Skip to content
Published on

Does Language Shape Thought? — The Sapir-Whorf Hypothesis

🇰🇷KO🇺🇸EN🇯🇵JA
Split View필사 모드
Authors

Opening — Can You Think Without the Word?

Let us begin with a thought experiment. If your language had no word for blue, would you see the blue sky differently? If it had no tense to point at the future, would you imagine the future differently? If you spoke a language that named directions only by north, south, east, and west instead of left and right, would your sense of way-finding change?

These questions gather into one enormous question. Does the language we speak mold the way we think? This question has long fascinated scholars, writers, and ordinary people alike.

On one side stands the romantic, powerful intuition that language defines our world. On the other stands the commonsense rebuttal that thought runs deeper than language, and that we think perfectly well without words. Both intuitions are plausible enough that we do not easily lean to either side.

At the very center of this fascinating tug-of-war sits an idea called the Sapir-Whorf hypothesis. Named after the early-twentieth-century linguist Edward Sapir and his student Benjamin Lee Whorf, it refers to the idea that language influences thought. Over a century, this hypothesis has wound its way through enthusiasm and criticism, revival and reinterpretation.

This essay follows that journey. We will examine step by step why the once-prevalent strong claim declined, what moderate picture took its place, and what fascinating research on color, space, and time reveals. To hint at the conclusion in advance: the truth lies somewhere between the two extremes. Language influences our thought, but it does not imprison us. Let us explore that subtle balance together.

1. The Strong and Weak Hypotheses — Two Forking Paths

First let us sort the terms. Claims about the relationship between language and thought commonly split into two strands. These are called the strong hypothesis and the weak hypothesis.

The strong hypothesis is called linguistic determinism. It is the claim that language determines thought, that the language we speak divides what we can and cannot think. Put to the extreme, if a language has no word for a concept, one cannot think that concept at all.

This claim is vivid and seductive. Novels and films have loved to borrow this notion. The fantasy that abolishing a word can abolish the thought it points to is both chilling and intriguing. Some dystopian novels have even portrayed a world in which power shrinks the vocabulary in an effort to narrow people's very thoughts.

The weak hypothesis is called linguistic relativity. It is the claim that language does not determine thought but does influence it. That is, the language we speak makes us attend more to certain things, makes certain distinctions easier, or gently nudges a certain mode of thinking. Determine versus influence. The difference between the two seems small but is in fact enormous. To determine is to imprison; to influence is to tilt.

Today the general consensus in the field is this. The strong hypothesis is wrong, and the weak hypothesis is correct to a degree. That is, language does not cage our thought like iron bars, but it does subtly tilt our attention and habits. Let us now look one by one at why the strong hypothesis collapsed and how the weak one survived.

2. Why the Strong Hypothesis Declined

The strong hypothesis, that language determines thought, once enjoyed considerable popularity. But over time this strong form of the claim gradually lost force. There are several reasons.

First, clear evidence accumulated that we can think a concept even without a word for it. That a language lacks a word for a certain color did not mean its speakers could not see or distinguish that color. They simply lacked the name; their eyes discerned the color all the same. When we meet a new concept, we coin a new word or string several words together to express it. Often the thought comes first, and language catches up to grasp it afterward.

Second, if language fully determined thought, translation should be impossible. The thoughts of one language should be untransferable to another, and learning a foreign tongue should never let one understand a new concept. Yet we translate, learn foreign languages, and absorb the ideas of other linguistic worlds. Not perfectly, but well enough. This shows that thought is not entirely caged within any one language.

Third, there is the thinking of infants and animals. A baby before learning to speak still perceives the world, faintly distinguishes quantities, and makes predictions. Animals without language solve problems and make plans. This suggests that thought stands on a foundation deeper and older than language. If language were the whole of thought, beings without language could think nothing, but reality is otherwise.

For these reasons, the strong claim that language determines thought gradually lost credibility in the field. Some of the stories once frequently cited as flagship examples of this hypothesis later turned out to be exaggerated or inaccurate. That does not mean the whole relationship between language and thought was a mirage. Where the strong claim collapsed, a more refined and interesting weak claim survived.

3. The Revival of the Weak Hypothesis — Language Tilts Thought

In the later twentieth century, more careful and refined research began to cast new light on the weak hypothesis. Rather than the grand claim that language determines thought, these studies focused on the testable question of whether language has a subtle effect on thought in particular tasks. And interesting results emerged.

The core insight is this. Language does not determine what we can think, but it tilts what we more often attend to. When a language grammatically forces a certain distinction, its speakers habitually pay more mind to that distinction. This is sometimes called thinking for speaking. To speak, we must always attend to certain information, and that habit faintly soaks into thought.

For instance, some languages assign a grammatical gender to every noun. Objects like a bridge, a key, an apple are classified as masculine or feminine.

In one fascinating study, even for the same object, the impression people called to mind tended to shift slightly depending on the object's grammatical gender. A bridge that was a masculine noun tended to be described as sturdy, while a bridge that was a feminine noun tended to be described as elegant. This suggests that a language's grammar can subtly reach even into our associations. Yet such studies, too, have drawn debate over replication, so how consistent and how large the effect is must still be handled with care.

The important word here is subtly. Such influences are usually small, confined to particular tasks, and easily overcome with conscious effort. A different grammatical gender does not change one's entire thinking about a bridge. It is only that, absent other cues, language tips the scale a little. The weak hypothesis speaks precisely of this slight tipping.

4. The Language of Color — Is Blue One or Two?

When studying the relationship between language and thought, color is a perennial subject. Color is a physically continuous spectrum, yet languages cut it differently. So color makes a good laboratory for testing whether language influences perception.

One fascinating case: some languages split what we simply call blue into two separate colors. They have distinct basic words for light blue and dark blue, much as we treat red and pink as separate colors. So do speakers of such languages distinguish light blue and dark blue faster?

Several studies tested this, and an interesting tendency appeared. People with separate names for the two kinds of blue tended to perform the task of distinguishing them slightly faster. It is as if language had planted a small signpost atop that color boundary. Here too, though, the difference was not large and stood out under particular conditions.

More intriguing is a clue that this effect relates to the left and right of the brain. In some studies, the effect of language on color distinction appeared more strongly mainly in the right visual field, the side connected to the left brain that processes language. This suggests that language really does slip into our color-perception process. Of course, such results are still under active research, and there is debate over the size and conditions of the effect.

The key point to mark here is that a different color vocabulary does not mean one sees different colors. Everyone with normal color vision receives the same physical light as the same signal. What language changes is the subtle processing: how fast we classify the color, how clearly we remember it, where we place our attention. Not the foundation of perception, but the habits of attention and classification layered atop it, shift slightly with language.

5. The Language of Space — Left or North?

Among studies of language and thought, the most striking domain is spatial expression. The way we point at directions differs by language, and there is evidence that this difference can affect actual cognition.

Most languages state direction relative to me: left, right, front, back. This is called a relative direction system. When I turn around, left and right turn with me. But some languages state direction only by absolute references, by north, south, east, and west. Instead of the cup is to the left of the desk, they say the cup is to the north of the desk. A speaker of such a language must always know which direction they face, because otherwise they cannot even speak.

Intriguingly, speakers of such absolute-direction languages are reported to tend to have a remarkable sense of direction. Even in an unfamiliar room, even in a windowless one, they will accurately point to the cardinal directions. For them, always being aware of the compass is as natural as our awareness of left and right. Because the language always demands that information, the ability to track it has been honed.

This is a fine example of the weak hypothesis. The absolute-direction language did not grant its speakers a superpower. Anyone can train a sense of direction. It is only that the language forces its speakers to constantly track compass information, so that ability develops in everyday life. Language did not determine thought; it diligently cultivated a particular habit.

There is research that differences in spatial expression can also affect how one imagines time. Some imagine the flow of time from left to right, others from east to west. When we picture time by analogy to space, which language's way of expressing that space we speak may exert a faint influence. Such findings show that language can cast a subtle shadow even over abstract thought.

6. The Language of Time — Is the Future Someone Else's Affair?

Temporal expression is also a fascinating research subject, because the grammar for handling time differs greatly across languages. Some languages have grammatical devices that sharply distinguish the future from the present; others make that distinction weakly.

Here a much-discussed hypothesis once arose. Might speakers of languages that grammatically distinguish the future strongly from the present feel the future as more distant and other, and thus tend to prepare less for the future through saving or health care? Conversely, might speakers of languages that treat future and present similarly feel the future as closer and prepare better? Such was the conjecture.

This hypothesis is very tempting, but also highly contentious. An analysis claiming to show a correlation between language and economic behavior was put forward, but much criticism followed. The biggest problem is the distinction between correlation and causation. Even if a correlation appears between groups speaking a certain language and a certain economic behavior, it is hard to assert that it is because of the language. Language is deeply entangled with culture, history, and institutions, so it is very tricky to isolate language and discern its effect alone.

This case shows both the appeal and the trap of linguistic-relativity research. The appeal lies in making us imagine the influence that something as familiar as language might have on our behavior. The trap lies in how much harder it is to prove that influence than one would think. A good scientist must test a tempting hypothesis all the more strictly. The language-of-time hypothesis posed an interesting question, but its answer can hardly be called settled.

The balanced attitude we should learn here is this. When you meet the appealing one-liner that language changes one's view of the future, rather than believing it at once, ask whether this is really an effect of language or of culture or some other factor. Linguistic relativity is a real phenomenon, but proving it demands caution.

7. Multilingual Speakers — Does Switching Languages Switch the Person?

The experience of people who speak several languages throws another light on this subject. Multilingual speakers often say that switching languages is like becoming a different person. What does this fascinating phenomenon tell us?

First let us note it carefully. There are several mixed reasons why a multilingual speaker feels differently in different languages. One language may be the emotional language learned with family in childhood, another the formal language used at work. In this case the change of feeling when switching languages is likely a product of the context, memory, and culture bound to the language, more than of the language itself. That is, language does not directly change thought so much as flip a switch of a particular situation and emotion.

Even so, there are interesting observations. In some research, when people judged a moral dilemma in a foreign language rather than their mother tongue, they were reported to show a tendency slightly different from judging in their mother tongue. Using a foreign language seemed to create emotional distance and a tendency toward cooler judgment. This suggests that language can subtly adjust the temperature of our emotional response. The mother tongue is more deeply entangled with emotion, while a later-learned foreign language holds a slightly calmer distance.

Such findings do not support the strong hypothesis either. Thinking in a foreign language does not give one an entirely different morality. It is only that language acts subtly on our emotion and attention, so the same problem can be cast in a slightly different hue. The experience of multilingual speakers supports the picture that language is closer to a filter that tints thought than to iron bars that cage it.

One clear fact is that knowing several languages broadens thought rather than narrowing it. A multilingual speaker can view the same object through the lenses of several languages. What one language misses, another catches. If language caged thought, knowing several languages would be holding several prisons; but in reality it is closer to holding several windows.

8. Criticism and Rebuttal — The Virtue of Caution

Linguistic-relativity research, appealing as it is, has drawn much criticism. Examining these criticisms is essential for a balanced view.

The most common criticism concerns effect size. The effects reported by studies claiming language influences thought are usually small: a slightly faster reaction time, a slightly different tendency on a particular task. Critics ask, may one draw the grand conclusion that language shapes thought from such small effects? Inflating a small effect into a big story should be guarded against. This is a fair point.

Another criticism is replicability. Even if some striking result emerged in one study, another researcher may fail to obtain the same result under different conditions. In science a single result is only a hypothesis; it becomes trustworthy fact only when repeatedly replicated. Some famous results of linguistic-relativity research stand in the midst of debate over replication.

A third criticism is the correlation-and-causation problem seen earlier. Even if a correlation appears between language and some thought tendency, it is hard to discern whether it is because of language or because of the culture and environment that travel with language. Language never exists in a vacuum. It is bound as one body with a community's history, way of life, and values. So isolating the pure effect of language requires very careful design.

These criticisms are not meant to negate linguistic relativity. Rather, they are meant to make it sturdier. Good science tests an appealing claim all the more strictly. The part that passes the criticism and survives becomes genuinely solid knowledge. When it comes to language and thought, we have arrived at the cautious conclusion, forged in the fire of criticism, that language clearly does influence, but the influence is small, subtle, and context-dependent.

9. Sapir and Whorf — The People Behind the Hypothesis

The story of the two men who lent the hypothesis their names is a good thread for understanding how this doctrine was first inflated and then pared back down. The intriguing point is that neither of them ever wrote down a strong hypothesis as a tidy formula. The very name Sapir-Whorf hypothesis is itself closer to a later summary that posterity drew from their writings and bundled together.

Edward Sapir was one of the foremost linguists of the early twentieth century. Under the influence of the anthropologist Franz Boas, he studied Native American languages deeply, and he was struck by how differently each language carves up the world. Yet his writing was largely cautious. He saw language as working like a background to our habitual thought, but he did not go so far as to assert that language thoroughly cages thought.

Benjamin Lee Whorf had an unusual biography: originally a fire-insurance engineer, he came to study under Sapir. Researching languages such as Hopi, he left behind bold writings arguing that a language's grammar seeps deeply into its speakers' worldview. His prose was captivating and poetic, and it was precisely that charm that got him widely quoted and also exaggerated. His passages suggesting that Hopi has almost no concept of time are especially famous, and later researchers strongly rebutted them, noting that Hopi in fact has rich devices for handling time.

Two lessons emerge here. First, the more appealing a doctrine, the more easily a wide gap opens between what the original texts actually said and what the public remembers. Second, one person's striking observation is only a starting point; for it to become real knowledge, it must pass down the rough road of criticism and re-verification. Sapir and Whorf posed a grand question beautifully, but refining the answer was left to a century of collective work.

10. The Language of Counting — Do Few Number Words Blur the Count?

One of the sturdiest pieces of evidence for linguistic relativity comes, unexpectedly, from number. Most languages have a rich set of exact number words, but a few languages have very few words for counting at all. One, two, and beyond that simply lumped together as many.

Studies of people who speak such languages reveal an interesting pattern. They handle small quantities exactly, but as numbers grow, they tend to struggle on tasks requiring an exact one-to-one match. Matching a row of seven objects with the same number of objects in another row, for instance, turned out to be surprisingly difficult for speakers without number words.

This must be interpreted carefully. It does not mean they cannot perceive quantity. Every human is born with an ability to grasp small quantities at a glance and to estimate the magnitude of large ones. It is only that to handle exact large numbers, one seems to need a linguistic tool for counting, a kind of cognitive ladder made of number words. A number word is at once a label pointing at a quantity and a tool that makes exact counting possible.

This case is interesting because it shows that language can act not merely as a slight tilt of attention but as a tool that makes a particular kind of precise thinking possible. If in the color and direction cases the effect of language was a faint tilting, in the number case language is closer to an outright instrument. That does not mean the strong hypothesis is correct, though. Someone without number words can soon handle large numbers exactly once they learn the new words. They merely lack the tool; the hand to grasp it belongs to everyone.

11. The Grammar of Evidentiality — Languages That Make You Say How You Know

When we say in English that it rained, we need not specify whether we saw it directly, heard it from someone, or inferred it from wet ground. But some languages grammatically force that information. The very form of the verb must mark how the speaker came to know the fact. This is called evidentiality.

A speaker of a language with evidentiality must mark the source of information every time they utter a sentence. Whether it was directly witnessed, heard secondhand, or inferred from a trace is distinguished by the verb ending. A speaker of such a language cannot help, in order to speak, always being aware of how I know this. Since weighing the source of information is etched into the grammar, it might well soak into habitual thought too.

This is a vivid case of the thinking for speaking seen earlier. When a language grammatically demands certain information, its speakers are trained always to attend to that information. The hypothesis that speakers of an evidential language are more sensitive to sources of information, or better at distinguishing rumor from witness, is appealing and plausible. Yet here too balance is needed. Speakers of a language without evidentiality can perfectly well say I saw it myself versus I heard so. The difference lies only in whether it is forced, not in whether it is possible.

Even so, evidentiality is a superb subject for thinking about linguistic relativity. If, for the same event, one language may omit the source while another must state it, the everyday habits with which speakers of the two languages handle information may come to differ in subtle grain. That is not a prison of thought but more like a small groove worn into the path along which thought flows.

12. The Names of Kin — Different Knives for Carving Up Family

Words for family relations also differ greatly by language. Some languages finely distinguish the father's side from the mother's, and the elder from the younger. They have separate words where English makes do with far fewer; the English uncle covers what another language splits into several distinct terms.

Such differences are a mirror reflecting which grain of family relations the community speaking that language has held important. A language that sharply distinguishes elder from younger makes its speakers habitually conscious of age and rank. A language that divides the paternal from the maternal line carries within it a history in which that distinction was socially meaningful. Kinship vocabulary is not a mere set of labels but a map of how a society has woven its human relationships.

So do such vocabulary differences affect actual thought? Studies give a cautious answer. There are reports that speakers of a language with separate words for elder and younger siblings tend to recall the relative age of siblings faster. This is the same grain as the color and direction cases. When a language always demands a certain distinction, attention to that distinction becomes a little faster and clearer.

Here too, though, overstatement is forbidden. A speaker of a language with only the one word uncle is not ignorant of the difference between a paternal and a maternal uncle. When needed, they distinguish it perfectly well by unpacking it, as in my father's elder brother. The difference in kinship vocabulary does not divide whether we can think about family; it merely tilts which distinctions we attend to by default. The same family is cast in a slightly different grain by each language, yet crossing that grain to understand one another's families is entirely possible.

13. The Untranslatable Words — What the Blank Tells Us

Some languages have a word hard to render in another in a single breath. Expressions that catch a particular longing, a particular ease, a particular shame in one word. Such untranslatable words are a charming subject that often appears when we speak of language and thought. People ask whether a speaker of a language without that word cannot feel that emotion.

The answer is clearly no. Lacking the word does not mean one cannot feel the emotion. We feel nameless emotions perfectly well, and when needed we explain them by stringing several words together. That a language has the word means the community has been conscious of that emotion often and clearly enough to name it, not that the emotion is their exclusive possession.

But the existence of the word does not do nothing, either. The moment an experience is named, it gains a clearer outline. A scattered feeling is bundled into a single thing, more easily recalled, more easily conveyed to another. A name condenses a vague mist into a form one can hold in the hand. So learning a new word is less like manufacturing an emotion that was absent than like sharpening one that was blurry.

What untranslatable words really tell us is not the prison of language but its abundance. The grain that one language has carefully named can be learned and made one's own by a speaker of another language. That is how loanwords enter, and how our emotional vocabulary widens. The blank in translation is not a closed wall but an open window. Through that window we come to see, together, the grain that another language saw clearly.

14. We Think in Metaphors — Time Is Money

Perhaps the most everyday channel by which language seeps into thought is metaphor. When we speak of the abstract, we almost always liken it to the concrete. Consider the expressions save time, waste time, invest time. We speak of time as though it were money, a resource that can be spent, saved, and lost. The metaphor time is money is steeped throughout our language.

There is a view that such metaphors are not mere ornaments of expression but mold the very way we understand abstract objects. We liken argument to war, saying we demolished his claim or an indefensible position, and in speaking so we come to treat argument as a fight one actually wins or loses. Metaphors that liken ideas to food, as in a hard idea to digest, or life to a journey, as in standing at a crossroads, form the skeleton laid beneath our thinking.

Here an interesting question arises. If each language favors different metaphors, does that difference affect thought? In one study, people were given the same social problem described through different metaphors. When crime was described as a beast preying on a city versus as a disease spread through a city, the solutions people preferred tended to shift slightly. The beast framing called to mind enforcement and punishment; the disease framing, diagnosis of causes and reform.

This is another face of linguistic relativity. Yet the influence of metaphor, too, is a tilt rather than a fate. A metaphor lights up one aspect while casting another into shadow. The metaphor time is money highlights time's scarcity but hides that, unlike money, time cannot be saved up. Good thinking is to become aware of the metaphors we use unthinkingly and, when needed, to swap in a different metaphor and light the same object anew. Metaphor does not cage us. It only quietly guides where we shine our light.

15. Artificial Intelligence as a New Mirror — The Question Language Models Pose

To the old question of language and thought, an unexpected new angle has recently been added: the rise of artificial intelligence trained on vast bodies of language, the large language model. These systems feed and grow on language alone. They do not see or touch the world directly; they learn only how words consort with one another. So what might they reflect about the relationship between language and thought?

On one hand, their abilities can look like evidence against the strong hypothesis. These models handle concepts that appeared little in their training data, move fairly smoothly between languages, and produce combinations they were never explicitly taught. If language thoroughly caged thought, such flexibility would be hard to explain. At least at the level of statistical pattern, meaning does not seem firmly bound to any one language.

On the other hand, they freshly recall the weak hypothesis that language molds thought as a tool. These models' thinking is drawn entirely from the statistical structure of language. So the biases and associations steeped in the language they learned are inherited wholesale by the model. Something like the way human thought is faintly tinted by the habits of language happens, far more nakedly, in the machine. This may become a new laboratory for peering into how language molds the material of thought.

Great caution is needed, though. The way artificial intelligence handles language and the way humans think with language are by no means the same. Human thought is rooted deeply in the body and senses, in emotion and social experience, whereas a language model works without such a foundation, on word statistics alone. So when we take a language model as a mirror of human cognition, we must carefully sort what is alike from what is different. Even so, this new technology is surely a fascinating mirror that makes us ask a century-old question anew from an entirely different angle.

16. So Does Language Shape Thought? — A Balanced Conclusion

Let us gather the long journey. To the opening question, does language determine thought, we can now give a far more refined answer.

The strong hypothesis, that language cages thought, is wrong. We think, translate, learn new concepts, and move between languages even without words. Thought is too deep and flexible to be caged in any one language. Infants and animals without language perceive the world and solve problems. Thought is older and broader than language.

But the weak hypothesis, that language influences thought, is correct to a degree. Language subtly tilts what we attend to, which distinctions we make more easily, which information we habitually track. The hand of language faintly touches the boundaries of color, the directions of space, the imagining of time. The influence is small but real.

So the most accurate picture is this. Language is closer to a filter that tints thought than to iron bars that cage it. The same world, passing through the filter of a different language, is cast in a slightly different hue. But that filter cannot stop us from seeing other colors. We can always become aware of the filter, swap it, and transcend it. Learning another language is gaining one more filter, and that broadens our view rather than narrowing it.

There is a strange beauty in this balanced conclusion. Had language fully determined thought, humanity would have been stranded on islands of separate languages, forever unable to understand one another. Conversely, had language no influence at all on thought, the diversity of languages would have been mere different name tags on the same thoughts. The truth lies between. We cast the world a little differently in different languages, yet through translation and learning we can cross that difference and reach one another. The diversity of languages is humanity's many windows onto the world, and we see a broader landscape by moving among those windows.

Closing — On Holding More Windows

Let us return to the opening thought experiment. Without the word blue, would you see the blue sky differently? The answer is this. You still see that blue. It is only that the way you name it, attend to it, and remember it clearly shifts slightly with language. Language cannot erase the blue of the sky. It only changes the shape of the finger pointing at that blue.

Exploring the relationship between language and thought leads, in the end, to understanding ourselves. We see the world wearing the spectacles of the language we speak. Those spectacles are usually transparent, so we forget we are even wearing them. Research on linguistic relativity lets us, for a moment, become aware of those spectacles' existence. The moment we are aware of the spectacles, we can wipe them, swap them, or take them off for a while.

The finest thing is that we are not caged within those spectacles. We can learn a new language and add new spectacles, and through translation peer beyond another's spectacles. Language tints our thought, yet does not cage us. And within that very fact dwell, together, the freedom and diversity of the human mind.

Things to Sit With

  • Is there a word unique to your language, hard to render in another? What thought or feeling does that word hold?
  • Have you ever felt a different self when speaking a foreign language than when speaking your mother tongue? Was that an effect of the language itself, or of the context bound to it?
  • When you hear the claim that speakers of this language think in such-and-such a way, how can you tell whether it is an effect of language or of culture?
  • If you could coin one new word and spread it widely, which thought of people would you want to make vivid?

A Small Quiz

  1. What is the strong claim called, that language determines thought?
  2. What is the moderate claim called, that language influences thought?
  3. What is the direction system called, that states direction by north, south, east, and west rather than by me?
  4. What problem is frequently noted as the reason a correlation cannot at once be deemed because of language?

(Answers: 1. Linguistic determinism (strong hypothesis) 2. Linguistic relativity (weak hypothesis) 3. Absolute direction system 4. The problem of distinguishing correlation from causation)

References

Discuss on Twitter • View on GitHub