HOW CHILDREN LEARN TO READ

The question of whether or not young children are aware of the existence of phonemes as units of speech has been studied extensively in relation to how children learn to read. It is obvious that awareness of the existence of phonemes as the “alphabet” of spoken language—what has come to be known as phonemic awareness—would be of enormous help in learning to use comparable elements in reading. The nature and degree of such awareness in children have been studied in several ways. Some studies, representing quite different approaches, are discussed here.

Bruce (1964) designed an experiment in which no reading was required. In total, 67 children, representing five mental age groups between 5 and 10 years, were presented orally with 30 words appropriate for their age. For each spoken

FIG. 4.5. Median growth scores for word comprehension, word production, and learning of grammar by infants as a function of age (redrawn from Bates & Goodman, 1997).

word, the child’s task was to pronounce the real word that remained if a particu-lar letter sound were to be taken away from the original word. For 10 test words, the sound removed was at the beginning, for another 10 a medial sound was omitted, and for 10 a final sound was removed. The residual words were always familiar words. Examples of words manipulated in this way, with the eliminated phoneme in parentheses, are: (n)ice, ha(n)d, star(t). The children were trained for the task by being asked to: (a) say a word, (b) make a sound, (c) indicate first, middle, and last of a group in a picture, (d) repeat the first, middle, and last of three digits spoken by the experimenter, (e) demonstrate which, and how many, bricks are left when others have been separated from them by the subject, and (f) observe modeling of several examples of the experimental task.

Figure 4.6 shows the mean correct scores, with their standard deviations, for the five age groups. It is clear that the children represented a wide range in the ability to analyze spoken words into phonemes. As the chronological ages of the children varied from 5.1 to 7.5 years, it seems possible that the results were partly, or perhaps fully, determined by differences in reading experience.

The effect of learning to read was investigated some years later more explic-itly in an experiment by Liberman et al. (1974). The test group included 46 pre-school children (mean age 4.9 years), 49 kindergarten children (mean age 5.8 years), and 40 first-grade children (mean age 6.9 years). Under the guise of a

“tapping game,” half of the children were required to repeat a word or sound

FIG. 4.6. Mean score and standard deviation of residual words, correctly pronounced by chil-dren as a function of mental age, if a particular letter sound of the words was taken away (based on data from Bruce, 1964).

spoken by the examiner and to indicate, by tapping a small wooden dowel on the table, the number of phonemes, whereas the other half indicated the num-ber of syllables. Test items varied from single vowels to three-phoneme words for the first subgroup and words of one to three syllables for the second sub-group. Figure 4.7 shows the percentage of children who, after some training, were able to tap correctly the number of phonemes or syllables in six consecu-tive items. These data suggest there is a dramatic increase in phoneme segmen-tation abilities from 0% of the preschoolers to 70% of children at the end of the first grade. Although preschool and kindergarten children are much better at identifying the number of syllables, these scores, too, improve with age, perhaps assisted by learning to read.

A quite different approach was used by Byrne and Fielding-Barnsley (1989).

They taught preliterate children to read the written equivalent of spoken words such as mat and sat; subsequently they presented a written word mow and asked the child to judge whether it stood for the pronounced word mow or sow. From the poor results, the authors concluded that both phonemic awareness and grapheme–phoneme knowledge are needed to perform this task successfully.

The experiment suggests that the ability to recognize the initial consonants of mat and mow as the “same” is not at all as “automatic” or self-evident as we

FIG. 4.7. Percentage of children of three groups able to tap correctly the number of pho-nemes or syllables of spoken words, as a function of their mean age (based on data from Liberman et al., 1974).

might think, but should be understood as an achievement related to the acquisi-tion of alphabetic reading.

These studies demonstrate clearly, as is now generally accepted, that young children learn to speak without any previous notion that words consist of pho-nemes. The studies indicate, too, not only that the awareness of phonemes co-incides chronologically with learning the alphabet but that the two are highly correlated. However, this correlation is no proof that phonemic awareness is in-duced by learning the alphabet—phonemic awareness may represent a typical stage in language acquisition that develops with age.

This question has been addressed by investigations in which adult nonreaders were presented with phoneme discrimination tasks. For example, Morais, Cary, Alegria, and Bertelson (1979) compared illiterate and literate adults in rural Por-tugal, and Read, Yun-fei, Hong-yin, and Bao-qing (1986) tested Chinese adults who had learned only Chinese characters versus others who had also learned Hanyu pinyin (writing Chinese words with alphabetic letters). In both experi-ments the subjects were requested to add individual consonants to, or delete them from, spoken real words as well as nonwords. The authors found that only subjects who had learned the alphabetic writing system in school could perform this task rather consistently (mean scores presented in Fig. 4.8). The fact that

FIG. 4.8. Percentage of adults unfamiliar (“nonalpha”) and familiar (“alpha”) with the writ-ten alphabet who were able to add or delete individual consonants from spoken words (based on data from Morais et al., 1979, in Portugal, and Read et al., 1986, in China).

some subjects defined as illiterates were partially successful might be due to the nature of the instructions or previous experience.

Further evidence that phonemic awareness may not emerge spontaneously but is the result of learning to read and write is indicated by the relative effec-tiveness of training procedures. The fact that accomplished readers seem to read words rather than strings of letters has led to the view that pho-neme–grapheme correspondences need not be taught explicitly but will be dis-covered more or less automatically by the child in the process of learning how words are written.⁵However, this has not proven to be the most effective proce-dure. More explicit attention to the one-to-one relation between phonemes and letters (at least as a first-order approximation) seems to be a crucial factor in teaching children to read (for a discussion, see Morais & Kolinsky, 1995).

It is also significant that many children who show no problems learning to talk appear to have considerable difficulties learning to read (Liberman et al., 1974). Such children (and adults) may be able to learn that the written b stands for the sound /b/, a for /æ/, and t for /t/, but cannot sound out the written word bat. On the other hand, they may be able to learn that this word as a whole is a visual symbol of the spoken word /bæt/. The number of such individuals may in fact be quite large, as much as 30–40% of the population (Lauter, 1999;

Lindamood, Bell, & Lindamood, 1997). Many of these individuals may com-pensate for their natural inability to hear out speech sounds inside syllables by an advanced proficiency for visually memorizing words like pictures, matched to spoken equivalents.

Thus beginning readers may read in a very different way than skilled readers (Gough, Juel, & Griffith, 1992). All children may begin by seeing printed words as pictographs representing spoken words comparable with pictures in a rebus.

If a language consisted of only a small number of different words, this would be a sufficient strategy. However, the difficulties of differentiating written words on the basis of their overall appearance increase dramatically as the number of words increases. Moreover, this strategy provides no means of connecting new printed words with their spoken equivalents. Hence, it is necessary to learn the letter–phoneme correspondences in order to acquire an expanding reading vo-cabulary. Happily, most children seem to have no problems in learning this.

Of course, such a strategy is useful only in so far as the phoneme–grapheme correspondence of a language provides a reliable guide—a quality that differs markedly from language to language. Finnish and Spanish are languages with a

5For an extensive review of this question, see Chall (1967).

close correspondence, whereas French and particularly English⁶are much less consistent (Downing, 1973). For instance, children have to learn that the “same” vowel in English may be spelled in different ways, as in out and crowd, and that in other cases letters appear that are not pronounced, as in know and doubt.

Alphabetic representation of speech may be misleading with regard to the

“natural” way that speech is represented internally. As Treiman and Baron (1981) said:

A second possible explanation, one that we have proposed here, is that young children do not represent speech in terms of segments, either consciously or unconsciously.

Rather, they represent the sound of a word or syllable “integrally,” as an indivisible whole. This possibility is consistent with children’s ability to hear and speak, and with their ability to judge that certain pairs of words sound different and that others are more alike. The fact that a child can do these things implies that he has some internal representation of speech—it does not tell what kind of representation. The possibility that young children represent spoken words wholistically is consistent with their ap-parent difficulties in hearing or repeating novel words (particularly those that are un-like any familiar words), and with their difficulties in telling how words sound aun-like or different. (p. 192)

In order to decide whether this explanation is right, we need to know more about the relation between speech and the alphabet. Traditionally, we are in-clined to consider the written letters as derived from the spoken phonemes, but is this all that can be said? Is it too bold to suppose that, conversely, the very con-cept of the phoneme is based, at least in part, on our familiarity with graphemes?

A study of the origins of our alphabet may help us obtain better insight into this question. Taking this historical journey may show us that the relation between written and spoken “alphabets” is much more complex than we thought.