原著 effects of phonological analysis, rapid naming, and
TRANSCRIPT
9KOBAYASHI et al.: Factors Affecting Reading in Japanese
コミュニケ ーション 障 害 学24, 9-21 (2007)
■ 原 著
Effects of Phonological Analysis, Rapid Naming, and Visual Search
Performance on 5TH Grade Reading in Japanese*1
Maya Shiho KOBAYASHI*2 Charles W. HAYNES*3
Pamela E. HooK*3 Paul MACARUSO*4
ABSTRACT: This study examined the relative validity of phonological analysis (mora deletion, sound
matching), rapid automatized naming (RAN), and visual search (cross out speed and accuracy) abilities for
predicting fluent text reading and reading comprehension within typically developing Japanese 5th graders (39in all). Analogous abilities have been identified as important predictors of reading skills in alphabetic
languages like English. Given the differences between English and Japanese orthography, it was hypothesized
that unique patterns of relationships would be obtained for children reading Japanese. The results
demonstrate that mora deletion contributes significantly to speed of oral reading. When mora deletion was
excluded from the regression model, RAN (number, hiragana, kanji) also explained variance in oral reading
speed. Furthermore, visual search speed predicts accuracy of oral reading. When there is no influence of
visual search speed in the regression model, sound matching and kanji RAN are also significantly associated
with oral reading accuracy. General knowledge influenced reading comprehension. Based on our previous
studies on reading of Japanese kindergartners and first graders, the findings suggest that the predictive power
of these measures varies depending on the nature of the reading task and stage of development.
Key Words: Japanese reading fluency, phonological analysis, Rapid Automatized Naming (RAN),
visual search
1. INTRODUCTION
There has been increased interest in reading fluency in
recent years because of a growing recognition of its
importance in reading comprehension. Fluency requires not
only the ability to recognize words automatically, but also to
apply appropriate rhythm, intonation, and phrasing to the
text so that attention can be allocated to comprehension
(Meyer, Felton, 1999).
According to the results of the Programme for
International Student Assessment (PISA), which is an
internationally standardized assessment administered to 15-
year-olds school children in approximately 30 to 40
participating countries, Japan's performance in readingdecreased significantly from 2000 to 2003 (DECD, 2000,
2003). There is no clear explanation for this result; however,
considering the fact that Japan maintained high scores in
mathematics and science in the same years, it appears that
the number of Japanese adolescents who struggle with
reading is increasing.
The current study examined relationships between
reading skills and selected cognitive-linguistic predictors
including phonological analysis (mora deletion, sound
matching), rapid automatized naming (RAN), and visual
search (cross out speed and accuracy) in Japanese fifth
graders. Teachers and researchers are concerned with theearly detection of children who are at risk for reading
disabilities and also looking for the early prediction of
children's later reading achievement; therefore, establishing
reliable and valid assessment tools is of critical importance.
*1 本 稿 の 一 部 は 日本LD学 会 第15回 大 会 (2006年10月7日,
札 幌), The American Speech-Language-Hearing
Association Conference (2006年11月16日, マ イ ア ミ,
FL, U. S. A.) で 発 表 し た. 2006年11月25日 受 理.
*2 上 智 大 学, 日 本 学 術 振 興 会 特 別 研 究 員 (〒102 -8554東 京
都 千 代 田 区 紀 尾 井 町7-1). Sophia University, Research
Fellow of the Japan Society for the Promotion of
Science (JSPS) for Young Scientists (7-1 Kioicho,
Chiyoda, 102-8554, Japan)
*3 MGH Institute of Health Professions, Boston,
Massachusetts
*4 Community College of Rhode Island and Haskins
Laboratories, Lincoln, Rhode Island and New Haven,
Connecticut
10 コミュニケーション障 害 学 Vol. 24 No. 1 2007. 4.
A large body of research has demonstrated that
phonological analysis, RAN, and visual search abilities areimportant for the acquisition of word recognition, fluency,
and reading comprehension in alphabetic languages such as
English (Badi 1994, 1995, 1998; Cornwall, 1992;
Cunningham, Perry, Stanovich, et al., 2002; Meyer, Wood,
Hart, et al., 1998; Scarborough, 1998). English is an
alphabetic language based on grapheme-phoneme
relationships, whereas Japanese orthographies are
comprised of two syllabaries-hiragana and katakana, and a
system of morphosyllabic symbols ("kanji"). In our
previous studies, in Japanese kindergartners, hiragana RANand number RAN best predicted reading accuracy and
speed. In our Japanese first graders, however, kanji RAN
and hiragana RAN were the best predictors of oral reading
speed, mora deletion best predicted oral reading accuracy,
and kanji RAN, mora deletion, and phonological memory
best predicted reading comprehension (Kobayashi, Haynes,
Macaruso, et al., 2005; Kobayashi, Kato, Haynes, et al., 2003).
Additionally, deficits in phonological awareness, speeded
naming, and visual/orthographic abilities successfully
differentiated Chinese dyslexic readers from typically
achieving readers (Ho, Chan, Tsang, et al., 2002). We hope
that our current study will add to our general understanding
of the cognitive-linguistic skills that underlie the ability to
read Japanese and also provide additional support for
developing tasks that facilitate early detection of childrenwho may struggle with learning to read.
1.1 PHONOLOGICAL ANALYSIS
Phonological analysis, or phonological awareness, refers
to the ability to detect and manipulate explicitly and
segmentally the sounds in one's oral language. This ability
is found to be critical for the efficient decoding of printed
words and the ability to develop correspondences between
sounds and letters when spelling or reading English
(Badian, 1995; Cornwall, 1992). While phonological analysisis a strong predictor of early reading ability, it has been
suggested that the relationship is reciprocal: the ability to
read also facilitates phonological analysis (e.g., Bowey,
Francis, 1991). Phonological analysis skills can be assessed
by comparing and contrasting the sounds of words for
rhyme and alliteration, blending and segmenting into
syllables, and manipulating sounds at the phoneme level
(Adams, 1990). Previous studies have found thatkindergarten tasks involving manipulation of phonemes
were some of the strongest predictors of later reading
development (Wagner, Torgesen, 1987). In Japan, past
studies found a strong relationship between phonological
analysis and acquisition of reading skills (Amano, 1970,
1988, 1989; Hara, 1998; Oishi, 1998). Mann (1986) reported
that Japanese first graders were found to be aware ofsyllables and less aware of phonemes in comparison to mostAmerican first graders who were found to be aware of bothsyllables and phonemes. Moreover, Japanese children'ssyllable (mora) and phoneme counting ability significantly
predicted speed and errors in oral reading of hiraganawords. Endo (1991) reported that six-year-old Japanesekindergarten children develop phoneme awareness andtheir phoneme awareness performance contributessignificantly to reading ability. In our more recent studies, amora deletion task, which required a child to hear a real
word, repeat the word, delete one specified mora, and saythe reconstructed word, did not predict kindergartners'
reading speed and accuracy (Kobayashi et al., 2003;Kobayashi et al., 2005). However, in first graders, moradeletion predicted reading accuracy and readingcomprehension (reading comprehension was notadministered for kindergartners in our previous study).These results indicate that the first grade text which wasrelatively complex and consisted of hiragana, katakana, and
kanji characters required higher phonological analysis skillsthan the kindergarten text which was relatively simple andmainly consisted of hiragana characters. We expect that themore complex text for fifth graders will require higher
phonological analysis skills. This study examines thecontributions to reading performance of two phonologicalanalysis tasks 1) mora deletion from our previous study, and2) sound matching which requires phoneme recognition.1.2 RAPID AUTOMATIZED NAMING (RAN)
Rapid automatized naming (RAN), as well as phonologicalanalysis, has been extensively proven to be related toreading performance in English (Badian, 2005; Manis,
Seidenberg, Doi, 1999). RAN tasks require one to name aseries of five to six randomly repeated stimuli (colors,number, letter, common objects) as quickly as possible,
(Denckla, 1972; Denckla, Rudel, 1974). RAN and reading tapsimilar underlying processes; they require recognition and
identification of symbols, retrieval of a pronunciation foreach symbol and articulation of it, and then careful rapidscanning to the next symbol (Miller, C., Miller, S., Bloom, etal., 2006; Stringer, Toplak, Stanovich, 2004; Wolf, Bowers,1999). Functional magnetic resonance imaging studies havealso indicated that RAN tasks require the same neural
network as is involved in reading (Misra, Katzir, Wolf, et al.,2004). However, reasons why the association betweeninefficient RAN performance and weak reading abilities
occur are not fully understood. There have been numeroushypotheses regarding this association. RAN may require
automaticity of underlying component skills, may represent
general difficulty general difficulty in speed of processing, and/or could
11KOBAYASHI et al.: Factors Affecting Reading in Japaanese
reflectpoorcoordinationofthevariousunderlying
processes (Korhonen, 1995; Snyder, Downey, 1997; Wolf,
Bowers, 1999, 2000).
RAN's predictive power has been replicated in many
studies (Denckla, Cutting, 1999; Schatschneider, Carlson,
Francis, et al., 2002; Wolf, Bowers, 1999, Wolf, O'Rourke,
Gidney, et al., 2002). However, not all RAN tasks are equally
powerful in prediction of reading. Letter and number RAN
are robust predictors of reading ability, while object and
color RAN are less predictive (Denckla, Cutting, 1999). In
addition, letter and digit forms of the RAN are strong
predictors of single word reading, but not of reading
comprehension (Meyeretal., 1998). It has also been
suggested that when typically developing children's reading
fluency is high suchas in grade 3 or4, RAN may contribute
only minimally to their reading performance (McBride-
Chang, Manis, 1996).
In a study of Japanese 6-year-olds, the speed of rapid
alternating stimulus naming for objects and numbers
significantly correlated with identification of 6 hiragana
characters (Kaneko, Uno, Haruhara, 2004). In our previous
studies, we constructed four RAN tasks including objects
(chair, fish, pencil, star, key, ship), numbers (2, 3, 4, 5, 7, 8),
hiragana (た, に, を, い, の, と), andkanji (金, 人, 上, 手,
月, 山). The results indicated that hiragana and number
RAN predicted oral reading speed and accuracy in
kindergarten, and hiragana RAN and kanji RAN predicted
reading speed in first graders (Kobayashi et al., 2005;
Kobayashi et al., 2003). Moreover, kanji RAN significantly
contributed variance in reading comprehension in first
graders. In the current study, we wish to explore the
relationship of RAN performance to reading in a more
advanced level of grade school reading in Japanese.
1.3 VISUAL PROCESSING
As previously noted, there is well-established evidence
that phonological analysis and RAN play critical roles in
reading acquisition (Manis, Doi, Bhadha, 2000; Miller et al.,
2006), but there is growing evidence that various visual and
orthographic tasks also add significant variance to reading
(Badian, 2005; Swanson, Trainin, Nechoechea, et al., 2003).
Badian defined visual-orthographic skills as the ability to
recognize the correct orientation of letters and numbers
(Badian, 2005). Bowers and Wolf (1993) suggested that
orthographic processing skills can beassessed with visual
matching tasks which require careful scanning and
comparing of symbols and symbol sequences. Badian
(1994) reported that a visual-orthographic task in
kindergarten, which required the child to scan for matching
letters and word-like groups of letters, was the best predictor
of reading and spelling in first graders when compared to
phonological awareness and serial naming speed. In longitudinal studies, visual symbol matching was found to
significantly predict reading comprehension, reading
vocabulary, and spelling (Badian, 1995). Orthographic
pattern recognition task performance has also been relatedto word-level decoding (Bourassa, reiman, 2003).
In our previous studies of Japanese first graders, visual
search speed significantly correlated with speed of oral
reading and reading comprehension. In a study of Chinese,
a lexical decision task requiring children to distinguish
between correctly oriented radicals and those that are
reversed did not differentiate good readers and poor readers
among third graders (Penney, Leung, Chan, et al., 2005).
Katzir, Wolf, O'Brien, et al. (2006) have pointed out that
contributions of visual or orthographic processing skills to
reading fluency and comprehension are still under-
investigated. In addition, it has been suggested that intact
phonological processing skills alone is not sufficient enoughto fully explain development of fluent reading and
comprehension (Duel, Griffith, Gough, 1986). In the present
study, we employ a visual search task to investigate the
hypothesis that visual search processes may predict later
Japanese grade school reading.1.4 GENERAL KNOWLEDGE
We previously stressed "fluency" as a key to lead children
to reading comprehension; however, fluent decoding alone
is not sufficient for reading comprehension (Hirsch, 2003).
General knowledge, which includes prior or background
knowledge about the topic, cultural understanding, and
world and vocabulary knowledge, is important to deepen
reading comprehension (Hirsch, 2003; McNamara, Kintsch,
Songer, et al., 1996). The "fourth grade slump" refers to
the finding that a child's reading achievement starts to fall off
in around fourth grade even though he or she appears to be
making adequate progress in basic decoding skills (Chall,
Jacobs, 2003). According to Chall's reading stage theory, inthe beginning of Stage 3 (grades fourth to eighth), children
need to use basic decoding skill as a tool for learning new
words and ideas beyond their own language and their
knowledge of the world (Chall, 1983). Japanese national
curriculum standards for fifth and sixth grader reading
target the enhancement of the student's ability to identify
and choose reading materials independently and to
comprehend the content in order to obtain new information,
deepen their knowledge, and construct further thinking
(Ministry of Education, Culture, Sports, Science andTechnology; MEXT, 2003). Even though a reader can
successfully decode a text, if he or she fails to make rapid
connections between new and previously leaned content or
prior knowledge, adequate reading comprehension might
12 コミュニケーション障 害 学 Vol.24 No.1 2007.4.
not be achieved. Therefore, the current study will explore
the contribution of general knowledge to the comprehension
of fifth grade text.
Investigations of relationships between cognitive-
linguistic factors and developing reading abilities in
Japanese are under-researched. The lack of research onnon-alphabetic orthographies limits our accurate
understanding of both typical and impaired reading function
and it is critical to know whether a particular model applies
across different orthographies or to just one type of
orthography. The ultimate goal of the present study is to
contribute to the current understanding of the underlying
structures of different aspects of Japanese reading by
examining cognitive-linguistic measures which have already
been found to be related to English reading. In this study,
we sought to investigate following questions;
1) What are the relationships between selected cognitive
linguistics measures (phonological analysis, RAN, visual
search) and types of reading performance in Japanese 5th
graders?2) What are the relative contributions of each of these
cognitive linguistic measures to reading performance in
Japanese 5th graders?Based on the previous studies in English and on our
studies of Japanese reading among kindergartners and first
graders, we hypothesized that phonological analysis, RAN,and visual search would each predict unique variance in fifth
grade reading performance in Japanese. Additionally, weexpected that the contribution of each to reading
performance would differ from patterns found in Japanesekindergartners and first graders. We hypothesized that,
because fifth grade text is more complex and comprises
more kanji characters, in addition to phonological analysis
and RAM's contribution, visual search would contribute
more toward prediction of reading accuracy and speed. In
addition, phonological analysis performance at the phoneme
level would contribute more to reading accuracy than mora
deletion, and general knowledge would contribute to
reading comprehension.
2. METHOD
2.1 PARTICIPANTS
The participants were 39 (23 boys, 16 girls) monolingual
typically developing Japanese speakers aged 10 to 11 years
in a small suburban elementary school in Chiba prefecture.
Mean age at the time of testing was 11. 24 years of age (SD=
3. 3 months), with a range of 130 to 141 months. There were
no participants diagnosed with ADHD or any other
developmental disabilities.
2.2 MEASURES
The following measures were included in this study.
2.2.1 Phonological Analysis
There were two types of tasks administered in this study,
mora deletion and sound matching.
(1) Mora DeletionThe Japanese mora deletion test was based on the
Elision subtest of the Comprehensive Test of
Phonological Processing (CTOPP) (Wagner, Torgesen,
Rashotte, 1999), and was the same task used in our
previous studies (Kobayashi et al., 2003; Kobayashi et al.,2005). There were three practice items followed by
twenty test items. Children hear a real word, repeat the
word, delete one specified mora, and say the
reconstructed word. The number of morae for test items
increased from four to eight. If a child was not able to
perform the practice item, the test was not given. Testingwas discontinued when a child missed three items in a
row. Scores are reported as the number correct.
(2) Sound Matching TasksFirst Sound: This task was taken from Endo's (1991)
study. The student is asked to look at 1 target and 3
testing pictures. The examiner points to each picture and
says a word corresponding to each picture. Then the
student is asked to point a picture which starts with the
same sound (phoneme) as the target picture's name. The
original task had two testing choices, but this task was
modified to offer three testing choices to decrease the
influence of children's guessing (see Appendix A).
There were four practice items followed by 10 test items.
For all testing items, no feedback was given by the
examiner and testing. Scores are reported as the number
correct.
Last Sound: In this task, children were asked to
identify which word out of three words has the first mora
ending with the same sound (phoneme) as the first mora
of a target word orally presented by the examiner. The
same procedure from the First Sound task was employed
for this task (see Appendix B). There were four practice
items followed by 10 test items. Scores are reported as the
number correct.
2.2.2 Rapid Automatized Naming
This task requires the child to name as quickly as possible
six symbols repeated six times in random order anddisplayedonachartina4×9format.Two RAN tasks
(objects, number) were taken from the CTOPP (Wagner et
al. 1999). Two additional RAN tasks (hiragana, kanji),
constructed to simulate the format of RAN tests on the
CIOPP, were taken from our previous studies (Kobayashi et
al ., 2003; Kobayashi et al ., 2005). The examiner says to the
13KOBAYASxi et al.: Factors Affecting Reading in Japanese
participant, "I want you to say the names of all of theseobjects (or letters) as fast as you can. Try not to skip any.
Ready? Begin." The examiner starts a stopwatch timing as
soon as the participant says the name of the first object (or
letter) and stop timing when the name of the last item is said.
The scores were reported as the time taken in seconds to
complete the chart.
2.2.3 Visual Search
The short version of the cross-out subtest of the
Woodcock Johnson-Revised Cognitive Battery (WJ-R)
(Woodcock, Johnson, 1989) was included as a measure ofvisual processing. This task requires the child to look at a
target symbol on the left side of the page, and search for
(circle) five identical symbols included randomly in a row of19 stimuli. There were two practice rows and 11 test rows.
Scores recorded included the time taken in seconds to
complete the task and the error score which consisted of the
total number of missed targets and circled incorrect items.
2.2.4 General Knowledge
The WISC-III Information subtest (Azuma, Ueno, Fujita,
et al., 1998) was administered to each child to obtain
measures of general knowledge. The examiner asks a
question regarding factual knowledge and historicallyfamous people and the child is required to produce the
answer. Raw scores were analyzed.
2.2.5 Reading Speed and Accuracy
The child was asked to read aloud a short fifth grade level
passage comprised of hiragana, katakana, and kanji. Thescore for passage reading speed was time in seconds taken
to read the passage. Error scores were defined as the
number of morae read incorrectly.
2.2.6 Reading Comprehension
Reading comprehension was tested by a standardized
reading comprehension measure (Okamoto, Muraishi,
1983). The Kuder-Richardson reliability coefficient was 0.93.
Validity, confirmed by good-poor analysis, was Z=3.662.
Twelve minutes were given to children to complete this test.
There were 28 multiple choice questions on seven short
passages including narratives (4) and expository texts (3).Children were required to read silently and circle the
answers. The score was the number of correct items.
2.3 PROCEDURES
The reading comprehension test was group-administered
to participants in their classrooms by classroom teachers.
The rest of the measures were administered individually in a
quiet room in their school, generally in two twenty minutessessions by the examiner. The sequence of tests was fixed in
an order that avoided consecutive administration of similar
tasks.
3. RESULTS
Means, standard deviations, and ranges for all variables
are given in Table 1 for the total sample.
3.1 CORRELATIONS
Question 1) What are the relationships between selectedcognitive linguistics measures (phonological analysis, RAN,
visual search) and types of reading performance in Japanese
5th graders?
Intercorrelations among scores on phonological analysis,
RAN, and visual search, and the three reading measures
were examined using Pearson correlation analyses,
controlling for children's age (see Table 2). Children's age
Table 1. Descriptive data for the total sample (N=39)
14 コミュニケーション障 害 学 V0l.24 No.1 2007.4.
was partialed out in order to examine the relationships
between variables after controlling for the age differences
(in months) among the participants in our sample.negative correlations of RAN, visual search (speed and
accuracy), and oral reading (speed and accuracy) with other
variables reflect the fact that higher scores indicate lower
performance.Mora deletion and all the RAN tasks except for object RAN
correlated with oral reading speed. Sound matching-last,
kanji RAN, and visual search speed correlated with oral
reading accuracy. WISC-III Information and sound
matching-last correlated with reading comprehension. All
the reading performance measurements correlated with
each other.
3.2 REGRESSION ANALYSIS
Question 2) What are the relative contributions of each
cognitive linguistic measure (phonological analysis, RAN,
visual search) to reading performance for Japanese 5th
graders?Hierarchical regression analyses were performed to
determine the unique variance contributed to reading
measures by each cognitive linguistic measure which
showed significant correlations with reading performance.
Each of the predictors was entered into the model
individually at the second step after participants's age and
Table2. Correlations between measures, partialing out age (Total=39)
*p<.05; **p<.01; ***p<.001
Note: WISC=WISC-III Information, MD=mora deletion, SM_F=sound matching (first sound), SM_L=sound matching (last sound),
RAN_O=Object RAN, RAN_N=Number RAN, RAN_H=Hiragana RAN, RANK=Kanji RAN, VS S=visual search (speed), VS A=visual search
(accuracy), OR S=oral reading speed, OR A=oral reading accuracy
Table3. Hierarchical regressions of variables separately
predicting reading performance
*p<.05; **p<.01; ***p<.001
Note: Each of the predictors was entered into the model
individually at the second step after participants's age and WISC-III
Information were entered into the model at the first step.
15KOBAYASHI et al.: Factors Affecting Reading in Japanese
WISC-III Information were entered into the model at the first
step to control for the effects of these variables on reading
performance. As shown in Table 3, the results demonstratethat participants' performance on mora deletion, number
RAN, hiragana RAN, and kanji RAN tasks explained variance
in oral reading speed. Participants' performance on sound
matching-last, kanji RAN, and visual search speed were
significantly associated with their performance on oral
reading accuracy. Participants' performance on sound
matching last did not significantly explain variance in
reading comprehension beyond the participants' age and
WISC-III Information variables.
Stepwise regressions were conducted to determine the
best predictors of the reading measures (oral reading speed
and accuracy, reading comprehension) among the
processing variables (mora deletion, sound matching firstand last, object RAN, number RAN, hiragana RAN, kanji
RAN, visual search speed and accuracy) (see Table 4). Mora
deletion remained significant in explaining variance in oral
reading speed. For oral reading accuracy, the only
significant predictor was visual search speed. There was no
single processing variable found to predict reading
comprehension. When WISC Information was included in
the model, however, its contribution to the prediction of
reading comprehension was significant.
Reading fluency is one of several critical components
necessary for reading comprehension in English. Because
there was no single processing variable that predicted
reading comprehension, stepwise regressions were
conducted to determine which component of oral reading
performance (reading accuracy and/or reading speed)contributed to fifth grade reading comprehension. Oral
reading speed was the only significant contributor to reading
comprehension (see Table 5).
4. DISCUSSION
In this study, a group of Japanese fifth graders were
administered a battery of tests that included phonological
analysis (mora deletion, sound matching first and last), four
kinds of RAN (object, number, hiragana, kanji), and visual
search (cross out speed and accuracy), as well as WISC
Information and measures of reading comprehension and
fluency. There were two major goals in this study. The first
was to examine relationships between selected cognitive-
linguistic skills and reading performance (oral reading
speed and accuracy, reading comprehension). The second
was to determine which measures best predicted fifth grade
reading performance. We also examined whether a stage
theory of advanced level reading acquisition that had been
developed primarily from English populations also held for
Japanese fifth graders.Mora deletion, number RAN, hiragana RAN, and kanji
RAN each contributed to prediction of oral reading speed
independently and the measure that most effectively
predicted oral reading speed was mora deletion. Withrespect to oral reading accuracy, sound matching last, kanji
RAN, and visual search speed contributed variance
independently; however, when all the dependent measures
were entered into the stepwise regression, only visual search
speed consistently predicted oral reading accuracy. Sound
matching of the last phoneme significantly predicted reading
comprehension; however, when the stepwise regression
included all the dependent measures, its predictive power
was canceled. When WISC-III Information was entered into
the stepwise regression along with other processing
Table4. Summary of stepwise multiple regression
analyses for reading performance
*p<.05; **p<.01; ***p<.001
Note: These stepwise regressions included all the processing
variables (mora deletion, sound matching first and last, object RAN,
number RAN, hiragana RAN, kanji RAN, visual search speed and
accuracy).
*p<.05; **p<.01; ***p<.001
Note: This stepwise regression model included all the processing
variables and WISC-III Information.
Table5. Summary of stepwise multiple regression
analyses for reading comprehension
*p<.05; **p<.01; ***p<.001
Note: This stepwise regression included oral reading speed and
accuracy.
16 コミュニケーション障害 学 Vol.24 No.1 2007.4.
measures, its contribution towards prediction of reading
comprehension was the largest among all the variables.
Mora deletion correlated with all the RAN tasks except for
object RAN. This result is largely consistent with findings
reported in studies of alphabetic reading, and supports the
view that phonological analysis and RAN tap into some of the
same underlying processing abilities (e.g., retrieval and
production of phonological information) (Miller et al., 2006).Unlike some studies of alphabetic reading that have
indicated weak relationships between phonological analysis
and RAN (Wolf, Bowers, 1999; Swanson et al., 2003), this
study's correlations between a phonological analysis
measure (mora deletion) and RAN speeds were relatively
high (r=-.52to-.57, p<.001) and lend support to Miller et
al.'s (2006) conclusions.
Despite the above correlations, hierarchical regression
analyses demonstrated that both phonological analysis
(mora deletion) and RAN also contributed independentvariance to oral reading speed. This finding suggests that
certain underlying processing differences in these tasks may
be significant in predicting text-level fluency (e.g.,
recognition of visual stimuli, association of visual with
phonological symbols, manipulation of phonologicalinformation, speed). The independent contributions of
phonological analysis and rapid naming to prediction ofreading performance is consistent with a double deficit
hypothesis as proposed by Wolf et al. (2002). Further
support for this view is underscored by the finding that
functional magnetic resonance imaging shows basic
differences in underlying processes required by
phonological analysis and RAN measures (Misra et al.,2004).
Within the limitations of a cross sectional analysis, several
interesting questions arise. The data from our previous
study of first grade beginning readers (Kobayashi et al.,
2005) suggest some complex developmental similarities and
differences when compared with results of our current study
of fifth graders. In first graders, high correlations were
found between mora deletion and all measures of reading
(speed, accuracy and comprehension). In the current studyof fifth graders, mora deletion correlated only with oral
reading speed. In addition, performance on all of the RAN
tasks (object, number, hiragana, kanji) in first grade
correlated with oral reading speed and comprehension,
while symbol-based RAN tasks in fifth grade correlated
significantly with oral reading speed but not with
comprehension. There were also differences in correlations
between mora deletion and RAN tasks. In first graders,
object and kanji RAN correlated with mora deletion, while in
fifth graders the three symbol-based RAN tasks correlated
strongly with mora deletion. The following questions arise:
1) Why is a task of phonological analysis (mora deletion)
more highly correlated with RAN performance in the fifth
grade than in the first grade?2) Why does mora deletion continue to have a significant
correlation with reading speed in grade five?
3) Why are there correlations between reading
comprehension and performance on cognitive linguistic
tasks of mora deletion and RAN in the first grade data, but
not in the fifth?
4) In the fifth grade data, why is there no significant
correlation between reading comprehension and RAN
performance, given the strong correlation between RAN
performance and oral reading speed and between oralreading speed and reading comprehension?
Potential answers to these questions may lie in
developmental differences in the processing requirements
of reading and comprehending text at the first versus fifth
grade. One explanation for the first question as to why
phonological analysis (mora deletion) is more highlycorrelated with RAN performance in the fifth grade than in
the first grade, may lie in the different processes tapped by
the mora deletion task for older children. At the fifth grade
level, children typically manipulate stimuli of increased
length (up to eight morae); in order to be successful at this
level, it is possible that children may need to be able to
rapidly coordinate phonological memory, retrieval and
orthographic (kana) visualization processes. As noted
earlier, RAN makes similar requirements for coordination ofvarious underlying processes (Korhonen, 1995; Snyder,
Downey, 1997; Wolf, Bowers, 1999, 2000).
Our second question, as to why mora deletion continues to
have a significant correlation to reading speed in grade five
may also be related to the discussion above which
emphasizes differing requirements of the mora deletion task
at different grade levels. The text used in our study of first
graders consisted of 5% kanji characters whereas the fifth
grade text in the current study contained 23% kanjicharacters. Kanji characters require the retrieval of correct
sounds that may have multiple meanings and/or multiple
pronunciations. The increase in the number of kanjicharacters coupled with increased syntactic and semantic
complexity of text at this stage seems to place more
demands on working memory, perhaps in a similar manner
to the mora deletion task. This may explain why the
phonological analysis task best predicted rapid decoding ofthe fifth grade connected text in Japanese.
The third question, as to why there arecorrelations
between reading comprehension and performance on
cognitive linguistic tasks of mora deletion and RAN in the
17KOBAYASXI et al.: Factors Affecting Reading in Japanese
first grade data, but not in the fifth, may be related to
developmental differences in the processing requirements
of first versus fifth grade reading. Early elementary text
contains familiar concepts and decoding is sufficient for
comprehension; the emphasis is primarily on learning to
map sounds onto hiragana and katakana characters, which
have mostly regular symbol-sound correspondences. By
fifth grade he focus in reading comprehension tasks is on
gaining new information and assumes adequate wordidentification. Our findings support Manis, Seidenberg, and
Doi's hypothesis that phonological analysis is more related
to the learning of systematic letter-sound correspondences,
whereas the RAN task involves routinized, automatic
symbol-sound associations (1990). Both of these processes
are largely automatic in the typically developing fifth grade
reader. As Schatschneider et al. (2002) suggested, RAN
appears to best predict the elementary stage of word
recognition, and RAM's contribution to reading performance
may lessen when children's reading fluency becomes higher
such as in grade 3 or 4 (McBride-Chang, Manis, 1996).
Our fourth question was why there is no significant
correlation between reading comprehension and RAN
performance in the fifth grade, given the strong correlationbetween RAN performance and oral reading speed and
between oral reading speed and reading comprehension.
There does not appear to be a simple explanation for these
findings. They suggest, however, that the specific aspects of
RAN that correlate with oral reading speed may involve
processes that are distinct from those tapped in therelationship between reading speed and comprehension.
Current research is emphasizing the complexity of the
reading task and supports the view that "in reading fluency,
the whole is more than its parts" (p. 77, Katzir et al., 2006).
From our perspective, this would certainly hold true for
reading comprehension as well.
Our previous study showed that mora deletion
contributed to prediction of the first grader's oral reading
accuracy (Kobayashi et al., 2005); however, in this study,
visual search speed consistently contributed variance to oral
reading accuracy even after accounting for the influence of
all other processing variables. Thirty nine percent of the
total oral reading errors (hiragana errors 50%, katakana
errors 11%) came from incorrect decoding of kanji
characters; this suggests that the ability to accurately
recognize and recall visually complex configurations
partially influences fifth graders' reading accuracy. Poorvisual search speed performance may be related to poor
orthographic memory and eventually leads to failure to
retrieve accurate pronunciations for each character. In our
previous studies, in kindergartners, visual search speed did
not correlate with any aspect of reading performance,
however, in first graders, it correlated moderately with
reading speed and reading comprehension (Kobayashi et al.,
2003; Kobayashi et al., 2005). As children's mastery level of
reading advances, visual or orthographic processing may
become more important in explaining more advanced
reading in Japanese. This study's findings suggest the value
of a visual search task for explaining accuracy of reading.
Incorporation of this task in assessment and prediction
studies may prove very useful for older children.
Whereas morn deletion explained variance in oral reading
speed, sound matching last only explained variance in
reading comprehension. These findings suggest that even
though both mora deletion and sound matching tasks are
phonological analysis measures, they may tap into differentcognitive and linguist requirements. This is also
demonstrated in the comparatively weak correlation
between these two tasks. Mora deletion involves
manipulation of speech sounds at the mora level while sound
matching last requires recognition of final phonemes in
words. The metalinguistic skills underlying manipulation at
the mora level is consistently found to be important across
three decades of research, but phoneme identification in
Japanese has been less investigated. Lack of predictive
power of sound matching first can be explained by previousfindings that phoneme awareness of the initial sound of a
word is easier than that of the final sound (Spector, 1995).
The mean score for sound matching first (m=8.56, SD=2.17)
was significantly higher than the mean for sound matching
last (m=6.85, SD=3.42), p=0.003. Thus, sound matching
first was a relatively easy task, which did not tap into
underlying cognitive linguistic processes required to
perform any of the reading outcome measures.Even though, there was no single cognitive-linguistic
measure that significantly predicted fifth grade reading
comprehension, our results indicate that general knowledge
plays an important role in this area. The WISC-IIIInformation subtest reflects the richness of the child's
background knowledge. Our results indicate that successful
reading comprehension of fifth grade materials requires not
only fluent decoding skills, but also general knowledge
which appears to enhance basic comprehension and leave
room for making connections between the new information
and previously learned knowledge, drawing inferences, and
thinking over implications (Hirsch, 2003).Moreover, the
findings from this study have implications for intervention;
instruction and practice in fluency by training for
phonological analysis, rapid decoding, and recognition oforthography need to be accompanied by instruction and
practice in general knowledge.
18 コミュニ ケーション障 害 学 Vol.24 No.1 2007.4.
In conclusion, our research offered an insight into
underlying cognitive-linguistic skills required by Japanese
fifth grade reading. This study indicates that Japanese fifth
graders learning to speak and read Japanese also developsome of cognitive-linguistic skills found to be important for
reading in English. However, performance on phonological
analysis, RAN, visual search, and general knowledge only
partially explained accuracy, fluency, and comprehension ofreading. It has been suggested that different cognitive-
linguistic skills tap different aspects of linguistic processing
and, therefore, relate differently to reading (Ramus, 2003).
Therefore, there will be more cognitive-linguistic skills
related to explaining reading performance in Japanese, and
these skills are expected to vary according to grade level,
children's reading experience, and components of reading
assessed. As noted, some of our analyses are based on cross
sectional data; a longitudinal study would help clarify the
predictive power of these skills as children advance through
grade school. Further research will also provide us with atheoretical framework and refined understanding of
cognitive-linguistic skills of reading in Japanese and may
suggest guidance for practical remediation approaches.
ACKNOWLEDGMENTS
We express our appreciation to the students and teachers
who participated in the study as well as to Emiko Saito for
her dedication in coordinating and supporting this fieldwork.
We also thank to Kyoko Iitaka and Mitsuko Shindo for their
continuous understanding and support. This work was
partly supported by a research fund of the Japan Society forthe Promotion of Science.
REFERENCES
Adams, M. J. Beginningtoread: Thinkingandlearning
aboutprint. Cambridge, MA, MITPress, 1990.
天 野 清. 語 の音 節 構 造 の 分析 行 為 の 形 成 と か な文 字 の 読 み
の 学 習 (Formation of the act of analyzing phonemic
structure of words and its relation to learning Japanese
syllabic characters). 教 育 心 理 学 研 究. 18, 76-89 (1970).
天 野 清. 音 韻 分 析 と 子 供 のliteracyの 習 得 (Phonemic
analysis of words and literacy acquisition among
children). 教育 心 理 学 年 報. 27, 142-164 (1988).
Amano, K. Phonological analysis and acquisition of reading
and wriitng in children. Psychologia. 32, 16-32 (1989).
東 洋, 上 野 一彦, 藤 田 和 弘, ほ か. 日本 版WISC-III知 能検
査 法 (Japanese Wechsler Intelligence Scale for
Children Third Edition). 東 京, 日本 文化 科 学社, 1998.
Badian, N. Does a visual-orthographic deficit contribute to
reading disability? Annals of Dyslexia. 55, 28-52 (2005).
Badian, N. A. Preschool prediction: Orthographic and
phonological skills, and reading. Annals of Dyslexia.
44, 87-100 (1994).
Badian, N. A. Predicting reading ability over the longterm:
The changing roles of letter-naming, phonological
awareness and orthographic processing. Annals of
Dyslexia. 45, 3-25 (1995).
Badian, N. A. A validation of the role of preschool
phonological and orthographic skills in the prediction of
reading. Journal of Learning Disabilities. 31, 472-481
(1998).
Bourassa, D., Treiman, R. Spelling in dyslexic children:
Analyses from the Treiman-Bourassa Early Spelling
Test. Scientific Studies of Reading. 7, 309-333 (2003).
Bowers, P. G., Wolf, M. Theoretical links among naming
speed, precise timing mechanisms and orthographic
skill in dyslexia. Reading and Writing: An
Interdisciplinary Journal. 5, 69-85 (1993).
Bowey, J. A., Francis, J. Phonological analysis as a function
of age and exposure to reading instruction. Applied
Psycholinguistics. 12, 91-121 (1991).
Chall, J. S. Stages of reading development. New York,
McGraw-Hill, 1983.
Chall, J. S., Jacobs, V. A. Poor children's fourth-grade slump.
American Educator. 27(1), 14-15 (2003).
Cornwall, A. The relationship of phonological awareness,
rapid naming, and verbal memory to severe reading and
spelling disability. Journal of Learning Disabilities. 25,
532-538 (1992).
Cunningham, A. E., Perry, K., Stanovich, K. E., Share, D. L.
Orthographic learning during reading: examining the
role of self-teaching. Journal of Experimental Child
Psychology. 82, 185-199 (2002).
Denckla, M. B. Color-naming defects in dyslexic boys.
Cortex. 8, 164-176 (1972).
Denckla, M. B., Cutting, L. E. History and significance of
rapid automatized naming. Annals of Dyslexia. 49, 29-
42 (1999).
Denckla, M. B., Rudel, R. G. Rapid automatized naming of
pictured objects, colors, letters, and numbers by normal
children. Cortex. 10, 186-202 (1974).
遠 藤 め ぐ み. 日 本 人 幼 児 の 韻 の 感 受 性 と 拗 音 表 記 法 の 習 得
(Phonological awareness of Japanese young children
and learning to readand write yoo-on). 教 育 心 理 学 研
究. 448-454 (1991).
原 恵 子. 健 常 児 に お け る 音 韻 意 識 の 発 達 (Development of
phonological awareness in normally developing
children). LD児 口米 比 較 研 究 報 告II. 117-124 (1998).
Hirsch, E. D. Reading comprehension requires knowledge-
of words and the world. American Educator. 27 (1), 10-
29 (2003).
19KOBAYASHI et al.: Factors Affecting Reading in Japanese
Ho, C. S. H., Chan, D. W. O., Tsang, S. M., et al. The
cognitive profile and multiple deficit hypothesis in
Chinese developmental dyslexia. Developmental
Psychology. 38(4), 543-553 (2002).
Juel, C., Griffith, P., Gough, P. B. Acquisition of literacy: A
longitudinal study of children in first and second grade.
Journal of Educatfonal Psychology. 78, 243. 255 (1986).
金 子 真 人, 宇 野 彰, 春 原 則 子. 就 学 前6歳 児 に お け るrapid
automatized naming (RAN) 課 題 と仮 名 音 読 成 績 の 関 連
(Relationship between rapid automatized naming and
kana reading tasks in preschool children). 音 声 言 語 医
学. 45, 30-34 (2004).
Katzir, T., Wolf, M., O'Brien, Kennedy, B., etal. Reading
fluency: the whole is more than the parts. Annals of
Dyslexia. 56(1), 51-82 (2006).
Kobayashi, M., Haynes, C., Macaruso, P., etal. Effects of
mora deletion, nonword repetition, rapid naming and
visual search performance on beginning reading in
Japanese. Annals of Dyslexia. 55(1), 105-128 (2005).
小 林 マ ヤ (志 帆), 加 藤 醇 子, チ ャ ー ル ズ・ヘ イ ン ズ, ほ か.
幼 児 の 読 み 能 力 に 関 わ る 認 知 言 語 的 能 力 (Cognitive-
linguistic factors in Japanese children's reading). LD研
究. 12(3), 259-267 (2003).
Korhonen, T. T. The persistence of rapid naming problems
in children with reading disabilities: Anine-year follow-
up. Journal of Learning Disabilities. 28, 232-239 (1995).
Manis, F. R., Doi, L. M., Bhadha, B. Naming speed,
phonological awareness, and orthographic knowledge
in second graders. Journal of Leaming Disabilities. 33,
325-333 (2000).
Manis, F. R., Seidenberg, M. S., Doi, L. M. See Dick RAN:
Rapid naming and the longitudinal prediction of reading
subskills in first and second graders. Scientific Studies
of Reading. 3, 129-157 (1999).
Mann, V. A. Phonological awareness: The role of reading
experience. Cognition. 24, 65-92 (1986).
McBride-Chang, C., Manis, F. Structural invariance in the
associations of naming speed, phonological awareness,
and verbal reasoning in good and poor readers: A test of
the double-deficit hypothesis. Reading and Writing: An
Interdisciplinary Journal. 8, 323-39 (1996).
McNamara, D. S., Kintsch, E., Songer, N. B., etal. Are good
texts always better? Interactions of text coherence,
background knowledge, and levels of understanding in
learning from text. Cognition and Instruction. 14, 1-43
(1996).
Meyer, M. S., Felton, R. H. Repeatedreading to enhance
fluency: Old approaches and new directions. Annals of
Dyslexia. 49, 283-306 (1999).
Meyer, M. S., Wood, F. B., Hart, L. A., etal. Selective value
of rapid automatized naming in poor readers. Journal of
Learning Disabilities. 31, 106-117 (1998).
Miller, C. J., Miller, S. R., Bloom, J. S., etal. Testing the
double-deficithypothesis inanadult sample. Annals of
Dyslexia. 56, 83-102 (2006).
文 部 科 学 省. 平 成15年 度 改 訂 小 学 校 学 習 指 導 要 領
(National Curriculum Standard for Elementary and
Lower Secondary School). 2003.
Misra, M., Katzir, T., Wolf, M., etal. Neural systems for
rapidautomatized naming in skilled readers:
Unraveling the RAN-reading relationship. Scientific
Studies of Reading. 8, 241-256 (2004).
OECD. The PISA 2000 Assessment Framework:
Mathematics, Reading, Science and Problem Solving
Knowledge and Skills. OECD Publications (2000).
OECD. The PISA 2003 Assessment Framework:
Mathematics, Reading, Science and Problem Solving
Knowledge and Skills. OECD Publications (2003).
大 石 敬 子. 読 み 障 害 児5例 の 障 害 機 構 の 検 討 と指 導 (Study
on the mechanism of dyslexia of five children and their
remediation). LD児 日 米 比 較 研 究 報 告, 50-54 (1998).
岡 本 奎 六, 村 石 昭 三. 教 研 式 全 国 標 準 読 書 力 診 断 検 査 高 学
年 (Reading Ability Analysis Test). 東 京, 図 書 文 化 社,
1983.
Penney, T. B., Leung, K. M, Chan, P. C., etal. Poor readers
of Chinese respond slower than good readers in
phonological, rapid naming, and interval timing tasks.
Annals of Dyslexia. 55, 9. 27(2005).
Ramus, F. Developmental dyslexia: specific phonological
deficit or general sensorimotor dysfunction? Current
Opinion in Neurobiology. 13(2), 212-218 (2003).
Scarborough, H. S. Predicting the future achievement of
second graders with reading disabilities: Contributions
ofphonemic awareness, verbalmemory, rapidnaming,
and IQ. Annals of Dyslexia. 48, 115-136 (1998).
Schatschneider, C., Carlson, C. D. Francis, D. J., etal.
Relationship of rapid automatized naming and
phonological awareness in early reading development:
implications for the double deficit hypothesis. Journal
of Learning Disabilities. 35, 245-256 (2002).
Snyder, L. S., Downey, D. Developmental differences in the
relationship between oral language deficits and reading.
Topics in Language Disorders. 17, 27-40 (1997).
Spector, J. E. "Phonemic Awareness Training: Application
of Principles of Direct Instruction." Reading and
Writing Quarterly: Overcoming Learning Difficulties.
11(1), 37-52 (1995).
Stringer, R., Toplak, M. E., Stanovich, K. E. Differential
relationships between RAN performance, behavior
ratings, and executive function measures: Searching for
20 コミュニ ケーション障 害 学 Vol.24 No.1 2007.4.
a double dissociation. Reading and Writing: An
Interdisciplinary Journal. 17, 891-914 (2004).
Swanson, H. L., Trainin, G., Nechoechea, D. M., et al. Rapid
naming, phonological awareness, and reading: A meta-
analysis of the correlation evidence. Review of
Educational Research. 73(4), 407-440 (2003).
Wagner, R. K., Torgesen, J. K. The nature of phonological
processing and its causal role in the acquisition ofreading skills. Psychological Bulletin. 101, 192-212
(1987).
Wagner, R., Torgesen, J. K., Rashotte, C. Comprehensive
Test of Phonological Processing. Austin, TX, Pro-Ed.,
1999.
Wolf, M., Bowers, P. G. The double-deficit hypothesis for
the developmental dyslexia. Journal of Educational
Psychology. 91(3), 124 (1999).
Wolf, M., Bowers, P. G. Naming-speed processes and
developmental reading disabilities: An introduction to
the special issue on the double-deficit hypothesis.
Journal of Learning Disability. 22, 322-324 (2000).Wolf, M., O'Rourke, A. G., Gidney, C., et al. The second
deficit: an investigation of the independence of
phonological and naming-speed deficits indevelopmental dyslexia. Reading and Writing: An
interdisciplinary Journal. 15, 43-72 (2002).
Woodcock, R. W., Johnson, B. M. Woodcock Johnson
Psycho-educational Battery Revised. Itasca, IL,
Riverside publishing, 1989.
APPENDIX A Stimulus Words Used for Sound Matching First
(Underlined letters indicate target sounds)
APPENDIX B Stimulus Words Used for Sound Matching Last
(Underlined letters indicate target sounds)
小 学5年 生 に お け る音 韻 認 識, RapidNaming, 視 覚 探 索 課 題 の 読 み 能 力へ の 影 響
小林 マヤ (志帆) CharlesW. HAYNES Pamela E. HooK Paul MACARUSO
近 年, 読 解の 重要 性 が明 らかに な るにつ れ, 読 み にお
け る “流暢性 (fluency)”が 着 目されて い る. “Fluency”
とは, 文 字 の音 声化 を 自動 的 かつ 容易 にで き るだ けで
は な く, そ の文脈 に合 った正 しい リズ ム, イ ン トネ ー
シ ョン, 間 (ポー ズ) を適用 しな が ら読 め る こと を意 味
す る. この “fluency”を 所 持す る読 み手 は, 自分 の注
意 力 を文書 の 内容, メ ッセ ー ジへ と向 け る ことが で き
るた め, 効 果 的 な読 解 が 可能 にな る と考 え られ て い る
(Meyer, Felton, 1999). 英 語 圏で は, この “fluency”に
関 わ る認知 言語 的能 力 と して, 音 韻分 析 (phonological
analysis), Rapid Automatized Naming (以下RAN), 視
覚 探索 (visual search) な どが報 告 され て い る (Badian,
1994, 1995, 1998; Cornwall, 1992; Meyer, Wood, Hart,
etal., 1998; Scarborough, 1998; Cunningham, Perry,
Stanovich, et al., 2002). 今 回の 調査 で は, 日本 の公 立学
校 の普通 学級 に在 籍 して い る小 学5年 生 (N=39) を対
象 に, これ らの認知 言 語 的能力 と読 み能 力 (流暢性 と読
解) との 関係, そ して それ ぞれ の読 み能 力 に対 す る予 測
精度 を調 査 した.
音韻 分析 (phonological analysis), また は音韻 認 識
(phonological awareness)は, 音 声言 語 の音韻 構 造 を分
析 し, それ を操作 す る能 力 を意 味 す る. 英 語 圏の研 究 で
は, 音韻 分析 能 力 と音読 ・スペ リング ・読 解 など との関
連 性 が明 らか にな って お り, そ して その能 力 は将来 の
読 み習 得度 を予測 す る ことが報 告 され てい る(Badian,
1995; Cornwall, 1992). 日本語 圏 にお いて も, 児 童 の音
韻 分析 発 達 と読 み能 力 との 関連 性 につ いて の研 究 が報
告 されて い るが(天 野, 1970, 1988, 1989; 原, 1998;
21KOBAYASHI et al.: Factors Affecting Reading in Japanese
大 石, 1998), 日本 語の音韻 分析 能力 を測 定 す る標 準化
され た検 査 は依 然 と して存在 しない. 今 回 の調 査 で は
以 前 の 調査 で使 用 したモ ーラ削 除 課題 と, モ ー ラ操作
よ り も難 易 度 が高 い と され る音 素 認 識 課 題 (遠 藤,
1991) を使 用 した.
RANは, 不規 則 かつ連続 的 に列 記 され た5~7種 類
の オ ブ ジ ェク ト, 文字, 数字, 色 な ど をで き るだ け早 く
読 み 上 げ る とい う課題 で ある (Denckla, 1972; Denckla,
Rudel, 1974). 英語 圏 で, 読 み障害 児 のRAN課 題 を行 う
ス ピー ドが健 常児 よ り も著 しく遅 い こ とが報 告 され て
以 来, RANと 読 みパ フォーマ ンス との 関連研 究 が多数
報 告 され て い る (Denckla, Cutting, 1999;
Schatschneider, Carlson, Francis, etal., 2002; Wolf,
Bowers, 1999, Wolf, O'Rourke, Gidney, et al., 2002).そ
して欧 米 で はRAN課 題 が音韻分析 課題 と並 び, 読 み障
害 児 (デ ィス レキシ ア) のス ク リーニ ング. テ ス トと し
て広 く使 用 され てい る. 我 々の調 査 で は, オ ブ ジ ェ ク ト
(椅子, 魚, 鉛 筆, 星, 鍵, 船), 数字 (2, 3, 4, 5,
7, 8), ひ らがな (た, に, を, い, の, と), 漢字 (金,
人, 上, 手, 月, 山) の4種 類のRAN課 題 を使 用 した.
視 覚 探 索 とは, 文 字や 図形の形 態 を迅 速 に弁別 す る
能 力 で あ る (Badian, 2005). これ は, 類 似 す る形 また は
文字 が並 べ られ てい る背 景群 の 中か ら, 目標 とす る形 ・
文字 を検 出す る課 題で測 定す る (Bowers, Wolf, 1993).
そ の処 理 能 力 が 自動 的かつ 容易 で あ るほ ど文 字 を流 暢
に読 む こ とが で き ると考 えられ て い る (Adams, 1990).
近年 欧 米 で は, 視 覚 探索能 力の読 み パ フ ォーマ ンスに
対 す る予 測精 度 に関す る報告 が増加 す る傾 向 にあ るが,
音韻 分析 とRANに 比 べ る とその報 告例 は少 な く, 今 後
さ ら な る研 究 が必 要 な領 域 で あ る (Katzir, Wolf,
O'Brien, et al., 2006).
効 果 的 な 読 解 を 達 成 す る た め に は,“流 暢 性
(fluency)”に 加 えて,“一 般 的知識 ”が重 要 な役割 を果
たす こ とが報 告 されて いる (Hirsch, 2003). 一般 的知識
とは, 主題 に まつわ る背景 的知識, 文化 に対 す る造詣,
世 界観 や語 彙 な どを含 める (Hirsch, 2003; McNamara,
Kintsch, Songer, et al., 1996). 文 字 を音声 化 す る とい う
基 本 的 な スキ ル をマ ス ター した と して も, そ の読 み手
の 一般 的知 識 が乏 しい もの だっ た場合, 読 書 か ら入手
しよ う と した新 しい知 識 と自分 が既 に有 す る知 識 との
関 連付 け に失敗 し, 結 局 は効 果的 な読 解 が達 成 で きな
い と考 え られ て い る (Hirsch, 2003). 今 回 の調 査 で は,
基 本 的 な読 字 ス キル をマ スター した レベ ル に あ る小学
5年 生 を対象 に して い るた め, 認知 言 語 的能 力 に加 え
て一 般 的知 識 の 読解 に対す る影響 を検 証 す る.
我 々の これ まで の調査 によ り, 英 語 圏 に お いて読 み
能力に関連するとされる認知言語的能力は日本人幼稚
園年長, 小学1年 生の読み能力にも関っていることが
確認された (Kobayashi, Haynes, Macaruso, et al., 2005;
小林, 加藤, ヘインズ, et al., 2003).
日本人の幼稚園年長児の音読 (速さと正確度) を予測
した課題はひらがなRANと 数字RANで あった. 日本人
の小学校1年 生を対象にした調査では, 音読の速さを
予測した課題はひ らがなRANと 漢字RANで あったが,
音読の正確度を予測したのは, 音韻分析課題であった.
さらに, 小学1年 生の読解を予測したのは, 漢字RAN,
音韻分析, 音韻記憶であった.
今回の調査では, 小学5年 生というやや高レベルの
日本語読みに関わる認知言語的能力を解明すること,
そして読みに困難を生じている児童を検出するために
必要な予測精度の高い認知言語的課題を構築するため
の手がかかりを示唆することを目的としている.
調査の結果, 小学5年 生の音読の速さにおいて一番
高い予測精度を示したのは, モーラ削除であった. モー
ラ削除の影響がない場合, 数字RAN, ひらがなRAN, 漢
字RANが 音読速さ予測に寄与した. 音読正確度におい
て, 一番高い予測精度を示 したのは視覚探索であった.
視覚探索の影響がない場合, 音素認識, 漢字RANが 音
読の正確度予測に寄与した. 音素認識とWISGIII知 識
の双方は読解との相関が有意であつたのにもかかわら
ず, 予測精度を示 したのはWISC-III知 識のみであった.
今回の調査によって, 日本語における小学5年 生の
読みが必要とする基盤的な認知言語的能力の理解を深
めることができた. また, 欧米で構築された読み能力獲
得のための理論体系が日本語の小学5年 生の読みにお
いてもほぼ応用可能であるということが示 された. そ
して幼稚園年長児, 小学1年 生の調査結果を踏まえる
と, 読み能力を予測する認知言語的能力は読む課題の
種類 (音読の速さ, 正確さ, 読解), 児童の学年などに
よって変化することが示された. 臨床への手がかりと
しては, 読解指導は文字の音声化だけではなく, 一般的
な知識の向上も含めて指導 していくことがより効果的
な読解につながることが示唆 された.
しかしながら, 今回取 り上げた音韻分析, RAN, 視覚
探索, そして一搬 的知識は, 小学5年 生の読みの流暢性
と読解の一部を予測 したのにすぎず, 他の認知言語的
能力が読み能力予測に関わっている可能性がある. 今
後, 横断的, 縦断的調査を実施することによって, 日本
人児童の読み能力に関わる認知言語的能力のさらなる
解明と, 治療教育へのガイダンスを示唆することが可
能であると考えられる.