Baby language: how children learn to speak

Every baby is a born linguist, capable of learning any of the 7,000 languages ​​in the world at a native level. 

Written by: Patricia K. Kohl.

in brief

• Babies' brains enter a "sensitive" phase at the age of six months, a period of time in which the child is able to optimally absorb the sounds of a language or two, in preparation for the development of the fluent tones and rhythms of native speakers.
• The built-in ability for language, when it is for itself, does not advance the child beyond the first murmurs of "mom" and "dad". Learning this essential social skill requires a lot of attention from the baby, for countless hours of parental talk.
• Insights from studies on early language acquisition have reached such a level of sophistication that researchers are thinking about the possibility of using brain activity recordings to check if a particular child is developing properly.

Small children have an amazing and fleeting talent: the ability to master language quickly. At the age of six months, they are able to learn the sounds that make up English words. If they are exposed to Quechua or Tagalog, for example, they will also pick up the unique acoustic features of these languages. At the age of three, toddlers are able to converse with parents, friends and strangers.

Even after forty years of researching child development, I am still amazed by their ability to go from random chatter to words and complete sentences in a matter of years, a mastery that is acquired faster than any other complex skill over a lifetime. Only in recent years have researchers in the field of neuroscience managed to begin to understand exactly what is going on in the brains of babies during this learning process, which turns them from purring newborns into young people who are interesting to talk to.

At birth, babies' brains are capable of grasping all 800 or so sounds needed to make all the words in every language in the world. These sounds are called "phonemes". Our research showed that in the second half of the first year of life, a mysterious door opens in children's minds: they enter a "sensitive phase", as neuroscientists call it, during which the young brain is ready to absorb the first basic lessons in the wonder of language.

The period of time when children's minds are most open to learning the sounds of the mother tongue begins at the age of six months for the vowel sounds, and at the age of nine months for the consonant sounds. The sensitive stage seems to last only a few months, but is prolonged in children who are exposed to the sounds of a second language. The ability to absorb a second language at a good level of fluency lasts until the age of seven.

The built-in ability to absorb language, when it is on its own, is not enough to bring the baby beyond the initial mumbles of "mom" and "dad". Language mastery, which is the most important of all social skills, is achieved through countless hours of listening to parents speak in the local "parental" language. The excessive inflections of this language ("Who's who?") serve the important purpose of providing daily lessons in the spelling and rhythm of the mother tongue. Our study settles the age-old debate, which determines more during the initial stage of language development: genetics or environment. Both have a central role.

Our knowledge about the early development of language has reached such a level of sophistication that it allows psychologists and doctors to create new tools to help children with learning difficulties. Research is beginning to show us how to determine, by measuring brain waves, whether children's language abilities are developing normally, or whether the children are at risk of autism, attention deficit disorder or other disabilities. It is possible that one day, the visit to the pediatrician will include, in addition to vaccinations against measles, mumps and rubella, also a brain scan.

The statistics of baby language

We can think of a test for language development because we have begun to understand how babies pick up language so easily. My lab and other labs have shown that infants use two distinct learning mechanisms in the earliest stages of language acquisition: one that recognizes sounds with the help of mental calculations, and one that requires constant social interaction.

To learn to speak, babies need to know which phonemes make up the words they hear around them. They must identify which 40 phonemes, approximately, out of the 800 possible phonemes they need to learn in order to speak the native language. This task requires the detection of subtle differences in the sounds in question. A difference of a single consonant can change the meaning of a word, for example "good" and "bear". A simple syllable like "ah" sounds different when different speakers say it in different contexts and speech rates. Such differences in phonemes are the reason why speech recognition software, such as Apple's "Siri", still does not work perfectly.

My research, and that of Jessica May (then at Northwestern University) and her colleagues, showed that statistical patterns—the frequency with which sounds appear—play a vital role in helping toddlers understand which phonemes are most important. Between the ages of eight and ten months, children still do not understand words, but they are very sensitive to the frequency of phonemes, what statisticians call frequency distribution. The most important phonemes in a language are the phonemes that are heard most often in speech. In English, for example, the sounds "r" and "l" are quite common and appear in many words. In Japanese, these sounds exist but appear less frequently, while the common Japanese "r" phoneme is almost unheard in the English language (to English speakers, the Japanese "r" sounds like "l" because it is halfway between "r" and "to" Americanism).

The statistical frequency of separate sounds affects the brain of babies. In one study, conducted in Seattle and Stockholm, we monitored the perception of vowel sounds at the age of six months, and showed that each group had already begun to focus on the vowels spoken in their native language. The culture of the spoken language had already permeated and influenced the way the babies' brains perceived sounds.

What is going on here, exactly? May showed that at this age, the brain is flexible enough to change the way sound is perceived. A Japanese baby who hears sounds from the English language learns to distinguish between American "r" and "l", and on the other hand a baby who grows up among English speakers will be able to pick up the typical Japanese sounds. It seems that learning sounds in the second half of the first year of life establishes connections in the brain that correspond to the mother tongue and not to other languages, unless the child is exposed during this period to multiple languages.

Later in childhood, and especially in old age, listening to a new language does not produce such dramatic results. A tourist in France or Japan will hear the statistical distributions of the sounds of the foreign language, but the brain will not change as a result. That is why it is so difficult to learn a second language later in life.

Another type of statistical learning allows infants to recognize whole words. As adults, we can tell when one word ends and another begins, but the ability to isolate words from the flow of speech requires complex mental processing. The spoken language reaches the ear as a continuous stream of sounds, without the separation between written words.

Jenny Sepran (now at the University of Wisconsin-Madison) and her colleagues, Richard Aslin of the University of Rochester and Alyssa Newport of Georgetown University, were the first to discover that infants use statistical learning to grasp the sounds of whole words. In the mid-90s, Safran's group published evidence that eight-month-old infants are able to learn word-like units based on the probability of the occurrence of a particular syllable after another. For example, because of the conjunction "cute baby" the baby will hear the syllable "thee" more often with "nu" after it, than with a syllable like "ha".

In an experiment she conducted, a librarian played sequences of meaningless, computer-generated made-up words to babies, with certain combinations of syllables more common than others. The infants' ability to focus on syllables associated with the invented language enabled them to recognize words.

The discovery of the statistical learning abilities of infants in the 90s caused much excitement because it hinted at a new theory of language learning. Until then, the prevailing idea was that children learn only thanks to conditioning and reinforcement from parents, who show them when a word is right or wrong. It turned out that the babies' learning happens before the parents are even aware of it. However, further experiments in my laboratory brought up a new and important finding that qualifies the above: the statistical learning process needs more than passive listening.

Babies meet

Our research has revealed that babies need to be more than just computational geniuses, processing clever neural algorithms. In 2003 we published the results of experiments in which nine-month-old babies from Seattle were exposed to Mandarin Chinese. We wanted to know if their static learning abilities would enable them to learn Mandarin phonemes.

The babies, in groups of two or three, listened to speakers whose native language was Mandarin while they played with them on the floor, with books and toys. Two other groups were also exposed to Mandarin, but one of them watched a film in which Mandarin was spoken, and the other listened to a recording. A fourth control group did not hear Mandarin at all, but instead listened to American students speak English while they played with the babies with the same books and toys. All this took place during twelve sessions over a month.

At the end of the process, babies from all groups returned to the laboratory to undergo psychological tests and brain monitoring to assess their ability to isolate Mandarin phonemes. Only the group exposed to Chinese in direct contact with the speakers learned to recognize the foreign phonemes. In fact, the performance of the infants in this group was identical to the performance of infants in Taipei, the capital of Taiwan, who listened to their parents speak Mandarin for the first eleven months of their lives.

The babies who were exposed to Mandarin through television or voice recording learned nothing. Their ability to distinguish phonemes was the same as that of the control group, whose performance, as expected, was no better than before the experiment.

The results of the study indicate that learning in babies is not a passive process. It requires human interaction: a need I call a "social gate". The hypothesis can even be extended to explain the way in which many other biological species learn to communicate: the experience of babies learning to speak is actually similar to the way birds learn to sing.

Before that, I worked with the late Alison Dupe of the University of California, San Francisco, and we compared the learning of babies to the learning of birds. We discovered that both in babies and in pinks, the social experience in the first months of life was essential. The babies and chicks constantly listen to their elders, and keep in their memory the sounds they heard. These memories shape the motor areas of the brain to produce the sounds often heard in the wider social community in which they grew up.

We still do not know how exactly the social context contributes to language learning in humans. My hypothesis is that parents and other adults provide both the motivation and the information necessary for learning. The motivational component is driven by the reward system in the brain, and especially by the areas of the brain that use the neurotransmitter dopamine during social interactions. Studies done in my lab have already shown that babies learn better in the presence of other babies, and we are currently conducting additional studies aimed at explaining why this is so.

Babies who look into their parents' eyes also receive important social cues that accelerate the next stage of language learning: understanding the meaning of actual words. Andrew Meltzoff of the University of Washington showed that young children who follow the gaze of adults develop a larger vocabulary in the first two years of their lives than children who do not follow eye movements. The connection between observation and speaking makes perfect sense, and also explains why watching an educational video is not enough.

In the group that learned directly from Mandarin speakers, the babies could see when the teacher looked at an object when she called its name: a tiny action that linked the word to the object. In a study published in July 2015, we also showed that when a Spanish teacher holds new toys and talks about them, the babies who alternately look at the teacher and the toy (instead of focusing on just one of them) learn both the phonemes and the words that appeared in the lesson. This example reinforces my theory, according to which the social skills of the babies strengthen, or "open a gate" for language learning.

These ideas, about the social component in the early learning of language, may also explain some of the difficulties faced by babies who later develop disorders such as autism. Babies with autism do not show a basic interest in speech. Instead, they focus on inanimate objects and ignore the social cues that are so important to language learning.

Say sha-a-lo-w-m!

A baby's ability to learn to speak depends not only on the ability to listen to adults, but also on how the adults speak to him. Anywhere in the world - Dhaka, Paris, Riga or the Tulalip Indian Reservation near Seattle - researchers who listened to people talking to babies discovered one simple thing: adults talk to babies differently from the way they talk to adults. Cultural ethnographers and haters call this "baby talk", and it is prevalent in most cultures. In the first stage, it was not clear whether baby talk could delay language development, but many studies have shown that the "mother" or "parental" language, as it is now called, actually helps children learn. This is not a modern invention either: a Roman expert on syntax, Varro (116 to 27 BC), noted that certain abbreviated words were only used when speaking to babies and small children.

My lab, as well as the labs of Anne Ferland at Stanford University and Lila Gleitman at the University of Pennsylvania, examined the sounds unique to "parenting" that intrigue toddlers: the high frequency, the slow tempo and the exaggerated playing. Babies will prefer to hear recordings of a "parent" over recordings of the same mothers talking to other adults. The high sounds are like an acoustic bait, which attracts and keeps the attention of the babies.

In "Horit", the differences between sounds are very emphasized, so it is easy to distinguish between phonemes. Our studies have shown that the excessive talking seems to help babies remember the sounds. In a recent study by my group, Nayron Ramirez-Esparza, now at the University of Connecticut, attached tiny, high-quality recording devices to baby vests. The babies wore the vests at home during the day. The recordings allowed us to enter the world of children's sounds, and we discovered that the babies whose parents spoke to them in "parenting" knew, a year later, more than twice as many words as babies whose parents spoke normally more often.

the traces of learning

With the accumulation of knowledge in the field of early child development, researchers in neuroscience are enthusiastic about the possibility of using this knowledge to identify traces of brain activity, also called "biological markers". Such traces can help detect difficulties in language learning.

In a recent study conducted in my laboratory, we played familiar and foreign words to two-year-olds with autism spectrum disorder while monitoring the electrical activity in their brains. We discovered that the degree to which a certain pattern of brain waves appeared in response to familiar words made it possible to predict the children's cognitive abilities and language abilities afterwards, at the ages of four and six. These measurements actually assessed the children's success in learning from other people. They showed that learning words in a social way is a measure of learning in general.

New tools have recently appeared that allow us to determine children's ability to recognize sounds, and this brings us closer to the day when we can measure their cognitive development. My research group started using magnetoencephalography (MEG), which is a safe and non-invasive imaging method, to show how the brain responds to speech. The device includes 306 SQUID sensors (superconducting quantum interference devices), located in a device that looks like a hairdresser's hair dryer. When the child is connected to the device and listens to speech, the sensors measure tiny magnetic fields that monitor the activity of specific nerve cells in the brain. With the MEG we have already shown that there is a critical window of time in which infants appear to mentally rehearse in preparation for speaking in their native language.

The MEG is too expensive and complex to use in the neighborhood clinic, but such studies pave the way for the identification of biological markers that can be measured, eventually, with the help of portable and cheap sensors, which will be operated outside a university laboratory.

If we can identify reliable biomarkers for language learning, they will help us determine whether children are developing normally or are at risk for language-related disabilities, including autism spectrum disorders, dyslexia, fragile X syndrome, and more. An understanding of the unique human talent for language, and when exactly it can be shaped, will open up for us, perhaps, the possibility of treating such disabilities at an early enough stage to change the child's life.

good to know

The sensitive period: culturally bound listening

At the age of six to eight months, babies are able to distinguish between phonetic units such as "ra" and "la", regardless of the culture in which they grow up. At ten months of age, this window begins to close and babies show the first signs of becoming culturally bound listeners. In a study conducted in Tokyo and Seattle, the ability of Japanese infants to distinguish between "ra" and "la" decreased, while at the same time the ability of American infants to distinguish between these sounds increased (red lines).

A study conducted in Taipei and Seattle showed that the ability of Taiwanese infants to distinguish between the Chinese characters "chi" and "shi" increased, while the ability of American infants to do so decreased (purple lines). Babies instinctively do exactly what is needed to advance their language skills.

The author of the article: Patricia Kohl, head of the chair of the Bezos Family Foundation for early childhood learning. She is also one of the directors of the University of Washington's Institute for Learning and Neuroscience, and director of the LIFE Center for Learning Sciences funded by the US National Science Foundation.