Seminar: University Seminar on Cognitive and Behavioral Neuroscience (603)
Date: January 23, 2003
Title: How and why young children acquire language
Speaker: Laura-Ann Petitto, Ph.D., Dartmouth College
Attendees: Herb Terrace, Co-Chair, Psychology Department, Columbia University
Peter Balsam, Co-Chair, Psychology Department, Barnard College
Ann Senghas, Psychology Department, Barnard College
Josh Davis, Psychology Department, Columbia University
Gina Cardillo, Psychology Department, Columbia University
Robert Krauss, Psychology Department, Columbia University
Lois Putnam, Psychology Department, Columbia University
Colin Beer, Psychology Department, Rutgers University
Jen Pardo, Psychology Department, Columbia University
Hadassah Paul, Psychology Department, Hofstra University
Dustin Merritt, Psychology Department, Columbia University
Tammy Moscrip, Psychology Department, Columbia University
Jon Horvitz, Psychology Department, Columbia University
Jacqui Rick, Psychology Department, Columbia University
Kate Lynch, Psychology Department, Columbia University
Anja Soldan, Psychology Department, Columbia University
Nate Kornell, Psychology Department, Columbia University
Robert L. Thompson, Psychology Department, Hunter College
Lisa Son, Psychology Department, Barnard College
Robert L. Thompson, Psychology Department, Hunter College
Julie Day, Psychology Department, Hunter College
Bridgid Finn, Psychology Department, Columbia University
Inge-Marie Eigsti, Psychiatry Department, CPMC
Peter Gordon, Biobehavioral Sciences Department, Teacher’s College
Rapporteur: Michael R. Drew
Summary:
Dr. Petitto began by describing the marvel of the linguistic stream. The marvel is our ability to parse the stream and assign meaning to the parts. This, she said, is one of the most complex computations in the universe. Her talk would focus on how this competency develops.
By 12 months of age, infants have mastered some linguistic skills, such as uttering words and expressing pragmatic meaning through gestures. Petitto asks, what happens during the first 12 months of life to effect the acquisition of these skills? Her talk would address this question using several experimental approaches. Her data point to the conclusion that language competency has a genetic basis and is mediated by brain tissue dedicated to detecting and processing the highly specific temporal patterns of language. Petitto also argues that understanding the neural basis of language acquisition is crucial to understanding our evolutionary origins and the uniqueness of the human species.
Petitto’s research career began in Herb Terrace’s lab, where she worked on the Nim Chimpsky project, in which researchers tried to teach sign language to a chimp. Although Nim was able to respond to and produce numerous signs, he failed to master human grammar. The project confirmed that not all aspects of language are teachable; in being able to use language, humans have a unique ability.
What is the psychological basis of this unique ability? Researchers had once argued that the bases were in the auditory perceptual and speech production apparatus. Petitto’s experiments on acquisition of sign language demonstrate that language acquisition is not tied to the auditory modality. The experiments asked this question: If speech development is governed by the maturation of the perceptual and motor apparatus for producing and perceiving speech, then how will development proceed when a non-spoken language is acquired? Will the nature and timing of developmental milestones be altered? Petitto examined children acquiring sign language and found that these children showed the same developmental milestones at the same time as speaking children. At about 10 months of age both speaking and signing children began babbling. At 12 months they spoke or signed their first words. At 24 months the children began to use morphological and syntactic markings, such as semantics, pragmatics, and discourse.
What if a child experiences both speech and sign language, as in a bilingual (speaking and signing) home? If there is a genetic preference for speech, then the developmental milestones for speech and signing should be asynchronous, with spoken milestones occurring first. On the contrary, the milestones occur at the same time. The children do not prefer speech.
Another set of experiments looked at hearing infants exposed exclusively to sign language, with no systematic exposure to spoken language early in life. Some children were monolingual for American sign language (ASL), others were bilingual for ASL and Langue des Signes Quebecoise (LSQ). All children exhibited the same developmental milestones, and they exhibited the milestones at about the same ages. These data converge on the conclusion that the time course and sequence of maturational milestones is not determined by the maturation of speech perception and production mechanisms. The data contradict the hypothesis that the brain’s language centers are set for sound and speech.
The next series of experiments examined silent babbling, which only recently was shown by Petitto to occur in deaf children. In hearing children, babbling is characterized by the use of a finite set of phonetic units with syllabic organization but without reference. In deaf children, babbling exhibits the same characteristics. Early accounts of babbling assumed that it is caused by rhythmic oscillations between the open and closed configurations of the vocal tract that are accompanied by phonation. The demonstration of babbling in deaf children suggests that there is a more basic, multimodal mechanism for babbling. The babbling experiment used Optitrack technology to electronically record hand and foot motion in 5-month-old children. There were two groups: hearing children exposed to spoken language and hearing children exposed only to sign language. The software blindly divided children into two groups. One group produced only fast rhythmic hand movements (2.5-3hz), and the other group produced both slow (1 hz) and fast rhythmic hand movements. The former group consisted only of children exposed to spoken language and the latter only of children exposed to sign language. The slow rhythmic movements are manual babbling occurring at about the same frequency as vocal babbling and normal speech. Petitto speculates that rhythm generator underlying vocal and manual babbling confers a sensitivity to the rhythms of speech, and this in turn allows children to begin to parse speech into meaningful units.
If this rhythm sensitivity is the basis for language acquisition, then the presence of rhythm sensitivity in young children conflicts with earlier theories that brain language centers do not come online until later in childhood. One of the arguments for delayed activation of the language centers is that lateralization of brain language centers is not established until at least the second year of life. To address whether lateralization might be established earlier, Petitto looked at whether manual and vocal babbling are lateralized. Her studies took advantage of an earlier finding by Gazzaniga and colleagues that people speak out of the right side of their mouth, presumably because language production centers are located on the left side of the brain. Similarly, smiling occurs predominantly on the left side of the mouth, presumably because emotional processing occurs predominantly in the right hemisphere. Petitto discovered that the slow rhythmic hand movements and vocalizations characteristic of babbling are indeed localized primarily on the right side of the mouth and hands. In contrast, the faster rhythmic movements and vocalizations, which are not classified as babbling, are symmetrical. This finding indicates that the language centers are lateralized by 5 months of age.
All the data point to the conclusion that language processing is driven by a central neural mechanism that is not specific to the spoken language. To explicitly test this hypothesis, Petitto used PET scanning. She compared deaf and hearing people who were exposed to meaningless signs and vocalizations. The signs and vocalizations could be either phonetic or nonphonetic, but in either case the were presented at the 1.5 hz rhythm that is characteristic of normal speech and signing. In hearing subjects the superior temporal gyrus (STG) was activated by phonetic vocalizations but not nonphonetic vocalizations or signing. In deaf subjects, the STG was activated by phonetic signing but not by nonphonetic signing or vocalizations. Extant theories of language posited that language processing is located in the STG because of the STG’s vicinity to auditory areas. The finding that the STG is also engaged in processing of sign language suggests that it is not reserved for spoken language. Petitto believes that the STG is instead reserved for processing the low frequency, high contrast rhythms characteristic of language, independent of modality.
Further evidence that the STG is multimodal comes from an fMRI morphometry study. Petitto examined the size of the primary and seconday auditory cortices in deaf and hearing adults. MRI images of their brains were taken, and the size of the STG white and gray matter was assessed by raters who were blind to both the category of the subject (deaf versus hearing) and the hemisphere (right versus left) they were scoring. Deaf adults were identical to hearing adults in terms of STG gray and white matter volume and location and in terms of hemispheric asymmetry. Petitto argued that if the STG were specialized for speech processing, then the STG in deaf adults would be atrophied. The absence of atrophy is consistent with her hypothesis that language processing areas are not pre-programmed to process a particular modality of input, but are instead sensitive to some aspect of language that is independent of modality. Specifically, Petitto hypothesizes that the language areas are dedicated to processing signals with a 1 to 1.5 s temporal window, and this tissue is essential for the acquisition of language. It is only through exposure to spoken or signed language early in development that these language areas become tuned to a single modality to the exclusion of others.
Discussion:
Lois Putnam: You started your talk by saying that humans have neural tissue dedicated to language processing, but you also said that sensory experience is crucial for the development of language centers. What happens to these centers in the absence of language exposure?
Dr. Petitto: Children are sensitive to the rhythmic patterns in language. They thrive on patterns. There is a marriage between what’s in the child and what’s in the environment. Children need language input.
Putnam: So without those patterns they will not develop?
Petitto: Yes. The STG is on a very strict maturational timetable with peak sensitivity during the first 12 months and with sensitivity dropping off after about 7 years of age. If there is no language exposure during that time period other systems will come in and take the child along, but the child will show telltale signs of having abnormal language development.
Ann Senghas: I have two questions. First, you would hypothesize that if a child were exposed to language input, but the input was speeded up or slowed down so that the rhythms are not within the 1 to 1.5 hz range, they would not acquire language.
Petitto: Yes.
Senghas: The other question is about the modality. Is language acquisition really modality independent? When people acquire language tactilely, are they really representing and processing language the same way as hearing people are? I wouldn’t be surprised if language acquisition were limited just to the manual and verbal modalities.
Petitto: Human language must be segmentable, perceivable, and inalienable. So the hands and tongue, which have high sensitivity, are the best options. But consider the Todoma Box, which is used by the deaf-blind to communicate tactilely. The box is placed on the stomach, and pins present tactile patterns representing words. Carol Chomsky has shown that these people form phonetic categories based on the tactile input. How much more removed from the vocal modality can you get?
Herb Terrace: I have two questions. First, do you see any implications for the evolution of language, vis-à-vis hand gestures and speech? Would you entertain the hypothesis that language could have developed initially through gestures? Second, why the 1.5 hz rhythm? Is there anything special about that frequency?
Petitto: First, yes language could have started gesturally. Indeed, it has to be so. The vocal apparatus are about 40 to 70 thousand years old. Now let’s go back to homo sapiens sapiens. One-hundred thousand years ago there was a spectacular flourishing of cave art. They wore loincloths. They buried their dead with symbols. They traveled in packs. They had coordinated hunts. They had fires. I don’t think they were mute. We know they could not have talked in the way we segment the speech stream. But they probably had a communicative trilogy. They grunted. They had face-to-face affect communication. And they had hands. When you add in the hands, then you are putting enough selection pressure on the brain to move away from the raw signal and set up a sensitivity to the patterns of the signal that do not hinge on the modality of the input. Why did it settle on 1.5 hz? I don’t know. But we do know that the brain keeps evolving toward increased processing speed, and this rhythm allows for fast processing.
Peter Balsam: The signer’s signal was subtracted from 1hz hand gestures in the PET study?
Petitto: No, the phonetic hand gestures were subtracted from the non-phonetic hand gestures.
Balsam: So how does the brain tell the difference?
Petitto: Temporally. The maximal contrasts are different. The babies are looking for a rhythm that is undulating at 1.5 hz that is in maximal contrast. The maximal contrast means open and closed.
Balsam: Could you demonstrate?
Petitto: I cannot because the stimuli were synthetic; that is, we smoothed the hands digitally to remove the maximal contrast. In hand babbling the hands go from maximally open to maximally closed. In mouth babbling, the mouth goes from maximally open to maximally closed. Initially it was thought that these oscillations were due to oscillations in the mandible. It was even thought that syllables are an artifact of these oscillations. But how can you explain hand babbling?
Jen Pardo: A second to 1.5 seconds is a little long for speech. You get about three or four syllables within that time frame, especially in natural running speech. I don’t know about in manual language, but in speech.
Petitto: Yes, in sign language it is the same. You can get the word and the basic phrase. So there are similar temporal constraints.
Hadassah Paul: Suppose you had a hearing child that was exposed only to sign language for the first three years. Then she is exposed to speech. What happens? Are there any ways in which that child’s spoken language will be different from a child who was not exposed to any language for the first 3 years?
Petitto: All the hearing children I showed you who did not have spoken language input never maintained this regime after age 3. For the first three years they were exposed only to sign language. Then at age 3 they were thrown into daycare and exposed to spoken language. Those children show absolutely no impact whatsoever of having had no systematic input of speech. When we compared them to hearing children exposed only to French until age 3, then thrown into an English environment, we saw no differences. We also had the reverse: children exposed only to English for the first three years, then exposed to French. There was normal bilingual development in all these groups of children. All acquired the new language in the same way. But they are very different from children who had no language exposure. In those children the portrait is very different.