Phonemic Awareness in Chinese L1 Readers of English: Not Simply an Effect of Orthography

MCDOWELL and MARJORIE PERLMAN LORCH Birkbeck College, University of London London, England The current study investigates the phonemic awareness and nonword processing of English as a fo

Phonemic Awareness in Chinese L1

Readers of English: Not Simply an Effect

of Orthography

HEATHER J MCDOWELL and MARJORIE PERLMAN LORCH

Birkbeck College, University of London

London, England

The current study investigates the phonemic awareness and nonword processing of English as a foreign language students from Hong Kong and Mainland China, with reference to factors considered the main facilitators of phonemic awareness: written language experience, spo-ken language experience, and metalinguistic training The Mainland Chinese students were literate in Pinyin, an alphabetic representation

of Chinese, and were first language (L1) speakers of Mandarin Half of the Mainland Chinese students had also been exposed to the Interna-tional Phonetic Alphabet (IPA) in their second language (L2) reading education The Hong Kong students were not Pinyin literate and spoke Cantonese The Mainland Chinese IPA-trained participants performed better than both the Hong Kong participants and the Mainland Chi-nese non-IPA-trained participants in initial phoneme deletion How-ever, both Mainland Chinese groups outperformed the Hong Kong group on a phoneme–grapheme nonword matching task This pattern

of results suggests that phonemic awareness in Chinese L1 readers of English is not simply an effect of orthography, but rather, may be in-terpreted in terms of access to explicit demonstration of phonemes Further, tests carried out in L2 which are intended to assess metalin-guistic awareness may be susceptible to artefacts introduced by the participants’ L1 spoken language

INTRODUCTION

Phonemic Awareness Development

Phonological awareness (of which phonemic awareness forms one part) refers to explicit awareness of and access to the sound structure

of oral language (Wagner, 1988) and is often operationalised in terms of manipulation of the component sound units of a word, for example, the

ability to segment the word lip into /l/, /I/, and /p/ or to transpose the

initial phonemes of two words Natural speech perception and produc-tion do not require the language user to possess conscious awareness of the phonological structure of words; however, analysis and manipulation

of sublexical units of sound can be developed into a conscious metalin-guistic skill Phonological awareness is a multilevel ability, comprising various differentially acquired subskills (Treiman, 1983) at syllabic, in-trasyllabic,1and phonemic levels (see Goswami & Bryant, 1990, p 2) Awareness at these three levels advances at different developmental stages, involving different cognitive demands Perceptual analysis at the syllabic and intrasyllabic levels is thought to be an innate ability which emerges in the course of normal language development (Liberman, Shankweiler, Fischer, & Carter, 1974) However, experiments carried out over the last 40 years suggest that the development of intentional ma-nipulation of single sounds (phonemic units) requires specific facilita-tion Three major sources of phonemic awareness facilitation have been proposed: written language experience, spoken language experience, and metalinguistic training

Much phonemic awareness research has been conducted with regard

to written language experience Early studies discovered that, whereas preschoolers are unable to perform phonemic segmentation, this ability improves dramatically with the onset of English reading acquisition (e.g., Bruce, 1964; Liberman et al., 1974) Thus, explicit awareness of pho-nemes appears to develop out of alphabetic literacy The argument for orthographic experience as a facilitator of phonemic awareness was strengthened by the finding that illiterate adults, like preliterate chil-dren, are also unable to analyse words into constituent phonemes (Mo-rais, Bertelson, Cary, & Alegria, 1986) Meanwhile, it is also argued that there is a reciprocal relationship between alphabetic literacy and pho-nemic awareness, with phopho-nemic awareness equally predicting reading acquisition (e.g., McDougall, Hulme, Ellis, & Monk, 1994) Recent re-search is driven by the idea that the development of phonemic awareness provides an important substrate for the acquisition of alphabetic read-ing

Research indicates that literacy based on nonphonemic orthographic representations does not give rise to elaborated phonemic awareness (e.g., Mann, 1986; Padakannaya, 2000; Read, Zhang, Nie, & Ding, 1986) This is consistent with the notion that explicit demonstration of pho-nemes is required for the development of this metalinguistic skill In addition, the increased attention accorded to individual phonemes in a phonics approach to teaching reading may be the source of explicit

1Intrasyllabic phonological description defines the initial unit as the onset and the postinitial component as the rime: for example, in the word /fæt/, /f/ is the onset and /æt/ is the

rime.

phonemic awareness development in most individuals, rather than the ability to read an alphabetic script per se In this regard, various studies demonstrate that, in English L1 literacy instruction, phonics training is superior to the whole-word approach in raising phonemic awareness (see Adams, 1990) Finally, there is evidence that, within written language exposure, the transparency with which graphemes are mapped to pho-nemes also affects phonemic awareness development For example, read-ers of languages with a transparent alphabetic orthography (such as Italian) have been shown to have better-developed phonemic awareness than readers of an opaque orthography (e.g., Cossu, Shankweiler, Liber-man, Katz, & Tola, 1988)

A smaller pool of literature suggests that various features of spoken language experience can also promote phonemic awareness develop-ment For example, Caravolas and Bruck (1993) found that, even prior

to literacy instruction, Czech children show a greater level of phonologi-cal awareness for complex onsets (word initial sound units consisting of more than one phoneme) than their English peers The authors sug-gested that this is because of the rich inventory of complex syllable onsets

in the spoken language, requiring more detailed auditory discrimina-tion Similarly, Cheung, Chen, Lai, Wong, and Hills (2001) found a specific language effect when comparing native speakers of English and Cantonese They suggested that the more complex syllable structure of English (which permits up to three consonants in the onset and coda) as compared with Cantonese (where only one consonant is permitted in onset or coda position) led to superior performance on rime and coda matching Finally, Durgunog˘lu and Öney (1999) discovered that young Turkish children are better able to manipulate syllables and final pho-nemes than their English-speaking peers; the authors attributed this to various characteristics of spoken Turkish, including its vowel harmony and syllabic structure This spoken language effect can be described in terms of increased saliency leading to more explicit awareness

There is also widespread acknowledgment that oral metalinguistic training, without reference to literacy skills or orthographic form, can facilitate the development of awareness of phonemes For example, Lundberg, Frost, and Petersen (1988) successfully trained a group of English-speaking kindergarten children to segment words into pho-nemes verbally Similarly, subsegmental oral language games, explicit sound segmentation training, and corrective feedback have also been demonstrated to be effective facilitators of phonemic awareness (e.g., Cheung, 1999; Maclean, Bryant, & Bradley, 1987; Morais, Content, Ber-telson, Cary, & Kolinsky, 1988)

In addition to supporting L1 alphabetic reading acquisition, phone-mic awareness is also important for L2 English learners Many EFL stu-dents have pre-existing awareness of phonemes from L1 alphabetic

lit-eracy, and there is evidence to suggest that once this has developed, it can be applied to later-learned languages (e.g., Durgunog˘lu, 1998) However, individuals who have not previously had any explicit demon-stration of phonemes in their L1 will require specific exposure in order

to develop the phonemic awareness necessary for efficient L2 English decoding Byrne, Freebody, and Gates (1992) demonstrated that al-though individuals with poor decoding skills may painstakingly acquire a large reading vocabulary, underspecified phonemic awareness can lead

to later difficulties in reading fluently, particularly when encountering unfamiliar words

Phonemic Awareness in Chinese EFL students

The linguistic, educational, and literacy backgrounds of Hong Kong readers of English differ from those of readers from Mainland China, providing an interesting opportunity to investigate phonemic awareness development in L2 learners Both Hong Kong and Mainland Chinese use the same Chinese script for reading The characters used in Hong Kong and Mainland China differ slightly, because Mainland China has adopted simplified characters However, the phonological/semantic composition of the characters remains the same Unlike an alphabetic script used in written English, whose letters notionally represent the language in terms of its phonemes, even if not on a one-to-one basis, the Chinese script uses orthographic symbols to represent monomorphemic syllables in a nonlinear layout (see Wang, Koda, & Perfetti, 2003) The majority of Chinese characters are compounds containing a phonetic radical and a semantic radical which give an approximate guide to pronunciation and meaning, respectively Thus, in the character (ma¯–

mother), the semantic radical performs the categorisation female,

whereas the phonetic radical represents the sound of the whole char-acter However, the phonetic radical does not always prove a reliable pronunciation guide, and many characters do not have exactly the same pronunciations as their phonetic (Zhou, 1978) Thus, learning to read the Chinese script does not provide any explicit experience of individual phonemes

Whereas Mainland China and Hong Kong share the same writing system, the process of literacy education differs In Hong Kong, each orthographic character and its pronunciation are learned by rote, and, when English is introduced as an L2 in school, a whole-word reading approach is commonly taken (Cheung, 1999) Thus, Hong Kong Chi-nese children typically learn to read an alphabetic system in the form of English L2 without extensive exposure to explicit phonemic-level traing However, in Mainland China, L1 reading instruction initially

in-volves Pinyin orthographic representations Pinyin is an alphabetic sys-tem with a transparent orthography which uses Roman symbols to rep-resent spoken Chinese and is used as a pedagogical aid; it is believed to foster the development of phonemic awareness (Read et al., 1986) Be-cause of their exposure to Pinyin in learning to read, Mainland Chinese EFL students can be assumed to have developed phonemic awareness skills in L1 to apply to the task of learning to read in English In contrast, Hong Kong students who learn to read the logographic (morsyllabic) Chinese script without Pinyin may not have developed the pho-nemic awareness necessary for optimal alphabetic decoding

To test this distinction, Holm and Dodd (1996) administered tasks of phonological awareness, decoding, and spelling to university students studying in Australia They found that, whereas the Mainland Chinese participants performed at ceiling on most nonword decoding and verbal phonological awareness tasks, those from Hong Kong obtained signifi-cantly lower scores These participants demonstrated particular difficulty

in creating a phonological representation for unknown words, and ap-peared to have gained functional literacy in English without elaborated phonemic awareness Holm and Dodd’s (1996) finding of low phonemic awareness in Hong Kong Chinese L1 students with high levels of L2 English reading proficiency has been supported by various studies (e.g., Cheung, 1999; Bialystok, McBride-Chang, & Luk, 2005; but see Huang & Hanley, 1995 for a study with somewhat mixed results)

These findings call into question the primary role of alphabetic lit-eracy in the development of phonemic awareness If alphabetic litlit-eracy were the only key to such development, the Hong Kong students’ exten-sive experience of reading in English should surely lead to good perfor-mance on phonemic awareness tasks In the discussion of their results, Holm and Dodd (1996) concluded that the individuals they studied had

acquired functional alphabetic literacy without phonemic awareness

de-spite many years of L2 alphabetic reading Consequently, Cheung et al (2001) suggested that it may be necessary for the first-learned script to be alphabetic in order for phonemic awareness to be acquired However, whereas literacy instruction method was taken into account by Holm and Dodd (1996), a more detailed examination of other potential sources of phonemic awareness facilitation may be useful In particular, explicit training in the International Phonetic Alphabet (IPA)2 is sometimes used in L2 teaching as a tool to improve segmental pronunciation, and this may prove particularly effective in fostering explicit phonemic

aware-2 In-class IPA exposure typically takes the form of presentation of each sound and its cor-responding character The focus is on distinguishing and articulating the different pho-nemes of English Students are not encouraged to learn to “read” in IPA per se, but to use the IPA characters as an aid to accurate pronunciation.

ness IPA exposure may develop phonemic awareness in a different man-ner to the exposure of the alphabetic Pinyin script used in early literacy education

Phonotactic patterns within the first language may also be a source of variation in phonemic awareness Although both Mainland Chinese and Hong Kong Chinese readers use the same script, they speak a number of distinct languages Whereas Mandarin Chinese (Putonghua) is widely spoken in Mainland China, Cantonese is the dominant spoken language

in Hong Kong There are substantial phonological differences between these two so-called dialects of the Chinese language (Norman, 1988): they have different consonants, vowels, tones, and phonotactic con-straints on syllable form In addition, the form of English used in Hong Kong is a nonstandard variety which is heavily influenced by Cantonese phonology (Hung, 2002) This is in contrast to the more standard variety taught on the Mainland Both of these spoken language factors may affect phoneme perception by Chinese students studying English in an immersion environment (e.g., England) and may also have a bearing on performance in phonemic awareness tasks related to reading Compo-nent analyses of phonological awareness have suggested that speech per-ception makes a unique and substantial contribution to this construct (McBride-Chang, 1995), and this may be especially relevant to the per-formance of L2 learners Phonological representations are derived from L1 experience, such that for later second language learners, L2 pho-nemes are frequently interpreted in terms of pre-existing L1 categories (Wade-Woolley, 1999) Therefore, the L1 spoken language difference between Mainland and Hong Kong Chinese may be relevant to perfor-mance in phonemic awareness tests administered in L2 English

An additional consideration is that previous investigations of phone-mic awareness in L2 learners have frequently used relatively complex tasks (e.g., Holm & Dodd, 1996; Wang, Koda, & Perfetti, 2003) Even native-English-speaker controls did not perform at ceiling on the tasks used in the study by Holm and Dodd (1996) At this demanding level, students from Hong Kong performed relatively poorly, but it is yet to be determined whether there is a minimum level of phonological awareness which might be sufficient to support alphabetic decoding Native-speaker control groups used in previous L2 phonological awareness studies tend

to be university students with an extremely high level of education and extensive reading experience It is possible that atypically well-developed levels of metalinguistic skills exist in such readers Readers more typical

of the general population may function efficiently in an alphabetic script without attaining such fine-grained levels of phonological awareness Many questions concerning phonemic awareness in Hong Kong and Mainland Chinese readers of English have been raised by Holm and Dodd (1996) and subsequent studies The current investigation aims to

extend and refine these findings in several ways First, with respect to participant variables, detailed language education histories are used to document students’ previous exposure to explicit demonstration of pho-nemes Second, the current study uses two groups of students from Mainland China who had been controlled for IPA exposure Third, the study is carried out with English high school students rather than uni-versity students as controls Fourth, tasks of phonemic awareness drawn from previous studies are used only if control participants’ performance was found to be at ceiling (100% accuracy) Because these tasks are designed to tap skills assumed to be acquired by young children, perfor-mance by older teenagers that was not at ceiling would indicate the presence of a task artefact

METHOD

Participants

Students born in Hong Kong and Mainland China attending an in-dependent boarding school in England were recruited for this study There is one group of Hong Kong Chinese and two groups of Mainland

Chinese students The Hong Kong (HK) group (n = 14; 10 males, 4 females) and one Mainland Chinese (MC) group (n = 8; 3 males, 5

females) had minimal IPA knowledge, reporting that they had never learned it or were unable to match IPA symbols to phonemes; the other

Mainland Chinese (MCI) group (n = 8; 3 males, 5 females) had received

extensive exposure to IPA training All participants had been initially educated in their home country and were self-reported proficient L1 readers of Chinese with no known reading impairments All are aged between 15 and 19, and studying alongside British pupils in years 9–13 following the standard curriculum, with EFL (English as a foreign lan-guage) support where necessary All students are deemed to have achieved International English Language Testing System (IELTS) level 5.5, with no substantial difference in language proficiency between the groups, as rated by the school EFL department’s internal tests Partici-pants who had learned other languages with alphabetic writing systems, and those who had lived in an English-speaking country from infancy were excluded from the study All Mainland Chinese participants had learned Pinyin and were L1 speakers of standard Mandarin; Hong Kong participants had not learned Pinyin and were L1 speakers of Cantonese Participant characteristics for the three groups are given in Table 1 A

group of age-matched native English speakers (n = 16) is used as controls

to test the stimuli to be included in each task

Materials and Procedure

The current study uses stimuli from Holm and Dodd (1996) and Cheung (1999), which have been used to investigate phonological aware-ness in Hong Kong users of English These are used to investigate meta-linguistic ability (a) to manipulate individual phonemes and (b) to tran-scribe nonwords using grapheme-phoneme correspondence rules In contrast to the method used by Holm and Dodd (1996), only those tasks are used for which the native-speaker-control-group performance was at ceiling The rationale is that any difference found at this level would suggest a robust distinction at a basic level of phonemic awareness, com-monly assumed to be a prerequisite for efficient decoding These tasks

were pretested on age-matched native speakers (n = 16), who all

per-formed at ceiling All testing took place in individual 20-minute sessions

in a quiet room during the school day (See Appendix for the items that were featured.)

Initial Phoneme Deletion

In the initial phoneme deletion task, participants were asked to re-spond with the word which would be left after the initial phoneme is removed Twenty high-frequency monosyllabic English words were se-lected from the list used by Cheung (1999), in which the deleted pho-neme was always a consonant The resultant strings were all words, al-though they varied in frequency Items included two types of stimuli:

simple onsets with single initial phonemes (n = 12) and complex onsets

of consonant clusters (n = 8), which are thought to provide a more

rigorous test of phonemic segmentation ability (McBride-Chang, 1995) These were presented together in a pseudorandom fixed order (as listed

in the Appendix) Two practice items were given prior to the beginning

of testing All items were presented orally by the researcher, who also

specified the phoneme to be deleted, as follows: “Say share Now say it

TABLE 1 Group Means and Standard Deviations for Age, Years of English Literacy Exposure, and

Years of Residence in England

Hong Kong

Mainland China

Mainland China + IPA

English literacy exposure (years) 12.32 1.56 6.38 2.00 5.56 1.80 Residence in England (years) 2.83 1.64 2.16 1.16 1.60 0.62

Note IPA = the International Phonetic Alphabet; M = mean; SD = standard deviation.

without the /ʃ/ sound.” Corrective feedback was provided for the prac-tice items only

Responses to the phoneme deletion task were tape recorded and marked by two independent scorers There was 99.3% agreement be-tween scorers; when they disagreed, the opinion of a third judge was accepted Because this task was intended to measure segmentation ability only, allowance was made for nonnative-like pronunciation Only first responses were accepted, but no time limit was used

Phoneme–Grapheme Nonword Matching

Reading nonwords aloud is considered a pure measure of how pho-nological rules are applied (see Nag-Arulmani, Reddy, & Buckley, 2003) The phoneme–grapheme nonword matching task from Holm and Dodd (1996) was used to assess ability to apply phonemic awareness to decod-ing Using nonwords prevents participants from relying on semantic information, forcing them instead to apply grapheme-phoneme corre-spondence rules In this task, participants are required to listen to a

series of monosyllabic nonwords (n = 20) and select the appropriate

match from an array of four written stimuli: (1) the target nonword, (2)

a distractor nonword which has a different vowel nucleus, (3) a distractor nonword which has a different postvocalic consonant coda, and (4) a completely dissimilar letter string Ten of the items used were taken from Holm and Dodd (1996), and 10 additional comparable items were added The target stimuli were tape recorded by a native English speaker, with a 6-second gap between items

RESULTS

Phoneme Deletion

The results of the phoneme deletion task (see left columns in Figure 1) were submitted to a one-way analysis of variance (ANOVA) There was

a significant effect of group: F(2,27) = 5.08, p < 0.05, ␻ = 0.46 Posthoc comparisons using Hochberg’s GT2 test indicated that the MCI group

(mean [M] = 18.37, standard deviation [SD] = 1.68) significantly outper-formed both the MC group (M = 15.12, SD = 1.55, p < 0.05) and the HK group (M = 15.21, SD = 3.06, p < 0.05) The MC and HK groups did not

differ significantly from each other

A posthoc item analysis was carried out to investigate whether those

items with consonant clusters (n = 8) were more difficult than items with single-initial phonemes (n = 12), as has previously been documented in

phonemic awareness research (the results are presented separately in Figure 2) The HK group appeared to have specific difficulty with the

consonant clusters (M = 4.92, SD = 2.30), showing relatively good per-formance on those items with single phoneme onsets (M =10.29, SD =

1.93) Surprisingly, the MC group performed relatively poorly on items

with simple onsets (M =8.50, SD = 1.60) On the items with complex consonant onsets, their performance was fairly good (M = 6.62, SD =

1.30) The MCI group performed almost at ceiling on both types (single

phoneme onsets: M =11.25, SD = 0.88; consonant clusters: M = 7.12,

SD = 1.45).

Phoneme–Grapheme Nonword Matching

An ANOVA conducted on the results of the phoneme–grapheme non-word matching task (see right hand bars in Figure 1) showed a

signifi-cant effect of group: F(2,27) = 6.99, p < 0.01,␻ = 0.53 Posthoc

compari-sons using Hochberg’s GT2 test indicated that the HK group (M = 16.85,

SD = 1.74) differed significantly from both the MC group (M = 18.75,

SD = 0.86, p < 0.05) and the MCI group (M = 19.00, SD = 1.41, p < 0.01).

There was no significant difference between the two Mainland Chinese groups

FIGURE 1 Percentage Performance Scores by Group on the Phoneme Deletion and Phoneme–

Grapheme Nonword Matching Tasks