Central to Bryant‘s argument is an understanding of just what is entailed in phonological awareness - awareness of the sounds which make up words. There are three ways of breaking up a word into constituent sounds and therefore three types of phonological awareness.

• Firstly, words can be divided up into syllables. Most language speakers, including young children (Liberman, Shankweiler, Fischer & Carter, 1974), have little difficulty in separating the words Daddy, bungalow and magnificent into two, three and four syllables respectively, and hence it can be argued that awareness of this type of phonological unit is early to emerge and quite well developed by the time children learn to read.

• A second way of dividing up words, and thus a second type of phonological awareness, is by phonemes, that is the smallest units of sound in words. The word dog, for example, has three phonemes, represented by the letters d, o and g. The word church also has three phonemes, but in this case these are each represented by letter strings (digraphs), ch, ur and ch. Advocates of phonics teaching methods lay great stress on children being taught the relationships between phonemes and the letter strings which can represent them, but the evidence suggests that young children are not actually aware of these sound units, as will be explained later.

• The third way of dividing up words, which has provided a rich source of interest for current understanding of children‘s phonological awareness, looks at units larger than phonemes but smaller than syllables. Each syllable can be divided into an opening and an end section. For example, the word throw has one syllable but two distinct parts, thr and ow. In the word chasing, each syllable also has two parts. ch-a and s-ing. These units are referred to as the onset and the rime. The use of the term, rime, for the end units makes obvious reference to the fact that words which finish with similar rimes do rhyme. It is quite clear that rhymes are a very significant part of young childrens‘ lives. They are taught them and make up their own long before they start school (Dowker, 1989) and can detect rhyme and alliteration before they begin to read (Bradley & Bryant, 1983). Their experience of and sensitivity to rhymes seems to be closely linked to their fluency in reading and to their experience of nursery rhymes in their early years (Bryant, Bradley, MacLean & Crossland, 1989).

The three types of sound-related word divisions are shown in the following table (taken from Goswami & Bryant, 1990):

Although there is plenty of evidence for a connection between children‘s reading and their awareness of sounds, this evidence does suggest that the direction of the connection is different for the different types of phonological awareness.

Research on Portuguese illiterates by Morais, Cary, Alegria & Bertelson (1979) found that these people were particularly bad at phonological tasks compared to a group of people who had been illiterates but had learned to read through an adult literacy programme. Learning to read seemed to have led to greater phonological awareness rather than phonological awareness enabling the learning of reading. Similar conclusions can be drawn from studies of readers who learned to read with alphabetic (as in the case of English) or logographic scripts (as in the case of traditional Chinese or Japanese scripts). Read, Zhang, Nie & Ding (1986) compared people who had learnt to read with traditional Chinese logographic orthography with others who had learnt with an alphabetic version of Chinese script. The logographic learners were much worse at phonological skills than the alphabetic script learners. Mann (1986) compared Japanese (logographic script learners) with American (alphabetic script learners) six year old children. The two groups were similar in their awareness of syllables but the American children much better at phoneme awareness tasks. It appears that it is learning to read with an alphabetic script that in particular leads to phonological skill.

Studies have also compared children before and after they began to learn to read. The original work of Bruce (1964), since confirmed many times, found that five and six year old children had great difficulty with tasks which involved removing a sound from a word (e.g. removing J from JAM; N from SNAIL; K from FORK). Seven year olds did better, but it was not until 8 or 9 that children managed a reasonable performance. These children‘s abilities to manipulate phonemes did not begin to develop until they had made a start on learning to read. Studies such as that of Calfee (1977) and Content, Morais, Alegria & Bertelson (1982) have, however, subsequently suggested that children do better in this type of experiment when the sound they have to remove is the onset, leaving the rime intact.

Work with tapping tasks, where children have to tap for each sound they hear in spoken words, has found that young children can tap correctly for syllables but not for phonemes (Liberman, Shankweiler, Fischer & Carter, 1974). They tend to tap out the number of letters rather than the number of sounds (Tunmer & Nesdale, 1985) which suggests that their way into understanding phonemes is through the letters which stand for them. This again supports the idea that awareness of phonemes is a result of reading experience rather than a precursor.

There is, however, also evidence that an awareness of onset and rime develops fairly early. Kirtley, Bryant, MacLean & Bradley (1989) report an experiment in which five, six and seven year olds were given tasks which involved saying which was the odd word out of groups such as:

• 1. doll, deaf, can (the oddity is in the first phoneme and the onset.)
• 2. mop, lead, whip (the oddity is in the last phoneme - a part of the rime.)
• 3. top, rail, hop (the oddity is in the rime.)
• 4. cap, doll, dog (the oddity is in the first two phonemes - crossing the onset/rime boundary.)

Groups 1 and 3 (where onset and rime were preserved) were found much easier, with even children who could not read able to have some success. Groups 2 and 4 (where onset and rime were disrupted) were only found possible by children who had begun to read. Phoneme awareness seems, therefore, to develop through learning to read, but some awareness of onset and rime seems to be already present before reading begins.

This view of the development of phoneme awareness is supported by evidence that beginning readers approach words without any analysis of phoneme-grapheme relationships, an inconvenient conclusion for advocates of phonic methods, but not damning if phonological awareness is conceived of as including onset and rime sensitivity. Bryant & Bradley (1983) found that children beginning reading seemed to adopt a visual strategy to the task, recognising words as logograms. As they became more proficient at reading they began to use phonological coding rather more. This is further evidence that reading causes phonological awareness rather than vice versa.

The interesting question arises of how, if children use a visual strategy for recognising words, can they manage to read new words? Giving children the tools to do this has long been one of the major arguments in favour of a phonics approach to reading teaching. Evidence suggests, however, that children do this, not by 'building up words' from their knowledge of phoneme-grapheme relationships, but by making analogies. Goswami (1986) has shown that even children of five years old who had not begun to read were able to do this. Her experiment involved teaching children new words such as beak, and then asking them to read other words which shared spelling features with the original word, either the same rime, e.g. weak, the same onset and part of the rime, e.g. bean, or some of the same letters in different orders, e.g. lake. It was found that children of five, six and seven years old all read words with the same rime better than the others and the difference was particularly marked in the case of the five year olds. Young children also seem to be able to make analogies on the basis of onset.

This seems to suggest that as soon as children begin to learn to read they do begin to adopt a phonological approach to the task, but that this is based on the phonological units which make most sense to them - onset and rime. Goswami (1990) also showed that there is a link between children's sensitivity to onset and rime and their abilities to use analogies in learning to read. This fits with the findings of Bradley & Bryant (1983, 1985) that there is a strong relationship between children‘s sensitivity to rhyme and alliteration when they begin school and their progress in learning to read over the following three years, even when the effects of intelligence are controlled for.

This kind of evidence is used by Goswami & Bryant (1990) to suggest a particular model for the development of reading in young children and the role of phonological awareness within it. The model proposes that children given experience of rhyme and alliteration develop an awareness of these before they begin to learn to read. They actually begin reading using only a visual approach to recognising words but very quickly their awareness of onset and rime allows them to apply analogies in order to recognise new words. Through experience of reading they begin to become aware of the phonemic basis of alphabetic script and to develop skill in using this new found awareness. Phonological awareness, in its various forms, is therefore both a precursor to and a product of reading. This model provides no support for an approach to the teaching of reading centred around phonics drills, but it does suggest that the traditional infant classroom emphasis upon rhymes (nursery or otherwise) is even more important than previously realised. There also seems to be support for the strategy of heightening beginning readers' awareness of sounds in words by pointing these out to them. Traditional 'I-spy' games and 'sound tables' etc. do seem to make a valuable contribution, but what is significant is the fact that these activities are designed actually to operate upon children's awareness rather than simply upon their implicit knowledge.