Again, this result looks as if it can be explained simply in terms of learning; as adults we learn to adapt to speech at different rates, and so we learn to be flexible in our interpretation of perceivable differences. But things are never that simple, and it turns out that babies do it too. Peter Eimas, with Joanne Miller at Northeastern University in Boston, found that babies also differentially interpret the relative durations of the different frequency components depending on the rate of the speech. It is probably just as well that babies can adjust to, or correct for, different rates of speech; otherwise they would hear two versions of the same syllable, uttered at different rates, and assume that they were different when in fact they were supposed to be the same.
We are left with a puzzle. The problem is that a word spoken quickly is not just a speeded-up version of a word spoken slowly. The actual sounds change: voice onset times differ, and slight changes in the frequencies occur which, were it not for the change in rate, would ordinarily be interpreted as different sounds. If the difference between fast and slow speech were the same as the difference between a tape played faster or slower, the problem would be easily solved, because although the absolute speed changes, the relative changes that occur between the individual sounds on the tape would be the same, irrespective of the speed. But in naturally produced speech, the difference between two sounds changes depending on the rate at which they were uttered. So being able to compensate for speaking rate is a complex process, and it would not be unreasonable to assume that it would require some considerable experience of spoken language. How could a baby manage this? It cannot be due to any knowledge about what words actually exist, as the baby does not yet have this knowledge.
We do not know whether babies are as adept at compensating for speech rate as adults are. Are babies and adults basically doing the same thing? The information we have so far suggests that the answer is `yes', but there are likely to be subtle differences that we have yet to discover. If the answer continues to be `yes', then perhaps this is the uniquely human ability that categorical perception was once thought to be. So the next obvious question is this: what about chinchillas? Do they compensate for speech rate? We do not know. But Macaque monkeys do, so probably chinchillas do too.
Whether babies, monkeys, chinchillas, and other inexperienced users of language all do the same thing as adult humans when compensating for speech rate is unknown. And whether these inexperienced language users can, like adult humans, compensate for different voices and different accents is also unknown. If they do not, then we need to explain how babies cope with so much variation between different speakers (and between the same speaker on different occasions). If they do, we need to explain how the hardware of the ear, and the wetware of the nervous system, allow such compensation to take place. Whatever the answers, the mystery remains. But at least we shall have a better idea of just how severe the problems are for the newborn baby and its speech recognition apparatus.
Phonemes and syllables are the building blocks of words. To communicate, the developing infant has to learn about words-what they mean, and how they are combined into meaningful sentences. It is to that mystery that we now turn.
Words, and what we learn
to do with them
Qua ngon ngu' chung to co' the' hieu' du'o'c y nghi cu'a con ngu'o'i. Unless you read Vietnamese, that sentence will mean nothing to you. Being able to distinguish between the different letters and the different words is of little use if you are unable to extract any meaning from those words. And even if you had the appropriate dictionary, you would still need to know something about the grammar of Vietnamese to be sure that you had understood the sentence in the way that was originally intended. But getting hold of the appropriate dictionary, and asking someone about the grammar of the language, is not an option for your average infant. So how do infants manage?
A problem with figuring out how children learn about words is that you just cannot ask an 18-month-old child what words it knows, or what it knows about those words, or how it came to know whatever it knows. And as any parent knows, even when children act as if they understand, they quite often do not.
One of the first words that my son Sam Altmann used was `yeah'. But the following is an example of a typical dialogue with him (aged around 20 months):
Question: Do you want a yoghurt?
Answer: yeah [and he proceeds to eat it]
Question: Do you want a juice?
Answer: yeah [but he ignores it]
Question: Do you want to eat worms?
Answer: yeah [they were not offered!]
It looks as if he had learned to say `yeah' in response to any `Do you want ...' question. But did the meaning Sam attached to `yeah' bear no resemblance at all to the meaning that we adults attach to it? In fact, it did bear some resemblance. Sam would, unprompted, point to something he wanted and say `yeah'. But only if it was out of reach. He would repeat it again and again, with increasing desperation, until he got what he wanted (or until he was removed from the object of his sudden affection). Presumably, then, `yeah' functioned in much the same way as `Give it to me'. But Sam could also sit in his toy car and say `yeah' until someone came over and pushed him around in it. So `Do this for me' would be a better translation. And the reason he might say `yeah' in response to just about any `Do you want . . .' question was probably that, like all children, he was simply being opportunistic; it is better to be offered something than to be offered nothing.
This example highlights two things. The first is that children do not learn by simple imitation, even though they often take great delight in imitating-Sam never saw anyone point at something and say `yeah'. The second is a methodological problem with research into language acquisition; when interpreting the ability of a child to use and understand words, it is important to understand the intentions of the child itself. How to do that is quite another matter. Still, ultimately it matters little whether the child's first words mean the same thing as the adult's. As long as they mean something (to the child at least), the child has started on the path to adult language use.
Broadly speaking, research into early language acquisition has pursued two goals; one has been to understand the acquisition of knowledge about individual words, and the other has been to understand the acquisition of knowledge about how words combine to form meaningful sentences. These two goals serve as the background from which to view research into language acquisition.
Learning words
One question of practical concern is whether there is any definition of what constitutes normal development. Although infants demonstrate an ability to comprehend (and even produce) individual words between around eight and 10 months of age, and to comprehend or produce combinations of words between 20 and 36 months, different children proceed at different rates. At 12 months, for instance, the average infant may have a comprehension vocabulary of around 100 words. But some infants have a vocabulary less than half this number, and some have a vocabulary almost twice this number. Of course, these numbers are only approximate because even large scale studies (the largest having investigated more than 1800 infants) rely on parental assessments (aided by checklists and guidelines) which are sometimes prone to subjective impression rather than objective fact. None the less, the numbers give an indication of the scale of individual differences.
Although there is considerable variation in the number of words that infants comprehend at around 12 months, there is much less variation in the number of words that these infants produce. They produce a lot fewer, and so there is less scope for variation. Without comprehension, there could not be any production, so it is unsurprising that the expressive vocabularies of these infants (used for production) lag behind their comprehension vocabularies. In any case, a 10-month-old infant may well understand the meaning of a word without being able to articulate that word; it takes some time before infants have sufficient control over their tongue, lips, and so on. At 12 months, the average infant may produce just 10 words
, with some infants not producing anything much until around 20 months. These late talkers seem to catch up in the vast majority of cases, and often their comprehension vocabulary is normal for their age (even before their expressive vocabulary has caught up). By 24 months, the average number of words produced goes up to in excess of 300 although, again, some toddlers produce less than half this number, and some produce almost twice as many.
Children know more words the older they get, and this is exactly what one would expect. However, although age is a relatively good predictor of vocabulary size, it by no means explains everything. There is so much variability across different infants that age cannot explain even the majority of this variation. Other factors play an important role in influencing the development of the child's vocabulary. These may include environmental factors (such as the degree to which it is necessary for the child to express what he or she wants) and biological factors. The first three years of a child's life are accompanied by significant changes in the structure of the brain (specifically, changes in the connections between individual nerve cells-synaptogenesis) which may influence the development of many of the child's abilities, including, of course, abilities not normally associated with language.
The fact that developments in brain function are correlated with developments in the child's more general abilities has led some researchers to look for, and find, correlations between the development of these other abilities and the development of language. Indeed, these researchers have argued that certain abilities may be prerequisites to early language development; that is, without the ability, language could not develop. Although a variety of abilities have been investigated, one that stands out as being particularly important is the ability to take part in pretend play, which involves one thing (for instance, a bath sponge) being used to symbolize another thing (for instance, a boat).
Pretend play, and the ability to represent one object with another, plays a central part in contemporary theories of how children's minds develop. The symbolic use of one object to represent another is also central to the development of language; after all, words are really just spoken symbols that represent whatever it is that the word means. If children could not represent one thing with another, they certainly could not use words to represent anything but the sounds they consist of. So what do words represent?
Acquiring meaning
The first words that a child tends to utter are the names of the observable objects in its environment; there are disproportionately more nouns than anything else in the early vocabulary (up until around three years). There are numerous anecdotes told by parents which suggest that the names of those first things are learned by simple association. More than one infant or young toddler has believed that the word for a telephone is a `hullo'. It sounds a simple enough thing to do, but it is no mean feat to be able to associate sounds with objects, even when those objects are being waved close to the child's face, or they are being pointed to, or they happen to be growling ominously. Actually, associating sounds with objects is easy-the accomplishment is to associate the right sounds with the right objects. If children simply associated the sounds they heard with whatever was in their field of vision, or whatever they were attending to, what would happen as they looked around with the radio on, or their parents chatting? They would surely associate whatever sound they were hearing with whatever object they were seeing.
Fortunately, children (that is, their brains) do not associate, willy- nilly, whatever they happen to hear with whatever they happen to be looking at at the time. And this is the puzzle: what prevents them from doing so? One theory is that the exaggerated intonation in childdirected speech serves to attract the attention of the child-the radio will not grab the child's attention in the same way, nor will two adults speaking. This attention-grabbing tactic means, in principle at least, that the child will not associate irrelevant speech with the things it happens to be looking at at the time. But unfortunately, this is not quite enough. It fails to explain how children learn the appropriate associations in societies which do not distinguish between child- and adult-directed speech (there are, admittedly, only very few). And it cannot explain how, even if children are directed to the relevant speech, they know which is the appropriate bit to associate with what they see. How is the child to know which bit of the utterance `Look how big that doggie is' corresponds to the furry beast in front of it? It is easy for us to say, because we already know what `look', `how', `big', and so on, mean. But what about the child who has yet to master these other words?
So one problem for the child is to figure out which bit of the listened-to utterance is the important bit-the bit that is supposed to be associated with whatever is being looked at. But this is only half the problem, because there will be a whole myriad of things that are in the child's line of sight at any one moment. And which of these is to be associated with whatever is being heard? In general, there is no guarantee that the child will be attending to the right object when the speaker utters the word for that object. In fact, the object might not even be in the line of sight-a parent or carer might utter `Look at the doggie' when the child is not looking at it. And by the time the child sees the dog, the sound corresponding to `doggie' will be long gone. And the child would still need to figure out what it was supposed to be looking at-the dog? Its owner? Its fur? Its teeth?
It seems an almost impossible task to figure out which of several possible things, or things going on, should be associated with which of several possible sounds. Especially when different kinds of words-assounds require different kinds of association. So whereas some words (corresponding to nouns) tend to associate with objects, other words (corresponding to verbs) tend to associate with the events that those objects play a part in. But which aspect of the event? Is it Daddy kicking the ball, the ball flying through the air, the ball hitting the ground, Daddy running, panting, going red in the face, or what? What constrains the child's interpretation of each word it hears?
Questions such as these have been addressed in laboratory conditions which allow the range of things that a new word could be associated with to be tightly controlled. By showing children objects they have never seen before, or never heard the name of before, the experimenter can look at what children think the meaning of a new word is when it is given to them. Typically, these experiments involve a puppet who speaks in `puppet talk', and the child is given the task of picking out the objects that the puppet wants. In effect, the child has to guess the meaning of a new word and demonstrate what that meaning is by picking out the appropriate object. The meanings of verbs can also be studied, by showing the puppet doing something (perhaps to another puppet), and giving that action a name, such as `gorping'. The child then has to say when it sees the puppet gorping again. Puppets are used because children are generally quite inclined to help a puppet-the fun element maximizes the likelihood that the children will cooperate, something that they are often loath to do.
A number of studies by Ellen Markman and colleagues, carried out in the 1980s at Stanford University, looked at cases where children were given novel words for things that the puppet was pointing to (in fact, not all these studies used puppets, but the principle is the same). The child's task was made a little more difficult, and consequently a little more realistic, by ensuring that there were several things in the line of sight, or that one thing contained another. What would children do? How would they interpret the new words they heard in puppet-speak? Two findings stand out. The first is that if children hear a new word for something, they assume that it applies to the whole thing, and not to any of that thing's parts (it applies to the dog as a whole, and not to its teeth). And the second is that if they already know the name of something, they assume that a new word cannot apply to that same thing-it must either apply to a part of that thing (to the teeth and not to the dog as a whole), or to another thing. This is just as well, as otherwise children might end up with hundreds of words for one thing, and no words for a hundred others.
>
Markman's findings illustrate some useful assumptions for children to make-they allow each child to restrict the range of things that a new word could refer to. But where do these assumptions come from? How does the child know to make them? In fact, it is possible that the child is not doing anything that rats and pigeons (and other animals) cannot also do. If a rat, for instance, is trained to expect some event (call it `A') when given some stimulus (call it `B'), and then experiences A again when given both B and a new stimulus C, it will not ordinarily associate that new stimulus C with A. This is similar in some respects to the child assuming that if it already knows the name of something, a new name probably does not refer to that same something. And generally, whole things are more salient than their parts (so if something moves, all of its parts move too, and the whole assembly of moving things will be more salient than any individual part of that assembly). When a child associates words with whole objects rather than to those objects' parts, it may do so on the basis of very primitive perceptual sensitivities. So it looks as if at least some of the `assumptions' that children make may have their origins in some very basic phenomena that can be seen in other animals-phenomena that have more to do with associative learning than with language.
The Ascent of Babel: An Exploration of Language, Mind, and Understanding Page 5