Brain (1987), 110, 1393-1417 ELECTROPALATOGRAPHIC ANALYSIS OF APRAXIA OF SPEECH IN A LEFT HANDER AND IN A RIGHT HANDER (From the Department of Rehabilitation, Tokyo Metropolitan Institute/or Neurosciences, Fuchu, Tokyo, Mihara Memorial Hospital, hesaki, Gunma, and the Department of Neurology, Chiba University, Chiba, Japan) SUMMARY Two cases with 'pure' apraxia of speech are reported. The articulatory disturbances were quite similar. One of the two cases was a left-handed male with a subcortical haemorrhage and the other a right-handed male with a cerebral infarct. The MRI and CT scans showed that the first case had a lesion that mainly involved the right precentral gyms and its deep white matter, and that the second had a lesion mainly affecting the lower parts of the left precentral and postcentral gyri and their deep white matter. These findings and a literature review suggest that a corticosubcortical lesion of the lower part of the left precentral gyrus in most right handers and a lesion of the symmetric region in the right hemisphere in some left handers cause apraxia of speech. The omission errors for sounds articulated by the tongue and the hard palate were analysed using electropalatography, which records visually the dynamics of the palatolingual contact. The results demonstrated that there were three kinds of omission errors: true omissions (no palatolingual contact); omissions with incorrect contact (palatolingual contact for a different sound or undifferentiated sound); and omissions with correct contact (correct palatolingual contact for a target sound). The latter two types of omission error were observed for initial consonants and they were probably caused by a delay in air flow. The patients also showed a tendency to substitute one of the two consonants /t, tj/ for other sounds, which suggested that they had difficulty in the inhibition of tongue activity. INTRODUCTION It has been known that a unilateral hemisphere lesion uncommonly causes disturbance limited to speech output involving neither auditory nor reading comprehension nor writing disturbance. The syndrome has been variously termed aphemia (Broca, 1861; Bastian, 1898), pure anarthria (Marie, 1906), pure word dumbness (Wernicke, 1906) or pure motor aphasia (Dejerine, 1914). We will use the term 'aphemia' for brevity. Correspondence to: Dr Morihiro Sugishita, Department of Rehabilitation, Tokyo Metropolitan Institute for Neurosciences, 2-6 Musashidai, Fuchu City, Tokyo, Japan. Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 by MORIHIRO SUGISHITA, KANAE KONNO, SUMIE KABE, KAZUTA YUNOKI, OSAMU TOGASHI and MITSURU KAWAMURA 1394 MORIHIRO SUGISHITA AND OTHERS Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 Aphemia represents at least two types of disturbance. Some patients with aphemia show an inability to articulate all except a very limited set of phonemes or words (Dufour, 1895; Ladame and von Monakow, 1908; Souques, 1928), while others are able to utter most phonemes but show many articulatory errors such as substitutions, additions, omissions and distortions (Denny-Brown, 1965; Lecours and Lhermitte, 1976; Tonkonogy and Goodglass, 1981). These errors are characterized by inconsistency in two ways: phonemes are articulated incorrectly at one time but correctly at another; and errors are not always the same. The second type of disturbance has been reported after recovery from the first (Denny-Brown, 1965; Lecours and Lhermitte, 1976) but not in others (Dufour, 1895; Souques, 1928). F. L. Darley (unpublished) designated the second type 'apraxia of speech'. These two types of disturbance are different from dysarthria arising from bilateral damage of the corticobulbar tracts, the cerebellum or the lower motor neuron whose speech has normal length and where errors tend to be consistent. Although the articulatory characteristics of apraxia of speech associated with aphasia have been thoroughly studied by systematic phonetic analysis (Shankweiler and Harris, 1966; Johns and Darley, 1970; Trost and Canter, 1974), those of'pure' apraxia of speech have not been well studied. Even in the well described case with pure apraxia of speech (Lecours and Lhermitte, 1976), the size of the speech sample was not large enough for systematic phonetic analysis. In the present paper, 2 cases of aphemia (pure apraxia of speech) will be reported. They uttered a very limited number of phonemes at an early stage. At a later stage, they began to produce most phonemes but with numerous articulatory errors. Thus their disturbance at its later stage is well covered by the term 'apraxia of speech'. The utterances of the 2 cases were studied systematically by auditory phonetic analysis to elucidate the articulatory characteristics of apraxia of speech. Among the various kinds of articulatory errors in apraxia of speech, substitution errors have attracted most attention in the literature. Omission errors, however, are also supposed to characterize apraxia of speech. Electropalatography was employed to observe directly the palatolingual contact during omission errors of our 2 cases in order to clarify the characteristics of this type of error. The location of the lesion responsible for aphemia or apraxia of speech is still uncertain. Lecours and Lhermitte (1976) have suggested that, in the right-handed, a lesion of the lower part of the left precentral gyms causes the syndrome. However, a number of cases are needed to confirm the site of the lesion responsible for the syndrome in the right-handed. In the left-handed, so far as we know, no study has established the site of the responsible lesion. One of the 2 cases in this study was left-handed, and the other right-handed. Magnetic resonance imaging (MRI) and computerized tomography (CT) were administered to localize the site of their lesions. APRAXIA OF SPEECH 1395 CASE HISTORIES Case 1 (K. O.) Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 A 27-yr-old left-handed man, a high school graduate, without prior medical problems, had a subcortical haemorrhage in the right frontal lobe on June 2, 1984. The patient was left-handed from infancy. The left hand was used for all activities except writing and eating with chopsticks. As for these two activities, he had used his left hand at first, but was forced to change to his right hand in childhood. He was left-handed except for writing on the Oldfield Handedness Battery (Oldfield, 1971). His parents and 3 siblings were all right-handed. Neurological examination on admission (June 2, 1984) revealed impaired consciousness, a left hemiplegia including the lower face, deviation of the tongue to the left, and a left hemihypaesthesia primarily affecting the left upper limb. Two days later he was alert, but with a mild hemiparesis predominantly in the left upper limb. He uttered only jo/, /N/ and /hai/, the latter two meaning 'yes' in Japanese. Angiography disclosed an arteriovenous malformation, which was fed from the right precentral artery and was drained to the right superficial sylvian vein. The patient was operated upon on June 6, 1984. A craniotomy was performed in the right frontoparietal region. A semiliquid haematoma ( 6 x 5 x 4 c m ) and the nidus were removed. Neurological examination 1 month after onset showed a mild left-sided lower facial weakness, deviation of the tongue to the left and a left arm monoparesis. There were no signs of left hemihypaesthesia. Two months after onset, neurological examination showed a mild left-sided lower facial weakness and a mild paresis of the left arm. Otherwise examination was normal. Tongue deviation was not present. Pure tone threshold audiometry showed normal hearing levels in both ears despite a 40 dB loss in the left ear and a 15dB loss in the right at 8000 Hz. Speech audiometry with monosyllabic speech sounds (67-A list a standardized single-syllable list for speech discrimination test by the Japanese Audiological Society) was within normal limits in both ears. Neurological findings. MRI (Picker International Ltd, VISTR-MR 0.5T) was carried out with a 0.26 Tesla superconducting magnet, using the inversion recovery technique, 10 months after onset. The images (fig. 1) showed an abnormal low intensity region mainly in the right precentral gyrus and its underlying deep white matter such as the middle region of the corona radiata and the right arcuate fasciculus. CT scans parallel to the orbitomeatal line were administered with Toshiba Medical Equipment TCT-60A 1 month and 18 months after onset. They revealed findings compatible with MRI although contaminated with artefacts from the surgical clips (fig. 1). Language testing. For 10 days after onset, his spontaneous speech was limited to /a/, /i/, /ui/, /o/, /N/, /uiN/ and /hai/ the latter three meaning 'yes' in Japanese. The patient could not even say his name. Oral repetition was severely disturbed. He could not repeat very familiar phrases such as /ohajo:/ (good morning), for which he uttered /ohooui/. However, he correctly understood questions about his illness and family. He could also comprehend oral commands, such as 'Close your eyes' and 'Open your mouth' and could carry out these commands. He had no difficulty in writing the names of common objects, although he showed a slight disturbance in writing sentences. The Japanese edition of the Western Aphasia Battery (WAB) (Kertesz, 1982; Sugishita el al., 1986) 0.8 month after onset revealed a marked discrepancy between his speech disturbance and other abilities, namely those of auditory comprehension, reading comprehension and writing (Table 1). In the spontaneous speech task, his disturbance was severe. His production was effortful and slow. Substitution, omission and distortion errors were numerous. He could utter only 9 different phonemes. Among them, 3 different phonemes happened to occur twice in his answers. These were articulated correctly at one time whereas they were misarticulated at another. However, his speech consisted of the correct number of syllables with an approximation of vowels when the target utterances were short. For example, when he was asked 'What is your name?', he uttered '/ooajajiieii/', the vowels of which were all correct, though his utterance included 3 omission errors on consonants. To the question 'What is your occupation?', he responded with /anikjaNN/ for /kaiJaiN/ (businessman). 1396 MORIHIRO SUGISHITA AND OTHERS Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 FIG. 1. A-C, CT scans from Case /, 18 months after onset showing an abnormal area of low density mainly affecting the right precentral gyrus extending deeply into the underlying white matter, although artefacts from the surgical clips make the scans unclear, D-L, MRI from this case, 10 months after onset, revealing an abnormal low intensity area in the right precentral gyrus including the middle region of the right corona radiata and the right arcuate fasciculus. Right central sulcus (large arrows); right precentral sulcus (small arrows). A P R A X I A OF SPEECH 1397 TABLE 1 WESTERN APHASIA BATTERY SUBSCORES (WAB) Case 2 Case 1 Interval postonset tested Subtest category (maximum) 2.8 mo. 1J mo. 9.6 mo. 3 1 9.6 3.6 6" 82 8.5 10 9 9* 10 6.7 6.8** 97 99 10 0 0 9.6 0 0.1 6.9 6 99 8 9* 10 5 6 5** 94 8.7 10 3 0.9 2.4 34 3 0.9 24 3.5 3 0.9 2.4 2.9 3 09 2.4 2.8 •Fluency evaluation in the WAB includes phrase length, melody, syntax, paraphasia and word-finding difficulty The high scores in fluency are due to lack of agrammatism and word-finding difficulty. **The relatively low scores in the naming task are due to substitution errors in sounds His abilities to repeat and to read aloud were also severely impaired. During the repetition task, he uttered /anana/ for /banana/ (banana) and /enwaanatte:a:sui/ for /denwaganatteimasui/ (The telephone is ringing). He had no response (naming difficulty) to 1 out of 20 objects in the naming task and articulatory errors to 19 out of the 20 objects. As illustrated in these examples, vowels were less affected than consonants. The patient showed many articulatory errors, such as substitutions, omissions, additions and distortions of sounds. Self-corTection was seldom observed. On the other hand, auditory comprehension, reading comprehension and writing were intact. Auditory comprehension of words was tested with 76 items, and in this he had a perfect score. With 'Yes-No questions' to assess auditory comprehension of sentences such as 'Do you eat a banana before you peel it?' and 'Is a horse larger than a dog?', he correctly responded to 19 out of 20 questions. Oral sequential commands varied from relatively short and easy commands, such as 'Raise your hand', to rather complicated ones such as 'Put the comb on the other side of the pen and turn over the book.' Ten of these 11 commands were correctly executed. His reading comprehension was also nearly intact. In the task where he read sentences and then chose the missing words or phrases, he correctly answered 7 out of 8 items. His writing ability was also quite well preserved, and he wrote 3 sentences describing 10 objects with only 2 spelling errors when requested to describe a presented picture. The WAB was retested 2.8 months after onset (Table 1). Though his speaking ability improved to the extent that he often uttered short sentences, articulatory errors such as substitution, omission, addition and distortion errors were still numerous. Compared with those of 0.8 month after onset, omission and substitution errors for consonants 2.8 months after onset decreased, while distortion errors for consonants increased. Substitution and distortion errors for vowels decreased. His speech showed prosodic abnormalities, such as the slow speaking rate resulting from a prolongation of sounds and inappropriate intersyllabic pauses, and monotonous pitch. These disturbances were observed not only in spontaneous speech but also in repetition and reading aloud. Syntax was normal. He had no response to 1 out of 20 objects in the naming task and had articulatory errors to 14 of these objects. On the other hand, he had perfect scores for auditory comprehension of sentences and of oral Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 Spontaneous speech Information content (10) Fluency (10) Comprehension (10) RepetiUon(lO) Naming (10) Reading (10) Writing (10) Praxis (10) Construction Drawing (3) Bloclc design (0 9) Calculation (2 4) Raven's score (3 7) OJi mo. 1398 M O R I H I R O SUGISHITA AND O T H E R S Case 2 (T. I.) A 58-yr-old right-handed man, a middle school graduate, suddenly suffered from weakness and numbness on the right side of his body and speech disturbance on January 4, 1984. He was righthanded, as were his parents, 5 siblings and 2 children. He had been hypertensive for 20 yrs. Neurological examination I month after onset showed right-sided lower facial weakness, deviation of the tongue to the right, a right hemiparesis predominantly affecting the right arm, and right hypaesthesia. The plantar response was extensor on the right. Neurological examination was otherwise negative. A gag reflex was present and there was no difficulty in swallowing. The visual fields were normal. The right hypaesthesia cleared by 2 months after onset. Fibrescope examination revealed that movements of the larynx and pharynx were normal. Six months after onset, neurological examination showed mild right-sided lower facial weakness and a mild right hemiparesis predominantly affecting the right arm. The plantar response remained extensor on the right. Sensory function and the visual fields were normal. Deviation of the tongue was no longer present. Pure tone audiometry and speech audiometry with the 67-A list were within the normal range. CT scans 1 month after onset and MRI 1 yr after onset showed an infarct mainly affecting the left precentral and postcentral gyri and their underlying deep white matter, such as the middle region of the corona radiata and the left arcuate fasciculus. The left insula was also involved (fig. 2). Language testing. For 1.3 months after onset, spontaneous speech was effortful and slow and consisted of only I to 3 syllables. It was limited to /a/, /i/, /o/, /N/, /afi/ (foot) and /ohajo:/(good morning). The patient was unable even to say his name. The WAB was carried out 1.3 months after onset (Table 1). The patient was able to utter only 1 or 2 /N/ to each of the 7 questions in the spontaneous speech task. In the naming task, he had no responses to any of the 20 objects. When the initial syllable of the names were given as cues, he said /boi/ for /bo:l/ (ball) and /maio/ for /naiCnu/ (knife). In the repetition task, substitution, omission and distortion errors were observed. For example he said /ahaha/ for /banana/ (banana). He uttered only 6 different phonemes. The phoneme /n/ occurred four times in his utterances and was articulated correctly once and variously misarticulated three times. There were few self-corrections in his utterances. He also showed a severe disturbance for reading aloud. On the other hand, auditory comprehension Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 sequential commands. He had no errors for reading comprehension at the two to four-sentence level or for executing written commands that included complicated commands such as 'Pick up the pencil, knock 3 times and put it back.' His written description for a presented picture was adequate and accurate using 3 sentences within 3 min. In the Token Test 2 months after onset the patient performed perfectly. Since his disturbance was limited to oral expression with auditory and reading comprehension and intact writing ability, he was diagnosed as having aphemia. His disturbance at the 2.8 months stage where most of Japanese phonemes appeared in his utterance constituted a pure form of apraxia of speech. His apraxia of speech was still observed 2 yrs after onset, though it had been moderately alleviated. Praxis and gnosis testing. At 0.8 month after onset, he performed perfectly the drawing, block design and praxis subtests of the WAB. There were no indications of constructional, ideomotor or ideational apraxia. He correctly performed the 5 items of the orofacial region in the praxis subtest and the other oral commands such as 'Swallow', 'Puff out your cheeks' and 'Make your teeth chatter'. In contrast, he had some difficulty in executing the commands 'Cough' and 'Click your tongue', indicating a mild orofacial apraxia. At 2.8 months after onset, he perfectly performed the latter two commands. He did not show unilateral hemineglect in drawing pictures. Other apraxias or agnosias were not observed. Intelligence testing. One month after onset, he attained a verbal IQ of 74 and a performance IQ of 85 on the Wechsler Adult Intelligence Scale. His best performance was on the block design subtest (scaled score: 14). APRAXIA OF SPEECH 1399 Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 FIG. 2. A-D, CT scans from Cos^ 2, I month after onset demonstrating an abnormal low density area in the lower parts of the left precentral and postcentral gyri extending deeply into the middle region of the corona radiata. The left arcuate fasciculus and the left insula were also affected, E-M, MRI from this case 1 year after onset revealing an abnormal low intensity area in almost the same area of the left hemisphere. Left central sulcus (large arrows); left precentral sulcus (small arrows). 1400 M O R I H I R O SUGISHITA AND O T H E R S EXPERIMENTAL INVESTIGATIONS Three articulation tests were employed to obtain speech samples of the 2 patients. Auditory phonetic analysis and electropalatography were used to analyse the speech samples. Articulation Three articulation tests were administered to assess the speech disturbance of the 2 patients. One was the Articulation Test developed by the National Rehabilitation Center for the Disabled (ATNR). The second test, the Articulation Test of the Tokyo Metropolitan Institute for Neurosciences (ATTM), had been used for patients before the ATNR was adopted. Most of the test items of the Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 was well preserved. He responded correctly by gesture to 19 out of 20 'Yes-No' questions and understood and performed correctly all 11 oral sequential commands. In the reading comprehension task, he executed correctly all of the 6 written commands, although he erred on 2 out of the 8 questions which required comprehension of a paragraph. When asked to write a description of a picture, he wrote only 9 nouns with no sentences. His written naming of a stamp, a watch, a pencil and woollen yarn was correct in ideograms (kanji), but showed a mild disturbance in phonograms (kana). However, there was a striking discrepancy between his writing disturbance and that of his speech. In summary, the patient showed severe speech disturbance and mild disturbances in writing and in reading comprehension 1.3 months after onset. The WAB was readministered 9 6 months after onset. He made no errors in reading comprehension task or dictation of words such as stamp and woollen yarn (Table I). In the written description of a presented picture he wrote 2 correct sentences having a total of 22 words in 3 min with 2 minor mistakes. He made no errors in writing 25 phonograms. He had perfect scores in the auditory comprehension task and in sequential commands. On the other hand, his speech disturbance was severe. Spontaneous speech and oral repetition was severely hampered by substitution, omission, distortion and addition of sounds, although he had improved to the extent that he often uttered short sentences. In the oral repetition task, omission errors for consonants 9.6 months after onset had decreased compared with the result 1.3 months after onset. Substitution and distortion errors for vowels also decreased. Abnormal prosody characterized by a prolongation of sounds, syllabification and monotonous pitch was apparent. In the naming task, he had no response to 1 out of 20 objects and had articulatory errors to 19 objects. Syntax was within normal limits. In the Token Test during the same period, the patient had errors on only 2 items in Section F. In summary, his deficits, characterized by isolated severe speech disturbance, was classified as aphemia. His disturbance at the 9.6 month stage, when most phonemes appeared in his speech, constituted an apraxia of speech. The WAB was also tested at 5.3 months after onset in addition to 1.3 months and 9.6 months after onset. The results obtained were very close to those at 9.6 months after onset except that he wrote only one sentence plus 1 word within 3 min in the written description of a picture. His disturbance at this stage is adequately termed a pure apraxia of speech. His apraxia of speech was still present 2 yrs after onset. Praxis and gnosis testing. At both 1.3 months and 9.6 months after onset, he performed the drawing, block design and praxis subtests in the WAB perfectly, which suggested that he did not have constructional, ideomotor or ideational apraxias. One month and 2 months after onset, he correctly performed commands such as 'Open your mouth' and 'Stick out your tongue.' However, he could not 'Cough' or 'Click his tongue', which suggested a mild orofacial apraxia. These two commands were correctly carried out 6 months after onset. Unilateral hemineglect was not observed when drawing pictures. Other apraxias or agnosias were not observed. Intelligence testing. He attained a performance IQ of 105 and a verbal IQ of 81, aided by writing, 1.3 months after onset. Five and a half months after onset, his verbal IQ was 96, and his performance IQwas 104. APRAXIA OF SPEECH 1401 Auditory phonetic analysis The responses of the patients were recorded with a cassette tape recorder (Sony TC-D5M) and narrow phonetic transcription using the International Phonetic Alphabet was carried out. The error responses were categorized into two types according to Trost and Canter (1974): inadequate responses and phoneme errors. Inadequate responses included (1) random phonemes, where phoneme strings failed to yield any kind of word approximation; (2) perseverative responses; (3) paraphasic responses, a semantic substitution such as 'hat' for 'coat'; (4) recurrent utterances; and (5) no response (in this study, naming difficulty). Phoneme errors were those responses which were close to the target word but had some errors in terms of phonemes. The phoneme errors were divided into the following four types of errors: (1) substitutions, as in /t/ for /p/; (2) omissions, as in /a/ for /na/; (3) additions, as in /to/ for jo/; and (4) distortions as in / 3 / for /d/. Compound errors, such as distorted substitution or distorted addition errors, were categorized as substitution or addition errors, respectively. The occurrence of sequential errors and self-corrections were also analysed separately. Sequential errors included anticipatory errors such as '/to:to/' for '/no:to/' (note book), perseverative errors like '/memane/' for '/megane/' (glasses) and metathetic errors like '/tsiukui/' for '/kiutsui/' (shoes). Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 ATTM were included in the ATNR. The third was a 100-syllable repetition test. The three tests will be briefly described. The A TNR. This test included 82 common nouns of from 2 to 6 syllables in length such as dog, glasses and newspaper. It was administered both as a naming task and for oral repetition. In the naming mode, 82 pictures representing the 82 words were presented, and the patient was asked to name each of them three times successively. In the oral repetition mode, the patient repeated each word three times after the examiner had pronounced it once. Japanese is an open-syllable language. That is, syllables consist of one consonant and one vowel with or without an intervening /j/ after the consonant. The syllable can, however, be closed with the nasal sound /N/. Therefore there are no final consonants except for /N/ in syllables. /N/ is a unique phoneme which is realized as [m, n, N] and the nasalization of the previous vowel. In this paper, therefore, /N/ was excluded from the analyses. The total number of phonemes in the 246 responses (82 words x 3) was 1176, with 537 consonants excluding /N/, 615 vowels and 24 consonant-semivowel clusters. The total numbers of initial consonants and medial consonants in test words were 210 and 327, respectively. All of the Japanese consonants (21 in number): /p, t, k, s, J, , <;, h, b, d, g, ts, tj, r, m, n, ji, dz, dj, j , w/ were tested both in initial and medial positions in words. Four consonant-semivowel clusters, /kj/, /bj/, /gj/, and /rj/, appeared 15 times in initial positions in words, and 3 consonant-semivowel clusters, /pj/, /bj/, and /gj/, appeared 9 times in noninitial positions. All of the Japanese vowels (5 in number): /a, o, e, ui, i/ were tested in initial, medial and final positions. The number of vowels tested was 615 (21 initial, 399 medial and 195 final vowels). The A TTM. This test consisted of 32 common words with 2 to 5 syllables. The consonants involved included all of those in the ATNR except for /tj, ji, dz/. The number of phonemes in the 32 words, excluding /N/, was 163. The patient was asked to name each word 3 times, so that the total number of phonemes tested was 489 (163 x 3). The number of consonants was 225 (81 initial and 144 medial consonants) and the number of vowels 255 (9 initial, 153 medial and 93 final). The number of consonant-semivowel clusters was 9. This test was administered only to Case 2 and only in the naming mode The 100-syllable repetition test. This test consisted of 100 monosyllables which included all of the possible syllabic combinations of Japanese phonemes. The patient was asked to repeat each syllable twice in succession. 1402 MORIHIRO SUGISHITA AND OTHERS FIG. 3. Artificial palate, A, photograph of an artificial palate. Receiving electrode (large arrow); one of the transmission electrodes (middle-sized arrow); lead wire (small arrow), B, artificial palate fitted against the patient's palate. Electropalatography Electropalatography (Rion Co. Ltd) was employed to observe the palatolingual contact during the entire ATNR and ATTM naming tests. The system of electropalatography used in this study included an artificial palate, a display panel with a memory system, a recording adapter connected to a dual-channel tape recorder (Sony TC-D5M) and a graphic printer. The artificial palate made of transparent plastic (1 mm in thickness) was formed after a plaster model of the palate of each patient. It had 63 gold transmission electrodes (2 mm diameter each) implanted on its under surface and 1 gold receiving electrode (8 mm diameter) on the upper surface (fig. 3A). The output from a signal source was applied to each transmission electrode through a distributor controlled synchronously by a clock oscillator. The artificial palate was fitted against the patient's hard palate (fig. 3B). When the tongue touched the transmission electrodes the current ran to the receiving electrode and was sensed by detectors. The signals were then transmitted to the memory system which processed the signals at a rate of 64 frames/s; 63 light emission diodes on a display panel showed the positions of the electrodes which the tongue touched. Both the signals and speech were stored on a dual-channel cassette tape via a recording adaptor. The data on the cassette tape could be reproduced and observed in slow motion on the display panel (fig. 4). The patterns of the palatolingual contacts for the 2 patients were then printed out and compared with those of 5 normal subjects who spoke each word 3 times, giving 15 normal patterns. The data for omission errors were classified into three types of palatolingual contact: correct contact, incorrect contact and no contact. The palatolingual contact patterns of the phonemes omitted by the patients were judged correct when they did not exceed the maximum or minimum contact range of the normal patterns of target sounds. They were judged incorrect when they were within the maximum or minimum contact range of different sounds or they were undifferentiated contact patterns, that is, exceeding the maximum or minimum contact range of any sound. They were judged no contact when they had no palatolingual contact. Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 B APRAXIA OF SPEECH 1403 RESULTS Case I Auditory phonetic analysis ATNR naming test. This test was carried out in the naming mode 2.3 months after onset. Out of the 82 test words, 2 (pencil and telephone) were not tested. The 80 pictures representing the 80 words were named 3 times, so that the patient produced 240 responses (80 words x 3) on which auditory phonetic analysis was performed. Out of the 240 responses, 55 responses were correct and 185 responses were judged as having phoneme errors. There were no inadequate responses such as no responses (naming difficulties). The 185 phoneme-error responses had 290 erroneous phonemes, that is, 25% of the total number of phonemes (1149). The phoneme errors included 63% (182/290) substitution errors, 22% (64/290) omission errors, 14% (41/290) distortion errors and 1% (3/290) addition errors. There were no omission or addition errors of syllable units. Self-correction responses were only 2% (7/290) of the total errors. Nine sequential errors (8 anticipatory, 1 perseverative) were observed. They accounted for only 3% (9/290) of the total Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 FIG. 4. Electropalatograms from Case I shown on the display panel. Transmission electrodes in contact with the tongue are indicated by the light emission diodes of the display panel which are small circles in the electropalatograms. Al, A2, correct contact of palatolingual pattern for the target sound /d^J AI is its closure and A2 IS its release following closure B, substituted pattern of palatolingual contact for the target sound /j/ to lingualalveolar sounds such as /t/, /d/ or /n/. Both (A, B) were perceived as omission errors 1404 MORIHIRO SUGISHITA AND OTHERS Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 errors. Further analysis for consonant, consonant-semivowel cluster and vowel errors was as follows. 1. Consonants. The patient made 261 consonant errors, involving 50% (261/525) of the total consonants used in the test (Table 2). There were no significant differences in the numbers and the types of errors between the first, second and third responses. The percentage of errors was 74% (153/207) for initial consonants and 34% (108/318) for medial consonants, indicating that initial consonants elicited more errors than medial consonants Of2 = 80.1, P < 0.01). Errors were observed for all 21 consonants. The consonant with the lowest percentage of errors was /ji/ (17%) followed by /tj/ (29%), /m/ (29%), /w/ (33%), /k/ (40%) and /t/ (41%). Those with higher percentages were /dz/ (95%) /ts/ (88%) /s/ (81%) /d/ (73%) and /d3 (73%) (fig. 5). This indicated that the most difficult consonants were affricates and fricatives. Out of 261 consonant errors, 63% were substitutions, 23% omissions, and 14% distortions, with no errors of addition (Table 2). Substitution errors occurred for all 21 Japanese consonants, and were highly inconsistent. Seventeen consonants which elicited more than 3 substitution errors (i.e., /j, p, t, k, b, d, g, m, n, s, h, ts, dz, r, J, tj", d3/) were substituted by more than 2 different consonants. These consonants were changed to 18 different consonants (/p, t, k, J", , c, h, b, d, g, ts, tj", r, n, ji, dz, d3, j/), with substitutions of /t/ (20), /tj/ (18), /k/ (14) and /n/ (13) occurring most frequently. When analysing the consonants for consonant-substitution errors using the four subphonetic dimensions of Trost and Canter (1974) — that is, place, manner, voicing and oral-nasal — 60% (91) were off by one dimension, followed by those off by two dimensions 29% (45), three dimensions 8% (12) and four dimensions 3% (4). Place was the most frequent error, 48% (111), followed by manner errors 26% (59), voicing errors 19% (43) and oral-nasal errors 7% (16). There were more voiceless substitution errors for voiced consonants, 81 % (35), than the reverse type of error, 19% (8), indicating a strong tendency toward devoicing. Sixty omission errors were observed for 17 different consonants among the 21 different test consonants (fig. 5). While the omission error percentages varied from 29% for /dz/ to 6% for /J/ and /j/, there were no significant differences between TABLE 2 CONSONANT, VOWEL AND CONSONANT-SEMIVOWEL CLUSTER ERRORS. AND E A C H T Y P E O F E R R O R IN CASE 1,2 3 MONTHS AFTER ONSET* Sound tested Errors (error rate) Error type Substitution Omission Addition Distortion Consonant Vowel Consonant-semivowel Total 525 261(50%) 600 I7( 3%) 24 12(50%) 1149 290(25%) 165(63%) 60(23%) 50 7 3 0 3? 9 9 78 2 3 5 44 8 a 96 11 30 FIG. 6. Confusion matrix of consonants, consonant-semivowel clusters and vowels for Case 2, 9.6 months after orfsct. Each line of the matrix shows the distribution of sounds which the patient produced in response to the target sounds as indicated in the left margin. Cells forming the diagonal are correct productions. SP= sounds not existent in Japanese language but possible to transcribe by International Phonetic Alphabet. OM = Omission; DIS = distortion; ADD-addition; T = total number of errors; O = total number of occurrences of each target sound. phonemes tested (Table 4, fig. 6). The phoneme errors included 78% (282/362) substitution errors, 13% (47/362) distortion errors, 8% (30/362) addition errors and 1 % (3/362) omission errors. Sequential errors were relatively small, accounting for 9% (32/362) of the total errors, with anticipatory errors predominant (25 anticipatory, 7 perseverative and no metathetic errors). There were no additions or omissions of syllable units. Further analysis for consonant, consonant-semivowel clusters and vowels were as follows. 1. Consonants. Consonant errors were made for 57% (301/525) of the consonants tested (Table 4). No significant differences were observed in error rates or types of errors between thefirst,second and third responses. The error rate for initial consonants was 80% (164/204) and that for medial consonants 43% (137/321), Downloaded from http://brain.oxfordjournals.org/ at Laurentian University on June 6, 2016 Targe sounds n J c 1/ d j PJ 1 1 7 w t Ik b d q rn n s h * ts dz r APRAXIA OF SPEECH 1409 TABLE 4 CONSONANT. VOWEL AND CONSONANT-SEMIVOWEL CLUSTER ERRORS, AND EACH TYPE OF ERROR IN CASE 2, 4.5 AND 9 6 MONTHS AFTER ONSET 4 5 months after onset Sound tested Errors (error rate) Error type Substitution Omission Addition Distortion Vowel Consonant-semivowel Total 225 131(58%) 255 16(6%) 9 9(100%) 489 156(32%) 121(92%) 7(5%) 1(1%) 2(2%) 4(25%) 0(0%) 9(56%) 3(19%) 7(78%) 0(0%) 0(0%) 2(22%) 132(84%) 7(5%) 10(6%) 7(5%) 525 301(57%) 600 41(7%) 24 20(83%) 1149 362(31%) 262(87%) 3(1%) 0(0%) 36(12*/.) 0(0%) 0<0%) 30(73%) 11(27%) 20(100%) 0(0%) 0(0%) 0(0%) 282(78%) 3(1%) 30(8%) 47(13%) •The ATTM naming test was employed 4 5 months after onset, and the ATNR naming lest 9 6 months after onset. indicating more errors for initial consonants than for medial ones (/ 2 = 72.5, P < 0.01). Errors were observed for all of the 20 consonants except for /ji/. The error rate was the highest for /g/ (27/27) and /