漢字習得和閱讀發展：部件覺識的角色

全文

(1)國立臺灣師範大學英語學系碩. 士. 論. 文. Master’s Thesis Department of English National Taiwan Normal University. 漢字習得和閱讀發展：部件覺識的角色. Chinese Character Acquisition and Reading Development: Role of Radical Awareness. 指導教授：陳純音博士 Advisor: Dr. Chun-yin Doris Chen 研究生：董子昀 Student: Tzu-yun Tung. 中華民國一百零三年六月 June, 2014.

(2) 摘要本研究旨在探討部件覺識在漢字習得和閱讀發展中所扮演的角色。關於漢字習得的討論包含透明度效應、脈絡效應、規則性效應、熟悉效應和年齡效應。另外，也嘗試為閱讀發展建立預測因子。施測的四項理解能力測驗包含字詞語意測驗、句子語意測驗、語音測驗和中文閱讀理解測驗，研究對象則是來自台灣的二十一位小二學生、二十位小四學生和二十二位小六學生。研究結果顯示，意符覺識能幫助孩童習得透明形聲字的字義，但無法輔助不透明形聲字的習得。而且，與字詞脈絡相比，句子脈絡更能輔助漢字字義的提取。另一方面，音符覺識能促進孩童習得規則形聲字的字音和字義，但對不規則形聲字則無影響。而在兩種覺識中，唯有意符覺識能預測孩童的閱讀發展。將熟悉度和年齡納入考量後，結果顯示部件覺識只有在孩童面對不熟悉的字詞時才會發揮顯著效果，在面對熟悉的字詞時則無顯著效果。另外，小二和小四學生皆已具備用於提取字義的意符覺識，和用於提取字音的音符覺識。但小六學生進一步發展出用於提取字義的音符覺識，且有足夠成熟的意符覺識作為閱讀發展的預測因子。最後，本研究亦嘗試在教學順序和重點上提出建議。首先，為了讓教學由簡入繁，透明形聲字的教學可先於不透明形聲字，而規則形聲字的介紹則可早於不規則形聲字。另外，為了提升意符覺識進而促成閱讀發展，教學可多著墨於透明形聲字。而不透明形聲字的字義和不規則形聲字的字音字義，則是掌握不熟悉字詞的關鍵，可由教師特別提點。. 關鍵詞：意符覺識、音符覺識、漢字習得、閱讀發展、第一語言習得. i.

(3) ABSTRACT The present study set to investigate the role of radical awareness in Chinese character acquisition and reading development. For character acquisition, the effects of transparency, context, regularity, familiarity and age were examined. And for reading development, a reading correlate was established. Four comprehension tasks (i.e., a semantic task in words, a semantic task in sentences, a phonetic task, and a reading comprehension task) were conducted. And the participants were twenty-one second graders, twenty fourth graders and twenty-two sixth graders in Taiwan. It was found that semantic radical awareness facilitated the meaning acquisition of transparent phonetic compounds but not that of the opaque ones, and that the semantic access of characters was enhanced in the sentence-level context compared to the word-level context. On the other hand, phonetic radical awareness benefited the participants’ pronunciation and meaning acquisition of regular phonetic compounds but not that of the irregular phonetic compounds. Concerning the two kinds of awareness, only semantic radical awareness was found to uniquely predict subsequent reading ability of the participants. Further consideration of familiarity and age reported that the effect of radical awareness was significant only in the unfamiliar characters, not in the familiar ones. And while semantic radical awareness in semantic access and phonetic radical awareness in phonological access already manifested themselves in the second and fourth graders, only the sixth graders showed phonetic radical awareness in semantic access, and had semantic radical awareness as a predictor of their reading ability. Finally, the pedagogical implications of the present study pertained to the teaching sequence and teaching emphases. To begin with, the transparent phonetic compounds are suggested to be introduced prior to the opaque ones, and the regular ii.

(4) phonetic compounds prior to the irregular ones. Furthermore, the transparent phonetic compounds can be especially stressed in class to enhance semantic radical awareness and subsequent reading ability. The general word knowledge of opaque and that of irregular phonetic compounds can also be emphasized because they have been found to be the key to master unfamiliar characters.. Keywords: semantic radical awareness, phonetic radical awareness, Chinese character acquisition, reading development, first language acquisition. iii.

(5) ACKNOWLEDGMENTS The completion of my graduate study relied largely on the help and support from numerous individuals. I would like to thank my advisor, Dr. Chun-yin Chen, who ignited my interest in language acquisition and guided me step by step in research and thesis writing. Without her incredible passion and patience, I would not have finished this thesis. My sincere appreciation also goes to my committee members—Dr. Shiao-hui Chan and Dr. Rueih-ling Fahn. Dr. Chan gave me valuable suggestions on experimental design and data analysis while Dr. Fahn helped solidify my theoretical background. Their insightful comments improved the thesis drafts greatly. I would also like to express my gratitude to all the professors who taught me in the graduate years—Dr. Yung-o Biq, Dr. Shiao-hui Chan, Dr. Chun-yin Chen, Dr. Miao-ling Hsieh, Dr. Chia-lin Lee, Dr. Jen-i Li, Dr. Chien-jer Lin, Dr. Hui-shan Lin, Dr. Hsi-yao Su, Dr. Jen Ting, Dr. Hsiao-hung Wu and Dr. Jing-lan Wu. They manifested to me the breadth of the linguistic field and the true spirit of a linguist. Their inspiring lectures and lively in-class and after-class discussions equipped me both academically and spiritually. Moreover, I am extremely thankful to all experiment participants—the 63 elementary school students, their parents and teachers. Their enthusiastic participation and thoughtful cooperation made the research possible. I am fortunate to have the company of several lovely classmates in National Taiwan Normal University—Becky, Brian, Edea, Gracie, Horace, Mark, Nik, Queenie, Tiffany, Vicky, and Vivi. Thank you for your constant encouragement and assistance. I am also grateful to all my friends, who lifted my spirit and provided generous help at all times. Finally, I am deeply indebted to my family. Their unconditional love and support iv.

(6) sustained me through the graduate study. Without them I would not have had the chance to embark on this journey.. v.

(7) TABLE OF CONTENTS CHINESE ABSTRACT .................................................................................................. i ENGLISH ABSTRACT .................................................................................................ii ACKNOWLEDGMENTS ............................................................................................ iv TABLE OF CONTENTS .............................................................................................. vi LIST OF TABLES ........................................................................................................ ix LIST OF FIGURES ....................................................................................................... x CHAPTER 1. INTRODUCTION .................................................................................. 1 1.1 Motivation ........................................................................................................ 1 1.2 Theoretical Framework .................................................................................... 2 1.3 Research Questions .......................................................................................... 3 1.4 Significance of the Study ................................................................................. 4 1.5 Organization of the Thesis ............................................................................... 4 CHAPTER 2. LITERATURE REVIEW AND A NEW CLASSIFICATION OF CHINESE CHARACTERS ........................................................................................... 5 2.1 Models of Chinese Character Recognition ...................................................... 5 2.2 Chinese Reading Acquisition ......................................................................... 12 2.3 Radicals in Chinese Reading Acquisition ...................................................... 18 2.3.1 Shu and Anderson (1997) ................................................................... 18 2.3.2 Shu, Anderson, and Wu (2000) .......................................................... 21 2.3.3 Ho, Ng, and Ng (2003) ....................................................................... 22 2.3.4 Summary ............................................................................................ 26 2.4 A New Classification of Chinese Characters ................................................. 28 2.5 Summary of Chapter Two .............................................................................. 37 CHAPTER 3. RESEARCH DESIGN .......................................................................... 39 vi.

(8) 3.1 Participants ..................................................................................................... 39 3.2 Materials and Methods ................................................................................... 41 3.3 Procedures ...................................................................................................... 46 3.3.1 Formal Study ...................................................................................... 46 3.3.2 Data Analysis ...................................................................................... 47 3.4 Summary of Chapter Three ............................................................................ 48 CHAPTER 4. RESULTS AND DISCUSSION ........................................................... 49 4.1 Overview ........................................................................................................ 49 4.2 The Acquisition of Transparent and Opaque Phonetic Compounds .............. 51 4.2.1 Comparisons between Transparent and Opaque Phonetic Compounds ......................................................................... 51 4.2.2 Discussion .......................................................................................... 52 4.3 Contextual Effect ........................................................................................... 56 4.3.1 Comparisons between ST-W and ST-S............................................... 56 4.3.2 Discussion .......................................................................................... 57 4.4 The Acquisition of Regular and Irregular Phonetic Compounds ................... 59 4.4.1 Comparisons between Regular and Irregular Phonetic Compounds .. 59 4.4.2 Discussion .......................................................................................... 62 4.5 Reading Correlate .......................................................................................... 68 4.5.1 Radical Awareness as Reading Correlate ........................................... 68 4.5.2 Discussion .......................................................................................... 70 4.6 Familiarity Effect ........................................................................................... 72 4.7 Age Effect ...................................................................................................... 75 4.8 Summary of Chapter Four ............................................................................. 76 CHAPTER 5. CONCLUSION..................................................................................... 77 5.1 Summary of the Major Findings .................................................................... 77 vii.

(9) 5.2 Pedagogical Implications ............................................................................... 78 5.3 Limitations of the Present Study and Suggestions for Future Research ........ 79 BIBLIOGRAPHY ........................................................................................................ 81 APPENDIX A: THE CLASSIFICATION AND DISTRIBUTION OF TARGET CHARACTERS ........................................................................................................... 84 APPENDIX B: THE COMPLETE SEMANTIC TASK IN WORDS (ST-W) ............ 87 APPENDIX C: THE COMPLETE SEMANTIC TASK IN SENTENCES (ST-S) ..... 93 APPENDIX D: THE COMPLETE PHONETIC TASK (PT) .................................... 100 APPENDIX E: AN EXAMPLE ITEM OF CHINESE READING COMPREHENSION TEST ....................................................................................... 106 APPENDIX F: CONSENT FORM ............................................................................ 108. viii.

(10) LIST OF TABLES Table 2-1 Major Findings and Limitations of Previous Studies .................................. 26 Table 2-2 The Classification of Target Characters ....................................................... 37 Table 3-1 Information of the Participants .................................................................... 40 Table 3-2 The Classification and Number of Target Characters .................................. 43 Table 3-3 A Test Item of Semantic Task in Words (ST-W) .......................................... 44 Table 3-4 A Test Item of Semantic Task in Sentences (ST-S) ...................................... 45 Table 3-5 A Test Item of Phonetic Task (PT) ............................................................... 46 Table 4-1 Summary of the Interaction Effects in Repeated Measures Four-way ANOVA’s ..................................................................................................... 50 Table 4-2 Summary of the Interaction Effects in Repeated Measures Two-way ANOVA’s ..................................................................................................... 51 Table 4-3 The Performance of Each Grade on Transparent and Opaque Phonetic Compounds at Different Familiarity Levels in ST-W and ST-S .................. 52 Table 4-4 The Performance of Each Grade in ST-W and ST-S.................................... 57 Table 4-5 The Performance of Each Grade on Regular and Irregular Phonetic Compounds at Different Familiarity Levels in PT....................................... 60 Table 4-6 The Performance of Each Grade on Regular and Irregular Phonetic Compounds (at Different Familiarity Levels) in ST-W and ST-S ............... 61 Table 4-7 Multiple Regression Explaining Reading Ability of Each Grade from Character Knowledge................................................................................... 70 Table 4-8 The Performance of Each Grade at F Level ................................................ 73 Table 4-9 The Performance of Each Grade at F-1 Level ............................................. 74 Table 4-10 The Performance of Each Grade at F-2 Level ........................................... 75. ix.

(11) LIST OF FIGURES Figure 2-1 The Lexical Constituency Model of Perfetti et al. (2005) ........................... 8 Figure 2.2 The Character Recognition System of Taft (2006) ..................................... 10 Figure 3-1 Procedures of the Formal Study ................................................................. 47 Figure 4-1 The Developmental Stages of Radical Awareness ..................................... 76. x.

(12) CHAPTER 1. INTRODUCTION 1.1 Motivation The Chinese language has a very different writing system compared to the alphabetic scripts. A Chinese character generally represents one syllable and may be a word or a part of a polysyllabic word. In modern Chinese, more than 80% of the characters are compound characters that can be divided into a semantic component and a phonetic component (Wan, 1991). The semantic component may reflect the meaning of the character while the phonetic component may represent the pronunciation. The validity of the two components can vary. For example, the character “枝” (zhī, ‘branch’) has the semantic component “木” (mù, ‘tree’), which is related in meaning to the character itself. Besides, its phonetic component “支” (zhī, ‘branch’) is pronounced the same as the character. Moreover, the character “特” (tè, ‘special’) has the semantic component “牛” (niú, ‘cow’), which is unrelated in meaning to the character itself. And its phonetic component “寺” (sì, ‘temple’) has a different pronunciation compared to the character. Under such a system, character decomposition might be a beneficial strategy for L1 learners when they process Chinese characters. The natural question is whether it actually takes place and if yes, since when. Do children acquire characters as a whole or as combinations of components? Are they aware of the function of the components and can they make productive use of the components? Is there any developmental difference between the semantic and phonetic components? Do they process familiar and unfamiliar characters similarly? And does the richness of context (i.e., a word or a sentence) influence children’s employment of the strategy? Finally, the relationship between the awareness of components and reading development can be explored. The investigation might demonstrate the advantage of character decomposition in reading 1.

(13) not only words or sentences, but also the whole texts. The results might enhance our understanding of the intricate process of reading and provide possible ways to facilitate it. Previous studies (Ho, Ng, & Ng, 2003; Shu & Anderson, 1997; Shu, Anderson, & Wu, 2000) have investigated the component knowledge in some details, but the semantic and phonetic components have not been considered together. Besides, the role of phonetic components and context in meaning access receives less attention, and the reading development has often been examined at the word-level or sentence-level, but not at the discourse-level. In terms of the participants, they mainly came from Mainland China and Hong Kong. Few studies have focused on children in Taiwan. Therefore, the present study will try to add to the width and depth of the investigation on character acquisition and reading development by addressing the above issues. 1.2 Theoretical Framework The Chinese characters are originally divided into Wén ‘simple characters’ and Zì ‘composite characters’ according to Shuōwénjiězì ‘Explanation on Simple and Composite Characters’ written by Shen Xu. Simple characters cannot be further decomposed into meaningful components. Composite characters, on the other hand, are formed by the combination of simple characters and they can thus be decomposed. To be more specific, simple characters include Xiàngxíngzì ‘pictographs’ and Zhǐshìzì ‘simple ideographs’ while composite characters include Huìyìzì ‘compound ideographs’ and Xíngshēngzì ‘phonetic compounds.’1 Pictographs, to begin with, are formed by the illustration of concrete objects. The character “日” (rì, ‘sun’), for example, depicts the shape of the sun itself. Simple ideographs are formed by the. 1. The English translation of the four types of characters follows Martin (1972). 2.

(14) symbolic signs. “上” (shàng, ‘up’) is an example that indicates the abstract concept of ‘up.’ Turning to the composite characters, compound ideographs are formed by combining two or more characters together. An example of it is “明” (míng, ‘bright’), which combines the form and meaning of “日” (rì, ‘sun’) and “月” (yuè, ‘moon’). Phonetic compounds, however, are formed by combining Yìfú ‘semantic radical’ with Yīnfú ‘phonetic radical,’2 which indicate the meaning and the pronunciation of the whole characters, respectively. For example, the character “松” (sōng, ‘pine’) receives its meaning from the semantic radical “木” (mù, ‘tree’) and its pronunciation from the phonetic radical “公” (gōng, ‘male’). Therefore, whereas compound ideographs have combination of meanings, phonetic compounds have union of meanings and sounds. As mentioned before, validity of the semantic and phonetic radicals in phonetic compounds varies from character to character. “枝” (zhī, ‘branch’), for example, is transparent with a reliable semantic radical, “木” (mù, ‘tree’), which hints at the meaning of the character. It is also regular with a trustworthy phonetic radical, “支” (zhī, ‘branch’), to provide the pronunciation of the character. “特” (tè, ‘special’), on the other hand, is opaque with an unreliable semantic radical, “牛” (niú, ‘cow’), which relates little to the meaning of the character. It is also irregular with an untrustworthy phonetic radical, “寺” (sì, ‘temple’), which is pronounced differently from the character (d’Arcais, 1992; Ho et al., 2003; Shu & Anderson, 1997; Shu et al., 2000). 1.3 Research Questions Under this theoretical framework, the research questions of this thesis are as follows:. 2. Different English translations for Yìfú and Yīnfú have been proposed in the literature. Martin (1972) used ‘radical’ and ‘phonetic’ while Ho et al. (2003) referred to the two as the ‘semantic radical’ and ‘phonetic radical.’ The current study mainly follows Ho et al. (2003) for the clarity of the translation. 3.

(15) 1). Are the transparent phonetic compounds easier to acquire compared to the opaque phonetic compounds?. 2). Is the meaning retrieval of characters easier in the word-level context or in the sentence-level context?. 3). Are the regular phonetic compounds easier to acquire compared to the irregular phonetic compounds?. 4). How is semantic or/and phonetic radical awareness related to reading development?. 5). Does the effect of radical awareness interact with character familiarity?. 6). What are the developmental stages of radical awareness?. 1.4 Significance of the Study The present study will initiate a new framework under which both semantic and phonetic radical knowledge can be examined systematically. The development of the two radicals can be compared, and meaning access can be studied not only from the use of semantic radical knowledge as in previous studies, but also from the use of phonetic radical knowledge in the present study. The effect of context on semantic radical activation can also be revealed. In addition, the role of radical awareness in reading can be established not only at the word-level and sentence-level as it has been reported before, but also at the discourse-level. 1.5 Organization of the Thesis This thesis is organized as follows: Chapter Two summarizes some previous studies on Chinese character recognition and reading acquisition. A new classification of Chinese characters is also provided. Afterwards, Chapter Three details the research design of the present study. Chapter Four then presents and discusses the results while Chapter Five concludes the whole thesis. 4.

(16) CHAPTER 2. LITERATURE REVIEW AND A NEW CLASSIFICATION OF CHINESE CHARACTERS In this chapter, previous studies on Chinese character recognition, reading acquisition, and the role of radicals in both character recognition and reading acquisition will be examined, followed by a new classification of Chinese characters. Section 2.1 presents three models of Chinese character recognition, and specifies how radicals are involved in character recognition within different models. Section 2.2 provides a theoretical background for reading acquisition, and summarizes the established correlates of reading acquisition. Section 2.3 focuses on how radicals contribute to reading acquisition according to previous studies, and Section 2.4 proposes a new classification of Chinese characters based on radicals. Finally, Section 2.5 summarizes the whole chapter. 2.1 Models of Chinese Character Recognition The first model of Chinese character recognition to be reviewed in this section is proposed by Zhou and Marslen-Wilson (1999). Beginning with the lexical representation of Chinese characters in their model, all orthographic forms of Chinese morphemes are represented at the same level in the orthographic lexicon. The Chinese morphemes include characters, phonetic radicals and semantic radicals, where the radicals may or may not be real characters on their own. The orthographic representations have direct links with the phonological and the semantic representations. When reading complex characters, visual input is automatically decomposed into different orthographic units of various sizes, and the units map in parallel to the orthographic representations in the lexicon. Automatic activation in the phonological and semantic systems then follows the activation of orthographic representations (Zhou & Marslen-Wilson, 1999). 5.

(17) According to this model, no fundamental difference exists between lexical processing of characters and sublexical processing of radicals in reading Chinese. In the initial lexical access, phonetic radicals are automatically decomposed from the whole characters, and employed to access their own phonological and semantic representations in the lexicon. In parallel, phonetic radicals also contribute to the access to the phonological and semantic representations of the whole characters. On the other hand, semantic radicals are also automatically decomposed and employed to access their own semantic representations, though it remains unclear whether semantic radicals are employed to access their own phonological representations as well when they are characters on their own with definite pronunciation. Therefore, sublexical processing (of phonetic radicals at least) resembles lexical processing in that they are both phonological and semantic in nature (Zhou & Marslen-Wilson, 1999). The Lexical Constituency Model advocated by Perfetti, Liu, and Tan (2005) is the second model to be delineated. This model seeks to explain word reading across different writing systems, including both alphabetic word naming and nonalphabetic Chinese single-character word naming. In the Lexical Constituency Model, words are represented in an idealized mental lexicon, and a word representation comprises three interlocking constituents, which are orthography (OR), phonology (PH) and semantics (SE). Therefore, the word (ORa, PHb, SEc) would have the orthographic form ORa, the pronunciation PHb, and the meaning range SEc. In order to identify a word, the values of the three constituents must be specified. Identification failures would emerge if any of the three values is underspecified. In written word recognition, the PH and SE of a word would be retrieved from its OR. Recognition failures can be incomplete or asynchronous retrievals, namely, not yielding one of the word’s constituents or yielding the 6.

(18) constituent untimely. And the cause of the failures can lie in the retrieval processes or word knowledge that is incomplete or unreliable. According to the Lexical Quality Hypothesis (Perfetti & Hart, 2001), this word knowledge directly reflects one’s reading skill since this skill is measured by the amount of fully specified lexical representations in one’s mental lexicon. Furthermore, in written word recognition, PH processing precedes SE processing. This priority of PH originates from the Determinacy Principle, which states that at the word-level, OR-PH relations are more specified than OR-SE relations in most writing systems. In English for example, words like bass (one OR, multiple PHs, multiple SEs) are rarer than words like spring (one OR, one PH, multiple SEs). Bass has one orthographic form, but has multiple pronunciations with multiple meanings. When used as a noun and pronounced as /beɪs/, its meaning is related to musical range. But when pronounced as /bæ s/, it can refer to a kind of fish. Spring also has one orthographic form and only one pronunciation with multiple meanings. With the same pronunciation /sprɪŋ/, it can refer to a season or a piece of curved metal. An OR usually corresponds to a PH, but an OR corresponds to multiple SEs in many cases. The OR-PH relations are thus more determined than the OR-SE relations. Therefore, when viewing an orthographic input, a word’s pronunciation is more determined than its meaning. This can be observed in both English and Chinese, and is the reason why pronunciation would be retrieved prior to meaning. A computational instantiation of the Lexical Constituency Model is also provided in Perfetti et al. (2005). Figure 2.1 presents the model in schematic form.. 7.

(19) Figure 2-1 The Lexical Constituency Model of Perfetti et al. (2005). This implemented model is a network of linked units of the three constituents (OR, PH, SE) with activation spreading across. The input units (146 in total) are radicals and spatial relationship between the radicals. The spatial relationships include Left-right as in “攏” (lǒng, ‘to gather’), Top-bottom as in “藝” (yì, ‘art’), Closed outside-inside as in “國” (guó, ‘nation’) and Open outside-inside as in “同” (tóng, ‘identical’). The input units are connected to the OR level, which consists of character units (204 in total). The character units have connections to the PH level, which comprises onset, vowel and tone units (63 in total). Character units also have connections to the SE level, which are constituted by meaning units (204 in total). In addition, SE units are also connected to PH units. The model succeeds in simulating the primed naming results from Perfetti and Tan (1998), showing that graphic, phonological and semantic information successively affect target naming. Basing on the implemented lexical constituency model, Perfetti et al. (2005) 8.

(20) suggested that the differences between writing systems lie in the relevant OR and PH units. Models for alphabetic writing systems must represent OR units as letters (or subsymbolic equivalents), and connect them to units at the PH level. On the other hand, models for Chinese must represent OR units as radicals. The characters composed of radicals at the OR level then link to PH units, which are syllables. And to allow for a stronger role of radicals in the models, the phonetic radicals at the OR level can also be connected to their pronunciations at the PH level. In addition, the authors also stressed the importance of phonology in reading, claiming that due to the relative indeterminate from-meaning connections compared to form-form connections, meaning access for single or pairs of characters is more likely to be mediated by phonology. In other words, OR units activate PH units, which in turn activate SE units. This use of phonology reflects the general preference of readers to use writing system as a link to oral language, not as a new semiotic system in which OR units establish new links to SE units. The third model of Chinese character recognition to be reviewed is developed by Taft (2006) to explain the representation and retrieval of characters by adult native Chinese readers. Taft (2006) started his discussion with a general lexical processing system that includes a representation of the form and the meaning of every word a language user knows. For a reader, specifically, not only phonological but also orthographic form is represented in the processing system. Moreover, a lemma level mediates between form and meaning, and each lemma represents one concept. The orthographic, phonological and semantic subsystems all have independent links to the lemma level. And in reading, the system is entered through the orthographic subsystem, which enables access to the phonological and semantic subsystem. This character recognition system is visualized in Figure 2.2.. 9.

(21) Figure 2.2 The Character Recognition System of Taft (2006). In particular, the orthographic subsystem has three hierarchical levels, namely, the features, the position-sensitive radicals and the characters. The feature units represent the strokes and the positions of strokes where the positions include left-hand, right-hand, top and bottom. Relevant stroke feature units can directly activate a simple character unit such as “馬” (mǎ, ‘horse’). But to activate a composite character unit such as “媽” (mā, ‘mother’), relevant simple character units (e.g. “馬”) along with positional feature units (indicated by arrows in Figure 2.2) have to first activate the position-sensitive radical units (e.g. the unit that includes “馬” plus a rightward arrow in Figure 2.2), which in turn activate the composite character unit. However, if the radical of a composite character cannot be a free-standing character by its own (hence without a corresponding simple character unit), the position-sensitive radical unit will be activated by a radical unit instead of a simple character unit. At the top of the hierarchy in the orthographic subsystem, the character units are directly linked to the lemmas, which provide access to pronunciations and meanings. 10.

(22) The lemmas have links to pronunciations in the phonological subsystem, and they also have links to sets of semantic features for different concepts in the semantic subsystem. In summary, the three models of Chinese character recognition share some similarities, but also differ in critical respects. Concerning lexical representation, they all posit orthographic, phonological and semantic representations for characters. However, they vary in the nature and organization of orthographic representations, as well as how orthographic representations are linked to phonological and semantic representations. Zhou and Marslen-Wilson (1999) represent Chinese morphemes including characters, phonetic and semantic radicals at the same level in the orthographic lexicon, and all orthographic representations are directly linked to their corresponding phonological and semantic representations. Perfetti et al. (2005), on the other hand, represent characters and their componential radicals at the orthographic level. But only the character units are linked to their matching phonological and semantic units while the phonetic radical units may be linked to their matching phonological units. Taft (2006) further differentiates three hierarchical levels from bottom to top, which are the features, the position-sensitive radicals and the characters, in the orthographic subsystem. Activation passes from units at the bottom to units at the top, and only the character units are directly linked to the lemmas, which provide access to the phonological and semantic subsystems. The differences in lexical representation among the three models have direct consequences on how they explain the role of radicals in the character recognition process. Zhou and Marslen-Wilson (1999) contend that radicals will activate their own phonological and semantic representations and also contribute to the access to the phonological and semantic representations of the whole characters. But Perfetti et 11.

(23) al. (2005) only suggest phonological representations for phonetic radicals and their contribution to the activation of the pronunciations of the whole characters. And in Taft’s (2006) model, radicals do not have phonological or semantic representations, and are merely used to activate character units in the orthographic subsystem. Therefore, the contribution of radicals to the access to the meanings and pronunciations of characters is best captured in the first model while the meanings and pronunciations of radicals and how they influence the retrieval of whole characters receive less attention in the other two models (Perfetti et al., 2005; Taft, 2006). Zhou and Marslen-Wilson’s (1999) model might thus best explain the results of the present study. 2.2 Chinese Reading Acquisition Ziegler and Goswami (2005) developed a psycholinguistic Grain Size Theory to explain reading and its development. Reading is seen as a process of understanding speech written down with the goal of acquiring meaning. And learning to read is viewed as a process of matching visual symbols to units of sound. The process of learning and applying these mappings is referred to as phonological recoding. For children to succeed in phonological recoding, they need to find shared grain sizes in the orthography and phonology of their language. However, availability, consistency and granularity of spelling-to-sound mappings are three problems for the beginning readers. Firstly, children do not have conscious access to all phonological units prior to reading. Secondly, some orthographic units have multiple pronunciations and some phonological units have multiple spellings. And thirdly, when the phonological system is accessed through bigger grain sizes, more orthographic units have to be learned since words outnumber syllables, which outnumber rimes, which in turn outnumber graphemes, and graphemes outnumber 12.

(24) letters. The Chinese language particularly illustrates the grain size problem. Because one Chinese character corresponds to a syllable rather than a phoneme, children have to learn by rote about 3,000 characters that constitute the morpheme stock of Chinese, which will cost them a minimum of three years. Although rote learning is taught in Hong Kong, a phonetic script with subsyllabic segments and tone markers, Zhùyīnfúhào, was introduced in Taiwan while Pīnyīn, a phonetic script with alphabets and tone markers, was introduced in China prior to character teaching. The learning problem posed by grain-size correspondence is thus reduced (Ziegler & Goswami, 2005). With the background of the Grain Size Theory (Ziegler & Goswami, 2005), several empirical studies have sought to establish the correlates of Chinese reading in children. To begin with, Li, Shu, McBride-Chang, Liu, and Peng (2012) studied the metalinguistic and cognitive correlates of Chinese character recognition, and the correlates included visual skills, orthographic knowledge, phonological awareness, morphological awareness, and speeded number naming. The visual skills were knowledge of visual-spatial relationship and visual memory, and the orthographic knowledge was knowledge of line drawings, noncharacters, pseudo-characters and real characters. While phonological awareness was measured by rime detection, syllable deletion (for kindergarteners only) and phoneme deletion (for primary school students only), morphological awareness was measured by homophone judgment, morphological construction (for kindergarteners only) and morpheme production (for primary school students only). Finally, Chinese character recognition of children was assessed by reading aloud a list of 150 single characters. The participants were 85 children in the second year of kindergarten, 99 children in the third year of kindergarten, 97 students in Grade 1, 87 students in Grade 2 and 13.

(25) 89 students in Grade 3. All of them were from Beijing. And the results showed that syllable deletion, morphological construction and speeded naming were unique correlates of Chinese character recognition for kindergarteners. On the other hand, orthographic knowledge, rime detection, homophone judgment, morpheme production and speeded naming were independent correlates of Chinese character recognition for primary school students. Therefore, some dimensions of phonological and morphological awareness played an important role in both early and intermediate Chinese reading acquisition. Beginning with phonological awareness, syllable awareness was crucial since a Chinese character corresponds to a syllable. And in line with the Grain Size Theory (Ziegler & Goswami, 2005), which suggested that larger grain size units might associate with reading more if the orthography represented print in larger units, awareness of larger units such as syllables and rimes, but not phonemes, might better predict reading in Chinese. Besides, the morphological awareness measured highlighted the importance of the ability to distinguish homophones and lexical compounding, given the abundance of homophones and words comprised of two or more morphemes in Chinese print. As for visual skills, although they were significantly correlated with character recognition for kindergarteners, they were not uniquely associated with character recognition at any age level. Orthographic knowledge, however, was uniquely associated with character recognition for the primary school students but not the kindergarteners, indicating the importance of orthographic knowledge as reading experience increased. Taken together, the authors suggested that the participants might recognize characters by overall shape in kindergartens only. After they acquired some orthographic knowledge, including the legal positions of radicals and common and uncommon stroke patterns, they seemed to forsake the basic visual strategy and 14.

(26) undertake orthographic analyses instead. And Zhang (2012) only focused on phonological and morphological awareness, and whether the two abilities can predict character and word reading within and across Chinese and English for first and fifth graders. Phonological awareness was defined as the ability to manipulate sounds, including the addition, deletion and blending of phonemes, as well as sensing riming and alliteration. And it was measured by a phoneme deletion test in both Chinese and English. Morphological awareness, on the other hand, was defined as the ability to perceive and manipulate morpheme structure, including recognizing morphemes and constructing new words. It was measured by compound construction and derivational morphology in English, and measured by compound construction and homophone sensitivity in Chinese. The results showed that Chinese morphological awareness, rather than Chinese phonological awareness, was the unique predictor of Chinese character reading. Similarly, English morphological awareness was the unique predictor of English word reading. Awareness of English derivational morphology was additionally associated with Chinese character reading. Extending the scope of Chinese reading assessment from character level to sentence level, Zhang et al. (2012) investigated the association of vocabulary knowledge, morphological awareness and orthographic-semantic awareness to Chinese sentence comprehension. The three correlates respectively corresponded to meaning acquisition at the word, morpheme, and semantic radical levels. While vocabulary knowledge included oral definitions of words, morphological awareness included homophone production and morphological construction. Besides, orthographic-semantic awareness was measured by picking one out of four nonsense characters based on the meaning of pictures. The correct answer was the nonsense character with the semantic radical in the same position as that of the target real 15.

(27) character. And reading comprehension was measured by a picture selection task which assessed the meaning of 30 sentences. Additionally, since character recognition, phonological awareness and speeded naming were also related to reading comprehension, they were tested to serve as statistical controls. Chinese character recognition was measured by reading aloud a list of two-character words, while phonological awareness included syllable and onset phoneme deletion. The participants were 164 primary school students from Hong Kong, who were tested twice at the age of seven (Time 1) and eight (Time 2). After controlling Chinese character recognition, phonological awareness and speeded number naming, the results showed that Time 1 morphological awareness uniquely explained Time 1 and Time 2 reading comprehension, while Time 2 orthographic-semantic awareness uniquely explained Time 2 reading comprehension. Thus, the meaning-building variables explained unique variance in reading comprehension both concurrently and longitudinally. Zhang et al. (2012) argued that as readers constructed the meaning of text, both bottom-up and top-down processing were used. When encountering new words, morphological awareness let children guess the meanings by known morphemes and rules of morpheme combination, and orthographic-semantic awareness let them guess the meanings by visual cues and orthographic rules. Background knowledge, context or their existing schema then came into play. And the potential path from print to meaning via semantic radicals might be unique to Chinese as compared to alphabetic languages. Apart from the establishment of correlates of Chinese reading in children, studies on reading pedagogy also emerged. Wen (2010) proposed a reading strategy based on the semantic radicals for the beginning foreign learners of Chinese character. It was found that characters sharing a semantic radical tended to have a particular part of 16.

(28) speech based on the semantics of the radical. Therefore, the author suggested that when facing new characters, learners could be taught to make use of the semantic cue of the semantic radical to determine the probable parts of speech of the characters containing the radical. Combining with the instruction of possible parts of speech of the subject, verb and object in a sentence, this reading strategy might benefit the readers. In sum, reading acquisition is a process of matching orthographic units to phonological units in the view of Ziegler and Goswami (2005). Therefore, children need to discover the shared grain sizes in orthography and phonology of their language, and they need to solve the availability, consistency and granularity problem of spelling-to-sound mappings. Given this theoretical background, several experimental studies sought to establish the correlates of reading acquisition in children. Li et al. (2012) found that in line with the Grain Size Theory (Ziegler & Goswami, 2005), phonological awareness of larger units such as syllables and rimes, but not phonemes, was uniquely correlated with Chinese character recognition in kindergarteners and first to third graders. Similarly, Zhang (2012) reported no unique role of phoneme awareness in Chinese character reading for first and fifth graders. However, other than phonological awareness, at least two other correlates had been found to influence reading acquisition (Li et al., 2012; Zhang, 2012; Zhang et al., 2012). One correlate was morphological awareness, which could also uniquely predict character recognition of children in the study of Li et al. (2012). This correlation was built on the abundance of homophones and lexical compounding in Chinese, and was found in the reading acquisition of kindergarteners and first to third graders. Zhang (2012) also established morphological awareness as the unique predictor of character reading for first and fifth graders. And extending the scope of reading acquisition from character recognition to sentence comprehension, Zhang et al. (2012) found that 17.

(29) morphological awareness of the seven-year-olds could uniquely explain sentence comprehension concurrently and longitudinally. Another correlate was orthographic knowledge, which was uniquely associated with first to third graders’ character recognition in the study of Li et al. (2012). The orthographic knowledge included the legal positions of radicals and common and uncommon stroke patterns. Furthermore, stressing the position and semantic information. of. the. semantic. radicals,. Zhang. et. al.. (2012). found. that. orthographic-semantic awareness of the eight-year-olds uniquely explained sentence comprehension concurrently. The semantic cue of the radicals also inspired Wen (2010) to propose a reading strategy for beginning foreign learners of Chinese character. This strategy made use of the semantics of the radicals to infer probable parts of speech of the characters containing radicals. 2.3 Radicals in Chinese Reading Acquisition This section focuses exclusively on the role of radicals in Chinese reading acquisition. Three studies investigating the awareness of semantic and phonetic radicals, and how the awareness relates to reading ability of children will be reviewed. Section 2.3.1 reviews Shu and Anderson (1997) while Section 2.3.2 discusses Shu et al. (2000). And Section 2.3.3 reports the study of Ho et al. (2003). 2.3.1 Shu and Anderson (1997) Shu and Anderson (1997) investigated semantic radical awareness in Chinese character acquisition with two experiments. In the first experiment, they addressed the question of whether children know the function of semantic radicals by manipulating the familiarity and morphology of the characters. Sixty-seven first graders, 71 third graders and 82 fifth graders were recruited. They were given a paper-and-pencil multiple choice test containing 90 two-character words, with the target character 18.

(30) appearing in Pīnyīn and the nontarget character appearing in character. The participants had to replace the Pīnyīn with one of four characters. The three distractors either had the same sound components but different semantic radicals from the composite target character, or were visually similar to the simple target character. For example, for the item tiào “望” (wàng, ‘to look afar’), “眺” (tiào, ‘to look from a high place’), “挑” (tiāo, ‘to pick’), “跳” (tiào, ‘to jump’) and “佻” (tiāo, ‘skittish’) were provided as options. Furthermore, in order to measure the participants’ semantic radical awareness, namely, their ability to predict the written form of words they had not seen in print before, at least two out of four characters were visually unfamiliar to them. It was found that their participants performed better on familiar characters than on less familiar ones. They also performed better on transparent characters than on opaque and simple ones, especially when the characters were unfamiliar to them. Shu and Anderson (1997) thus argued that children were aware of radicals and that radical analysis was helpful in character learning. A developmental progression was also reported, showing that first graders did not have clear semantic radical awareness, whereas third and fifth graders were able to use the semantic information of the radicals to derive unfamiliar characters and to remember recently learned ones. For each grade, the participants were further assigned to high, average and low Chinese reading levels based on their teachers’ ratings. The test results revealed that children from different reading levels performed equally well on familiar characters, but when facing less familiar characters, those children at a higher reading level scored statistically significantly higher. This better performance of the children at a higher reading level might be attributed to either their broader general knowledge of characters, or their better ability of lexical decomposition and morphological analysis. In the second experiment, Shu and Anderson (1997) explored the question of 19.

(31) how the participants used radicals to derive the meanings of unfamiliar characters by varying the morphology of the characters as well as the conceptual difficulty of the words. Thirty-nine third graders and 33 fifth graders took a test similar in form with Experiment One. Each grade received 60 two-character words and had to select the right target character written in Pīnyīn from four options. The results showed that the participants performed better on characters that were constituents of easy words than the ones that were constituents of difficult words. They also performed better on transparent characters with familiar radicals than on transparent characters with unfamiliar radicals and opaque characters. This control of familiarity of semantic radicals led Shu and Anderson (1997) to confirm that the participants’ better performance on transparent characters with familiar radicals was due to radical analysis of the children. A further interaction between conceptual difficulty and morphology revealed that the participants scored the highest when the characters were transparent with familiar radicals and were contained in conceptually easy words. Therefore, both conceptual difficulty and morphology were suggested to be taken into account when children derived the meanings of unfamiliar characters. Moreover, the participants were also rated by their teachers as having low, average or high reading level. Those with higher reading ability only performed better on transparent characters with familiar radicals while no difference of performance was obtained with transparent characters with unfamiliar radicals or with opaque characters. Consequently, Shu and Anderson ascribed the advantage of better readers to better morphological analysis instead of broader general knowledge of words. Shu and Anderson (1997) concluded that third and fifth graders already had semantic radical awareness. They could infer the meanings of unfamiliar characters by extracting the semantic information of the radicals and integrating the information with the meanings of the whole words. Specifically, successful morphological 20.

(32) analysis required knowledge of semantic radicals, knowledge of word meanings and the ability to carry out morphological analysis. And instruction in Chinese morphology or strategies in using morphology might benefit poor readers. 2.3.2 Shu, Anderson, and Wu (2000) Shu et al. (2000) examined the development of phonetic radical awareness during the elementary school years. Forty-two second, 36 fourth and 35 sixth graders were asked to write down Pīnyīn for a set of 60 characters. They were found to perform better on familiar characters than on less familiar ones, better on regular characters than on irregular and bound ones, especially when the characters were unfamiliar to them. Therefore, Shu et al. (2000) claimed that children could and did use the phonological information in the phonetic radical to memorize the pronunciations of known characters and to anticipate the pronunciations of unknown characters. The phonetic radical awareness was already evident in second graders, and continued to develop in fourth and sixth graders. However, compared to the performance of fourth and sixth graders, the performance of second graders was still influenced more by their familiarity of characters, showing that the younger children tended to learn characters by rote. Fourth and sixth graders nevertheless made more use of the phonetic information, and could better pronounce unfamiliar characters. The participants were also rated by their classroom teachers as at high, average, or low reading ability level. And the results of the test showed that when facing familiar characters, the performance of the children at the three reading ability levels did not differ in terms of character regularity. But when facing unfamiliar characters, those children at a higher reading ability level could infer the pronunciation of regular characters from their phonetic radicals more than those at a lower reading ability 21.

(33) level. Moreover, three types of errors made by the participants were discussed. The first type was a phonetic error, which resulted from the misuse of phonetic radicals in the pronunciation of irregular characters. The second type was an analogy error, which occurred when the participants mispronounced a character the same as another character sharing the same phonetic radical. The last type of error was nonphonetic, which included no response and overt errors. Shu et al. (2000) found that those children with a higher grade also had a higher proportion of phonetic errors and analogy errors, which they referred to as phonetic-related errors. The errors of second graders at all reading ability levels indicated that the younger children made little use of phonetic analysis. In contrast, the errors of sixth graders at all reading ability levels showed their more robust use of phonetic analysis. Fourth graders, on the other hand, appeared to be in the middle. The lower reading ability ones behaved more like second graders, whereas the higher reading ability ones shared similar error pattern as sixth graders. Shu et al. (2000) concluded that phonetic radical awareness could be found in children as young as the second grade, and continued to develop in fourth and sixth graders, who were influenced more by phonetic regularity and made more phonetic-related errors. This knowledge of orthography-phonology relationships provided a nonarbitrary way of encoding characters, and the decomposition of characters into their phonetic and semantic radicals brought about successful retrieval of the pronunciations and meanings of characters. 2.3.3 Ho, Ng, and Ng (2003) Ho et al. (2003) conducted two studies to examine the influence of radical knowledge in Chinese reading development. The first study was devoted to the 22.

(34) semantic radical knowledge, which included the position, function and specific semantic categories of semantic radicals. Sixty Cantonese-speaking children in Hong Kong, 20 each in Grades 1, 3 and 5, were recruited. They first received an IQ test, Raven’s Standard Progressive Matrices, which measured children’s nonverbal reasoning ability. Afterwards, two reading tests were administered. The first was Chinese Word Reading Test, which required the participants to read aloud 65 Chinese two-character words of the primary school level. The second was a Chinese sentence comprehension test with 30 different sentences for each grade. The participants had to choose from four options a character that best completed the sentence. Finally, there were five semantic radical tasks for them to complete. The first one was a character decision. task,. including. 16. familiar. characters,. 16. rare. characters,. 16. pseudocharacters and 16 noncharacters. They had to judge whether a stimulus looked like a Chinese character. The second task was a radical position judgment task, which required the participants to indicate the legal position of 40 semantic radicals as left, right, top or bottom. The third task was a semantic-relatedness judgment task with 30 pseudocharacters as targets. For each target, the participants selected from three options the character that related semantically to the target character. The three options share the semantic, phonetic, or no radical with the target respectively. The fourth task was a semantic category judgment task, in which the participants chose from four pictures the one that represented the meaning or semantic category for 40 semantic radicals. And the last task was a pseudocharacter meaning judgment task, which contained 40 pseudocharacters. The participants had to select from four pictures the one that represented the meaning of the target characters. The results of the character decision task showed that most first graders could already judge frequent characters as real characters and noncharacters as illegal. But only fifth graders could judge pseudocharacters as character-like significantly above 23.

(35) chance level. Ho et al. (2003) thus suggested that the first graders already acquired some knowledge of character structure, and this knowledge grew with grade. Concerning the scores of all grades on other tasks, only first graders’ score on the semantic-relatedness judgment task was not significantly above chance level. The results indicated that first graders already had some knowledge of the position and semantic category of semantic radicals, but it was not until the third grade did they understand the function of the semantic radical. Therefore, positional knowledge seemed to be acquired earlier than functional knowledge. For the semantic category judgment and the pseudocharacter meaning judgment task, the participants of all grades performed better on lexical radicals than nonlexical ones, showing that lexical radicals provided better semantic cues. As for the Chinese sentence comprehension test, only fifth graders performed significantly better on semantically transparent characters than on opaque ones, which indicated that fifth graders could utilize the meaning cues of semantic radicals when processing a sentence. Furthermore, character decision, semantic category judgment and pseudocharacter meaning judgment were found to correlate significantly with Chinese word reading and sentence comprehension while radical position judgment correlated significantly with Chinese word reading only. Ho et al. (2003) thus proposed that character structure, radical position and semantic category played an important role in reading Chinese. And the semantic radical knowledge influenced word reading more than sentence comprehension. In their second study, phonetic radical knowledge, which included the position, function and sound value of the phonetic radicals were investigated. The participants were 60 Cantonese-speaking children in Hong Kong, 20 each in Grades 1, 3 and 5. They also received Raven’s Standard Progressive Matrices and Chinese Word Reading Test. Five phonetic radical tasks were also administered. The character 24.

(36) decision task was similar to the one used in Study 1 and contained 24 familiar characters, 24 rare characters, 24 pseudocharacters and 24 noncharacters. The second task, radical position judgment task, was also similar to Study 1 and contained 20 phonetic radicals. The third task was a phonological-relatedness judgment task with 20 pseudocharacters as targets. The participants had to choose among three alternatives one character that had the same pronunciation with the target. The three alternatives share the phonetic, semantic, or no radical with the target respectively. The fourth task was a phonetic radical naming task, in which the participants named 22 phonetic radicals. And the last task was a pseudocharacter naming task, in which the participants guessed the pronunciation of 25 pseudocharacters. Scores. of. the. character. decision,. radical. position. judgment,. and. phonological-relatedness judgment task were significantly different from chance. Ho et al. (2003) found that even first graders acquired some knowledge of the position, function and sound value of phonetic radicals, and the knowledge advanced with grade. Concerning the phonetic radical naming task, lexical phonetic radicals were named significantly better than nonlexical ones. Moreover, character decision, phonological-relatedness judgment, phonetic radical naming (lexical items) and pseudocharacter naming correlated significantly with Chinese word reading. Therefore, they argued that character structure, function and sound value of phonetic radicals had an influence on reading familiar and novel Chinese characters. Ho et al. (2003) concluded that radical knowledge was important in Chinese reading development. Knowledge of the position and semantic category of the semantic radicals and knowledge of the function and sound value of the phonetic radicals correlated significantly with Chinese word reading. On the other hand, knowledge of semantic category had a significant correlation with sentence comprehension. Developmentally, Chinese children progressed from a visual and 25.

(37) arbitrary way of reading characters to an analytic and rule-based manner. They first acquired knowledge of character structure and radical positions, and then developed knowledge of radical function and regularities, which helped them understand and remember characters more efficiently. 2.3.4 Summary Table 2-1 summarizes the major findings and limitations of the empirical studies reviewed in Section 2.3. Table 2-1 Major Findings and Limitations of Previous Studies Major findings Limitations Shu and Anderson (1997). 1. Radical knowledge: Children knew the 1. Participants: Three age function of semantic radicals, and the groups (Grades 1, 3 radical analysis aided in character and 5) learning. 2. Task: Only one (a 2. Development: First graders did not comprehension task) show clear semantic radical awareness, but third and fifth graders did. 3. Reading: Children with higher reading ability had better semantic radical awareness.. 1. Radical knowledge: Children knew the 1. Participants: Three age Shu, function of phonetic radicals, and the groups (Grades 2, 4, Anderson, radical analysis aided in character and 6) and Wu learning. 2. Task: Only one (a (2000) 2. Development: The phonetic radical production task) awareness was already evident in second graders, and continued to develop in fourth and sixth graders. 3. Reading: Children with higher reading ability had better phonetic radical awareness. Ho, Ng, and Ng (2003). 1. Radical knowledge: Children knew the Participants: Three age position, function and meaning/ groups (Mean ages=7;2, pronunciation of radicals. 9;2, 11 in Study 1 and 7;2, 9, 11;2 in Study 2) 26.

(38) 2. Development: First graders acquired some knowledge of the position and semantic category of semantic radicals, and the position, function and sound value of phonetic radicals. But only third graders started to exhibit knowledge of the function of semantic radicals. The radical knowledge continued to advance with grade. 3. Reading: Knowledge of the position and semantic category of the semantic radicals and knowledge of the function and sound value of the phonetic radicals correlated significantly with Chinese word reading. On the other hand, knowledge of semantic category had a significant correlation with sentence comprehension.. In sum, previous researchers have investigated children’s radical knowledge from different perspectives yielding different results. Shu and Anderson (1997) focused on the function of semantic radicals and how the radical analysis helped character learning. Shu et al. (2000) examined the function of phonetic radicals and the facilitation of character learning by radical analysis. And Ho et al. (2003) included the position, function and specific meaning or pronunciation of the radicals in their study. Concerning the development of radical knowledge, all these researchers described a progression with grade, but they did not agree on the onset age of different aspects of the radical knowledge. Shu and Anderson (1997) found that third graders started to exhibit semantic radical awareness while Shu et al. (2000) reported phonetic radical awareness in second graders. Ho et al. (2003) suggested that first 27.

(39) graders already showed certain knowledge of semantic and phonetic radicals. Furthermore, these researchers have sought to establish the relationship between radical knowledge and reading development in diverse ways. Shu and Anderson (1997) reported a correlation between semantic radical awareness and reading ability while Shu et al. (2000) found phonetic radical awareness to correlate with reading ability. However, Ho et al. (2003) documented a correlation between certain knowledge of semantic and phonetic radical and word reading as well as a correlation between certain knowledge of semantic radical and sentence comprehension. These previous studies also have some limitations. First of all, they recruited only selected age groups. While Shu and Anderson (1997) found semantic radical awareness in third graders and Ho et al. (2003) found certain semantic radical knowledge in first graders, second graders were not recruited and examined. Secondly, the task employed was not without flaw. In the comprehension task of Shu and Anderson (1997), the test items were isolated two-character words. Without further contextual cues, the answers could be diverse, but all plausible, especially given the creative mind of children. 2.4 A New Classification of Chinese Characters In this new classification of the Chinese characters, familiarity, semantic radicals and phonetic radicals will be the criteria. Each of the criteria has been used in the previous studies (Ho et al., 2003; Shu & Anderson, 1997; Shu et al., 2000), but has not been considered all together. Presently, the familiarity of characters to children will be the first criterion. Both Shu and Anderson (1997) and Shu et al. (2000) judged familiarity according to the sequence of introduction of the characters in textbooks. Familiar characters were present in the textbook one or two grade below the current grade of the students, and unfamiliar characters were in the textbook one or two grade 28.

(40) above. But in the present study, familiarity will be determined by a more objective measure. The familiarity of the characters is based on Grade 1-9 Curriculum Guidelines laid out by the Ministry of Education (MOE) of Taiwan in 2011. According to the guidelines, the first and second graders need to recognize 700 to 800 frequently used characters, the third to fourth graders 1,500 to 1,800 frequently used characters, the fifth to sixth graders 2,200 to 2,700 frequently used characters, and the seventh to ninth graders 3,500 to 4,500 frequently used characters. All characters can be found in Guóyǔcídiǎn jiǎnbiānběn biānjízīliào zìcípín tǒngjìbàogào [Report on the Frequency of Characters and Words of the Editing Material for the Concise Edition of Mandarin Dictionary] published by the MOE in 1997. In this report, 5,731 characters are included. And the current study will select from these characters and categorize them into Type 1 (F), Type 2 (F-1) and Type 3 (F-2) for each grade according to their frequency and appropriateness to the grade level. The three types represent three familiarity levels, which are F as familiar, F-1 as one level below the familiar level, and F-2 as two levels below the familiar level. Therefore, in the spectrum from familiar to unfamiliar, both F-1 and F-2 are unfamiliar levels, only the latter is more unfamiliar than the former. Take the second grade as an example, their F level characters come from the first 800 frequent characters; F-1 level characters come from the 801st to 1,800th characters; and F-2 level characters come from the 1,801st to 2,700th characters out of the 5,731 characters, respectively. Type 1: F Level In Type 1, the characters are further categorized into two subtypes according to their semantic radicals. Type 1-1 refers to transparent characters, whose semantic radicals are helpful in figuring out the meaning of the whole characters. And Type 1-2 includes opaque characters, whose semantic radicals contribute little or no cue to the 29.

(41) meaning of the whole characters (d’Arcais, 1992; Ho et al., 2003; Shu & Anderson, 1997). Type 1-1: Transparent For Type 1-1 characters, two subtypes are distinguished according to their phonetic radicals. Type 1-1-1 includes regular characters, whose phonetic radicals have identical pronunciation as the whole characters. And Type 1-1-2 involves irregular characters, whose phonetic radicals have different pronunciation from the whole characters (Shu et al., 2000). Type 1-1-1: Regular An example of this type is “清” (qīng, ‘clear’) as in (1), which has the semantic radical “水” (shuǐ, ‘water’) related in meaning to the whole character. Besides, its phonetic radical “青” (qīng, ‘green’) is pronounced the same as the whole character. Therefore, “清” is transparent with a semantic radical related in meaning and regular with a phonetic radical having the same pronunciation as the whole character.. (1). Type 1-1-2: Irregular An example of this type is “跑” (pǎo, ‘to run’) as in (2). The character is transparent since its semantic radical “足” (zú, ‘foot’) is related in meaning to the character. The character is also irregular since its phonetic radical “包” (bāo, ‘to wrap’) has a different pronunciation compared to that of the character itself. 30.

(42) (2). Type 1-2: Opaque In this type, the characters are also divided into regular (Type 1-2-1) and irregular (Type 1-2-2) characters according to their phonetic radicals (Shu et al., 2000). Type 1-2-1: Regular An example of this type is “停” (tíng, ‘to stop’) as in (3), which is opaque since its semantic radical “人” (rén, ‘people’) is unrelated in meaning to the whole character. “停” is also regular because its phonetic radical “亭” (tíng, ‘pavilion’) has the same pronunciation as the whole character.. (3). Type 1-2-2: Irregular “始” (shǐ, ‘beginning’) as in (4) is an example to this type since its semantic radical “女” (nǚ, ‘female’) is unrelated in meaning to the whole character, and it makes the character opaque. “始” is also irregular since its phonetic radical “台” (tái, ‘platform’) has a different pronunciation compared to the character itself. 31.

(43) (4). Type 2: F-1 Level In Type 2, the characters are also divided into two subtypes, transparent (Type 2-1) and opaque (Type 2-2), based on their semantic radicals (d’Arcais, 1992; Ho et al., 2003; Shu & Anderson, 1997). Type 2-1: Transparent Type 2-1 characters can be further distinguished by their phonetic radicals, and form two subtypes: regular (Type 2-1-1) and irregular (Type 2-1-2) (Shu et al., 2000). Type 2-1-1: Regular An example of this type is “泥” (ní, ‘mud’) as in (5), which has the semantic radical “水” (shuǐ, ‘water’) related in meaning to the whole character. Besides, its phonetic radical “尼” (ní, ‘nun’) is pronounced the same as the whole character. Therefore, “泥” is transparent with a semantic radical related in meaning and it is regular with a phonetic radical having the same pronunciation.. (5). 32.

(44) Type 2-1-2: Irregular “泡” (pào, ‘bubble’) as in (6) is an example of this type because its semantic radical “水” (shuǐ, ‘water’) is related in meaning to the whole character and makes it transparent. “泡” is also irregular since its phonetic radical “包” (bāo, ‘to wrap’) differs in its pronunciation from the whole character.. (6). Type 2-2: Opaque Characters in Type 2-2 can also be further categorized into regular (Type 2-2-1) and irregular (Type 2-2-2) according to their phonetic radicals (Shu et al., 2000). Type 2-2-1: Regular An example of this type is “伸” (shēn, ‘to stretch’) which is opaque because its semantic radical “人” (rén, ‘people’) provides little cue to the meaning of the whole character. “伸” is also regular because its phonetic radical “申” (shēn, ‘to state’) is pronounced the same as the whole character.. (7). 33.