The short version

Why none of the advice you have read fits your household

Open any guide to raising a bilingual child and count how long it takes before you hit an instruction you cannot carry out. Usually it is the first one. Speak the minority language consistently. Make the home a minority-language zone. Do not answer in English when your child speaks English to you. Read aloud in the language every night. One parent, one language.

Every one of those instructions assumes the language is already inside an adult in the house and the problem is getting it out of that adult and into the child. That is a real problem, and most of the literature is about it, because most bilingual families are that shape. It is not your problem. In your house the language is not in anybody. There is nothing to be consistent with.

This produces a specific and demoralising experience, which we hear from parents often enough that it is worth naming: you read six articles, you implement the one or two things that are technically possible for you — buy picture books you cannot read, put on Japanese cartoons, download an app — and then you watch nothing happen, and you conclude that you have failed at something other people manage. You have not. You implemented the leftovers of a plan designed around a resource you do not have.

The useful move is to stop reading heritage-maintenance advice for tips and start from the structure of your own household. That is what the rest of this article does.

The number that is missing from the most-cited table in bilingual parenting

If you have read anything serious about this, you have probably met Annick De Houwer’s 2007 study. It asked 1,899 families in Flanders, all of them with at least one parent who spoke a language other than Dutch, a simple question: does the child actually speak the minority language? Then it sorted the families by which parent spoke what. The spread is large enough that it gets quoted everywhere, usually as an argument against “one parent, one language.”

Look at that table as a slope rather than as five facts. The thing that moves the number is not method, not effort, not discipline. It is how much of the household’s adult speech is in the minority language. The drop from 96.9% to 35.7% tracks the amount of the language present in the parents, and the famous “one parent, one language” arrangement sits near the bottom of the five precisely because it halves that amount by design.

Now extend the slope downwards by one step, to a household where the figure is zero. We do not know what number sits there, and we are not going to invent one. But we can say something that matters more than the number: whatever is at the end of that slope cannot be reached by the method the slope is measuring. There is no amount of parental consistency available to you, because consistency with nothing is nothing. If your child is going to speak Japanese, the input has to come from outside the household, and your job is not to supply it. Your job is to arrange it, protect it, and make it matter. Those are three different skills, and nobody writes about them because in most families they are a side-show.

Your child’s Japanese will have a different shape, not just less of it

This is the part that surprises parents, and it is the single most useful thing on this page.

Japanese is not one register. At a minimum there are three that a child needs: the plain, intimate speech used at home with family; the polite desu/masu speech used with people outside the circle; and the written, formal language of school and documents. A child raised in a Japanese-speaking household abroad gets the first one in enormous quantity and the second and third barely at all. That is the standard heritage profile, and it is why those children sound warm and natural at the dinner table and then sound blunt, or oddly childish, the moment they are in front of a teacher or a shopkeeper.

Your child will arrive at the opposite profile, and for a reason that is well evidenced.

A 2021 study of Japanese families recorded and transcribed everyday home speech and counted politeness forms. Out of 311,933 parent utterances recorded at home, only about 4% were in polite form; in the children’s own speech the figure was around 2%. In other words, the Japanese that fills a Japanese home is almost entirely the plain register. It is not taught, it is not written down, and it is the first casualty when a language is acquired in a classroom instead of a kitchen.

Everything you can buy for your child is a public situation. A lesson is a public situation. A tutor is a stranger. An app speaks in the register of an announcement. So your child’s Japanese will be built almost entirely from the register that a heritage child is missing — and will be missing almost entirely the register that a heritage child has in abundance.

Two practical consequences follow, and they are worth more than a year of app subscriptions.

First, stop preparing for the rudeness problem. If you have read that children raised abroad shock their relatives by speaking too casually, that is a real phenomenon, and we have written about it at length — but it is a problem of households where plain Japanese is the home language. Your child’s likely failure mode is the mirror image: addressing a grandparent the way one addresses a teacher. Relatives find this touching rather than offensive, which is a kinder failure to have, but it is also a signal that something is missing, and it will not fix itself by adding more lessons, because lessons are where it came from.

Second, the register you cannot buy is the one you have to engineer. Plain-form Japanese lives in exactly one place: relaxed talk between people who are not being evaluated. Not a lesson. Not a presentation. Not an adult asking a child a question the adult already knows the answer to. This is why the single highest-value thing you can arrange is not more teaching — it is unsupervised-feeling conversation with other children, which we return to below.

What you actually control, and it is not input

Here is the reframe. In a household with a Japanese-speaking parent, the parent is the input source and almost everything else is secondary. In your household, you are not an input source at all and never will be. Trying to become a weak one — learning alongside your child so you can speak to them in Japanese — is a nice thing to do for other reasons, but as a language-transmission strategy it has a low ceiling and takes years to reach it.

What you control instead are three variables that the research suggests matter independently of how many minutes of exposure a child gets. None of them is about you speaking.

Lever one: count speakers, not minutes

Parents in your position almost always start by counting time. Thirty minutes a day on an app, one hour a week with a tutor, two films at the weekend. It feels measurable, so it feels like progress.

There is reason to think the count that matters is different. A 2011 diary study of 29 two-year-olds growing up with two languages found that the relative amount of exposure predicted how well each language developed — which is the obvious result — but also that two further things accounted for variance beyond what the hours explained: the number of different people the child heard the language from, and the proportion of the input that came from native speakers.

If you take that seriously, the arithmetic of what you can buy changes completely. A weekly one-to-one lesson, which is what most families in your situation buy first, delivers one speaker. It is the most expensive way to purchase the smallest number in the category. And because that one person is likely to be the only Japanese speaker your child ever interacts with, your child’s entire model of the language becomes one adult’s voice, vocabulary and habits. Children in heritage households at least hear two or three.

Run the same arithmetic over the options and it reorders your shopping list without you having to know anything about teaching.

• An app: zero speakers. It cannot be spoken to, so nothing the child says is ever repaired.
• Videos and anime: zero speakers in the sense that matters, though many voices. Useful for ears, useless for mouths.
• A one-to-one tutor: one speaker, always the same, always an adult, always maximally patient.
• A small group class with the same children each week: several speakers, including children, and the teacher is not the only person your child has to be understood by.
• A Saturday supplementary school: many speakers, if your child can survive the entry conditions — see the caution below.
• Family in Japan on a video call: one or two speakers, enormously motivating, and almost always the cheapest thing on this list. Most families with no Japanese heritage do not have this one, which is exactly why the others matter more for you.

Lever two: other children are a separate input type

Adults talk to children the way adults talk to children. They slow down, simplify, repeat, forgive, and — crucially — they work out what the child meant and respond to that rather than to what the child said. A paid teacher does this harder than anyone, because being understanding is the service.

Other children do none of it. A seven-year-old who does not understand another seven-year-old simply stops playing. That is the pressure that produces speech, and it is the pressure your household structurally cannot generate, because there is no one at home a child must be understood by in Japanese.

There is direct evidence that peer speech is not just a motivational nicety but a distinct kind of input. A 2025 open-access study that recorded 12 children through daily life found, as expected, that speech directed at the child predicted their production patterns best — and additionally that, across both speech aimed at the child and speech merely happening around them, speech from other children was the strongest predictor of how often a child produced a given form. The children in that study were learning a single language in a village setting, not maintaining a heritage one, so take it as evidence about the nature of the input, not about Japanese schools. But it points in an unambiguous direction.

Translated into decisions for your household: between two options at the same price, take the one with other children in the room, even if the teaching looks less sophisticated. Between a prestigious one-to-one tutor and a mediocre group of four, take the four. Your child does not primarily need better instruction. They need somebody their own age who will not switch to English for them.

Lever three: continuity, and why you may be unusually good at it

This is the lever nobody mentions, and it is the one where a non-Japanese household sometimes has an advantage.

In a heritage family, language transmission is woven into family life, which sounds like a strength and is often a weakness. It has no schedule, so it can be eroded by any busy month. It carries emotional weight, so a bad week turns into a row. The Japanese parent is simultaneously the teacher, the quality control and the person who has to put the child to bed, and when those roles conflict the language loses.

Your situation is cleaner. Japanese in your household is an arrangement, not a relationship. It has a time, a cost and a person attached to it. You are not qualified to judge your child’s accent and everyone knows it, which removes the single most common source of conflict in heritage homes — the parent correcting, the child retreating. And protecting a recurring commitment across years, against sport, exams, house moves and everything else, is an ordinary parenting skill that has nothing to do with speaking Japanese.

We cannot give you a figure for how much continuity is worth, because we did not find a study that measures it for this population, and we are not going to borrow one from somewhere else and dress it up. What we can tell you is the shape of the failure we see described most often: families in your position do not usually quit because the method was wrong. They quit in the third year, in a month when something else needed the slot. Decide now what that slot is, and decide that it is not negotiable, and you have done the thing you are actually best placed to do.

The status question works differently when both parents are the non-speaker

There is a well-established point about the parent who does not speak the language: children work out what a language is worth less from the parent who speaks it than from the parent who does not. A father who leaves the room during Japanese calls teaches one lesson; a father who knows thirty words and defends the Saturday slot teaches the opposite one. We have set that argument out in full in our article for households with one Japanese parent, and we will not repeat it here.

What is different in your house is worth saying, though, because it is not obvious. In a one-Japanese-parent family, the non-speaking parent is signalling an attitude toward something the child can already see: Mum’s language, Mum’s family, Mum’s phone calls. The signal has a referent. In your family there is no referent. Your child has never seen an adult in their own home use Japanese for anything real. So the question your child is quietly answering is not “do my parents value this?” but something harder: “is this a thing adults actually do, or is it a thing children are made to do?”

That reframes your job from modelling to casting. You cannot be the adult who uses Japanese for real purposes. Somebody has to be, in your child’s sight, repeatedly, and not in a teaching role. Practical versions, in rough order of how easy they are:

• Let your child watch you fail at it, cheerfully and often. Try the greeting, get it wrong, let them correct you. This is the one piece of modelling available to you, and the content of it is not Japanese — it is that an adult can be bad at something and keep going.
• Put your child in the room when Japanese is being used for a purpose that is not their lesson. A Japanese-speaking colleague on a call. A restaurant where you let your child order. A shop where the staff speak Japanese. Anything where the language is doing a job rather than being practised.
• Find one adult who is not paid to speak Japanese to your child. A neighbour, a colleague’s parent, a university student, somebody at a local Japanese cultural association. Twenty minutes a month of a real person is worth more as a status signal than any number of lessons, because a lesson is evidence that this is a thing children are made to do.
• Never apologise for the schedule. “Sorry, I know it is a lot” tells a child the activity is a burden the family tolerates. Nobody apologises for swimming lessons.

What does not work in this household specifically

The honest comparison below is not about which product is best. It is about which one supplies the three things your household cannot: more than one speaker, at least one of them a child, and a reason to keep going for years. We have put our own kind of provision in the shaded column and marked where it loses.

Two of those rows deserve a sentence each, because they are where we lose and we would rather say so.

A class cannot give your child plain-form Japanese. It is a public setting with a teacher in it, so it runs in the polite register, which is exactly the half your child is already going to over-develop. The only fixes are outside any school: relatives, friendships that outlive the class, and time in Japan. If somebody sells you a course on the promise that your child will sound like a Japanese child at home, they are selling you something no course can produce.

And you are the single point of failure. In a heritage household, if the lessons stop, some Japanese keeps happening by accident. In yours, if the arrangement stops, the language stops the same week. That is not an argument for buying more. It is an argument for buying something small that you can keep for a decade, rather than something impressive you will abandon in year two.

A caution about Saturday schools before you apply

Japanese supplementary schools — the Saturday schools that exist in most cities with a Japanese community — look like the obvious answer, and in the table above they score well. For many families in your position they are the wrong first step, and it is better to know why now than in October.

Most of these schools exist to keep children from Japanese-speaking families on track with the Japanese national curriculum, usually so that a return to Japan stays possible. The teaching is conducted in Japanese, at the pace of children who speak it at home, in subjects rather than as a language course. They are also small: across the world the Japanese government counts roughly 20,822 children of compulsory-school age at supplementary schools, spread over a couple of hundred schools, so a typical school has well under a hundred children across nine year groups.

A six-year-old with no Japanese dropped into that room is not in a beginners’ class. They are in a Japanese first-grade classroom conducted entirely in Japanese. Some schools have a separate stream for children from non-Japanese-speaking homes and some do not, and this varies school by school. Ask that specific question before you apply — not “do you take beginners” but “is there a separate class, who is in it, and what happens at the end of it.” If the answer is that your child joins the mainstream class, the realistic outcome is a term of incomprehension followed by a child who now believes they are bad at Japanese. That belief is far more expensive than the fees.

None of this means never. It means that for a household starting from zero, a Saturday school is usually something you graduate into, not something you start with.

How big is this world, actually?

One of the quieter costs of your situation is the feeling of being an oddity. Heritage families have a visible internet: forums, blogs, communities, people posting about their bilingual toddlers. You have none of that, so it is easy to conclude that families like yours are rare and that the whole enterprise is slightly eccentric.

The numbers say the opposite, and they are not close.

Questionnaires for that survey were distributed between September and December 2024, with the period extended to May 2025 for some countries and regions, and the report was published in March 2026. The counts are of people enrolled at institutions, so children learning at home are not in them. The survey does not record whether a learner has Japanese heritage, so the split between heritage and non-heritage learners is not something it — or we — can tell you.

Most of those two million children are not heritage learners. They are children in Indonesia, Australia, South Korea, Vietnam, China and elsewhere studying Japanese as a school subject, from households with no Japanese in them. That is the category your child is in. It is the largest category there is, and the materials, teachers and qualifications of the world are built around it rather than around heritage maintenance.

This is not consolation. It is a strategic point. You should be shopping in the foreign-language market, not the heritage market, and most of your frustration so far has probably come from shopping in the wrong one. Heritage provision assumes a child who can already speak and needs reading, writing and polish. Foreign-language provision assumes a child who starts at nothing. The second thing is what you need, and there is an enormous amount of it.

Setting a target you can actually hit

Parents in your position tend to arrive with an unspoken target of “bilingual,” usually meaning “like my friend’s child whose mother is Japanese.” Held against that, every year will feel like failure, and in the fourth year you will stop.

We are not going to tell you that target is impossible, because occasionally families reach something close to it and we do not know what distinguishes them — there is no research we could find on outcomes for households with no Japanese-speaking parent, and we would rather leave a hole here than fill it with a guess. What we will say is that it is the wrong thing to steer by, because it is not under your control and it gives you no information month to month.

Better targets, in the order we would set them:

Where an online Japanese school fits, and where it does not

Given everything above, the thing that fits a household like yours is narrower than the market suggests. You need somewhere that assumes no Japanese at home, puts your child with other children rather than alone with an adult, sorts children by what they can say rather than by age or by school year, and is not structurally dependent on your city containing a Japanese community.

That is what TOMODACHI JAPAN is built as. Classes are small and run online with children in Japan and teachers who are not relatives, children are placed by how much they can actually say rather than by age, and nothing about it assumes a Japanese-speaking parent at the other end. For a family with zero Japanese at home it solves the two hardest variables — more than one speaker, and other children — without requiring you to live near a Japanese community.

Who this is not right for

We would rather you did not spend money on the wrong thing, so here is where we would tell you to look elsewhere.

• If you are living in Japan. Your child is surrounded by the language and by Japanese children all day. Your problem is the opposite one — school survival and written Japanese — and the advice on this page is close to useless for you.
• If one parent does speak Japanese but has stopped using it. That is a much better situation than this article describes, and a different fix. Start with the one-Japanese-parent article instead.
• If your child is already a teenager and is not the one asking. Starting a language from zero at fourteen because a parent wants it is a project with a poor record. If the teenager wants it, the record is excellent. Our article on teenagers and Japanese covers the difference.
• If you need a qualification on a specific timetable — a national school examination in Japanese, for example. That is a different purchase: an examination course, usually in your own country, with past papers.
• If you cannot protect a weekly slot for several years. Then do not start with lessons. Start with the free things in the plan below, and start lessons in the year when the slot exists.

Age by age, starting from zero

Starting from nothing at three and starting from nothing at thirteen are not the same project with different timescales. They are different projects.

One thing is constant across all five rows. The written language is a reward, not a foundation. In a heritage household it is reasonable to start kana early, because the child already speaks and reading simply unlocks what they know. In your household, starting with kana means the child’s first experience of Japanese is a worksheet, and a worksheet with no spoken language behind it is indistinguishable from homework. Most of the children we see refuse Japanese within a year were started that way.

A realistic ninety-day plan

This is written for a household where no adult speaks Japanese, and most of it costs nothing. The paid item appears in week five, deliberately, because the free parts determine whether the paid part survives.

Weeks 1 to 4. Pick the slot first: one fixed time each week that is now Japanese, written on whatever your family uses for the calendar, before you have decided what goes in it. Put Japanese audio in the background of ordinary life — music while cooking, a children’s programme at a fixed time — without asking your child to do anything with it. Learn five words yourself, badly, and use them in front of your child. Do not buy a textbook. Do not start hiragana. If your child is under eight, do not ask them to repeat anything.

Weeks 5 to 8. Add human beings. In order of preference: a small group class with other children; failing that, any class with at least one other child in it; failing that, a tutor, chosen for warmth rather than credentials, with an explicit instruction that the lessons are to be conversation rather than worksheets. At the same time, go looking locally — a Japanese cultural association, a university Japanese department, a community noticeboard, a neighbour. You are looking for one unpaid adult and, if you can get it, one child. Nothing on the internet substitutes for a person in your city who says hello to your child in Japanese.

Weeks 9 to 12. Find the use. Something your child wants that is in Japanese: a game with a Japanese setting, a series without dubbing, a message thread with somebody, a video call with the one adult you found. One use is enough; the point is that the language becomes a means rather than an end. Then run two checks, and only these two. Has my child spoken Japanese to a human being who was not paid to listen? Is the weekly slot still intact after twelve weeks of ordinary life? If both are yes, you are in a far stronger position than most families who have been at this for a year. If the second is no, fix the slot before you change anything about the method — the method was almost certainly not the problem.

What not to measure in the first ninety days: vocabulary count, kana recognition, lesson attendance, app streaks, and whether your child can say anything to you. All of those will move, and none of them predicts anything.

Frequently asked questions

Sources used for this article

Your child needs other children, not another adult

TOMODACHI JAPAN runs small online classes with children in Japan and teachers who are not relatives, placing children by what they can actually say rather than by age. Nothing about it assumes a Japanese-speaking parent at your end. The school guide is free to read.

Visit TOMODACHI JAPAN