The first time a mother and her infant arrive for an appointment at Michael Goldstein’s lab in the psychology building on the Cornell University campus, they’re not tested at all. They’re simply put in a quiet room with a few toys, for half an hour, to get used to the setting. The walls are white, decorated with Winnie the Pooh stickers. The carpet is light brown and comfortable to sit on. On the floor are many of the same playthings the infant might have at home—a brightly colored stuffed inchworm, stacking rings, a play mat with removable shapes, and a toy chest to explore. At three points around the room, videocameras extend from the wall, draped in white cloth to be inconspicuous. The mom knows full well that she is being watched, both on camera and through a large one-way glass pane. But this is otherwise a nice moment to interact with her baby—she can’t be distracted by the cell phone or household chores. Her baby pulls himself to his mom’s lap, puts the toys nearby in his mouth, and if he can crawl, perhaps pulls himself up to look inside the toy chest.
The next day, mother and baby return. In Goldstein’s seminal experiment, the nine-month-old infant is put in denim overalls that carry a very sensitive wireless microphone in the chest pocket. The mother is given a pair of wireless headphones that still allow her to hear her baby. They are put back in the playroom, and again asked to play together naturally. After ten minutes, a researcher’s voice comes over the headphones with instructions. When the mom hears the prompt “Go ahead,” she’s supposed to lean in even closer to her baby, pat or rub the child, and maybe give him a kiss.
The mom doesn’t know what triggers this cue. The mom just knows that, over the next ten minutes, she hears “Go ahead” a lot, almost six times a minute. She might notice that her baby is vocalizing more—or that he’s waving his arms or flapping his feet—but she won’t know what’s triggering this. For the final ten minutes, she’s asked to simply play and interact naturally with her child again.
When mother and infant leave, she has almost no idea what the researchers might have been up to. For two half-hour periods, she merely played and talked with her child.
But here’s what it was like on the other side of the one-way glass: during those middle ten minutes, every time the child made a voiced sound (as opposed to a cough, grunt, or raspberry), it could be heard loudly over speakers in the observation room. Immediately, the researcher told the mother to “go ahead,” and within a second the mother had affectionately touched the child. Later that night, a graduate student would sit down with the session videotape and take notes, second by second, tracking how often the baby babbled, and what quality of sounds he made.
While all baby babble might sound like gibberish, there’s actually a progression of overlapping stages, with each type of babble more mature and advanced than the one prior. “No less than eighty muscles control the vocal tract, which takes a year or more to gain control of,” Goldstein explained. From birth, children make “quasi-resonant” vowel sounds. They use the back of the vocal tract with a closed throat and little breath support. Because the larynx hasn’t yet descended, what breath there is passes through both the mouth and nose. The result is nasal and creaky, often sounding like the baby is fussy (which it’s not).
While the child won’t be able to make the next-stage sound for several months, there’s still a very important interaction with parents going on. They basically take turns “talking,” as if having a mock conversation. The baby coos, and the daddy responds, “Is that so?” The baby babbles again, and the daddy in jest returns, “Well, we’ll have to ask Mom.”
While most parents seem to intuit their role in this turn-taking pattern spontaneously—without being told to do so by any handbook—they don’t all do so equally well. A remarkable study of vocal turn-taking found that when four-month-old infants and their parents exhibited better rhythmic coupling, those children would later have greater cognitive ability.
According to Goldstein, “Turn-taking is driving the vocal development—pushing the babies to make more sophisticated sounds.”
Parents find themselves talking to their baby in the singsongy cadence that’s termed “parentese,” without knowing why they’re strangely compelled to do so. They’re still using English, but the emotional affect is giddily upbeat and the vowels are stretched, with highly exaggerated pitch contours. It’s not cultural—it’s almost universal—and the phonetic qualities help children’s brains discern discrete sounds.
Around five months, a baby has gained enough control of the muscles in the vocal tract to open her throat and push breath through to occasionally produce “fully resonant” vowels. “To a mother of a five-month-old,” Goldstein said, “hearing a fully resonant sound from her baby is a big deal. It’s very exciting.” If her response is well-timed, the child’s brain notices the extra attention these new sounds win. At this point, parents start to phase out responding to all the old sounds, since they’ve heard them so often. That selective responsiveness in turn further pushes the child toward more fully-resonant sounds.
Soon the baby is adding “marginal syllables,” consonant-vowel transitions—rather than “goo” and “coo,” more like “ba” and “da,” using the articulators in the front of the mouth. However, the transition from the consonant to the vowel is drawn out, since the tongue and teeth and upper cleft can’t get out of the way fast enough, causing the vowel to sound distorted. (This is why so many of a baby’s first words start with B and D—those are the first proper consonants the muscles can make.)
As early as six months, but typically around nine months, infants start producing some “canonical syllables,” the basic components of adult speech. The consonant-vowel transition is fast, and the breath is quick. The child is almost ready to combine syllables into words. “We used nine-month-olds in our study because at that age, they are still commonly expressing all four types of babble,” Goldstein said. Quasi-resonant vowel babble might still be in the majority, and canonical syllables quite rare.
With this developmental scale in mind, it’s shocking to hear the difference in how the baby vocalizes over the course of Goldstein’s experiment. In the first ten minutes (that baseline natural period when the mom responded as she might at home), the average child vocalized 25 times. The rate leapt to 55 times in the middle ten minutes, when the mom was being coached to “go ahead” by Goldstein. The complexity and maturity of the babble also shot up dramatically; almost all vowels were now fully voiced, and the syllable formation improved. Canonical syllables, previously infrequent, now were made half the time, on average.
To my ear, it was stunning—the children literally sounded five months older, during the second ten-minute period, than they had in the first.
“What’s most important to note here is that the infant was not mimicking his parent’s sounds,” Goldstein noted.
During those middle ten minutes, the parent was only caressing the child, to reward the babble. The child wasn’t hearing much out of his mother’s mouth. But the touching, by itself, had a remarkable effect on the frequency and maturity of the babble.
Goldstein reproduced the experiment, asking parents to speak to their children as well as touch them. Specifically, he told half the parents what vowel sound to make, the other half he fed a consonant-vowel syllable that was wordlike, such as “dat.” Not surprisingly, the tots who heard vowels uttered more vowels, and those who heard syllables made more canonical syllables. Again though, the babies weren’t repeating the actual vowel or consonant-vowel. Instead, they adopted the phonological pattern. Parents who said “ahh” might hear an “ee” or an “oo” from their baby, and those who said “dat” might hear “bem.” At this tender age, infants aren’t yet attempting to parrot the actual sound a parent makes; they’re learning the consonant-vowel transitions, which they will soon generalize to all words.