Transcript
Reading and Writing: An Interdisciplinary Journal 17: 59–78, 2004.© 2004 Kluwer Academic Publishers. Printed in the Netherlands. 59 Processing phonological information in a semi-syllabic script:Developmental data from Telugu DUGGIRALA VASANTA Department of Linguistics, Osmania University, India Abstract. Three experiments were undertaken to examine second and fifth grade Telugu-speaking children’s awareness of phonological and orthographic properties of familiar Teluguwords. Experiment 1 focused on the strategies the children used in completing word frag-ments. Experiment 2 examined the children’s ability to judge and generate rhyming words,and Experiment 3 examined the children’s strategies in comprehending meanings of ortho-graphically similar rhyming vs. non-rhyming word pairs in a sentence completion task. Theresults demonstrated that specific features of semi-syllabic alphabets such as Telugu interactwith phonological knowledge during the processing of meaningful words such that childrenwith more formal instruction in reading are able to access phonological information betterthan younger children with less well developed orthographic knowledge. Some pedagogicimplications of the results are discussed. Key words: Orthographic knowledge, Phonological awareness, Syllabic alphabet, Telugu Introduction The earliest and the most common pedagogic activity school childrenencounter is learning to represent speech sounds and segments graphically.The proponents of the linguistic approach to beginning reading have arguedthat this involves learning a code that maps print to speech (e.g., Perfetti,1985, 1994). During their primary school years, children acquire not only thewriting system associated with the language in which they receive instruction,but also a high degree of attention-free and automatic use of that code, in partbecause their scholastic achievement is chiefly measured in terms of how wellthey can read and write.The rate of acquisition and mastery of the writing system in a givenlanguage may in turn depend upon the size of the speech unit representedin the script. Writing systems differ in the size of the unit represented. Thus,in the morphosyllabic writing system of classical Chinese, morphemes arethe units of reference, whereas in alphabetic writing systems (e.g. English),phonemes are the unit of reference and in syllabic scripts such as theJapanese kana script, the syllable is the unit. Hebrew, Persian, and manyIndic scripts, including Telugu, make use of a modified alphabetic system in 60 DUGGIRALA VASANTA which consonants are more reliably represented than vowels in the script. Thesmaller the unit of speech that is represented in a script, the more constrainedthat script is by the spoken language. Examination of the regularities betweenthe script and the sound, and between the script and lexical meaning is funda-mental to the psychological study of the performance of readers and spellers(Henderson, 1982).Considerable evidence exists in relation to the processes underlyingacquisition of beginning reading skills in English and other alphabetic scripts.Much of the evidence, based on correlational as well as training studies,shows that phonological awareness contributes to decoding of nonwords aswell as comprehension of common words (see Goswami & Bryant, 1990;Treiman, 1991; Goswami, 1999a). Phonological awareness includes aware-ness of phonemes, syllables and intra-syllabic features such as onsets andrimes. It is generally agreed that syllable awareness precedes phoneme aware-ness as well as that of sub-syllabic units. There has been much debateregarding the relationship between phonemic awareness and success in begin-ning reading in English (Bradley & Bryant, 1983; Bryant, Maclean, Bradley& Crossland, 1990). In addition, there has been some discussion on whetherthe relationship is bi-directional, that is, whether early phonological aware-ness predicts later reading development or whether knowledge of spellinginfluences children’s conceptualization of the phonemic structure of words.Reading ability has been shown to have a positive correlation with knowl-edge of orthographic structure (see Ehri & Wilce, 1980; Massaro & Hestand,1983). Proponents of the view that phonological awareness is an importantaspect of literacy acquisition have also argued for early assessment andinstruction in phonological awareness by parents and teachers (Rappaport &Hunter-Carsch, 1999).Treiman, Mullennix, Bijeljac-Babic and Richmond-Welty (1995) con-ducted a statistical analysis of the links between spellings and sounds inEnglish words with consonant, vowel, consonant (CVC) syllable structure.The results revealed that orthographic units consisting of a vowel and finalconsonant (VC) had a more stable pronunciation than either individual vowelgraphemes or initial consonant plus vowel (CV) units. The authors concludedthat characteristics of the English orthography encourage readers to use anonset-rime parsing strategy when reading words. They further commentedthat the link between orthographic structure and phonology is no accident aswriting systems are designed to encode the spoken language.Miller (1994) argued that the syllable structure and word structure in allnatural languages are motivated by a sonority hierarchy. According to thisprinciple, segments can be ranked on a sonority scale (stops < fricatives < nasals < liquids < glides < vowels) in such a way that segments ranking TELEGU DEVELOPMENTAL STUDY 61higher in sonority (e.g. liquids and glides) stand closer to the center of thesyllable whereas segments ranking lower insonority (e.g. stops and fricatives)stand closer to the margin (see Clements, 1990, for details). In the courseof phonological acquisition, young English speaking children reportedlyproduce the most optimal or preferred syllables (those that honor the sonorityprinciple) when reducing clusters (Ohala, 1995). Sonority of phonemes hasalso been shown to influence English-speaking children’s segmentation of syllable constituents (Yavas & Gogate, 1999). A study based on native andbilingual Filipino readers of English demonstrated that sonority constraintsunderlie onset–rime cohesiveness in addition to playing a role in visual wordrecognition of monosyllabic English words containing clusters in the initialposition. Despite properties in their native language which appear to beincongruent with the onset-rime division, native Filipino speakers processedEnglish words similarly to native English speakers (Alonzo & Taft, 2002).Studies which seek to understand how two very different writing systems,such as English and Chinese, make use of phonological information in visualword recognition suggest that, in each writing system, higher frequencywords tend to be recognized on a visual basis, without phonological medi-ation, and that phonology enters into processing primarily in the case of low frequency words (Seidenberg, 1985). However, recent research hasshown that both familiarity and speed of naming affects decoding of letters(see Joshi & Aaron, 2002). Another study of cross-linguistic comparisonsof English- vs. Chinese-speaking children revealed that English speakingchildren developed an awareness of rimes and initial consonants (onsets)earlier than their Chinese counterparts (Ho & Bryant, 1999). The authorssuggest that this finding has something to do with the disparities in theorthographies of the two languages. Ho and Bryant concluded that bothphonological and orthographic systems of a language have an impact on therate and pattern of development of phonological awareness.Using the nonsense word reading technique with English, Greek, Spanishand French speaking children, Goswami (1999a, b) hypothesized thatchildren who are reading very transparent orthographies (e.g. Greek andSpanish) develop robust phoneme-grapheme relationships much earlier thanthose learning less transparent orthographies such as English and French.Support was obtained for this suggestion but it remains to be determined howmuch of the difference noted is attributable to literacy instruction.Drawing on spoken word segmentation data from English, French andJapanese, Otake, Hatano, Cutler and Mehler (1993) demonstrated thatlisteners’ segmentation of spoken words relies on procedures determined bythe characteristic phonology of their native language. This point is elaboratedin another cross-linguistic study involving English, Japanese and Korean 62 DUGGIRALA VASANTA speakers reported by Kubozono (1996). This study suggested that speakers’knowledge of orthography plays a crucial role in determining the preferredmanner of speech segmentation, with VC-based segmentation in English andCV-based segmentation in Japanese. However, the reason for this prefer-ence might be related to the fact that English is rich in closed syllableswhereas in Japanese most words end with a vowel. Kubozono’s research hasdemonstrated that stress-based rhythm in English and mora-based rhythm inJapanese constrain the segmentation performance of speakers of these twolanguages. Korean, which makes use of a syllable-based native orthography(referred to as Hangul) reportedly behaves like Japanese in that the preferredsegmentation pattern is CV/C. This permits onset consonant and nuclearvowel to be written on the upper half, and the coda consonant on thelower half. More recently, Silva and Alves-Martins (2002) cited evidencethat children segment words better when the words are presented in awritten rather than a spoken format. Their own study, based on ninetyPortuguese-speaking five- to six-year-old children, showed a complex inter-action between the processes involved in becoming aware of the units inspeech and understanding the way in which the written code works. When theinfluence of the script is eliminated, phonological structure, especially speechrhythm, has been shown to affect the speech segmentation performance of native and non-native listeners of Japanese and Telugu (see Cutler, Murty &Otake, 2003, for details).Turning to studies relating to languages spoken in India, the limitedevidence confirms the existence of a bidirectional influence of phonologyand orthography on reading acquisition in general (see Patel & Soper, 1987,with respect to Gujarati), and speech segmentation in particular (Sailaja,2000). Three separate studies aimed at examining phonological knowledgeof Kannada-speaking elementary school children, monolingual Hindi literateand illiterate adults, and Kannada-English biliterate adults were reportedby Prakash, Rekha, Nigam and Karanth (1993). Based on these studies theauthors concluded that one’s ability to manipulate the structural featuresof language is facilitated by literacy in general, and by the features of thescript employed in particular. Vaid and Gupta (2002) reported a study dealingwith word recognition in Hindi, which makes use of the Devanagari script.Like most Brahmi-derived scripts in India, Devanagari has properties of bothsyllabic and alphabetic writing systems. Forty college age native speakers of Hindi served as participants in their first experiment, ten child native readersof Hindi (age range 7.5–10 years) took part in the second experiment andthe same ten children and twenty other college age adults participated in thethird experiment (Vaid & Gupta, 2002). Exploiting a rare disparity in Hindibetween spoken and written sequencing of letters (i.e., the case of the short TELEGU DEVELOPMENTAL STUDY 63“i” vowel), Vaid and Gupta (2002) investigated naming speed, accuracy andwriting order responses to determine if this mismatch in auditory (temporal)and visuo-spatial sequencing would incur a processing cost. The results infact revealed a cost. Furthermore, the study suggested that participants madeuse of a partly phonemic and a partly syllabic level of segmentation inprocessing Devanagari, mirroring the structural hybridity of the script. Amore recent study by Vaid (2002) examined Hindi–English bilingual adultson a task in which they were given cross-language homophones (i.e., wordsthat could “pass” as Hindi or English words) in separate language blocks.Participants were asked to delete the “first sound” of the word and pronouncethe remaining part of the word. The results showed a strong influence of orthographic knowledge in participants’ judgments of the first sound. Forexample, the initial sound for the word hum “we” in Hindi was perceivedto be CV whereas that for the same word presented as an English word wasperceived to be C (Vaid, 2002).The research reviewed above shows that oral language input helpschildren develop phonological awareness and that literacy instruction andscript-specific factors can influence that awareness. Miller (1994) cautionedresearchers against assuming that structural properties of scripts can impartall linguistic knowledge (i.e., knowledge required to segment continuousspeech into sentences, phrases, words, syllables, and phonemic units).According to Miller, the most one can say is that alphabetic knowledge helpsto transfer implicit segmental knowledge to a level of conscious awareness,which is still not the same as explicit metalinguistic knowledge. There isno point, Miller argues, in giving privileged significance to segmentation,as defined by the separate linear representation of consonants and vowelsin the case of alphabetic scripts, while ignoring syllable structure and thesonority principle. Instead, one mustexamine theentire range ofphonologicalknowledge from segments to syllables that is represented in scripts classifiedas primarily syllabic or primarily segmental. Vaid and Gupta’s (2002) studyillustrates this point with reference to Devanagari script.In what follows, I discuss how the orthographic system of Telugu, alanguage whose script is characterized as semi-syllabic, encodes phono-logical information and how children, in the course of reading acquisition,make use of this information in processing Telugu words. Characteristics of Telugu orthography Telugu is one of the four major Dravidian languages (the others beingKannada, Tamil and Malayalam) spoken primarily in South India. Thesouthern group of Indian scripts associated with these four languages datesback to the sixth to eighth century and are linked to Grantha, Kalinga and
