In 1979, a young Nicaraguan girl named Isela arrived at the newly opened Melania Morales Special Education School in Managua. She could not speak. She had grown up largely isolated, without a shared language, communicating through home gestures she and her family had invented between them. The school gathered dozens of deaf children together for the first time in Nicaragua's history. With no shared language, no teachers who knew sign language, and no model to imitate, the children began doing something extraordinary. They invented their own.
By the 1980s, linguist Judy Kegl, working with Ann Senghas and their colleagues, was documenting what the children had created. The first cohort — those who arrived as older children or adolescents — had developed a pidgin-like gestural system: communicative, expressive, but grammatically simple. Then a second generation of younger children joined the school and learned from the first cohort. These younger children, as they acquired the language from the older ones, did something that their teachers had not done and that the first cohort had not consistently done: they added grammatical structure. They added recursion — the ability to embed one clause inside another that gives language its expressive power. They added verb agreement and spatial grammar. They added the features that distinguish a full language from a pidgin. Nicaraguan Sign Language had emerged spontaneously from children's interaction with each other. It remains the only documented case of a complete language being created from nothing within a single generation.
The case is astonishing for what it reveals. The children were not taught grammar. No adult modeled the grammatical structures they created. The structures came from the children themselves — specifically from the younger children who were still in the optimal developmental window for language acquisition. Whatever the capacity for language is, it is not merely imitation. Something in the human developmental system takes linguistic raw material and restructures it, adding the abstract architecture of grammar whether or not that architecture is present in the input.
"Language is not a cultural artifact that we learn the way we learn to tell time or how the federal government works. Instead, it is a distinct piece of the biological makeup of our brains." — Steven Pinker, The Language Instinct (1994)
Key Definitions
Language acquisition device (LAD) — Noam Chomsky's hypothesized innate mental structure that enables children to acquire language. The LAD is posited to contain Universal Grammar: the abstract principles shared by all human languages.
Universal grammar (UG) — In Chomsky's theory, the set of structural principles common to all human languages that constrain the possible forms a grammar can take. Children are hypothesized to be born knowing these principles.
Critical period hypothesis — The hypothesis, first formulated by Eric Lenneberg in 1967, that there is a biologically determined window of development (roughly birth to puberty) during which the brain is maximally sensitive to language input for full acquisition.
Poverty of the stimulus — The argument that the linguistic input children receive is insufficient to account for the grammatical knowledge they acquire, implying that some grammatical knowledge must be innate.
Statistical learning — The ability to extract patterns from experience by tracking statistical regularities — how often items co-occur, follow each other, and cluster. Demonstrated in 8-month-olds by Saffran, Aslin, and Newport (1996).
Joint attention — The shared focusing of two individuals on the same object or event, mediated by gaze-following and pointing. Tomasello argues joint attention is the foundation of word learning and social cognition.
Word learning constraints — Cognitive biases that help children narrow down possible word meanings: the whole-object assumption, the mutual exclusivity principle, and the taxonomic assumption.
Mutual exclusivity principle — The assumption that each object has only one name, which helps children assign new words to unnamed objects rather than to objects they already have words for.
Motherese / child-directed speech — The simplified, high-pitched, repetitive speech register that adults in most cultures use when addressing infants, which may facilitate phonological and lexical development.
Overgeneralization — The application of a grammatical rule to forms that are exceptions, such as "goed" for "went" or "mouses" for "mice," providing evidence that children are learning rules rather than memorizing individual forms.
Bilingualism — The ability to use two languages with some degree of proficiency. Simultaneous bilingualism refers to acquiring two languages from birth; sequential bilingualism refers to acquiring a second language after the first is established.
Language acquisition vs language learning — In Krashen's influential distinction, acquisition is unconscious, implicit, and resembles first language development; learning is conscious, explicit, and involves knowing about language rules. The two processes are hypothesized to be distinct.
The Nativist Position
The dominant framework for thinking about language acquisition from the 1960s through the 1980s was the nativist theory developed by Noam Chomsky. Chomsky's 1959 review of B.F. Skinner's "Verbal Behavior" was a landmark in the history of cognitive science: it demolished the behaviorist account of language as a conditioned response system and set the agenda for decades of research. Skinner had argued that children learn language through reinforcement — they produce vocalizations, some are rewarded by caregivers, and the rewarded ones are retained and refined. Chomsky showed that this account was hopelessly inadequate. The speed, accuracy, and creativity of language acquisition cannot be explained by reinforcement schedules. Children produce sentences they have never heard and cannot have been reinforced for. They acquire abstract grammatical rules, not just sequences of words.
Chomsky's alternative was the nativist account. The central argument is what he called the "poverty of the stimulus." Children acquire rich, detailed grammatical knowledge — knowledge of complex rules that apply to abstract sentence structures they may rarely or never encounter explicitly. They do not receive explicit instruction in these rules. The input they receive is impoverished relative to the grammatical knowledge they end up with: it is incomplete, contains errors, and provides few, if any, direct examples of the abstract principles at stake. Conclusion: children must bring some grammatical knowledge to the task of language acquisition — they must be born with it.
Chomsky hypothesized a Language Acquisition Device: an innate mental structure dedicated to language acquisition that constrains the possible grammars a child will consider and allows the child to converge on the correct grammar of their language on the basis of limited exposure. The LAD contains Universal Grammar — the principles shared by all human languages, which linguists have worked to identify. All human languages have nouns and verbs. All have ways of asking questions. All have recursive sentence structure (you can always add another clause). All have rules that refer to abstract structural properties of sentences, not just linear sequences of words.
Steven Pinker's 1994 "The Language Instinct" popularized these ideas for a general audience, arguing that language is a biological adaptation as natural as a bird's wing — not a cultural invention but a species-specific capacity shaped by natural selection.
The Critical Period Hypothesis
Eric Lenneberg's "Biological Foundations of Language" (1967) formulated the critical period hypothesis in its classic form. Drawing on evidence from brain injury studies, cases of late language acquisition, and the biology of brain development, Lenneberg argued that language acquisition is tied to a window of developmental plasticity that closes around puberty. Before puberty, the brain is reorganizing and lateralizing — the left hemisphere is becoming specialized for language — and this plasticity allows language to be acquired relatively automatically from exposure. After puberty, the window closes, and full acquisition is no longer automatic.
The most dramatic evidence — and the most difficult to interpret — comes from cases of children who were deprived of language exposure during the critical period. "Genie" is the pseudonym of a girl discovered in Los Angeles in November 1970 at the age of 13. She had been kept in near-total isolation by her parents, strapped to a potty chair during the day and a crib at night, physically punished for making sounds, and exposed to almost no language. After her discovery, she received intensive language training and therapeutic care from a team of researchers including psycholinguist Susan Curtiss. She made remarkable progress: she learned vocabulary rapidly and could communicate effectively. But she never acquired the grammatical rules of English — she could not reliably form questions, negate sentences, or use relative clauses. Curtiss's 1977 monograph documented this in detail.
The Genie case is consistent with the critical period hypothesis but not decisive proof of it, because she had also experienced severe trauma, malnutrition, and social deprivation. A cleaner test comes from immigrant studies. Jacqueline Johnson and Elissa Newport's 1989 study, published in Cognitive Psychology (doi: 10.1016/0010-0277(89)90044-6), is among the most cited in the field. They tested 46 Chinese and Korean immigrants who had moved to the United States at ages ranging from 3 to 39 and had been living in the country for an average of 10 years. All had had extensive exposure to English; the question was whether age of arrival predicted grammatical proficiency. The results were striking: immigrants who arrived before age 7 performed indistinguishably from native speakers on a grammaticality judgment task. Performance declined steadily for those who arrived between ages 7 and 15. Those who arrived after 17 showed the lowest performance and the most individual variation. Newport and Johnson interpreted this as evidence for a critical period for syntax, declining gradually from age 7 and largely closed by late adolescence.
The critical period is not all-or-nothing, and it differs for different aspects of language. The evidence for a critical period for phonology — accent-free pronunciation — is the strongest. Almost no adult acquires a second language without a detectable accent; adults who arrived before age 5 or 6 are much more likely to be taken for native speakers. The evidence for a syntactic critical period is robust but somewhat less absolute: there are documented cases of adults reaching very high levels of grammatical proficiency in a second language. Vocabulary has no such critical period; people learn new words throughout their lives.
Statistical Learning
In 1996, psycholinguist Jenny Saffran, working with Richard Aslin and Elissa Newport, published a study in Science (doi: 10.1126/science.274.5294.1926) that transformed thinking about language acquisition. The study asked how infants identify word boundaries in speech — a non-trivial problem, because the speech stream has no pauses between words; it is a continuous flow of sound, not a series of discrete units separated by spaces like written text.
Saffran and colleagues hypothesized that infants could use statistical regularities in the speech stream to identify words. Within a word, syllables follow each other predictably — the syllable "ba" in the word "baby" is almost always followed by "by." Across word boundaries, the predictability is lower — "by" at the end of "baby" might be followed by many different syllables depending on what word comes next. This difference in transitional probability could, in principle, allow a learner to segment the speech stream into word-like units.
The experiment exposed 8-month-old infants — before they produce words, before they have any English vocabulary — to a continuous, monotone, two-minute stream of synthesized speech containing four three-syllable "words" (such as "bidaku" and "tupiro") repeated in random order with no pauses or prosodic cues. The only information about word boundaries was statistical: syllables within words had transitional probability of 1.0 (they always followed each other); syllables across word boundaries had average transitional probability of 0.33. After just two minutes of exposure, infants listened longer to novel syllable sequences (sequences that crossed word boundaries) than to familiar words, indicating that they had extracted the word structure from the statistics.
The implications are wide-ranging. First, 8-month-olds are performing implicit statistical computations on speech input — tracking how often syllables co-occur across the stream. Second, they do this rapidly, after just two minutes of exposure. Third, this mechanism could in principle allow infants to begin segmenting words from the speech stream before they have semantic knowledge — a prerequisite for word learning. Fourth, the mechanism is not specific to language: similar statistical learning has been demonstrated for musical patterns, visual sequences, and action sequences, suggesting it is a general-purpose learning mechanism.
This finding complicated the nativist picture. If children can extract word structure from statistics, perhaps grammatical knowledge more broadly can be acquired from statistical regularities in input rather than from innate grammatical principles. The debate between nativist and empiricist accounts of language acquisition now turns significantly on whether statistical learning mechanisms are powerful enough to account for the full complexity of grammatical knowledge.
Word Learning
The vocabulary explosion that begins around 18 months is one of the most striking phenomena in child development. At 18 months, the average child knows roughly 50 words. By 24 months, 200-300 words. By 6 years, 10,000 or more. This represents an average rate of nearly 5 new words per day across the preschool years.
The word-learning problem is non-trivial. When a child hears a new word in a situation, the word could in principle refer to any of countless features of that situation: the whole object, a part of it, its color, its action, the type of thing it is, the specific instance, the relationship between objects. Yet children converge on the right meaning — usually the whole object — remarkably quickly and without explicit instruction.
Psycholinguist Susan Carey documented "fast mapping" in a 1978 study: children heard a new word in natural conversation and formed a partial mapping to its meaning after a single exposure, retaining it over days. This initial mapping is rough but sufficient to get started, with the mapping refined over subsequent encounters.
Several cognitive biases help children constrain the space of possible word meanings. The whole-object assumption leads children to interpret new nouns as referring to whole objects rather than to parts or properties. The mutual exclusivity principle (also called the principle of contrast) leads children to assume that different words have different meanings: when they hear a new word, they look for a referent that does not already have a name. The taxonomic assumption leads children to extend a new word to other objects of the same kind rather than to objects that are merely situationally associated.
Michael Tomasello's social-pragmatic approach emphasizes that children are not just matching words to objects but understanding communicative intentions. In a series of elegant studies, Tomasello and colleagues showed that children use speaker gaze, pointing, and inferred intention to determine word meaning. In one paradigm, an adult searched for an object while using a new word; children mapped the word to the object the adult was searching for rather than to a different object that happened to be present. Children as young as 18 months follow gaze and infer intention to solve the word-meaning problem — they are not just learning words but learning what people mean.
How Children Learn Grammar
Children's acquisition of grammar follows predictable sequences that have been documented across dozens of languages. Roger Brown's 1973 longitudinal study of three English-speaking children — Adam, Eve, and Sarah — documented the order in which grammatical morphemes (markers like plural -s, past tense -ed, the progressive -ing, and the articles a and the) are acquired. The order was remarkably consistent across children and was predicted not by the frequency with which morphemes appeared in parents' speech but by their grammatical complexity: simpler forms were acquired before more complex ones.
Overgeneralization errors — forms like "goed" for "went," "mouses" for "mice," "foots" for "feet," or "broked" for "broke" — are among the most telling phenomena in language acquisition. Children who initially produce the correct irregular forms ("went," "broke") begin producing incorrect but rule-governed forms after they have acquired the regular past tense rule. This U-shaped development — correct, then incorrect, then correct again — demonstrates that children are not imitating what they hear. No adult says "goed"; the child could not have heard it. The child has inferred the regular past tense rule and is applying it productively, temporarily overriding the irregular forms they had memorized.
This evidence strongly supports the view that children are learning grammatical rules, not just memorizing sequences. The question is whether the rules are abstracted from input through learning mechanisms or whether they are specified in an innate Universal Grammar. Tomasello's usage-based account, developed in his 2003 "Constructing a Language," argues that children build grammatical abstractions from frequent, concrete linguistic patterns in their input, using general cognitive skills of pattern recognition and analogy. This account does not require innate grammatical structure. Chomskyan accounts argue that the speed and accuracy of grammatical development, particularly for abstract syntactic properties, still requires innate grammatical knowledge to explain.
Bilingualism
The evidence on bilingualism consistently refutes the folk concern that exposing children to two languages produces confusion or delay. Bilingual children do not systematically confuse their two languages in ways that impair communication. Code-switching — alternating between languages within or across utterances — is not confusion but a skilled rhetorical strategy with its own grammatical rules, used by fluent bilinguals to achieve specific communicative effects.
Bilingual children may have slightly smaller vocabularies in each individual language at a given developmental stage, because vocabulary is divided across two languages. But when total conceptual vocabulary is measured — the number of distinct concepts a child has words for across both languages — bilingual children are comparable to or ahead of monolinguals. Bilingual children also show advantages in metalinguistic awareness: they understand earlier that words are arbitrary symbols, that different languages use different words for the same object, and that language itself is a system that can be reflected on.
Ellen Bialystok's research from the early 2000s proposed a broader cognitive advantage for bilinguals. In a series of studies, she found that bilingual children and adults outperformed monolinguals on tasks requiring executive function — the cognitive control systems involved in selective attention, inhibition of distracting information, and task-switching. The hypothesized mechanism was that managing two languages requires constant practice in selecting the intended language and suppressing the other, strengthening executive control more broadly. Her 2006 paper reported that bilingual Alzheimer's patients showed symptom onset approximately 4.1 years later than monolingual patients with comparable degrees of brain pathology.
These findings attracted considerable attention and generated a large follow-up literature. Subsequent studies produced mixed results. Some replicated executive function advantages; others, including several large-scale studies, found no advantage. A 2014 meta-analysis by Paap and colleagues concluded that the bilingual advantage in executive function was not robust. The current consensus is that bilingualism does not harm development, clearly produces advantages in metalinguistic awareness, may produce some executive function benefits under some conditions, and should be supported for its social, cultural, and communicative benefits regardless of specific cognitive advantages.
Second Language Acquisition
Second language acquisition differs from first language acquisition in ways that matter for language education and immigration policy. The most robust difference is the age effect on phonology: adults who begin learning a second language after puberty almost never achieve accent-free pronunciation comparable to native speakers. The mechanism is unclear but appears to involve reduced neural plasticity for phonological categories: the adult brain has already committed to the phonological system of the first language, making it harder to form new categories for the sounds of the second.
For grammar, the story is more complex. Krashen's input hypothesis, influential in the 1980s, proposed that language is acquired through comprehensible input — exposure to language just slightly beyond the learner's current level. Explicit grammatical instruction, on Krashen's view, can help learners monitor their output but does not drive acquisition, which is fundamentally an implicit, unconscious process. The interaction hypothesis, developed by Michael Long and others, emphasized that conversational interaction and negotiation of meaning — the back-and-forth that happens when communication breaks down and participants work to restore it — drives acquisition by providing learners with feedback on their output and modifying input to be more comprehensible.
Research since the 1990s has supported a more nuanced picture: comprehensible input is necessary, interaction accelerates learning, and form-focused instruction helps learners notice and acquire specific grammatical features they might otherwise ignore. Motivation matters enormously in second language acquisition in ways it does not in first language acquisition: adults' success is strongly predicted by their reasons for learning, their identification with the target language community, and their anxiety in using the language. Adults who must use a second language for survival or professional advancement often achieve impressive proficiency; those with only instrumental or classroom motivation often plateau at intermediate levels. Fossilization — the stabilization of grammatical errors that do not improve despite continued exposure — is one of the most puzzling phenomena in second language acquisition and may reflect the limits of post-critical-period language learning.
References
Chomsky, Noam. "A Review of B.F. Skinner's Verbal Behavior." Language, 35(1), 26-58, 1959. https://doi.org/10.2307/411334
Lenneberg, Eric H. Biological Foundations of Language. Wiley, 1967.
Saffran, Jenny R., Richard N. Aslin, and Elissa L. Newport. "Statistical Learning by 8-Month-Old Infants." Science, 274(5294), 1926-1928, 1996. https://doi.org/10.1126/science.274.5294.1926
Johnson, Jacqueline S., and Elissa L. Newport. "Critical Period Effects in Second Language Learning: The Influence of Maturational State on the Acquisition of English as a Second Language." Cognitive Psychology, 21(1), 60-99, 1989. https://doi.org/10.1016/0010-0277(89)90044-6
Tomasello, Michael. Constructing a Language: A Usage-Based Theory of Language Acquisition. Harvard University Press, 2003.
Kegl, Judy, Ann Senghas, and Marie Coppola. "Creation Through Contact: Sign Language Emergence and Sign Language Change in Nicaragua." In M. DeGraff (Ed.), Language Creation and Language Change. MIT Press, 1999.
Brown, Roger. A First Language: The Early Stages. Harvard University Press, 1973.
Carey, Susan. "The Child as Word Learner." In M. Halle, J. Bresnan, and G.A. Miller (Eds.), Linguistic Theory and Psychological Reality. MIT Press, 1978.
Pinker, Steven. The Language Instinct. William Morrow, 1994.
Bialystok, Ellen. "Bilingualism in Development: Language, Literacy, and Cognition." Cambridge University Press, 2001. https://doi.org/10.1017/CBO9780511605963
Frequently Asked Questions
How did Nicaraguan Sign Language prove that children create language spontaneously?
Nicaraguan Sign Language (NSL) is the only documented case of a complete language being created from nothing within a single generation, offering extraordinary evidence about the human capacity for language. Before 1979, deaf people in Nicaragua had no shared language and lived largely isolated lives. When the Nicaraguan government opened special education schools in Managua, deaf children were brought together for the first time. The teachers, who did not know sign language, tried to teach lip-reading and spoken Spanish — with minimal success. But the children began using gestural communication with each other in the schoolyard and on school buses, outside the official curriculum. What linguists Judy Kegl, Ann Senghas, and their colleagues documented from the 1980s onward was remarkable. The first cohort of children (who arrived as older children or adolescents) developed a pidgin-like system of gestures with limited grammatical structure. But the second cohort — younger children who learned from the first cohort — transformed this pidgin into a fully grammatical language. The younger children added spatial grammar, verb agreement, and most importantly, grammatical devices for embedding clauses within clauses (recursion), which had been absent or inconsistent in the first cohort's signing. No teacher introduced these grammatical structures. No adult model provided them. The children created them spontaneously through the process of language transmission and acquisition. This finding supports Chomsky's claim that children have an innate grammatical capacity — they did not just imitate what they saw but added grammatical structure that was not in their input. NSL is now a living language with thousands of users and continues to evolve.
What is the critical period hypothesis and what is the evidence for it?
The critical period hypothesis holds that there is a biologically determined window of development, roughly from birth to puberty, during which the brain is maximally sensitive to language input. After this period, acquiring a language to native-speaker competence becomes significantly harder or impossible. Neurologist Eric Lenneberg first formulated this hypothesis in his 1967 book 'Biological Foundations of Language,' arguing that brain lateralization for language completed around puberty, after which language acquisition was no longer fully automatic. The most dramatic evidence comes from feral and isolated children. Genie, a child discovered in Los Angeles in 1970 at the age of 13 after spending her entire childhood in extreme isolation, could understand and learn individual words but never acquired the grammatical structures of adult language despite years of intensive training and care. Victor, the 'Wild Boy of Aveyron' found in 1800, similarly failed to acquire normal language. These cases are traumatic, confounded by abuse and general deprivation, so their interpretation is uncertain. More controlled evidence comes from immigrant studies. Jacqueline Johnson and Elissa Newport's influential 1989 study published in Cognitive Psychology (doi: 10.1016/0010-0277(89)90044-6) tested 46 Chinese and Korean immigrants who had arrived in the US at different ages. Those who arrived before age 7 performed at native-speaker level on English grammar tests; those who arrived between 8 and 15 showed gradual decline; those who arrived after 17 showed the weakest performance. Importantly, the critical period appears to be different for different aspects of language: phonology (accent-free pronunciation) has a particularly strong early window; syntax (grammar) has a somewhat less absolute one; vocabulary can be learned throughout life.
What is the poverty of the stimulus argument and why is it controversial?
The poverty of the stimulus argument, central to Noam Chomsky's theory of language acquisition, claims that the linguistic input children receive is insufficient to explain the grammatical knowledge they acquire. Children hear incomplete, ungrammatical, and fragmentary speech — yet they acquire grammar that goes well beyond what they have been explicitly exposed to, including abstract structural properties that are never explicitly modeled or taught. Chomsky concluded that children must have innate, language-specific grammatical knowledge — a Language Acquisition Device containing the principles of Universal Grammar — that allows them to go beyond the evidence. The argument is controversial for several reasons. Empirically, corpus studies of child-directed speech have found that it is more grammatically regular than conversational speech, that parents repeat and correct certain structures, and that statistical regularities in input are rich enough for learning mechanisms to extract grammatical patterns. Computationally, machine learning researchers have shown that statistical learning algorithms — neural networks trained on language data — can acquire impressive linguistic competencies, suggesting that innate grammar may not be necessary. Theoretically, Chomsky's critics argue that the universals he identifies as evidence for Universal Grammar may reflect universal features of human cognition, communication, and social interaction rather than language-specific innate structure. Tomasello's usage-based approach, building on cognitive linguistics, argues that grammar is abstracted from patterns of use rather than instantiated from an innate template. The debate remains one of the most contested in cognitive science, with genuine evidence on multiple sides, and its resolution will determine how we understand the relationship between human biology and human culture.
How do infants learn words so rapidly, and what does 'fast mapping' mean?
One of the most striking facts about child language development is the speed of vocabulary acquisition. At 18 months, the average child knows approximately 50 words. By age 2, the vocabulary has expanded to 200-300 words. By age 6, children understand approximately 10,000 words — an average of roughly 5 new words per day through the preschool years. This 'vocabulary explosion' raises the question of how children solve the word-learning problem: a new word could refer to any of countless features of a situation, yet children consistently map new words to the right referents. Psycholinguist Susan Carey, in her 1978 study with preschoolers, coined the term 'fast mapping' to describe children's ability to form an initial, partial mapping between a word and its meaning after a single exposure. In her study, children heard a new word ('chromium,' describing the color olive green) in natural conversation and retained a rough mapping of the word to its meaning a week later, without explicit instruction. Several learning constraints help children narrow down word meanings. The 'whole-object assumption' leads children to interpret new nouns as referring to whole objects rather than parts or properties. The 'mutual exclusivity principle' (or 'principle of contrast') leads children to assume that a new word refers to an object that does not yet have a name — so if they see a dog and a novel object and hear a new word, they assume it refers to the novel object, not the dog they already have a name for. Social cues are also crucial: Michael Tomasello's research showed that children use speaker gaze and intention to determine word meaning, learning words even when an adult uses them to refer to objects the child cannot see, based on inferring what the adult is looking for.
Do bilingual children get confused or fall behind monolingual children?
No. This is one of the most persistent myths about bilingualism, and research consistently refutes it. Bilingual children do not become confused between their two languages in any way that impairs development, and they do not fall behind monolingual peers in language development overall. They may have somewhat smaller vocabularies in each individual language at a given age (though total vocabulary across both languages is comparable or larger than monolinguals), and they may slightly delay the emergence of some linguistic milestones, but these are transient differences that disappear by school age. The appearance of confusion comes from code-switching — switching between languages within a conversation or even within a sentence. But code-switching is not confusion; it is a sophisticated linguistic skill that follows grammatical rules specific to bilingual speech and is used by fluent bilinguals for communicative effect, not because they cannot access the right word in one language. Bilingual children who code-switch are demonstrating linguistic flexibility, not deficit. A more contested area is the 'bilingual advantage' hypothesis, associated primarily with researcher Ellen Bialystok, who published studies from 2004 onward suggesting that lifelong bilingualism enhances executive function — the cognitive control processes involved in attention, inhibition, and task-switching — potentially delaying dementia onset by several years. These findings attracted enormous attention and generated a large follow-up literature. Many subsequent attempts to replicate the executive function advantages found reduced or null effects, and the field is now more cautious. What is clear is that bilingualism does not harm cognitive development, that it produces genuine advantages in metalinguistic awareness (understanding of language as a system), and that maintaining a heritage language has significant social and cultural benefits.
What is statistical learning in language acquisition and why does it matter?
Statistical learning is the ability to extract patterns from experience by tracking how frequently items co-occur, transition, and repeat. In language acquisition, it refers to infants' ability to use statistical regularities in speech input to identify word boundaries, learn sound patterns, and eventually extract grammatical structures. The landmark study demonstrating statistical learning in language was published by Jenny Saffran, Richard Aslin, and Elissa Newport in Science in 1996 (doi: 10.1126/science.274.5294.1926). They exposed 8-month-old infants to 2 minutes of a continuous stream of synthesized speech containing no pauses, punctuation, or other cues to word boundaries — just an unbroken string of syllables from an artificial language with four three-syllable 'words.' The only information available to identify word boundaries was statistical: syllables within words always followed each other (transitional probability = 1.0), while syllables spanning word boundaries followed each other less predictably (average transitional probability = 0.33). After just 2 minutes of exposure, the infants showed longer listening times to novel words than to familiar words in a subsequent test, demonstrating that they had extracted the word structure purely from statistical regularities in the input. This was remarkable for several reasons. First, 8-month-olds — well before they produce words themselves — were performing sophisticated statistical computations on linguistic input. Second, they did it in 2 minutes, suggesting a fast, efficient, implicit learning mechanism. Third, the finding implied that infants could segment the speech stream into word-like units before they have any semantic knowledge — a prerequisite for word learning. Statistical learning has since been documented for musical patterns, visual sequences, and action sequences, suggesting it is a domain-general learning mechanism rather than one specific to language.
How does second language acquisition differ from first language acquisition?
Second language acquisition (SLA) differs from first language acquisition (FLA) in several important ways. Age of acquisition matters more in SLA: while first languages are acquired to native-speaker competence regardless of individual differences (given normal exposure and no neurological impairment), second languages acquired after puberty rarely reach native-like phonological competence — the accent-free pronunciation that characterizes native speakers. The critical period for accent appears to be the strongest and most absolute of the language-specific sensitive periods. Grammar in L2 acquisition is more variable: some adults acquire near-native-like grammatical competence, but many show 'fossilization' — persistent grammatical errors that stabilize and do not improve despite continued exposure and instruction. Stephen Krashen's influential (and controversial) input hypothesis proposed that language is acquired through comprehensible input — exposure to language just beyond the learner's current level ('i+1') — and that explicit grammatical instruction plays a minor role. The competing interaction hypothesis, associated with Michael Long, argues that interaction and feedback — particularly negotiation of meaning when communication breaks down — are essential drivers of acquisition. Both capture some truth: comprehensible input is necessary, and interaction accelerates learning, particularly for grammar. The role of motivation is larger in SLA than in FLA: children acquire their first language universally, but adults' success in second language acquisition varies substantially with instrumental motivation (needing the language for practical goals), integrative motivation (wanting to participate in the target language community), and anxiety. Contrary to common belief, adults are in some respects faster and more efficient language learners than children — they understand instruction, use explicit memory effectively, and have richer world knowledge. The disadvantage is biological: the sensitive period means that high proficiency, especially in pronunciation, is harder to achieve.