatically influence perception: Evidence from neural encoding of colors belonging to different categories. PNAS. 2026;123(24):e2538139123.
- Fedorenko E, Piantadosi ST, Gibson EAF. Language is primarily a tool for communication rather than thought. Nature. 2024;630:575-586.
