LISTENING TO a child who is just learning to talk, one is most aware of the child's limited command of the language. What one tends to overlook is the sheer magnitude of the child's achievement. Simply learning the vocabulary is an enormous undertaking. The fact is that for many years after starting to talk a child learns new words at a rate of more than 10 per day! Yet little is known about how children do it.

The word-learning process becomes even more complex during the school years. In the early grades schoolchildren are expected to learn to read and write. At first they read and write familiar words they have already learned by means of conversation. In about the fourth grade they begin to see written words they have not heard in conversation. At this point it is generally assumed that something special must be done to teach children these unfamiliar words.

This educational assumption runs into serious problems. Although children can recognize that they have not seen a word before, learning it well enough to use it correctly and to recognize it automatically is a slow process. Indeed, learning a new word entails so much conceptual clarification and phonological drill that there simply is not enough classroom time to teach more than 100 or 200 words a year in this way. Since learning runs so far ahead of teaching -- some 5,000 words learned in a year compared with 200 taught -- it is hard to avoid the question: How do schoolchildren learn so much more than they are taught?

Many words are acquired through reading. Children learn words at school in the same way as they do at home: by observing how the words are used in intelligible contexts. The difference is that the academic environment depends more on written contexts. Both public opinion and scientific evidence are converging on the view that the best way to facilitate vocabulary growth in schoolchildren is to have them read as much as possible.

Learning words by reading them in context is effective but not efficient. Some contexts are uninformative, others misleading. If the word in question expresses an unfamiliar concept, a single context of use will seldom support more than one hypothesis about the word's meaning. In order for reading to have any substantial effect on vocabulary a great deal of reading must be done.

Turning the Pages

How much? A child who spent 50 minutes of every school day reading at, say, 200 words per minute would read one million words in a 100-day school year. A million running words of English prose would typically contain no more than 50,000 distinct word types, representing roughly 10,000 word families. Schoolbooks would probably contain fewer different words. Even among 10,000 different words, it is unlikely that more than 1,000 would be totally new lexical items. Since multiple encounters are required in order to learn a new word, it is clear that reading one million words per year in not enough. In order to account for a growth rate of 5,000 words in a year {see box}, it seems necessary to think about continued learning from conversational interactions supplemented by reading several million words per year. Indeed, children who read little outside the classroom generally do poorly on vocabulary tests.

The fact that children learn many more words than anyone has time to teach them also carries implications for the role of teachers in this learning process. Learning new words from purely literary contexts of use -- from the contexts provided on the printed page -- is harder than learning them through interaction with a person. In conversation it is usually possible to ask the speaker what an unfamiliar word means. Moreover, in most conversations visual information supplements the linguistic information. Such help is missing from the printed page.

One way to figure out what an unfamiliar word means is to use a dictionary. In about the fourth grade, therefore, most schools begin to teach dictionary skills: spelling, alphabetizing, pronunciation, parts of speech and a little morphology and etymology. The idea, which is perfectly reasonable, is that children should learn how to find unfamiliar words in a dictionary and how to understand what they read there.

One trouble with this approach is that most healthy, right-minded children have a strong aversion to dictionaries. There may be good reson. We have looked at some of the tasks techers assign in order to get students to use dictionaries. In our opinion these exercises do not merit the faith that teachers and parents have put in them.

Two tasks are often assigned when children are being taught how to use a dictionary. One task entails disambiguation: The child is given a sentence that contains an ambiguous word -- a word with two or more senses -- and told to find it in the dictionary and to decide which sense the author of the sentence had in mind. The other task calls for production: The child is given a word and told to look it up in the dictionary and to write a sentence incorporating it. On the face of it both tasks look as though they should be instructive. It is therefore surprising to discover how ineffectual they are.

Learning from a dictionary requires considerable sophistication. Interrupting your reading to find an unfamiliar word in an alphabetical list, all the while keeping the original context in mind so that you can compare it with the alternative senses given in the dictionary, then selecting the sense that is most appropriate in the original context -- that is a high-level cognitive task. It should not be surprising that children are not good at it. Even when most of the complications are removed, children are still not good at it. On a simplified disambiguation task, in which fourth-grade students were given just two senses and asked to choose the one that was intended in a particular sentence, the students did little better than chance.

The second task, producing a sentence incorporating a new word, has the virtue of requiring the student to use the word and so, presumably, to think about its meaning. We have studied this production task extensively. After reading several thousand sentences that were written by children in the fifth and sixth grades we have concluded that it too is a waste of time.

Typical of the curious sentences we encountered was "Mrs. Morrow stimulated the soup." It illustrates the most frequent kind of error made by children in that age range. If they already know the word, their sentences are usually all right. If the word is unfamiliar, however, the results are often mystifying. In order to understand what the child did, you have to read carefully the same dictionary definitions the child read. The child who looked up stimulate found stir up among the definitions.

The example provides a key to what happens when children consult a dictionary. They find the unfamiliar word and then look for a familiar word or phrase among the definitions. Next they compose a sentence using the familiar word or phrase and substitute the new word for it. One of our favorite examples came from a fifth-grader who looked up the unfamiliar word erode, found the familiar phrases eat out and eat away in the definition and thought of the sentence "Our family eats out a lot." She then substituted erode for eats out; the resulting sentence was "Our family erodes a lot."

Looking It Up

If children are so good at learning new words when they hear or see them used in context, why do they have trouble learning new words when they see them in a dictionary? We decided to look more closely at what goes on when an unfamiliar word is encountered in the context of a typical sentence. A preliminary study indicated that children can write better sentences when they are given a model sentence employing the word than when they are given a definition of the word. Since many of the sentences they wrote were patterned on the models, this result could not be interpreted to mean that the children learned more about the meaning of a word from illustrative sentences than they learned from definitions. Nevertheless, the observation was encouraging, and we pressed on.

The next step was simple: If one example is good, three should be better. When we made this comparison, however, we found that the number of examples made little difference. The acceptability ratings of sentences written after seeing one model sentence were the same as the ratings of sentences written on the basis of three examples. That observation made us think again about what was going on. Apparently three unrelated sentences are hard for children to integrate, and so they simply focus on one of three examples and ignore the others. This behavior resembles what children do in reading dictionary definitions.

We were surprised by one result, although perhaps in retrospect we should have expected it. Mistakes resembling simple substitutions appeared even when model sentences were given instead of dictionary definitions. For example, given the model sentence "The king's brother tried to usurp the throne" to define the unfamiliar word usurp, the children wrote such sentences as "The blue chair was usurped from the room," "Don't try to usurp that tape from the store," "The thief tried to usurp the money from the safe" and so on. They had gathered from the model sentence that usurp means take, and so they composed sentences using take and then substituted usurp for it.

Children can appreciate at least part of the meaning of an unfamiliar word from its context, as in the case of take as one component of the meaning of usurp. That is to say, if usurp is incompletely defined as take, it can be said of anything takable: chairs, tape, money or whatever. When it is seen from this perspective (as an overgeneralization), the behavior of these children in the fifth and sixth grades is merely a later stage in the development of a word-learning process employed by preschool children.

The substitution strategy therefore seems to be quite general. In the context of a model sentence, however, something more than a simple substitution error appears. The children cannot search through an illustrative sentence for a familiar word as they could in a dictionary definition. First they must abstract a familiar concept from the context of the unfamiliar word. Only then can they apply the substitution rule.

The Video Solution

Might there be a better way to foster the growth of vocabulary? What we and others have found out about the word-learning process will support some plausible suggestions. Put at the front of your mind the idea that a teacher's best friend in this endeavor is the student's motivation to discover meaning in linguistic messages. Then the problems with the traditional modes of instruction will begin to make sense. Drill on arbitrarily preselected lists of words seldom takes place at a time when the student feels a need to know those words; it fails to draw on the natural motivation for learning the associations between word and meaning. Learning through reading faces the opposite problem: not enough information about the word is available at the moment the student is motivated to learn its meaning.

What is needed is reading, which can make students curious about unfamiliar words, supplemented by immediate information about the meaning and use of those words. Ther important thing is to provide the information while the reader still wants it. Dictionaries are too slow. Recourse to a dictionary may help a mature and well-motivated student, but for the average child in the elementary grades it is likely to compound interruption with misunderstood information. A human tutor -- someone immediately available to detect and resolve lexical misunderstandings -- would be much better than a dictionary.

Given the shortage of attentive tutors to sit at every young reader's elbow, it is natural to wonder how much of the tutoring task might be carried out by a suitably programmed computer. For example, suppose reading material was presented to the student by a computer that had been programmed to answer questions about the meanings of all the words in the material. No alphabetical search would be needed: The student would simply point to a word and information about it would appear. No sophisticated disambiguation would be necessary: The computer would know in advance which particular sense of a word was appropriate in the context. Indeed, no definition would be necessary: The phrase or sentence containing the word could be rephrased to show what the word meant in the context. {Software programs of this sort are now available on the retail market.}

We are exploring some of these possibilities with a setup in which children in the fifth and sixth grades interact with a video display. They are asked to read a text that describes an episode from a motion picture they have just seen. Included in the text are certain marked words the reader is expected to learn. When one of them comes up, the child can ask for information about its meaning in any or all of three forms: definitions, sentences and pictures.

We found that providing information when it is wanted can significantly improve the children's grasp of unfamiliar words, as is demonstrated by their ability to recognize the meanings and to write acceptable sentences incorporating the words. The results reinforce our belief that much can be done with computers to make learning words easier.