Background：The Japanese writing system associates imported Chinese characters with the sounds of Japanese speech according to their original Chinese meanings, and developed kana to represent Japanese syllables for reading Chinese text as Japanese. In Japanese-speaking patients with aphasia, interventions often exploit the distinctive properties of each script；among these, kana, which correspond one-to-one with syllables, are commonly used in training for phonological disorders. We report the training process in a patient with conduction aphasia in whom impaired kana-syllable correspondence made kana-based phonological training using kana difficult, and for whom kanji readings were used therapeutically.
Case Presentation：The patient was a 64-year-old right-handed male who developed conduction aphasia following a hemorrhagic infarction in the posterior portion of the left middle and inferior frontal gyri and the supramarginal gyrus. Kana-phonology correspondence was poor, precluding conventional kana-based training for phonological deficits. We therefore established a kana-phoneme correspondence by pairing kana characters with kanji characters that had corresponding pronunciations.
Discussion：Mapping logographic characters (kanji) onto the phonemes of phonetic characters (kana) may provide a useful therapeutic strategy for phonological disorders in Japanese-speaking patients with aphasia.
背景：日本語の文字言語は，伝来した漢字を従来の中国語の意味と日本語の音声言語を対応させ，さらに中国語の文章を日本語として読むために日本語の音節に対応した仮名を生み出した。そのため日本語話者の失語症患者では各文字特徴を活かした訓練が成され，そのなかで仮名は音節と1対1対応であるため音韻障害に対する訓練に活用される。今回は仮名と音節の対応が障害され仮名を用いた音韻訓練が困難な伝導失語例に対する漢字を活用した訓練経過を報告する。
症例提示：患者は64歳，右利き男性，左中・下前頭回後部と縁上回の出血性梗塞後に伝導失語を呈した。仮名と音韻の対応が不良で，仮名を活用した音韻障害に対する訓練はできなかった。そこで仮名の読み（音韻）に対応する漢字を用いて，仮名と音韻の対応を定着させた。
考察：表語文字である漢字を表音文字の仮名の音韻に対応させることで，音韻障害に対する訓練として漢字が活用できる可能性を示唆した。
Japanese has a distinctive writing system that uses two types of characters, kanji and kana. Kanji are logograms that represent morphemes, the smallest units of meaning, unlike phonograms such as alphabetic characters. Kana, in contrast, are phonograms that correspond one-to-one with syllables (morae), whereas alphabetic letters correspond to phenomes. Historically, kanji, which originated from pictographs, were introduced to Japan from China. Native Japanese words (wago) were assigned to kanji while retaining the characters’ original Chinese meanings. Kana, a uniquely Japanese script corresponding to Japanese syllables and derived from kanji used to represent Chinese pronunciations, was subsequently developed to read Chinese texts (kanbun) in Japanese. Kana later differentiated into katakana, used to render Chinese texts in Japanese, and hiragana, used mainly for private letters and diaries written in cursive style, whereas kanji were used for official documents.. Currently hiragana is primarily used for wago, particles, and inflections, whereas katakana is used for loanwords and onomatopoeia. Although both scripts can be read aloud because they are syllabic, more familiar scripts and words with greater orthographic plausibility are processed more rapicly in oral reading tasks (Takayama et al., 2017). In Japanese-speaking patients with aphasia, dissociations between kanji and kana have been observed in reading comprehension and oral reading (Iwata, 1984, Sasanuma, 1986, Kawahata et al., 1988), and neuroimaging studies have shown that kanji and kana involve different neural processing mechanisms (Sakurai et al., 2000, Ino et al., 2009, Horie et al., 2012). These orthographic and neural processing differences underlie the use of kanji and kana in aphasia treatment. In particular, information about the number of syllables and phonological structure is explicitly represented in kana because kana is a syllabic script in which each character consistently corresponds to one syllable without exception.
Patients with conduction aphasia exhibit phonemic paraphasias and phonemic paragraphias caused by disruption of phonological selection and sequencing during language processing (Shizuka et al., 2002, Soma, 1993). For these impairments, oral reading of kana and the provision of kana readings for kanji（ruby annotation）are recognized as effective and are commonly used in clinical intervention (Tanemura, 2006). As noted previously, kana is a syllabic script in which each character corresponds to one syllable, thereby making phonological information explicit for patients with unstable phonological representations of words. Case reports of conduction aphasia have shown that interventions using kana can facilitate phonological processing and reduce phonemic paraphasias. These interventions include oral reading of kana words and ruby annotation for kanji words (Imamura, 1994), oral repetition combined with oral reading of kana words (Uno et al., 1997), and selecting and sequencing kana characters from a set of options to form words corresponding to pictures before reading them aloud or naming them (Tanaka, 2006). Although kana-based intervention is effective for phonological deficits in conduction aphasia, the ability to read or write single kana characters to a some extent is a prerequisite for this approach, as Imamura stated, “The mildness of language impairment enabled this patient to select and sequence kana from the mental kana lexicon” (translated by the authors)(Imamura, 1994).
In the present case, we encountered a patient with conduction aphasia who had difficulty reading and writing single kana characters, which prevented us from implementing kana-based language therapy targeting phonological deficits. In contrast, despite impaired oral reading of single kana characters, his oral reading of some kanji with a single phonetic value (kanji with a single-syllable reading) was preserved. We therefore created a modified gojūon chart in which each basic Japanese kana character was paired with a single-syllable kanji character with the corresponding pronunciation, thereby facilitating the association between kana characters and syllables through kanji. As a result, the patient became able to read aloud and write single kana characters to dictation, which enabled kana-based language therapy for phonological deficits. Subsequently, phonemic paraphasias decreased across overall speech production. This report describes the intervention and clinical course.
64-year-old right-handed man with 12 years of education.
Past Medical History. type 2 diabetes mellitus, hypertension, and hyperlipidemia
History of Present Illness. The patient presented to his primary care physician with word-finding difficulty and weakness of the right upper and lower extremities. He was admitted on the same day with a diagnosis of hemorrhagic infarction. On post-stroke day 13, he was transferred to our hospital for rehabilitation, with sensory impairment of the right upper and lower extremities and aphasia.
Neuroradiological Findings. A head CT scan obtained on the day of transfer to our hospital (post-stroke day 13) is shown in Figure 1. The scan demonstrated a hemorrhagic infarct involving the cortical and subcortical white matter of the posterior portions of the left middle and inferior frontal gyri and the left supramarginal gyrus.
A hemorrhagic infarction was observed in the cortex and subcortical white matter of the posterior portion of the left middle and inferior frontal gyri and the left supramarginal gyrus.
Neurological Findings (At the Time of Transfer). The patient was alert, but demonstrated mild right hemiparesis involving the upper and lower extremities (Brunnstorm Stage：upper extremity V, hand/fingers V, lower extremity V), severe sensory disturbance, and mild right facial palsy. His activities of daily living in the hospital were largely independent, as indicated by a Functional Independence Measure motor subscale score of 71/91 and a cognition subscale score of 34/35.
Neuropsychological Findings (At the Time of Transfer). The patient was oriented, and demonstrated appropriate social behavior. He presented with moderate conduction aphasia, characterized by frequent phonemic paraphasias and neologisms. Although he repeatedly attempted self-correction, he was unable to produce the target words. Detailed language dysfunctions are described below. Ideational apraxia was not evident, whereas mild oral apraxia and ideomotor apraxia were observed. His score on the Japanese translation of Raven’s colored progressive matrices (Sugishita et al., 1993) was 31/36 (mean score in the 60-69-year-old group：29.2±5.4) and completion time of Part A of the Japanese version of the Trail Making Test (Brain Function Test Committee, Japan Society for Higher Brain Function, 2019) was 36 seconds without errors, indicating performance within normal limits. Performance Intelligence Quotient on the Japanese version of the WAIS-III (Wechsler Adult Intelligence Scale Third Edition)(Japanese WAIS-III publication committee, 2006) was 76 (Perceptual Organization Index 87, Processing Speed Index 60), and his Visual Memory Index on the Japanese version of the Wechsler Memory Scale-Revised (WMS-R) (Sugishita, 2001) was 79.
The profile of the Standard Language Test of Aphasia (SLTA)(Brain Function Test Committee, Japan Society for Higher Brain Function, 2003), administered on post-stroke days 14-16 at the time of transfer to our hospital is shown in Figure 2, and the results of detailed assessments are presented in Table 1.
（Brain Function Test Committee, Japan Society for Higher Brain Function. Standard Language Test of Aphasia (SLTA). 2nd rev ed. Tokyo：Shinkoh Igaku Shuppansha；2003．https://www.higherbrain.or.jp/publication/test/slta/）
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/46_151_te1.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
＊1 A task was conducted using pictures from the SLTA object naming task (20 items) to assess whether the patient could retrieve the correct number of morae of the target word, by asking the patient to state the mora count of each word.
＊2 Rhyme judgment task：A task was conducted to assess whether the mora of the target word could be appropriately retrieved, by asking the patient to judge whether the final syllable of two words was identical (i.e., whether they rhymed)(Miyazaki, 2018).
＊3 Mean correct responses were obtained from each test manual.
＊4 The number in parentheses indicates the number of responses in which a phonemic paraphasia was initially produced but a correct response was subsequently achieved through self-correction.
Auditory and written comprehension were possible at the word-to-simple-short-sentence level. He demonstrated difficulty understanding complex content, but showed no substantial difficulty in daily conversation. His speech was fluent；however, frequent phonemic paraphasias and phonemic errors were observed in naming, repetition, and oral reading. In the word repetition subtest of the SLTA, phonemic paraphasia occurred in 4 out of 10 words (e.g.,［megane] “eyeglasses”→［mekine],［tamago] “egg”→［temago]). He also exhibited phonemic paraphasia and neologisms in sentence repetition (e.g.,［watashi no ie ni ookina kozutsumi ga todoita] “A large package arrived at my house”→［watashi no ie nu ookina komutsumi ga to todoita]) and in oral reading of short sentences (e.g.,［otoko no ko ga basu ni noru] “A boy gets on a bus”→［otoko no toshi kafui … tantsu …notteru]), with no correct responses (sentence repetition：0/5, oral reading of sentence：0/5). Although he repeatedly attempted to self-correct these phonological errors, most corrections were unsuccessful.
Object naming on the SLTA and Supplementary Tests for Standard Language Test of Aphasia (SLTA-ST)(Brain Function Test Committee, Japan Society for Higher Brain Function, 2011) indicated moderate impairment. On the SLTA, phonemic paraphasia occurred in 11 words (e.g., [shinbun] “newspaper”→［kin kinbunshi],［kusuri] “medicine”→［kisu, kushiri…kusurin]), and only one of these responses was successfully self-corrected, resulting in a score of 11/20. On the SLTA-ST, 31 words were produced with phonemic paraphasias (e.g.,［kagami] “mirror”→［kama kama…kaama…kamami],［shoukaki] “fire extinguisher”→［see…sho shoukachi]), of which 12 were self-corrected, resulting in a score of 56/80 (high frequency words 41/55, low frequency words 15/25).
In a mora-counting task using the 20 picture stimuli from the SLTA object naming subtest, he answered 19/20 items correctly. He also correctly identified the number of morae in words that he failed to produce during picture naming.
His score for oral reading of the basic kana characters (hereafter referred to as seion) was 40/45, but responses often required repeated self-correction and were therefore prolonged. His score was 64/90 on the OR 35 oral reading of wordsII；the stimuli consists of 30 words each for hiragana, katakana, and kanji. The 30 words comprised 10 two-morae, three-morae, and four-morae words in Sophia Analysis of Language in Aphasia (Fujibayashi et al., 2004), which was below the mean score of patients with aphasia. Phonemic errors in the reading task were observed in 24 words, 6 of which were self-corrected. His score on the OR37 Oral reading of nonwords (14 nonwords each of 2 to 5 morae) was 9/56, including self-corrected responses, which was markedly lower than the mean score of patients with aphasia. Phonemic errors were observed across all script types, including kanji, katakana, and hiragana, and were accompanied by a word-length effect.
In spontaneous kanji writing, formal paragraphia (substitution of visually similar kanji characters), distorted character formation, and stroke omissions were observed. In spontaneous kana writing, he had difficulty retrieving characters. Similar errors were observed in dictation of kanji words. In dictation of kana words, character retrieval difficulty was accompanied by incomplete strokes and nonexistent characters. In dictation of kana letters (seion), he scored 23/45, with impaired character retrieval and literal paragraphia.
Although to the patient correctly read aloud 40 of 45 single kana characters (Table 1), most correct responses were achieved only after repeated self-correction, and response times were prolonged even when the correct response was eventually produced. Thus, the difficulty in reading kana characters made kana-reading and ruby annotation ineffective as treatments for phonological deficits. In addition, although he could read kanji words aloud, he had difficulty extracting a specific syllable from the syllable sequence of a word and producing it in isolation. For example, although he could read the kanji word “雨”([ame], “rain”) aloud, extracting its first syllable was difficult. The training procedure is shown in Figure 3.
※Tasks included oral reading of kana and providing kana readings to kanji (ruby annotation) using the gojūon chart. Tasks were initially performed in gojūon order；from approximately post-stroke day 26, they were administered in a randomized order. In the oral reading task, stimulus presentation was gradually shifted from combined kanji-kana stimuli to kana-only stimuli.
4.1. “Construction of a gojūon chart” （Post-Stroke Days 18-19）
We focused on his relatively higher accuracy in reading kanji characters with a single phonetic value (kanji characters with a single-syllable reading such as “田” [ta]), and used such characters as phonemic cues. Based on the methodology of Miyazaki et al.（Miyazaki et al., 2010), we selected kanji characters with a single phonetic value for each syllable in the gojūon chart, and placed them alongside the corresponding kana characters (hereafter referred to as the gojūon chart)(Figure 4). Only kanji characters that the patient could read aloud consistently were selected, when oral reading responses were unstable because of hesitation or response variability, multiple kanji characters were written（for a single syllable for the kana character “い” [i], multiple kanji characters such as “井” and “胃” were presented.)(Figure 4). The gojūon chart included only seion and training was not conducted for dakuon (kana with diacritical marks indicating voicing), hatsuon (the moraic nasal /N/), or sokuon (the geminate consonant marker).
Kanji characters that the patient could read aloud stably were selected. In some cases, multiple kanji were assigned to a single kana character (e.g., “井” and “胃” for “い” [i]), or a kanji with a two-mora reading was selected (e.g., “桃” [momo］for “も” [mo]).
4.2. “Training Period Using the Gojūon Chart” （Post-Stroke Days 20-48）
We repeatedly administered oral reading and ruby annotation tasks using the gojūon chart to establish stable correspondence between kana and morae. In the oral reading task, kanji characters with a single phonetic value and the corresponding kana characters were presented simultaneously, and the patient was asked to read them aloud. In the ruby annotation task, after reading aloud a single kanji character from the gojūon chart, the patient was asked to write its reading in kana. Initially, both tasks were conducted following the gojūon chart order；from post-stroke day 26 onward, the stimuli were presented in randomized order. In addition, the oral reading task was changed from simultaneous kanji and kana presentation to kana-only. When kana reading was difficult, the patient was encouraged to refer to these corresponding kanji character in the gojūon chart. He carried the gojūon chart at all times outside speech-language therapy sessions and repeatedly reviewed it as self-training. In parallel with these tasks, we conducted oral reading of kanji words (2-4 morae), a kana selection and sequencing task, and blending and mora-extraction tasks. In the kana selection and sequencing task, the patient was asked to select all kana characters constituting a word from among the presented options and copy them. In the mora-blending task, the patient blended multiple morae presented auditorily one by one, and chose a corresponding kanji word from some options (Miyazaki, 2018). For example, the morae [to],［ke], and [i] were auditory presented one by one and the patient selected “時計” ([tokei], “clock”) from the kanji word options in front of him. In the mora-extraction task, a kanji word and circles representing its constituent morae were visually presented, and the patient was asked to produce the mora corresponding to the indicated position. For example, when the first circle of “時計 ([tokei]) / 〇〇〇” was indicated, the patient was asked to produce [to]. In every task, when reading a single kana character was difficult, he was advised to refer to the gojūon chart.
After approximately one month of training, the number of correct responses increased from 40/45 to 44/45 in oral reading of single kana characters, and from 23/45 to 42/45 in dictation of kana letters. In addition to the increased number of correct responses, the frequency of self-corrections decreased, and the patient was generally able to respond immediately. Accordingly, we discontinued tasks using the gojūon chart.
4.3. “Phonological Training Period Using Kana Characters” （Post-Stroke Days 49-69）As described above, the patient became able to read and write single kana characters, which enabled intensive phonological treatment to facilitate word-level production using kana reading and writing. In kana-reading training, we used ruby-annotated yojijukugo (four-character-kanji idioms), words atypically written in kana, and kana words including dakuon, hatsuon, or sokuon, and gradually increased the number of morae in the words. Phonemic errors could be corrected by checking and reading each kana character aloud one by one.
From post-stroke day 62 onward, the task was advanced to sentence-level oral reading. The patient also performed ruby annotation for kanji words consisting of two to three kanji characters as a self-directed task.
4.4. At Discharge （Post-Stroke Days 70-71）Language assessment results at discharge are shown in Figure 2 and Table 1. The SLTA score showed overall improvement. In object naming on the SLTA and SLTA-ST, four words were initially produced with phonemic paraphasias；however, after self-correction, all items were answered correctly. Oral reading of single kana characters was 45/45, with immediate responses；SALA OR35 oral reading of wordsII was 90/90；OR37 oral reading of nonwords was 49/56；and kana letter dictation was 45/45. Phonemic errors in overall oral output decreased, and communication in daily conversation became smooth. He was discharged home on post-stroke day 72.
Frequent phonemic paraphasias were observed. Although the patient noticed these errors, he was unable to correct them succesfully. Although phonemic errors prevented him from reaching the target word, he was able to generate the correct number of morae, suggesting impairment in the speech production processes of mora extraction and sequencing. Furthermore, because oral reading of nonwords and single kana characters was difficult, non-lexical kana-to-phoneme conversion was considered impaired.
Table 2 and Figure 5 summarize the scores for oral reading and kana letter dictation, as well as the training content, in the present case and in previously reported cases of conduction aphasia in which phonological impairment improved through kana-based language therapy, such as kana writing, ruby annotation, and oral reading of kana, and for which SLTA profiles were available. The previously reported cases comprised four patients with different clinical profiles：a case of mild conduction aphasia (Imamura, 1994), a case that evolved from Wernicke’s aphasia characterized predominantly by neologistic jargon to conduction aphasia (Funayama et al., 2010), and two cases in which the impairment was largely confined to phonological processing from the early stage after onset (Ota et al., 1998). Despite these varied clinical presentations, oral reading and dictation of kana letters were preserved to some extent across all cases, and the training content focused on kana-based phonological training from the early stage of intervention. In contrast, the present patient showed insufficient performance in oral reading and dictation of kana letters, making kana-based language training difficult. The SLTA scores for oral reading and dictation of kana letters were lower in the present case than in previously reported cases (Figure 5). We interpreted this finding as indicating bidirectional impairment in kana-to-mora and mora-to-kana conversion. This bidirectional impairment was considered the reason why kana-based phonological language training, as used in previously reported cases, was not effective in the present patient.
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/46_151_te2.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
＊The severity of oral reading and dictation of kana letters was determined with reference to profile B of the Standard Language Test of Aphasia (SLTA).
The present case showed lower performance in both oral reading and kana letter dictation compared with previously reported cases.
5.2. Treatment Course
As described above, the present patient exhibited impaired kana-to-mora and mora-to-kana conversion；therefore, phonological training using oral reading of kana was considered inappropriate. Because oral reading of kanji with a single phonetic value was relatively preserved compared with oral reading of single kana characters, we aimed to establish kana-phoneme correspondence through kanji reading. Kanji characters corresponding to each kana in the gojūon chart were selected and written alongside the kana（Figure 4), and the chart was used for oral reading and ruby annotation tasks. By reading the accessible kanji written alongside the kana and learning the paired association between each kanji’s mora and the corresponding kana, kana-mora correspondence was established. Through readable kanji, kana-to-mora conversion was repeatedly trained using the oral reading task, and mora-to-kana conversion was repeatedly trained using the ruby annotation task. As a result, the patient became able to read and write single kana characters (seion) immediately, allowing training to progress to kana-based phonological training, such as oral reading. Kana-mora correspondence was established early, likely because the use of readable kanji with single-mora readings enabled spontaneous production of the target sound and because frequent self-training was possible outside speech-language therapy sessions.
Because the kanji-annotated gojūon chart (Figure 4) included only seion, training initially focused on establishing kana-to-mora and mora-to-kana conversion (i.e., kana-mora correspondence) for seion. Oral reading of words containing dakuon, hatsuon, or sokuon was introduced during the “Phonological Training Period Using Kana Characters” (post-stroke days 49-69), after oral reading of single seion kana characters had been established. Although no kanji-based training was conducted for dakuon, hatsuon, or sokuon, the patient became able to read them aloud without specific training once kana-mora correspondence for seion had been established to some extent. This finding suggests that the patient did not establish phoneme conversion through kanji per se；rather, kanji appeared to compensate for the impaired conversion process between kana characters and morae. This processing mechanism is discussed below.
5.3. The Processing Mechanism of Kanji Oral ReadingTwo routes are assumed to underlie oral reading：lexical and non-lexical processing For kana, in which letter-sound correspondence is regular, nonexistent kana letter strings are considered to be processed via the non-lexical route (kana-to-mora conversion)(Fushimi et al., 2000). In contrast, kanji originated as logograms and convey meaning at the level of a single character；thus, lexical processing is generally assumed. However, some kanji characters lack a kun-reading (native Japanese pronunciation), are unfamiliar as individual characters or character strings, or do not form lexical items independently and are used only as components of compound words. Among these, some kanji are used as phonograms (ateji) by applying their on-reading (Chinese-derived pronunciation), as in “市俄古”([shikago], the American city of Chicago). In such cases, oral reading of kanji can be interpreted as involving the non-lexical route.
As with kana, oral reading of kanji in the present patient showed a word-length effect and was difficult；however, some kanji with single-mora readings could be read stably. The kanji that the patient could read aloud included both characters that form lexical items as single characters (e.g., “絵” “picture”, “手” “hand”) and characters that are typically used as components of compound words rather than as independent lexical items (e.g., “宇”, “佐”). Therefore, success or failure in oral reading could not be attributed solely to differences between lexical and non-lexical processing. Rather successful reading was presumed to reflect the fact that single-mora readings could draw on both lexical and non-lexical processing, and did not require extraction of a specific mora from a sequence of multiple sounds. In any case, although the patient had difficulty reading kana nonwords and single kana characters, he was considered able to read some kanji with a single phonetic value by using the non-lexical processing route for kanji.
The present approach, which used kanji to establish kana-mora correspondence, represents a strategy specific to Japanese speakers, who use two distinct scripts in combination―kana and kanji―with kana itself comprising two phonographic scripts (hiragana and katakana) that differ slightly in form but in which each character corresponds to a single sound. This approach leverages characteristic features of the Japanese language in a manner analogous to the use of grammatical gender as a phonological cue in languages in which gender is assigned according to phonological criteria. Therefore, no similar previously reported approaches were identified in other languages. However, analogous approaches may be possible in speakers in multilingual settings by leveraging scripts from a language other than the acquired spoken language, or by using Latin script in addition to the native script, to establish correspondence with the phonology of the native language.
However, although the patient became able to read aloud and write single seion kana characters after training with the gojūon chart, the intervention was conducted relatively early, around post-stroke day 50, and the possible contribution of spontaneous recovery cannot be excluded. As this is a single-case report, the effect of the present training cannot be demonstrated definitively.
The present patient, who presented with phonological impairment, showed poor kana-to-mora and mora-to-kana conversion, which made language training using kana characters difficult. We therefore created a gojūon chart composed of readable kanji and used kanji to compensate for impaired processing in single-kana-character reading. This approach enabled the patient to read kana characters aloud, and kana-mora correspondence was interpreted as having been reestablished.
The main content of this paper was presented at the 48th Annual Meeting of the Japan Society for Higher Brain Dysfunction (2024), and this manuscript was submitted at the recommendation of the session chair.
Written informed consent was obtained from the patient for presentation at academic conferences and publication, with anonymity of personal information ensured.
The authors declare no conflicts of interest, including those involving any company, organization, or group, in relation to this paper.
This study received no specific funding.
Drafting of the Manuscript：Yuki Takeda
Figure and Table Preparation：Yuki Takeda
Critical Revision of the Manuscript：Yasuhiro Miyazaki, Hidetoshi Takahashi
Supervision：Hiroshi Kigawa
The English version of this article has been translated from the original Japanese manuscript, and is based on an English draft prepared by the authors or an individual with expertise in the field. Preparation of the English draft may have included AI-assisted translation. The manuscript then underwent professional English language editing by Editage. The final English version has been reviewed and approved by the authors, who accept complete responsibility for the content and accuracy of both the Japanese and English versions.
日本語は漢字と仮名の異なる特徴の文字形態を使用する，世界でも稀有な言語である。漢字はアルファベットなどの表音文字と異なり，言語の意味的な最小要素である形態素を表す表語文字である。一方，仮名は表音文字であるが，音素を表すアルファベットとは異なり，規則的に音節（モーラ）と1対1で対応する音節文字である。これらの文字の歴史的な背景は，象形文字から派生した漢字が中国大陸より伝来した。この漢字を中国語の意味のまま用いて，その意味に該当する日本語の音声言語（和語）を割り当てた。その後，中国語の文章（漢文）を日本語で読むために，日本語の音節に対応した独自の文字（従来の中国語の音に対応する漢字を崩した文字）である仮名を誕生させた。この仮名は漢文を日本語で読むために付加する文字として誕生したカタカナと，私的な手紙や日記など（公的な文書は漢字）に用いる，書きやすさを求めた草書体のひらがなの2種類に派生した。この2種類の仮名文字は，現在ではひらがなが主に和語や助詞・活用語尾に，カタカナは外来語・擬音語などの表記に使い分けられている。ひらがなとカタカナはともに音節文字であるため，いずれの表記であっても音読は可能であるが，馴染みのある文字種，適切な表記（表記妥当性が高い）の方が音読の処理速度が速いとされる（髙山ら 2017）。また漢字と仮名は，日本語話者の失語症者の読解や音読において乖離し（Iwata 1984，Sasanuma 1986，Kawahataら 1988），脳イメージング研究にて脳内の処理過程が異なることが示されている（Sakuraiら 2000，Inoら 2009，Horieら 2012）。この文字特性，脳内の処理過程の相違から，失語症者の言語治療では漢字・仮名のそれぞれの特徴を活用する。特に仮名は1文字と1音節が例外なく規則的に対応する音節文字であるから，文字表記により音節数や音韻の構成などの情報が明確となる。
伝導失語では音韻性錯語や音韻性錯書が出現し，これは音韻の選択・配列の言語処理過程の障害が起因するとされる（志塚ら 2002，相馬 1993）。この音韻の障害に対する一般的な言語訓練として仮名の音読や仮名振りが有効とされている（種村 2006）。これは先に述べたように，仮名文字が原則的に1文字と1音節が対応する音節文字である特徴から，単語の音韻列が不安定な音韻障害に対して仮名文字の活用で音韻が明確になるためである。伝導失語例の報告では，仮名単語の音読，漢字単語の仮名振り（今村 1994）や復唱に加えて仮名単語の音読（宇野ら 1997），仮名文字をいくつかの選択肢から選択・配列し，絵に対応した単語を作成したのち，音読・呼称する方法（田中 2006）など，音読・書字による仮名文字の活用により音韻処理過程に働きかけることで音韻性錯語の減少が認められている。伝導失語例の音韻障害に対する言語訓練においても仮名文字の活用が効果的であるが，今村が自身の症例について「本例は，軽度であるため，頭の中にある仮名文字の辞書の中から想起して選択，配列が可能であった。」と述べているように，仮名1文字の音読や書取が一定水準は保持されていることが前提となる（今村 1994）。
今回，我々が経験した伝導失語例は仮名1文字の音読や書取が困難で，音韻障害に対する言語訓練として仮名文字を効果的に用いることができなかった。ただ仮名1文字の音読が困難な一方で，1音価の漢字（1音節の読みの漢字）の読みが一部，保持されていたことが特徴であった。そこで仮名1文字に対応する音節の読みの漢字を割り当てた50音表を作成し，漢字を介して仮名と音韻の対応の定着を促した。この結果，仮名1文字の音読・書取が可能となり，仮名文字を活用した音韻障害に対する言語訓練が実施できるようになった。その後，発話全般において音韻性錯語が減少したので，その経過を報告する。
【症例】64歳，男性，右利き。教育歴は12年。
【既往歴】2型糖尿病，高血圧，高脂血症。
【現病歴】言葉の出づらさと右上下肢の動かしにくさを主訴に，かかりつけ医を受診した。出血性梗塞の診断にて即日入院となった。右上下肢の感覚障害と失語症が残存し，リハビリテーション目的で第13病日に当院へ転院となった。
【神経放射線学的所見】当院転院時の頭部CT（第13病日）を図1に示す。左中・下前頭回後部と左縁上回の皮質および皮質下白質に出血性変化を伴う梗塞巣を認めた。
左中・下前頭回後部と左縁上回の皮質および皮質下白質に出血性変化を伴う梗塞巣を認めた。
【神経学的所見（当院転院時）】意識は清明で，右上下肢に軽度の運動麻痺（Brunnstrom stage：上肢V，手指V，下肢V）と重度の感覚障害，軽度の右顔面麻痺を認めた。FIM（Functional Independence Measure）の運動項目は71/91点，認知項目は34/35点で病棟内ADLはおおむね自立していた。
【神経心理学的所見（当院転院時）】見当識は正常で礼節は保たれていた。中等度の伝導失語を呈しており，音韻性錯語が頻発し，一部は新造語の出現を認めた。これらの誤りに対して，自己修正をくり返すものの，適切な単語の表出には至らなかった。言語障害の詳細は後述する。観念失行は明らかでなかったが，軽度の口腔顔面失行，観念運動失行を認めた。レーヴン色彩マトリックス検査（杉下ら 1993）は31/36点（60～69歳の平均29.2±5.4点），Trail Making Test日本版（日本高次脳機能学会Brain Function Test委員会 2019）のPartAは所要時間36秒，誤反応0回で正常判定，日本版WAIS-III成人知能検査（Wechsler Adult Intelligence Scale Third Edition）（日本版WAIS-III刊行委員会 2006）の動作性知能指数は76（知覚統合87，処理速度60），日本版ウエクスラー記憶検査（Wechsler Memory Scale-Revised：WMS-R）（杉下 2001）の視覚性記憶は79であった。
当院入院時（第14～16病日）の標準失語症検査（Standard Language Test of Aphasia：SLTA）（日本高次脳機能学会Brain Function Test委員会 2003）のプロフィールを図2に，掘り下げ検査の結果を表1に示した。
（日本高次脳機能学会Brain Function Test委員会．標準失語症検査．改訂第2版．東京：新興医学出版社；2003．https://www.higherbrain.or.jp/publication/test/slta/）
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/46_151_tj1.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
※1 目標語のモーラ数が想起可能であるかを評価する目的で，SLTAの呼称（20語）の絵を用いて当該語のモーラ数を問う課題を実施した。
※2 押韻判断課題：目標語の音韻が適切に想起可能であるかを評価する目的で，2単語の最終音節の音が同一か（韻を踏んでいるか）の判断を求める課題を実施した。（宮﨑 2018）
※3 平均正答数は各検査マニュアルによる（日本高次脳機能学会Brain Function Test委員会 2003，日本高次脳機能学会Brain Function Test委員会 2011，藤林ら 2004）。
※4 括弧内は音韻性錯語の出現後，自己修正で正答に至った数を示す。
言語理解は音声，文字とも単語や簡単な短文の課題は可能であった。複雑な内容の理解で低下を認めたが，日常会話場面におけるやり取りでは大きな支障がなかった。発話は流暢だが，呼称，復唱，音読のいずれでも音韻性錯語・錯読が頻発した。SLTAの単語の復唱で，眼鏡→“めきね”，卵→“てまご”と10語中4語で音韻性錯語を認めたほか，文の復唱で，私の家に大きな小包が届いた→“私のいえぬ，大きなこむつみが，と，届いた”との反応や，短文の音読で，男の子がバスに乗る→“男のとし，かふい・・たんつ・・乗ってる”と，音韻性錯語，新造語が出現し，正答には至らなかった（文の復唱：0/5，短文の音読：0/5）。これらの音の誤りに対しては自己修正を繰り返すものの，大半は適切に修正できなかった。
呼称はSLTA，標準失語症検査補助テスト（Supplementary Tests for Standard Language Test of Aphasia：SLTA-ST）（日本高次脳機能学会Brain Function Test委員会 2011）とも中等度の低下であった。SLTAでは，新聞→“きん，きんぶんし”や薬→“きす，くしり・・くすりん”のような音韻性錯語が11語にみられ，そのうち自己修正で正答したのは1語のみで，正答数は11/20であった。SLTA-STでは，鏡→“かま，かま・・かーま・・かまみ”，消火器→“せー・・しょ，しょうかち”のような音韻性錯語が31語にみられ，そのうち自己修正で正答に至ったのは12語で，正答数は56/80（高頻度語41/55，低頻度語15/25）であった。
モーラ分解は，SLTAの呼称（20語）の絵を用いて当該語のモーラ数を問う課題で，結果は19/20の正答であった。呼称で不正答の語彙もモーラ数は正答した。
音読は仮名1文字音読（清音）が40/45の正答であったが，大半が自己修正を繰り返し正答の反応は遅延した。SALA失語症検査（Sophia Analysis of Language in Aphasia）（藤林ら 2004）のOR（oral reading）35単語の音読II（「ひらがな」「カタカナ」「漢字」が各30語でそれぞれ2モーラ語，3モーラ語，4モーラ語が10語ずつから成る）が64/90の正答で，失語症者の平均正答数を下回っていた。音韻性錯読は24語にみられ，そのうち自己修正で正答に至ったのは6語であった。OR37無意味語の音読（2，3，4，5モーラの無意味語が各14語から成る）は自己修正の正答を含めて9/56で失語症者の平均正答数と比べ明らかな低下を認めた。漢字，カタカナ，ひらがなのいずれの文字種も変わらず語長効果を伴う音の誤りは出現した。
書字は漢字単語の自発書字で形態性錯書や字形の崩れ，字画の脱落がみられ，仮名単語の自発書字で文字想起が困難であった。漢字単語の書取は自発書字と同様の誤りにて困難であり，仮名単語の書取は文字想起ができない他に線分の途切れや非実在字の誤りで困難であった。仮名1文字の書取（清音）は23/45の正答で，文字の想起困難や字性錯書の誤りを認めた。
仮名1文字音読は40/45の正答（表1）であったが，誤りに対する修正を繰り返すことが大半を占め，最終的に正答に至っても反応が遅延していた。そのため，仮名単語の音読が困難であり，音韻障害に対する言語訓練として仮名の音読や仮名振りの課題は効果的に実施できなかった。また漢字単語の音読は可能であっても，単語を構成する音節列のうち特定の1音節を抽出できず，特定の音節を産生させることは困難であった。例えば，「雨」の音読が可能であっても「雨」の1音節目の抽出を促すと，産生することができなかった。本患者の訓練経過を図3に示した。
※50音表を利用した仮名音読と漢字の仮名振り課題。両課題とも当初は50音順に，第26病日頃よりランダムに実施した。また，音読課題は漢字・仮名の呈示から仮名のみの呈示へと移行した。
1． 「漢字で構成された50音表の作成」（第18～19病日）
仮名1文字の音読に比べ1音価の漢字（「田（た）」のような1音節の読みの漢字）の音読が安定して可能なことに着目し，1音価の漢字を音韻の手掛かりとして用いることとした。宮﨑らの方法（宮﨑ら 2010）を参考に，50音の各音節に対応した1音価の漢字を選出し，仮名50音表に併記した（以下，50音表）（図4）。なお，選出された漢字は本患者が安定して音読可能な文字とし，音読の反応によどみ，浮動性があるような不安定な場合は複数の漢字を選定した（い：井・胃など）（図4）。また，50音表は清音のみで構成し，濁音，撥音，促音については訓練を実施しなかった。
漢字は，本患者が安定して音読可能な文字を選出し，仮名1文字に対し，複数の漢字（い：井，胃）や2音節から成る漢字（も：桃）を選定する場合もあった。
2． 「50音表活用期」（第20～48病日）
50音表を利用した音読・仮名振りの2課題を反復的に実施し，仮名と音韻の対応の定着を図った。音読課題は1音価の漢字と対応する仮名文字を同時に呈示し，音読を求めた。仮名振り課題は50音表の漢字1文字を呈示し，音読後にその読みを仮名文字で書字するよう求めた。両課題とも当初は50音順に，第26病日頃よりランダムに実施した。また，音読課題は漢字と仮名の同時呈示から仮名のみの呈示へと移行した。仮名のみで音読が困難な際は，50音表にて該当する漢字を確認するよう促した。また，患者は言語訓練の時間以外にも，50音表を常時携帯して自主的に繰り返し確認していた。また上記の課題に並行して，漢字単語の音読（2～4モーラ），仮名の選択・配列課題と音の合成・抽出課題を実施した。仮名の選択・配列課題は漢字単語の音読後に，複数呈示した仮名文字から単語を構成する仮名文字をすべて選択し写字するよう求めた。音の合成課題は1音ずつ提示した複数音を合成し，複数の漢字単語から該当するものを選択する課題（宮﨑 2018）である。例えば，「と」「け」「い」の音を1つずつ聴覚提示し，眼前の漢字単語の選択肢から「時計」を選ぶ。また，音の抽出課題は単語の漢字とモーラを丸で示したものを視覚的に呈示して，指定した丸の位置の音を表出してもらう課題である。例えば，「時計 / 〇〇〇」で1番目の○を指定したら「と」と答える。いずれの課題も仮名1文字の音読が困難な場合には（手元の）50音表を参照するよう指導した。
約1ヵ月の訓練期間で，正答数は仮名1文字の音読が40/45から44/45，仮名1文字の書取は23/45から42/45となった。正答数の向上に加え，誤りに対する自己修正が減り，おおむね即答できるようになった。そのため，50音表を活用したこれらの課題は終了とした。
3． 「仮名文字による音韻訓練期」（第49～69病日）上記の通り，仮名1文字の音読，書取は可能となり，仮名文字の音読，書字を利用しての単語レベルの表出を促す音韻訓練を集中的に実施できるようになった。単語の音読課題は，ルビを振った四字熟語や仮名文字の表記妥当性の低い仮名単語，濁音，撥音，促音を含む仮名単語を用い，徐々に目標語のモーラ数を増やして実施した。音の誤りには仮名を1文字ずつ確認，音読することで修正できた。
その後，第62病日以降は文章の音読へと展開した。また2～3文字の漢字単語の仮名振りを自主課題で行った。
4． 「退院時」（第70～71病日）退院時の言語機能評価を図2，表1に示した。SLTAは全般的に改善を認めた。SLTA，SLTA-STの呼称では音韻性錯語が4語みられたものの，自己修正でき全問正答した。仮名1文字の音読は45/45と即答，SALA OR35単語の音読IIは90/90，OR37無意味語の音読は49/56，仮名1文字の書取は45/45と全問正答した。発話全般における音の誤りは減少し，日常会話のコミュニケーションは円滑になった。第72病日に自宅退院となった。
入院時の本患者は音韻性錯語が頻発し，音の誤りに気づいていても修正には至らなかった。音の誤りにて目標語に至らないものの適切なモーラ数は想起できていたことから，音の抽出・配列の発話処理過程が障害されていたと考えられた。さらに無意味語，仮名1文字の音読は困難であったことから，非語彙処理の仮名から音韻への変換も障害されていたと考えられた。
仮名書字や仮名振り，仮名音読などの仮名文字を利用した言語訓練により音韻障害が改善した伝導失語のうち，SLTAのプロフィールが開示されている既報告例と本患者の仮名1文字の音読・書取の成績と言語訓練の内容を表2・図5に示した。既報告例は，軽度の伝導失語例（今村 1994），新造語ジャルゴン優位のウェルニッケ失語から伝導失語への収束例（船山ら 2010），発症初期から障害が音韻処理過程にほぼ限局していた2例（大田ら 1998）の4症例で障害像は様々であるが，仮名1文字の音読・書取の成績は一定水準，保持されており，言語訓練の内容は介入初期から仮名文字を利用した音韻訓練が中心である。一方，本患者は仮名文字の音読・書取が不十分で仮名文字を活用した言語訓練が困難であった。SLTAにおける仮名1文字の音読・書取の成績は，既報告例と比べ低値であった（図5）。これは本患者が仮名から音韻，音韻から仮名の変換障害を呈していたことを示し，既報告例のような仮名文字を用いた音韻に対する言語訓練が効果的ではなかったことの理由と筆者らは解釈した。
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/46_151_tj2.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
※仮名1文字の音読・書取の重症度はSLTAのプロフィール（B）を参考に判断
本患者の仮名1文字の音読，書取は既報告例よりも低値であった。
2． 訓練経過
本患者は前述の通り，仮名から音韻，音韻から仮名の変換障害を呈していたために，仮名文字の音読を用いた音韻訓練は適応でなかったと考えられた。一方で本患者は仮名1文字の音読に比べ1音価の漢字の音読が保持されており，漢字の音読を介して仮名と音韻の対応の定着を図ることとした。50音と対応する漢字を選出，併記（図4）し，この50音表を用いて音読・仮名振り課題を実施した。併記された音読可能な漢字を読み，その音節に対応する仮名を対連合学習することで，仮名と音韻の対応を定着させた。音読可能な漢字を介して，音読課題により仮名から音韻，仮名振り課題により音韻から仮名の変換を繰り返し実施した。これにより仮名1文字の清音の音読・書取が即答できるようになり，仮名文字の音読などの音韻の訓練へと移行できた。仮名と音韻の対応が早期に定着したのは，音読可能な1音価の漢字を用いることで自発的に目標音が産生できたこと，また言語訓練の時間以外にも頻回な自己学習が可能であったためと思われる。
漢字を併記した50音表（図4）は清音のみのため，まずは清音の仮名から音韻，音韻から仮名の変換（仮名と音韻の対応）の定着を促した。濁音や撥音，促音を含む単語の音読は，清音のみの仮名1文字の音読が定着後の「仮名文字による音韻訓練期（第49～69病日）」に実施した。濁音や撥音，促音については漢字1文字との対応訓練を実施しなかったが，清音における仮名と音韻の対応が一定程度定着したあとに，特別な訓練なしに音読が可能になった。これは本患者が漢字を用いて音韻への変換を確立したのではなく，障害された仮名文字と音韻の変換過程を漢字が補ったと解釈できる。この処理過程に関しては以下に考察する。
3． 漢字音読の処理過程音読の処理過程には語彙処理と非語彙処理が想定されており，特に文字と音の対応が規則的な仮名の場合では実在しない仮名文字列は非語彙経路（仮名から音韻の変換）で処理されると考えられている（伏見ら 2000）。一方，表語文字である漢字は1文字で意味を成し，この場合は語彙処理が想定される。ただ漢字のなかには，訓読みがなかったり，一般的には馴染みのない文字・文字列，1文字では語彙を形成せず，もっぱら熟語の一部として使われる文字がある。例えば「市俄古（アメリカの都市シカゴ）」のように音読みを使って音節文字（当て字）として用いる場合の漢字の音読は非語彙処理されていると解釈できる。
本患者における漢字の音読は仮名文字の音読同様に語長効果を認め困難であったが，一部の1音価の漢字は安定して読むことができた。本患者が音読可能な1音価の漢字は，1文字で語彙として成立する漢字（例：絵，手など）と，熟語の一部として用いられることが多い漢字（例：宇，佐など）があり，語彙処理，非語彙処理の相違が音読の可否と関係しているとは言い難かった。ただ語彙処理と非語彙処理の双方を利用でき，かつ複数の音列から特定の音韻を抽出する必要がない1音価の読みであることが奏功したと推察される。いずれにしても本患者は仮名の非語，1文字の音読が困難であったが，漢字の非語彙処理過程を利用することで，一部の1音価の漢字の読みが可能であったと考えられた。
今回の仮名と音韻を定着させるために漢字が有効であった報告は，仮名と漢字の特徴の異なる2種類の文字（ひらがなとカタカナは形態がわずかに異なるものの各文字が対応する表音文字）を併用する日本語話者における特異的なアプローチであった。これは音韻的な基準によって性を付与する言語における性を手掛かりとした音韻の促進など，各言語の特徴を活用したものである。そのため，他言語における類似する既報告は見当たらなかった。ただし多言語国家の話者であれば習得された音声言語とは別の言語の文字や，母国語の文字に加えてラテン文字が使用されていることを活用して母国語の音韻に対応させるアプローチが可能かもしれない。
ただし，本患者は50音表を活用した訓練実施後に仮名1文字の清音の音読・書取が可能となったが，訓練時期が発症50病日前後と，比較的急性期に近い時期であったことから，改善には自然回復による影響も否定できない。単一の症例報告であることから，今回の訓練の効果を明確に示すには限界がある。
音韻障害を呈した本患者は，仮名から音韻，音韻から仮名への変換が不良で，仮名文字を利用した言語訓練が困難であった。そこで音読が可能な漢字で構成した50音表を作成し，漢字を介して仮名1文字の読みの処理を補った。これより仮名文字の音読が可能となり，仮名と音韻の対応が再構築されたと解釈できた。
本論文の要旨は，第48回日本高次脳機能学会総会（2024）で発表したもので，座長推薦により論文投稿するものである。
個人情報の十分な匿名性確保を前提に，患者本人から専門の学術大会および論文での発表の承諾を得た。
著者全員に本論文に関連し，開示すべき利益相反状態にある企業，組織，団体はいずれもありません。
本研究に対する特別な資金提供はありません。
論文執筆（初稿）：武田有希
図表の作成：武田有希
論文校閲・編集：宮﨑泰広，高橋秀寿
監修：木川浩志
