Japanese infants hear foreign words by Japanese rules at 14 months, study shows
A baby who has not yet said a word, does not know what a noun is and cannot reliably sit up has already made an irreversible decision about which sounds count. That is the uncomfortable finding of a study on Japanese and French infants by researchers at the RIKEN Brain Science Institute and their collaborators, which produced the first clues about how the machinery for perceiving foreign words gets built and, more to the point, how early it gets locked.
By only 14 months old, Japanese infants were unable to distinguish between words with sound sequences foreign to the Japanese ear. They had tuned their perception to the way sounds are sequenced in their native language before learning that language's words or its grammar.
Sequencing, not just sounds
The finding is subtler than the familiar story about consonants. Most people have heard that Japanese speakers struggle with the English distinction between r and l, and that this is a perceptual problem rather than a muscular one. The RIKEN work points at a different layer: not which sounds a language uses, but which sequences it permits.
Every language has rules about what can follow what. Japanese is unusually strict. Consonant clusters are largely forbidden; syllables want a consonant followed by a vowel. English tolerates pile-ups that would be unpronounceable in Japanese. This system of permitted combinations is what linguists call phonotactics, and the study suggests infants absorb it astonishingly early, then use it as a filter on everything they hear afterwards.
The consequence is that a Japanese infant does not merely find a foreign cluster hard to say. By 14 months, the infant is not fully hearing it as a cluster at all. The brain is inserting the vowel the language expects, because a sequence that cannot occur is a sequence the perceptual system has stopped listening for.
Why this matters for how we think about learning
The question of how infants learn to perceive and segment speech sits at the centre of our understanding of the origins and development of language. Earlier studies had shown that young infants can already distinguish patterns common to their language from patterns that are not. What was not clear was how that early ability connects to the highly tuned, and highly restricted, speech perception found in adults. The RIKEN work draws a line between the two.
The mechanism at work has a name: perceptual narrowing. Newborns are, in effect, universal listeners, capable of discriminating contrasts from languages they have never encountered. Over the first year that capacity is pruned. Distinctions the ambient language does not use are discarded, and the ones it does use are sharpened. It is not damage. It is specialisation, and it is what makes fast, effortless native comprehension possible in the first place. But specialisation has a price, and the price is paid decades later in an evening class.
The adult learner's real problem
This reframes an experience millions of people have had and misdiagnosed. The adult learner who cannot hear a distinction after months of practice is not lazy or unmusical. That learner is running perceptual software optimised for a different input, software installed before they could walk.
Several practical implications follow, and they are better supported than most of the advice in the language-learning market:
- Listening is not passive. Perceptual training, in which a learner is drilled on minimal pairs with immediate feedback, measurably improves discrimination in adults. It is slow, dull, and it works.
- Production follows perception. A learner who cannot reliably hear a contrast has no target to aim at. Pronunciation drills before perceptual drills waste effort.
- Early exposure is disproportionately valuable. Not because children are magically better at grammar, but because the perceptual window is genuinely narrower later.
- Accent is not laziness. A persistent foreign accent is largely a perceptual artefact, not a failure of effort, and treating it as a moral matter helps nobody.
The wider science of second language acquisition has spent decades circling the question of whether adults can ever reach native-like perception. The honest answer is that some do, most do not, and the difference is only weakly explained by talent.
How you test a 14-month-old
The methodology deserves a moment, because it is the part that makes the finding credible. You cannot ask a toddler whether two words sound different. Researchers infer it from behaviour, typically by measuring how long an infant looks towards a sound source, or by habituating the child to one stimulus until attention drops and then switching to another. If attention rebounds, the infant noticed the change. If it does not, the infant did not.
Running that design across two language groups, Japanese and French, is what gives the result its force. French permits sequences Japanese forbids. If Japanese infants fail to detect a contrast that French infants of the same age detect easily, the difference cannot be maturation, attention span or general auditory ability. It has to be the language in the room.
The bilingual case
Infants raised with two languages face a harder computational task and appear to solve it by maintaining two sets of rules in parallel, at the cost of a slightly slower start in each. That early cost disappears. The retained flexibility does not, which is one reason children raised bilingual find a third language easier: their perceptual system was never allowed to narrow quite so far.
None of this makes monolingual adults hopeless. It makes them late, which is a different condition with a different treatment. Understanding language acquisition as a process of narrowing rather than accumulation changes what a serious learner should actually do with their first hundred hours, and it explains why the people who make fastest progress tend to be the ones who spent those hours listening rather than memorising, a pattern that comes up constantly in communities like r/languagelearning.
The baby in the RIKEN lab was not learning Japanese words. She was building the ears she would use to hear everything else.