192 J. Segui and J. Grainger

Segui, J & Grainger J. (1990). Priming word recognition with orthographic neighbors: effects of relative prime-target frequency. Journal of Experimental Psychology: Human Perception and Performance, 16,1, 65-76.

Analytic Approaches to Human Cognition J. Alegria, 1). Holender, J. Junça tic Morais and M. Radeau (e<)s.) © 1992 lilsevicr Science Publishers B.V. AH rights reserved. 193

CHAPTER 12

The Causes of Phonemic Awareness

José Morais and Philippe Mousty Laboratoire de Psychologie Experimentale Université Libre dc Bruxelles 1050 Brussels - Belgium

ABSTRACT

We propose that phonemic awareness has two main causes: linguistic (phonological) development and alphabetic instruction. In this chapter, wc deal with the relations between phonological development on the one hand, and phonemic awareness and literacy abilities on the other hand. The main trend of phonological development is the progressive build up of segmentally structured representations in both the input and the output lexicons. These segmental representations, together with phonological sensitivity and phonological awareness are necessary conditions for phonemic awareness. These proposals arc consistent with an increasing number of experimental findings and clinical observations which are reviewed here. A conceptual framework, shared by a large number of researchers, is available and may fruitfully guide numerous investigations on highly specific issues.

A FOREWORD

The work the Brussels group has done, now for more than 15 years, on phonological awareness owes much to the personal contribution of Paul Bertelson. For one of the present authors (J.M.), it has been a great joy to share this intellectual enterprise with Paul, which followed almost immediately their collaborative work on the determinants of perceptual lateral differences.

Fruitful collaboration may benefit from some divergent views. Paul and J.M. disagreed a few times, just the right portion, and they adopted the usual scientific rule of resolving disagreements on empirical grounds. Thus, after Morais (1982) showed more optimism than Bertelson (1982) concerning one particular categorisation of hemispheric specialization, the facts forced both the first to co-author an embarrassing finding for that categorisation (Bertelson & Morais, 1983), and the latter to co-author another finding that supported it (Peretz, Morais, & Bertelson, 1987).

The present chapter puts up a strong case for the perceptual roots of phonemic awareness. We suspect that Paul would not put it up as we do. In a recent paper with B. de Gelder, he discusses the question of the origin of

194 J. Murai': and P. Mousty

phonological awareness, and he considers that the data showing differences in speech recognition as a function of reading ability did not necessarily arise at the perceptual level hut "may have arisen at the postperceptual level" (Bertelson & de Gelder, 1991). Will this apparent disagreement lead to new findings? To raise a discrepancy was the best way we found to dedicate a chapter to Paul, and we believe he will like the way if not the claim we make.

INTRODUCTION

The Brussels group found that ex-illiterate adults can manipulate phonemes intentionally, but illiterate adults cannot (Morais, Cary, Alegria, & Bertelson, 1979). This demonstrated that learning to read and write in the alphabetic system is a cause of phonemic awareness. In Morais et al. (1979), we put forward the interactionist position (as we coined it later on in e.g., Morais, Alegria, & Content, 1987) on the relationships between phonemic awareness and alphabetic literacy: Alphabetic literacy is both a cause and a consequence of phonemic awareness.

The relations between alphabetic literacy and phonemic awareness are so intricate that to disentangle the precise causal influences of each skill on the other is a hard task. In the present paper we will not deal with the host of literacy factors (degree of precocity of the acquaintance with letters, instruction methods, degree of regularity of grapheme-phoneme correspondences, etc.) which may be considered to have a causal influence on the acquisition of phonemic awareness. We will rather be concerned with the conditions that must be present to allow the child to take advantage of being exposed to alphabetic material and receiving instruction on the alphabetic code in order to construct conscious representations of phonemes.

Such conditions may in principle be related to either language capacities or general cognitive abilities, or even both. In a recent paper. Fowler (1991) indicates that, according to the traditional view of phonemic awareness, words are implicitly represented as sequences of phonemes, and therefore becoming able to represent phonemes explicitly is strictly a matter of metalinguistic development. Fowler argues against this view; we have done so too (e.g., Morais et al., 1987) and we will continue the argument in the present paper.

Phonemic awareness is a metalinguistic capacity. However, rather than evolving from other, more holistic, metalinguistic capacities as a consequence of the gradual improvement of a general cognitive ability of analysis, phonemic awareness may he contingent both on representational changes within ihe language system, and on instruction of the alphabetic code. To put it in simple words, we claim that the causes of phonemic awareness are mainly two: linguistic development and alphabetic instruction. Both are necessary, but not sufficient.

Vie Causes of Phonemic A warenexs 195

There is empirical evidence against the notion that phonemic awareness is simply a sophisticated analytic accomplishment. For instance, we found that dyslexic children are much poorer than normal readers in the deletion of the initial consonant of a short utterance. However, the dyslexies performed at the same level as the normal readers on a task requiring them to delete the initial note of a short sequence of notes (Morais, Cluytens, & Alegria, 1984). This finding suggests that dyslexies (who have not developed phonemic awareness) "are not deficient in general analytic ability, but in analytic skills that are specific to phonology" (as Kitz & Tarver, 1989, phrased our conclusion). Indeed, we never heard that dyslexies cannot display highly analytic capacity in any functional domain.

The question of whether low-intelligence children can develop phonemic awareness should deserve thorough investigation. As far as we can infer the presence of phonemic awareness from the ability to read nonwords, we may say that the answer is yes. There are many low-intelligence children who can read nonwords (this is the case, for instance, of hyperlexic children), as well as many severely disabled readers who are highly intelligent.

Thus, we take for granted that phonemic awareness is not a matter of intelligence or of general cognitive abilities. Difficulties in developing phonemic awareness in spite of appropriate instruction in the alphabetic code are difficulties in cognitive access to the outputs of the phonological processing system. They are however not related to the cognitive capacities involved in this access. Rather, they must stem from some deficiency in the phonological representations available to consciousness.

In this chapter, we will attempt to discuss how phonemic awareness may depend on language capacity, and more specifically on phonological development. We will do this by referring to the most relevant empirical data on the basis of which one can tackle the following issues:

1. Are difficulties in developing phonemic awareness related to deficiencies in current phonological processing?;

2. how specific is the association between phonemic awareness and literacy on the one hand, and speech processing on the other hand?:

3. what kind of representations of speech are necessary to acquire phonemic awareness?;

4. are the anomalies in speech processing which are observed in dyslexies a consequence rather than a cause of literacy level?: and

5. how does the beginning reader proceed to acquire conscious representations of phonemes?

196 J. Morais and P. Mousty

QUESTIONS

/. Are difficulties in developing phonemic awareness related to deficiencies in current phonological processing ?

In the last ten years, evidence has been accumulated suggesting that dyslexies and, more generally, poor readers may have phonological deficits. Of course, we are not referring here to deficits in metaphonological processing or in the phonological processing of written language, but to deficits in the phonological system involved in the comprehension and production of spoken language, so that in the rest of this paper, for the sake of convenience, we will mean the latter system when we use the expressions phonological system, phonological processing or phonological representations.

Differences in speech recognition between good and poor readers, or between dyslexies and normal readers, have been observed in several investigations.

Some of these studies dealt with the issue of categorical perception. Identification and discrimination curves for the stop place of articulation continua were found to be less steep in dyslexies and poor readers than in control subjects (Godfrey, Syrdal-Lasky, Millay, & Knox. 1981; Reed, 1989; Werker & Tees. 1987), though the differences between the groups were not always significant or consistent (Brandt & Rosen, 1980; De Weirdt, 1988).

Other studies used word, nonword or phrase repetition tasks. They showed higher scores in the normal readers than in the reading-impaired subjects (Brady, Shankweiler, & Mann, 1983; Brady, Poggie, & Rapala, 1989; Snowling, J981; Snowling, Goulandris, Bowlby & Howell, 1986; Cans, 1989; Kamhi, Cans, Mauer, Apel, & Gentry, 1988; Taylor, Lean, & Schwartz. 1989).

Phonological discrimination as involved in same-different classification of two speech episodes also yields convergent findings. Hurford and Sanders (1990) used pairs of CV syllables differing or not in the consonant (e.g., /gi-bi/ followed by /gi-bi/ or /bi-gi/) and found that second-grade normal readers performed at a much higher level than second-grade children with reading disabilities.

A task involving judgment of the temporal order of two stop consonant syllables differing in the place of articulation and presented at short interstimulus intervals showed a detrimental effect of decreasing interval in the reading disabled group, not in the normal readers group (Reed. 1989). The temporal order judgment may not be the critical factor of the difference between the groups, since this difference was observed even at the longest interval. Moreover, the reading disabled group was also poorer than the normal readers in a picture-word matching task implying to discriminate two

T7tt> Causes of Phonemic Awareness 197

picture names differing in the place of articulation of the initial stop. Similar findings were obtained by Tallal and Piercy (1974) in children with developmental aphasia.

Problems in perception of time-compressed sentential speech also contribute to identify children with reading problems (Freeman & Beasley, 1978) and to distinguish between good and poor readers from the same first-grade class (Watson, Stewart, Krause, & Rastatter, 1990). In the latter study, reading level did not interact with the type of phonetic contrast. This is an issue that should be explored further, since all the categorical perception studies we have mentioned earlier, as well as the phonological discrimination studies of Reed (1989) and Hurford and Sanders (1990), only focused on the place of articulation. Moreover, even though recognition of place of the stop consonants may be a major source of perceptual anomaly in dyslexic children, this may not be due to a general deficiency ofthe categorisation process. As a matter of fact, Lieberman (1986) could identify subgroups of dyslexies characterized by the particular place distinction that made problem in speech perception.

Finally, de Gelder and Vroomen (personal communication) found much worse lip-reading performance in dyslexic children than in reading-age controls, indicating that the processing of visual speech information may be relevant to reading ability.

All the findings described so far clearly point to an association between reading (and presumably phonemic awareness) on the one hand, and speech processing on the other hand. It should be remarked that the speech abilities tested in the studies mentioned above are only part, perhaps a small part, of the whole process of language comprehension. In other words, a child who suffers from some anomaly in one or more of those abilities may not display an apparent impairment in language comprehension, which would be detectable by the parents, the school teacher, or the clinician. It remains that subtle anomalies in speech representation may cause difficulties for the child who ought to analyze the speech representation as a sequence of phonemic units.

Furthermore, one should not overlook the fact that most children diagnosed as dyslexies present speech anomalies, especially in production, which are much more apparent than those described above. In Belgium, for instance, the great majority of the children who attend special classes for dyslexia remediation had been referred for speech therapy.

Stackhouse and Wells (1991), in England, illustrated perfectly this type of population with the case of Michael. Michael's speech was distorted and inconsistent. Although he had been taught phonics, he was unable to use it in reading, and he actually read words on the basis of their visual appearance. Of course, "his auditory discrimination, rhyme, syllable and consonant segmentation skills were all very poor" (p. 187). Hut, as (he authors warn.

I9H J. Morals and P. Mousty

most dyslexies have a speech disorder which is much less severe than Michael's and which can be neglected. In these cases, it is often après coup that the speech anomaly, if noticed, is related to the reading difficulties. Stackhouse and Wells illustrate this kind of dyslexic with Richard, a child of above average intelligence but with literacy skills lagging by three years behind his chronological age. His speech was apparently normal. He had no difficulty on the usual tests of auditory discrimination, nor on rhyme identification. However, his performance on consonant segmentation and on spoonerisms was poor, and his use of phonological decoding in reading was inefficient. The authors' investigation of Richard's speech showed that he had abnormal difficulty repeating multisyllabic words and nonwords. The perseveration, substitution and transposition of consonants in speech strikingly resembled the errors made in reading nonwords.

Indeed. Michael and Richard are only two cases, but they are representative of large numbers of dyslexies. Thus, both the experimental data and the analysis of clinical data compel one to give the question heading this section a definitely affirmative answer.

2. How specific is the association between phonemic awareness and literacy on the one hand, and speech processing on the other hand ?

There are several indications that the differences between reading disabled children and normal readers are not limited to the processing of speech stimuli. Tallal (1980) found reading disabled children to be much worse than normal readers in making temporal order judgments for very brief nonspeech tones. In a similar vein, Reed (1989) found group differences for very brief tones as well as for stop consonants, but not for vowels which have more long-lasting acoustic cues. Group differences did not come out for vowels even when recognition was made more difficult by using a white noise background. Interestingly, group differences did not come out either when visual stimuli of the same duration as the tones were used in an analogous test. This finding suggests that ihe deficiency of reading disabled readers in processing rapidly presented information is limited to the auditory modality. In this context, the observation by de Gelder and Vroomen (see above) of an association between reading and lip-reading may be considered as being due not to the short duration of the facial cues but rather to the phonetic nature of this information.

There are other indications of an association between processing of the temporal cues of nonspeech sounds and reading level. Both Zurif and Carson (1970), and McGivern. Berka. Languis, and Chapman (1991) found the ability to discriminate the temporal patterns of the Seashore Rhythm Test to be impaired in reading disabled children. Unfortunately, neither of these studies provides the precise duration of tones and intervals. Anyway, it seems that the

Tlie Causes of Phonemic Awareness 199

deficiency of reading disabled children in processing transient auditory cues is not speech-dependent.

Based on this notion, one might be tempted to suggest that the deficiency in processing brief sounds is the cause of difficulties in acquisition of both phonemic awareness and alphabetic literacy. This might be a misleading suggestion. Reading disabled children might have trouble in consciously identifying the phonemes contained in their perceptual representations of speech, not because they have a general auditory impairment, but because their phonological representations are not sufficiently adequate to the conscious identification of phonemes. An analogy with the effects of deafness might help to illuminate this reasoning. Deaf children would have difficulty in acquiring phonemic awareness, not because they are deaf, but because, being deaf, they have not developed a phonological processing system capable of elaborating representations of speech of enough quality. In more epistemological terms, remote causes lose explanatory power.

The different subsystems concerned with the processing of information coming through the ears, although possibly sharing similar or identical processes, shall at least differ in the specific knowledge they employ to deliver an output that is informationally different from the input they receive. Each of these subsystems may suffer from some transmission deficit or from some very early processing deficit. Thus, a deficiency in. say, accurately processing onset time may have detrimental effects in localizing sound sources, in segregating them, as well as in distinguishing voiced from unvoiced speech sounds. In this case a pre-phonetic anomaly may have a phonetic effect, and this phonetic effect may be cause of further deficits in later processing.

Alternatively, a simultaneous impairment of several susbsystems may also occur independent of sequential constraints in processing. Neurological damage may affect regions (or neurons of a particular kind) serving several subsystems. Some types of processing may thus be affected for reasons that have nothing to do with sequential constraints on processing. To take again the case of onset time processing, difficulties in source segregation and in voicing identification may not result from an antecedent processing anomaly. If this is true, then there would be no pre-phonetic cause of the phonetic effect.

It is presently impossible to say whether the origin ofthe anomaly affecting speech perception in reading disabled children is auditory (pre-phonetic) or phonetic. What we know for sure is that these children present phonological impairments, and we can easily understand that it may be difficult to elaborate conscious representations of phonemes from poor speech representations. Although there may be concomitant deficits in processing tone patterns, the perceptual representations of these patterns entai! a different processing line; they underlie musical abilities, but not metaphonological ones.

200 J. Morais and P. Mousty

The notion that difficulties in acquiring phonemic awareness may be due to anomalies in speech perception has important practical implications. It means that, by improving speech perception, one might obtain some progress in conscious phonemic segmentation ability in those children. The test has been made recently (see Hurford & Sanders, 1990; Hurford, 1990). First, Hurford and Sanders showed that phonemic discrimination can be improved in children with reading disabilities. They trained them to discriminate speech stimuli, initially sounds without transitions, and later on sounds with transitions progressively briefer (vowels, then cv syllables beginning with liquids, and finally cv syllables beginning with stops). The post-training discrimination performance of the children with reading disabilities was significantly better than their pre-training performance. More importantly, their post-training scores were not lower than the scores obtained in an assessment test by a group of children without reading disabilities.

In a further study, Hurford (1990) examined the effect of phonemic discrimination training on conscious phonemic segmentation. Phonemic segmentation was assessed with a deletion test of the type .Toy bug without the sound Ibf. The results showed that, contrary to control reading disabled subjects, the reading disabled children trained on phonemic discrimination strongly improved in conscious segmentation from assessment to post-training.

This finding is both important and encouraging. On the theoretical side, it strengthens of course the proposal of a relationship between speech perception and phonemic awareness. Unfortunately, by itself, it does not allow us to identify the origin ofthe perceptual effect at either the phonological processing system or a previous stage. A transfer effect of discrimination training might be observed even if there were an early deficit in processing either brief sounds or short temporal relations, or both, before the intervention of speech-specific processes. However, this issue might be examined. It would be particularly relevant to determine if training on tone discrimination also has a favourable effect on conscious phonemic segmentation. No such effect would be predicted on the basis of the specificity hypothesis. If the deficits in the speech perception system are highly specific, then their treatment must also be specific.

On the practical side, the finding that training discrimination has an effect on conscious segmentation means that helping reading disabled children to overcome their difficulties in speech processing and, consequently, in acquiring phonemic awareness, is not hopeless. Indeed, we are not saying that the whole problem of dyslexia is a matter of training speech discrimination. Not only may there be other reasons of difficulty, but in addition the effects of discrimination training on the quality of the perceptual representations may be more limited than suggested by Hurford's results. It should be recalled that dyslexic adults, in spite of extensive training of different sorts, remain impaired in some metaphonological tests which are particularly difficult. For instance, both error rate and response time in reversal of three-phoneme words

The Causes of Phonemic Awareness 201

(say tap for pat) are much higher in dyslexic university students than in their nondyslexic peers, even though the dyslexies had received extensive remedial training (Kitz & Tarver, 1989). The best dyslexic on this task was still worse than the worst nondyslexic.

The reversal task is probably difficult for dyslexies because they have both to discriminate between the two consonants of the CVC syllable and to encode them in an abstract form so that their exchange can be obtained. It would be interesting to know whether the difficulty of such reversals increases with the perceptual confusability ofthe consonants. Anyway, the extreme difficulty of dyslexies in reversing phonemes may be the consequence of a relatively holistic way of recognizing spoken words. As we will discuss in the next section, the holistic kind of processing may characterize very young children, but be inappropriate for children approaching the traditional age of literacy onset.

3. Wfutt kind of representations of speech are necessary to acquire phonemic awareness ?

One idea developed by Luce and Luce (1990) about spoken word recognition in young children is that either they perform nothing but a gross encoding which may ignore much ofthe fine acoustic-phonetic detail, or, more likely, that they represent this detail as a set of only loosely organized phonetic features. If the phonetic information is not organized in a linear sequence of segment-sized bundles, then it may be difficult not only to distinguish words like tap and pat but also to analyze their phonemic constituents. Luce and Luce (1990) showed that words in the 5-year-olds' lexicon have fewer similar neighbours than the same words analyzed in the 7-year-olds' lexicon, and of course than in the adults' lexicon. The increase of the lexicon requires the ability to represent words as linear sequences of segments so that discrimination among neighbour words can be optimized.

There are several empirical indications of a developmental change, perhaps associated with the increase of the lexicon, from less segmental toward more segmental representations in both speech perception and production. For instance, Nittrouer and Studdert-Kennedy (1986) found that 3-year-old, and to a lesser extent 5-year-old children are less sensitive than adults to the steady-state friction of fricatives as well as to the vocalic formants, and more sensitive to the transitions that depend on both segments. In production, Nittrouer, Studdert-Kennedy, and McGowan (1988). and McGowan and Nittrouer (1988) found greater effects of vowel context within the fricative noise of 7-year-old children and younger than of adults. Thus, children would contrast phonetic segments less clearly than adults.

The presence of less sharply defined segments in the speech representations of dyslexies and poor readers than in normal readers might explain some facts observed in spoken word and nonword recognition. For instance. Snowling et

202 J. Morals and P. Mousty

al. (1986) showed that the reading disabled children are inferior to their age-matched controls at recognizing auditorily presented nonwords and low-frequency words, but not at recognizing high-frequency words. Thus, knowledge of words may compensate for some ambiguity in the phonological structure derived from the input on the basis of the sole acoustic-phonetic features. When compared to reading-age controls, the inferiority of the reading disabled children should not occur with low-frequency words, but only with nonwords, and this is actually the case. The analogy with written word recognition should be noticed: With spoken as with written items, the inferiority of dyslexies and poor readers is the smallest for high-frequency words, the largest for nonwords, and intermediate for low-frequency words.

In a similar vein, Reed (1989) found a large effect of lexical status on phonetic categorisation near the category boundary (displacing the boundary toward the word in a word-nonword continuum, for example the continuum between /task/ and /dask/) in reading disabled children, but not in normal readers. In other words, the reading disabled children are more sensitive than the normal readers to the lexical information, and they use it to disambiguate their perceptions, probably because their phonetic distinctions are less sharply defined. It should be noticed, in addition, that this finding does not concern the organization of phonetic information into a segmental line, but rather the extraction of the phonetic features themselves. Dyslexies might thus be impaired in both the specification and the organization of features.

So, what kind of speech representations are necessary to become aware of phonemes ? The answer seems to be representations in which the phonetic featural information is organized into segmental bundles, so that the child can use these bundles, when provided with external symbols for them, as cues for the conscious extraction of phonemes.

4. Are the anomalies in speech processing which are observed in dyslexies a consequence rather than a cause of literacy level ?

The simple observation of an association between speech recognition and literacy is not enough to conclude that poor abilities in speech perception must be responsible for literacy impairments. After all, the benefit of literacy proficiency is to be able to handle more information more efficiently. Literates may have developed more efficient strategies to process speech. For instance, they may use their phonemic awareness to focus on the phonemic structure of the words they listen to (see Morais, Castro, Scliar-Cabral, Kolinsky, & Content, 1987). There is thus some reason to admit the possibility of an influence of literacy on speech processing rather than the reverse, although that possibility is usually neglected.

One way to assess the direction of the possible causal influence is to •"are literates and illiterates on the same tasks that reveal an inferiority in ;xics as compared to normal readers. Indeed, if the same inferiority is not

Vie Causes of Phonemic Awareness 203

observed in illiterates, then the possibility of an influence of literacy on speech perception could be discarded.

Our research group has been gathering some data on the speech recognition abilities of illiterates. Here, we will briefly review findings obtained for three kinds of ability, namely categorical perception, lip-reading, and repetition, which are especially relevant in the present context.

Identification of stimuli along the acoustic continuum from /ba/ to /da/ did not show a significant difference between illiterate and literate people of about the same age in the steepness of the curve (Morais, de Gelder, & Verhaeghe, in preparation). This absence of a difference was replicated by Castro and Morais (in preparation) for voicing contrasts (see Morais, Castro, & Kolinsky, 1991, for further details).

In the experiment in which we examined the categorical perception of the continuum /ba-da/, we also presented the same subjects with a face silently pronouncing ba or da. Identification of these sounds through lip-reading yielded similar scores in the two literacy groups.

Thus, one may conclude that categorical perception and lip-reading, in other words abilities on which dyslexies and normal readers show striking differences, are not a consequence of literacy level. This finding strengthens the idea that anomalies in categorical perception and lip-reading play a causal role in reading disabilities, most probably by making the acquisition of phonemic awareness particularly difficult.

In collaboration with Ruth Campbell and Luz Cary, we also compared literates, illiterates, and ex-illiterates of the same age, on word and nonword repetition. Literate people obtained higher scores than the other groups on the average (see Table 1). Unfortunately, we could not perform an audiometric assessment ofthe subjects. If one makes the reasonable assumption that literate subjects, who belong to a higher social class than the illiterates, are in a better audiological condition, then the observation of a difference can be interpreted in a trivial way. More interesting is the fact that between illiterates and ex-illiterates there was no significant difference either in word or in nonword repetition. Ex-illiterates are people who can read and write, although in a rudimentary way. They can analyze utterances into phonemes, as indicated by

Table I

I l l i t e r a t e s E x - i t l i t e r a t e s L i te ra tes

Words 94.8 94.4 100.0 Nonwords 51.1 55.1 77.8

Percentages of correct word and nonword repetition by literates, ex-illiterates, and illiterates.

?nt J. Morais and P. Mousrv

the relatively high scores this population usually obtains in tasks such as phoneme deletion and addition (Morais et al., 1979; Morais, Bertelson, Cary, & Alegria, 1986). Thus, phonemic awareness per se does not seem to be instrumental in ameliorating repetition ability.

The deficiency in repetition displayed by dyslexies and poor readers compared to normal readers should not be attributed to an inferiority in conscious phonemic segmentation. Snowling et al. (1986) interpreted that deficiency in repetition as resulting from difficulty with conscious phoneme segmentation. According to them, "the dyslexies' difficulty does not appear to be one of speech perception" (p. 504). However, the lack of difference in repetition between illiterates and ex-ill iterates, coupled with the huge difference in phonemic segmentation between the same populations, undermines this view. On the contrary, the deficiency of dyslexies in repetition would reflect a more basic phenomenon, presumably at the level of phonological representations. Fowler (1991) proposes that the segmental organization helps to assign novel stimuli into a recoverable representation and thus helps nonword repetition.

The findings described here suggest that there are speech processing anomalies which are the cause, not consequence, of difficulty in acquiring phonemic awareness. This does not imply that the causal relations between speech processing and phonemic awareness are unidirectional, always from the former to the latter (for some evidence of the reverse causality, see Morais et al.. 1991).

5. How does the beginning reader proceed to acquire conscious representations of phonemes ?

In order to answer this question, the evidence obtained by studying illiterate people can be of some help. Illiterates are. if anything, very poor at all sorts of phonemic manipulation tasks. However, they display some sensitivity to phonetic properties. One piece of evidence of this phonetic sensitivity is provided by the detection task, which can in principle be accomplished without any formal analysis. Our group has used so far several versions of the detection task, for example, to detect a phoneme target within a sentence, and to detect it in an isolated word, but in this case with the target varying from trial to trial. With these procedures, we always observed relatively poor performance in illiterates. In collaboration with Inacio Fiadeiro and Luz Cary. we tried a much easier version ofthe detection task. Not only the stimulus was an isolated word, but in addition the target was always the same in a block of trials and it occupied always the same position in the stimulus when it occurred. When this position was word-initial, performance was clearly above chance (see Table 2). The successful matching of the phoneme target with the word stimulus indicates that there is some sensitivity to phonetic cues spanning one phoneme. Sensitivity to more extended

Vie Causes of Phonemic Awareness 205

Table 2

Position Target In i t ia l Non-initial

/pa/ 92 66 / p / 83 66

Percentages of correct detection of /pa/ and Ipl targets in initial and non-initial position in the carrier-word (chance level = 50%).

phonological similarity was already documented by the illiterates' ability to appreciate rhyme (see Morais et al., 1986).

Cues for segments might be obtained not only at the phonetic level, but also by paying attention to the articulatory gestures. There is dramatic evidence of dyslexies being very poor in articulatory awareness. Montgomery (1981) asked dyslexic children and 9-year-old normal readers matched for reading to indicate which of several schematic drawings corresponded to the position of their tongue, teeth, and lips for a given phoneme. The normal readers performed at about 80% correct, whereas the dyslexies were much poorer (22% and 51% in two different samples). Our group, in collaboration with Luz Cary, has recently employed this test with illiterate adults. They also display clear evidence of articulatory awareness (72% correct, on the average, in the same test). This relatively high level of performance in awareness of articulation contrasts strikingly with the illiterates' inability to manipulate phonemes intentionally. Thus, the inability of dyslexies to interpret their articulatory gestures may reflect a cause rather than a consequence of their low reading level. Both replication of the dyslexies' performance in this test and more detailed analysis of their lack of articulatory awareness should, indeed, be sought.

The resort to both phonetic and articulatory cues is emphasi7ed in several training programs, for instance in the one developed hy Lindamood and Lindamood (1967), which seems to produce interesting clinical results. In this program, the reading disabled subject is led to focus on the oral-motor information and then on the combination of this information with the corresponding phonetic and visual information, in an attempt to help the learner recognize that syllables are sequences of phonemes.

CONCLUSIONS

The original paper of our group on phonemic awareness (Morais et al.. 1979) claimed that phonemic awareness does not arise spontaneously from linguistic experience. We still strongly support this discontinuity view. On the other hand, phonemic awareness is not a mental creation ex nihilo. Moreover,

111!) J. Morais and P. M ousts

one cannot explain conscious representations of phonemes by invoking letters, nor letters by invoking conscious representations of phonemes. Both have to be explained together as two aspects of a single, non-obligatory step in a developmental process. Thus, we also hold the idea that phonemic awareness is a metalinguistic capacity with both linguistic and metalinguistic precursors.

The experimental findings described in the previous five sections of this paper suggest that, in order to understand both failure and success in the acquisition of phonemic awareness, we have to take into account the underlying phonological representations, and especially the information contained in the conscious perceptual representations of speech. Much of the understanding of the causes of phonemic awareness will come from the advances to be made in the study of how children represent and process speech before they start learning to read.

What are the developmental precursors of phonemic awareness ? The most important one must be the developmental changes in the mode of perceiving and producing speech, from less segmental toward more segmental representations.

Why does this segmental organization help in the acquisition of conscious representations of phonemes ? Probably because the conscious representations of speech contain cues of this segmental organization. This means that we do not need to hypothesize that the preconscious representations of the speech system are accessed or made conscious. Fricatives, for instance, are less encoded in the acoustic stream than stops, and therefore the phonetic and articulatory cues of fricatives are more apparent than those of stops. Accordingly, Content (1985) found in prereaders much better performance on the segmentation of fricatives than on the segmentation of stops.

The fact that the representations of speech progressively acquire a segmental format contributes both to lexical expansion and to the development of phonological sensitivity. We call phonological sensitivity the capacity to distinguish between phonologicaliy similar words while listening for meaning. Phonological awareness is more than mere phonological sensitivity, because it implies attention to the phonological forms of utterances. Self-corrections and . language play, present even before two years of age, may be examples of relatively crude phonological awareness. David, the first author's child, loved to distort words by substituting vowels in a systematic way (e.g.. /pi3ama/. /pi3imi/ . /pi3umu/) (around 2 years). More reflexive thought on phonological relations appeared much later (3;5 years): lor example, "/buJV is like /duj/". 'Ihese manifestations of phonological interest may depend on the constitution of segmental representations of words, and they may contribute en retour to the consolidation of such representations.

Phonological awareness is necessary, for example, to perform usual phonetic discrimination tasks, in which nonwords differing on a single phoneme or even on a single feature are presented for same-different

7he Causes of Phonemic Awareness 207

classification. This kind of capacity is present in illiterates. Our group recently conducted two investigations of minimal pairs discrimination and phonemic analysis, one in Brazil in collaboration with Leonor Scliar and Luisa Nepomuceno, and the other in Spain in collaboration with José Antonio Adrian. The results are clearcut. Performance was null or almost null on phonemic analysis, but it was perfect or approached the possible maximum score on phonetic discrimination.

Phonemic awareness is thus not the kind of phonological awareness that is required to discriminate nonsense minimal pairs. The acquisition of phonemic awareness may not depend critically on the development of a particular form of phonological awareness, but it is difficult to figure out how phonemic awareness could ever be developed if the child never focused his/her attention on pronunciation instead of on meaning.

Thus, the developmental changes preparing the conditions for successful acquisition of phonemic awareness in the alphabetic literacy setting are segmentally structured representations of speech, phonological sensitivity, and phonological awareness. All these factors may interact in complex ways.

Since the segmentally structured representations of speech provide the most basic condition of phonemic awareness, it is important to discuss briefly at least one issue, namely the place of these representations within the speech processing system. Most ofthe experimental evidence described earlier in this paper concerns speech perception rather than production. However, we have seen that articulatory cues may be as useful as phonetic ones for the acquisition of phonemic awareness. Our belief, supported by recent advances in the study of phonological development, is that perception and production shape each other from an early age. Segmentally structured representations develop most probably both in the input lexicon and in the output lexicon.

There is increasing evidence ofthe influence ofthe specific characteristics ofthe ambient language on the child's babbling repertoire. For instance, there would be a very early development of isosyllabicity in the babbling of French infants, as suggested by one case studied by Levitt and Utman (1992). Furthermore, de Boysson-Bardies has demonstrated the existence of language differences in the babbling by 10 months, both for vowels (de Boysson-Bardies. Halle. Sagart, & Durand. 1989) and for consonants (de Boysson-Bardies & Vihman. 1991). The articulatory patterns created during the babbling period may in turn create templates for the purpose of speech identification. In later periods, marked by the representation of intra-syllabic constituents, the input and the output lexicons may be supposed to develop similarly.

The existence of two separate, though presumably not independent lexicons is supported in particular by neuropsychological data obtained in the adult (the most dramatic case is word meaning deafness in the presence of intact spontaneous speech, see Bramwell, 1897). This separation allows selective

208 J. Morais and P. Mousty

impairments to occur also in development. For instance, J.M., a phonological dyslexic described by Snowling and Hulme (1989), performed at normal level on nonword discrimination but was severely impaired in nonword repetition. According to Snowling, Chiat, and Hulme (1991), he has specific difficulties with output phonological processes.

Word recognition impairments may be at the origin of difficulties at developing phonemic awareness, even at an elementary level. The extreme case is the deaf child, who can only develop phonological representations, including phonological awareness, by relying heavily on articulation and on external aids like lip-reading and Cued Speech (see Alegria, Leybaert, Charlier, & Hage, this volume). Extreme deprivation of articulatory possibilities does not seem to be so harmful as extreme auditory deprivation for phonemic representation (see Bishop & Robson, 1989).

On the other hand, less severe output deficits, apparently in the absence of input deficits, may lead to disorders in both high forms of phonemic segmentation abilities and literacy abilities. Richard, the dyslexic studied by Stackhouse and Wells (1991), had problems with sequencing consonants in nonword production; he also had problems with some types of phonemic awareness tests and with phonological assembling in reading. Brady et al. (1989) found that repeating quickly short sequences which can be produced easily in isolation is a more difficult task for poor readers and young children than for good readers and older children. Children who make deviant errors in speech may both acquire phonemic awareness and be able to use simple rules of grapheme-phoneme correspondence in reading, but they have trouble with the use of context-dependent rules (Dodd, Leahy & Hambly, 1989). Briefly, all these data seem to converge upon the idea that, behind specific deficits in literacy skiUs and in the abilities of phonemic manipulation there may be specific deficits in the primary phonological representations of speech, particularly in their segmental organization. It is the exact correspondences between the abilities at these two levels ofthe language system that one should look for.

We cannot better conclude this chapter than borrowing Fowler's (1991) words: "it may be that extending our investigation to basic phonological development will provide a more complete account of the ontogenesis of phonological awareness and of reading disability than has been available to date" (p. 98).

ACKNOWLEDGEMENTS

Preparation of this paper was supported by the Belgian "Fonds de la Recherche Fondamentale Collective" (convention n° 2.4562.86) and "Fonds National de la Recherche Scientifique-Loterie Nationale" (conventions n° 8.4527.90 and 8.4505.92).

77ur Causes of Phonemic Awareness 209

REFERENCES

Alegria, J-, Leybaert, J., Charlier, B., & Mage, C. (1992). On the origin of phonological representations in the deaf: Hearing lips and hands. In J. Alegria, D. Holender, J. Morais, & R. Radeau (Eds.), Analytic approaches to human cognition (pp. 107-131). Amsterdam: North Holland.

Bertelson, P. (1982). Lateral differences in normal man and lateralization of brain function. International Journal of Psychology, 17,173-210.

Bertelson, P., & de Gelder, B. (1991). The emergence of phonological awareness: Comparative approaches. In I. (J. Mattingly & M. Sluddcrt-Kcnncdy (Eds.), Modularity and the Motor Theory of Speech Perception (pp. 393-412). Hillsdale, N J.: Lawrence Erlbaum Ass.

Bertelson, P., & Morais, J. (1983). A Ponzo-likc illusion left and right of fixation: A failed prediction. Neuropsychologia, 21,105-109.

Bishop, D. V. M., & Robson, J. (1989). Accurate non-word spelling despite congenital inability to speak: Phoncmc-graphcmc conversion does not require subvocal articulation. British Journal of Psychology, 80,1-13.

Brady, S., Shankweiler, P., & Mann, V. (1983). Speech perception and memory coding in relation to reading ability. Journal of Experimental Quid Psychology, 35, 345-367.

Brady, S., Poggic, E., & Rapala, M. M. (1989). Speech repetition abilities in children who differ in reading skill. Language and Speech, 32(2), 109-122.

Bramwcll, B. (1897). Illustrative cases of aphasia. I^ancct, /, 1256-1259. (Reprinted in Cognitive Neuropsychology, 1984, 1, 245-258).

Brandi, J., &. Rosen, J. J. (1980). Auditory phonemic perception in dyslexia: Categorical identification and discrimination of slop consonants. Brain and Language, 9, 324-337.

Castro, S. L., & Morais, J. (in prepcration). The development of dichotic word recognition: An error analysis.

Calls, H. (1989). Speech production deficits in developmental dyslexia. Journal of Speech and Hearing Disorders, 54,422-428.

Content, A. (1985). L'analyse segmentate de la parole chez l'enfant. Unpublished Doctoral Dissertation. Université Libre de Bruxelles,

dc Boysson-Bardies, B., Halle. P., Sagart, L„ & Durand, C. (1989). A cross-linguistic investigation of vowel formants in babbling. Journal of Child Language, 16, 1-17.

dc Boysson-Bardies, B-, & Vihman, M. M. (1991). Adaptation to language: Evidence from babbling and first words in four languages, language, 67 (2), 297-319.

Dc Wcirdt, W. (1988). Speech perception and frequency discrimination in good and poor readers. Applied Psycholinguistics, 9, 163-183.

Dodd, B., Leahy, J„ &. Hambly, (',. (1989). Phonological disorders in children: Underlying cognitive deficits. British Journal of Developmental Psychology, 7, 55-71.

Fowler, A. E. (1991). How early phonological development might set the stage for phoneme awareness? In S. A. Brady & D. P. Shankweiler (Eds.), Phonological Processes in Literacy: A tribute to Isabelle Y. Liberman (pp. 97-117). Hillsdale, N.J.: Lawrence Erlbaum Associates.

210 J. Morals and P. Mousty

Freeman, B. A., & Bcaslcy, D. S. (1978). Discrimination of lime-ailcred scntcnlial approximations and monosyllables by children with reading problems. Journal of Speech and Hearing Research, 21,497-506.

Godfrey, J. J., Syrdal-Lasky, A. K., Millay, K. K., & Knox, C. M. (1981). Performance of dyslexic children on speech perception lests. Journal of Experimental Child Psychology, 32, 401-424.

Hurford, D. P. (1990). Training phonemic segmentation ability with a phonemic discrimination intervention in second- and third-grade children with reading disabilities. Journal of Learning Disabilities, 23, 564-569.

Hurford, D. P., & Sanders, R. E. (1990). Assessment and remediation of a phonemic discrimination deficit in reading disabled second and fourth graders. Journal of Experimental Child Psychology, 50,396-415.

Kamhi, A., Calls, 11., Mauer, D., Apel, K., & Gentry, B. (1988). Phonological and spatial processing abilities in language- and reading-impaired children. Journal of Speech and Hearing Disorders, 53,316-327.

Kilz, W. R„ & Tarver, S. G. (1989). Comparison of dyslexic and nondyslexic adulls on decoding and phonemic awareness lasks. Annals of dyslexia, 39,196-205.

Levitt, A. G„ Utman, J. (ï. (1992). From babbling towards the sound systems of English and French: A longitudinal iwo-casc study. Journal of Child Language, 19, 19-49.

Lieberman, P. (1986). On the genclic basis of linguislic variation. In J. S. Pcrkcll & I>. II. Klall (Eds.). Invariance and v-inability in speech processes. Hillsdale, NJ.: Lawrence Erlbaum Associates.

Lindamood, C. H„ & Lindamood, P. C. (1967). 77ie A.D.D. program: Auditory discrimination in depth. Boston: Teaching Resources Corp.

Luce, J., & Luce, P. A. (199()). Similarity neighbourhood of words in young children's lexicons. Journal of Child Language, 17,205-215.

McGivcrn, R. F., Berka, C , Languis, M. L., & Chapman, S. (1991). Detection of deficits in temporal pattern discrimination using the Seashore Rhylhm Tesl in young children wilh reading impairments. Journal of Learning Disabilities, 24, 58-62.

McGowan, R. S., & Nittrouer, S. (1988). Differences in fricative production between children and adulls: Evidence from an acoustic analysis of /J/ and / s / . Journal of the Acoustical Society of America, 83,229-136.

Montgomery, D. (1981). Do dyslexies have difficulty accessing articulatory informal ion? Psychological Research, 33, 235-243.

Morais, J. (1982), The two sides of cognition. In J. Mehler, M. Garrett, & E. Walker (Eds.), Perspective on mental representation (277-309). Hillsdale, N.J.: Lawrence Erlbaum Ass.

Morais. .)., Alegria. J., & Contciil. A. (1987). The relationships between segmental analysis and alphabetic literacy: An interactive view. Cahiers de Psychologie Cognitive, 7, 415-438.

Morais, J., Bertelson, P., Cary, L., & Alegria, J. (1986). Literacy training and speech segmental ion. Cognition, 24, 45-64.

Morais, J., Cary, L., Alegria, J,, & Bertelson, P. (1979). Does awareness of speech as a sequence of phones arise simultaneously? Cognition, 7,323-331.

Vie Causes of Phonemic A wareness 211

Morais, J., Castro, S. L., & Kolinsky, R. (1991). La reconnaissance des mots chez les illettrés. In R. Kolinsky, J. Morais & J. Segui (Eds.), La reconnaissance des mots dans les différentes modalités sensorielles: Etudes dc psycholinguistique cognitive. Paris: Presses Universitaires dc France.

Morais, J., Castro, S. L., Scliar-Cabral, L., Kolinsky, R., & Content, A. (1987). The effects of literacy on the recognition of dicholic words. Quarterly Journal of Experimental Psychology, 39A, 451-465.

Morais, J., Cluytens, M., & Alegria, J. (1984). Segmentation abilities of dyslexies and normal readers. Perceptual and Motor Skills, 58, 221-222.

Morais, J., de Gelder, B., & Vcrhacghc, A. (in preparation). Visual effects in speech perception and literacy.

Nittrouer, S., & Sluddcrt-Kcnncdy, M. (1986). The role of coarticulatory effects in ihe perception of fricatives by children and adulls. Haskins Laboratories: Status Report on Speech Research, SR-88,73-93.

Niltroucr, S., Studdert-Kennedy, M., & McGowan, R. S. (1988). The emergence of the phonetic segments: Evidence from the spectral structure of fricative-vowel syllables spoken by children and adults. Haskins Laboratories: Status Report on Speech Research, SR-93/94,1-21.

Peretz, I., Morais, J., & Bertelson, P. (1987). Shifting car differences in melody recognition through strategy inducement. Brain and Cognition, 6, 202-215.

Reed, M. A. (1989). Speech perception and the discrimination of brief auditory cues in reading disabled children. Journal of Experimental Child Psychology, 48, 270-292.

Snowling, M. (1981). Phonemic deficils in developmental dyslexia. Psychological Research, 43, 219-234.

Snowling, M., Chial, S., & Hulme, C. (1991). Words, nonwords, and phonological processes: Some commcnls on Galhcrcolc, Willis, Emslic, and Baddeley. Applied Psycholinguistics, 12,369-373.

Snowling M., Goulandris, N., Bowlby, M., & Howell, P. (1986). Segmentation and speech perception in relation lo reading skill: A developmental analysis. Journal of Experimental Child Psycholog}', 41,489-507.

Snowling, M., & Hulme, C. (1989). A longitudinal case study of developmental phonological dyslexia. Cognitive Neuropsychology, 6, 379-401.

Stackhousc, J., & Wells, B. (1991). Dyslexia: The obvious and hidden speech and language disorder. In M. Snowling & M. Thomson (Eds.), Dyslexia: Integrating Tlteory and Practice (pp. 185-194). London: Whurr Publishers Ltd.

Tallal, P. (1980). Auditory temporal perception, phonics and reading disabilities in children. Brain and Language, 9, 182-198.

Tallal, P., & Piercy, M. (1974), Developmcnlal aphasia: Rate of auditory processing and selective impairment of consonant perception. Neuropsychologia, 12,83-93.

Taylor, H. («., Lean, D., & Schwartz, S. (1989). Pseudoword repetition ability in learning-disabled children. Applied Psycholinguistics, 10,203-219.

Watson, M., Stewart, M., Krause, K., & Rastaltcr, M. (1990). Identification of time-compresscd sentential stimuli by good vs poor readers. Perceptual and Motor Skills, 71, 107-114.

Werker, J. F., & Tees, R. C. (1987). Speech perception in severely disabled and average reading children. Canadian Journal of Psychology, 41(1), 48-61.

212 J. Morais and P. Mousty

Zurif, E. B., & Carson, G. (1970). Dyslexia in relation to cerebral dominance and temporal analysis. Nettropsychologia, 8, 351-361.

Anahtii Appimtlirs tt> llinnaii *\'j*,rniiini J. Ali-j:ria, I). IIHICIKJVI. J. .1IIII\:I >Ic Minais ami M Kaikaii («Is.) (?) \')')2 l-kcvicr Scirncc Puhlislirrs U.V. Ml vKh\* resaveii 213

CHAINER 13

Single and Multiple Process Models of Print to Speech Conversion

Alain Content and Ronald Peereman laboratoire dc Psychologic Expérimentale Université Libre de Bruxelles Bruxelles - Belgium

ABSTRACT

Theories of prinl-lo-specch transcoding diverge on three main issues: the nature of knowledge sources, the need to posit several distinct processes, and, within multiple-process theories, the interactions between the routines. We first present a study showing thai distinct sources of information and separable processes are available to skilled readers. Then, we describe data from a second experiment, which demonstrate the existence of interactions between lexical and sublcxical processing. In the same experiment, we observed performance patterns that cannot bc explained by statistical regularities in the mapping from print to speech. Finally, we discuss the findings in relation lo current models.

The description of the mechanisms by which readers transcode strings of letters into speech has been the object of intensive research for about 20 years. These efforts have progressively led to more sophisticated experimental techniques, to more detailed investigations and to refined theoretical and computational models. Yet, the largest part of the endeavour has been conducted in English, leading to concerns about the universal validity of the findings and conclusions. Under the impetus of Paul Bertelson, we have been pursuing experimental studies of French-speaking adult performance. Our primary aim has been to examine to what extent current models could be appropriately applied to French. We also tried to take advantage of specific characteristics of the French orthography to explore unanswered or neglected issues. In this chapter, we report recent investigations with skilled readers. We also explore how the data can help constrain the choice among current models, and possibly extend their scope.

French orthography, like English, is characterised by inconsistencies in the mapping of spelling to sound, although it differs from English in various ways. The French spelling system is an ancient, conservative orthography, marked by its history. French is often considered unpredictable, inconsistent and ambiguous, but this reputation is mostly due to characteristics of the