What's Wrong with the Linguistic Relativity Hypothesis

Few ideas in cognitive science have had as eventful a life as the hypothesis of linguistic relativity — the claim that the language we speak influences, or perhaps even determines, the way we perceive and think about the world. Originally associated with the linguist Benjamin Lee Whorf and his mentor Edward Sapir, the hypothesis fell out of favour in the cognitive revolution of the 1960s, only to experience a remarkable comeback in popularised form over the past two decades.

Today the «Sapir–Whorf hypothesis» is a staple of popular science articles, TED talks, and social media content. The claim seems almost intuitively appealing: of course the words we have for something affect how we think about it. But as so often in science, the intuitive appeal and the evidence are two different things.

What the Hypothesis Actually Claims

First, a clarification of terms. The hypothesis comes in two versions, which are rarely distinguished in popular accounts:

• Strong version (linguistic determinism): Language determines thought. You cannot think about something for which your language has no word. Some concepts are literally unthinkable in certain languages.
• Weak version (linguistic relativity): Language influences thought. Speakers of different languages may tend to think about certain domains differently, but can still grasp the same concepts.

The strong version is almost universally rejected by linguists and cognitive scientists. It is easily falsified: people routinely think about, perceive, and communicate concepts for which they initially lacked words. Language learning itself would be impossible if you couldn't understand new concepts. Translation would be impossible. Neither is true.

The weak version is more defensible and is what modern researchers actually study. The question becomes: does having a word for X make you faster, more accurate, or more likely to notice X? That is an empirical question, and the empirical answers have been much more complicated than the popular accounts suggest.

The Colour Studies: The Most Cited Evidence

The most famous modern evidence for linguistic relativity comes from colour perception studies. The key finding: languages differ in how they carve up the colour spectrum. Russian has obligatory separate words for light blue (голубой, goluboy) and dark blue (синий, siniy) where English uses a single word «blue». Do Russian speakers therefore perceive blues differently?

Studies by Lera Boroditsky and colleagues in the 2000s suggested yes: Russian speakers were faster at discriminating shades of blue that crossed the light/dark boundary than shades that fell within one category. This was interpreted as evidence that language shapes perceptual processing.

However, this interpretation has serious problems:

• Effect size: The effects are real but small. We are talking about differences of tens of milliseconds on reaction time tasks. The claim that «language shapes thought» based on 30ms RT differences is a significant rhetorical escalation.
• Replication problems: Several attempts to replicate the key findings have produced inconsistent results. The field of colour cognition has been among those affected by the replication crisis in psychology.
• Left/right asymmetry: The effect is stronger for stimuli presented in the right visual field (processed primarily by the left hemisphere, which handles language) than the left visual field. This suggests language is being used as a strategy during the task rather than fundamentally shaping pre-linguistic perception.
• Verbal interference: When participants are given a verbal shadowing task (repeating words while doing the colour task), the effect largely disappears. This strongly suggests the effect is mediated by online language use, not by deep perceptual change.

Spatial Reasoning: The Guugu Yimithirr Example

Another frequently cited example involves spatial cognition. The Guugu Yimithirr people of Australia use absolute spatial directions (north, south, east, west) for all spatial reference — including describing positions on a table in front of them — rather than the egocentric system (left, right, in front of, behind) used in most languages. Speakers of such languages reportedly maintain extraordinary compass-like orientation abilities.

This is genuinely interesting. But the causal story is complicated. Do they maintain better orientation because their language requires it? Or does their language require absolute terms because their culture and environment make maintaining such orientation normal and useful? Separating language from culture and habitual practice in such cases is extremely difficult.

The Methodological Problem

This last point generalises to a deeper methodological challenge for the entire research programme. Languages do not exist in isolation from cultures, environments, and practices. Speakers of different languages typically also have different cultural backgrounds, different educational experiences, and different habitual practices. Showing that they differ on some cognitive task does not demonstrate that the difference is caused by language rather than these other factors.

True causal evidence for linguistic relativity would require showing that changing language — while holding everything else constant — changes cognition. This is enormously difficult to do experimentally. The most convincing approach would be longitudinal studies of people learning a second language with different features, measuring cognitive changes before and after. This kind of research exists but is limited and produces modest results.

The Pop-Science Distortion

The gap between the careful, qualified claims of actual researchers and the sweeping assertions in popular accounts is remarkable. A researcher might publish: «Russian speakers show a 28ms reaction time advantage in cross-boundary blue discrimination tasks, an effect that disappears under verbal interference, suggesting language plays an online strategic role in colour discrimination.» The pop-science version becomes: «The language you speak determines how you see colours.»

This distortion matters because it creates a false picture of what we know. It also fuels a related popular claim — that having more words for something makes you better at it — which is typically stated without any evidentiary basis. The Inuit supposedly have 50 words for snow (a massively inflated figure based on a misunderstanding of how Inuit languages work morphologically) and are therefore better at perceiving snow differences. The reasoning is circular: we infer perceptual superiority from lexical abundance, and then use the lexical abundance as evidence for the perceptual superiority.

What We Can Reasonably Believe

Setting aside the overclaims, what does the evidence actually support?

• Language can influence performance on cognitive tasks, particularly tasks where verbal strategies are available and useful. This is an «online» effect of language use, not a deep rewiring of perception.
• Having a commonly used label for a concept makes it easier to communicate and remember within a community. This is almost definitionally true and not very surprising.
• Different languages draw attention to different aspects of the world through their obligatory grammatical distinctions. Languages that require speakers to mark evidentiality (how do you know what you're saying?) may make evidential thinking more habitual. This is plausible but difficult to study cleanly.

What the evidence does not support:

• That speakers of some languages are incapable of thinking certain thoughts
• That language fundamentally rewires perceptual systems
• That the number of words for X in a language reliably predicts perceptual or cognitive superiority in X

Why Does It Matter?

This might seem like an academic dispute, but the stakes are real. Overclaims about linguistic relativity feed into broader narratives about the fundamental incomprehensibility between speakers of different languages — that there are concepts in Japanese that simply cannot be translated, that Russian speakers live in a fundamentally different perceptual world. These claims are romantic but misleading.

Human beings share a common perceptual and cognitive architecture. We are all equally capable of understanding that the sky is blue, that yesterday preceded today, or that some promises were kept and others weren't — regardless of whether our languages have one word or six for these distinctions. Translation is hard, but it is not impossible in principle. Cross-cultural understanding is difficult, but it is not blocked by linguistic incommensurability.

The hypothesis of linguistic relativity, in its weak and responsible form, is interesting and worth studying. In the strong and popular form it usually takes in non-specialist accounts, it is mostly mythology.