languages. Since the Japanese writing style is further from English than. Korean, it is shown, ..... extensive reading does not form part of the approved syllabus in. Japan. ... 148 students; SHS2, 116 students; SHS 3, 70 students. The students in ...
CHAPTER THREE THE DEVELOPMENT OF WORD READING SKILL IN SECONDARY SCHOOLS IN EAST ASIA DAVID COULSON UNIVERSITY OF NIIGATA PREFECTURE, JAPAN
Abstract The issue of word recognition efficiency in secondary school students in Japan and Korea is considered. The relative distance between language writing systems can exert a strong influence on reading in second languages. Since the Japanese writing style is further from English than Korean, it is shown, through the use of a quick-and-easy test of word recognition, that Koreans have greater ease in learning to read English high-frequency words and score significantly higher on this measure by the end of secondary schooling. Low scores on this measure can be considered as very likely linked to weaker comprehension ability in English. In addition to this, the paper suggests that the emphasis on word recognition is too weak in secondary education, especially in Japan. The implications of this situation for English education in East Asia are considered. Keywords: decoding, word-recognition, secondary-school, skill development.
orthography,
phonology,
Introduction Learning to read in second languages is an increasingly important objective for many people in this globalized world. It helps them to achieve their aims, be they professional, academic or recreational. Even for learners lacking direct interaction with speakers of their target language, skill in reading under timed conditions is often tested, often for high-stakes purposes such as entrance examinations to schools and
Development of Word Reading Skill in Secondary Schools in East Asia
43
universities or for promotion. In all cases, if reading is not well developed, this can have a serious impact on people’s futures. Efficient reading, offering exposure to printed material from the target language, is also the most convenient means by which learners can manage and review the input essential for acquiring their new language. Nevertheless, the reading skill does not emerge as a byproduct of increasing general competence (Koda, 2005). Therefore, in the early stages of English education, in addition to the standard focus on grammar and vocabulary, much attention should also be paid to the skills of decoding, using, for example, phonics for basic word reading. On standardized proficiency tests such as TOEFL and TOEIC, Japanese people tend to score more points on the listening than the reading section and to have lower overall scores than their counterparts in neighboring Korea. Furthermore, my experience shows that many Japanese students fail to complete the final reading section of TOEIC in the allotted time. Word recognition is both a crucial element in reading and an important predictor of future reading ability. According to Grabe, “fluent reading comprehension is not possible without rapid and automatic word recognition of a large vocabulary” (2009, p.23). Japanese test scores may be comparatively low due to the particular demands of English word recognition, which are considerably more different from Japanese than they are from Korean. The impact of this on reading performance will be discussed below in the section on cross-linguistic effects.
Literature Review The components of reading Reading can be separated into lower-level (or bottom-up) and topdown processes. The former refers to the combination of the processes of orthography, phonology and semantic activation. The latter refers to stages in the formation of comprehension, such as directing attention, setting strategies, inferencing, activating background knowledge and interacting with the text. Effective reading involves all of these processes operating smoothly at the same time. It is not the case that one moves from bottomup processing to become a uniquely top-down reader. Rather, decoding, in which connections between orthographic units and the sounds they represent are made, is fundamental. Failure to retrieve the phonological information is a major impediment to comprehension. This does not match the idea underlying the conception of reading as a predictive guessing game, as described by Goodman (1967), which long remained influential in language education. Much research has updated
44
Chapter Three
this view. For example, Rayner & Pollatsek (1989) demonstrated how our eyes are involved in very complicated movement over text, in which the majority of words are momentarily settled upon, at the rate of about 5 words a second. If reading were a primarily top-down activity, the ends of sentences should be read faster because more accurate predictions can be made from the earlier part of the sentence. However, Paran (1997) reports that this kind of acceleration isn’t seen and for this reason he suggests (p.29) that learners should rely on top-down processes less. This should be reflected in classroom practice, emphasizing early decoding skills. Since all component processes are at work simultaneously, reading is a very intensive activity, and this explains why we get tired or distracted by heavy texts, or when we are reading in a foreign language that we are not fluent in. Automatic functioning of the reading complex is necessary to counter lagging attention and to maintain a high level of comprehension. The recommended method agreed upon by experts (e.g. Nation, 1991, Grabe, 2009) is for extensive, and repetitive, amounts of text experience. This allows for the emergence of rapid reading. Grabe (p.23) recommends this should at the rate of 250 to 300 words a minute. However, my experience with Japanese university EFL learners suggests a speed of only around 180 words a minute. This is similar to the speed of children in grade 5 (11 year-olds) in America, according to Carver (1992). Such labored reading must have a negative effect on comprehension. Teachers may assume that, if members of a class are motivated, or willing to speak, the reading skill must have taken care of itself. This idea is explicitly ruled out by Koda (2005, p.10), who states that progress in second language reading fluency cannot be taken for granted. Grabe (2009, p.78) also makes a related point on the importance of maintaining skills. “In reading, word recognition, vocabulary, and fluent use of language, learning gains are fragile and need to be reinforced consistently over a period of time.” Overall, in foreign language learning the early development of bottomup skills is essential. The question in this chapter is how to assess these skills and factors that may influence them.
Orthographic and phonological influences Sight recognition of words is one solution to the challenge of reading English. Comprehension would suffer if readers confused similarly pronounced words such as ‘meat’ and ‘meet’. Similarly, orthographic knowledge allows us to recognize ‘peeple’ as a misspelling of ‘people’. Readers of languages with very opaque spelling such as Danish, and even
Development of Word Reading Skill in Secondary Schools in East Asia
45
more so Chinese, rely strongly on this very fast, efficient strategy. But, in the slightly less opaque case of English, there has been a vigorous debate about the pros and cons of whole-word reading and the use of phonics, in which the relationship between letters and sounds is systematically taught. Dehaene (2009, p.220) describes the very negative effect on reading attainment in California in the 1980s when the education policy deemphasized the teaching of decoding skills in schools. This shows that, in fact, the process of becoming a fluent reader involves a staged process of learning how to combine both orthographic and phonological information. Research on children learning their L1 is instructive. According to Ehri (e.g. 2005), there are four steps on the path towards automatic sight-word reading: 1) pre-alphabetic; 2) partial alphabetic; 3) full alphabetic with full awareness of the relationship between graphemes and phonemes; 4) consolidated alphabetic with connections made between syllabic and morpheme units e.g. ‘ing’ in ‘sing’ and ‘bring’ etc. Juel (1988) found that grade 4 children who were asked to read words with common letter patterns could do so faster than words with less common spelling patterns. Knowledge of legal letter patterns is clearly one of the key stages in orthographic skill. Conversely, Koda (2005) explains that the ability to fluently name non-words (e.g. ‘nerm’) indicates strong reading competence. Such words do not exist in memory so only phonological decoding can lead to correct pronunciation. It seems that although many Japanese university students have reached 4) on Ehri’s stages, they may still have insufficient skill at stage 3). The reasons for this will be described below. So, in addition to orthographical skill, activating phonological information in English reading is one of the vital contributory factors in decoding. Koda insists on the importance of phonology for reading: “Seamless performance is not attributable to whole-word retrievals, but rather to children’s accumulated knowledge of their writing system, sound-symbol relationships in particular.” (p.32). Further, Koda (2005, p.33) argues that short-term memory works by holding the phonological representation of words while the higher-level process of comprehension continues. Therefore, without phonology, understanding would be impossible.
The writing systems of English, Japanese and Korean Before comparing the word-recognition ability of Japanese high-school learners of English with those of their Korean counterparts, we should contrast the decoding requirements of these languages’ orthographies. The way each writing system represents words shapes a reader’s cognitive
46
Chapter Three
processes for reading and this has an effect on the ease of reading in foreign languages. Although Japanese and Korean have syntactic and lexical similarities, their writing systems are quite different. However, Korean is typologically closer to English. We will consider how this influences reading in English later, but first let’s examine the properties of each language’s writing system. i) English, like most European languages, uses the Roman alphabet, although all these languages differ in their degree of consistency between spelling and pronunciation, or more precisely graphemes and phonemes. While most English high-frequency words are consistent (e.g. ‘cat’, ‘mat’ etc.), with less common words inconsistency in pronunciation is commonplace (e.g. ‘bored’, ‘board’) and the same spelling can represent different sounds (e.g. the infinitive and past tense of ‘read’). This lack of transparency makes it hard for native learners and especially non-native English speakers who are used to reading in a regular orthography. Cook and Bassetti (2005, p.7) point out that the English letter ‘o’, for example, corresponds to at least 10 phonemes. One reason for this variation is that English has a much wider range of sounds to represent through its writing systems than other languages such as Spanish or Italian. On the other hand, Birch (2002) argues that English spelling does have patterns. A non-word such as ‘tun’ could be pronounced with the ‘u’ similar to the sound in ‘pup’, ‘put’, or ‘truth’. However, processing experience allows people to know that the most likely pronunciation is /tݞn/ since in consonant-vowel-consonant groups, the probability of this pronunciation is the highest, at 63%. Birch stresses the relation of this probabilistic system to reading skill: “Once the orthographic processor is trained to the point of automaticity to recognize these graphemes, it is not hard for it to associate the correct pronunciation with them because they are highly predictable.” (p.81) Nevertheless, many have argued for a thorough standardization of English spelling of the kind that started with the amended conventions in American-English spelling (e.g. ‘center’). Concerning this, Dehaene (2009, p.32) argues that full standardization is not feasible since this would require the invention of many new letters to represent diphthongs, for example. Dehaene further argues that present English spelling has the advantage that, once mastered, the spelling of words “points straight at meaning.” (p.34) ii) Next, Japanese writing uses a rather complicated system of logographic characters (‘kanji’) and two syllabaries (‘hiragana’ and
Development of Word Reading Skill in Secondary Schools in East Asia
47
‘katakana’). Kanji are used to represent words but they have virtually no phonetic value. This can be seen in how the same character can be read in completely different ways depending on context. The kanji for ‘mountain’ (ᒣ) is conventionally pronounced ‘yama’, but can also be read as ‘san’, as in ‘Fuji-san’, Japan’s famous volcano. By contrast, hiragana and katakana, which both total 46 phonetic characters, have a direct, almost completely transparent correspondence to the spoken sounds of Japanese. (They are used mainly for prepositions, grammatical elements and foreign words.) Hiragana can also be used when one does not know a kanji. One could write ‘yama’ (mountain) in hiragana as ࡸࡲ, where the single characterࡸ represents ‘ya’, and ࡲ represents “ma”. This can be pronounced only one way, unlike the kanjiᒣ. One may wonder why Japanese does not use hiragana for all words and save children the effort of memorizing thousands of kanji. One reason calls to mind the proliferation of many opaquely spelled words in English such as ‘yacht’; once learned they allow direct access from memory without intermediary phonological activation. In sum, Japanese has a curious blend of almost completely transparent syllable characters and almost completely opaque ‘kanji’ characters. This results in a system famed for how long it takes for Japanese people, and L2 learners, to master. iii) Korean uses a unique writing system called Hangul, invented several centuries ago. In this system, graphemes represent consonant and vowel phonemes (absent in Japanese) and they are packaged together into syllable blocks, albeit with various restrictions (Park, 2008). The correspondence between these graphemes and phonemes is very consistent, although the syllable blocks may have some degree of inconsistency with spoken syllables (p.202). Korean is, therefore, an alphabetic system. One distinguishing point from English, where the graphemes are analyzed linearly from left to right, is that in Korean words the Hangul letters are piled up in a more two-dimensional arrangement. Finally, logogram characters are not alien in Korean, but they are not used in ordinary situations. In this way, writing systems in different languages have very distinct systems for representing words. In response to this, we develop appropriate, specific cognitive processes to decode the writing system of our L1. A systematic way of describing these differences is explained next.
Chapter Three
48
The minimum size of sound represented in orthographies According to Wydell and Kondo (2003), writing systems are described with two dimensions – “transparency” and “granularity”. We have already reviewed the transparency dimension above. Granularity refers to the unit (or “grain”) size at which sound is represented in a language. The diagram showing the different grain size of various languages is reproduced from their paper below (Figure 1). In the bottom left corner, Italian is shown as a phonemically transparent language, whereas English, and more so Danish, have increasingly lower transparency. Half-way up the grain scale, Japanese Kana (i.e. hiragana and katakana) appear as transparent but rather coarse, and in the top right, Japanese Kanji are both coarse and opaque. In this diagram, Korean would appear somewhere between Italian and German, according to Perfetti and Dunlap (2008). 㻳㼞㼍㼚㼡㼘㼍㼞㻌㻿㼕㼦㼑
㻯㼛㼍㼞㼟㼑 㼃㼛㼞㼐
㻿㼥㼘㼘㼍㼎㼘㼑
㻷㼍㼚㼖㼕 㻯㼔㼍㼞㼍㼏㼠㼑㼞㻛㼃㼛㼞㼐 㻌㻌㻷㼍㼚㼍 㻌㻌㻯㼔㼍㼞㼍㼏㼠㼑㼞 㻌
㻼㼔㼛㼚㼑㼙㼑
㻌㻌㻵㼠㼍㼘㼕㼍㼚㻌㻳㼑㼞㼙㼍㼚㻌㻌㻌㻱㼚㼓㼘㼕㼟㼔㻌㻌㻌㻰㼍㼚㼕㼟㼔
㻲㼕㼚㼑 㻌㻌㼀㼞㼍㼚㼟㼜㼍㼞㼑㼚㼠
㻌㻌㻌㻌㻌㻌㻌㻌㻌㻻㼜㼍㼝㼡㼑
Figure 1: Hypothesis of granularity and transparency and orthography-tophonology correspondence
According to Wydell and Kondo (2003), granularity has significant implications for decoding. Reading disabilities, such as dyslexia, more readily appear in speakers of languages where phonological information is not always directly recoverable from the script. These are languages, such as English and others, which appear towards the right side of the diagram. On the other hand, although Japanese is opaque, the writing system has a rather large grain size (kana and kanji), so reading disabilities are not prevalent. The rate of dyslexia in Japan is reported by Wydell (2000) as being very low, at about 0.1% of the population, whereas in Britain it
Development of Word Reading Skill in Secondary Schools in East Asia
49
affects between ten and fifteen percent of the population, according to the British Dyslexia Association.
Cross-linguistic effects in foreign-language reading As described above, reading in our native language forms specific processing styles which become strongly automatized through repeated experience. Much research using judgment tasks has demonstrated how these processing styles have strong and persistent effects on reading efficiency in the L2. Ryan & Meara (1991) reported that Arabic spelling does not normally represent vowels. Therefore, Arab readers have less sensitivity to vowels in English spelling. When presented in a test with a correctly spelled English word followed after a short interval by the same word, or the same word with a missing vowel, they took much longer to decide, and scored less highly, than another control group. This can lead to reading miscues over words with a similar consonant structure. For example, ‘break’ and ‘broke’ have the same syllable structure (brk), which is most salient for Arabic L1 readers, and this can lead to confusion. Much research in this area with Korean and Japanese L1 readers has been done. Koda (1999) compared Koreans and Chinese on their ability to recognize legal and illegal strings in English. She found that the Koreans could reject low-frequency illegal letter strings more easily than the Chinese. This showed the Koreans had more quickly mastered probable letter combinations, and this was due to their L1 alphabetic experience. Similarly, Wang et al. (2003) compared Chinese and Korean L1 learners of English and found a similar processing advantage in phonology for Korean L1 subjects, whereas the Chinese attended more to orthography. However, the situation is not always so straightforward. Wade-Wolley (1999) compared intermediate proficiency-matched Japanese and Russian English learners. The Japanese were faster in some word judgment tasks than the Russians. Wade-Woolley argued that Russians use the reading style of their L1 and that, although Russian has an alphabet, the demands of reading in English are different. The Japanese, having no L1 phonemic experience, simply resorted to orthographical skills to do the task, which proved faster in this case. This demonstrates that the linguistic effect from L1 is indeed significant, although some L1 effects may prove advantageous, such as the Japanese ability to read words by analyzing their visual features. Nevertheless, the importance of phonological decoding is emphasized by reading experts. As with Koda’s claim for the importance of reading nonwords, Stanovich (1982) states that one of the tasks that most clearly
50
Chapter Three
differentiates good and poor readers is the speed and accuracy with which they name pseudo-words. As we have seen, reading skill development cannot be assumed from general proficiency. Fluent bottom-up decoding is essential for comprehension. Weaker readers over-rely on contextual support to compensate for this (Grabe, 2009, p.73), but this is not an ideal strategy and certainly not one employed by very fluent L2 readers or accomplished L1 readers.
Approaches to assessing word recognition skill Based on the evidence we have seen, it is clear that some kind of practical test of word-decoding skill would help teachers to assess their learners’ progress. The research designs of the papers reviewed above are sophisticated and their testing methods inappropriate for busy teachers. A quick-and-easy test would be very helpful for teachers to address secondlanguage reading deficiencies. Several approaches have been tried in L1 reading research. These often depend on altering the conditions under which words are usually read. One way of doing this is to hide words among strings. Adams (1990, p.127) used this approach with items such as and . Here the surrounding letters are all consonants and so they contain no syllables. In this kind of test, capable English readers quickly recognize the presence of the target word ‘back’, thanks to their ability to form syllables around the vowel and also by relying on their knowledge of likely letter combinations. This recalls Birch’s explanation earlier. Further, words are easier to process than when they are presented in syllabic chunks. Therefore, is read faster than . In this way, children gain the ability to know where syllables start and end. Adams (1990, p.127) also used a test style very similar to the one used in this paper. With young English L1 readers as her subjects, she showed strings of words that form a meaningful idea such as and meaningless strings of words such as . She found that the performance on meaningless strings separated stronger and weaker readers better than the meaningful strings. These results show just how vital the skill of word recognition is for successful reading. Weaker readers took several years to match the performance of initially stronger readers on this test, underlining the importance of a solid start in bottomup decoding skills. This applies as much to L2 English learners as it does to L1 learners, especially in light of the fact that one’s L2 can have such a strong impact on foreign language reading styles.
Development of Word Reading Skill in Secondary Schools in East Asia
51
The Word Recognition Index: “WRI” Following the style of Adams’ investigations using word chains, a similar format devised by Jacobson (1995), which produces an index of word recognition skill, was selected as an appropriate testing tool. WRI has been used to assess decoding skill in people with dyslexia (Jacobson, 1995) and the relative processing skills of L1 English and Swedish children (e.g. Guron & Lundberg, 2004). For the investigation reported below, it is used to assess the decoding development of Japanese and Korean learners of English. The test instrument contains two sections called Word Chains and Letter Chains. The relative performance on these tests is used to calculate an index of word recognition (WRI), with a low score indicating problematic reading. To calculate WRI, subjects take the Word Chain test first. Each item is a nonsense string of three high-frequency words, which have to be separated with pencil strokes. The item boygomeet would be correctly answered like this: boy/go/meet. The words are selected from the highest-frequency vocabulary of English. The Word Chain test contains 120 items and the time limit is three minutes, so a very large sample (up to 360 words) can be quickly checked. This is far more comprehensive than is usual in psycholinguistic-oriented reading or vocabulary research. Next, subjects take the Letter Chain test. They have to divide a string of capitalized letters, where the same letter appears twice. For example, OUCCNEMHHE would be correctly answered as OUC/CNEMH/HE. The Letter Chain test contains 80 items and the time limit is 90 seconds. A low score on the Word Chain test and a normal score on the Letter Chain test indicates word recognition problems. Scores are calculated by the formula ((WCh-LCh)/LCh). A score of 90 word chains and 60 letter chains would give a WRI of 50; i.e. ((90㸫60)/60) × 100=50. Individuals with the same score on both tests get a score of 0. This would indicate that they can process word strings no better than random letters and that therefore they have severe reading disability. Negative scores are also possible. Jacobson first tested 150 school children ranging in age from 8-16 (grades 1-9). Data on adults were also included. The results showed that WRI values rose rapidly from grade 1 to 5 and then were rather stable. In grade 2, the WRI was near zero, whereas, in normal adult groups, WRI was about 90, indicating that word recognition was largely automatic. His second experiment was on American families who self-reported as having reading difficulties. A normal group was also tested as a control. The least affected had a mean WRI value of 66 and the most affected a mean value of 4. The control group obtained a mean WRI value of 83.
52
Chapter Three
For this investigation, the WRI test would be used to assess primarily the development of word recognition skill among Japanese students in the six years of secondary school education. The same test would be used for all grades. Students across three years of senior high schools in South Korea would also take the test. The predictions of this research were as follows: a) In Japan, the rate of increase on WRI would be slow but stable. Stability was predicted since the school in question is an academic feeder school for university-bound students, who must pass entrance examinations in English. Slowness was predicted since extensive reading does not form part of the approved syllabus in Japan. Consequently, the mean WRI might not reach the level of even mildly challenged English L1 readers (around WRI 66, according to Jacobson). Nevertheless, the initial rate of increase should be similar to the rate of increase seen in elementary (primary) school level pupils, outlined above. b) The Korean students should attain a higher level of WRI than the Japanese at the end of secondary education because Korean has an orthography closer to that of English than Japanese, and this should facilitate, through cross-linguistic effects, the development of more fluent word decoding. English education is rigorous in both countries, at least in high-achieving schools, but the advantage for the Koreans should be clear, at least by the age of graduation (around the age of 18).
Method Subjects The students in a single junior/senior high school in a provincial city in Japan were the subjects of the test. Their ages ranged from around 12 to 18. There were six separate groups. First, the Junior High School (JHS) consisted of years 1 to 3: JHS1, 83 students; JHS2, 80 students; JHS3, 74 students. The Senior High School (SHS) consisted of years 1 to 3: SHS1, 148 students; SHS2, 116 students; SHS 3, 70 students. The students in the three years of the Junior High School were not streamed by proficiency. The higher number of students in the Senior High years was partly due to the intake of additional students by competitive intake at the age of 15. In each year of the Senior High School, students were divided into three levels of proficiency (mixed, high, and very high) for English classes by
Development of Word Reading Skill in Secondary Schools in East Asia
53
an unspecified placement test. In addition to these subjects, the test was given to two classes in the Japanese humanities faculty of a provincial university. Subsequently, data were collected from two high schools in Seoul, South Korea. The first was a girls’ high school, and reportedly very competitive to enter. Thirty-two second-year students took the test in Seoul. Subsequently, data were collected from each grade of another good, but reportedly less high-ranking, high school. The number of students was 59 in year 1 (15 to 16 years of age) 60 in year 2 and 58 in year 3. Although there were no comparative data on students’ proficiency level from either country, from explanations from teachers at both schools it seemed likely that the high school in Japan and the second high school in Korea were roughly similar: competitive to enter, located in urban areas but below the level of the most prestigious local schools. Finally, eleven native speakers also took the test to serve as a baseline. They all took the test twice with an intervening period of two weeks to investigate the re-test validity of the assessment test instrument.
Materials A new version of the WRI test was made and 360 high-frequency words were selected from the JACET 8000 word list. (Aizawa, Ishikawa & Murata, 2005) All test items were prepared to be divisible into three separate words with only two pencil strokes, leaving no extra letters. All words were between 2 and 6 letters long and came from the top 1000 frequency band. The mean rank order of the words was 335. One hundred and twenty word chain items were printed in three columns on a single sheet of A4 paper. The font was 10.5 Times New Roman. A serif typeface was chosen as this is typical of fonts used in approved textbooks. On the next page, 80 letter item chains were printed in three columns. The wordchain items were carefully checked so that there was only one way of dividing the chain into three words using two pencil strokes. The first two items from the wordchain quiz are shown below. In the case of 2) manyfighthouse, the word ‘fig’ is contained within ‘fight’ but segmenting at this point leaves a nonsense word ‘thouse’. 1) aboveundermake 2) manyfighthouse The wordchain test is marked out of 120. In each item, any mistake results in a score of zero, even if another word has been segmented correctly.
54
Chapter Three
The test sheets were distributed to students by the teacher in charge at the beginning of a class. The sheets had a top page attached with an instruction printed clearly not to turn over the page. An explanation of the procedure in either Japanese or Korean was written below. Five examples each of wordchains and letterchains were shown and students were allowed to try them. After the brief explanation, the students were given three minutes to finish the wordchains, and, immediately after, 90 seconds to complete the letterchains The papers were collected and returned to the author for checking. Each paper took about 1 minute to check. Once all the data were collated, the word recognition index for each individual was calculated using the formula 100 × (wordchains㸫letterchains) / letterchains), as described above. The number of mistakes was also tallied. Finally the school year means were calculated.
Results In Figure 2, the results for wordchains are shown in striped bars, letterchains in block colour, and the Word Recognition Index in dotted bars. The results for the native speakers are shown first on the left. Their Word Recognition Index (WRI) was 53. This is somewhat lower than the value of 90 reported by Jacobson for normally-reading English L1 adults. Their wordchain score was 89.6 (sd 11.4) and letterchain score 59.5 (sd 8.6). The range of values for WRI went from 14 to 113. The test-retest correlation was strong, r=0.73.
Figure 2: The results of all test takers on the Word Recognition Index (“WRI”) test
Looking at the Junior High School results, the number of wordchains the first years (JHS1) completed is a little more half the number of
Development of Word Reading Skill in Secondary Schools in East Asia
55
letterchains. This produces a strongly negative WRI value of 㸫45.3. By JHS3, this had improved to 㸫7.4. At this stage, the number of wordchains is almost the same as letterchains. It is notable that, by this age, the children’s letter decoding skill is fully automatized since their letterchain score is almost the same as that of the native speaker group. After the beginning of Senior High School, the number of wordchains starts falling, and this continues until the end of high school. This depresses WRI, which settles into a negative, persistent value of around 㸫15. The year-on-year decreases in wordchains do not reach significance in an analysis of variance, but the fall in WRI from JHS1 to SHS3 is significant, t=1.15, p
