Most parents who want to teach a child a second language at home begin with the wrong question. They ask which method is the correct one, pick a system, then feel they have failed when the system slips after three weeks. The largest research on the subject points somewhere else entirely: what predicts whether a child ends up actually speaking the language is not the elegance of the method. It is how much of ordinary home life genuinely runs in that language, and whether the child has real reasons to answer back.
That is good news, because the thing you control most is also the thing that matters most. Below: what the evidence says about family language strategies, how much daily exposure a child needs, why a language that was fluent at six often fades by sixteen, and what a realistic week looks like for a family with jobs and school runs. For the wider picture, our guide to helping a child build language covers the ground around this one.
What does it actually mean to teach a child a second language at home?
It means arranging daily life so that the child hears the language from people who matter to them, in situations they care about, and has reasons to reply in it. For children between 3 and 12, a second language is acquired through ordinary interaction rather than instruction. Lessons help later, once the child can already hold a conversation.
The distinction changes what you measure. A parent running "lessons" tracks worksheets and vocabulary lists. A parent running exposure tracks hours, people and situations: who speaks the language to the child, for how long, doing what. The American Speech-Language-Hearing Association's developmental milestones describe the same mechanism for a first language, which is conversation with responsive adults, not drilling.
One consequence is worth stating plainly. If the second language exists only during a twenty minute activity slot, the child will understand it and refuse to speak it. Comprehension is cheap; production is expensive, and children only pay that cost when the language is the normal way to get something they want.
Which family language strategy actually works?
The pattern with the highest success rate is the one where both parents use the second language at home, and at most one of them also uses the majority language. Strategies where one parent speaks only the majority language succeed far less often. The famous "one parent, one language" rule works reasonably well, but it is neither necessary nor sufficient.
The evidence here is unusually solid. Annick De Houwer surveyed 1,899 families in Flanders in which at least one parent spoke a language other than Dutch, each with a child aged 6 to 10, and published the results in Applied Psycholinguistics. Every child in the sample spoke the majority language. What varied, sharply, was whether they also spoke the minority one.
De Houwer's conclusion was blunt: the one parent, one language strategy "did not provide a necessary nor sufficient input condition." Families who mixed both languages did about as well as strict OPOL families, and families where the second language had only one part-time speaker did badly whatever label they used.
| Strategy | How it works | Where it breaks |
|---|---|---|
| One parent, one language (OPOL) | Each parent consistently uses one language with the child | The minority-language parent is often the one who travels, works late or does bedtime only twice a week |
| Minority language at home | The whole household runs in the second language; the majority language comes from school and outside | Hard when one parent does not speak it, and awkward with guests |
| Time and place | The second language owns specific slots: weekends, mealtimes, holidays at the grandparents' | Slots erode first when life gets busy, and the child learns the language is optional |
Most families end up with a hybrid, and that is fine. The question is not "which box am I in" but "how many waking hours a week does my child hear this language from a person who responds to them." Our piece on raising a bilingual child when only one parent speaks the language covers the single-speaker case.
How much of the day does the second language really need?
Roughly a fifth of the child's waking language input is the working minimum, and comfortably more than that is what produces a child who speaks rather than only understands. Below about 20 percent, children very often become receptive bilinguals: they follow everything and answer in the majority language.
This threshold comes from Barbara Zurer Pearson, Sylvia Fernández, Vanessa Lewedeg and D. Kimbrough Oller, who tracked 25 simultaneously bilingual infants and correlated each child's vocabulary in each language against the estimated share of input they received in it. The correlation was strong, r = .82, and the finding was published in Applied Psycholinguistics in 1997. It has been replicated often enough that 20 to 25 percent is now the standard cut-off researchers use.
The arithmetic is worth doing honestly. A child awake thirteen hours a day, at school for six of them in the majority language, has about seven hours of home language time. Twenty percent of the waking day is roughly two and a half hours, so the second language needs to own a large share of the evening, not a corner of it. We worked through the weekly numbers in how many hours a week a child needs to become bilingual.
Does the kind of input matter, or is any exposure equal?
The kind of input matters a great deal, and the clearest dividing line is whether a human being is responding to the child in real time. Recorded audio and video produce very little language learning in young children compared with the same content delivered by a person who reacts to them.
Researchers at the University of Washington's Institute for Learning and Brain Sciences ran one of the cleanest demonstrations of this. In a study of infants in Spain who received one hour a day of English sessions for eighteen weeks, the children who got live, interactive, exaggerated speech from a tutor showed substantial English learning, while comparison conditions did not produce the same effect. The university's summary of the research is worth reading in full.
The practical translation is unglamorous: twenty minutes on a video call with a grandparent beats an hour of cartoons, and narrating the supermarket trip beats a vocabulary app. The NCT's guidance on raising a bilingual child makes the same point from the parenting side.
Why do children who spoke the language at six stop speaking it at sixteen?
Because transmission and retention are two different problems, and most families solve only the first. A language can be passed on successfully in early childhood and still erode through adolescence if it never acquires a function outside the parent-child relationship.
The French data on this is sobering and rarely quoted. In the Ined and Insee Trajectoires et Origines 2 survey of 2019 and 2020, 52 percent of adults born in France to at least one immigrant parent were raised by their parents in French and in at least one other language. So transmission clearly happened. But among those who report a foreign family language, only 16 percent read, write, speak and understand it very well, and 22 percent have very good command of the spoken language only. A third had taken formal classes in it, and those who had read and wrote it noticeably better.
The honest limit. Nothing in this research promises that a language kept alive at home will survive into adulthood. What the Insee figures suggest is that the families where it survives are the ones where the language does something: regular trips, relatives who speak no French, reading and writing taught deliberately, a community that uses it. Speaking it at the dinner table is the floor, not the ceiling.
That same survey found that 40 percent of respondents with a foreign family language now speak it with their own children. The chain continues in roughly four families out of ten.
What does a realistic week look like for a busy family?
A realistic week protects two or three fixed anchors in the second language and lets everything else float. Anchors survive busy periods because they are attached to things that happen anyway: the meal, the bath, the car, the bedtime story.
A workable pattern for a child of 4 to 9: the second language owns the evening from the front door to lights out, which is where most of the hours are. One video call a week with a relative who does not speak the majority language, because a child who can switch will. One book a night in the second language, chosen by the child, reread as often as they want. And one outing a month where the language is unavoidable: a shop, a community group, a cousin, a film without subtitles.
What to drop is correction. Children between 3 and 12 stop producing a language when production is expensive, and public correction makes it expensive. The ASHA guidance on speech and language development points to recasting instead, which means repeating what the child said correctly inside your own reply and moving on. Mixing languages within a sentence is normal and is not a sign of confusion.
Can a talking toy help, and where does it stop?
A talking toy can add conversation practice in a second language, which is the resource most households are short of, and it cannot replace a person. Anyone claiming otherwise is overselling. The Pearson threshold is about hours of real input, and a toy contributes a modest number of them.
Where it does help is narrow but real. A child who is shy about speaking will often talk more freely to an object than to an adult who might correct them, which addresses the production problem described above. A child with one minority-language parent gets a second voice in that language rather than one. And unlike a screen, a talking toy keeps the child's hands and eyes free, so it fits into the evening anchors rather than competing with them.
Where it stops: a toy does not know your family and cannot take your child to see their grandmother. It is one input among several, and it works best when the parents are already doing the rest. Ted is an interactive plush toy built on that assumption, with no camera, no subscription and a parent app where the adult chooses the language, the topics and the limits. The language is set by the parent in the app and does not change on its own during a conversation, which matters if you are protecting a fixed evening slot in one language. Ted costs 129 euros as a one-off purchase. Parents who want the detail on what is collected and what is not will find it on our security page.
Frequently asked questions
My partner does not speak the second language. Is it still worth trying?
Yes, but plan for the harder version. In De Houwer's data, families where one parent speaks only the majority language succeed around 36 percent of the time, so the minority-language parent needs more hours and more outside support: relatives, a weekend school, trips.
Will two languages delay my child's speech?
No. Bilingual children reach the standard milestones on the same schedule as monolingual ones when you count words across both languages together. A child who is genuinely late should be assessed, and the assessment should look at both languages, not just the school one.
My child understands but refuses to answer in the second language. What now?
This is receptive bilingualism, and it usually signals not enough input and not enough need. Increase the hours first, then create situations where the majority language does not work, most reliably a person who does not speak it. Do not make speaking it a condition of getting what they asked for.
Is it too late to start at eight or nine?
No, though the route is different. Older children learn faster explicitly and benefit from structure, reading and a peer group, where a four-year-old just needs hours. What is genuinely harder later is accent, not competence.
Should I correct my child's mistakes in the second language?
Rarely and gently. Recast instead: reply using the correct form naturally in your answer, without drawing attention to it. Explicit correction reliably reduces how much a child is willing to say.
