A Bilingual Perspective on the Possible Universality of Phonological Awareness Skills Across Two Languages

Reading comprehension relies on the integration of phonological, semantic, syntactic and pragmatic language abilities. The current study investigated phonological awareness in six-year-old children’s mastery of reading in Maltese and English. The researchers recruited eighty-two bilingual participants attending bilingual schools in Malta and administered two parallel batteries comprising parallel word reading tests and phonological tasks in the two languages. Principal components analysis identified clear componential structures in both of the phonological batteries (Maltese and English). A statistical regression analysis identified similar phonological underpinnings across the two single word reading measures. Specific measures of phonological awareness constituted common phonological underpinnings of reading performance in both Maltese and English, if to different degrees. The results support the notion of similarity in the patterns of association of skills sustaining reading across Maltese and English in bilingual children. The view that the phonological skills underpinning reading development across alphabetic languages may not differ substantially between different orthographies is supported.

view of the European Council which considers bilingualism to be continuous rather than dichotomous (Gazzola, 2016). Although all schools in Malta teach in both languages, literacy instruction varies across school types (Xuereb, Grech, & Dodd, 2011). Independent schools use English as their language of instruction (Bonnici, 2010), but State schools tend to use Maltese (Vallejo & Dooly, 2009). Church schools are generally more balanced using both languages according to need (Firman, 2007;Sammut, 2014).
Although bilingual learners may express the differences they encounter in the orthographic depth through their rate of reading development, the process remains similar across languages (Vaessen, Bertrand, Denes, & Blomert, 2010). Even if most predictors of reading performance are deemed to be universal across alphabetic languages, their precise weight varies systematically as a function of script transparency (Saiegh-Haddad & Geva, 2010;Ziegler et al., 2010). English and Maltese differ in their orthographies with the former having a deep, opaque orthography in which phonology does not always clearly guide word spelling or reading (Gorman & Gillam, 2003) and the latter having a shallow, transparent orthography through which word phonology maps consistently onto orthography (Xuereb, 2009).

Materials and Method
The current study investigated the underlying phonological skills in six-year-old children's mastery in reading Maltese and English. Grade 2 is considered to be critical to children's reading development since, at this time, word-attack ability develops rapidly (Logan et al., 2013). The church school sector in Malta with its balanced language approach and bilingual context was best suited for this study. The researchers explored if in the Maltese-English bilingual context, the phonological skills that children brought to bear on early reading development differed. This investigation was intended to identify those processing components that were common to both language systems and those that were language-specific. The second author collected the data during the first term of the 2016-2017 scholastic year. After due piloting, pruning and modification of the broader test battery, this author compiled shorter parallel batteries of phonological tests in each language. These tests validly investigated the specificity of the phonological awareness tasks underpinning reading and yielded maximum variance.
At this stage, one must address the issue of the comparability of phonological awareness tasks across languages. Eviatar, Taha and Shwartz (2018) used virtually identical tests of phonological awareness in Hebrew and Arabic because Hebrew and Arabic are both Semitic languages with similar semantic and phonological features. Maltese has an Arabic base (written in the Roman alphabet) with the grafting of lexical elements of Italian and English in particular, but it retains a Semitic morphological and syntactic structure (Fabri, 2010). While English constitutes many nouns with a CVC construction, Maltese has few such nouns; many monosyllabic nouns have CCVC and CVCC constructions. The various tasks used in the testing procedures reflected the phonological characteristics of both languages and were parallel but not fully phonologically comparable (Share, 2008).
The researchers adhered strictly to all the procedures laid down by the institutional ethics committee for recruiting participants. Eighty 6-year-old, bilingual, typically developing pupils of Maltese heritage were recruited from six church schools. Two reading tests were administered to each child individually to assess word decoding ability in Maltese and English, and to select average readers for this study. To guard against floor or ceiling effects, only participants with average word reading accuracy participated in the study. The participants completed the phonological test battery in two 20-minute sessions.
For reading, the authors adopted the Maltese Word Reading Test (Bartolo, 1988) normed on a locally representative population of 1,160 children aged 6:00-10:06 and the York Assessment of Reading for Comprehension Single Word Reading Test (Snowling et al., 2009) for English. Both these tests provided a measure of children's word reading skills. Identical administration and discontinuation procedures of eight consecutive mistakes were adopted. The parallel phonological tasks in Maltese and English constituted tests of syllable segmentation, rhyme awareness and generation, phoneme segmentation, elision, sound matching (initial and final) and phoneme substitution. All the tests were preceded by two or three practice trials to help instruct the participants in the task, followed by the administration of the test items. The second author provided corrective feedback during the practice trials but not during the test itself. The order of administration of the tests was randomised for each child to minimise practice and order effects. Some of the items used pictures to reduce memory load.

Assessing at the Syllabic Level -Syllable Counting
The two parallel tasks devised for the current study employed the use of Maltese and English real words. Children repeated up to four-syllable words and represented the syllables with counters. This procedure was based on the procedure used by Cossu et al. (1988).

Assessing at the Rhyme Level -Rhyme Oddity Task and Rhyme Generation
In the Rhyme Oddity task, the assessor uttered three words in succession, and the participants identified the two that shared the rhyming sound (e.g., fish, cap, tap). The second measure of rhyme awareness, the Rhyme Generation task required participants to produce words that rhymed with the stimulus word, thus demonstrating their ability to access, retrieve and provide words with similar rhymes. Participants completed ten items in both Maltese and English rhyme oddity and generation tasks with discontinuation set at four consecutive errors according to the established convention (Muter, 1994 The Phoneme Substitution task was adopted from the Phonological Assessment Battery 2 (Gibbs & Bodman, 2014). Participants were asked to replace the first sound of a word with a new sound (e.g., "cot" with a /g/ gives "got") in ten test items. A parallel task in Maltese accompanied the first. Five of the six single phonemes in the two tasks were phonetically identical but, it was not possible to replicate the four consonant blends in the English test if the words used were to possess the appropriate currency for the children's age. The authors adhered to the guidelines in the test manual advising discontinuation after three consecutive mistakes or after three minutes had elapsed since the presentation of the first item for both language versions.
In the Letter-Sound Knowledge task, each child was asked to identify the phonetic sound of ten lowercase letters in random order to assess letter-knowledge in Maltese and English. The individual letters were selected for their frequency in each language. The phonetic sounds of the letters "h", "b", and "e" were common to both tests.

Results
Data obtained from this study were processed using IBM SPSS Statistics Version 23. Normality of distribution of the dependent variables was assessed statistically and visually according to procedures suggested by Maxwell and Delaney (2004 Mean age in years 6 years 5 months 6 years 5 months Mean raw score in word reading (SD) 24 (9.8) 16 (6.9) Table 2 shows the scores for the whole group on the phonological measures in Maltese and English.  One notable exception is the issue of the Maltese and English Phoneme Counting tasks that featured prominently in the Maltese language pattern matrix but weakly in the English language pattern matrix.
In a bid to establish the contribution of each of the independent variables (tasks of phonological awareness) to the dependent variable (reading task), a series of stepwise regressions was used to determine how strongly each subtest was associated with reading achievement. Collinearity among the variables was weak enough to introduce only minimal imprecision to the regression model. The variance inflation factor (VIF) for each variable never exceeded the value of 3, which is acceptable (Hair, Black, Babin, & Anderson, 2010). Zero-order correlations were generally low among the variables as shown in Tables 5 and 6 below, only once reaching .7 with this being between the initial and final sound matching measures in both languages. The inter-item correlation of the Maltese and English phonological tasks was computed to be .240 and .204 respectively.

Discussion
The current study explores the specific relationship between children's phonological skills and reading.
In the first set of stepwise regression analyses, Maltese Elision (β =.420) emerged as the strongest  Mannai & Everatt, 2005, Caravolas et al., 2012 language assessed by this measure, support the case for bilingual children using underlying parallel skills to read both languages, these parallel skills being assessed for by manipulation skills (elision) and phoneme awareness (sound matching).
Here, one must refer to the similarity of the PCA of both batteries of phonological awareness measures.
The fact that the two analyses were almost identical with only minor differences between the two, attests to similar componential structures. This suggests that the phonological processing skills of the children participating in the study were similar for Maltese and English, as explained earlier.
The argumentation presented so far would have made a reasonable case for some degree of universality for specific phonological awareness skills underlying literacy development across the two languages (al Mannai & Everatt, 2005;Caravolas et al. 2012Caravolas et al. , 2013 had the participants been perfectly matched but growing up as monolinguals in their respective countries. However, increasingly, a bilingual continuum is becoming more of a reality (Gazzola, 2016) and such separate groups of monolinguals may not be easily found. In this present study, the fact that the participants were balanced users of both languages probably resulted in a cross-language effect. This could mean that whatever phonological strategies they developed to decipher words in one language could have been adopted to decipher words in the other language, resulting in a possible transfer of phonological skills from one language system to another. One may also argue that this result is due to the use of phonics instruction in English and in Maltese too, which led the children to develop explicit sensitivity to phonemes across languages in the first year of schooling (Duncan et al., 2013). The results of this study appear to support at least the first stage of Comeau et al.'s (1999) theory that phonological awareness of the first language transfers to the second language during literacy development and back to the first language, assisting the development of literacy across both languages.
This study has theoretical and practical implications. The general findings suggest that some predictors of reading performance may be universal, irrespective of the orthographic depth of the written language.
These are relevant to practical pedagogical issues relating to literacy and support the direct link between phonological awareness and reading in the two languages. In line with Gillon's (2018) recommendations, phonological awareness training should focus on the development of skills at the phoneme level (Brennan & Ireson, 1997;Cary & Verhaeghe, 1994;Lundberg, Frost, & Petersen, 1988;Yopp, 1988). Furthermore, phonemic awareness training should be integrated into children's learning routines, because phonological attack skills were deemed to be employable in reading (Al-Bataineh & Sims-King, 2013;Cunningham, 1990;Hatcher et al., 1994;Yeung et al., 2013).
The study has some limitations. Studies involving larger numbers of participants are recommended with matched monolingual speakers if one is to confirm the causal significance of these findings. The issue of the phonological awareness tests used not being strictly comparable down to the phonological unit may prove somewhat more challenging to address. This study adds to a small but growing indeed transfer between languages. The predictive relationship between phonological skills and children's word reading skills in this study supports both a small language-specific component and a significantly larger common set of phonological skills driving reading development across the two alphabetic languages. Share's (2008) contention that reading science cannot be founded on a single orthography is to be heeded because such studies are likely to offer a better approximation to the global norm of reading development.