Almost every parent who asks this question is really asking two things at once. Is my child still young enough? And have I already missed the window? The honest answer is reassuring on both counts, but it is not the answer the internet usually gives.
There is no single cut-off age. What exists instead is a set of overlapping windows, each closing at a different speed, each governing a different part of language. Knowing which window is closing and which is wide open is what lets you make a sensible decision at 18 months, at 4 years or at 9.
What is the short answer?
Earlier is better for accent and for effortlessness, but almost everything else stays learnable for years. If your child is under 10, you have not missed anything. If your child is under 3, you have an advantage that is real but smaller than the marketing suggests.
The practical answer for most families is: start whenever you can sustain it. Consistency over years beats an early start that fizzles out after two months, and that is the part nobody puts on a poster.
What actually happens in the first year of life?
Babies start out able to hear the sound distinctions of every language, then lose the ones they do not need. Between roughly 6 and 12 months, an infant gets better at telling apart the sounds of the language spoken around them, and measurably worse at telling apart sounds that language does not use.
Patricia Kuhl and colleagues documented this narrowing in Developmental Science, and the sharpness of the effect is why the first year gets so much attention. A baby who never hears the difference between two sounds stops treating them as different sounds at all. That is what makes a foreign accent hard to shed decades later.
The practical implication is narrow. This window governs pronunciation, not vocabulary, not grammar, not the ability to hold a conversation. A child who starts a second language at 5 can reach a very high level. They are simply more likely to keep a trace of an accent.
Is there a hard cut-off after which it is too late?
No, and the single most cited study on the subject is often misread. Hartshorne, Tenenbaum and Pinker analysed the English of roughly 670,000 native and non-native speakers and found that grammar-learning ability stays close to its peak until about 17.4 years, then declines.
That 2018 paper in Cognition is frequently quoted as proof of a wall at 17. It is closer to the opposite. It says a 12-year-old is still working with nearly the same machinery as a 6-year-old.
And it has been challenged. A 2022 reanalysis in Language Learning by Frans van der Slik and colleagues found no sharp break at all, just a steady decline from early childhood onward, with differences between types of learner. Reviews in Bilingualism: Language and Cognition land in similar territory. The field agrees that something changes with age. It does not agree on when, or how abruptly.
Does starting at 3 really beat starting at 8?
For final accent, yes. For speed of early progress, no, and this surprises people. Older children learn grammar and vocabulary faster per hour of exposure, because they can use analogy, memory strategies and literacy.
The advantage of the young child is not efficiency. It is time and attitude. A 3-year-old will happily repeat the same twenty words for a month without deciding they are bad at languages. A 9-year-old covers more ground per hour but is far more self-conscious about mistakes.
| Starting age | What comes easily | What it demands from you |
|---|---|---|
| 0 to 3 years | Native-like sounds, no fear of speaking | Daily live exposure, years of it |
| 3 to 6 years | Very good accent, vocabulary through play | Routine and repetition, low pressure |
| 7 to 11 years | Fast grammar, reading transfers across | Motivation and a reason to use it |
| 12 years and up | Rapid progress with explicit teaching | Tolerating the awkward stage out loud |
What do European school systems actually do?
They have been moving the start date earlier for two decades. According to the European Commission's Key Data on Teaching Languages at School in Europe, 2023 edition, learning at least one foreign language is compulsory before age 8 in most systems, and before age 6 in six of them. Around two thirds of systems with comparable data have lowered the starting age since 2002.
Eurydice itself, in a piece bluntly titled "Is younger always better?", is careful not to oversell the change. Starting earlier only pays off if the hours are real and the teaching is good. An early start of one 45-minute lesson a week produces very little, at any age.
Does exposure from a screen count?
Much less than people hope, and the evidence on this is unusually clean. Kuhl's team exposed 9-month-old American infants to Mandarin, some through a live person, some through video and audio of the same material.
The infants who sat with a real person learned the new sound contrasts. The infants who got identical content through a screen learned essentially nothing, as reported in PNAS. Something about being addressed, and about turn-taking, switches the learning on.
That finding lines up with later work on older children. Rachel Romeo and colleagues found that the number of back-and-forth conversational turns a 4 to 6-year-old experiences predicted activity in the brain's language regions better than the sheer number of words heard, published in Psychological Science. Volume of input is not the point. Exchange is.
How much does a young child actually need each day?
Less than you fear, more often than you think. Fifteen to thirty minutes of genuine two-way use, most days, sustained over years, does more than a weekend immersion camp once a term. We went through the numbers in more detail in how many hours a week a child needs to become bilingual.
The hard part is rarely the child. It is finding a second speaker who is available at 6pm on a Tuesday. That is the gap a talking companion can honestly fill: an interactive plush like Ted answers in more than 60 languages and adapts to the child's level, so the back-and-forth keeps happening on the days when no adult is free to provide it. It does not replace a person, and the research above is the reason we say so plainly.
If only one parent speaks the target language, the pattern that works is consistency of person and place rather than volume, which we covered in raising a bilingual child when only one parent speaks the language.
What if nobody at home speaks the language?
You can still do it, but you have to build the input deliberately instead of relying on it happening. Roughly a third of the bilingual families we hear from are in exactly this position: two parents who speak the same language at home, and a genuine wish for their child to grow up with a second one.
The mistake is to think the parent has to be fluent. Children do not need a perfect model to get started, they need a reliable one. A parent working from a hundred words of Spanish, used consistently at the same moment every day, gives a 4-year-old something a flawless once-a-month lesson never will. Accuracy can be corrected later. The habit cannot be installed later nearly as easily.
Three things do most of the work in a household like this. Anchor the language to a fixed slot, such as bath time or the walk to nursery, so nobody has to decide each day whether it is happening. Find at least one real interlocutor beyond the parent, whether that is a grandparent on a video call, a neighbour, a weekend group or a talking companion at home, because a child who only ever hears one voice learns one voice. And protect the child from performance pressure. Asking a preschooler to demonstrate their second language for visiting relatives is the fastest way to make them refuse it, a pattern we looked at in what to do when a child refuses to speak your language.
One more piece of realism. Progress in a non-native household is slower and more visibly effortful than in a household where the language is simply the air. That is not failure. Comprehension almost always runs years ahead of production, so a child who understands everything and answers in French is not stalled, they are at a completely normal stage of the same journey.
Frequently asked questions
At what age should a child start learning a second language?
Any age under 10 is a good starting point, and before 3 gives the clearest advantage for accent. What matters far more than the exact age is whether the child gets regular two-way practice over several years rather than a short burst.
Is 5 years old too late to start a second language?
No. A child starting at 5 can reach a very high level in grammar, vocabulary and fluency. The main thing they are statistically less likely to achieve is a completely native accent, and even that is not ruled out.
Will learning two languages delay my child's speech?
No. Bilingual children reach the major speech milestones on the same schedule as monolingual children. Their vocabulary in each single language may look smaller for a while, but their total vocabulary across both languages is comparable.
Can my child learn a language from videos alone?
Not well. Infants exposed to a new language through video learned almost nothing, while infants exposed to the same material through a live person learned the sound contrasts. Interaction, not content, appears to be the active ingredient.
How many minutes a day is enough for a preschooler?
Fifteen to thirty minutes of active use on most days is a realistic target, provided the child is speaking and not only listening. Frequency matters more than session length at this age.