What is the regularization approach in language learning?
Evidence-based answer · Last updated · How it’s made
The short answer
"Regularization" refers to a well-documented pattern where language learners make an inconsistent, messy language input more consistent than it actually was — essentially cleaning up the "noise" they hear rather than reproducing it exactly. Research consistently finds that young children do this strongly and automatically, while adults mostly reproduce the inconsistencies they're exposed to, only regularizing under specific conditions like very high inconsistency, memory strain, or ambiguity about whether errors matter.
Answered for language learning in general.No specific language was set for this question, so the findings are drawn from general second-language research rather than one language’s literature.
Children regularize far more than adults
When exposed to inconsistent grammatical patterns (such as a word sometimes used one way and sometimes another with no clear reason), young children reliably iron out the inconsistency and produce a more regular, consistent version, while adults tend to faithfully copy the inconsistency they heard, including the messy parts.1,6,7,8,9
Example
In one set of experiments, adults and children were taught made-up mini-languages where a grammatical marker (like 'a' or 'the') was used unpredictably; five- and six-year-olds ended up using the most common form almost all the time, while adults kept mixing forms in roughly the same messy proportions they'd been taught.
Adults do regularize, but only under specific pressure
Adults aren't immune to regularizing — they do it when the inconsistency in what they hear is extremely high, when retrieving the right word or form at the moment of speaking is made harder, or when they don't feel pressure to be 'correct' and assume the inconsistency isn't meaningful.1,3,5,14
Example
In one study, simply making a follow-up test harder (forcing people to recall forms under more difficult conditions) pushed adults to smooth out inconsistencies they had otherwise learned accurately.
Real-world evidence: children outperforming their input
This isn't just a lab phenomenon — a deaf child learning sign language solely from parents who were imperfect, inconsistent late-learners of that language ended up producing the language's grammar more consistently and correctly than his own parents did.18,20
General cognitive overload isn't the main driver
Overloading adults' short-term memory or attention doesn't reliably make them regularize the way children do, suggesting the child/adult difference isn't simply explained by children having weaker general thinking capacity under strain.2,4,10,22
Regularization pressure comes from producing, not just hearing, language
Multiple studies suggest that regularizing behavior emerges specifically when learners have to produce (speak or write) the language under difficult conditions, rather than simply from difficulty absorbing or encoding what they hear.4,11,22
Predictable inconsistency is learned accurately, not smoothed over
Regularization mainly kicks in when the variation in the input seems random and unexplainable; if the same inconsistency instead follows a learnable pattern (tied to a specific word or context), both children and adults tend to learn the actual pattern rather than flattening it out.1,7,8,15,21
Bilingual and second-language learners also show regularizing errors
Bilingual children and second-language learners sometimes overregularize grammar in their non-native or heritage language, applying a regular pattern (like a standard verb ending) to cases that are actually irregular exceptions.16,17
Example
English-Spanish bilingual children learning Spanish verb endings sometimes applied the regular present-tense pattern to irregular verbs, more so than children raised with Spanish as their primary language.
Regularization shows up beyond grammar rules, including pronunciation and word order
The tendency to smooth out inconsistency isn't limited to grammatical markers — it also appears in how learners handle unpredictable word order patterns, sound patterns like vowel harmony, and even non-language tasks involving tracking frequencies, suggesting a broad, general-purpose bias rather than something unique to grammar.10,11,12,13,23
Regularization can shape whole languages over generations, not just individuals
When a made-up language is passed from one learner to another repeatedly (simulating how languages evolve across generations), even a small individual bias toward regularizing can snowball into a fully regular, consistent language over time.13,19,24
What to do with this
For an English-speaking adult learner, these findings mean you should expect to naturally preserve, not automatically 'fix,' the messy exceptions and inconsistent patterns you hear in your target language — unlike a child, your brain doesn't reflexively smooth over irregularity, which is actually useful because it means you can learn genuine exceptions accurately if you pay attention to them. However, the research also shows adults DO start to flatten out inconsistencies when input is extremely erratic, or when they're forced to produce language quickly under time or memory pressure — so if you find yourself defaulting to the 'regular' pattern and dropping irregular forms (like regularizing an irregular verb), it may be a sign you're being pushed to produce language faster than you've solidified it, and slowing down or reviewing irregular forms explicitly (through flashcards or targeted grammar study) may help lock in the real pattern instead of your own simplified version. Since predictable variation (tied to a specific context, word, or situation) is learned much more accurately than random inconsistency, exposure to input where irregular forms are used in consistent, recognizable contexts will help you absorb them correctly rather than overgeneralizing a 'regular' rule to everything.
Now pick something to do it with
Each of these opens an overview of the apps, courses and immersion material worth a look for that — filtered to exactly what the advice above points you at:
Almost all of this evidence comes from laboratory experiments using artificial mini-languages (invented for research purposes) rather than natural second-language classrooms, so caution is needed in applying it directly to, say, learning French or Japanese from textbooks.
Much of the research also focuses on comparing children versus adults learning a first language or a new sign language system, rather than adults learning a second language for communicative purposes; the bilingual/heritage-language studies included here are the closest real-world parallel to adult L2 (second-language) learners specifically.
Findings are also somewhat mixed on exactly why adults sometimes regularize (memory retrieval difficulty, production pressure, or beliefs about whether accuracy matters), so no single tidy explanation exists yet.
[1]Carla L. Hudson Kam, Elissa L. Newport. Getting it right by getting it wrong: When learners change languages. Cognitive Psychology 2009. doi.org/10.1016/j.cogpsych.2009.01.001
[2]Andrew Perfors, Nicholas R. Burns. Adult language learners under cognitive load do not over-regularize like children. Adelaide Research & Scholarship (AR&S) (University of Adelaide) 2010. http://hdl.handle.net/2440/64268
[3]Amy Perfors. Probability matching vs over-regularization in language: participant behavior depends on their interpretation of the task. Adelaide Research & Scholarship (AR&S) (University of Adelaide) 2012. http://hdl.handle.net/2440/77552
[5]Hudson Kam, Carla L.. Reconsidering Retrieval Effects on Adult Regularization of Inconsistent Variation in Language. Language Learning and Development 2019. doi.org/10.1080/15475441.2019.1634575
[6]Austin, Alison C., Schuler, Kathryn D., Furlong, Sarah, Newport, Elissa L.. Learning a Language from Inconsistent Input: Regularization in Child and Adult Learners. Language Learning and Development 2022. doi.org/10.1080/15475441.2021.1954927
[9]Alison Austin, Kathryn Schuler, Sarah Furlong, Elissa Newport. Learning a language from inconsistent input: Regularization in child and adult learners. 2020. doi.org/10.31234/osf.io/kgz9x
[10]Vanessa Ferdinand, Bill Thompson, Simon Kirby, Kenny Smith. Regularization behavior in a non-linguistic domain. eScholarship (California Digital Library) 2013. http://www.escholarship.org/uc/item/8fx246sv
[11]Carmen Saldana, Kenny Smith, Simon Kirby, Jennifer Culbertson. Is Regularization Uniform across Linguistic Levels? Comparing Learning and Production of Unconditioned Probabilistic Variation in Morphology and Word Order. Language Learning and Development 2021. doi.org/10.1080/15475441.2021.1876697
[13]Reali, Florencia, Griffiths, Thomas L.. The Evolution of Frequency Distributions: Relating Regularization to Inductive Biases through Iterated Learning. Cognition 2009. doi.org/10.1016/j.cognition.2009.02.012
[14]Hudson Kam, Carla L., Chang, Ann. Investigating the Cause of Language Regularization in Adults: Memory Constraints or Learning Effects?. Journal of Experimental Psychology: Learning, Memory, and Cognition 2009. doi.org/10.1037/a0015097
[15]Anna Samara, Kenny Smith, Helen Brown, Elizabeth Wonnacott. Acquiring variation in an artificial language: Children and adults are sensitive to socially conditioned linguistic variation. Cognitive Psychology 2017. doi.org/10.1016/j.cogpsych.2017.02.004
[17]Ana Fernández-Dobao, Julia Herschensohn. Acquisition of Spanish verbal morphology by child bilinguals: Overregularization by heritage speakers and second language learners. Bilingualism: Language and Cognition 2020. doi.org/10.1017/s1366728920000310
[18]Singleton, Jenny L., Newport, Elissa L.. When Learners Surpass Their Models: The Acquisition of American Sign Language from Inconsistent Input. Cognitive Psychology 2004. doi.org/10.1016/j.cogpsych.2004.05.001
[19]Kenny Smith, Amy Perfors, Olga Fehér, Anna Samara, Kate Swoboda, Elizabeth Wonnacott. Language learning, language use and the evolution of linguistic variation. Philosophical Transactions of the Royal Society B Biological Sciences 2016. doi.org/10.1098/rstb.2016.0051
[20]Elissa L. Newport. Statistical language learning: computational, maturational, and linguistic constraints. Language and Cognition 2016. doi.org/10.1017/langcog.2016.20
[21]Carla L. Hudson Kam. The Impact of Conditioning Variables on the Acquisition of Variation in Adult and Child Learners. Language 2015. doi.org/10.1353/lan.2015.0051
[22]Aislinn Keogh, Simon Kirby, Jennifer Culbertson. Predictability and Variation in Language Are Differentially Affected by Learning and Production. Cognitive Science 2024. doi.org/10.1111/cogs.13435
[23]Do, Youngah, Mooney, Shannon. Variation Awaiting Bias: Substantively Biased Learning of Vowel Harmony Variation. Journal of Child Language 2022. doi.org/10.1017/s0305000920000719
[24]Thomas L. Griffiths, Michael L. Kalish. Language Evolution by Iterated Learning With Bayesian Agents. Cognitive Science 2007. doi.org/10.1080/15326900701326576
Does this mean adults are worse at learning a new language than children?
Not necessarily — the research shows a difference in how children and adults handle inconsistent, messy input, not an overall verdict on who learns language better. Children tend to smooth over irregularities almost automatically, while adults tend to accurately copy whatever inconsistency they're exposed to; this makes adults arguably better at capturing genuine irregular exceptions, though it also means adults don't get the 'free cleanup' of noisy input that children seem to get.
Why do people sometimes say 'goed' instead of 'went' or apply a regular verb ending to an irregular verb?
This is a classic example of overregularization, where a learner applies a common, predictable pattern (like adding '-ed' for past tense) to a word that is actually an exception to that pattern. Studies of English past-tense errors and of Spanish verb morphology in bilingual children both found this kind of error, especially when the correct irregular form is harder to retrieve or hasn't been solidified yet.
Does being bilingual or learning two languages at once make this regularizing tendency worse?
One study of English-French bilingual children found that bilingual children reduced variation more than monolingual children, but only in the language that was naturally more variable (French) — not simply because managing two languages taxes the brain more. This suggests the effect has more to do with how variable each specific language is than with a general bilingual 'overload.
