LITERACY ACQUISITION AND LINGUISTIC SKILLS
2.1. PHONOLOGICAL PROCESSING
2.1.1. Phonological awareness
2.1.1.1. The concept of linguistic awareness
Before proceeding to review the empirical data, the concept of linguistic awareness and its relationship to other aspects of language processing needs explanation. Linguistic awareness (also referred to as: metalinguistic awareness or metalinguistic skills) implies explicit access to linguistic structures, and the possibility of conscious (voluntary) control over them. Different authors (Tunmer, Pratt & Herriman, 1984; Gombert, 1992; Krasowicz-Kupis, 1999) emphasise different attributes of being “linguistically aware”; some common threads, however, can be identified:
• Objectification of language: it ceases to be merely a (transparent) medium or tool of communication, but also becomes an object of reflection and play.
• Ability to analyse and manipulate linguistic structures.
• Control over language that is conscious and deliberate, as opposed to the tacit knowledge necessary to speak and understand speech.
• Insight, control, analysis and manipulation that are instance-free, i.e. presuppose the grasp of underlying rules.
All authors also agree in stressing the functional role of linguistic awareness. Children acquire explicit control over linguistic structures as it helps them to handle growing demands of communication - especially using written language.
Given the above characteristics, an analogy may be drawn between the child gaining linguistic insights, and the work of an adult linguist (Krasowicz-Kupis, 1999). Both treat language as the subject of inquiry, and each of them build, in their own way, a theory of language. This analogy is implicit in endowing children with meM-linguistic abilities, since the term ‘metalinguistic’ was originally used only in the context of formal scientific investigations. The analogy has its limitations, too. Whereas a linguist has to use some meta-language (a set of formal notions used to describe language), a linguistically aware child may have insights that enable her to control and manipulate language at will, without necessarily being able to utilise such formal descriptions (she might be able to verbalise her insights but informally, e.g. through prosodic manipulation: “Not WAbbitt but RAbbitt”).
The concept of linguistic awareness may be construed as a developmental continuum that is orthogonal to other aspects of linguistic processing. Thus, processes within different language sub-modules (phonology, morphology, syntax, etc.) and having different locations along the processing path (reception, production), may be subjected to explicit control. Some disagreement exists as to what tasks (and what ordinary manifestations of language use) may be interpreted as true manifestations of linguistic awareness (Roth et. al., 1996; Krasowicz-Kupis, 1999). Those disagreements are not surprising if we assume that linguistic awareness is indeed a continuum or it develops gradually across a number of levels or stages (e.g. Gombert, 1992). Table 2-1 lists some aspects of linguistic functioning that are commonly seen as the manifestations of linguistic awareness (cf. Gombert, 1992; Krasowicz-Kupis, 1999).
language domain
language units
manifestations of metalinguistic skills
phonology syllables,
intrasyllabic units, phonemes
• making up poems
• sound-based games (e.g. secret languages) • identification of target phonological units • segmentation and blending
• manipulation: deletion, substitution and replacement of target phonological units m orphology & syntax words phrases sentences
• correction and self-correction of grammatical errors • judging the grammatical acceptability of sentences semantics &
pragmatics
words sentences texts
• understanding and appreciation of figurative language and linguistic humour (e.g. puns)
• detecting contradictions and inconsistencies • adjusting the style to the listener/reader • structuring the written/spoken stories • using appropriate stylistic conventions
Table 2-1. The scope of metalinguistic skills. Adapted from Krasowicz-Kupis, 1999, p.56.
2.1.1.2. Development of phonological awareness
Phonological awareness is the subset of linguistic awareness skills that enables one to access and manipulate the sound structure of language.
The cumulative body of evidence shows clearly that the developmental course of phonological awareness varies with the size of phonological elements involved (syllables, intrasyllabic units, phonemes).
The awareness of syllables seems to occur first. The classic studies of Liberman and her colleagues (e.g. Liberman, Shankweiler, Fisher & Carter, 1974) showed that some 4-year olds and many 5-year olds were able to tap the number of syllables in a word (or lay out tokens corresponding to syllables: Treiman & Baron, 1981). This achievement occurred spontaneously, without any prior instruction and in children who were pre-literate. The availability of syllable level awareness to many children before the onset of formal literacy instruction has since been confirmed in many languages (Italian: Cossu, Shankweiler, Liberman, Katz & Tola, 1988; Japanese: Mann, 1986; Chinese: Ho & Bryant, 1997a, b; Polish: Krasowicz & Bogdanowicz, 1997; Krasowicz- Kupis, 1999). The ‘natural’ character of this ability is confirmed by the finding that many illiterates and near-illiterates can also segment words on the level of syllables (Morals, Bertelson, Cary & Alegria, 1986; Morals, Content, Bertelson, Cary & Colinsky, 1988; Lukatela, Carello, Shankweiler & Liberman, 1995).
The awareness of intrasyllabic units also emerges early. It is widely agreed that a syllable is not merely a string of phonemes, but a hierarchical structure that clusters phonemes into two intra-syllabic units. A syllable divides either between the initial consonant or consonant cluster {onset) and the vowel with following consonants {rime),
e.g. TW - 1ST; or between initial consonants with the vowel {body) and the remaining consonants {coda), e.g. TWI - ST. There are arguments from linguistic theory, experimental data and the analysis of spontaneous language use (speech errors, word games) for either onset-rime and body-coda models of syllables (Duncan, Seymour & Hill, 1997). However, most experimental studies with pre-school and early school children found onset-rime division to be most naturally available to a child (Kirtley, Bryant, MacLean & Bradley, 1989; Treiman, 1992). This division corresponds to poetic devices of alliteration and rhyme' (Goswami & Bryant, 1990). Classic studies on that topic, carried out by Bryant and colleagues (Bradley & Bryant, 1978, 1983) employed sound categorisation tasks (also known as oddity tasks) in which participants had to say which one of three or four monosyllables was the “odd one” because it did not share a complete rime or onset with the others. Many 4- and 5-year olds were found to be sensitive to onset-rime division; a finding corroborated by a number of subsequent studies (for review and discussion see Goswami & Bryant, 1990; Goswami, 1999). Even some 3-year olds could perform this task when pictures were used in order to remove memory load (MacLean, Bryant & Bradley, 1987). The validity of the sound categorisation paradigm for the assessment of phonological awareness is disputable, however. It seems that the detection of an odd-sounding item in the stream of monosyllables may be accomplished with a low-level phonemic discrimination that does not necessarily require explicit access to sound structure of the stimuli. Thus, oddity tasks, although undoubtedly measuring phonological processing, may not measure awareness. A recent validity study (Schatschneider, Francis, Foorman, Fletcher & Mehta, 1999) seemed to confirm these reservations, as it found the oddity task to be a poor indicator of a general latent factor of phonological awareness. However, early developmental emergence of rime-onset awareness has also been confirmed with other methods. Some 4-year old children are capable of forced-choice rime detection (they can say which two of three or four words rhyme: Muter, Hulme & Snowling, 1998) as well as onset and rime production (generating exemplars that rhyme or share alliteration
with a target word: MacLean et al., 1987; Stuart, 1986, reported in Bunn, 1995). Most importantly, studies of pre-school language play demonstrated the children enjoy, remember and make poems, both spontaneously and in experimental settings (Dowker, 1989). These are often based on phonological devices of rhyme and alliteration, although marked cross-linguistic differences are observed in terms of frequency of their use (Dowker, Pinto, 1993; Dowker, Krasowicz, Pinto, Roazzi & Smith, 1998; see also chapter 3). The ability to rhyme seems, again, at least partially independent of literacy. This is demonstrated by the superior rhyming abilities of illiterate or semi-literate Portugese and Brasilian oral poets (Morals, 1991; Roazzi, Dowker & Bryant, 1993).
In contrast, the ability to perceive words in terms of their phonemic structure, and the ability to manipulate those phonemic strings - that is, phonemic awareness - is linked to instruction in alphabetic reading and spelling. Numerous studies have confirmed the seminal findings of Bruce (1964) and Liberman et. al. (1974) that pre literate children find it very difficult to delete a specific phoneme from a word, or tap a number of phonemes in a word. Such lack of explicit control over phonemic structure is particularly striking in children who are fully pre-literate (i.e. lack not only reading ability but also knowledge of letter names or sounds), although such children may perform with some degree of success on the tasks that involve phoneme-level representations, yet do not require explicit segmental manipulation (phoneme oddity, detection and blending: Burgess & Lonigan, 1998; Naslund & Schneider, 1996; Wimmer, Mann & Signson, 1999). However, even short acquaintance with alphabetic writing, (evidenced by partial letter knowledge and minimal decoding skills) is associated with significant growth of phonemic awareness. This was demonstrated in developmental studies (e.g. Bowey, 1994; Wimmer, Landerl, Linortner & Hummer, 1991) as well as those involving semi-literate adult participants (Lukatela et. al., 1995). The specific connection between alphabetic literacy and phonemic awareness is also evidenced by studies of people who read non-alphabetic scripts. The comparison of native speakers of Chinese using only the traditional, morpho-syllabic script, with those taught to use the alphabetic orthography (Pinyin) found the former very poor on phoneme deletion and addition (Read, Zhang, Nie & Ding, 1986). Likewise, Japanese six-year old children were much poorer than their American counterparts on phoneme tapping, whereas no difference was observed for tapping or deleting syllables. Although the users of syllable-based Japanese orthography do eventually become aware of
phonemes, the development of these skills is very protracted in the absence of alphabetic facilitation (Mann, 1986).
The picture of phonological awareness development, which I sketched separately for different size-levels of phonological units, becomes more complicated when task- specific factors are taken into consideration. Performance on all levels is influenced by the nature of the required operation, and the characteristics of the phonological material. The detection of difference or similarity is generally easier than operations requiring explicit segmentation, deletion, addition or transposition; words are easier than nonwords; performance is also affected by the phonological properties of the target unit, such as its position within a word, phonemic category (vowel, consonant, consonant type), the presence of consonant clusters or sonority hierarchy (Goswami, Bryant, 1990; Krasowicz-Kupis, 1999; Schatschneider et. al., 1999, Stahl & Murray, 1994; Yopp, 1988; Goswami, 2000). It must also be stressed that the performance of pre-literate children is far from perfect on all levels of phonological awareness. Induction into alphabetic writing consolidates and enhances performance across the board - although it has the most direct and dramatic impact on the awareness of phonemes (e.g. Cossu et. al., 1988).
The internal structure of phonological awareness is a matter of some debate. Factorial analyses showed syllable awareness and phonemic awareness to emerge as two separate, yet correlated, latent variables (Wagner & Torgesen, 1987). Rime-onset level skills and phonemic level skills were shown, in some studies, to load on the same factor (Wagner et. al, 1993, Stahl & Murray, 1994, Schatschneider et. al, 1999) whereas others found them to emerge separately (Muter et. al., 1998). Other factorial analyses suggested that the division should be drawn not between the size-levels of phonological units, but between the type of operations involved. Thus, two-factor solutions were found which correspond to phonological analysis and synthesis (Wagner et. al, 1993, Schatschneider et. al, 1999) or ‘simple’ and ‘compound’ operations (Yopp, 1988). Most studies agree, however, that the emerging factors are highly intercorrelated (especially in pre-school children: Wagner et. al., 1993) and that phonological awareness can be described as a single broad domain. Individual differences in phonological awareness were observed to be relatively stable, at least over a medium-time span (kindergarten to grade four: Wagner, Torgesen & Rashotte, 1994; Wagner et. al., 1997).
2.1.1.3. The role of phonological awareness in the acquisition of literacy.
It is widely agreed that phonological awareness plays a central role in the acquisition of word decoding, recognition and spelling. The specific mechanisms linking phonological awareness and literacy are, however, a matter of some controversy. Any successful theory has to accommodate a paradoxical set of findings: phonological awareness is necessary for reading and spelling, yet the ability to read and spell is also necessary for phonological awareness.
The main evidence for the latter claim (becoming literate leads to phonological awareness) was discussed in the previous section. Phonemic awareness, in particular, seems to occur only after the onset of alphabetic literacy, though even limited exposure (such as the knowledge of the alphabet: Burgess & Lonigan, 1998) is usually sufficient to elicit it. Syllabic and intrasyllabic level awareness, although they occur earlier, also improve greatly with the acquisition of reading.
The reverse causal connection (phonological awareness is necessary for the acquisition of literacy) is evidenced, first of all, by a number of longitudinal studies which showed that the pre-school awareness of syllables, rimes and phonemes is related to subsequent reading and spelling performance, even when likely general confounds, such as IQ, chronological age or mother’s educational level are controlled (see Wagner & Torgesen, 1987; Goswami, Bryant, 1990; Share, 1995 for a review of the literature; for recent replication see Wagner et. al. 1994, 1997; Torgesen, Wagner, Rashotte, Burgess & Hecht, 1997; de Jong & van der Leil, 1999). However, a correlation between two variables (even the longitudinal one) is never an unequivocal proof of causation since it can always be a product of a third factor (Dancey & Reidy, 1999; Goswami & Bryant, 1990). In this case, letter knowledge is likely to be such a third factor, causal to both pre-school phonological awareness and subsequent reading progress. Knowledge of letter names and sounds is a very strong predictor of later reading (Denckla & Rudel, 1967/1997, Adams, 1990) and also correlates with pre-school phonological awareness (Burgess & Lonigan, 1998). Unfortunately, very few studies have controlled for this possibility. Those which did take letter knowledge into account gave somewhat ambiguous results: in some the relationship between pre-school phonological awareness and first-grade reading achievement disappeared (Lundberg, Olofsson & Wall, 1980, reanalysed in Wagner & Torgesen, 1987; de Jong & van der Leil, 1999) whereas others (Wagner, et. al., 1994; Naslund & Schneider, 1996) still found a significant longitudinal
link. The latter outcome is consistent with Maclean’s et.al. (1987) data showing that rime and alliteration sensitivity at age 3 could predict children’s ability to recognise words a year later, but was not related to the knowledge of the alphabet or to early arithmetic skills. Cross-linguistic differences may be important here, as both studies that failed to find the predictive role of pre-school phonological awareness involved non- English participants (see chapter 3).
Longitudinal studies provide more consistent support for the idea that phonological awareness measured during early school years exerts a causal influence on reading and spelling progress. Individual differences in phonological awareness measured at the beginning of first grade predict later reading achievement even when differences in IQ, phonological working memory and - most importantly - concurrent reading skills are partialled out (de Jong & van der Leil, 1999; Wagner et. al, 1994; Torgesen et. al., 1997). A unique contribution of phonological awareness to reading during the first year of school was observed even when pre-school differences in phonological awareness were not predictive (de Jong & van der Leil, 1999). In English- speaking children phonological awareness exerts such unique influence on subsequent reading achievement until the 4'^ grade, and possibly longer, although the effects tend to diminish with age (Wagner et. al., 1994, 1997; Torgesen et. al., 1997). This suggests that the role of phonological awareness may be most crucial in the early phase of reading acquisition (in the context of learning to decode) and may decrease as children become more skilled and make greater use of lexical-based reading strategies (Bus & Ijzendoorn, 1999). However, this outcome may also be an artifact of a ceiling effect on the experimental measures of phonological awareness, which are typically tailored to differentiate the lower range of performance.
The second, most direct type of evidence for a causal role of phonological awareness in the acquisition of literacy comes from training and intervention studies. Explicit instruction in phonological awareness is generally found to benefit subsequent word reading, spelling, and reading comprehension attainment. The effect is specific to literacy and does not generalise to mathematical achievement (National Reading Panel, 2000). A meta-analysis of existing studies (Bus & van Ijzendoorn, 1999) showed that experimental intervention on phonological awareness explains about 12% of variance in later reading skills. However, the benefits were observed mainly in the studies that combined phonological and letter training, while ‘pure’ phonological awareness training brought smaller and often non-significant results (e.g. Bradley & Bryant, 1983). It may
be argued, therefore, that it is an early phonics reading instruction, rather than phonological awareness training per se, that is beneficial. The benefits of training are usually transient (below 6 months) for word identification. However, longer-lasting (if small) effects occur for spelling and reading comprehension. This (together with the results of longitudinal studies discussed previously) suggests that phonological awareness may directly facilitate the process of learning to read (especially the self teaching mechanism based on decoding) but only indirectly the learning outcomes
(Share, 1995; Bus & van Ijzendoorn, 1999).
The third piece of evidence for the causal role of phonological awareness in reading acquisition comes from studies of developmental dyslexia employing reading level-match design (i.e. comparing poor readers with younger normal children of the same reading age). A number of studies have found that dyslexies perform worse than reading-age controls on phonological awareness tasks. These include: rhyme and alliteration oddity (Bradley & Bryant, 1978), the T ig Latin’ game (in which the initial phoneme of a word has to be moved to the end and the sound ‘ay’ added, eg. pig - igpay: Olson, Wise, Conners, Rack & Fulker, 1989), or phoneme deletion (Olson, Rack & Forsberg, 1990, reported by Rack, 1994). The reading-level-match design rules out the possibility that such differences are simply the result of the limited reading skill and experience of dyslexic participants, and suggests that the phonological deficit is a primary cause of dyslexic difficulties. The results are mixed, however, and several studies (e.g. Beech & Harding, 1984) failed to find a phonological awareness deficit in dyslexia - instead, they observed a developmental delay (performance commensurate with reading ability: worse than in chronological-age matched, but the same as in reading-age matched controls). This inconsistency may be resolved if we take the size of phonological unit into consideration. At least in older dyslexies, the awareness deficit seems to be limited to the phonemic level, whereas the syllabic and onset-rime deficit may be accounted for by the word-finding difficulties also observed in dyslexies (Goswami, 1997; Swan & Goswami, 1996). Differences may also be observed regarding the type of dyslexic difficulties. A deficit in phonological awareness can characterise poor readers who have specific problems with alphabetic skills (phonological dyslexies), but may not occur in surface dyslexies, whose difficulties are predominantly orthographic (Stanovich, Siegel & Gottardo, 1997).
Overall, the pattern of findings presented above provides a strong case for reciprocal causation: phonological awareness is a condition for reading, but reading also
engenders phonological awareness. A number of theories have been put forward to explain this reciprocality, most of which share the same general structure. They assume that the development of phonological awareness can be divided into at least two main stages, phases or levels. The earlier one(s) occur as a natural part of language development, and constitute the prerequisite of successful reading and spelling acquisition. The later one(s) are a direct product of becoming literate. Diverse