What If Language Completely Shaped Thought?

Speakers of Guugu Yimithirr, an Aboriginal language from north Queensland, do not say "the cup is to your left." They say something closer to "the cup is to the north." Cognitive scientist Stephen Levinson found that Guugu Yimithirr speakers, when blindfolded, spun in a chair, and asked to point toward a location they'd seen minutes earlier, pointed with an average error of just 13.6 degrees (Levinson, 1997). Most English speakers, spun in the same chair, would be lost.

That result is real, published, and replicated in outline by later work. It is also, on its own, a modest finding — a group of people got very good at an orientation task because their language forces them to track compass direction constantly. It is not evidence that Guugu Yimithirr speakers live in an alternate universe of geometry. But it's exactly the kind of finding that sits at the center of one of linguistics' oldest and most misunderstood debates: does the language you speak change what you are capable of thinking?

This piece asks a more extreme version of that question as a thought experiment. What if the strongest possible version of that idea were true — if language didn't just nudge thought but completely constructed it, so that two people speaking different languages inhabited two different, mutually inaccessible realities? We'll look at what real research actually shows, walk through what a world of total linguistic determinism would look like, and then explain why virtually every linguist who has studied this seriously rejects the extreme version while still taking the moderate one very seriously.

The Sapir-Whorf hypothesis, and why "hypothesis" undersells the mess

The idea traces to linguist Edward Sapir and his student Benjamin Lee Whorf, working in the early-to-mid twentieth century. Whorf, an insurance-fire inspector by day and a linguist by night, studied Hopi and other Indigenous American languages and argued that their grammatical structures — especially their treatment of time — shaped a fundamentally different worldview than English. Whorf never actually wrote a formal, testable "hypothesis." What later scholars call the Sapir-Whorf hypothesis is really a retrospective label applied to a cluster of Whorf's essays, and it comes in two strengths that get conflated constantly.

Linguistic determinism (the strong version) claims language determines thought: you cannot think a thought your language has no way to express. Linguistic relativity (the weak version) claims language influences thought: the categories your language makes obligatory or easy will bias your attention, memory, and judgment, without locking out other ways of thinking.

Think of it like a hiking trail through a forest. Linguistic determinism says the trail is the only path that exists — you physically cannot walk anywhere the trail doesn't go. Linguistic relativity says the trail is just the path of least resistance — you can absolutely bushwhack off it, but you'll walk the trail more often because it's there, worn smooth, and requires less effort. Almost the entire modern evidence base supports something close to the second picture, not the first.

What the actual evidence shows

Modern linguistic-relativity research, much of it led by Stanford psychologist Lera Boroditsky, has moved past armchair speculation into controlled experiments. The picture that emerges is: real, measurable, but bounded effects — never a hard wall.

Space and cardinal direction. Levinson's Guugu Yimithirr work above is one data point. A parallel line of research on Kuuk Thaayorre, spoken in the Pormpuraaw community of Cape York, found that speakers spontaneously arrange time along an absolute east-to-west axis rather than a body-relative left-to-right one. Boroditsky and Alice Gaby gave Kuuk Thaayorre speakers shuffled photo sequences (a man aging, a banana being eaten) and asked them to lay the cards in temporal order. Depending on which cardinal direction the person was facing, the "correct" order ran left-to-right, right-to-left, toward the body, or away from it — always tracking east to west (Boroditsky & Gaby, 2010, Psychological Science). English speakers, tested the same way, always lay time left to right regardless of which way they're facing, matching the direction their writing system runs.

Color perception. Russian obligatorily splits what English calls "blue" into two basic terms: siniy (dark blue) and goluboy (light blue). Jonathan Winawer and colleagues, working with Boroditsky, found Russian speakers were measurably faster at distinguishing two blue swatches when they fell into different Russian categories than when both fell into the same category — an advantage English speakers didn't show at all. Crucially, making Russian speakers repeat a verbal task while judging colors erased the advantage; a spatial distraction task did not. That's a strong clue the effect runs through language itself, not general visual acuity (Winawer et al., 2007, PNAS).

Number and counting. The Pirahã, an Amazonian group studied extensively by Daniel Everett, have no words for exact quantities, only terms roughly meaning "one/few," "some," and "many." Psycholinguist Peter Gordon tested Pirahã speakers on quantity-matching tasks and found their accuracy degraded systematically as quantities grew past three or four items, consistent with the two general numerical systems humans share (fast, precise tracking of very small sets, and fuzzy estimation for large ones) — but without any exact counting to bridge the gap (Gordon, 2004, Science; Frank et al., 2008, replicating and refining the picture). Everett has argued this reflects cultural priorities as much as linguistic ones, and the Frank et al. follow-up found Pirahã speakers could perform exact one-to-one matching when memory load was removed — meaning the deficit shows up specifically when a task requires holding an exact number in mind without a word for it, not as a total inability to reason about quantity.

Every one of these findings is genuinely interesting. None of them show that speakers of a different language live in a wholly separate, untranslatable reality. Russian speakers can still perceive that two blues look similar; they're just faster at noticing they're different. Pirahã speakers can still reason about "more" and "less"; they just lack a word-based tool for exact tallying. That distinction — bias and difficulty versus absolute impossibility — is the entire ballgame, and it's the hinge this whole thought experiment turns on.

The thought experiment: what if determinism were literally true?

Set the real evidence aside for a moment. Imagine a world where the strong version wins completely: language doesn't just nudge cognition, it constructs it, full stop. There is no thought outside language, and there is no shared conceptual bedrock underneath different languages. What would that actually look like, mechanically, at the level of everyday reasoning?

Case 1: Tense versus aspect, and the shape of causality

Language A (modeled loosely on English) grammatically requires you to mark every verb for tense — you cannot say "I eat" without committing to whether you mean past, present, or future. Speakers of Language A, on this hypothesis, would be forced to conceptualize time as a strict linear sequence: past causes present, present will cause future, one direction, one arrow.

Language B (modeled loosely on Mandarin, which doesn't grammatically require tense marking) treats verbs by aspect instead: is the action complete, ongoing, or habitual, regardless of when it happened. A determinist reading would say speakers of Language B experience events less as points on a timeline and more as states that simply hold or don't hold, with "when" left to context rather than baked into the grammar.

Under full linguistic determinism, this wouldn't just be a stylistic difference — it would be an unbridgeable cognitive one. Language A speakers would be constitutionally incapable of processing Language B's timeless, aspect-based worldview, and vice versa. A historian speaking Language A and one speaking Language B couldn't even agree that "event X caused event Y," because Language B's grammar wouldn't force a causal, sequential reading onto the events in the first place.

Case 2: Classifier languages and the grain of categorization

Languages like Mandarin, Japanese, and Thai require a "classifier" word whenever you count a noun — you can't just say "three books," you say something closer to "three [flat-bound-object] books." English drops this requirement almost entirely. In the extreme version of the thought experiment, classifier-language speakers would be forced to constantly categorize the physical shape and texture of every object they discuss, making them reflexively concrete thinkers. Non-classifier speakers, never prompted this way, would default to abstraction, since nothing in their grammar makes object-shape mandatory information.

Push this to its logical end and you get two cognitive populations that process nothing at the same "resolution." The classifier speaker's mental model of the world would be dense with sensory categories that the non-classifier speaker's brain never generates in the first place, since — under strict determinism — a concept can't exist for you unless your grammar routinely forces you to produce it.

Case 3: Agency, evidentiality, and who's responsible for what

Some languages, like Tariana and Tsafiki, grammatically require you to mark every statement with an "evidential" — a suffix or particle indicating whether you saw something happen, heard about it secondhand, or inferred it. English lets you skip that entirely: "It rained" is grammatically complete whether you watched it happen or just saw wet pavement. Under linguistic determinism, evidential-marking speakers would be forced to constantly track and weigh the reliability of information, a habit built into every sentence they utter, while English speakers would have no comparable built-in epistemic discipline unless they consciously chose to add it.

Now imagine two courtroom witnesses giving testimony, one in each type of language. Under strong determinism, the evidential speaker literally cannot make a flat assertion without first classifying its evidentiary basis, and a legal system built around the English speaker's language — which allows unqualified assertions — would be structurally incompatible with how the evidential speaker's mind organizes claims about reality. Translation between the two testimonies wouldn't just be imperfect; it would misrepresent the witness's actual epistemic state every time, because the target language has no grammatical slot to carry that information at all.

What a fully language-determined civilization would look like

If you take all three cases and scale them up to civilizational size, several consequences follow directly, and none of them are subtle.

Science stops working. The scientific method depends on an assumption that's easy to take for granted: that observers using different vocabularies are still describing one shared, mind-independent reality, and that their observations can in principle be reconciled. Strong linguistic determinism removes that assumption by definition. If "the Earth orbits the Sun" is a claim constructed entirely within Language A's grammar of causality and sequence, and Language B has no grammatical machinery to represent causality and sequence the same way, there's no neutral ground on which to check whose claim is "true." Peer review, replication, and cross-lab verification, all of which assume observers can converge on the same facts through different descriptions, would be replaced by parallel, non-comparable Language-A-science and Language-B-science.

Translation degrades from imperfect to fictional. Right now, translation is genuinely lossy — poetry, puns, and culturally loaded idioms resist perfect transfer, and every working translator will tell you something is always sacrificed. But translation works, well enough that international treaties, medical research, and literature move across languages every day. Under total determinism, that functional translation would be an illusion: a translated sentence wouldn't approximate the original thought, it would simply generate a different thought in the target language, with no way to check how far the two had drifted, because checking would itself require a shared frame of reference that, by hypothesis, doesn't exist.

Political negotiation between language communities collapses. Diplomacy, trade agreements, and international law all assume that, with enough good-faith translation and back-and-forth clarification, two parties can eventually converge on shared terms. If thought is wholly bound by grammar, "shared terms" is not a communication problem to be solved with better interpreters, it's a metaphysical impossibility. Two governments speaking incompatible grammatical systems for causality, evidentiality, or agency wouldn't misunderstand each other occasionally; they would be constitutionally unable to construct a jointly intelligible treaty, full stop.

Bilingualism becomes cognitive fracture rather than cognitive flexibility. Roughly half the world's population is bilingual or multilingual by most linguistic estimates, and the everyday experience of bilingual people is that switching languages feels like switching registers, not switching realities. Under strong determinism, that everyday experience would have to be wrong: a bilingual person would be two separate cognitive agents sharing a skull, unable to fully translate their own English-thoughts into their own Mandarin-thoughts, with identity and memory potentially bifurcating along linguistic lines. This is close to what philosophers call a Ship-of-Theseus problem for personal identity, except applied to a single living brain rather than a slowly-replaced wooden boat.

Why the extreme version almost certainly fails, and what survives

Three separate lines of evidence push hard against total linguistic determinism, and they converge from very different directions.

Translation empirically works, imperfectly but really. If two languages constructed genuinely incommensurable realities, technical translation, treaty law, and scientific publishing across languages should be riddled with silent, undetectable failures. Instead, translated engineering specifications produce buildings that stand, translated drug trial protocols produce replicable results across countries, and translated legal contracts get enforced in court. That's strong indirect evidence of a conceptual substrate that survives translation, even when idiom and connotation don't.

Preverbal infants already have core cognitive categories. Developmental psychologists have shown that infants only months old, well before they have functional grammar, already expect objects to be solid, to persist when hidden, and to move continuously rather than teleport. Elizabeth Spelke's "core knowledge" research documents these expectations forming before language could plausibly be doing the constructing. If foundational physical reasoning predates language acquisition, language cannot be the sole source of that reasoning.

Universal grammatical categories recur across unrelated language families. Every documented human language distinguishes nouns from verbs in some form, marks questions differently from statements, and allows recursive embedding ("the man who saw the woman who owns the dog"). Linguist Joseph Greenberg catalogued dozens of such implicational universals across genetically unrelated languages. That kind of convergence, across languages with no shared ancestry and no contact, is much easier to explain if there's a shared cognitive architecture underneath all of them than if each language independently constructs an unrelated reality from scratch.

Put those together and the field's actual consensus looks like this: language reliably shapes which features of a situation you notice fastest, remember best, and reach for first — the "trail through the forest" from earlier — without shutting off your ability to notice, remember, or reason about anything your language doesn't happen to have a dedicated word for. Boroditsky herself, whose experiments produced some of the strongest relativity evidence on record, has been explicit that her findings support relativity, not determinism: speakers of different languages think differently in measurable, replicable ways, but they are never cognitively trapped.

Where people still get this wrong

Two persistent misreadings keep the strong version alive in pop-science circles, and both are worth naming directly.

The "Eskimo words for snow" myth. The claim that Inuit languages have dozens (or hundreds) of words for snow, and that this proves language reshapes perception, has been circulating since a mangled 1911 aside by anthropologist Franz Boas. Linguist Geoffrey Pullum dismantled the myth's exaggerated versions decades ago, pointing out that the actual count depends heavily on how you define "word" in polysynthetic languages, and that English speakers, given the motivation, generate plenty of their own snow vocabulary (powder, slush, sleet, corn snow, packed powder). Having more specific vocabulary for something you care about reflects attention and utility, not a different cognitive apparatus.

Confusing "my language lacks a word for X" with "I cannot think about X." A language without a single word for schadenfreude doesn't stop its speakers from recognizing the specific pleasure of someone else's misfortune — English absorbed the German word precisely because English speakers could already conceive of the feeling and just wanted a shorter label for it. Borrowing itself is evidence against determinism: if a concept were truly inaccessible without the right grammar, you couldn't import the word and have it click immediately, the way loanwords routinely do.

Frequently asked questions

Q: Is the Sapir-Whorf hypothesis considered true by modern linguists? A: Not in its strong form. Most linguists and cognitive scientists accept "linguistic relativity" — language measurably influences perception, memory, and attention — while rejecting "linguistic determinism," the claim that language fully controls or limits what you can think.

Q: Did Sapir or Whorf actually claim language completely determines thought? A: Not explicitly. Whorf's original essays described tendencies and habitual patterns of thought associated with grammar, not an absolute cognitive prison. The strongest deterministic framing was largely built up by later commentators and popularizers, then knocked back down by subsequent research.

Q: What's the strongest real-world evidence for linguistic relativity? A: The Winawer et al. (2007) Russian blue-discrimination study is often cited as the cleanest experimental evidence, because it showed the effect appearing and disappearing depending on whether a verbal task interfered with it during the same session, isolating language as the mechanism rather than culture or general perception.

Q: Do bilingual people think differently depending on which language they're using? A: Some measurable shifts have been documented in mood, risk assessment, and moral judgment when bilinguals switch languages, an effect researchers call the "foreign language effect." But this reflects a shift in framing and emotional distance, not a change in fundamental identity or a total swap of cognitive reality.

Q: Does the Pirahã lack of number words mean they can't count at all? A: Not exactly. Studies found Pirahã speakers struggle specifically with tasks requiring exact quantity to be held in memory without number words as a tool, but they perform accurate one-to-one matching when the task doesn't rely on memory for an exact count. It's a gap in a specific cognitive tool, not a total absence of numerical reasoning.

Q: Could artificial intelligence trained on different languages develop different "ways of thinking"? A: This is an active open research question rather than settled science. Large language models trained predominantly on one language do show measurable differences in reasoning benchmarks and cultural bias across languages, but whether that constitutes something analogous to human linguistic relativity, or is simply a reflection of training-data imbalance, is unresolved.

Q: What's the difference between linguistic relativity and linguistic determinism in one sentence? A: Relativity says language biases which thoughts come easily; determinism says language is the only thing that makes thought possible at all, and the evidence overwhelmingly supports the former.

Q: Is color perception really different across languages? A: Basic color perception (the physiology of the eye and visual cortex) is essentially universal, but categorization speed and memory for color can shift measurably based on whether your language has a dedicated word for the boundary between two shades, as the Russian siniy/goluboy research showed.

Q: Why does the "Eskimo words for snow" claim keep coming up if it's been debunked? A: It's an intuitive, memorable shorthand for a real phenomenon (vocabulary tracking cultural attention), even though the specific numeric claims about word counts were exaggerated and the underlying logic — more words means fundamentally different perception — doesn't hold up under scrutiny.

The takeaway

Total linguistic determinism, the idea that language doesn't just influence thought but fully manufactures it, fails every empirical test researchers have thrown at it: translation works well enough to run governments and hospitals, infants reason about physical objects before they can talk, and grammatical universals recur across languages with no shared history. What survives instead is something more interesting than either extreme: language as a set of default settings, not a locked cage. It makes some distinctions effortless and others require deliberate work, the way a compass-direction grammar turns Guugu Yimithirr speakers into reflexive navigators and a color-splitting vocabulary turns Russian speakers into faster blue-discriminators. None of that erases a shared human cognitive floor underneath every language on Earth. The honest version of "language shapes thought" was never about prisons. It was always about which doors swing open first.

Read next:

• What If Humans Developed a Hive Mind Consciousness?
• What If We Discovered Intelligent Aliens Tomorrow?
• What If Consciousness Could Be Uploaded to a Computer?