Language

Abstract

Language is the human system of communication that pairs structured sound or sign with meaning, the faculty that lets a finite vocabulary and a small set of grammatical rules generate an unbounded range of expressions. This article sets out what distinguishes human language from other communication systems, the levels at which linguists analyse it, and the long debate over whether it is a dedicated biological faculty or a product of general cognition and cultural transmission. It surveys how children acquire language within an early sensitive period, how the brain produces and comprehends it in real time, and how the language a person speaks may shape thought. Three interactive demonstrations let a reader compare the design features of communication systems, trace how age of first exposure bounds ultimate attainment, and see how spatial metaphors for time differ across languages.

Keywords: language, universal grammar, design features, critical period, linguistic relativity

Language is the structured system through which humans encode meaning in sound, sign, or writing and exchange it with one another. In the Medical Subject Headings vocabulary the descriptor names a means of communication through a system of symbols, and that phrasing captures the two properties that matter most: language is symbolic, in that its units stand for things by convention rather than resemblance, and it is systematic, in that those units combine under rules. Ferdinand de Saussure, whose lectures founded modern structural linguistics, drew the first property sharply. A linguistic sign, he argued, joins a signifier (a sound pattern) to a signified (a concept), and the bond between them is arbitrary: there is nothing tree-like about the word tree, and different languages attach quite different signifiers to the same concept (Saussure, 1959). Arbitrariness is what frees language from the here-and-now and lets it name anything a community agrees to name.

Types of Language

MeSH files the descriptor Language beneath Communication and, on a second branch, beneath Information Science, and gives it three narrower descriptors. These subtypes are not mutually exclusive strata of one hierarchy so much as different cuts through the same phenomenon: one names the object of scientific study, one names the skills through which language is taught and used, and one names a modality in which a full language can be realised. They can and do overlap — a linguist studies sign languages, and the language arts are taught about signed as well as spoken tongues — so the table below is an indexing classification for retrieval, not a claim that the three carve language into disjoint kinds.

Design Features of Human Language

If arbitrariness is the first thing that sets language apart, it is not the only one. The linguist Charles Hockett proposed a checklist of design features meant to specify exactly what a communication system must have to count as a human language, and to make precise how animal signalling falls short (Hockett, 1960). Four of the features do most of the work. Semanticity is the use of signals that carry fixed meanings. Arbitrariness, as Saussure had already stressed, is the absence of resemblance between signal and meaning. Displacement is the capacity to refer to things remote in time and space — yesterday's hunt, an imagined future, an abstraction that was never present at all — a power the honeybee's dance approaches only for the location of food and no other topic. Productivity (or openness) is the ability to produce and understand utterances never encountered before, which follows from combining units by rule. A fifth, duality of patterning, is the organisation of language on two levels at once: a small set of meaningless sounds (phonemes) combines into a large set of meaningful units (morphemes and words), which is what lets a few dozen speech sounds yield hundreds of thousands of words. Against this checklist, the vervet monkey's alarm calls have semanticity but almost no displacement or productivity, and birdsong has elaborate patterning but little semanticity — no natural system but human language has all of the features together. Figure 1 makes the most basic feature, arbitrariness, concrete: a single concept is paired with the unrelated sound patterns four languages happen to use for it. The first demonstration then lets a reader select a communication system and see which design features it possesses.

Language as a Human Faculty

The productivity of language raises a question that has organised the field for seventy years: how can a finite mind command an infinite language? Noam Chomsky's answer, which launched the cognitive study of language, is that speakers internalise a generative grammar — a finite system of rules capable of producing and interpreting an unbounded number of sentences, including wholly novel ones. In a review that helped end the dominance of behaviourism, Chomsky argued that the speed, uniformity, and creativity of child language acquisition cannot be explained by reinforcement of imitated responses, because children routinely produce sentences they have never heard and could not have been rewarded for (Chomsky, 1959). The positive proposal is universal grammar: an innate endowment that constrains the form a human grammar can take and lets the child converge on the adult system from limited and imperfect input. Charles Yang and colleagues, restating the case, argue that the child's rapid, error-guided growth of grammar reflects exactly such principled constraints on the space of possible languages, learned from experience but not reducible to it (Yang et al., 2017).

What, precisely, is innate has narrowed over time. Marc Hauser, Chomsky, and Tecumseh Fitch drew an influential distinction between the faculty of language in the broad sense (FLB) — the sensory-motor and conceptual systems that language recruits, much of it shared with other species — and the faculty of language in the narrow sense (FLN), which they conjectured might reduce to a single capacity: recursion, the ability to embed structures within structures of the same kind, yielding the hierarchical, potentially unbounded syntax that other animals lack (Hauser et al., 2002). Not everyone accepts that the faculty is so specialised or so recent. Steven Pinker and Paul Bloom argued that the language faculty shows the hallmarks of a complex biological adaptation — intricate, functional design for communication — and so must have arisen by ordinary Darwinian natural selection rather than as a sudden by-product, placing language firmly within evolutionary biology (Pinker & Bloom, 1990). The disagreement over how much of language is special, and how much of that is unique to our species, remains live.

Acquiring Language

However much is innate, language must still be acquired, and it is acquired on a schedule. Infants begin as universal listeners, able to discriminate the speech-sound contrasts of every language, and then narrow: over the first year, perception retunes toward the contrasts of the ambient language and away from the ones it does not use, so that by twelve months a baby raised on English no longer hears distinctions that a Hindi- or Japanese-learning baby keeps. Patricia Kuhl's work on this perceptual narrowing shows that early experience commits the perceptual system to the native language, and that social interaction, not mere exposure to recordings, drives the learning (Kuhl, 2004). This early plasticity is bounded. Eric Lenneberg proposed that language is acquired most readily within a critical period extending from infancy to roughly puberty, after which the capacity for effortless, native-like acquisition declines as the maturing brain loses plasticity (Lenneberg, 1967). The evidence for a sharp boundary is debated, and later work favours a gradually closing sensitive period over an abrupt cutoff, but the central regularity is robust: the earlier a first or second language is encountered, the higher the ultimate attainment tends to be. The second demonstration traces a schematic attainment curve as a function of the age of first exposure.

Language in the Mind and Brain

Using language means solving hard computational problems fast. In production, a speaker must, in a fraction of a second, select a concept, retrieve the word that expresses it, assemble its sounds, and articulate them. Willem Levelt, Ardi Roelofs, and Antje Meyer's model of lexical access decomposes this into ordered stages — conceptual preparation, selection of an abstract word (the lemma), retrieval of its morphological and phonological form, and articulation — each supported by chronometric evidence about what speakers can and cannot do under time pressure (Levelt et al., 1999). Comprehension faces the mirror problem, and a still tighter constraint. Morten Christiansen and Nick Chater argue that language processing is dominated by a Now-or-Never bottleneck: the sensory trace of speech decays within moments, so the system must recode incoming material into more abstract chunks almost immediately or lose it, and this pressure shapes the structure of language itself toward what can be processed incrementally (Christiansen & Chater, 2016).

Where in the brain these operations live has been substantially revised. The nineteenth-century picture localised production in Broca's area and comprehension in Wernicke's area, connected by a single tract. Angela Friederici's synthesis of the neurobiology replaced this with a distributed network in which frontal and temporal regions of the left hemisphere are linked by distinct dorsal and ventral pathways that mature over development and support syntactic and semantic processing respectively (Friederici, 2011). Peter Hagoort presses the same point for meaning beyond the single word, arguing that comprehension recruits a broad fronto-temporo-parietal network for combining words into messages and integrating them with world knowledge, not a dedicated sentence module (Hagoort, 2019). Evelina Fedorenko and Idan Blank go further, marshalling functional-imaging evidence that Broca's area is not a single functional unit at all but a piece of cortex shared between the language network and domain-general control systems — so that even the field's most iconic language region is not a natural kind (Fedorenko & Blank, 2020).

Language and Thought

Does the language a person speaks shape how they think? The strong claim that language determines thought has few defenders, but a weaker linguistic relativity — that habitual patterns in a language nudge habitual patterns in cognition — has real experimental support. Lera Boroditsky's studies of time are the clearest case. English speakers talk about time mostly horizontally (the deadline is ahead of us; the worst is behind us), whereas Mandarin speakers also use a vertical metaphor (an earlier event is up, a later one down). Boroditsky found that this difference tracks a difference in thought: Mandarin speakers were faster to judge temporal relations after seeing a vertical spatial arrangement, and English speakers after a horizontal one, indicating that the spatial metaphors ingrained in each language shape how its speakers implicitly represent time (Boroditsky, 2001). The effect is a bias, not a cage — bilinguals shift with the language in use, and everyone can think outside their language's default — but it shows that the grammar and vocabulary a person grows up with leave a measurable trace on non-linguistic cognition. The third demonstration lets a reader see how a horizontal or vertical framing speeds a matching temporal judgment.

Worked Example

Consider how the age of first exposure bounds ultimate attainment in a language, following the logic of the critical-period account. Represent ultimate attainment as a score U on a 0–100 scale, where 100 is indistinguishable from a native speaker, and model it as a schematic function of the age of first exposure a in years. Before the sensitive period begins to close, attainment is at ceiling; after it, attainment declines at a roughly constant rate until it reaches a floor. A simple piecewise model captures the shape: U(a) = 100 for a ≤ 7, and U(a) = 100 − 3 × (a − 7) thereafter, not falling below a floor of 30.

Take a child first exposed at age 5. Since 5 ≤ 7, U = 100: full native attainment is expected. Take a learner first exposed at age 15. Here U = 100 − 3 × (15 − 7) = 100 − 3 × 8 = 100 − 24 = 76. Take a late learner first exposed at age 30: U = 100 − 3 × (30 − 7) = 100 − 3 × 23 = 100 − 69 = 31, just above the floor. Now apply a native-like threshold of 85: a learner counts as native-like only if U ≥ 85. The age-5 learner (100) clears it comfortably; the age-15 learner (76) falls short; the age-30 learner (31) is far below. Solving 100 − 3 × (a − 7) = 85 gives 3 × (a − 7) = 15, so a − 7 = 5 and a = 12 — the model's crossover age, beyond which native-like attainment is no longer predicted. The numbers are illustrative, not measured effect sizes, but they reproduce the central regularity: attainment is flat and high in early childhood and then falls with the age of first exposure. The second demonstration lets a reader move the age of exposure and watch U cross the threshold at exactly this point.

Discussion

The study of language sits at the centre of cognitive science because it is the clearest case of a mental capacity that is at once universal across the species, effortlessly acquired, computationally intricate, and plausibly unique to humans. Its defining property is the one Saussure and Hockett first isolated and Chomsky formalised: an arbitrary, dually-patterned symbol system made productive by a generative grammar, so that finite means yield infinite ends (Saussure, 1959; Hockett, 1960; Chomsky, 1959). The enduring disagreement is over the nature of the machinery behind that productivity — whether a narrow, possibly recursion-centred faculty (Hauser et al., 2002) or an evolved adaptation continuous with other cognition (Pinker & Bloom, 1990) — and the question is not settled.

What has changed most is the picture of how language is implemented and acquired. Acquisition is now understood as an experience-driven narrowing within an early sensitive period rather than the flipping of an innate switch (Kuhl, 2004; Lenneberg, 1967), and the brain basis has moved from a two-box, two-region diagram to a distributed, pathway-connected network in which even Broca's area is not a dedicated language organ (Friederici, 2011; Fedorenko & Blank, 2020). Meanwhile the real-time constraints on processing (Christiansen & Chater, 2016) and the measurable back-pressure of language on thought (Boroditsky, 2001) show that language cannot be studied in isolation from memory, perception, and general cognition. The trajectory of the field is from treating language as an autonomous module toward treating it as a deeply integrated, and deeply human, system.

Current Directions

The most active current questions concern how far language depends on machinery specific to it. Fedorenko and Blank's argument that Broca's area is not a natural kind is one front in a broader reassessment of the language network's boundaries and its relation to domain-general control, using the spatial and temporal resolution of modern imaging to ask which computations are genuinely language-specific (Fedorenko & Blank, 2020). Hagoort's programme for meaning beyond the single word pushes in the same direction, treating comprehension as the integration of linguistic input with memory and world knowledge across a wide cortical network (Hagoort, 2019). A second front is theoretical: Christiansen and Chater's Now-or-Never bottleneck reframes many properties of language — its reliance on local structure, its redundancy, its incremental parsability — as consequences of a universal processing constraint rather than of an autonomous grammar, an account that invites direct comparison with the generative tradition (Christiansen & Chater, 2016). The rise of large language models trained only on text has sharpened all of these debates, offering a working system that acquires striking linguistic competence without a built-in universal grammar, a critical period, or a body — and so serving as an unexpected test bed for claims about what language really requires. Whether such models illuminate human language or merely mimic its surface is now among the field's most contested questions.

Common Misconceptions

Language is just a system of naming things.

Naming is only the semantic surface. What makes language language is the combinatorial machinery — a grammar that assembles a finite vocabulary into an unbounded set of structured, novel expressions (Chomsky, 1959).

Animals such as bees and apes have language too.

Animal communication is real but lacks the full set of design features together — especially displacement and open-ended productivity — that define human language (Hockett, 1960).

Adults can learn a second language just as easily as children.

Later learners can become highly proficient, but native-like attainment becomes progressively harder after an early sensitive period, as the evidence on age of acquisition shows (Lenneberg, 1967).

Language lives in Broca's and Wernicke's areas.

Language is supported by a distributed left-hemisphere network linked by multiple pathways; even Broca's area is shared with domain-general systems rather than being a dedicated language organ (Friederici, 2011; Fedorenko & Blank, 2020).

Glossary

Arbitrariness.

The absence of any natural resemblance between a linguistic signal and its meaning; the signifier–signified bond is fixed by convention alone.

Critical period.

A maturational window, extending roughly from infancy to puberty, within which language is acquired most readily; often reframed as a gradually closing sensitive period.

Design features.

Hockett's checklist of properties — semanticity, arbitrariness, displacement, productivity, duality of patterning, and others — that jointly characterise human language.

Displacement.

The capacity to refer to things remote in time or space, or that do not exist at all, freeing language from the immediate situation.

Duality of patterning.

The organisation of language on two levels: meaningless sounds combine into meaningful units, so a small phoneme inventory yields a vast vocabulary.

Faculty of language.

The mental capacity underlying language, divided by Hauser, Chomsky, and Fitch into a broad sense (systems language shares with other cognition) and a narrow sense (what may be unique to it).

Generativity.

The property of a grammar that lets a finite rule system produce and interpret an unbounded number of sentences, including entirely novel ones.

Grammar.

The system of rules and representations governing how a language's units combine, spanning phonology, morphology, and syntax.

Lexical access.

The retrieval of a word's abstract identity and its sound form during speaking or listening, modelled as a sequence of timed processing stages.

Linguistic relativity.

The hypothesis that habitual patterns in a language influence habitual patterns in the thought of its speakers; the weak, evidence-supported version of the Sapir–Whorf idea.

Morphology.

The level of language concerned with the internal structure of words and the units of meaning (morphemes) from which they are built.

Phonology.

The level of language concerned with the sound system: the inventory of speech sounds and the rules that govern how they pattern.

Productivity.

Also called openness; the ability to produce and understand an unlimited number of new utterances by combining units according to rule.

Recursion.

The embedding of a structure within another structure of the same type, yielding the hierarchical, potentially unbounded syntax proposed as the core of the narrow language faculty.

Semantics.

The level of language concerned with meaning: the meanings of words and the way they combine into the meanings of phrases and sentences.

Syntax.

The level of language governing how words combine into phrases and sentences; the locus of the hierarchical structure that distinguishes grammatical from ungrammatical strings.

Universal grammar.

In the generative tradition, the innate endowment that constrains the class of possible human grammars and enables acquisition from limited input.

Key Researchers

Lera Boroditsky (contemporary). Professor of cognitive science at the University of California, San Diego; her experimental work on language and thought, especially conceptions of time and space, is the leading modern evidence for linguistic relativity. Google Scholar - Wikipedia

Noam Chomsky (b. 1928). Institute Professor emeritus at MIT and laureate professor at the University of Arizona; founder of generative grammar and the theory of universal grammar, and the figure most responsible for the cognitive turn in the study of language. Google Scholar - Wikipedia - Faculty page

Angela D. Friederici (b. 1952). Founding director at the Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig; her work mapped the distributed brain network and the dorsal and ventral pathways that support syntactic and semantic processing. ORCID - Google Scholar - Wikipedia

Eric H. Lenneberg (1921-1975). Psychologist and linguist; his Biological Foundations of Language (1967) grounded language in biology and introduced the critical-period hypothesis for its acquisition. Wikipedia - Wikidata

Steven Pinker (b. 1954). Johnstone Family Professor of Psychology at Harvard University; argued that the language faculty is an evolved biological adaptation and wrote widely on language acquisition and its place in the mind. ORCID - Google Scholar - Wikipedia

Ferdinand de Saussure (1857-1913). Swiss linguist whose posthumously published Course in General Linguistics founded structural linguistics and introduced the arbitrary sign as the basic unit of language. Wikipedia - Wikidata

Frequently Asked Questions

What is language?
Language is a structured, symbolic system of communication that pairs arbitrary signals (sounds, signs, or written marks) with meanings and combines them under grammatical rules to produce an unbounded range of expressions (Saussure, 1959).

What makes human language different from animal communication?
Human language uniquely possesses the full set of design features together, above all displacement (reference to the remote and the abstract) and open-ended productivity (endless novel utterances built by rule), which animal signalling systems have only in part (Hockett, 1960).

What is universal grammar?
Universal grammar is the generative tradition's proposal that humans are born with an innate endowment constraining the form a grammar can take, which lets children acquire language rapidly from limited and imperfect input (Chomsky, 1959; Yang et al., 2017).

Is there a critical period for learning language?
Language is acquired most readily within an early sensitive period; native-like attainment becomes progressively harder with age of first exposure, though the boundary is gradual rather than an abrupt cutoff (Lenneberg, 1967; Kuhl, 2004).

Where is language processed in the brain?
Language relies on a distributed left-hemisphere network connecting frontal and temporal regions through distinct dorsal and ventral pathways, not a simple pair of centres; even Broca's area is shared with domain-general systems (Friederici, 2011; Fedorenko & Blank, 2020).

Does a speaker's language shape how they think?
In its weak form, yes: habitual patterns in a language nudge habitual cognition. Boroditsky's studies show, for example, that English and Mandarin speakers differ in how readily spatial framings speed their judgments about time (Boroditsky, 2001).

Why is language hard to process in real time?
The sensory trace of speech decays within moments, so comprehension must recode incoming material into abstract chunks almost immediately or lose it; this Now-or-Never bottleneck shapes the structure of language itself (Christiansen & Chater, 2016).

Did language evolve by natural selection?
Pinker and Bloom argue that the language faculty shows the intricate functional design characteristic of a biological adaptation and so must have evolved gradually by natural selection, rather than appearing as a sudden by-product (Pinker & Bloom, 1990).

References