DISCUSSION

The first research question addressed the effect of the lexical status of the word on SAV phoneme isolation. The results of the study showed that when SAV phonemes were targeted, the lexical status of the stimulus word (SAV, MSA, or pseudoword) did not affect phoneme isolation performance. These findings provide apparent counter evidence to the predictions of the lexical restructuring model (Metsala, 1999; Metsala & Walley, 1998; Walley, 1993; Walley et al., 2003), according to which the phonological analysis of words should be better than that of pseudowords and the phonological analysis of familiar words should be better than that of unfamiliar words. Two unique features of Arabic are proposed to explain this unexpected finding, the phonological/prosodic structure of Arabic and Arabic orthography. As to the phonological/prosodic structure, it is suggested that the rather simple syllabic structure of Arabic may account for the absence of a word familiarity effect on phoneme segmentation. As such, the basic and most frequent syllable structure in Arabic is the CV, followed by the CVC structure. Consonantal cluster rime coda structures (CVCC) are very rare in MSA (Ababneh, 2000), as they are restricted to monosyllabic words and to short vowels. Except for the CVCC structure, no complex onset or coda subsyllabic units exist in MSA. Contrary to Arabic, the prosodic structure of English features a variety of complex onset and coda subsyllabic units that may embody as many as three consonants, as in the word “strands” /strandz/, and even a marginal length of four consonantal phonemes, including the appendix, as in “sixths” /siksθs/. This is believed to make the segmental analysis of English words harder, more variable, and perhaps, as a consequence, more strongly dependent on lexical aspects of the word than is the case in Arabic. Thus, it is possible that the more complex the prosodic structure of the language is, the more difficult the phoneme segmentation task will be and the more dependent it will be on the lexical status of the stimulus word. In order to probe the validity of this argument, research into the role of lexical involvement in segmental analysis should take a linguistic perspective on the prosodic structure of the stimuli used. It is possible that lexical involvement would figure more prominently in the phonemic analysis of complex prosodic structures. As such structures are rare in Arabic, the present study tested the segmental analysis of CVCVC words only. It was found that the lexical status of the word did not affect phoneme isolation. This finding, however, does not automatically invalidate the relevance of the lexical status of the word to phonological analysis in other languages, or even in Arabic but among younger children. The role of the lexical status of the word may interact with its prosodic structure in affecting phoneme segmentation. This issue remains for future research to pursue, and it would be best tested in a language that features a variety of prosodic structures and with varying degrees of complexity.

Work in English has shown that even first graders' phoneme segmentation benefited from familiarity with the stimulus word (Metsala, 1999). The present study demonstrated the absence of a word familiarity effect on the segmental isolation of SAV phonemes even among kindergarteners. A second and tentative interpretation for this finding in Arabic relates to the nature of vowelized Arabic orthography. Not only is the prosodic structure of MSA simple, but the orthography that beginning readers are exposed to is highly transparent: shallow vowelized Arabic. Upon the inception of reading, children receive positive and reliable feedback from the orthography as to word pronunciation. This results in greater confidence in the orthography and a sharper awareness of the phoneme, as decoding is predicated primarily on a grapheme–phoneme conversion mechanism. In line with this argument, crosslinguistic research that compared the development of phonemic awareness in beginning readers of a shallow orthography, like Turkish and German, with that of beginning readers of English showed that exposure to a shallow orthography accelerates the acquisition of phonemic awareness (Caravalos & Bruck, 1993; Cossu et al., 1988; Durgunoglu & Oney, 1999; Mann & Wimmer, 2002; Oney & Goldman, 1984). Such enhanced phoneme awareness may explain the absence of lexical involvement in phoneme isolation in Arabic. The extent to which children from different oral language background and different orthographies might make use of lexical processes in phonological analysis, and the point in the acquisition process when this ceases to constitute a factor is an issue for future crosslinguistic research to explore.

The second research question addressed the effect of a word's lexical status on the isolation of phonemes with different linguistic affiliation: SAV versus MSA phonemes. The results of the study showed that the linguistic affiliation of the target phoneme, SAV versus MSA, had a main effect on phoneme isolation; MSA phonemes, that were absent in the oral language experience of children, were more difficult for both kindergarten and first grade children to isolate than SAV phonemes. At the same time, we noted a developmental effect, as MSA phonemes proved significantly more challenging for kindergarteners than for first grade children. As to the effect of a word lexical status on phoneme isolation, the results again failed to support a main effect of word familiarity on phoneme isolation: children showed comparable levels of phoneme isolation performance for MSA and pseudowords. The results also showed that, contrary to expectations, kindergarten children had more difficulty with the analysis of MSA words than pseudowords and with the isolation of MSA phonemes embedded within MSA than within pseudowords. Such a reversed effect of word familiarity than would be predicted from theories of lexical involvement in phonological analysis is argued to result from another unique linguistic feature of a diglossic context, and it sheds light on the linguistic components that go into word familiarity in Arabic. This feature is the phonological distance, yet relatedness between the phonological representations of the same word in SAV and MSA. As such, first graders, as they have had a whole year of consistent daily exposure to MSA, have internalized a rather robust phonological representation for MSA phonemes and MSA words, especially as the latter had been drawn from the first grade primer. Such high quality phonological representation (Elbro, 1996, 1998) of MSA phonemes explains the developmental effect reflected in the significantly higher MSA phoneme isolation scores among first graders. Such robust phonological representation is probably also responsible for the absence of a word familiarity effect on phoneme isolation among first graders. Contrary to first graders, the phoneme isolation performance of kindergarteners was found sensitive to both the linguistic affiliation of the target phoneme and the lexical status of the stimulus word. The results showed that kindergarteners had particular difficulty isolating MSA phonemes. Further, while they were shown to have particular difficulty with the analysis of MSA words in general, they found MSA phonemes that were embedded within MSA words more difficult to isolate than those embedded within pseudowords. Children at the kindergarten level, and in the absence of consistent formal exposure to MSA, have probably acquired a rather vague, inaccurate, and unstable phonological representation of MSA phonemes and MSA words, the kind that is implicitly gained from occasional exposure to TV programs that use MSA, or from story reading either at school or at home. As a result, when they were required to analyze an MSA word, they had automatically activated the word's SAV phonological representation. As the two forms of the word are phonologically related, such phonological relatedness interfered with phonological analysis and resulted in lower phoneme isolation scores. For instance, when children were required to isolate the initial MSA phoneme /θ/ from the MSA word /θa

/ab/ (“a fox”), they automatically activated the SAV representation of this word /

a

/ab/. This resulted in a phoneme isolation error. In the absence of a related competing phonological representation in the mental lexicon of children, as in the case of pseudowords, kindergarten children achieved higher phoneme isolation scores.

Thus, while first graders may have used their rather robust knowledge of the phonological representation of MSA phonemes and MSA words to facilitate MSA phoneme isolation, kindergarteners showed a reversed pattern, with MSA words posing a particular stumbling block. The absence of a robust phonological representation of MSA words, especially given a competing overlapping phonological representation for some MSA words, disrupted the phonetic recoding of the word in working memory (Wagner & Torgesen, 1987) and led to a phoneme isolation error. In the analysis of pseudowords, and in the absence of a competing phonological representation, children achieved higher phoneme isolation scores. This finding has important implications for phonological analysis in bilingual contexts as well, especially when the two languages of bilingual children are typologically related and, thus, share a large set of phonologically overlapping cognates. The results of the study suggest that in the phonological analysis of cognates both forms of the cognate may be activated. This may interfere with the outcome of the phonological analysis process. In order to tap into the processes that underlie the outcome of phonological analysis, future research should employ introspective methods of data collection. Only such an approach would allow an examination of the factors that underlie phonological analysis and outcome. Such an approach would also inform about the mechanisms that underlie the analysis of phonologically related linguistic forms, especially as transfer may surface more prominently when the reader can make “equivalence classifications” (Fledge, 1987) on the two phonological forms available to him.

The present findings demonstrate the centrality of two manifestations of the linguistic distance between MSA and SAV in understanding the phonological sensitivity of children in a diglossic context toward the language they learn to decode. The first is the phonemic distance between MSA and SAV. The second is the lexical distance between the two forms. The results of the present study replicate previously reported findings of the role of phonemic distance in the acquisition of phonemic awareness in a diglossic context (Saiegh–Haddad, 2003). However, rather than testing the phonological analysis of pseudowords only, the present study tested the effect of phonemic distance on the phonological analysis of different types of words: SAV, MSA, and pseudowords. Here too, the linguistic affiliation of the phoneme had a main effect on phoneme isolation, hence strengthening previously reported evidence and consolidating the contention that the availability of the phoneme in the oral language of children facilitates phoneme isolation, irrespective of the lexical status of the stimulus word within which the phoneme is embedded. These findings demonstrate the unique complexities that inhere in the acquisition of reading in a phonologically distant language (Charity et al. 2002; Saiegh–Haddad, 2003, 2004; Wade–Woolley & Geva, 2000; Wang & Geva, 2003). Children acquiring reading in a language that is phonologically distant from their oral language, as in reading in a bilingual context, in a standard dialect, and in a diglossic context, are required to construct mental representations for a set of novel phonological structures. Until such novel structures are acquired, and until their phonological representation becomes robust, and their activation automatic (LaBerge & Samuels, 1974; Perfetti, 1985; Wolf & Katzir–Cohen, 2001), novel phonological structures may well constitute an additional challenge for beginning readers.

It is noteworthy that the study tested the isolation of phonemes from both an initial and a final word position. The results showed that the isolation of initial onset phonemes was consistently more difficult than the isolation of final phonemes. This pattern was evident among both kindergarten and first-grade children, across both SAV and MSA phonemes, and across all types of words: SAV, MSA, and pseudowords. Further, initial phoneme isolation was found more sensitive than final phoneme isolation to variations in the linguistic affiliation of the target phoneme (SAV vs. MSA) and in the lexical status of the word (SAV, MSA, and pseudo). These findings replicate and extend previously reported findings from the phonemic awareness of Arabic-speaking children (Saiegh–Haddad, 2003) and show that Arab children find it more difficult to isolate the prevocalic consonantal phoneme than a postvocalic final consonant. Saiegh–Haddad (2003) found that final phonemes were easier to isolate than initial phonemes even when they were embedded in a complex consonantal cluster rime coda. The particular difficulty that Arab children revealed in isolating initial-onset phonemes than final rime coda is at odds with the patterns thus far revealed among English speaking children (Treiman, 1985, 1988), and, thus, casts doubt on the validity of the onset–rime dichotomy in explaining phoneme isolation performance in Arabic. Thus, it appears that, in Arabic, initial phoneme isolation is significantly more challenging for children than final phoneme isolation. This task is also more sensitive to variations in the linguistic affiliation of the target phoneme and of the stimulus word. Future research should address this issue directly using various types of phonemic awareness tasks that tap into the representation and processing of intrasyllabic units. It is possible that language-specific prosodic constraints may affect the ease with which various intrasyllabic units are orally produced as the outcome of phonological analysis. This may explain crosslinguistic variations in the ease with which initial versus final phonemes are isolated.

In summary, the results show that phoneme isolation is not an all-or-none phenomenon. Various aspects of the phonological and lexical representation of the word may affect the ease with which children can operate on phonemes. Previous work has shown that at least two major factors affect children's ability to perform mental operations on the phonological structure of words: the size of the phonological unit and the linguistic context (Treiman & Weatherson, 1992; Treiman & Zukowski, 1991). The results of the study show that at least two more factors are pertinent to explaining individual variations in phonological analysis in a diglossic context. These are the linguistic affiliation of the target phoneme and the lexical status of the stimulus word. Thus, in order to explain crosslinguistic variations in the route and the rate with which phonological awareness skills are acquired, current conceptualizations of the construct of phonemic awareness must be extended to incorporate other factors that are particularly pertinent to linguistic contexts characterized by linguistic disparity between the oral language of children and the written language.