Language Is a Subset of Natural Reality
Between Chomsky, Wittgenstein, and the World We Name
We abbreviate reality in language and science, intensify it in art, and catalyse it in poetry, always from within the world we describe.
We name the world, sort it into kinds, describe it and try to explain it. We build theories about where it came from, write laws for how it behaves, and argue about our place in it. Underneath all of that sits a question that never quite goes away: are we describing reality as it is, or only the version of it that our minds and languages let us grasp?
It is harder to answer than it looks, mainly because we never get to stand outside the world and look at it. We are organisms inside the thing we are studying. Our brains are made of the same stuff as everything else, our senses work through physical processes, and our languages grew out of people dealing with each other and with the places they lived.
So we are not peering at reality through a window made of words. We are part of reality, trying, through consciousness, to make sense of itself.
I should be careful with that last sentence. It is a philosophical way of putting things, not a scientific result. The claim I actually want to defend is smaller: human inquiry is a natural process, and it happens inside the world it investigates.
Language is one of the main tools that make that inquiry possible. It is enormously powerful. But it isn’t the whole of human experience, and it isn’t a clear pane of glass either. Most of the trouble starts when we forget the difference between our descriptions and the things they describe.
1. Is Language Simply the Naming of Things?
An old and stubborn idea holds that words are basically names. A tree has one, a river has one, so does a person. You point, you attach a sound, and from then on you can talk about the thing.
There is truth in this, but it runs out quickly. Try pointing at justice, or possibility, or nothingness, or infinity, or tomorrow. Some of these are abstractions, some are relations, some are categories of time or logic. You can’t explain any of them by holding up an object.
We also use language to ask, order, promise, tell stories, guess, grieve and imagine how things might have gone otherwise. We can talk about unicorns without believing in them. We can picture a universe with different laws of physics, and prove theorems about infinite sets without ever meeting an infinite pile of anything.
Two things follow. Being able to refer to something doesn’t show that it exists. And not having a word for something doesn’t show that the thing, or the experience, isn’t there.
So language is much more than a list of names. It is a whole system of symbols, grammar, social habits and mental work that people use to communicate and to organise meaning.
That matters as soon as we ask whether the structure of language tells us anything about the structure of reality. Languages differ in how they sort objects, mark time, describe movement and arrange relationships. Do their speakers experience the world differently as a result? If they do, is that a fact about the world, or about the conceptual systems that make it intelligible? These are partly empirical questions and partly philosophical ones. Neither gets answered if we simply assume that words mirror things.
2. Chomsky and the Search for Linguistic Universals
Few modern thinkers have shaped the scientific study of language as much as Noam Chomsky. He rejected the idea that children learn language through imitation, reward and general learning alone. Instead he argued for an innate capacity for grammar, specific to our species (Chomsky, 1965).
The theory he called Universal Grammar has been rebuilt many times since, but the puzzle behind it hasn’t gone away. Children hear a limited, messy sample of speech, and yet they end up producing and understanding sentences nobody has ever said to them. They pick up grammatical patterns and become fluent without being taught most of the rules they follow. Chomsky’s great contribution was to turn that ability into a serious scientific problem.
There is a distinction worth holding onto, though. Sharing a biological capacity for language is one thing. Sharing a detailed grammar is another. People everywhere may have the same equipment for learning language and still build very different languages with it.
Evans and Levinson (2009) made exactly this argument in The Myth of Language Universals. Drawing on research across many languages, they showed that languages vary far more than is often assumed: in their sound systems, in their grammar, and in how they relate ideas and events to one another. They didn’t show that there are no universals at all. Their point was that we can’t take widespread similarity for granted without comparative evidence to back it up.
The article drew sharp replies. Critics argued that differences on the surface might still hide deeper common ground. That kind of argument is healthy. You can’t establish a universal because it shows up in a few familiar languages, and you can’t throw one out just because languages obviously differ.
Getting this right takes discipline. A proposed universal has to be defined precisely, with a clear idea of what would count against it. It has to be tested on a genuinely varied set of languages, and real structural similarity has to be kept apart from the kind that loose definitions manufacture. Relying on languages from one historical family is especially risky. Build a theory on a narrow sample and you may end up describing the sample rather than language.
None of this questions whether Chomsky understood something important. He changed the field. What matters is which particular claims about universals hold up once all the cross-linguistic evidence is in. You can respect a scientist without surrendering to him.
3. Wittgenstein: Meaning Is More Than Naming
Ludwig Wittgenstein came at the problem from somewhere else entirely. In the Tractatus Logico-Philosophicus (1921/1922) he asked how a proposition could picture a possible state of affairs, and how the structure of a representation relates to what it represents.
By the time of Philosophical Investigations (1953) he had given up on the idea that one general theory of representation could explain language. He turned instead to how varied our use of language actually is. Words get their meaning, he argued, from the part they play in our activities, conventions and forms of life. He called these practices language-games, situations where talking is tangled up with doing.
Take promise. It isn’t a label stuck on an object. Say the right words in the right setting and you have put yourself under an obligation.
Take help. What it means depends on what people are doing, what they need, and what job the word is doing in that moment.
Or take Why? Depending on who asks and when, it can be curiosity, a demand for justification, a challenge to authority, a request for a cause, or a doubt about an assumption. Same word, different work.
This makes the later Wittgenstein a real problem for any theory that tries to explain all meaning with one mechanism. Still, it would be too much to treat his work as a refutation of Chomsky’s, which came later, or to say he alone understood language. The two men were after different things.
Chomsky wanted to know about the mental structures behind our linguistic competence. Wittgenstein wanted to know how language comes to mean anything within human life. Their approaches can clash, but they aren’t rival answers to the same question. A theory of how children acquire grammar won’t explain every aspect of meaning. A philosophy of meaning won’t tell you how a child’s brain picks up grammar. A full account of language needs both levels, and has to keep them distinct.
What Wittgenstein left us wasn’t a set of answers. He made it very hard to go on assuming that one model could explain everything meaningful we do with words.
4. Language and Reality: Which Comes First?
At first glance, the idea that language represents reality seems simple enough. Things exist, we perceive them, and language gives us symbols to describe them. It gets murkier once we look at the concepts we use to make sense of what we perceive.
Start with the word reality. We might mean the physical universe, or the world of everyday experience, or everything that exists, or the world as consciousness interprets it. These overlap, but they aren’t the same. And whichever we mean, we can only say it in language. Even the claim that reality exists independently of language is, after all, a sentence.
Does that make reality something language constructs? I don’t think so. Needing language to state a claim doesn’t mean the thing the claim is about needs language to exist. The conditions under which we describe something are different from the conditions under which it exists. A mountain wears down without having a name. A virus infects whether or not anyone understands it. A child falls without any concept of gravity.
Our understanding of mountains, viruses and gravity is another matter. That is shaped by observation, by categories and instruments, by what we have inherited from others. We never meet scientific reality raw. We work our way towards more reliable descriptions of it, using methods that let us test and correct ourselves.
So the distinction that matters is between reality’s existence and its intelligibility. Reality may well exist independently of what we say about it, while our ability to understand it depends partly on the concepts and words we have to hand. Language doesn’t need to create the universe in order to shape the universe we are able to describe.
The other side has a real point too. We aren’t handed a world already sorted into categories, waiting for labels. We learn to tell objects apart, classify experiences and read events through social and cultural practice. Science invents concepts that let researchers see patterns they couldn’t otherwise put into words. The idea of the gene didn’t create inheritance, but it changed how inheritance could be studied: what researchers noticed, what they measured, which questions seemed worth asking.
The relationship, then, is neither simple mirroring nor free invention. Language is a natural human capacity for representing the world we live in, and it also helps organise how that world comes to mean something to us.
In its strongest form the question is still open. What could we know about reality apart from every possible way of representing it? We shouldn’t pretend to have an answer. But we also shouldn’t treat the limits of our knowledge as proof that there’s nothing beyond what we can say.
5. We Are Reality, Not Spectators Outside It
Talk of our “relationship” with reality can make it sound as though we stand apart from the world we study. We don’t. The brain is an organ. Consciousness happens in living bodies. Speech is air pushed through a throat, shaped by muscles and decoded by someone else’s ears. Writing needs hands, materials and tools.
Language isn’t some immaterial bridge between a free-floating mind and a universe it has nothing to do with. It belongs to the natural and social world. When we use it to investigate nature, there’s a sense in which nature is investigating itself, through one particular kind of living thing.
That’s an interpretation of where we stand, not evidence that the universe has a mind or a purpose. What it does undercut is the idea that knowing something requires an observer standing completely outside it. We study our own minds, the biology of our own senses and the social history of our own concepts, and we do it while remaining part of what we study.
So the problem isn’t that we have no access to reality. It’s that our access is partial, embodied and fallible, and it runs through the very faculties we use to perceive. Science makes progress because our first attempts can be checked against observation and revised when they turn out wrong.
If language and thought had no contact with a world able to push back against our expectations, there would be no difference between an explanation that works and one that fails. Reality isn’t just whatever we say it is.
6. Do Languages Reveal Universal Features of Human Experience?
The debate about universals gets sharper once we separate what people experience from how they put it into words. Everywhere, people deal with objects, movement, relationships, hunger and changing surroundings. That doesn’t mean every language organises those experiences in the same way. One language builds into its grammar what another leaves to separate words or to context. Languages divide up space and time along different lines.
The interesting question, then, isn’t whether people everywhere can express similar ideas. They usually can. It’s whether languages show recurring patterns in how those ideas get encoded, and that is exactly what careful comparison has to find out.
Evans and Levinson (2009) argue that linguistic diversity runs deep enough to undermine strong claims about a universal grammatical structure. They urge researchers to study each language on its own terms, rather than as a variation on the familiar European ones.
Diversity doesn’t rule out every kind of universality, though. All languages are learned and used by one species, working within the same limits of perception, memory, learning and social life. Those shared limits could easily produce recurring patterns without any common blueprint behind them.
A pattern that keeps turning up might have several different causes:
- biological constraints on what human minds can learn;
- functional pressures, since some structures are easier to learn, process or pass on;
- history, through shared ancestry or contact between languages;
- method, as an artefact of how the data were sampled.
So spotting a pattern doesn’t tell you where it came from. Researchers have to tell a true universal apart from a statistical tendency, a historical accident, a functional preference or a sampling artefact. The same care applies to anyone claiming there are no universals. Define a universal too broadly and it tells you very little. Define it precisely and you can test it.
We don’t have to choose between naive universalism and anything-goes relativism. The job is to find out which properties recur, which vary, why, and what that says about how human minds work. The debate stays open because language is biological, cognitive, social, historical and cultural all at once, and no single level of explanation can settle the rest.
7. Does Language Shape Thought?
If languages organise experience differently, do their speakers end up paying attention to the world in different ways? That’s the question of linguistic relativity.
In his 1997 review, John A. Lucy set out several ways of studying how linguistic structure relates to thought. Instead of assuming that language determines thought, researchers can ask whether a particular linguistic distinction affects a particular cognitive task. Wolff and Holmes (2011) surveyed the evidence across motion, colour, space, number and reasoning about other minds. They found support for some kinds of linguistic influence, but not for the stronger claim that language fixes the basic categories of thought.
That difference matters. A language may steer its speakers towards certain distinctions, make some descriptions habitual, or hand them useful conceptual tools. None of that stops anyone thinking beyond their language’s usual categories. People learn new concepts, pick up second languages, master scientific vocabularies and question the classifications they grew up with. The fact that we can criticise our own language at all shows that linguistic habits aren’t prisons.
Language can guide attention without dictating every thought. It can make some kinds of reasoning easier without making others impossible.
This has real consequences in classrooms. A teacher who takes unfamiliar vocabulary as a sign that a student doesn’t understand is confusing how someone speaks with what they know. One student may understand a scientific process perfectly well without knowing the technical terms. Another may recite the terms without understanding the process at all. Language is a tool for thinking, but fluency and understanding are different things.
8. When Language Becomes a Political Instrument: Examples From Four Continents
All of this becomes concrete when linguistic categories start deciding things in schools.
Take Cook Islands children in Australia. Many don’t come directly from the Cook Islands. They come via New Zealand, as New Zealand citizens; immigration data counted 5,269 New Zealanders in Australia who were born in the Cook Islands (The Conversation, 2014). Their language situation doesn’t fit neatly under one label. Cook Islands Māori is classed as endangered, and more and more young people speak only English, or understand Māori without speaking it (Endangered Languages Project, n.d.). Yet some outer-island schools still teach in Māori up to Year 3 and only begin formal English at Year 4 (Cook Islands News, n.d.).
The eligibility rules can turn that route into a barrier. In New South Wales, a student can’t receive intensive New Arrivals Program support if they use English at home, or if English was the main language of instruction at their previous school (NSW Department of Education, n.d.). A child who spent a few years in an English-medium school in New Zealand can fail that test regardless of how well they actually cope with academic English. Once they are in mainstream classes, Pacific children are counted in the broad category of children for whom English is an additional language, but the data aren’t broken down by Pacific language. That makes it hard to see how they are being supported with literacy, or how they could be (The Conversation, 2014).
Nobody has to make a mistake for this to happen. Home language, previous language of instruction and visa status are each sensible criteria. Put them together and they can leave a child with real English-language needs outside the system built to meet them. That’s the problem at its plainest: a label can be correct by its own rules and still miss the child.
The underlying problem reaches well beyond one community. Administrative language categories shape how institutions notice students’ needs and how they share out resources.
Picture three students. One chats easily in English but struggles with academic English. Another uses English alongside Cook Islands Māori at home and in the community. A third is highly literate in one language and still building literacy in English. No yes-or-no label covers all three.
Conversational skill, academic proficiency, literacy and home language are related, but they aren’t the same thing. A student who talks comfortably may still find the language of mathematics, scientific explanation, historical argument or formal exams hard going. Academic language asks for technical vocabulary, complicated grammar, abstract argument and the particular writing conventions of each subject.
A system that labels a student English-speaking just because English is all around them may miss the help that student needs. The opposite mistake happens too: a multilingual student can be treated as weak in English despite being fluent and academically strong. Labelling as such isn’t the problem. Whether the label fits the student’s actual situation is.
Australian funding shows why this matters. The Australian Government’s Schooling Resource Standard includes a loading for students with low English-language proficiency, subject to eligibility rules and calculated from administrative data. Those rules have changed over the years, so any claim about a particular group’s eligibility has to be checked against the rules that applied at the time.
The way that loading is designed shows how much rides on a definition. A student attracts it only if they come from a language background other than English and at least one parent finished school at Year 9 (or equivalent) or below (Australian Government Department of Education, n.d.). The formula never measures the student’s English. A parent’s schooling stands in for the child’s proficiency.
It isn’t hard to imagine how this plays out. A newly arrived student with very little English, whose parents both finished secondary school, adds nothing to the loading. A fluent bilingual student whose parent left school early does. A review by the NSW Department of Education noted that the measure is widely acknowledged to be a poor approximation of limited English proficiency, and that it was adopted because it was the only measure available nationally (Centre for Education Statistics and Evaluation, n.d.).
The scale says something too. The loading is worth an estimated $74.1 million in 2026, about 0.2 per cent of Commonwealth recurrent school funding. New South Wales does things differently for its own English-proficiency funding, using each student’s assessed phase of English, time in Australian schools and residency status. One system counts a proxy. The other at least tries to describe the child.
Tanzania offers an African example of the same thing, working through the language of instruction instead of a funding formula. Under the Education and Training Policy (2014, 2023 edition), Kiswahili is the medium of instruction in pre-primary and primary school, and English takes over from lower secondary, with exceptions for schools allowed to choose otherwise (Daily News, 2023). For years the system rested on an assumption: that children who had studied English as a subject from Standard 3 to Standard 7 were ready to learn everything else through it. The 2023 reforms cut primary school to six years and introduce English as a subject from Standard One (The Guardian, 2025), but the switch to English-medium secondary school is still there.
The research doesn’t support that assumption. Testing students as they made the move, Qorro (2010) found that new secondary students weren’t competent enough in English to study the whole curriculum through it, even though they wrote reasonably well in Kiswahili. A study drawing on 383 questionnaire responses and focus groups found that students arrived at secondary school with high hopes that were then worn down by bullying and punishment, and teachers saw students losing hope as one reason for truancy (Joyce-Gibbons et al., 2018).
Here the label isn’t a funding category. It’s an assumed ability. A child who is fluent and literate in Kiswahili is treated, from the first day of Form 1, as someone who ought to think in English already. These students don’t lack language. The system simply treats one of their languages as the language of serious learning and the other as a stepping stone towards it.
In the Philippines, the official standing of a child’s language changed almost overnight. From 2013, the Enhanced Basic Education Act required instruction, teaching materials and assessment from kindergarten to Grade 3 to be in the learners’ regional or native language (Philippine Daily Inquirer, 2024). In October 2024, Republic Act No. 12027 ended that requirement. Teaching reverts to Filipino and English, with regional languages in a supporting role (Republic Act No. 12027, 2024).
The mother tongue survives as an option, but only in monolingual classes, and only where the language meets conditions such as having an official orthography published by the Komisyon sa Wikang Filipino. Whether a child can now learn in their home language depends on two classifications: whether their class counts as monolingual, and whether their language has the paperwork the state recognises. Supporters said many schools weren’t equipped to teach in the mother tongue well. Critics called the change a step backwards. Either way, the language a child brings to school was the same before and after 2024. What changed was the box it goes in.
Peru, in Latin America, attaches the label to the school rather than the child. Intercultural Bilingual Education (EIB) is mandatory policy for the country’s 55 Indigenous peoples, and some 26,397 schools operate under it, serving about 1.24 million children and adolescents. A school’s place on the National Registry of EIB Institutions is what entitles it to bilingual teachers, materials in Indigenous languages and dedicated funding (Hidalgo Collazos & Ordoñez Hidalgo, 2025).
That made the registry itself a target. In 2022 the Ministry of Education asked regional directorates for lists of schools to be “exempted” from EIB. Once off the registry, those schools would no longer need bilingual teachers. Reporting at the time linked the request to a shortage of bilingual teachers, which reclassification would have let schools fill with Spanish-only staff (Jara, 2022). After protests from Indigenous organisations and civil society, the Ministry dropped the plan.
The shortage is real. Of 72,701 EIB teaching positions, only 52,253, or 72 per cent, are held by teachers formally trained in EIB, and around 11,736 of the rest are Spanish monolinguals who don’t speak their students’ language. The label promises a bilingual education the system can’t always provide. Removing the label would have made the shortfall disappear from the records without changing a single classroom.
The general principle holds on every continent: when resources depend on classification, the quality of the classification has material consequences. A label affects what a school gets, what a school gets affects the help a child receives, and that help affects the child’s chances. A category that looks purely descriptive ends up helping to produce social outcomes.
This is where the politics of language runs into the philosophy of knowledge. Institutions don’t just observe children and write down neutral facts. They define categories, set thresholds, collect data and make decisions on that basis. Often they have to. The danger comes when a category is taken as the whole picture of a complicated human being.
These cases raise an awkward question: who decides which linguistic characteristics count, and whose experience gets lost when the categories are applied? It bites harder in societies shaped by colonial history, where the language of government and schooling isn’t every community’s first language.
A fair education system keeps the language of administration separate from the languages in which children live their family life, culture and everyday knowledge. Being multilingual shouldn’t be read as a deficiency. And English proficiency shouldn’t stand in for intelligence, potential, or a student’s whole linguistic ability.
Assessment exists to find out what a learner can do, what they’re still developing and what support they need. Labels should serve that purpose, not replace it.
9. Language Does Not Contain Everything That Exists
There’s another reason not to reduce reality to language. We meet the world through our senses, through movement, emotion and practical action, and much of that needs no words, at least at first.
A baby reacts to discomfort long before learning the word pain. We recognise a familiar face without being able to say which features gave it away. A musician can play a passage beautifully and still not be able to explain how. None of this shows that experience is free of concepts. It shows that putting something into words isn’t a precondition for every kind of perception, discrimination or skill. Research on embodied cognition looks at how our bodily engagement with the world contributes to thinking, though how much it contributes, and how, is still argued over.
So it’s hard to defend the assumption that anything we experience must first be represented in language. We recognise things we can’t explain. We notice relationships before we have names for them. We learn patterns through practice long before we can state the rules.
Language makes experience shareable, lets us reflect on it, and lets knowledge build up across generations. It doesn’t follow that language exhausts what we can experience. And failing to express something precisely doesn’t make it unknowable. Sometimes our vocabulary falls short. Sometimes our concepts are still crude. Sometimes the thing itself is just hard to observe. Those are different limitations, and it helps to keep them apart.
10. Science as an Abbreviation of Reality
Calling science an abbreviation of reality is a way of admitting the gap between the world and the models we use to study it. A model picks out some features of a phenomenon and represents them in a form that lets us explain, predict, compare or intervene.
A map doesn’t reproduce every feature of the ground it covers. A model of planetary motion doesn’t contain any planets. An equation doesn’t have to look anything like the process it describes in order to capture the relationships that matter.
Abstraction isn’t a flaw in science. It’s what makes systematic inquiry possible in the first place. Trouble starts when a useful abstraction gets mistaken for a complete account of what exists. A model can explain a whole range of observations and still leave much of the phenomenon out of reach. A theory can be enormously successful without answering every philosophical question about reality.
Scientific language is selective on purpose. It names variables, defines concepts, specifies relationships and leaves out whatever doesn’t bear on the question. Its strength lies in making complicated things manageable, but the selection has to stay visible. Theories need operational definitions, measurements need instruments and procedures, and explanations need evidence that sets them apart from the alternatives. Without those, technical vocabulary can look precise while delivering very little.
Philosophy is open to the same temptation. A striking phrase isn’t a sound argument, and a memorable metaphor isn’t evidence. A concept that seems to explain everything may be so broad that it explains almost nothing. An abstraction is worth exactly as much as it helps us understand.
11. Art as an Intensification of Reality
If science abbreviates reality by selecting and formalising, art intensifies it. A painting doesn’t need to reproduce the whole world to show us something about it. A poem doesn’t need to cover every dimension of grief to make grief newly understandable.
Art selects, transforms, sets things side by side, exaggerates and rearranges. It can suddenly make some overlooked corner of ordinary life visible. It can hold together feelings that resist being expressed one at a time. It can let us inhabit someone else’s point of view. In that sense art concentrates attention on what would otherwise slip past.
Poetry goes further, because it changes the relationships between words. A metaphor lets one area of experience light up another. Rhythm sets up expectations and then breaks them. Ambiguity lets several readings live together. An image can carry an emotional or conceptual connection without flattening it into a single literal statement.
That’s why I think of poetry as a chemical catalyst. A catalyst changes the conditions under which a reaction happens without becoming part of what the reaction produces. A poem changes the reader’s relationship to an experience in a similar way, without handing over a full explanation of it. Familiar words in an unfamiliar arrangement can make us see what we had stopped noticing.
Consider this sentence:
We do not stand outside reality to describe it; we describe it from within.
It states no new law of nature. What it does is rearrange a relationship between ideas, nudging the reader to reconsider where the observer stands in relation to the observed. That’s one of the things poetic and philosophical language can do: it reorganises attention.
Art doesn’t give us more accurate knowledge than science, though. The two have different methods and different aims. A poem may capture what an experience feels like and say nothing reliable about its causes. A scientific model may explain a process and miss what it’s like to live through it. These aren’t competing achievements. Understanding needs many kinds of representation, and each should be judged by what it can honestly reveal.
12. Can Contradiction Contain Truth?
The claim that reality depends on language, and that language depends on reality, sounds like a contradiction. One half stresses the world’s independence from how we represent it. The other stresses language’s role in making experience meaningful.
It’s tempting to say both are true and leave it there. But a contradiction doesn’t establish anything. Two claims that really are incompatible can’t both hold in exactly the same sense and under exactly the same conditions. It’s more useful to pull the claims apart:
- The existence of physical processes does not appear to depend on our naming them.
- Our knowledge of those processes depends on perception, concepts, evidence and representation.
- Language influences which distinctions we habitually draw and how we communicate what we understand.
- Institutions use linguistic categories to sort people, allocate resources and shape opportunity.
These four claims are compatible because they’re about different relationships. The physical world can exist independently of language while our interpretations of it remain shaped by language and culture. Language can shape understanding without creating what it describes. Institutions can build categories with real consequences without creating the characteristics those categories try to capture.
That’s a more defensible kind of pluralism than simply embracing a contradiction. The work lies in identifying which aspect of reality each claim is about. A lot of the apparent conflict comes from using one word, reality, for three different things: the world that exists, the world as we experience it, and the world as we represent it. Separating them doesn’t solve every difficulty, but it clears away a confusion we don’t need. The relationship between language and reality can be genuinely complex without being incoherent.
13. Signals Before Speech: Language on a Natural Continuum
If language is part of natural reality, it ought to have relatives. It does. Long before anyone spoke, living things were exchanging signals: bacteria through chemical cues, bees through dances that show where food can be found, primates through calls and gestures. Whether nature communicates isn’t really in doubt. What’s worth asking is where human language sits along that range, and what about it is genuinely new.
The history of that question is a lesson in itself. In 1980, Seyfarth, Cheney and Marler reported that vervet monkeys give acoustically distinct alarm calls for different predators, and that each call sends listeners into a different escape. The calls seemed to work like words, and the study became the textbook case of semantic communication in animals. Four decades on, its strongest claims have been substantially revised, including by the authors themselves, and evidence for symbolism or syntax in the natural communication of other species remains thin (Rendall, 2021).
That history pulls in two directions. We’re inclined to find our own language in nature, because language is the lens we look through. But the alarm calls still did something real: they carried information that changed what other monkeys did. Perhaps human language is best seen as an extraordinary intensification of something nature does all the time, letting one event stand for another.
14. The Law of Abbreviation: When the Metaphor Turns Literal
I began this essay by saying that language abbreviates reality. It turns out that abbreviation is more than a figure of speech. The linguist George Zipf noticed that the more often a word is used, the shorter it tends to be. Compare the, a and is with photosynthesis. This law of abbreviation has since turned up in the vocal and non-vocal behaviour of other animals.
Ferrer-i-Cancho and colleagues showed that the pattern follows from compression, the principle in information theory of keeping the expected length of a code as short as possible. They argued that compression is a general principle of animal behaviour, reflecting selection for efficient coding (Ferrer-i-Cancho et al., 2013).
If they’re right, language abbreviates reality in two senses. In the philosophical sense, it selects and leaves things out. In a physical sense, its very form is shaped by the same pressure towards economy that shapes signals across the living world. The shortness of our commonest words may owe less to cultural accident than to that economy. Language is about nature, and its shape carries nature’s fingerprints too.
15. Languages as Ecosystems
Languages aren’t only about environments. They’re spread across them in much the way living things are. Gorenflo and colleagues mapped more than 6,900 languages against the world’s biodiversity hotspots, regions rich in species found nowhere else and under heavy threat. Those regions are home to people speaking 3,202 languages, nearly half of all the languages on Earth, and many of those languages, like the species around them, exist in only one place. The reasons for the overlap differ from place to place, but the authors argue that the strong geographic match points to some kind of shared relationship (Gorenflo et al., 2012).
That suggests another way to read this essay’s title. If language is a subset of natural reality, then linguistic diversity may be a kind of biodiversity, one of the ways human communities adapt to particular places. A language spoken in one valley carries names for its plants, its seasons and its soils, distinctions sharpened over generations of living there. When that language dies, a description of an ecosystem, written from inside it, dies too.
Looked at this way, the classification problems in Section 8 aren’t only about fairness. Tanzania alone has more than 130 languages (Penn State, 2012), and its schools teach in two. When a school system treats a local language as a stage to get through, whether in Tanzania, the Philippines or Peru, it is also deciding which descriptions of the world get carried forward and which are left to fade.
16. Language as a Constructed Niche
Evolutionary biology offers another way in. Organisms don’t just adapt to their environments. They also change them. Beavers dam rivers, earthworms rework soil, and the altered environment then shapes the pressures on later generations. Laland, Odling-Smee and Feldman call this niche construction, and they argue that culture amplifies the human capacity for it to the point where it raises new questions about how humans adapt (Laland et al., 2000).
Language may be the most powerful niche humans have ever built. Every child is born into surroundings already full of words, categories and grammatical habits laid down by earlier generations, and adapts to them as surely as to climate or diet. Those surroundings include institutions: funding formulas, language-of-instruction laws, registries of schools. An administrative category works like a dam in the linguistic river. It redirects resources, attention and opportunity, and the next generation grows up in the landscape it leaves behind.
Put this way, the essay’s political argument becomes an argument about nature. Classifications aren’t unreal just because people made them. They’re as real as a beaver’s dam: built, contingent, and able to reshape the lives that grow up around them. Which is exactly why they need to be built carefully.
17. Language Without a Body?
One last frontier has only opened recently. Large language models now produce fluent text without bodies, childhoods or forms of life in Wittgenstein’s sense. They learn from huge amounts of human writing and produce language that people find meaningful. What follows is speculation rather than settled knowledge, and it points in several directions at once.
For Chomsky’s tradition, these systems raise the question of how much linguistic structure can be pulled out of exposure alone, although they’re trained on far more text than any child ever hears, so the comparison isn’t straightforward. For Wittgenstein’s, they raise the question of whether meaning can survive cut off from the practices that produced it, or whether these systems borrow all their meaning from the human activity recorded in what they were trained on. For the embodiment research in Section 9, they offer something new: words without the perception and practical experience that, in humans, usually come first.
From where this essay stands, one thing seems clear. These systems aren’t outside natural reality either. They run on physical machines, they’re built by people, and they learn from the accumulated record of human speech and writing. If language is a subset of natural reality, machine language is a new branch of it, an abbreviation of an abbreviation. Whether it can also intensify, or catalyse, the way art and poetry can, is a question we’re only just learning how to ask.
Conclusion: The World Exceeds Its Description
Language is one of the most remarkable things human beings have produced. It lets us remember beyond a single lifetime, share experiences other people have never had, form abstract ideas, keep knowledge, build scientific theories and imagine what doesn’t exist yet. Even so, it is a human capacity working inside a world bigger than any description of it.
Chomsky pushes us to investigate the biological and mental conditions that make language learnable. Wittgenstein reminds us that meaning grows out of the many things people do, and can’t be reduced to naming. Research across languages puts our assumptions about grammatical universals to the test, and work on linguistic relativity asks how language shapes attention and thought. None of these cancels the others out.
Newer lines of research take things further. Communication runs all through living nature. The economy that shortens our commonest words shapes animal signals too. Languages are spread across the planet like species, and they disappear alongside them. The categories we build become the environments the next generation grows up in. Even language produced by machines stays, on this view, part of the natural world rather than an escape from it.
That leaves three questions. A scientific one: which claims about language does the evidence actually support? A philosophical one: how do language, consciousness and representation relate to the world we live in? And a political one: what happens when institutions treat linguistic categories as complete descriptions of people?
The last may be the most pressing. Label a child English-speaking, and the label can hide the differences between chatting fluently, reading and writing academically, being multilingual, and needing help. Treat one language as the standard for all the others, and difference starts to look like deficiency. Let administrative categories decide who gets support, and the consequences go far beyond words.
That doesn’t mean language creates every reality. It means language helps create social realities while still answering to facts it doesn’t control.
Perhaps the deepest distinction is the one between the world and what we claim about it. The world pushes back against our descriptions. Evidence shows us where we went wrong. Other languages point out distinctions we missed. Other people challenge the categories we use to see them. And sometimes a poem shows us what our usual descriptions had hidden.
So we shouldn’t worship language as the maker of everything, and we shouldn’t dismiss it as a passive set of labels either. It is the natural human capacity through which reality becomes something we can describe, argue about and, in many ways, understand. What it gives us is always partial and open to revision. It doesn’t contain reality. It opens ways of understanding it, and now and then it hides the very things it helps us see.
Thought has to keep both of those possibilities in view. Science has to keep testing our descriptions, and philosophy has to keep questioning their assumptions. The job of imagination is to find out what becomes visible when we learn to see past the edges of our familiar words.
References
Australian Government Department of Education. (n.d.). Schooling Resource Standard. Retrieved October 27, 2026, from https://www.education.gov.au/recurrent-funding-schools/schooling-resource-standard
Centre for Education Statistics and Evaluation. (n.d.). Improvements and alternatives to the disadvantaged Language Background Other Than English (LBOTE) measure. NSW Department of Education. https://education.nsw.gov.au/about-us/education-data-and-research/cese/publications/research-reports/improvements-and-alternatives-to-the-disadvantaged-lbote-measure.html
Chomsky, N. (1965). Aspects of the Theory of Syntax. MIT Press. https://mitpressbookstore.mit.edu/book/9780262527408
The Conversation. (2014). Pacific Islanders and education: Is Australia an ‘unlucky country’? The Conversation. https://theconversation.com/pacific-islanders-and-education-is-australia-an-unlucky-country-21920
Cook Islands News. (n.d.). [Article on Māori-medium instruction at Apii Mauke]. https://www.cookislandsnews.com/?p=281177
Daily News. (2023). Year ender. 2023: New dawn for country’s education. Daily News (Tanzania). https://dailynews.co.tz/year-ender-2023-new-dawn-for-countrys-education/
Endangered Languages Project. (n.d.). Southern Cook Islands Māori. https://endangeredlanguages.com/lang/2519?hl=en
Evans, N., & Levinson, S. C. (2009). The myth of language universals: Language diversity and its importance for cognitive science. Behavioral and Brain Sciences, 32(5), 429–448. https://doi.org/10.1017/S0140525X0999094X
Ferrer-i-Cancho, R., Hernández-Fernández, A., Lusseau, D., Agoramoorthy, G., Hsu, M. J., & Semple, S. (2013). Compression as a universal principle of animal behavior. Cognitive Science, 37(8), 1565–1578. https://doi.org/10.1111/cogs.12061
Gorenflo, L. J., Romaine, S., Mittermeier, R. A., & Walker-Painemilla, K. (2012). Co-occurrence of linguistic and biological diversity in biodiversity hotspots and high biodiversity wilderness areas. Proceedings of the National Academy of Sciences, 109(21), 8032–8037. https://doi.org/10.1073/pnas.1117511109
The Guardian. (2025, May 21). Iringa trains primary schools English teachers on new teaching programme. IPP Media. https://ippmedia.co.tz/the-guardian/news/local-news/read/iringa-trains-primary-schools-english-teachers-on-new-teaching-programme-2025-05-21-202155
Hidalgo Collazos, L., & Ordoñez Hidalgo, Á. (2025, March 1). The challenge of ensuring the right to education for Indigenous peoples: Intercultural bilingual education in Peru. Debates Indígenas. https://debatesindigenas.org/en/2025/03/01/the-challenge-of-ensuring-the-right-to-education-for-indigenous-peoples-intercultural-bilingual-education-in-peru/
Jara, M. (2022, June 16). Bilingual intercultural education, an endangered Indigenous right in Peru. Inter Press Service. https://www.ipsnews.net/2022/06/bilingual-intercultural-education-endangered-indigenous-right-peru/
Joyce-Gibbons, A., Galloway, D., Mollel, A., Mgoma, S., Pima, M., & Deogratias, E. (2018). Successful transition to secondary school in Tanzania: What are the barriers? Journal of International Development, 30(7), 1142–1165. https://doi.org/10.1002/jid.3304
Laland, K. N., Odling-Smee, J., & Feldman, M. W. (2000). Niche construction, biological evolution, and cultural change. Behavioral and Brain Sciences, 23(1), 131–146. https://doi.org/10.1017/S0140525X00002417
Lucy, J. A. (1997). Linguistic relativity. Annual Review of Anthropology, 26, 291–312. https://doi.org/10.1146/annurev.anthro.26.1.291
NSW Department of Education. (n.d.). New Arrivals Program. Retrieved October 27, 2026, from https://education.nsw.gov.au/teaching-and-learning/multicultural-education/english-as-an-additional-language-or-dialect/new-arrivals-program
Penn State. (2012, May 7). Endangered species, languages linked at high biodiversity regions. ScienceDaily. https://www.sciencedaily.com/releases/2012/05/120507154112.htm
Philippine Daily Inquirer. (2024, October 12). It’s back to Filipino and English, from kinder to Grade 3 classes. https://newsinfo.inquirer.net/1992165/its-back-to-filipino-and-english-from-kinder-to-grade-3-classes
Qorro, M. (2010). Testing students’ ability to learn through English during the transition from primary to secondary schooling. In B. Brock-Utne, Z. Desai, M. Qorro, & A. Pitman (Eds.), Language of instruction in Tanzania and South Africa: Highlights from a project (pp. 157–188). Sense Publishers. http://hdl.handle.net/20.500.11810/5337
Rendall, D. (2021). Aping language: Historical perspectives on the quest for semantics, syntax, and other rarefied properties of human language in the communication of primates and other animals. Frontiers in Psychology, 12, 675172. https://doi.org/10.3389/fpsyg.2021.675172
Republic Act No. 12027. (2024). An Act discontinuing the use of the mother tongue as medium of instruction from kindergarten to Grade 3. Republic of the Philippines. https://lawphil.net/statutes/repacts/ra2024/ra_12027_2024.html
Seyfarth, R. M., Cheney, D. L., & Marler, P. (1980). Monkey responses to three different alarm calls: Evidence of predator classification and semantic communication. Science, 210(4471), 801–803. https://doi.org/10.1126/science.7433999
Wittgenstein, L. (1922). Tractatus Logico-Philosophicus (C. K. Ogden, Trans.). Routledge & Kegan Paul. (Original work published 1921) https://www.gutenberg.org/ebooks/5740
Wittgenstein, L. (1953). Philosophical Investigations (G. E. M. Anscombe, Trans.). Blackwell. https://catalogue.bnf.fr/ark:/12148/cb31723790k
Wolff, P., & Holmes, K. J. (2011). Linguistic relativity. Wiley Interdisciplinary Reviews: Cognitive Science, 2(3), 253–265. https://doi.org/10.1002/wcs.104
Thank you for reading!